early risk assessment of long- term sick leave among...

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2016

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Medicine 1194

Early risk assessment of long-term sick leave among patients inprimary health care

risk factors, assessment tools, multidisciplinaryintervention, and patients’ views on sick leaveconclusion

ANNA-SOPHIA VON CELSING

ISSN 1651-6206ISBN 978-91-554-9508-4urn:nbn:se:uu:diva-280414

Dissertation presented at Uppsala University to be publicly examined in A1:107a,Biomedicinskt Centrum, Husargatan 3, Uppsala, Friday, 29 April 2016 at 09:15 for thedegree of Doctor of Philosophy (Faculty of Medicine). The examination will be conductedin Swedish. Faculty examiner: Associate Professor Sören Brage.

Abstractvon Celsing, A.-S. 2016. Early risk assessment of long-term sick leave among patientsin primary health care. risk factors, assessment tools, multidisciplinary intervention, andpatients’ views on sick leave conclusion. Digital Comprehensive Summaries of UppsalaDissertations from the Faculty of Medicine 1194. 85 pp. Uppsala: Avhandlingsproduktion.ISBN 978-91-554-9508-4.

Background. Long-term sick leave is one of the main risk factors for permanent exit out of thelabour market. The longer the duration of sickness absence, the less likely sick leave conclusion.

Objectives and Methods. The aims were to analyse possible determinants of sick leaveconclusion and their relative impacts, to analyse the properties of two models for the assessmentof sick leave conclusion, to study the impact of a multidisciplinary vocational intervention forsick leave conclusion in a high-risk group for long-term sick leave compared to a matched-control group, and to compare the patients’ own assessment on chance to sick leave conclusionwithin 6 months with the assessment of a team of rehabilitation professionals. A prospectivecohort study of 943 patients aged 18 to 63 years, sickness certified at a Primary Health CareCentre in Sweden during 8 months in 2004, and follow-up for three years.

Results. Significant determinants increasing time to sick leave conclusion were numberof sick leave days the year before baseline, age and a psychiatric diagnosis (F in ICD-10).Concordance between actual sick leave conclusion and that predicted by a computer-basedmodel was 73-76% during the first 28-180 days in a manual model, and approximately 10%units higher in a computer based model. Three nomograms provided detailed information onthe probability on sick leave conclusion. Before intervention started, the rehabilitation grouphad a 73% higher sick leave conclusion rate than the control group but during the rehabilitationprogramme period, a 51% lower conclusion rate, and after there were no significant differencesbetween the groups. The patients’ and the rehabilitation teams’ assessment scores were highlycorrelated (r=0.49).

Conclusions. Previous sick leave was the most influential variable associated with sick leaveconclusion. A computer- based assessment model gave more detailed information on sick leaveconclusion than a manual model. A multidisciplinary intervention declined sick leave in a high-risk group for long-term sick leave but after intervention there was no difference betweengroups. Patients’ own view on sick leave conclusion was highly correlated to the assessmentof professionals’.

Keywords: sick leave, sick leave conclusion, return to work, risk factors, assessment tool,multidisciplinary intervention, risk assessment, gut feeling

Anna-Sophia von Celsing, Department of Public Health and Caring Sciences, FamilyMedicine and Preventive Medicine, BMC, Husargatan 3, Uppsala University, SE-752 37Uppsala, Sweden.

© Anna-Sophia von Celsing 2016

ISSN 1651-6206ISBN 978-91-554-9508-4urn:nbn:se:uu:diva-280414 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-280414)

To my family

”Nothing is impossible.The impossible just takes a little longer”

Winston Churchill

List of Papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals

I von Celsing A-S, Svärdsudd K, Eriksson H-G, Björkegren K,

Eriksson M, Wallman T. Determinants for return to work among sickness certified patients in general practice. BMC Pub-lic Health 2012, 12:1077

II von Celsing A-S, Svärdsudd K, Wallman T. Predicting return to work among sickness- certified patients in primary health care. Properties of two assessment tools. Upsala Journal of Medical Sciences 2014, 119: 268-277

III von Celsing A-S, Svärdsudd K, Wallman T. Unexpected effects of a vocational rehabilitation programme on return to work among sick-listed primary health care patients. A population-based matched case-control study. Submitted

IV von Celsing A-S, Svärdsudd K, Eriksson H-G, Björkegren K, Wallman T. Patients’ versus medical professionals’ assessment of possibilities to return to work. A longitudinal study of pa-tients with a high risk of long-term sick leave. Manuscript

Reprints were made with permission from the respective publishers.

Contents

Prologue .......................................................................................................... 9

Introduction ................................................................................................... 11

Background ................................................................................................... 12Sick leave .................................................................................................. 12The Swedish sickness insurance system ................................................... 13The sickness certificate ............................................................................. 13Long-term sick leave ................................................................................ 14Determinants of long-term sick leave ....................................................... 14The sickness certification consultation in primary health care ................. 15Rehabilitation of patients at risk of long-term sick leave ......................... 16Return to work versus sick leave conclusion ............................................ 17Prediction of time to sick leave conclusion .............................................. 17The patients’ own assessment of ability to return to work ....................... 18Study background ..................................................................................... 18

Aims of the study .......................................................................................... 20

Material and Methods .................................................................................... 21Setting ....................................................................................................... 21Study population ....................................................................................... 21Data collection .......................................................................................... 23Statistical Analysis – general information ................................................ 24

Follow-up and outcome ....................................................................... 24Statistical analysis – specific information ................................................ 25

Paper I .................................................................................................. 25Paper II ................................................................................................. 26Paper III ............................................................................................... 28Paper IV ............................................................................................... 31

Ethical considerations ............................................................................... 33

Results ........................................................................................................... 35Paper 1 ...................................................................................................... 35

Characteristics of the study population ................................................ 35Potential determinants of sick leave conclusion .................................. 35Degree of explanation .......................................................................... 40

Paper II ..................................................................................................... 41Sick leave conclusion according to the manual model ........................ 41Sick leave conclusion according to the computer-based model .......... 42

Paper III .................................................................................................... 45Benefit status at three years after baseline ........................................... 50

Paper IV .................................................................................................... 51Influence of gender in the assessments ................................................ 54

Discussion ..................................................................................................... 57Study design and methodological considerations ..................................... 57Return to work – sick leave conclusion .................................................... 58Discussion by findings .............................................................................. 59

Determinants ........................................................................................ 59Easy-to-use screening tool ................................................................... 61Multidisciplinary rehabilitation intervention ....................................... 62Timing of intervention ......................................................................... 64Patients’ and rehabilitation team’s assessment .................................... 66Red flags, signs of alert in sick leave ................................................... 67

Conclusions ................................................................................................... 69General conclusions .................................................................................. 69Implication for practice ............................................................................ 69Suggestions for further research ............................................................... 69

Swedish summary ......................................................................................... 71

Acknowledgements ....................................................................................... 73

References ..................................................................................................... 75

9

Prologue

My journey into medical social insurance research started in 2002 with a feeling of frustration, staring at piles of sickness certificates on my desk. At that time, Sweden experienced a substantial increase in sick leave, and at the primary health care centre where I was working as a consultant in orthopae-dics and rehabilitation there were a number of Locum Doctors handling the difficult task to assess their patients need of sickness certification.

In the rehabilitation team, we fought our way to find out which patients needed our help to regain work ability and who would return to work spon-taneously. It was obvious that we had to find a mode of classification to identify the high-risk patients for not concluding sick leave. One of the expe-rienced physiotherapists and I were convinced − based on many years of professional experience and our “gut feeling” − that we could tell which patients would conclude their sick leave at the expected time or not. Moreo-ver, we believed that a fairly good prediction could be made, by considering just a few significant determinants for long-term sick leave. Intense discus-sions about those determinants and our “gut feeling” followed and we felt uncertain about various aspects, for example; why is it that out of two pa-tients with the same diagnosis, one returns to work and the other one ends up with a disability pension? What significance do sex, age, unemployment and previous periods of sick leave have on time until sick leave conclusion? What constitutes enough information to be able to predict sick leave conclu-sion?

My curiosity increased and I found that it was necessary to conduct some research to be able to tell whether the hypothesis was correct or not. Fur-thermore, why not construct a very simple prediction tool that would be use-ful to identify the patients who are in need of rehabilitation? I believed that it would be easy and would not take very long to find out. My old mathemat-ics teacher used to say, “Anna-Sophia, you cannot see the wood for the trees”. He was right and I was wrong, it has indeed been a difficult process to enter the world of research to finally see the wood and it took a very very long time! But it has been so much more interesting, rewarding and chal-lenging than I ever imagined.

However, my most important reason for writing this thesis has been to gain more knowledge and understanding of the determining factors in the tortuous path of the sick leave process.

11

Introduction

Sick leave and incapacity to work have increasingly been recognized as con-siderable health problems, with consequences for individuals, their families, and workplace and for society (1). Early detection of patients at risk of long-term sick leave is of importance for the identification of individuals in need of rehabilitation measures to enable them to regain their work ability. The medical literature indicates generally that the timing of an intervention is important, with a less favourable outcome with increasing sick leave dura-tion (2-4). The harm of being long-term sickness absent for the individual and society is demonstrated in a number of studies (5-8).

The physician is the key person to assess not only a patients’ medical problem but also the possible consequences of the condition, e.g., whether it effects work ability or not. If so, the physician issues a sickness certificate. Thereby, a suitable length of sick leave must be noted in the sickness certifi-cate. Today most physicians are well aware of the risk of possible no return to work when being long-term absent from work. The difficulty in the time limited every day consultation, is to identify at a glance a patient who is at risk for long-term sick leave.

Research in the last decades found many risk factors for long-term sick leave, but we do not know if just a few more significant ones would be suffi-cient to take into account when assessing length of sick leave. In other fields of medicine various instruments for the prediction of outcomes have been forwarded, of which the most common are so-called nomograms, where the combined effects of various predictors for the risk of a specific outcome may be arrived at. No similar instrument for the prediction of length of sick leave has been presented so far. Having access to the individual risk factor pattern of patients on sick leave may provide a possibility of early identification of patients who may not conclude sick leave as expected.

12

Background

Sick leave That work is generally good for health and well-being, and absence from work is generally detrimental is a widely supported concept today (9-11). In several western countries sickness absence has increased in the last decades to become a major public health and economic problem. In Sweden, sickness absence has since the 1980s, been characterized by large variations without any correspondence of variations in sickness and health in the population (12).

The total expenditure on the allowances and benefits administered by the Swedish Social Insurance Agency amounted to around SEK 218 billion in 2014, or 5.6 per cent of Sweden’s gross domestic product (GDP) (12). Just over half of the expenditures were to persons who were sick and persons with disabilities. Long-term sick leave contributes disproportionately much to these numbers.

Figure 1. Number of cases on sick leave for 29 days or longer in December, 1974-2014. Data from the Swedish Social Insurance in Figures 2015 (12).

13

The Swedish sickness insurance system The Swedish sickness insurance system is regulated by the Social Insurance code 2010:110 (13). Sickness benefit is income-based and covers all people who are living in Sweden, are 16 years or older, and who have annual in-come from work. The primary requisite for being eligible for sickness bene-fit is that the individual has reduced workability due to disease or injury by at least 25%. Depending on the degree of work ability reduction, the rate of compensation provided by the insurance scheme may be 25%, 50%, 75% or 100%. Sickness benefit may also be granted to individuals undergoing medi-cal treatment or rehabilitation intended to prevent or shorten illness. The decision on whether a sickness benefit is granted or not, is taken by the Swe-dish Social Insurance Agency. Disability pension may be granted if worka-bility is permanently reduced with at least 25%.

Employed individuals receive sickness compensation from their employ-ers during the first 14 days of sickness absence, however the first day consti-tuting a qualifying day. If the workability is still reduced after the period of employers’ sickness compensation is over, the employee may receive sick-ness benefit from the Swedish Social Insurance Agency. A sickness certifi-cate is required after seven days of a patient’s self certified sickness absence.

Unemployed individuals are paid sickness benefit by the Swedish Social Insurance Agency after one qualifying day. Self-employed individuals have a default qualifying period of 7 days, but may choose an optional period of 1, 14, 30, 60 or 90 days.

Up until 2007, there was no official limit on how long an individual could be granted sickness benefit. In July 2008, the National Board of Health and Welfare introduced national guidelines to be used by physicians, providing advice regarding reasonable time on sick leave for specific diagnosis (14). The aim was to make the sickness certification consistent and coherent, and to safeguard a legally secure process. In an amendment to the 2008 legisla-tion, a stricter interpretation of laws and regulations was adopted regarding how long sickness benefits could be granted, The Rehabilitation chain (15). Such compensation was limited to one year, although it might be prolonged under certain circumstances. During the study period of this thesis there was no upper time limit for sickness benefit.

The sickness certificate All Swedish physicians are entitled to issue certificates for sickness absence and disability pension. The sickness certificate must provide information on patient identity, diagnosis, a brief sickness history, a description on the med-ical condition, in what way the condition affects function and workability, treatments and other recommended measures to improve health and restore

14

workability, and the duration and degree of the sickness absence spell. While issuing the certificate, the physician is required to consider the appli-cation of the sickness insurance rules and collaboration with external parties.

Long-term sick leave The reason for ‘early’ interventions is based on research from the last dec-ades indicating that a history of prolonged or recurrent sickness absence, makes it less likely that the individual will return to work (2, 3, 16). The history of previous sick leave predicts future sick leave (17-19). Further-more, prolonged periods of sick leave are associated with an increased risk of being granted a disability pension, which in itself has been found to have future negative consequences, such as increased mortality rates among disa-bility pensioners (20-25). Stapelfeldt et al. studied sick leave patterns in elder care workers, and found sick leave length was a better indicator of future workability than spell frequency (26). Waddell et al. summarized in a review, ‘the longer anyone is off work, the less likely they are to get back – ever’ (10). Hence, duration of sick leave is one of the most powerful risk factors and predictors for long-term incapacity.

There is no generally established definition of long-term sick leave in the literature, and the number of days, regarded as long-term sick leave, varies considerably between different studies (27). In the Ten Town study of mu-nicipal employees in Finland sick leave of more than three days and also ten days was defined as long-term sick leave (20). In a Danish study of employ-ees, and in prospective cohort studies in Norwegian counties, sick leave ex-ceeding eight weeks was considered as long-term sick leave (28-31). In the Whitehall study of British civil servants, seven to 21 days, and in the French GAZEL cohort of employees from the National Gas and Electricity Compa-ny, 14 days was defined as long-term sick leave (11, 32, 33). In Sweden, the cut off levels used in various studies have ranged from 21 days (34) to 28 days (35), 56 days (36), more than 59 days (37), and 90 days or more (38, 39). The reason for choosing a specific cut off level of days might be the availability of sick leave data in a specific setting or adjustment to the cur-rent social insurance legislation.

Determinants of long-term sick leave Early detection of patients at risk for long-term sick leave is of importance for identification of individuals in need of rehabilitation measures in order to regain their work ability (24, 40, 41). A number of potential determinants have been identified for long-term sick leave and possible transition to disa-bility pension, such as female gender, high age, low socio-economic status,

15

and previous spells of sick leave (40, 42-46). The most consistent results within socio-demographic factors have been reported for age, which tends to be negatively related to sick leave, and for gender − women have higher rate of absence and more days of absenteeism than men, but may differ between diagnosis (47). Some sick-leave diagnoses appear to be more prone to long-term sick leave periods than others, such as psychiatric diseases (31, 48-51), and musculoskeletal disorders (45, 52-55). Work-related factors identified are physically heavy work, high work demands, low work control, relational problems at the workplace, and a stressful work situation, which all tending to prolong sick leave. Low job satisfaction was in the Whitehall study strongly related to sickness absence with higher rates in those who reported low job satisfaction (32). A Dutch study found that job satisfaction was sig-nificantly related to total sickness absence duration (56). Still other determi-nants are for instance unemployment (43, 57, 58), fatigue and sleeping dis-orders (59-63), abuse of alcohol (64-67), sickness presence (68, 69), and being a frequent attender in primary health care (70).

The sickness certification consultation in primary health care Sickness certification is a common task for many physicians in Sweden as well as in other European countries (71, 72). Although all physicians are allowed to issue sickness certificates, most certificates are issued by GPs (73). At the time of the present study, a review by Wynne-Jones et al. on sickness certification rates in European primary care found a wide variation, ranging from 18 per 100 person years in Norway to 239 per 100 person years in Malta (74). In Sweden, a survey in 2008 of physicians of all specialities showed that 67% had sickness certification consultations, and of these, 83% had it at least once a week (75). There was variability between specialities regarding the proportion of who had such consultations more often than 5 times a week, ranging from 3% in dermatology, 43% in primary care, to 79% in orthopaedics. In a European perspective there is large variability in sickness certification policy and hence sickness certification rates. It is also difficult to obtain data relating to sickness absence due to insufficient availa-bility of comprehensive, reliable, and comparable data (76).

Physicians’ difficulties with the sickness certification procedure are well documented in previous research (77, 78). Physicians experience sickness certification as a complex and problematic task (71, 73, 79-82), however, Ljungquist et al., found that the rates of physicians finding sickness certifica-tion task problematic varied much between clinical setting (83). In a study by Lindholm et al. the largest problems were reported by general practition-ers and rheumatologists (75).

16

Studies exploring sickness certification in primary care have provided greater insight into GP’s perceptions of their role in the management of sickness absences issues (84). Important factors influencing GPs in their attitude to the management of return to work issues were: the doctor-patient relationship, patient advocacy, pressure on consultation time, limited occu-pational expertise, and barriers to good practice (85-88). GPs in the UK and elsewhere often feel that managing work and health issues per se goes be-yond their role (85, 86). Hussey et al. illustrated how GPs manage the certi-fication process to overcome some of challenges they encounter (85). Löfgren et al. looked at the rate of problems GPs encounter in managing certification issues (73). Physicians even consider sickness certification as a work environmental problem (89).

In the study by Lindholm et al. around 60% found it problematic to assess patients’ work capacity and to provide a prognosis regarding the duration of work incapacity. Peterson et al. in 1997 found that if the GP issues a first sickness certificate with a long sick leave duration that is an intuitive predic-tion of long-term sick leave in itself (90). Reiso et al. investigated the accu-racy of GPs’ return to work predictions for their patients (91). They noted that GP prediction accuracy was much higher for short-term than for long-term absences. Shiels et al. investigated patient factors that influenced the duration of sickness absence and transition to state incapacity benefits, and found factors such as diagnoses of mild mental disorders, old age and social deprivation (92). Some studies explored the nature of the sickness consulta-tion itself and both found that communication as well as system issues im-pacted on GPs attitudes to work and health (93, 94).

Rehabilitation of patients at risk of long-term sick leave It has been suggested that early multidisciplinary medical rehabilitation and vocational intervention are generally effective and recommended to be in-cluded in all interventions to enhance returning to work (95, 96). In this con-text, multidisciplinary means that rehabilitation team members from several rehabilitations fields, not only physicians but also physiotherapists, occupa-tional therapists, social workers, employers, etc., cooperate and provide their competence in the rehabilitation process.

Vocational rehabilitation is by definition a multidisciplinary intervention in a process linked to the facilitation of return to work or to the prevention of loss of work (97), and in a definition by Waddell et al. ‘whatever helps someone with a health problem to stay at, return to and remain in work’ (98). Thus, vocational rehabilitation addresses both health care problems and work issues requiring a combination of healthcare and workplace interven-tions. Commitment and coordinated action from all stakeholders is found to be crucial for successful vocational rehabilitation: especially important is

17

communication between healthcare professionals, employers and employees, which should be initiated at an early stage of absence (99).

In Sweden, recommended method to promote return to work is multidis-ciplinary vocational rehabilitation in cooperation with authorities. Several European countries have varieties of return to work coordination pro-grammes that aim at facilitating and hastening return to work after sickness absence (100). These programmes differ due to variations in settings, target groups and content, but they usually include medical management, physical rehabilitation, worker-job match, and managed care.

A number of these programmes have been evaluated. Several studies with specific treatment interventions at target diagnostic or other homogeneous groups show a positive effect (101), and there is support to recommend an early start of rehabilitation in a sick leave period to achieve sick leave con-clusion (100).

However, it has also been found that interventions may prolong sickness absence (102, 103). There are few intervention studies targeted at a general patient population in primary health care, and to the best of our knowledge, no multidisciplinary rehabilitation and vocational intervention has shown to be effective regarding return to work (102, 103).

Return to work versus sick leave conclusion In this research context the term ‘return to work’ was usually used to indi-cate that the sick leave period had ended. However, since not all patients in the present study went back to work when the sick leave period was over, the term ‘sick leave conclusion’ is used to indicate the conclusion of the sick leave period, and the term ‘return to work’ is used for those who actually did so when the sick leave period was over.

Prediction of time to sick leave conclusion An early estimate of the prognosis for patients on sick leave is of importance to identify individuals less likely to recover and conclude their sick leave. GPs are requested to estimate the duration of sickness absence, and their ability to predict the length of sickness absence is important for the initiation of therapy and rehabilitation in patients with increased probability of long-term sick leave (72, 104). The assessment as to whether the patient will con-clude sick leave at a defined time is difficult and the GPs’ accuracy to pre-dict the further course of long-term sick leave is low (105). Fleten et al. found that sick-listed individuals predicted their length of sick leave more accurately than professionals did (106). Furthermore, it is shown that work-

18

ers’ expectation of return to work is positively associated with likelihood (107, 108) and time to return to work (109).

There is a lack of assessment tools for the purpose of predicting time to sick leave conclusion. In a study by Duijts et al., a screening instrument for early identification of employees at risk for sickness absence due to psycho-social health complaints was developed (110).

Examples of prediction tools from other areas in medicine are the risk nomograms for myocardial infarction (111, 112), the WHO Fracture Risk Prediction Tool (FRAX) for evaluation of the probability of a fracture (113), and the asthma risk nomogram to assess the risk in preschool children of developing asthma during the next 5 years (114). No similar tool regarding the assessment of time to sick leave conclusion has been presented so far.

The requirement for a screening tool would be to accurately assess pa-tients’ risk for long-term sick leave, i.e., who is not likely to conclude sick leave at expected time, and it should be easy to use by GPs’ and other medi-cal professionals’. The question though, is whether there are determinants more significant than others and if they are easily available to the GP in a busy and ordinary practice. It must be easy to use information that automati-cally is already known or immediately accessible without long time search-ing in the patients’ medical records or deep penetration into medical history.

The patients’ own assessment of ability to return to work There are few studies on the prediction of sick leave duration focused on general practitioners’ ability to forecast this duration (41, 105). Some studies have focused on the patients’ view of expected sick leave duration. Howev-er, the results are at variance, some claim patients’ view to be the most in-formative (106), others that professionals’ view carry the best information (105).

Study background This study is based on all sick-listed patients, sampled during an eight months wide time window at a primary health care unit in Eskilstuna. The analyses have been focused on four objectives where previous research ei-ther has shown conflicting results or where no previous research has been published.

These objectives were first to analyse not only what factors are of im-portance for sick leave conclusion, but also their relative importance, i.e., their rank as risk factors, second to create a prediction instrument, third to

19

make an effort to evaluate the effect of multidisciplinary rehabilitation, and fourth to analyse the predictive power of professionals’ and patients’ views on time of sick leave conclusion.

20

Aims of the study

The overall aim of this thesis was to improve the early identification of pa-tients at risk for long-term sickness absence in primary health care, to im-prove prediction of time on sick leave and to evaluate the effects of interven-tion.

The specific aims were

• Paper I – to analyse possible determinants of sick leave duration and their relative impacts, in order to arrive at a simple model by which sick leave conclusion might be estimated early in the sick leave process in the everyday clinical primary health care practice.

• Paper II – to compare the predictive properties of a traditional GP clinical assessment of sick leave conclusion, with those based on a computer based multivariate analysis of potential determinants presented as nomograms for sick leave conclusion.

• Paper III – to evaluate the efficacy of an early medical assessment and a vocational intervention in patients on sick leave with a high risk of not concluding sick leave within expected time as com-pared to a matched non-programme group.

• Paper IV – to compare the scores of the high-risk patients’ own assessment of chance to sick leave conclusion within 6 months, with the corresponding assessment performed by a team of reha-bilitation professionals.

21

Material and Methods

Setting The study was performed in Eskilstuna, Sweden, an industrial city, located 110 kilometres west of Stockholm, with 91,000 residents as of 2004. It was performed at one of the County Council-operated primary health care cen-tres, with ten GPs serving a population of approximately 25,000 residents.

Study population The study was designed as a three-year prospective cohort study with re-cruitment from 1 January until 31 August 2004. During the recruitment peri-od copies of all sickness certificates, whether new or prolongation ones, issued at the primary health care centre, were obtained. All patients aged 18 to 63 years, who were sickness certified by a physician at the centre at any time during the recruitment period, and who were not already included in a medical or vocational rehabilitation programme, were included, 482 women and 461 men, altogether 943 patients. The aims, study design, study population, outcome and statistical methods used in the four papers are summarized in Table 1. In Paper I the complete study population was used as a cohort. In Paper II the study population was classified into a high-risk group (n=170), a moderately high-risk group (n=277), and a low-risk group (n=496) for not concluding sick leave at expected time. The high and moder-ately high-risk groups were amalgamated into a high-to-moderately high-risk group (n=447). The classification was performed at baseline by Rehabilita-tion team I, which included experienced rehabilitation professionals − a phy-sician, a physiotherapist, and a clerk from the local Social Insurance Agency office. The members of the team were the same during the eight months inclusion period. In Paper III the high-risk group (n=170) for not concluding sick leave at expected time was compared to a matched control group (n=340) sampled from the moderately-high-to-low risk groups, see Statistical analysis below. Flow-chart of the study population is shown in Figure 2. Paper IV was based on the high-risk patient group (n=170).

22

Table 1. Overview of aims, study designs, study populations, outcome, and statistical methods in Papers I−IV.

Paper I Paper II Paper III Paper IV

Aim To analyse possible deter-minants of sick leave conclusion and their rela-tive impacts, in order to arrive at a simple model by which sick leave conclusion might be esti-mated early in the sick leave process

To compare two models for the assessment of sick leave con-clusion after sickness certifi-cation, one manual and one computer-based

To evaluate the efficacy of a multidiscipli-nary vocational intervention between a group of high-risk sick-listed for no sick leave conclusion within expected time and a matched-control group

To compare the sick-listed high-risk patients own assessment of chance to sick leave conclusion within 6 months, to the assess-ment of a team of rehabilitation professionals’

Study design

Prospective cohort study 2004-2007

Prospective cohort study 2004-2007

Matched case-control 2004-2007

Prospective cohort 2004-2005

Study population

Sick-listed Jan-Aug 2004, age 18-63, at a primary health care centre, n=943

High-risk (n=170) + mod-erately high-risk group (n=277) n=447 Low-risk group n=496

High-risk group n=170 Matched control group n=340 out of 773 (moder-ately high-risk + low-risk group)

High-risk group n=170

Outcome Sick leave con-clusion of the sick certification in effect at baseline during 3 years of fol-low-up

Sick leave con-clusion of the sick certification in effect at baseline during 3 years of fol-low-up

Sick leave con-clusion of the sick certification in effect at baseline during 3 years of fol-low-up

Sick leave conclusion of the sick certifi-cation in effect at baseline during 6 months of follow-up

Statistical methods

Nominal logistic regression Pro-portional haz-ards regression

Binominal logistic regres-sion Proportional hazards regres-sion

T-test, chi-square Binominal logistic regres-sion Proportion-al hazards re-gression

Binominal logistic regres-sion Proportion-al hazards re-gression

23

Figure 2. Flow-chart of the classification in risk groups.

Data collection Baseline data regarding sex, age, sick leave diagnosis, employment, and the degree of sick leave (25%, 50%, 75%, or 100%) were obtained from the sickness certificate. Sick leave data were checked in the primary health care medical records in order to get valid information on sick leave during the first 14 days paid by the employer. The sick leave diagnoses were coded according to the International Classification of Diseases (ICD-10) (115).

Data on number of sick leave days during three years after baseline, first and last day of each sickness spell, degree of sick leave (25%, 50%, 75% or 100%) type of sickness benefit (compensation or rehabilitation), and whether a disability pension was granted during follow-up was obtained from the Swedish Social Insurance Agency database. Further information obtained

24

from the database was whether being born in Sweden, citizenship, marital status (classified as never married, married, divorced or widowed), occupa-tional status (being employed or not), salary at baseline (expressed in Euro), and data on number of sick leave days during the year before baseline. The same data as above was also available in the local Social Insurance Agency office.

Data on days of unemployment benefits during the three-year follow-up were obtained from the Swedish Employment Agency. Data on social bene-fits to the patients’ household during the three-year follow-up period was obtained from the Eskilstuna County Council. Data on vital status, including date of death for those who died (n=6) was obtained from the National Cause of Death Register. Data on patient score of expected time to sick leave con-clusion within 6 months were obtained from question number 16 in the Örebro Musculoskeletal Pain Screening Questionnaire (116). Data on the Rehabilitation team II assessment of sick leave conclusion in 6 months was obtained at baseline.

Statistical Analysis – general information Data were analysed with the Statistical Analysis System (SAS), version 9.3, and Statistical Package for the Social Sciences (SPSS) version 22.0. There was no missing data in Paper I−III. In Paper IV the maximum proportion of missing data was 3.5%. For analyses of baseline relationships binominal logistic regression providing odds ratio (OR) and its 95% confidence limit (95%CI), Wald’s χ2 (a measure of exposure impact on outcome), and p-values were used. For follow-up analyses proportional hazards regression analysis (Cox’s analysis) was used, providing the same statistics as in the logistic regression except that hazards ratio (HR) is produced rather than odds ratio. Simple differences between groups were tested with Student’s t-test regarding continuous variables and the χ2 test regarding discrete varia-bles. All tests were two-tailed and the significance level was set at p<0.05.

Follow-up and outcome The three-year sick leave follow-up data were converted into a day-by day matrix starting with variable day 0 (baseline day) and ending with variable day 1095 (end of follow-up). Each variable measured whether the patient was on sick leave (=1) or not (=0) on that day. For each sick spell the follow-ing two sick leave conclusion criteria, proposed by Bogefeldt et al. (117), were applied. Criterion 1: the sick spell was followed by a sick leave free interval of more than 28 days, regardless of the length of any following sick spell. Criterion 2: the sick spell was followed by a sick leave free interval of more than 7 days, and that interval was longer than the next sick spell. When

25

at least one of the criteria was satisfied, sick leave conclusion was presumed to have occurred on the first non-sick leave day. If none of the criteria were satisfied at end of follow up, no sick leave conclusion was presumed to have occurred. Follow-up time from baseline to sick leave conclusion or end of follow-up was measured as number of days from baseline. Outcome in all Papers was conclusion of the sick certification period in effect at baseline.

Statistical analysis – specific information

Paper I Determinants for sick leave conclusion were tested with SAS version of Cox’s proportional hazards regression analysis with sick leave conclusion and the day when this occurred as outcome, and age at baseline, sex, number of days of sick leave during one year before baseline, degree of sick leave, marital status, employment, born in Sweden, citizenship, salary, and sick leave diagnosis included as potential determinants. The procedure provides hazards ratios, 95% confidence intervals and Wald’s χ2, the latter, being the test parameter and computed with one degree of freedom for all variables regardless of grading, and therefore used as measure of determinant impact on outcome.

The analyses were performed in two steps. First, orienting bivariate anal-yses were performed, one for each potential determinant, followed by multi-variate analyses. In the latter, variables with mutually exclusive responses, such as marital status and sick leave diagnoses, were analysed with dummy variables. For marital status, being married, which had the hazard ratio clos-est to 1, was chosen as reference for the effect of marital status, and regard-ing sick leave diagnoses, respiratory system disease (ICD-10 code J, (in most cases upper respiratory tract infections) had the same characteristics and was used as reference for the effects of the diagnoses. To arrive at interpretable hazard ratios, number of sick leave days during one year before baseline was recomputed as number of weeks, and age as five-year age groups.

The analyses were performed straightforwardly with sex as determinant as well as stratified for sex. In the latter, the results for women and men were similar. For this reason only results from the former type of analysis are shown.

The content of Figure 5 was computed with proportional hazards regres-sion technique using the same multivariate analysis model as above. Data on the effects on sick leave conclusion of age, sick leave days during one year before baseline, having a psychiatric diagnosis (ICD-10 code F) having a respiratory disease variable was also obtained from the analysis model. Re-garding age data for persons aged 20, 30, 40, 50, or 60 years, and for sick

26

leave during one year before baseline days 0, 28, 90, and 180 were obtained from the model. The specific days were chosen accordingly to the Swedish Social Insurance Agency as times for instituting rehabilitation. Regarding the diagnoses data on those with a diagnosis of F or J, respectively, and those with no such diagnosis were also obtained from the analysis model.

The degree of explanation, i.e. how well the determinants could explain sick leave conclusion, was measured with nominal logistic regression across the follow-up period during the first 7, 14, 28, 90, 180, 270, 365, 455, and 545 days, using concordance index, an estimate of the area under the receiver operating curve characteristic (ROC) curve (118), and a standard option in SAS ‘Logist’ procedure, and with the same variable set up as in the propor-tional hazards regression analysis model (Figure 6).

Paper II

The manual model Based on baseline sick leave information together with an element of intui-tion of the assessor’s experience (gut feeling) as well, the Rehabilitation team I classified the sick listed patients into a high-to-moderately-high risk group (n=447) or a low risk group (n=496) for not concluding sick leave within expected time.

The manual based classification was based on sick leave days during one year before baseline of more than 28 days, sickness certified at baseline be-cause of a psychiatric disease (F) or a musculoskeletal disease (M), being unemployed, being more than 45 years old, being a woman, and degree of sick leave. The three first factors were given a high weight, the fourth somewhat less weight, and the two last were valid only in borderline cases as reinforcement of risk. Typically, a woman aged 55, unemployed and with sick-listing periods in the previous year for a diagnosis of fibromyalgia, would be categorised as high risk, while a man, 37 years old, employed, and presently having a full sick leave might be categorised as low-risk. However, if this patient had many days of sick leave last year with a diagnosis of stress, he might be classified as high risk (Figure 2).

The manual model was analysed using Cox’s analysis with sick leave conclusion and the day from baseline when it occurred as dependent varia-bles and high or low risk group assignment as the independent variable.

27

Table 2. Sick leave diagnosis according to the International Classification of diseas-es (ICD-10) and diagnostic weights (-2 - +3) from proportional hazards regression analysis.

ICD-10 code Diagnosis Diagnostic weight

F Mental and behavioural disorders -2

G Diseases of the nervous system -2

I Diseases of the circulatory system -1

K Diseases of the digestive system -1

M Diseases of the musculoskeletal system and connective tissue

-1

S Injuries and poisoning 0

R Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified

0

E Endocrine, nutritional and metabolic diseases 0

D Diseases of the blood 0

C Neoplasm 0

A-B Certain infectious and parasitic diseases + 1

O Pregnancy, childbirth and the puerperium + 1

L Diseases of the skin and subcutaneous tissue + 2

N Diseases of the genitourinary system + 2

H Diseases of the eye and ear + 3

J Diseases of the respiratory system + 3

The computer-based model The computer-based model was also based on the proportional hazards re-gression technique in its multivariate form. Independent variables were the same significant determinants for sick leave conclusion, as in Paper I, i.e., age at baseline, sex, sick leave days one year before baseline, sick leave di-agnosis, degree of sick leave, and occupational status. Sick leave conclusion and the day from baseline when it occurred was entered as dependent varia-bles. All sick leave diagnoses were used as demonstrated in Table 2.

In order to facilitate the use of sick leave diagnoses in the analyses the di-agnoses were given weights. These were obtained proportional hazards re-gression analyses with sick leave conclusion and the time when it occurred as outcome (y-variables) and single sick leave diagnostic groups (letter), one at a time, and age and sex as exposure (x-variables). Based on these analyses the diagnostic groups were ranked in order of importance for rapid sick leave

28

conclusion and given weights, -2 indicating a protracted sick leave period and +3 a short one.

The proportionality assumption in both models was checked by the SAS ‘Lifetest’ procedure and found acceptable. During the analyses of the com-puter-based model there were signs of collinearity between occupational status, sick leave diagnosis, and number of sick leave days during the last year. After elimination of the variable with the weakest association with outcome, occupational status, no signs of collinearity remained.

From the Cox’s analysis of the computer-based model, data were obtained showing proportion of sick leave conclusion for combinations of age, num-ber of sick leave days one year before baseline and sick leave diagnosis groups on day 28, day 90 and day 180 from baseline. These data were en-tered into the nomograms in Figure 7. Degree of explanation was measured as agreement between the real (crude) outcome and outcome according to the analysis models with concordance c index derived from logistic regres-sion (118).

Paper III The high-risk group (n=170) for no sick leave conclusion within expected time passed an examination and assessment of medical needs and workabil-ity regarding a rehabilitation programme by the Rehabilitation team II, com-prising a physician, a physiotherapist, a social worker and an occupational therapist (Figure 3). After this procedure, recommendations regarding a re-habilitation programme from a medical point of view were given. In the next step there was a meeting between the Rehabilitation team II and stakeholders (non-medical clerks) from the local Social Insurance Agency, the local Em-ployment Agency and a social worker from the city of Eskilstuna, as rec-ommended by the Swedish Social Insurance Agency, since these authorities are or might be financially responsible for the rehabilitation. In this meeting an individually tailored rehabilitation and vocational programme with focus on sick leave conclusion was decided on and thereafter the rehabilitation process started. The 773 subjects in the moderately-high-to-low-risk group received standard treatment as recommended by their general practitioner (Figure 3).

29

Figure 3. Flow-chart of assessment and intervention in Paper III.

The study was not randomised into an intervention and a control group, since it initially was not intended to be a scientific study. A post hoc control group, as similar to the high-risk group as possible was created by means of a propensity score. Matching control subjects to cases is usually limited to, at the most, three matching variables, unless a very large pool of potential control subjects is available. Propensity score, first proposed by Rosenbaum and Rubin in 1983 (119), implies that matching may be performed based on an unspecified number of matching variables that are weighed together.

A prerequisite for matching in this case was that the manual classification of individuals into a high-risk group versus a moderately-high-to-low-risk group was not perfect, as it usually is when no explicit weights are used.

30

When logistic regression was used to compute predicted risk based on the same variables that were used in the manual model, a substantial overlap of risk score was found, primarily between the high-risk and the moderately high-risk group.

This circumstance then allowed computation of a propensity score with binominal logistic regression using the rehabilitation (high-risk) group (code 1) and all others (code 0) as dependent variable, and age, sex, number of sick leave days last year, sick leave diagnosis, degree of sick leave, whether born in Sweden, whether Swedish citizen, employment status, salary and marital status as independent variables. Sick-leave diagnoses were weighted and grouped according to Paper II. In this way all variables entered into the lo-gistic regression, assumed to carry a risk for not concluding sick leave at expected time, were collected into one measure; a propensity score.

Patients in the non-rehabilitation group were then matched to subjects in the rehabilitation group by propensity score to form potential control groups. Mean (SD) propensity score in the rehabilitation group was 0.6977 (SD 0.1866), in the first matched potential control group 0.6638 (0.1669), in the second control group 0.5721 (0.1062), in the third control group 0.5258 (0.0865), and in the fourth control group 0.1855 (0.1243). The propensity scores of the first and second control groups were thus fairly similar to the score of the rehabilitation group, and were combined into a common control group. Thus, the rehabilitation group (n=170) and the control group (n=340) constituted the study population of this report.

Simple differences between the rehabilitation and the control group were tested with Student’s t-test for continuous variables and the χ2-test for dis-crete variables. Follow-up time from baseline to sick leave conclusion or end of follow-up was measured as number of days from baseline. The hazard rate in the rehabilitation and control group was computed with the SAS ‘lifetest’ procedure. There were no close proportional hazards across the total follow-up time. Rehabilitation activities usually started after about 14 days, the me-dian duration of the programme was 77 days and the 75th percentile was day 112. Follow-up time was therefore divided into the pieces day 1-14 (when rehabilitation activities had not started), days 15-112 (when most rehabilita-tion activities were performed), days 113-365 days (when most rehabilitation activities were finished), and day 366-1095 (long-term follow-up). In each of these partial follow-up periods, there were approximately proportional haz-ards rates.

The effect of the vocational rehabilitation programme was evaluated with proportional hazards regression, one analysis for each partial follow-up peri-od, where conclusion of sick leave and the time when it occurred were en-tered as dependent variables, and group allocation was entered as the inde-pendent variable and so was individual propensity scores to further adjust for the potential remaining risk differences between the groups. Moreover, being unemployed, being disability pensioner, or having activity support during the

31

first 30 days after the sick leave period was concluded, were also entered as independent variables (co-variates) to further adjust for group differences.

To test the changing hazard rates during follow up, a proportional hazards regression analysis was preformed for each follow-up time period with sick leave conclusion and the time when it happened entered as dependent varia-bles and group (rehabilitation or control) and propensity score entered as independent variable, adjusted for status after sick leave conclusion (return to work, unemployment, disability pension, and activity support).

The analysis provided hazards ratio (rehabilitation group versus control group) and its 95% confidence limit, Wald’s χ2 (a measure of exposure im-pact on outcome), and p-values. All tests were two-tailed, and the signifi-cance level was set at p < 0.05.

Paper IV After the classification by Rehabilitation team I, patients in the high-risk group (n=170) passed a standard medical examination within 1-3 weeks after baseline, performed by the Rehabilitation team II (Figure 4). The patients were informed about the study prior to the medical examinations and asked to participate by completing the Örebro Musculoskeletal Pain Screening Questionnaire (116). After informed consent was obtained, an independent physiotherapist provided the questionnaire to be completed by the patient at the Primary Health Care Centre. The same physiotherapist collected the completed questionnaire and the responses were kept unknown to the team. The Rehabilitation team II was thus blinded to the patients’ responses.

Rehabilitation team II assessment After medical examination the Rehabilitation team II discussed weekly their findings and in a joint discussion assessed the patients’ possibility for sick leave conclusion within 6 months. The patients in the high-risk group (n=170) were classified into five levels of “rehabilitation potential” by the team. The levels where numbered 1-5 where 1 meant very low potential and 5 meant very high potential for sick leave conclusion within 6 months.

The rehabilitation intervention thereafter was individually tailored and could for instance include medication, physiotherapy, and cognitive behav-ioural therapy in addition to various vocational measures taken by one or several stakeholders from the authorities i.e., the local Social Insurance Agency, the employer, the Employment Agency or a social worker from the city of Eskilstuna (Figure 4).

In the questionnaire from which the patients’ assessment scores were ob-tained, the time frame for possible sick leave conclusion was set as the next 6 months (180 days). Since no specific time for sick leave conclusion was requested, nominal logistic regression might be a suitable analytical ap-proach with the outcomes sick leave 1=was concluded or 0=was not con-

32

cluded within 6 months. On the other hand, since exact time for sick leave conclusion was at hand proportional hazards regression with sick leave con-clusion and the time when it happened as outcome variables, might provide a higher precision in the analyses.

For this reason both regression methods were used. In all regression anal-yses the influence of age and sex was taken into account. The cumulative life table curves were computed based on proportional hazards regression.

Assessment by patients The Örebro Musculoskeletal Pain Screening Questionnaire, is intended for use in individuals who are experiencing regional pain problems affecting performance at work, taking repeated short spells of sickness absence or are currently off work and have been so for up to 12 weeks. It is a self-administered instrument containing 25 items. The reliability and validity of the questionnaire has been tested previously and found satisfactory, and so has the predictive validity (116).

There are 21 scored questions concerning attitudes and beliefs, behaviour in response to pain, affect, perception of work and activities of daily living. It can usually be completed in five minutes before the patient meet the health professional. The 21 questions are scored on a 0-10 scale, yielding a total score range of 0-210, low risk (-89), moderately risk (90-105) and high risk (106-201). A score of 105 and below has been found to predict, with 95% accuracy, those who will recover and, with 81% accuracy, those who will have no further sick leave, in the next 6 months. Prediction of long-term sick leave (>30 days within the next 6 months) was found to be 67% accurate. A score of 130 and above correctly predicted 86% of those who failed to return to work (116). This assists the clinician to apply interventions (including the use of activity programmes based on cognitive behavioural strategies) to reduce the risk of long-term-pain-related disability. Evidence indicates that these factors can be changed if they are addressed.

The patients completed the Örebro Musculoskeletal Pain Screening Ques-tionnaire. Question number 16, used for this study, was worded “in your estimation, what are the chances that you will be working within 6 months?”, was used. Responders marked one of the numbers (0−10) on a visual analogue scale (VAS) where 0 meant no chance at all and 10 meant very large chance. In this study, the VAS was then converted into a 5−level scale where 1 corresponded to 0−2 on the original VAS, 2 corresponded to 3−4, 3 corresponded to 5−6, 4 corresponded to 7−8, and 5 corresponded to 9−10.

33

Figure 4. Flow-chart of professionals’ and patients’ assessment in Paper IV.

Ethical considerations The study was approved by the Regional Ethics Review Board, Stockholm, Sweden, Dno 2008/980-31/3.

Categorization into riskgroups by assessment team based on any of 1) sickness absence last year >28 days 2) sick leave diagnosis ICD-10 codes M or F 3) unemployed [4) age >45 years, 5) female, 6) sick leave degree] + ”gut feeling”

Individually tailored rehabilitation and vocational programme in cooperation with authorities

All subjects on sick leave, n=943

High-risk group (n=170)

Moderately-high -low-risk group (n=773)

Completion of OMPSQ questionnaire by

patient

Medical examination by multidisciplinary rehabilitation

team

Professional assessment of rehabilitation potential

Patient assessment of rehabilitation potential

Follow-up of sick leave at 6 months

34

Table 3. Baseline characteristics of the study population (n=943). Women Men n Mean (SD)

or % n Mean (SD)

or % Age, years Mean 482 39.1 (12.04) 461 39.2 (11.66) Inter-quartile range 29-49 30-48 Employed,% 365 75.7 356 77.2 Born in Sweden,% 389 80.7 342 74.2 Swedish citizen,% 454 94.2 429 93.1 Marital status,% Never married 195 40.5 218 47.3 Married or cohabiting 183 38.0 162 35.1 Divorced 90 18.7 79 17.1 Widowed 12 2.5 2 0.4 Annual salary at baseline, € Mean 373 20,127 (5,386) 342 24,241 (8,195) Inter-quartile range 17,333-23,333 20,000-26,667 Sick leave before baseline, days Mean 55.1 (105.82) 48.8 (94.69) Inter-quartile range 0-49 0-35 Full-time sick leave at baseline 431 89.4 430 93.3 Sick leave diagnoses, % Musculoskeletal disease 150 31.1 152 33.0 Psychiatric disease 131 27.2 90 19.5 Respiratory disease 88 18.3 70 15.2 Injury, poisoning 17 3.5 34 7.4 Symptoms and signs 33 6.9 31 6.7 Infectious-parasite disease 12 2.5 20 4.3 Dermatology disease 9 1.9 9 2.0 Cardiovascular disease 8 1.7 10 2.2 Genitourinary system disease 7 1.5 8 1.7 Digestive system disease 5 1.0 13 2.8 Eye or ear disease 5 1.0 6 1.3 Nervous system disease 5 1.0 5 1.1 Endocrine-metabolic disease 4 0.8 3 0.7 Blood disease 3 0.6 1 0.2 Pregnancy, childbirth 1 0.2 0 - Miscellaneous 4 0.8 9 2.0

35

Results

Paper 1 Characteristics of the study population Slightly more than half the subjects were women, mean and median age were 39 years, more than 75% were employed, more than 93% were Swe-dish citizens, 44% were never married and somewhat more than one third were married, mean annual income was 22,000€, mean number of days of sick leave during the year preceding baseline was 51, and the vast majority were on full-time sick leave (Table 3). The most common sickness certifica-tion diagnoses in rank order were musculoskeletal disease, psychiatric dis-ease, and respiratory system disease. Half of the study population returned to work within 14 days after baseline, and after three years only 15 subjects were still on sick leave.

Potential determinants of sick leave conclusion The effects of potential determinants for sick leave conclusion are shown in Table 4. In bivariate analyses previous sick leave, age, present part-time sick leave, being divorced, being born in Sweden, and being a Swedish citizen all decreased the probability of sick leave conclusion, while male sex, never being never married, and being employed increased it. Annual salary had no significant influence.

Among the diagnoses some were positively and other negatively associat-ed with the probability of sick leave conclusion. The most influential ones were respiratory system disease, infectious disease, eye or ear disease, and unspecific symptoms and signs that increased probability, while psychiatric and musculoskeletal disorders decreased the probability of sick leave con-clusion.

In multivariate proportional hazards analysis previous sick leave, age, present part-time sick leave, and being employed kept their significance for sick leave conclusion, while marital status, male sex, being born in Sweden, citizenship, and annual salary did not. Among the diagnoses psychiatric and musculoskeletal disorders, cardiovascular diseases, unspecific symptoms and signs, nervous system disease, digestive system disease, injury and poison-

Tabl

e 4.

Sic

k le

ave

conc

lusi

on d

eter

min

ants

.

B

ivar

iate

ana

lysi

s M

ultiv

aria

te a

naly

sis

H

R

95%

CI

Wal

d’s χ

2 p

HR

95

%C

I W

ald’

s χ2

p

Prev

ious

sick

leav

e by

wee

k 0.

97

0.96

-0.9

7 15

0.5

<0.0

001

0.97

0.

97-0

.98

92.0

<0

.000

1

Age

by

5-ye

ar g

roup

s 0.

91

0.88

-0.9

4 38

.6

<0.0

001

0.95

0.

92-0

.99

6.9

<0.0

1

Empl

oyed

1.

38

1.18

-1.6

2 16

.3

<0.0

001

1.28

1.

02-1

.59

4.7

<0.0

5

Part-

time

sick

leav

e 0.

57

0.45

-0.7

2 22

.4

<0.0

001

0.75

0.

58-0

.98

4.5

<0.0

5

Mar

ital s

tatu

s

mar

ried

1.10

0.

96-1

.25

1.8

0.17

1.

00

refe

renc

e -

-

neve

r mar

ried

1.16

1.

02-1

.32

5.1

<0.0

5

2.

3 0.

13

divo

rced

0.

74

0.62

-0.8

7 12

.7

<0.0

005

0.9

0.33

wid

owed

0.

61

0.35

-1.0

6 3.

1 0.

08

0.3

0.56

Bor

n in

Sw

eden

0.

82

0.70

-0.9

5 6.

6 <0

.05

1.5

0.22

Mal

e se

x 1.

16

1.02

-1.3

2 4.

7 <0

.05

1.3

0.26

Ann

ual s

alar

y by

10,

000 €

0.99

0.

90-1

.08

0.1

0.80

0.

7 0.

39

Swed

ish

citiz

ensh

ip

0.77

0.

59-0

.99

4.0

<0.0

5

0.

03

0.85

Dia

gnos

is

Res

pira

tory

dis

ease

2.

36

1.97

-2.8

2 88

.3

<0.0

001

1.00

re

fere

nce

- -

Psyc

hiat

ric d

isea

se

0.70

0.

60-0

.82

20.6

<0

.000

1 0.

31

0.25

-0.4

0 94

.9

<0.0

001

Mus

culo

skel

etal

dis

ease

0.

79

0.69

-0.9

1 11

.2

<0.0

01

0.41

0.

33-0

.51

67.6

<0

.000

1

Car

diov

ascu

lar d

isea

se

0.72

0.

45-1

.14

1.9

0.16

0.

30

0.17

-0.5

2 18

.4

<0.0

001

Sym

ptom

s and

sign

s 1.

33

1.03

-1.7

1 4.

7 <0

.05

0.53

0.

38-0

.75

12.8

<0

.000

5

Ner

vous

syst

em d

isea

se

0.68

0.

35-1

.32

1.3

0.26

0.

26

0.12

-0.5

7 11

.6

<0.0

01

Dig

estiv

e sy

stem

dis

ease

0.

82

0.51

-1.3

2 0.

7 0.

41

0.41

0.

24-0

.71

10.1

<0

.005

Inju

ry, p

oiso

ning

1.

12

0.84

-1.4

9 0.

6 0.

44

0.56

0.

39-0

.80

9.9

<0.0

05

Mis

cella

neou

s 1.

02

0.59

-1.7

7 0

0.93

0.

48

0.23

-0.9

8 4.

1 <0

.05

Endo

crin

e-m

etab

olic

dis

ease

1.2

9 0.

61-2

.71

0.4

0.50

3.

0 0.

08

Eye

or e

ar d

isea

se

1.83

1.

01-3

.31

3.9

<0.0

5

2.

8 0.

09

Infe

ctio

us-p

aras

ite d

isea

se

1.62

1.

14-2

.31

7.2

<0.0

01

2.2

0.14

Gen

itour

inar

y sy

stem

dis

ease

1.2

1 0.

73-2

.02

0.6

0.46

1.

5 0.

22

Blo

od d

isea

se

0.72

0.

27-1

.91

0.4

0.50

1.

0 0.

32

Der

mat

olog

y di

seas

e 1.

63

1.02

-2.6

1 4.

2 <0

.05

0.6

0.43

Preg

nanc

y, c

hild

birth

5.

53

0.77

-39.

52

2.9

0.09

-

- Po

tent

ial d

eter

min

ants

of s

ick

leav

e co

nclu

sion

with

in th

e ne

xt 1

095

days

38

Figure 5. Sick leave conclusion during three years of follow-up in groups according to age (A), number of sick leave days the year before baseline (B), sick leave for to psychiatric diagnosis (C), and sick leave for to respiratory disease diagnosis (D).

ing, and miscellaneous diseases were significantly associated with decreased probability for sick leave conclusion. The effects over time of age, previous sick leave, having a psychiatric diagnosis and having a respiratory system diagnosis are shown in Figure 5.

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600 700 800 900 1000 1100

Still

on

sick

leav

e, %

Follow-up time, days

A

60

50

40

30

20

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600 700 800 900 1000 1100

Still

on

sick

leav

e, %


C

Psychiatric disease diagnosis

No psychiatric disease diagnosis

39

For each 10 years of age from age 30 the sick leave conclusion was slow-er than in the previous decade. On the 100th day 7% of the 20-year olds were still on sick leave versus 25% of the 60-year old. Among those who had no sick leave days during the last year 8% were still on sick leave on the 100th day versus 35% of those with 180 sick leave days during the last year. Among those with no psychiatric diagnosis 11% were on sick leave on the 100th day versus 27% among those with such a diagnosis, and among those with an upper respiratory tract diagnosis 4% were still on sick leave on the 100th day as compared with 17% among those with other diagnoses.

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600 700 800 900 1000 1100

Still

on

sick

leav

e, %


B

180 days 90 days 28 days 0 days

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600 700 800 900 1000 1100

Still

on

sick

leav

e, %


D

No respitatory disease diagnosis

Respiratory disease diagnosis

40

Figure 6. Cumulative degree of explanation of twelve significant sick-leave conclu-sion determinants, given in ranking order according to importance (Wald’s χ2) and measured on days 7, 14, 28, 90, 180, 270, 365, 455, and 545. 1 = sick certified for psychiatric disease, 2 = number of sick leave days the year be sick certi-fied fore baseline, 3 = sick certified for musculoskeletal disease, 4 = sick certified for cardio-vascular disease, 5 = sick certified due to unspecified symptoms or signs, 6 = sick certified for nervous system disease, 7 = sick certified for digestive system disease, 8 = sick certified for injury or poisoning, 9 = age at baseline, 10 = employed, 11 = extent of sick leave, 12 = sick certified for miscellaneous disease.

Degree of explanation The degree of explanation across the first one and a half year of follow-up is shown in Figure 6. Using only the most influential variable, sick leave due to psychiatric diagnosis, the degree of explanation would be approximately 55% throughout the follow-up period. Including also the previous sick leave track record increased the degree of explanation to approximately 85%, and by including all other significant determinants degree of explanation ap-proached 88% - 90%.

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400 500 600

Deg

ree

of e

xpla

natio

n, %


+12 +11 +10 +9 +8 +7 +6 +5 +4 +3 +2 +1 Total

41

Table 5. Characteristics of the risk groups and effects of the risk grouping on return to work.

Low risk group High risk group

n Mean (SD) or %

HR (95% CI) n

Mean (SD) or %

HR (95% CI)

p for group difference

n 496 52.6 447 47.4

Age, years 38.2 (12.6) 40.2

(10.9) <0.005

Women, % 233 47.0 249 55.7 <0.01 Employed, % 440 88.7 281 62.9 <0.0001 Full-time sick leave, % 481 97.0 380 85.0 <0.0001 Sick leave >28 days last year, %

68 13.7 210 47.0 <0.0001

Sick leave diagnosis

Musculoskeletal disorder 111 22.4 191 42.7 <0.0001

Psychiatric disorder

59 11.9 162 36.2 <0.0001

Effect on return to work

Within 28 days 1.00 (–)

0.29 (0.24-0.34)

<0.0001

Within 90 days 1.00 (–)

0.32 (0.28-0.38) <0.0001

Within 180 days 1.00 (–) 0.35

(0.30-0.40) <0.0001

Paper II Sick leave conclusion according to the manual model According to the manual model those classified as being at high risk of not concluding sick leave as expected were significantly older, were women and unemployed to a larger extent, were on full-term sick leave to a lesser extent, had more often more than 28 days of sick leave last year, and had musculo-skeletal or psychiatric diagnoses to a larger extent than those at low risk (Table 5). The hazards ratio of concluding sick leave in the high risk group in rela-tion to the low risk group was 0.29 (0.24-0.34) after the first 28 days, 0.32 (0.28-0.38) during the first 90 days, and 0.35 (0.30-0.40) during the first 180 days of follow up, indicating 71%, 68% and 65% less sick leave conclusion during the three time periods in the high risk group then in the low risk group (Table 5). The agreement between the real (crude) outcome and out-come according to the analysis model was 76.0%, 74.0%, 73.2% for the three time periods, respectively.

42

Sick leave conclusion according to the computer-based model The results are shown in Table 6. During the first 28, 90 and 180 days in-creasing age, increasing number of sick leave days last year, and being on full-time sick leave, decreased the probability for sick leave conclusion. Moreover, there was a gradient among the sick leave diagnoses with a pro-gressively larger decrease of sick leave conclusion probability with more ‘heavy’ diagnoses. According to Wald’s χ2 the most influential determinants during all the three time periods were number of sick leave days last year and sick leave diagnosis followed by age, while degree of sick leave was of minor importance. The agreement between the real (crude) outcome and analysis model outcome was 86.1%, 88.0%, and 89.4%, approximately 10% units higher than in the manual model.

Table 6. Effects of possible determinants on return to work during three time peri-ods.

Return to work

HR 95% CI Wald’s χ2 p Concordance c index

First 28 days 86.1

Age, years 0.987 0.981-0.994 13.0 <0.0005

Sick leave last year, days 0.990 0.987-0.992 87.2 <0.0001

Full time sick leave 0.45 0.29-0.68 13.8 <0.0005

Sick leave diagno-sis

F or G (-2) 0.35 0.27-0.45 66.6 <0.0001

I, K, or M (-1) 0.48 0.38-0.60 44.7 <0.0001

all other dia-gnoses (0) 0.59 0.46-0.75 17.2 <0.0001

A, B, O, L (+1) 0.69 0.50-0.97 4.7 <0.05

N (+2) 1.00 0.56-1.81 0 0.99

43

First 90 days 88.0

Age, years 0.983 0.977-0.989 27.8 <0.0001




F or G (-2) 0.35 0.27-0.44 81.2 <0.0001

I, K, or M (-1) 0.44 0.35-0.54 61.4 <0.0001


A, B, O, L (+1) 0.67 0.48-0.92 6.0 <0.05

N (+2) 1.03 0.58-1.82 0 0.92

First 180 days 89.4

Age, years 0.982 0.976-0.988 32.9 <0.0001




F or G (-2) 0.33 0.27-0.42 91.9 <0.0001

I, K, or M (-1) 0.43 0.35-0.53 65.9 <0.0001


A, B, O, L (+1) 0.67 0.49-0.93 5.8 <0.05

N (+2) 0.99 0.56-1.74 0 0.96

The proportion of the study population concluding sick leave in relation to age at baseline, days of sick leave, and sick leave diagnosis is shown in Ta-ble 6. During the first 28 days sick leave conclusion ranged from 10% among the oldest subjects with the most ‘heavy’ diagnoses (F and G), and a long track record of previous sick leave, to 98% among the youngest with the ‘lightest’ diagnoses (H and J) and no sick leave during the last year.

Figure 7 also shows sick leave conclusion after 90 and 180 days. On day 90 from baseline the sick leave conclusion ranged from 17% among the old-est with the most ‘heavy’ diagnoses (F and G), and a long track record of previous sick leave, to 100% among the youngest with the lightest diagnoses (H and J) and no sick leave during the last year. The corresponding range of proportions on day 180 was 23% to 100%.

45

Figure 7. Nomograms showing proportion (%) return to work on day 28, 90, and day 180 from baseline. Age is attained age at baseline, sick leave days during the last year are the total number of days of sick leave, and ICD-10 diagnostic code are the ICD-10 chapter codes (F=psychiatric disease, G=nervous system disease, I=cardiovascular disease, K=gastrointestinal disease, M=musculoskeletal disease, A=infectious disease, B=parasite disease, O=obstetric disorders, L=dermatology disease, N=genitourinary system diseases, H=eye or ear disease, J=respiratory disease. The return to work probability is obtained by drawing a line connecting the actual age and the number of sick leave days last year. Where the line crosses the bold vertical line in the middle of the nomogram, the return to work probability (%) is found by moving horizontally to the sick leave diagnosis group. For example, the proportion returning to work within the first 28 days in an 54-year-old subject with 73 sick leave days last year, and an F diagnosis is approximately 38% and 78% if the diagnosis was J. The corresponding proportions at day 90 are approximately 56% and 92%, respectively.

Paper III The demographic variables showed no significant differences between the groups (Table 7). However, among sick leave data there were consistent significant differences, generally favouring the control group. These differ-ences are reflected in the propensity score differences. During the first 14 days, 21 sick-listings in the rehabilitation group were concluded, yielding a hazard rate of conclusion of 9.4% at the start of the period and 5.5% at the end, or a change of hazard rate of -41.7% (Table 8).

46

Table 7. Characteristics of the study population in Paper III. Study groups

Rehabilitation group Control group

n Mean (SD) or % n Mean (SD) or % p

N 170 340 Age at baseline, years 41.1 (10.4) 39.8 (11.8) n.s. Male sex, % 74 43.5 170 50.0 Born in Sweden, % 135 79.4 259 76.2 n.s. Swedish citizen, % 161 94.7 320 94.1 n.s. Marital status, % n.s. Never married 63 37.1 143 42.1 Married/Cohabiting 57 33.5 123 36..2 Divorced 47 27.7 68 20.0 Widowed 3 1.8 5 1.5 Sick leave days last yr 113.4 (132.2) 57.4 (108.9) <0.0001 Present sick leave Sick leave diagnosis*) =0.01 Score <0, % 148 87.1 254 74.7 Score ≥0, % 22 12.9 86 25.3 Full-time, % 138 81.2 304 89.4 =0.01 No. of days <0.0001 Mean (SD) 268.9 (316.9) 88.4 (191.4) Median 117.5 17.0

Sick leave completed at end of follow-up, % 162 95.3 336 98.8 =0.03

Propensity score 0.70 (0.19) 0.62 (0.15) <0.0001 *) Based on separate analyses the sick leave diagnoses were given weights according to their association with duration of the sick leave period, low weights indicating protracted sick leave period. ICD-10 codes F (psychiatric disorders) and G (neurological disorders) were given the weight –2, codes I (cardiovascular disorders), K (gastrointestinal disorders), and M (musculo-skeletal disorders) weight –1, codes A and B (infectious disorders), O (obstetric disorders), and L (dermatological disorders) weight +1, code N (uro-genital disorders) weight +2, codes H (ophthalmologic or otology disorders) and J (pulmonary disorders) weight +3, and all other diagnoses code 0. In the control group the corresponding numbers were 155 conclusions among the 340 patients yielding a hazard rate of 40.8% at the start of the period and 31.9 at the end, yielding a hazard rate change of -21.8%. The rehabilitation group thus had a faster decline than the control group of pa-tients remaining on sick leave.

During the next period, days 15-112, the hazard rate change was -41.8% in the rehabilitation group, and -85.8% in the control group. In the third

Tabl

e 8.

Haz

ard

rate

ana

lysi

s of s

ick-

listin

g co

nclu

sion

in th

e re

habi

litat

ion

grou

p an

d th

e co

ntro

l gro

up.

R

ehab

ilita

tion

grou

p C

ontro

l gro

up

Follo

w-u

p,

days

Sick

-list

ed

at st

art o

f pe

riod,

n

Sick

-list

ings

co

nclu

ded

durin

g pe

riod,

n

HR

, %

Sick

-list

ed

at st

art o

f pe

riod,

n

Sick

-list

ings

co

nclu

ded

durin

g pe

riod,

n

HR

, %

Star

t of

perio

d E

nd o

f pe

riod

Cha

nge

Star

t of

perio

d En

d of

pe

riod

Cha

nge

1-14

17

0 21

9.

40

5.48

-4

1.70

34

0 15

5 40

.78

31.9

0 -2

1.78

15

-112

14

9

61

5.48

7.

77

-41.

79

185

124

31.9

0 4.

54

-85.

77

113-

365

88

40

6.89

2.

92

-57.

62

61

37

4.84

2.

75

-43.

18

366-

1095

48

40

3.

04

2.61

-1

4.14

24

20

6.

21

2.63

-5

7.65

*)

HR

=haz

ard

rate

, per

cent

sick

-list

ings

con

clud

ed p

er d

ay

Tabl

e 9.

Effe

ct o

f reh

abili

tatio

n ve

rsus

no

reha

bilit

atio

n on

sick

leav

e co

nclu

sion

.

Con

ditio

nal p

ropo

rtion

al h

azar

ds re

gres

sion

ana

lysi

s St

atus

whe

n pr

esen

t sic

k le

ave

perio

d en

ded.

Follo

w-u

p,

days

n1)

Pa

ram

eter

es

timat

e (S

D)

Wal

d’s

χ2 H

R2)

95

% C

I p

Ret

urn

to

wor

k,

R/C

3) %

Une

mpl

oy-

men

t, R

/C3)

%

Dis

abili

ty

pens

ion,

R

/C3)

%

Act

ivity

su

ppor

t, R

/C3)

%

1-14

17

6 0.

55 (0

.24)

5.

1 1.

73

1.08

-2.7

9 <0

.05

90.5

/82.

6 9.

5/16

.8

0/0.

7 0/

0 15

-112

18

5 -0

.72

(0.1

7)

19.0

0.

49

0.35

-0.6

7 <0

.000

1 67

.7/7

9.9

29.0

/17.

7 3/

1.6

0/0.

8 11

3-36

5 77

-0

.21

(0.2

4)

0.8

0.81

0.

51-1

.29

0.37

42

.5/6

4.9

37.5

/21.

6 20

.0/1

3.5

0/0

366-

1095

60

-0

.07

(0.2

8)

0.06

0.

93

0.54

-1.6

0 0.

80

25.5

/58.

34)

23.4

/12.

5 48

.9/2

9.2

2.1/

0 0-

1095

49

8

52.9

/77.

95)

27.1

/17.

44)

19.4

/4.4

5)

0.6/

0.3

1) n=

num

bers

con

clud

ing

sick

leav

e du

ring

the

time

perio

d 2)

Haz

ards

ratio

, adj

uste

d fo

r mat

chin

g, p

rope

nsity

scor

e, a

nd st

atus

at t

he e

nd o

f the

si

ck

leav

e pe

riod

(ret

urn

to

wor

k,

unem

ploy

men

t, di

sabi

lity

pens

ion,

ac

tivity

su

ppor

t an

d so

cial

se

curit

y su

ppor

t).

3)R

/C=r

ehab

ilita

tion

grou

p

ve

rsus

con

trol g

roup

. 4) p

<0.0

5 5)

p<0

.000

1.

Tabl

e 10

. Int

erve

ntio

n st

udie

s with

out

com

e re

turn

to w

ork

(RTW

). RC

T= ra

ndom

ised

con

trol

led

tria

l. St

udy

Stud

y de

sign

In

terv

entio

n Fo

llow

up

Out

com

e H

aldo

rsen

et

al.

[120

] 20

02, N

or-

way

RC

T, in

subj

ects

with

a jo

b, si

ck-li

sted

≥8

wee

ks b

ecau

se o

f mus

culo

skel

etal

pai

n (n

=654

) sub

divi

ded

into

goo

d, m

ediu

m o

r po

or p

rogn

osis

for R

TW

Usu

al tr

eatm

ent (

n=26

3), l

ight

m

ulti-

disc

. (n=

222)

or e

xten

-si

ve m

ultid

isci

plin

ary

(n=1

69)

14 m

onth

s A

ll tre

atm

ents

wer

e eq

ually

goo

d in

goo

d pr

ogno

sis g

roup

, Lig

ht a

nd e

xten

sive

tre

atm

ents

wer

e m

ost e

ffec

tive

in m

ediu

m

prog

nosi

s gro

up, E

xten

sive

trea

tmen

t was

m

ost e

ffec

tive

in p

oor p

rogn

osis

gro

up

Ane

ma

et

al. [

122]

20

07, T

he

Net

herla

nds

RC

T in

low

bac

k pa

in su

bjec

ts si

ck li

sted

2-6

w

eeks

. W

orkp

lace

inte

rven

tion

(n=9

6) v

s usu

al c

are

(n=1

00).

for 8

w, t

hen

rand

omis

atio

n to

gr

aded

act

ivity

vs u

sual

car

e (n

=55,

57)

12, 2

6 an

d 52

wee

ks

Wor

kpla

ce in

terv

entio

n ha

d ef

fect

. Gra

d-ed

act

ivity

had

reve

rsed

eff

ect,

and

com

-bi

ned

inte

rven

tion

had

no e

ffec

t.

Bog

efel

dt e

t al

. [11

7]

2008

, Sw

e-de

n

RC

T in

sick

-list

ed lo

w b

ack

pain

subj

ects

(n

=160

) U

sual

car

e (n

=71)

vs m

anua

l th

erap

y (n

=89)

2

year

s Fa

ster

RTW

in m

anua

l the

rapy

gro

up

durin

g fir

st 1

0 w

eeks

of f

ollo

w-u

p, fr

om

then

on

no si

gnifi

cant

diff

eren

ces b

e-tw

een

the

two

grou

ps

Bül

tman

n et

al

. [12

1]

2009

, D

enm

ark

RC

T of

subj

ects

sick

list

ed 4

-12

wee

ks d

ue to

m

uscu

losk

elet

al d

isor

ders

, Int

erve

ntio

n gr

oup

(n=6

6), c

ontro

l gro

up (n

=47)

Coo

rdin

ated

tailo

red

wor

k re

habi

litat

ion

1 ye

ar

Inte

rven

tion

grou

p ha

d fe

wer

day

s of s

ick

leav

e th

an c

ontro

l gro

up d

urin

g 12

m

onth

s of f

ollo

w-u

p

Joha

nsso

n et

al.

[102

] 20

12, S

we-

den

RC

T of

subj

ects

at r

isk

for l

ong-

term

sick

ness

ab

senc

e. In

terv

entio

n gr

oup

(n=2

1), c

ontro

l gr

oup

(n=2

4) a

nd o

bser

vatio

nal s

tudy

with

a

reha

b gr

oup

n =1

076

and

cont

rols

n=3

7938

Early

mul

tidis

cipl

inar

y pr

o-gr

amm

e

1 ye

ar

Inte

rven

tion

prol

onge

d si

ckne

ss a

bsen

ce

spel

ls b

y 3

mon

ths

Car

lsso

n et

al

. [10

3]

2013

, Sw

e-de

n

RC

T in

subj

ects

sick

list

ed <

4 w

eeks

for

psyc

hiat

ric a

nd m

uscu

losk

elet

al d

iagn

oses

. R

ehab

ilita

tion

grou

p (n

=18)

vs c

ontro

l gro

up

(n=1

5)

Early

mul

tidis

cipl

inar

y as

-se

ssm

ent w

ithin

one

wee

k in

re

habi

litat

ion

grou

p vs

usu

al

care

in th

e co

ntro

l gro

up

3 m

onth

s , 1

ye

ar

A la

rger

num

ber o

f ind

ivid

ual o

n si

ck

leav

e af

ter 3

and

12

mon

ths a

nd a

larg

er

num

ber o

f sic

k le

ave

days

in th

e re

hab

grou

ps a

s com

pare

d w

ith th

e co

ntro

l gr

oup

Tabl

e 10

con

tinue

d.

Stud

y St

udy

desi

gn

Inte

rven

tion

Follo

wup

O

utco

me

Ejeb

y et

al.

[123

] 201

4,

Swed

en

RC

T of

subj

ects

with

men

tal d

isor

ders

(n

=245

) A

lloca

tion

to C

BT

(n==

84),

mul

ti-m

odal

inte

rven

tion

(n=8

0), u

sual

car

e (n

=81)

2 ye

ars

No

sign

ifica

nt p

ositi

ve im

pact

s but

a

tend

ency

for w

orse

ning

eff

ect o

f mul

ti-m

odal

inte

rven

tion

or c

ogni

tive

beha

v-io

ural

ther

apy

as c

ompa

red

to u

sual

car

e Je

nsen

et a

l. [1

24] 2

013,

D

enm

ark

RC

T of

subj

ects

sick

-list

ed 4

-6 w

eeks

due

to

low

bac

k pa

in

Ran

dom

allo

catio

n (n

=325

) to

mul

tidis

cipl

inar

y in

terv

entio

n or

usu

al c

are

1 ye

ar

No

effe

ct o

f int

erve

ntio

n

Mar

tin e

t al.

[125

] 201

5 D

enm

ark

Stud

y of

subj

ects

sick

-list

ed 4

-12

wee

ks fo

r m

enta

l pro

blem

s. In

terv

entio

n gr

oup

recr

uite

d ce

rtain

wee

kday

s, co

ntro

l gro

up o

ther

day

s

Reh

ab g

roup

(n=

88) g

ot

coor

dina

ted,

tailo

red

wor

k re

hab,

con

trol g

roup

(n=8

0)

got u

sual

car

e

2 ye

ars

No

effe

ct o

f int

erve

ntio

n

Pres

ent

stud

y M

atch

ed p

rimar

y he

alth

car

e ba

sed

stud

y in

51

0 pa

tient

s sic

k-lis

ted

from

any

cau

se

Hig

h-ris

k gr

oup

(n=1

70) a

nd

prop

ensi

ty sc

ore

mat

ched

co

ntro

l gro

up (n

=340

)

3 ye

ars

Initi

al h

ighe

r deg

ree

of R

TW in

hig

h-ris

k gr

oup

befo

re in

terv

entio

n, h

ighe

r RTW

in

cont

rol g

roup

dur

ing

inte

rven

tion,

from

th

en o

n no

eff

ect.

Kär

rhol

m e

t al

. [12

6]

2008

, Sw

e-de

n

Mat

ched

two-

coho

rt de

sign

with

subj

ects

w

ith m

uscu

losk

elet

al a

nd p

sych

iatri

c di

agno

-se

s (n=

64) a

nd c

ontro

ls (n

=64)

. No

popu

la-

tion

base

Late

voc

atio

nal c

oord

inat

ed

reha

bilit

atio

n vs

usu

al c

are

6 ye

ars

Sick

leav

e w

as c

onst

ant d

urin

g si

x ye

ars

of fo

llow

-up

in th

e re

habi

litat

ion

grou

p,

incr

ease

d in

the

cont

rol g

roup

.

Suoy

rjö e

t al

. [95

] 20

09, F

in-

land

Mat

ched

two-

coho

rt de

sign

with

subj

ects

un

derg

oing

reha

bilit

atio

n an

d m

atch

ed c

on-

trols

with

no

reha

bilit

atio

n

Voc

atio

nal m

ulti-

disc

iplin

ary

reha

bilit

atio

n (n

=223

6) v

s no

ne in

con

trols

(n=8

944)

7 ye

ars

Inte

rven

tion

grou

p ha

d m

ore

sick

leav

e da

ys d

urin

g th

e 6-

year

follo

w-u

p

50

period, days 113-365, the corresponding numbers were -57.6% decline ver-sus -43.2%, and during the last period, days 366-1096, -14.1% versus -57.7%.

There was thus evidence that the rehabilitation group had a faster decline than the control group of sick leave conclusions in the first follow-up period, but these relations changed during the further follow-up. When sick leave was concluded the patient might return to work, or to unemployment, or be granted a disability pension, or receive activity support. As shown in the right hand panel of Table 9, ‘return to work’ dominated during the first time period and then successively became less prevalent, while ‘unemployment’ and ‘disability pension’ successively increased, and ‘activity support’ re-mained low in all time periods. Furthermore, there were moderate differ-ences between the groups, the only significant one being ‘return to work’ during the last period.

Effect of rehabilitation versus no rehabilitation on sick leave conclusion is shown in the left hand panel of Table 9. During the period 1-14 days the rehabilitation group had a significantly faster conclusion of sick leave than the control group (p<0.05). During the second period, days 15-112, the con-trol group had a faster conclusion rate than the rehabilitation group (p<0.0001), and in the third and fourth follow-up time periods there were no significant differences between the groups (p=0.37 and p=0.80).

Furthermore, a comparison was performed of 11 other studies and the present study, evaluating the effects of multidisciplinary rehabilitation and vocational intervention with outcome on return to work (Table 10). Four studies showed a better effect of intervention than of usual care (117, 120-122), two showed a better effect of usual care than of intervention (95, 102), and five showed equal effect of intervention and usual care (103, 123-126), similar to the present study.

Benefit status at three years after baseline Benefit status was registered for the three study groups at the end of the study period, day 1095, which showed that 35.7% had some kind of benefit (Table 11). It was possible receiving two (25 patients) or three (5 patients) different benefits on the same time.

Table 11. Benefit status at follow-up three years after baseline (n=943). Group n SB UB SS AS DP No B Retired Deceased B% High-risk 170 25 24 3 10 40 87 0 0 48.8 Mod. high-risk 277 29 37 4 37 35 159 2 2 41.8 Low-risk 496 43 71 9 28 14 359 3 4 26.6 All 943 97 132 16 75 89 599 5 6 35.7

SB = Sickness benefit, UB = Unemployment benefit, SS = Social support, AS = Activity support, DP = Disability pension, No B = No Benefit, Retired = Old age retirement at 65 years, B%= per cent having any benefit at follow-up three years after start. AC is benefit from the Employment Agency when taking part in various labour market activities.

51

Table 12. Characteristics of the study population in Paper IV. Women Men

n Mean (SD) or %

n Mean (SD) or %

p-value

N 96 56.5 74 43.5

Age at baseline, years 96 40.0 (11.0)

74 42.7 (9.5)

0.09

Born in Sweden 83 86.5 52 70.3 0.009 Education 0.80 Elementary school 31 32.2 23 31.1 High school/upper secondary school 54 56.3 39 52.7 University 10 10.4 10 13.5 Marital status 0.34 Never married 40 41.7 23 31.1 Married/Cohabiting 30 31.3 27 36.5 Divorced 24 25.0 23 31.1 Widowed 2 2.1 1 1.4 Employment 55 57.3 48 64.9 0.32 Sick leave full-time 75 78.1 63 85.1 0.25

Sick leave days last year 119.3 (135.5)

105.7 (128.4)

0.51

Sick leave diagnosis 0.10 Psychiatric 34 35.4 15 20.3 Musculoskeletal 50 52.1 43 58.1 Other 12 12.5 16 21.6 Expected (Scored) sick leave conclusion Patients’ view, score 93 3.4 (1.5) 71 2.7 (1.5) 0.007 Rehabilitation team’s view,score 96 3.4 (1.4) 74 3.8 (1.2) 0.06

Time to sick leave conclusion, days 96 116.9 (71.3)

74 101.5 (65.0)

0.038

Paper IV

In the high risk population slightly more than half were women, mean age was 41 years, 86.5% of the women and 70% of the men (p=0.009) were born in Sweden, approximately one third had compulsory education only, one third was married or cohabiting, and 60% had a job (Table 12). The most common sick leave diagnoses were psychiatric and musculoskeletal diagno-ses.

The Rehabilitation team II mean assessment score for expected sick leave conclusion within the next 6 months (range 1-5) was 3.6. The corresponding

52

mean for the patients’ assessment score was 3.4 among women and 2.7 among men (p=0.007). Time until sick leave conclusion was 117 days among women and 102 days among men (p=0.038).

The Rehabilitation team II assessment scores and the patient assessment scores were closely correlated. (Spearman correlation coefficient=0.49). This correlation is visualised in Table 13. In 62 (37.8%) observations patients’ and Rehabilitation team II’s scores were identical. In 114 (69.5%) observa-tions the scores agreed within a range of ±1 score.

Table 13. Patients’ assessment and team’s assessment of possibilities to conclude sick leave within 180 days. Team=assessment team.

Team assessment score

1 2 3 4 5 Total

Patients’ assessment score

1 11 7 12 7 5 42

2 2 4 7 2 3 18

3 0 5 10 7 8 30

4 0 4 9 5 8 26

5 1 2 6 7 32 48

Total 14 22 44 28 56 164

Further evidence of the close correlation between the scores is given in Figure 8, which shows cumulative sick leave conclusion curves for the vari-ous scores of patients’ assessment and of Rehabilitation team II assessment. The two types of assessment performed rather similar.

53

Figure 8. Conclusion of sick leave by follow-up time in relation to patients’ assess-ment scores (upper part) and rehabilitation team’s assessment score (lower part).

54

In Table 14 the results from the nominal logistic regression and the propor-tional hazards regression after taking the influence of age and sex into ac-count are shown. In the univariate binominal logistic regression analysis the Rehabilitation team II’s assessment score was highly significantly associated with sick leave conclusion within 180 days (OR=1.61, 95%CI=1.23-2.11, p=0.0005, Wald’s χ2=12.0, c=74.1%). In a similar analysis patients’ assess-ment score was also highly significantly associated with sick leave conclu-sion within 6 months (OR=1.60, 95%CI=1.25-2.03, p=0.0001, Wald’s χ2 =14.4, c=74.8%), thus indicating the importance of both assessments.

Table 14. Effect of patients’ and rehabilitation team’s assessment score on sick leave conclusion within 180 days adjusted for the influence of age and sex, in nomi-nal logistic regression and in proportional hazards regression.

Wald’s χ2

OR or HR 95% CI p

Univariate nominal logistic regression Rehabilitation team’s assessment score 12.0 1.61 1.23-2.11 0.0005 Patients’ assessment score 14.4 1.60 1.25-2.03 0.0001 Multiple nominal logistic regression Patients’ assessment score 14.4 1.60 1.25-2.03 0.0001 Rehabilitation team’s assessment score 2.0 – – 0.16 Univariate proportional hazards regression

Patients’ assessment score 13.0 1.29 1.12-1.48 0.0003 Rehabilitation team’s assessment score 12.3 1.37 1.15-1.63 0.0005 Multivariate proportional hazards regression

Patients’ assessment score 13.0 1.29 1.12-1.48 0.0003 Rehabilitation team’s assessment score 2.3 – – 0.13

In a multivariate logistic regression analysis patients’ assessment score were significant at the same level as in univariate analysis, while rehabilitation team’s score became non-significant. The proportional hazards regression analyses gave similar results. In univariate analyses the two scores were highly significantly associated with sick leave conclusion, while in multivar-iate analysis patients’ assessment score remained significant but not Rehabil-itation team II’s score.

Influence of gender in the assessments If the patients, both men and women, and the rehabilitation team scored an equally high score, the outcome at day 180 was correct for two out of three patients. This was also true for plus/ minus one step from the diagonal in Table 13. If equally low score, three out of four women were still on sick

55

leave at day 180 after baseline. However, among men less than 50% were on sick leave on day 180. If the scores were at least two steps from each other, one part scored high and the other low, then 50% had concluded sick leave at day 180. Scoring low, only 25% of women and 40% of men concluded sick leave at day180. Spearman rank correlation test showed a significant gender difference between the assessments, r=0.59 for women and r=0.43 for men.

The comparison between the patient’s and the rehabilitation teams’ as-sessment of chance to sick leave conclusion, with outcome whether on sick leave or not at day 180 after study start is visualized in Figure 9.

The Örebro Musculoskeletal Pain Screening Questionnaire, was complet-ed by 159 (11 missing 6.5%) patients in the high-risk group. Out of 89 wom-en, 14.6% scored low risk, 13.5% scored moderately risk and 71.9% scored high risk. The corresponding scores for 70 men were 5.7%, 18.6% and 75.7%.

Fi

gure

9. C

ompa

rison

bet

wee

n th

e w

omen

’ and

the

reha

bilit

atio

n te

ams’

ass

essm

ent a

nd th

e m

ens’

and

the

team

s’ a

sses

smen

t with

out

com

e at

da

y 18

0.

W

omen

M

en

No sick leave

Sick leave

57

Discussion

In this primary health-care study population of 943 patients on sick leave, a number of determinants easily accessible in the general practitioners’ office were found positively and negatively associated to a sick leave conclusion. Previous days of sick leave, and a psychiatric diagnosis F (ICD-10) and old-er age were the most influential determinants. When analysing the properties in two models for the assessment of sick leave conclusion after sickness certification, a computer-based assessment model had a higher precision and gave more detailed information compared to a manual model. Furthermore, based on the latter, three nomograms were constructed for assessment of the probability of sick leave conclusion. The efficacy of a multidisciplinary vo-cational programme in sick-listed high-risk patients for not concluding sick leave at the expected time, compared to matched-non-programme patients, showed a significant sick-leave conclusion in the rehabilitation group during the rehabilitation programme and no compensatory increase was observed during follow-up. Before the intervention programme started, the rehabilita-tion group concluded sick leave somewhat faster, however, during the inter-vention the control group had a faster sick leave conclusion, and after inter-vention there was no difference between groups. In a comparison of the as-sessment scores of a rehabilitation team and high-risk patients’ on the possi-bility to sick leave conclusion within 6 months, the scores were found to be highly associated.

Study design and methodological considerations This 3-year prospective population-based longitudinal study, originated from the actual working situation in the primary health-care centre in 2004, with an overwhelming number of patients on sick leave. Therefore, it was not possible to perform the optimal study design, a randomised controlled trial (RCT), since the study initially was not meant to be a scientific one. This is a limitation of this study; however, in conformity with research on sick leave in general, which shows a scarcity of RCT studies (127), the study popula-tion was large enough for its purpose, and it was based on all patients pass-ing through the ‘time window’ of data collection, making it equivalent to a random patient sample. In Sweden, the study population might be similar to a population at any primary health-care centre, including both sexes, differ-

58

ent ages, all diagnoses, employed and unemployed, and in this case it was sociodemographic equal to the population of the city Eskilstuna (128). How-ever, in 2004 there was about a 15% difference in the employment rate be-tween the study population and the population in Eskilstuna, aged 18-63 (n=1666), where 74% were employed (78% of men and 72% of women) (128). In Paper III, some sort of matching was necessary in order to obtain a control group as similar to the intervention group as possible. The matching procedure was based on the propensity score method, providing an individu-al score for not returning to work as expected, based on well-known and generally accepted risk factors. By adjustment in the analyses for the small remaining average differences in the propensity score between the interven-tion and control group, results similar to random allocation were obtained.

The exposure and outcome data used in the analyses in Paper I−III were obtained from the Swedish Social Insurance Agency sick leave database, and from medical records regarding the first two weeks of sick leave. Data was complete with no data loss in Papers I−III. In Paper IV 3.5% of data was missing.

The study was performed in 2000–2007 when there was no time limit for sickness benefit. In 2008 the benefit system changed to recommendations for length of sick-leave periods for different diagnoses. Would the new reg-ulation change the results? An evaluation in 2011 by the Swedish Social Insurance Agency shows that the sick-listing pattern has changed. The length of sickness absence has been shortened while the spread in length and the number of new sickness absence spells has been reduced since the introduc-tion of the recommendations (129). However, the three important determi-nants, − previous sick leave history, common mental health disorders (ICD-10 diagnosis F) and increasing age − remain described in the literature as significant risk factors for long-term sick leave (130, 131).

Would the outcome have been different with other criteria applied for sick-leave-free intervals? The results for sick leave conclusion during the first year may have been somewhat different, but probably not for the three years of follow-up, whereas patients at risk of long-term sickness absence show recurrent periods of sick leave over time (16, 17).

Return to work – sick leave conclusion The term ‘return to work’ is frequently used in the literature but differently defined and measured, depending not only on the actual study setting, but also on the scientific discipline e.g. medicine, psychology, sociology and economics (27). For employees it often means actual return to their present work in part- or full time, or in an altered work situation, or in new employ-ment. So far there is no clear agreement among researchers about how sick

59

leave and a successful return to work should be measured (27, 132). The return to work might be a long-lasting process, where the patient may be in several and recurrent states; for instance, receiving various social security benefits or working, and alternating between these states (133). In this study the term sick leave conclusion was introduced in order to clarify that the sick leave spell in effect at baseline was concluded. However, in Paper III the actual outcome, i.e., return to work, unemployment, disability pension or activity support, was used to describe the situation at sick leave conclusion (Table 9).

Discussion by findings Determinants In the analysis of possible determinants of sick leave conclusion and their relative impacts, the choice of determinants was primarily based on the fol-lowing three requirements; 1) the determinants should be of significant im-portance for risk of long-term sick leave and disability pension; 2) infor-mation on the determinants should be available from the sickness certificate; and 3) the determinants should be easy to use as a screening/assessment tool.

Out of a number of potential determinants identified as predicting long-term sick leave, easily available determinants from the sickness certificate were age, sex, sick leave diagnosis, employment status and degree of sick leave. The intention was to help the GP, in a time-limited consultation, to focus on necessary tasks such as medical history taking, examination, diag-nosis, tests and referrals. Therefore, the information on determinants should be easily available with no further searching in medical records or deeper medical history penetration.

However, information on sick leave during the year preceding the base-line examination had to be searched for in the patient medical records. This was also automatically done by the GP when issuing a sickness certificate in 2004 in this primary health-care centre. In the last few years, the sickness certification system has been changed to an Internet-based system by the Swedish Social Insurance Agency, which means a technical step has been taken towards easier access to sick leave data history for a specific patient, irrespective of where the sickness certificate was issued. However, the ac-cess for GPs to previous sick leave data is still limited to the GPs actual pri-mary health care centre.

In the multivariate analysis of all potential variables for delayed sick leave conclusion collected from the study population, sex had no significant influence, when other factors were taken into account. It was simply out-competed by variables with much stronger impact. Performing the analyses with sex as determinant or stratified to sex made no difference.

60

This may seem a somewhat surprising finding, since the female sex has been reported in other studies to be a significant risk factor for long-term sick leave and disability pension (44, 47), therefore it was included in the classification performed by the Rehabilitation team I. However, the analyses in most previous studies were generally bivariate, assessing the importance of one factor at a time. In that situation, female sex is a significant determi-nant, but in multivariate analysis it is out-competed.

In accordance with many other studies, psychiatric disease and musculo-skeletal disorders prolonged time to sick leave conclusion. However, other diagnoses also proved to be significant determinants, some facilitating, oth-ers delaying sick leave conclusion, which not having been previously report-ed at the time of the study was a novel finding. In a recent register study by Lidwall, on sick leave diagnosis and return to work, distinct differences were found across diagnosis within all diagnosis chapters (134). This finding is also applicable for mental disorders and musculoskeletal diseases, which are the most common causes of sick leave. The knowledge that some diagnoses, i.e. infectious respiratory disease, viral infections, eye or ear disease, and unspecific symptoms and signs increased the probability of sick leave con-clusion is of importance when considering length of sick leave, and whether costly rehabilitation measures are needed or not. In times of economic strain in public health care it is increasingly necessary to give priority to patients in need of support to conclude sick leave rather than spending time and effort on those patients who will conclude sick leave spontaneously.

It might be questioned whether the outcome would have been different if other determinants had been picked out? Are these determinants alone sig-nificant enough for the purpose of early prediction? The degree of explana-tion in the regression-based model, close to 90%, means that the possibility that any overlooked strong determinant might alone provide 10% explana-tion is unlikely (Figure 6).

The multivariate analysis confirmed that previous sick leave, age, present part-time sick leave, and being employed were significant determinants for sick leave conclusion. Positive effect of part-time sick leave compared to full-time sick leave is in line with a RCT study by Viikari-Juntura et al., who found a faster and more sustainable return to work among patients with mus-culoskeletal disorders (135). In addition, previous days of sick leave was one of the mot informative factors and the possible use of administrative sick leave data as a risk marker of future sick leave and disability pension was demonstrated in a Danish study (2), and confirmed in Dutch and British studies (17, 136). The significance of sick leave diagnosis ICD-10 F is not surprising, whereas since the 1990s, common mental health disorders in-creased, and they have a high degree of being recurrent and of long durations (12, 130). In Sweden 2011−2014, sick leave of longer than 60 days due to F diagnosis, was 37% for women and 27% for men. The findings regarding the determinants of previous days of sick leave, psychiatric diagnosis and age

61

are in line with previous studies (137, 138). However, their use in an easy-to-use assessment tool predicting time to sick leave conclusion, as in the present study, is unique.

Easy-to-use screening tool Determinants tested in the two models in Paper II were supposed to have different underlying properties, in spite of the fact that they basically relied on the same variables. The manual model mimicked the way GPs assess the risk of long-term sick leave, based on the presence or absence of known risk factors but also on intuitive elements. General practitioners often face com-plicated, vague problems in situations of uncertainty, which have to be solved in a brief consultation. In these situations, ‘gut feelings’ have been shown to play a significant role in the diagnostic process, also described as a third track beside medical decision-making and medical problem-solving (139). Furthermore, gut feelings are a specific form of intuition based on the interaction between patient information and the GP’s own knowledge and experience. In a qualitative Dutch GP study, two types of gut feelings were distinguished, a sense of reassurance, and a sense of alarm (140). The GPs explained the first as being the type of gut feeling that made them confident about the prognosis and therapy, although they might not have a clear diag-nosis in mind. The latter type of gut feeling was described as a feeling of something being wrong even though arguments for this circumstance were lacking. In another study the presence of gut feelings in diagnostic reasoning among GPs was measured with a validated questionnaire (141). In the man-ual model tested in the present study, the intuitive element − the gut feelings of the highly experienced rehabilitation team − was similar to a sense of alarm in cases classified into the high-risk group (‘something is wrong with the sick-listing course of this patient’).

The computer-based model was based on the standard method for the construction of nomograms. A priori, the manual model, being more flexible in its structure and perhaps catching important non-measurable information by its intuitive component, was assumed to perform at least as well or better than the computer-based model. However, the opposite turned out to be the case.

In both models sick leave conclusion was inversely related to age, number of sick leave days during the last year, and being on full-time sick leave, and increased across the diagnosis code groups from F and G to H and J. In the manual model female sex was used as a factor associated with a moderate delay in sick leave conclusion. In the computer-based model sex did not affect sick leave conclusion, since it was out-competed by other factors. The manual model provides only a yes/no grading of increased risk of delayed sick leave conclusion, while the computer-based model provided a graded probability of sick leave conclusion, ranging from only a few per cent for

62

those with the slowest sick leave conclusion up to 100% for those with the fastest.

The outcome in both models was the actually occurring sick leave conclu-sion. The numbers of subjects concluding sick leave according to the two models were the same as the actual concluded sick leave. The difference between the models refers to their precision in correctly identifying subjects that actually concluded sick leave and those that did not. The proportion of correctly identified patients concluding or not concluding sick leave was 73−76% with the manual model, indicating that a quarter of the subjects were misclassified. In the computer-based model 86−89% were correctly classified, the proportion misclassified thus being less than half that of the manual method.

The results in Paper IV indicate that one possibility of improving both models might be to add the patients’ own assessment. In order to maximise practical value, only predictive factors that are potentially available to the general practitioner during the first consultation were taken into account. Consequently, information that would require more effort than interviewing the patient or looking into medical records was disregarded. To our knowledge, no other study has focused on the time-limited first consultation when the GP needs a tool to assess the patients’ time to sick leave conclu-sion.

Multidisciplinary rehabilitation intervention Early multidisciplinary vocational rehabilitation decreased sick leave signifi-cantly in the rehabilitation group and there was no compensatory increase during the three years of follow-up. However, the matched controls receiv-ing treatment as usual had a faster sick leave conclusion during the period of intervention, and when the intervention was concluded there were no differ-ences in sick leave conclusion between the groups.

Delayed sick leave conclusion may be due either to time-consuming reha-bilitation measures organised by stakeholders or to programme failure, or both. Different stakeholders have different perspectives, agendas and budg-ets, which are not always compatible with each other. Improving our under-standing of the different perspectives of the various stakeholders appears to be a possible avenue for facilitating cooperation and commitment to the goal of work-disability reduction. In particular, given their central roles in the sick leave conclusion process, enhancing collaboration between the sick-listed and stakeholders is important.

Ekberg et al. recently investigated early (within three months), and late (between three and twelve months) return to work in employed patients on sick leave due to common mental disorders (142). One interesting finding was that no health measures were associated with return to work, but rather reflected associations with work conditions. This highlights the importance

63

of including workplace conditions and relations at the workplace in return to work interventions.

Another possibility for delayed sick leave conclusion might be the pro-longation of sickness certificates for the sole purpose of receiving rehabilita-tion measures from the Swedish Social Insurance Agency. At the time of the study, a prerequisite for access to rehabilitation measures was that the patient was on sick leave. Moreover, patients who previously were granted social benefits also needed a sickness certificate when ill but were given low priori-ty.

In Sweden, multidisciplinary vocational rehabilitation in cooperation with authorities is the recommended method to promote return to work. However, there is also evidence that interventions may prolong sickness absence. As pointed out by Johansson et al. (102) and Carlsson et al. (103) the rehabilita-tion team may focus more on rehabilitation than on encouraging the sick-listed individual to return to work, and the early intervention programme may therefore have a locking-in effect, i.e., actually prevent the sick-listed from returning to work. The results from the present study may be interpret-ed as being due to such an effect.

The comparison of intervention studies with outcome on return to work (Table 10), displays a diversity regarding study populations and interven-tions. There was a wide range of study populations, such as employees in various occupations, hospital-based patients, primary health-care patients, etc., with low back pain, neck pain, musculoskeletal pain, mental health problems and other disorders (117, 120, 122, 143). Furthermore, rehabilita-tion programmes are often carried out in special rehabilitation centres while vocational rehabilitation measures are mostly performed in the workplace. However, given the inconsistencies of published results from rehabilitation programmes, and the scarcity of intervention studies in similar primary health-care populations to those in this study, this individually tailored inter-vention may be a sensible working mode with existing resources.

Since 2008, when the Swedish government introduced financial initiatives to improve rehabilitation and sick leave management, a multitude of reha-bilitation programmes have been put forward. The results vary as to whether multimodal rehabilitation is an effective intervention programme for those on sick leave due to common mental health disorders and musculoskeletal pain disorders. The effects of a Swedish national programme (the Rehabilita-tion Guarantee) was analysed in a report by the Swedish Social Inspectorate in the Swedish region of Skåne (144). Cognitive behavioural therapy CBT) was provided to patients with light to moderate mental and behavioural dis-orders, and multimodal rehabilitation (MMR) for patients with musculoskel-etal-related pain in the back, neck and shoulders. The results showed that CBT reduced but MMR increased sickness absence regardless of the pa-tient’s status at start of the programme. Stigmar et al., found in a primary health-care setting, that patients with musculoskeletal disorders significantly

64

improved their health- related quality of life and functional ability after re-habilitation, especially those with no sick leave (145). However, the study was inconclusive as to what components were effective in the multimodal rehabilitation. Wåhlin et al. investigated the association between clinical and work-related interventions and return to work (146). In this questionnaire study of 699 newly sick-listed patients in the primary health-care or occupa-tional health service, a combination of interventions increased the probabil-ity of return to work for patients with mental disorders, but not for patients with musculoskeletal disorders. For the latter, better health, positive expecta-tions of return to work and better work ability was associated with return to work.

Research in recent years gives some understanding of the complexity of various issues to consider in sick leave conclusion interventions in the fu-ture. Some aspects that should be addressed are, for example, that learning from the patients’ negative experiences of being on long-term sick leave may be important (6, 7), that the sick listed are not to be considered as a homoge-neous group, in spite of having the same sick leave diagnosis (4, 134, 138, 147), and that the possible barriers or facilitators due to previous encounters with health care, employers, the Social Insurance Agency and other stake-holders should be removed (148, 149).

The Rehabilitation team II in the present study experienced similar find-ings to those in a qualitative study by Eriksson et al., i.e., that the unem-ployed fell between the cracks due to weak cooperation between the local Social Insurance Agency and the Employment Agency (150). Unemployed individuals on sick leave were declared to be too sick to work and were re-moved from the unemployment register. The Employment Agency also gave low priority to individuals difficult to find a job for, such as the unemployed on sick leave. In fact, a labour market problem was turned into a medical problem, often with long-term sick leave as a result.

Timing of intervention What is the perfect time for sick leave? Is there an optimum timing for an intervention to achieve sick leave conclusion? If so, when is that? When is it too late and is there maybe even a too early? There is no direct evidence on a ‘perfect’ timing of interventions. However, the concept of early interven-tion is central to vocational rehabilitation, because the longer anyone is off work, the greater the obstacles to sick leave conclusion and the more diffi-cult vocational rehabilitation becomes (9).

In a UK review report by Waddell, Burton and Kendall, it was found that in the first 3−6 weeks of sick leave, the likelihood of recovery and rapid sick leave conclusion is high, with or without healthcare intervention (98). The great majority of workers with common health problems concluded sick leave quickly with minimal healthcare and interventions may be limited to

65

good clinical care and workplace management and practice. It is mentioned that ‘formal vocational rehabilitation programmes might be unnecessary at this stage, because they are unlikely to have any significant impact on what is already a good natural history, and they are unlikely to be cost-effective’ (page 300) (98). In addition ‘They may even obstruct and delay natural re-covery due to a combination of prescribed inactivity, physical and mental deconditioning, ‘labelling’ and ‘attention’ effects that may encourage illness behaviour, and delaying reactivation’ (page 300) (98). Furthermore, the authors conclude that the priorities in the beginning are to support and en-courage restoration and function, including remaining at work or early sick leave conclusion, and to avoid iatrogenic disability. In workers, after 6 weeks of sickness absence the risks of long-term incapacity increased by 10−20%, and after 6 months, there was only a 50% chance of returning to a previous job (98). After one year on incapacity benefit, the probability of remaining on benefits for years or until retirement was significant.

The present study intended to evaluate an early intervention, which for most patients in the rehabilitation group, meant the earliest start of interven-tion was 15 days after inclusion. The results give no answer as to whether this is the best time for intervention due to the lack of difference to the matched-control group in the follow-up time 113-1096 days after inclusion. In a study focused on the timing of rehabilitation, Marnetoft and Selander investigated 612 long-term sick leave cases (90 days or more) regarding long-term effects of early versus delayed vocational rehabilitation. The out-come was no social benefit or lowered benefit. In a four-year follow-up, early intervention was found more effective than late, but only for women, and more so for younger women than for the older women (151). The time before the start of rehabilitation decreased the probability of successful reha-bilitation.

Furthermore, there is support in the literature that the timing of interven-tions for those in employment needs to begin early enough to prevent the transition to disability pension, and a ‘window of opportunity’ for effective clinical and vocational management seems to be between 1 and approxi-mately 6 months off work (98). Within this critical window, earlier interven-tions are likely to be simpler, more effective, cheaper and more cost-effective. For people with common health problems, the deleterious effects of further time out of work and delaying intervention are likely to outweigh and deadweight savings.

The lack of good evidence of effective interventions for patients on sick leave for more than 1 year reinforces the need to intervene before individuals become trapped on benefits. Thus, there is still reason to emphasise early involvement in a case to guarantee that the timing of rehabilitation measures is not missed.

66

Patients’ and rehabilitation team’s assessment The close correlation between the patients’ assessment score and the team’s assessment score for expected sick leave conclusion within 6 months, is in line with a study by Sandström and Esbjörnsson in 1986 (152). They found that patients with chronic low-back pain, in a follow-up at 1 and 4 years, predicted the outcome of a vocational rehabilitation programme correctly in 69%, with a sensitivity of 68% and a specificity of 71%. There was a signifi-cant correlation between the patients’ prediction and the recommendations given by the rehabilitation unit.

One interesting finding in this study is that the men assessed their chance to sick leave conclusion as being significantly lower than the team did (Fig-ure 9). The men seemed to have poor self-confidence or maybe they were more realistic about their chances to conclude sick leave due to experiences from the labour market? The rehabilitation team on the other hand, were maybe realistic in a different way, as they were well aware of the possible rehabilitation measures that could be used. However, when the outcome was negative, the men predicted more correctly than the team.

Different aspects of the patients’ own assessment on return to work indi-cates that the patients’ perception might be the best predictor of return to work (107, 153-155). In a review by Mondloch et al. 2001, all 16 included studies showed that a patient’s expectations had an influence, differing from large, moderate to small, on how well they recovered (156). Expectations were found to influence the recovery of all patients, regardless of differences in things such as the severity of their condition, their social position or men-tal and physical health.

Patients’ assessment of self-efficacy, based on Banduras’ theory − the be-lief that one can achieve what one sets out to do (157) − has been investigat-ed and found to have a significant impact as a predictor of sick leave conclu-sion (158-160). Lagerveld et al. developed a questionnaire, ‘the return to work self-efficacy scale’, specifically measuring work-related self-efficacy for workers on sick leave with mental disorders (161). In a recent study by Brouwer et al. evidence was found supporting the predictive validity of the ‘return to work self-efficacy scale’ within 12 months after musculoskeletal injury (162) and Volker et al. found the scale to be a predictor of time to return to work in sick-listed employees with all- cause sickness absence (163).

Another possible tool for patient assessment on sick leave conclusion might be self-rated health, which has been shown to be associated with the number of days on sick leave, of being granted a disability pension, and with mortality (164-166). In a study by Post et al., self-rated health was found to predict return to work for employees on long-term sickness absence due to musculoskeletal or other physical health complaints, and psychological complaints (167). Worse subjective severity of complaints was associated

67

with a low return to work rate, while a faster return to work rate was associ-ated with better physical and social functioning, and a better perception of general health. Josephson et al. found that good self-rated health and work ability were associated with full-time return to work in a study of public-sector employees in Sweden (168). This might also be a way of better under-standing the patients’ perspective (169). In addition the SF-36 bodily pain scale might be helpful in identifying patients at risk of long- term sick leave (170), and the screening instrument Balansmeter was found to predict sick-ness absence within employees with depressive complaints (171).

In this study one single question seemed to serve the purpose well in ob-taining the patients’ perception on sick leave conclusion. Maybe one single question gives enough valid prediction to be used as a fourth item (patients’ own assessment) in an easy-to-use assessment tool? Bowling discusses dif-ferent aspects of shorter instruments versus longer measures, and asks ‘if one question works, why ask several?’ (172). There is positive support to this suggestion in the thesis by Heijbel, based on the HAKul study on employees in four county councils and local authorities in six municipalities in Sweden, who concludes that the most important finding was the impact of the sick-listed individuals’ own perception of their future sick leave conclusion and that only one question was required (173). Another tool is the Work Ability Index (WAI), long since used in rehabilitation, occupational health examina-tions and research to assess workability (174). In a prospective study of women on long-term sick leave, the Work Ability Index and a single-item question showed strong predictive value for the degree of sick leave and health-related quality of life (HRQoL) (175).

In summary, patients’ capacity of predicting the length of the sick-listing period should be taken into account in the sickness certification procedure. Several studies have shown that physicians often issue the certificate on the request of the patient and rely on the patients’ own assessment of workabil-ity (176-178). This might be regarded as a less professional way of handling sickness certifications, but given the evidence listed above it might rather be the mode of sickness certification with the highest precision as to sick leave duration.

Red flags, signs of alert in sick leave One possibility for drawing attention to alert signs and symptoms indicating a patient at risk of long-term sick leave might be to introduce so-called ‘red or yellow flags’. Most physicians are familiar with red flags, originally de-signed for use in acute low-back pain, but the underlying concept can be applied more broadly as clinical indicators of possible serious underlying conditions requiring further medical intervention. Yellow flags are psycho-social indicators suggesting increased risk of progression to long-term dis-tress, disability and pain. They indicate psychosocial barriers to recovery. In

68

principle, yellow flags can also be applied to assess the likelihood of the development of persistent problems from any acute pain. A suggestion for further research is to investigate whether the determinants − previous sick leave days, age, sick leave diagnosis, and the patients’ assessment of sick leave conclusion − might serve the purpose of red flags in the sickness certi-fication consultation. The most important finding in this study was the significance of only these three determinants; previous sick leave, diagnosis, and age predicting time to sick leave conclusion as well as the possible use of these determinants in an easy-to-use screening tool, a nomogram. Consequently, GPs’ access to sick leave track records is a prerequisite for solving this task. The patients’ own assessment on time to sick leave conclusion might be a fourth significant determinant to include in the prediction. However, there are other factors than an early start to rehabilitation of importance in primary health care for a successful outcome for sick leave conclusion, for example the right measures, at the right time, patients men-tally prepared for the measure, and measures of high quality. In other words, an individual pathway must be tailored to enhance sick leave conclusion, in a collaborative teamwork with stakeholders.

69

Conclusions

General conclusions A few high-impact determinants for long-term sick leave, number of sick leave days in the previous year, sick leave diagnosis and age, can be applied in a screening tool to predict time to sick leave conclusion with a high de-gree of explanation (86-89%) for patients in primary health care. The pa-tients’ own assessment on sick leave conclusion is valuable information for GPs in the sickness certification consultation while assessing the length of sick leave. Multidisciplinary interventions for sick leave conclusion among patients on sick leave in primary health care need to be individually tailored.

Implication for practice An implication for clinical practice is that the nomograms can be a valuable tool not only for GPs when assessing their sickness-certified patients time for sick leave conclusion, but also for rehabilitation teams and stakeholders in vocational rehabilitation. Patients found to be at high risk of long-term sick leave might be referred to a rehabilitation team for a thorough medical examination and the necessary vocational rehabilitation can be initiated at an early stage. In a next step, systematic rehabilitation with collaboration be-tween employer, occupational health service, and the local Social Insurance Agency is necessary (32).

Suggestions for further research Future studies should be designed in order to allow conclusions to be drawn with respect to the generalisability of the findings. Such studies, preferably using a randomised controlled design, combined with a well-designed inter-vention, may answer the question of whether these nomograms are suffi-ciently sensitive and specific to enable early identification of patients at risk of long-term sick leave. Are they valid and reliable tools for everyday use, not only for GPs and the rehabilitation team but also for physicians in other specialities and stakeholders in the Swedish Social Insurance Agency, the Employment Agency and the Community?

70

This study also highlights the need to address the importance of the pa-tients’ own assessment on sick leave conclusion. A suggestion for the above-mentioned research is to combine the nomograms with the patients’ prediction, which might improve the degree of explanation on sick leave conclusion.

Another possibility for drawing attention to alert signs and symptoms in-dicating a patient at risk of long-term sick leave might be to introduce red flags in the sickness certification process.

71

Swedish summary

Sjukfrånvaro och arbetsoförmåga är ett problem i Sverige med stora konsekvenser för individen, socialt, i arbetslivet och för samhället. Ju längre sjukfrånvaro desto mindre sannolikhet att individen kommer tillbaka i arbete men det är svårt för en läkare vid ett tidspressat mottagningsbesök att veta vem som är en riskpatient för långtidssjukskrivning eller inte.

Övergripande syfte med avhandlingsarbetet var att tidigt identifiera pa-tienter i primärvård med risk för långtidssjukskrivning och därmed komma igång med behandlingar och arbetsrehabiliterande insatser för att möjliggöra snabbare återgång i arbete eller att kunna vara arbetssökande. I denna treåri-ga prospektiva kohortstudie ingick 943 patienter som sjukskrevs på en vårdcentral i Eskilstuna under 8 månader 2004.

Först analyserades kända riskvariabler för långtidssjukskrivning och deras relativa betydelse för tid till avslut av sjukskrivning (Studie I). Störst risk för längre sjukskrivningstid visade tidigare antal sjukskrivningsdagar, psykisk diagnos och högre ålder. Diagnoser som övre luftvägsinfektioner och ögon- och öronsjukdomar hade minst risk.

Två modeller testades för skattning av tid till avslut av sjukskrivning. Den ena baserades på en klinisk bedömning inklusive ”magkänsla” och den andra på en databaserad multivariat analys med de viktigaste riskvariablerna. Ett prognosverktyg (nomogram) utvecklades för sannolikhetsbedömning av tid till avslut av sjukskrivning (Studie II). Skattningsmodellerna visade 73-76 % överensstämmelse mellan den skattade och den verkliga arbetsåtergången inom 28-180 dagar, dock med ca 10 % högre precision i den databaserade modellen.

Effekten utvärderades av en tidig intervention med multidisciplinär ar-betsrehabilitering i samverkan med myndigheter mellan en grupp högrisk patienter (n=170) för långtidssjukskrivning och en matchad kontrollgrupp sjukskrivna (n=340) som fick sedvanlig behandling (Studie III). Under de första 14 dagarna avslutade interventionsgruppen sjukskrivning i högre grad än den matchade kontrollgruppen men dag 15-112, när interventionen pågick, avslutade kontrollgruppen snabbare. Efter dag 113-1095 sågs ingen skillnad mellan grupperna.

De sjukskrivnas egen skattning av chans till återgång i arbete inom sex månader jämfördes med ett rehabiliteringsteams skattning (Studie IV). Överensstämmelsen var god mellan patienternas och rehabiliteringsteamets skattningar men en könsskillnad noterades då kvinnornas och rehabilitering-

72

steamets skattningar överensstämde i högre grad än skattningarna mellan männen och teamet. Männen hade en klart lägre tilltro till sin möjlighet att återgå i arbete än vad rehabiliteringsteamet hade.

Slutsatser: Tidig prediktion för tid till avslut av sjukskrivning är möjlig baserad på ett fåtal riskvariabler och för bättre prediktion bör denna komplet-teras med patientens egen skattning. Tidig multidisciplinär arbetsrehabiliter-ing i samverkan med myndigheter av sjukskrivna i primärvård med hög risk för långtidssjukskrivning, minskade sjukskrivningen men visade ingen skillnad mellan interventionsgrupp och en matchad kontrollgrupp efter avslutad intervention.

73

Acknowledgements

I wish to express my appreciation and deepest gratitude to all who, in so many different ways, have supported me throughout the process of writing this thesis. I particularly want to thank:

Associate Professor, MD, Thorne Wallman, my main supervisor and co-author, for believing in me, for guidance in the research process, for out-standing support and your never-ending enthusiastic encouragement. Always fun to work with! Thank you for fruitful discussions about research and eve-rything else in life during years of car trips from Eskilstuna to Uppsala. And yes, now I agree, it is great fun reading scientific articles!

Professor Emeritus Kurt Svärdsudd, my second supervisor and co-author, for scientific teaching, answering questions of ignorance with patience, for giv-ing “focus, focus” advices, for finding brilliant solutions to statistical prob-lems and scientific text reviews. A warm thank you for generously spending your time in retirement on my thesis.

Hans G Eriksson, my statistician at Centre for Clinical Research Sörmland, and co-author, words are to small to thank you for all those years of statisti-cal teaching, analysing, sharing frustration while waiting for registered data from authorities, and long hours of discussing results and figures.

Margaretha Eriksson, PhD, my co-supervisor and co-author, thank you for guidance in the administrative jungle of the University world, for help with your eminent review of manuscripts for all the small details, and for always being there also for ‘plattfots-frågor’.

Karin Björkegren, MD, Senior lecturer, my co-supervisor and co-author, thank you for support and valuable comments on manuscripts and thesis.

Staffan Norlander, PhD, my former supervisor at R&D Centre Sörmland County Council, thank you for believing in my study idea, for your help with my first report and for introducing me to “forskningskaffe”.

Lotta Jansson and Pirjo Teckenberg, thank you for a fantastic cooperation in the classification of more than 3000 sickness certificates! Lotta, the first

74

“rehabilitation coordinator” in Sörmland and Pirjo for being an excellent representative of the Swedish Social Insurance Agency! Thank you to the whole staff for your administration of the sickness certificates during the study period. The rehabilitation team, Viktoria Zander, Thomas Persson, Mats Henningsson, and Helena Ekman, my warmest thank you for outstand-ing work with all patients on sick leave, and for dedicated discussions about treatments, rehabilitation potential, gut feelings and for laughter. The present rehabilitation team at Vårdcentralen City, thank you for your dedicated man-ner to continue the work we started with patients on sick leave.

Linda Lännerström, research colleague, for inspiring work together and to all research colleagues in Uppsala and at Centre for Clinical Research Sörm-land, in Eskilstuna, for interesting scientific dialogues in seminars and coffee breaks.

Landstinget Sörmland (Sörmland County Council), Centre for Clinical Re-search Sörmland, RAR Sörmland, Collegues in Försäkringsmedicinska Kommittén, The Medical Library of Sörmland for making it possible to combine clinical work with research, and for or all the practical and financial support a researcher could ask for. Hildur Nordins minnesfond, thank you for grants and years of kind support from Anna-Brita Rudbeck, Elisabeth Rudbeck and her lovely family.

Rebecka Millhagen Adelswärd, editor, Dick Norberg, photographer, Mats Rehnström, antiquarian advisor and all researchers included in the parallel research project with the Biby book, thank you for sharing your profound knowledge in the field of art history and for an excellent work. Claes Essén, legal advisor, for outstanding support in so many ways. My colleagues and dear friends, Eva Basilier, Eva Hesse-Sundin, Anna-Lie Nielsen and Annika Rymark, thank you for encouragement and sharing joy and hardships.

My extended family, my late brother Peter and his family and my brothers Klas, Otto and their families, my lovely cousins, close to my heart Pillan Nilsen Moe, Carin Björne, Miriam Kipfer, Catharina De Geer, my god- children and all my friends for making my non-work arena so positive.

Finally and foremost, my beloved family, my husband Poul, for always being there and for your unconditional love and patience. Ebba, for always caring and so nice to share the field of medicine with you, Sofia for introduc-ing different ways of thinking and making the world of new technology so fun, and Ulrika for sharp observations and humorous comments. Erik, appre-ciated son in law and always helpful as “computer doctor”. Thank you all for your patience sharing my life through times of difficult changes during the last eight years, for all your love and support to finalize my thesis.

75

References

1. Sundquist J, Al-Windi A, Johansson SE, Sundquist K. Sickness absence poses a threat to the Swedish Welfare State: a cross-sectional study of sickness absence and self-reported illness. BMC Public Health. 2007;7:45.

2. Lund T, Kivimaki M, Labriola M, Villadsen E, Christensen KB. Using administrative sickness absence data as a marker of future disability pension: the prospective DREAM study of Danish private sector employees. Occup Environ Med. 2008;65(1):28-31.

3. Labriola M. Conceptual framework of sickness absence and return to work, focusing on both the individual and the contextual level. Work. 2008;30(4):377-87.

4. Roelen CA, Norder G, Koopmans PC, van Rhenen W, van der Klink JJ, Bultmann U. Employees Sick-Listed with Mental Disorders: Who Returns to Work and When? Journal of occupational rehabilitation. 2012.

5. Vingard E, Alexanderson K, Norlund A. Swedish Council on Technology Assessment in Health Care (SBU). Chapter 9. Consequences of being on sick leave. Scandinavian journal of public health Supplement. 2004;63:207-15.

6. Lannerstrom L, Wallman T, Holmstrom IK. Losing independence--the lived experience of being long-term sick-listed. BMC Public Health. 2013;13:745.

7. Floderus B, Goransson S, Alexanderson K, Aronsson G. Self-estimated life situation in patients on long-term sick leave. J Rehabil Med. 2005;37(5):291-9.

8. Virtanen M, Kivimaki M, Vahtera J, Elovainio M, Sund R, Virtanen P, et al. Sickness absence as a risk factor for job termination, unemployment, and disability pension among temporary and permanent employees. Occup Environ Med. 2006;63(3):212-7.

9. Wadell G, Burton AK. Is work good for your health and well-being? Norwich: The Stationary Office; 2006.

10. Waddell G, Burton K, Aylward M. Work and common health problems. J Insur Med. 2007;39(2):109-20.

11. Kivimaki M, Head J, Ferrie JE, Shipley MJ, Vahtera J, Marmot MG. Sickness absence as a global measure of health: evidence from mortality in the Whitehall II prospective cohort study. BMJ. 2003;327(7411):364.

12. Swedish Social Insurance Agency. Social Insurance in Figures 2015. Stockholm: Försäkringskassan, 2015.

13. The Social Insurance Code 2010:111, (2010). 14. The National Board of Health and Welfare (Socialstyrelsen). Försäkrings-

medicinskt beslutsstöd (Decision support for insurance medicine). 2008. 15. Swedish Social Insurance Agency. Assessment of work capacity. 2012. 16. Koopmans PC, Roelen CA, Groothoff JW. Risk of future sickness absence in

frequent and long-term absentees. Occup Med (Lond). 2008;58(4):268-74. 17. Roelen CA, Koopmans PC, Schreuder JA, Anema JR, van der Beek AJ. The

history of registered sickness absence predicts future sickness absence. Occup Med (Lond). 2011;61(2):96-101.

76

18. Wallman T, Wedel H, Palmer E, Rosengren A, Johansson S, Eriksson H, et al. Sick-leave track record and other potential predictors of a disability pension. A population based study of 8,218 men and women followed for 16 years. BMC Public Health. 2009;9:104.

19. Hultin H, Lindholm C, Malfert M, Moller J. Short-term sick leave and future risk of sickness absence and unemployment - the impact of health status. BMC Public Health. 2012;12:861.

20. Vahtera J, Pentti J, Kivimaki M. Sickness absence as a predictor of mortality among male and female employees. J Epidemiol Community Health. 2004;58(4):321-6.

21. Wallman T, Wedel H, Johansson S, Rosengren A, Eriksson H, Welin L, et al. The prognosis for individuals on disability retirement. An 18-year mortality follow-up study of 6887 men and women sampled from the general population. BMC Public Health. 2006;6:103.

22. Karlsson NE, Carstensen JM, Gjesdal S, Alexanderson KA. Mortality in relation to disability pension: findings from a 12-year prospective population-based cohort study in Sweden. Scandinavian journal of public health. 2007;35(4):341-7.

23. Ferrie JE, Vahtera J, Kivimaki M, Westerlund H, Melchior M, Alexanderson K, et al. Diagnosis-specific sickness absence and all-cause mortality in the GAZEL study. J Epidemiol Community Health. 2009;63(1):50-5.

24. Gjesdal S, Bratberg E. Diagnosis and duration of sickness absence as predictors for disability pension: results from a three-year, multi-register based* and prospective study. Scand J Public Health. 2003;31(4):246-54.

25. Mittendorfer-Rutz E, Kjeldgard L, Runeson B, Perski A, Melchior M, Head J, et al. Sickness absence due to specific mental diagnoses and all-cause and cause-specific mortality: a cohort study of 4.9 million inhabitants of Sweden. PLoS One. 2012;7(9):e45788.

26. Stapelfeldt CM, Nielsen CV, Andersen NT, Krane L, Borg V, Fleten N, et al. Sick leave patterns as predictors of disability pension or long-term sick leave: a 6.75-year follow-up study in municipal eldercare workers. BMJ Open. 2014;4(2):e003941.

27. Hensing G, Alexanderson K, Allebeck P, Bjurulf P. How to measure sickness absence? Literature review and suggestion of five basic measures. Scand J Soc Med. 1998;26(2):133-44.

28. Christensen KB, Labriola M, Lund T, Kivimaki M. Explaining the social gradient in long-term sickness absence: a prospective study of Danish employees. J Epidemiol Community Health. 2008;62(2):181-3.

29. Lund T, Labriola M, Villadsen E. Who is at risk for long-term sickness absence? A prospective cohort study of Danish employees. Work. 2007;28(3):225-30.

30. Gjesdal S, Ringdal PR, Haug K, Maeland JG, Vollset SE, Alexanderson K. Mortality after long-term sickness absence: prospective cohort study. Eur J Public Health. 2008;18(5):517-21.

31. Bratberg E, Gjesdal S, Maeland JG. Sickness absence with psychiatric diagnoses: individual and contextual predictors of permanent disability. Health Place. 2009;15(1):308-14.

32. Marmot M, Feeney A, Shipley M, North F, Syme SL. Sickness absence as a measure of health status and functioning: from the UK Whitehall II study. Journal of epidemiology and community health. 1995;49(2):124-30.

77

33. Kivimaki M, Head J, Ferrie JE, Singh-Manoux A, Westerlund H, Vahtera J, et al. Sickness absence as a prognostic marker for common chronic conditions: analysis of mortality in the GAZEL study. Occup Environ Med. 2008;65(12):820-6.

34. Fahlen G, Goine H, Edlund C, Arrelov B, Knutsson A, Peter R. Effort-reward imbalance, "locked in" at work, and long-term sick leave. International archives of occupational and environmental health. 2009;82(2):191-7.

35. Lindberg P, Josephson M, Alfredsson L, Vingard E. Promoting excellent work ability and preventing poor work ability: the same determinants? Results from the Swedish HAKuL study. Occup Environ Med. 2006;63(2):113-20.

36. Karlsson NE, Carstensen JM, Gjesdal S, Alexanderson KA. Risk factors for disability pension in a population-based cohort of men and women on long-term sick leave in Sweden. Eur J Public Health. 2008;18(3):224-31.

37. Blank N, Diderichsen F. Short-term and long-term sick-leave in Sweden: relationships with social circumstances, working conditions and gender. Scand J Soc Med. 1995;23(4):265-72.

38. Westerlund H, Ferrie J, Hagberg J, Jeding K, Oxenstierna G, Theorell T. Workplace expansion, long-term sickness absence, and hospital admission. Lancet. 2004;363(9416):1193-7.

39. Vaez M, Rylander G, Nygren A, Asberg M, Alexanderson K. Sickness absence and disability pension in a cohort of employees initially on long-term sick leave due to psychiatric disorders in Sweden. Social psychiatry and psychiatric epidemiology. 2007;42(5):381-8.

40. Gjesdal S, Bratberg E. The role of gender in long-term sickness absence and transition to permanent disability benefits. Results from a multiregister based, prospective study in Norway 1990-1995. Eur J Public Health. 2002;12(3):180-6.

41. Tellnes G. Duration of episodes of sickness certification. Scand J Prim Health Care. 1989;7(4):237-44.

42. Lidwall U, Bergendorff S, Voss M, Marklund S. Long-term sickness absence: changes in risk factors and the population at risk. Int J Occup Med Environ Health. 2009;22(2):157-68.

43. Allebeck P, Mastekaasa A. Swedish Council on Technology Assessment in Health Care (SBU). Chapter 5. Risk factors for sick leave - general studies. Scand J Public Health Suppl. 2004;63:49-108.

44. Karlsson N, Borg K, Carstensen J, Hensing G, Alexanderson K. Risk of disability pension in relation to gender and age in a Swedish county; a 12-year population based, prospective cohort study. Work. 2006;27(2):173-9.

45. Hansson T, Jensen I. Swedish Council on Technology Assessment in Health Care (SBU). Chapter 6. Sickness absence due to back and neck disorders. Scand J Public Health Suppl. 2004;63:109-51.

46. Asplund R, Marnetoft SU, Selander J, Akerstrom B. Place of residence as a correlate of sickness absence in Sweden. Int J Rehabil Res. 2007;30(2):147-51.

47. Sandanger I, Nygard JF, Brage S, Tellnes G. Relation between health problems and sickness absence: gender and age differences--a comparison of low-back pain, psychiatric disorders, and injuries. Scand J Public Health. 2000;28(4):244-52.

48. Koopmans PC, Roelen CA, Bultmann U, Hoedeman R, van der Klink JJ, Groothoff JW. Gender and age differences in the recurrence of sickness absence due to common mental disorders: a longitudinal study. BMC Public Health. 2010;10:426.

78

49. Hensing G, Wahlstrom R. Swedish Council on Technology Assessment in Health Care (SBU). Chapter 7. Sickness absence and psychiatric disorders. Scand J Public Health Suppl. 2004;63:152-80.

50. Gjesdal S, Haug K, Ringdal P, Maeland JG, Hagberg J, Roraas T, et al. Sickness absence with musculoskeletal or mental diagnoses, transition into disability pension and all-cause mortality: a 9-year prospective cohort study. Scand J Public Health. 2009;37(4):387-94.

51. Kivimaki M, Ferrie JE, Hagberg J, Head J, Westerlund H, Vahtera J, et al. Diagnosis-specific sick leave as a risk marker for disability pension in a Swedish population. J Epidemiol Community Health. 2007;61(10):915-20.

52. Du Bois M, Szpalski M, Donceel P. Patients at risk for long-term sick leave because of low back pain. Spine J. 2009;9(5):350-9.

53. Lotters F, Burdorf A. Prognostic factors for duration of sickness absence due to musculoskeletal disorders. Clin J Pain. 2006;22(2):212-21.

54. Alexanderson KA, Borg KE, Hensing GK. Sickness absence with low-back, shoulder, or neck diagnoses: an 11-year follow-up regarding gender differences in sickness absence and disability pension. Work. 2005;25(2):115-24.

55. Bergstrom G, Bodin L, Bertilsson H, Jensen IB. Risk factors for new episodes of sick leave due to neck or back pain in a working population. A prospective study with an 18-month and a three-year follow-up. Occup Environ Med. 2007;64(4):279-87.

56. Roelen CA, Koopmans PC, Notenbomer A, Groothoff JW. Job satisfaction and sickness absence: a questionnaire survey. Occup Med (Lond). 2008;58(8):567-71.

57. Stover M, Pape K, Johnsen R, Fleten N, Sund ER, Claussen B, et al. Unemployment and disability pension--an 18-year follow-up study of a 40-year-old population in a Norwegian county. BMC Public Health. 2012;12:148.

58. Selander J, Marnetoft SU, Bergroth A, Ekholm J. The process of vocational rehabilitation for employed and unemployed people on sick-leave: employed people vs unemployed people in Stockholm compared with circumstances in rural Jamtland, Sweden. Scand J Rehabil Med. 1998;30(1):55-60.

59. Janssen N, Kant IJ, Swaen GM, Janssen PP, Schroer CA. Fatigue as a predictor of sickness absence: results from the Maastricht cohort study on fatigue at work. Occup Environ Med. 2003;60 Suppl 1:i71-6.

60. Roelen CA, Heymans MW, van Rhenen W, Groothoff JW, Twisk JW, Bultmann U. Fatigue as prognostic risk marker of mental sickness absence in white collar employees. Journal of occupational rehabilitation. 2014;24(2):307-15.

61. Bultmann U, Nielsen MB, Madsen IE, Burr H, Rugulies R. Sleep disturbances and fatigue: independent predictors of sickness absence? A prospective study among 6538 employees. Eur J Public Health. 2013;23(1):123-8.

62. Akerstedt T, Kecklund G, Selen J. Disturbed sleep and fatigue as predictors of return from long-term sickness absence. Ind Health. 2010;48(2):209-14.

63. Salo P, Oksanen T, Sivertsen B, Hall M, Pentti J, Virtanen M, et al. Sleep disturbances as a predictor of cause-specific work disability and delayed return to work. Sleep. 2010;33(10):1323-31.

64. Spak F, Hensing G, Allebeck P. Sick-leave in women with alcohol dependence or abuse: effects of additional psychiatric disorders. Soc Psychiatry Psychiatr Epidemiol. 1998;33(12):613-9.

79

65. Vahtera J, Poikolainen K, Kivimaki M, Ala-Mursula L, Pentti J. Alcohol intake and sickness absence: a curvilinear relation. Am J Epidemiol. 2002;156(10):969-76.

66. Salonsalmi A, Laaksonen M, Lahelma E, Rahkonen O. Drinking habits and sickness absence: the contribution of working conditions. Scand J Public Health. 2009;37(8):846-54.

67. Upmark M, Moller J, Romelsjo A. Longitudinal, population-based study of self reported alcohol habits, high levels of sickness absence, and disability pensions. J Epidemiol Community Health. 1999;53(4):223-9.

68. Kivimaki M, Head J, Ferrie JE, Hemingway H, Shipley MJ, Vahtera J, et al. Working while ill as a risk factor for serious coronary events: the Whitehall II study. Am J Public Health. 2005;95(1):98-102.

69. Bergstrom G, Bodin L, Hagberg J, Aronsson G, Josephson M. Sickness presenteeism today, sickness absenteeism tomorrow? A prospective study on sickness presenteeism and future sickness absenteeism. J Occup Environ Med. 2009;51(6):629-38.

70. Bergh H, Baigi A, Mansson J, Mattsson B, Marklund B. Predictive factors for long-term sick leave and disability pension among frequent and normal attenders in primary health care over 5 years. Public Health. 2007;121(1):25-33.

71. Alexanderson K. Sickness absence: causes, consequences, and physicians' sickness certification practice. A systematic literature review by the Swedish Council on Technology Assessment in Health Care (SBU). Scandinavian journal of public health Supplement. 2004;63:3-263.

72. Tellnes G. Sickness certification in general practice: a review. Family practice. 1989;6(1):58-65.

73. Lofgren A, Hagberg J, Arrelov B, Ponzer S, Alexanderson K. Frequency and nature of problems associated with sickness certification tasks: a cross-sectional questionnaire study of 5455 physicians. Scand J Prim Health Care. 2007;25(3):178-85.

74. Wynne-Jones G, Mallen CD, Welsh V, Dunn KM. Rates of sickness certification in European primary care: A systematic review. Eur J Gen Pract. 2009:1-10.

75. Lindholm C, Arrelov B, Nilsson G, Lofgren A, Hinas E, Skaner Y, et al. Sickness-certification practice in different clinical settings; a survey of all physicians in a country. BMC Public Health. 2010;10:752.

76. Gimeno D, Benavides FG, Benach J, Amick BC, 3rd. Distribution of sickness absence in the European Union countries. Occup Environ Med. 2004;61(10):867-9.

77. Letrilliart L, Barrau A. Difficulties with the sickness certification process in general practice and possible solutions: a systematic review. Eur J Gen Pract. 2012;18(4):219-28.

78. Foley M, Thorley K, Denny M. 'The sick note': a qualitative study of sickness certification in general practice in Ireland. Eur J Gen Pract. 2012;18(2):92-9.

79. Engblom M, Alexanderson K, Rudebeck CE. Characteristics of sick-listing cases that physicians consider problematic--analyses of written case reports. Scand J Prim Health Care. 2009;27(4):250-5.

80. Kiessling A, Arrelov B. Sickness certification as a complex professional and collaborative activity--a qualitative study. BMC Public Health. 2012;12:702.

81. Wahlstrom R, Alexanderson K. Swedish Council on Technology Assessment in Health Care (SBU). Chapter 11. Physicians' sick-listing practices. Scandinavian journal of public health Supplement. 2004;63:222-55.

80

82. Swartling MS, Hagberg J, Alexanderson K, Wahlstrom RA. Sick-listing as a psychosocial work problem: a survey of 3997 Swedish physicians. J Occup Rehabil. 2007;17(3):398-408.

83. Ljungquist T, Hinas E, Nilsson GH, Gustavsson C, Arrelov B, Alexanderson K. Problems with sickness certification tasks: experiences from physicians in different clinical settings. A cross-sectional nationwide study in Sweden. BMC Health Serv Res. 2015;15:321.

84. Wynne-Jones G, Mallen CD, Main CJ, Dunn KM. What do GPs feel about sickness certification? A systematic search and narrative review. Scand J Prim Health Care. 2010;28(2):67-75.

85. Hussey S, Hoddinott P, Wilson P, Dowell J, Barbour R. Sickness certification system in the United Kingdom: qualitative study of views of general practitioners in Scotland. BMJ. 2004;328(7431):88.

86. Gerner U, Alexanderson K. Issuing sickness certificates: a difficult task for physicians: a qualitative analysis of written statements in a Swedish survey. Scand J Public Health. 2009;37(1):57-63.

87. Hiscock JaR, J. The role of GPs in sickness certification. Pension DfWa; 2001 148.

88. Arrelov B, Alexanderson K, Hagberg J, Lofgren A, Nilsson G, Ponzer S. Dealing with sickness certification - a survey of problems and strategies among general practitioners and orthopaedic surgeons. BMC Public Health. 2007;7:273.

89. Ljungquist T, Arrelov B, Lindholm C, Wilteus AL, Nilsson GH, Alexanderson K. Physicians who experience sickness certification as a work environmental problem: where do they work and what specific problems do they have? A nationwide survey in Sweden. BMJ Open. 2012;2(2):e000704.

90. Peterson S, Eriksson M, Tibblin G. Practice variation in Swedish primary care. Scand J Prim Health Care. 1997;15(2):68-75.

91. Reiso H, Nygard JF, Jorgensen GS, Holanger R, Soldal D, Bruusgaard D. Back to work: predictors of return to work among patients with back disorders certified as sick: a two-year follow-up study. Spine. 2003;28(13):1468-73; discussion 73-4.

92. Shiels C, Gabbay MB, Ford FM. Patient factors associated with duration of certified sickness absence and transition to long-term incapacity. Br J Gen Pract. 2004;54(499):86-91.

93. von Knorring M, Sundberg L, Lofgren A, Alexanderson K. Problems in sickness certification of patients: a qualitative study on views of 26 physicians in Sweden. Scand J Prim Health Care. 2008;26(1):22-8.

94. Swartling M, Peterson S, Wahlstrom R. Views on sick-listing practice among Swedish General Practitioners - a phenomenographic study. BMC Fam Pract. 2007;8:44.

95. Suoyrjo H, Oksanen T, Hinkka K, Kivimaki M, Klaukka T, Pentti J, et al. The effectiveness of vocationally oriented multidisciplinary intervention on sickness absence and early retirement among employees at risk: an observational study. Occup Environ Med. 2009;66(4):235-42.

96. Kuoppala J, Lamminpaa A. Rehabilitation and work ability: a systematic literature review. J Rehabil Med. 2008;40(10):796-804.

97. Gobelet C, Luthi F, Al-Khodairy AT, Chamberlain MA. Vocational rehabilitation: a multidisciplinary intervention. Disabil Rehabil. 2007;29(17):1405-10.

98. Waddell G, Burton, K., Kendall, N. Vocational Rehabilitation: What works, for whom, and when? Pensions DfWa; 2013.

81

99. Franche RL, Cullen K, Clarke J, Irvin E, Sinclair S, Frank J, et al. Workplace-based return-to-work interventions: a systematic review of the quantitative literature. Journal of occupational rehabilitation. 2005;15(4):607-31.

100. Hoefsmit N, Houkes I, Nijhuis FJ. Intervention characteristics that facilitate return to work after sickness absence: a systematic literature review. J Occup Rehabil. 2012;22(4):462-77.

101. Norlund A, Ropponen A, Alexanderson K. Multidisciplinary interventions: review of studies of return to work after rehabilitation for low back pain. J Rehabil Med. 2009;41(3):115-21.

102. Johansson P, Lindahl E. Locking-in effects due to early interventions? An evaluation of a multidisciplinary screening programs for avoiding long-term sickness. Evaluation review. 2012;36(5):323-45.

103. Carlsson L, Englund L, Hallqvist J, Wallman T. Early multidisciplinary assessment was associated with longer periods of sick leave: a randomized controlled trial in a primary health care centre. Scand J Prim Health Care. 2013;31(3):141-6.

104. Reiso H, Nygard JF, Brage S, Gulbrandsen P, Tellnes G. Work ability and duration of certified sickness absence. Scandinavian journal of public health. 2001;29(3):218-25.

105. Reiso H, Gulbrandsen P, Brage S. Doctors' prediction of certified sickness absence. Fam Pract. 2004;21(2):192-8.

106. Fleten N, Johnsen R, Forde OH. Length of sick leave - why not ask the sick-listed? Sick-listed individuals predict their length of sick leave more accurately than professionals. BMC Public Health. 2004;4:46.

107. Heijbel B, Josephson M, Jensen I, Stark S, Vingard E. Return to work expectation predicts work in chronic musculoskeletal and behavioral health disorders: prospective study with clinical implications. J Occup Rehabil. 2006;16(2):173-84.

108. Marhold C, Linton SJ, Melin L. Identification of obstacles for chronic pain patients to return to work: evaluation of a questionnaire. J Occup Rehabil. 2002;12(2):65-75.

109. Nielsen MB, Madsen IE, Bultmann U, Christensen U, Diderichsen F, Rugulies R. Predictors of return to work in employees sick-listed with mental health problems: findings from a longitudinal study. Eur J Public Health. 2011;21(6):806-11.

110. Duijts SF, Kant IJ, Landeweerd JA, Swaen GM. Prediction of sickness absence: development of a screening instrument. Occup Environ Med. 2006;63(8):564-9.

111. Oppenheimer GM. Becoming the Framingham Study 1947-1950. Am J Public Health. 2005;95(4):602-10.

112. Wilhelmsen L, Wedel H, Conroy R, Fitzgerald T. [The Swedish SCORE chart for cardiovascular risk. Better possibilities for prevention of cardiovascular diseases]. Lakartidningen. 2004;101(20):1798-801.

113. Kanis JA, McCloskey EV, Johansson H, Oden A, Strom O, Borgstrom F. Development and use of FRAX in osteoporosis. Osteoporosis international : a journal established as result of cooperation between the European Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA. 2010;21 Suppl 2:S407-13.

114. Broms K, Norback D, Sundelin C, Eriksson M, Svardsudd K. A nationwide study of asthma incidence rate and its determinants in Swedish pre-school children. Eur J Epidemiol. 2012;27(9):695-703.

115. WHO. International Classification of Diseases (ICD-10). 1997.

82

116. Linton SJ, Boersma K. Early identification of patients at risk of developing a persistent back problem: the predictive validity of the Orebro Musculoskeletal Pain Questionnaire. Clin J Pain. 2003;19(2):80-6.

117. Bogefeldt J, Grunnesjo MI, Svardsudd K, Blomberg S. Sick leave reductions from a comprehensive manual therapy programme for low back pain: the Gotland Low Back Pain Study. Clin Rehabil. 2008;22(6):529-41.

118. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143(1):29-36.

119. Rosenbaum PRR, D.B. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41-55.

120. Haldorsen EM, Grasdal AL, Skouen JS, Risa AE, Kronholm K, Ursin H. Is there a right treatment for a particular patient group? Comparison of ordinary treatment, light multidisciplinary treatment, and extensive multidisciplinary treatment for long-term sick-listed employees with musculoskeletal pain. Pain. 2002;95(1-2):49-63.

121. Bultmann U, Sherson D, Olsen J, Hansen CL, Lund T, Kilsgaard J. Coordinated and tailored work rehabilitation: a randomized controlled trial with economic evaluation undertaken with workers on sick leave due to musculoskeletal disorders. Journal of occupational rehabilitation. 2009;19(1):81-93.

122. Anema JR, Steenstra IA, Bongers PM, de Vet HC, Knol DL, Loisel P, et al. Multidisciplinary rehabilitation for subacute low back pain: graded activity or workplace intervention or both? A randomized controlled trial. Spine (Phila Pa 1976). 2007;32(3):291-8; discussion 9-300.

123. Ejeby K, Savitskij R, Ost LG, Ekbom A, Brandt L, Ramnero J, et al. Symptom reduction due to psychosocial interventions is not accompanied by a reduction in sick leave: results from a randomized controlled trial in primary care. Scand J Prim Health Care. 2014;32(2):67-72.

124. Jensen OK, Stengaard-Pedersen K, Jensen C, Nielsen CV. Prediction model for unsuccessful return to work after hospital-based intervention in low back pain patients. BMC Musculoskelet Disord. 2013;14:140.

125. Martin MH, Nielsen MB, Pedersen J, Rugulies R. Stability of return to work after a coordinated and tailored intervention for sickness absence compensation beneficiaries with mental health problems: results of a two-year follow-up study. Disabil Rehabil. 2015:1-7.

126. Karrholm J, Ekholm K, Ekholm J, Bergroth A, Ekholm KS. Systematic co-operation between employer, occupational health service and social insurance office: a 6-year follow-up of vocational rehabilitation for people on sick-leave, including economic benefits. J Rehabil Med. 2008;40(8):628-36.

127. Axelsson I, Marnetoft SU. Benefits and harms of sick leave: lack of randomized, controlled trials. Int J Rehabil Res. 2010;33(1):1-3.

128. CDUST. Liv och hälsa 2004. 2005. 129. Swedish Social Insurance Agency. Socialförsäkringsrapport (Social Insurance

Report) 2011:7. Försäkringsmedicinska beslutsstödet 2011. 130. Knudsen AK, Harvey SB, Mykletun A, Overland S. Common mental disorders

and long-term sickness absence in a general working population. The Hordaland Health Study. Acta Psychiatr Scand. 2013;127(4):287-97.

131. Koopmans PC, Bultmann U, Roelen CA, Hoedeman R, van der Klink JJ, Groothoff JW. Recurrence of sickness absence due to common mental disorders. Int Arch Occup Environ Health. 2011;84(2):193-201.

132. Pransky G, Gatchel R, Linton SJ, Loisel P. Improving return to work research. Journal of occupational rehabilitation. 2005;15(4):453-7.

83

133. Oyeflaten I, Lie SA, Ihlebaek CM, Eriksen HR. Multiple transitions in sick leave, disability benefits, and return to work. - A 4-year follow-up of patients participating in a work-related rehabilitation program. BMC Public Health. 2012;12:748.

134. Lidwall U. Sick leave diagnoses and return to work: a Swedish register study. Disabil Rehabil. 2015;37(5):396-410.

135. Viikari-Juntura E, Kausto J, Shiri R, Kaila-Kangas L, Takala EP, Karppinen J, et al. Return to work after early part-time sick leave due to musculoskeletal disorders: a randomized controlled trial. Scand J Work Environ Health. 2011.

136. Whittaker W, Sutton M, Maxwell M, Munoz-Arroyo R, Macdonald S, Power A, et al. Predicting which people with psychosocial distress are at risk of becoming dependent on state benefits: analysis of routinely available data. BMJ. 2010;341:c3838.

137. Alexanderson K, Kivimaki M, Ferrie JE, Westerlund H, Vahtera J, Singh-Manoux A, et al. Diagnosis-specific sick leave as a long-term predictor of disability pension: a 13-year follow-up of the GAZEL cohort study. J Epidemiol Community Health. 2012;66(2):155-9.

138. Hubertsson J, Englund M, Hallgarde U, Lidwall U, Lofvendahl S, Petersson IF. Sick leave patterns in common musculoskeletal disorders--a study of doctor prescribed sick leave. BMC Musculoskelet Disord. 2014;15:176.

139. Stolper E, Van de Wiel M, Van Royen P, Van Bokhoven M, Van der Weijden T, Dinant GJ. Gut feelings as a third track in general practitioners' diagnostic reasoning. J Gen Intern Med. 2011;26(2):197-203.

140. Stolper CF, Van de Wiel MW, De Vet HC, Rutten AL, Van Royen P, Van Bokhoven MA, et al. Family physicians' diagnostic gut feelings are measurable: construct validation of a questionnaire. BMC Fam Pract. 2013;14:1.

141. Stolper E, van Bokhoven M, Houben P, Van Royen P, van de Wiel M, van der Weijden T, et al. The diagnostic role of gut feelings in general practice. A focus group study of the concept and its determinants. BMC Fam Pract. 2009;10:17.

142. Ekberg K, Wahlin C, Persson J, Bernfort L, Oberg B. Early and Late Return to Work After Sick Leave: Predictors in a Cohort of Sick-Listed Individuals with Common Mental Disorders. Journal of occupational rehabilitation. 2015;25(3):627-37.

143. Karrholm J, Ekholm K, Jakobsson B, Ekholm J, Bergroth A, Schuldt K. Effects on work resumption of a co-operation project in vocational rehabilitation. Systematic, multi-professional, client-centred and solution-oriented co-operation. Disabil Rehabil. 2006;28(7):457-67.

144. The Swedish Social Insurance Inspectorate (Inspektionen för socialför-säkringen). Rehabiliteringsgarantins effekter på hälsa och sjukfrånvaro, 2014:12. The Rehabilitation Guarantee. 2014.

145. Stigmar KG, Petersson IF, Joud A, Grahn BE. Promoting work ability in a structured national rehabilitation program in patients with musculoskeletal disorders: outcomes and predictors in a prospective cohort study. BMC Musculoskelet Disord. 2013;14:57.

146. Wahlin C, Ekberg K, Persson J, Bernfort L, Oberg B. Association between clinical and work-related interventions and return-to-work for patients with musculoskeletal or mental disorders. J Rehabil Med. 2012;44(4):355-62.

147. Eriksson UB. An explorative interview study of men and women on sick leave with a musculoskeletal disgnosis seeking an acceptable role. Nordic Journal of Social Research. 2013;4.

84

148. Hubertsson J, Petersson IF, Arvidsson B, Thorstensson CA. Sickness absence in musculoskeletal disorders - patients' experiences of interactions with the social insurance agency and health care. A qualitative study. BMC Public Health. 2011;11:107.

149. Stahl C, Edvardsson Stiwne E. Narratives of sick leave, return to work and job mobility for people with common mental disorders in Sweden. J Occup Rehabil. 2014;24(3):543-54.

150. Eriksson UB, Engstrom LG, Starrin B, Janson S. Falling between two stools; how a weak co-operation between the social security and the unemployment agencies obstructs rehabilitation of unemployed sick-listed persons. Disabil Rehabil. 2008;30(8):569-76.

151. Marnetoft SU, Selander J. Long-term effects of early versus delayed vocational rehabilitation - a four-year follow-up. Disabil Rehabil. 2002;24(14) :741-5.

152. Sandstrom J, Esbjornsson E. Return to work after rehabilitation. The significance of the patient's own prediction. Scand J Rehabil Med. 1986;18(1):29-33.

153. Sampere M, Gimeno D, Serra C, Plana M, Lopez JC, Martinez JM, et al. Return to work expectations of workers on long-term non-work-related sick leave. J Occup Rehabil. 2012;22(1):15-26.

154. Hansen A, Edlund C, Henningsson M. Factors relevant to a return to work: a multivariate approach. Work. 2006;26(2):179-90.

155. Ebrahim S, Malachowski C, Kamal El Din M, Mulla SM, Montoya L, Bance S, et al. Measures of patients' expectations about recovery: a systematic review. Journal of occupational rehabilitation. 2015;25(1):240-55.

156. Mondloch MV, Cole DC, Frank JW. Does how you do depend on how you think you'll do? A systematic review of the evidence for a relation between patients' recovery expectations and health outcomes. CMAJ. 2001;165(2):174-9.

157. Bandura A. Self-efficacy: the exercise of control. New York, USA: W.H. Freeman and Company; 1997.

158. Franche RL, Krause N. Readiness for return to work following injury or illness: conceptualizing the interpersonal impact of health care, workplace, and insurance factors. Journal of occupational rehabilitation. 2002;12(4):233-56.

159. Brouwer S, Krol B, Reneman MF, Bultmann U, Franche RL, van der Klink JJ, et al. Behavioral determinants as predictors of return to work after long-term sickness absence: an application of the theory of planned behavior. Journal of occupational rehabilitation. 2009;19(2):166-74.

160. Brouwer S, Reneman MF, Bultmann U, van der Klink JJ, Groothoff JW. A prospective study of return to work across health conditions: perceived work attitude, self-efficacy and perceived social support. Journal of occupational rehabilitation. 2010;20(1):104-12.

161. Lagerveld SE, Blonk, R.W.B., Brenninkmeijer, V., Schaufeli, W. B.,. Return to work among employees with mental health problems: Development and validation of a self-efficacy questionnaire. Work & Stress. 2010;24(4):359-75.

162. Brouwer S, Amick BC, 3rd, Lee H, Franche RL, Hogg-Johnson S. The Predictive Validity of the Return-to-Work Self-Efficacy Scale for Return-to-Work Outcomes in Claimants with Musculoskeletal Disorders. Journal of occupational rehabilitation. 2015;25(4):725-32.

85

Among Long-Term Sick-Listed Employees. Journal of occupational rehabilitation. 2015;25(2):423-31.

164. Eriksson HG, von Celsing AS, Wahlstrom R, Janson L, Zander V, Wallman T. Sickness absence and self-reported health a population-based study of 43,600 individuals in central Sweden. BMC Public Health. 2008;8:426.

165. Halford C, Wallman T, Welin L, Rosengren A, Bardel A, Johansson S, et al. Effects of self-rated health on sick leave, disability pension, hospital admissions and mortality. A population-based longitudinal study of nearly 15,000 observations among Swedish women and men. BMC Public Health. 2012;12:1103.

166. Idler EL, Benyamini Y. Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav. 1997;38(1):21-37.

167. Post M, Krol B, Groothoff JW. Self-rated health as a predictor of return to work among employees on long-term sickness absence. Disabil Rehabil. 2006;28(5):289-97.

168. Josephson M, Heijbel B, Voss M, Alfredsson L, Vingard E. Influence of self-reported work conditions and health on full, partial and no return to work after long-term sickness absence. Scand J Work Environ Health. 2008;34(6):430-7.

169. Unden AL, Elofsson S. Health from the patient's point of view. How does it relate to the physician's judgement? Fam Pract. 2001;18(2):174-80.

170. Atroshi I, Andersson IH, Gummesson C, Leden I, Odenbring S, Ornstein E. Primary care patients with musculoskeletal pain. Value of health-status and sense-of-coherence measures in predicting long-term work disability. Scandinavian journal of rheumatology. 2002;31(4):239-44.

171. Lexis MA, Jansen NW, van Amelsvoort LG, Huibers MJ, Berkouwer A, Tjin ATG, et al. Prediction of long-term sickness absence among employees with depressive complaints. Journal of occupational rehabilitation. 2012;22(2):262-9.

172. Bowling A. Just one question: If one question works, why ask several? J Epidemiol Community Health. 2005;59(5):342-5.

173. Heijbel B. Return to work after long-term sick leave: Barriers and facilitators [Thesis for doctoral degree]. Stockholm: Karolinska Institutet; 2012.

174. Ilmarinen J. The Work Ability Index (WAI). Occup Med (Lond). 2007(57):160.

175. Ahlstrom L, Grimby-Ekman A, Hagberg M, Dellve L. The work ability index and single-item question: associations with sick leave, symptoms, and health--a prospective study of women on long-term sick leave. Scand J Work Environ Health. 2010;36(5):404-12.

176. Carlsson L, Lannerstrom L, Wallman T, Holmstrom IK. General practitioners' perceptions of working with the certification of sickness absences following changes in the Swedish social security system: a qualitative focus-group study. BMC Fam Pract. 2015;16:21.

177. Englund L, Svardsudd K. Sick-listing habits among general practitioners in a Swedish county. Scand J Prim Health Care. 2000;18(2):81-6.

178. Stigmar K, Grahn B, Ekdahl C. Work ability--experiences and perceptions among physicians. Disabil Rehabil. 2010;32(21):1780-9.

163. Volker D, Zijlstra-Vlasveld MC, Brouwers EP, van Lomwel AG, van der Feltz-Cornelis CM. Return-to-Work Self-Efficacy and Actual Return to Work

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Medicine 1194

Editor: The Dean of the Faculty of Medicine

A doctoral dissertation from the Faculty of Medicine, UppsalaUniversity, is usually a summary of a number of papers. A fewcopies of the complete dissertation are kept at major Swedishresearch libraries, while the summary alone is distributedinternationally through the series Digital ComprehensiveSummaries of Uppsala Dissertations from the Faculty ofMedicine. (Prior to January, 2005, the series was publishedunder the title “Comprehensive Summaries of UppsalaDissertations from the Faculty of Medicine”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-280414

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2016

early risk assessment of long- term sick leave among...

Documents