personal factors associated with sickness absence' · the length ofservice from aninitial...

Brit. J. industr. Med., I968, 25, io6.

Personal Factors Associated with Sickness Absence'A Study of I94 Men with Contrasting Sickness Absence Experience in

a Refinery Population

P. J. TAYLORFrom Shell U.K. Limited, Shell Haven, Essex

Men with different patterns of sickness absence behaviour have been identified from a refinerypopulation by simple epidemiological techniques. A detailed clinical study is described of four groups:56 men with five or more sickness spells in I964 and a matched control of 56 men; 35 men who had 6oor more days of sickness absence in I964; and finally 47 men who had not had one day off sick for at leasteight years.Whereas the men who were frequently sick tended to be younger and mostly on day work, those with

long periods of sickness were reasonably representative of the whole population, and the men withoutany sickness absence were older and mostly on shift work. An analysis of records both before and sincei964 showed that the groups had maintained a consistent pattern of sickness absence, but when individualswere considered their behaviour was less consistent. Nevertheless there appeared to be states of sicknessabsence 'liability' and also 'resistance' which persisted for a variable length of time from a year or two upto many years.The pre-employment medical examination proved in retrospect to have been of little predictive value.

Absenteeism, lateness, and also occupational injuries were all strongly associated with sickness spells,although the level of overtime was not. Previous episodes of neurotic illness, peptic ulceration, and lossof work due to back pain were also associated with frequent sickness spells, so also were frequent colds andtroublesome constipation. An unexpected finding from the physical examination was that over one quarterof those who were never sick had some organic disease.Although neither the social nor economic circumstances differed between the groups, the attitude of

the men towards themselves and their work proved to be of major importance. A memory of an unhappychildhood was more common in both groups with a lot of sickness absence, whilst dislike of the job orfrustrated ambition was common in men with frequent spells. Those who were never sick denied allsuch problems as they denied illness. Personality testing revealed that extroversion was more markedin the frequently sick group, neuroticism in the long sick, and introversion in the never sick.The validity and significance of the results are discussed and suggestions are made for further

investigation.

Absence from work due to illness or injury can Exchequer alone is well over £200 million. Morrisnot only produce difficulty and sometimes hardship (i965) has shown that sickness absence has increasedfor the sick person and his family but it also brings dramatically in this country since World War II,problems and expense to the employer. Further- and Enterline (i964) indicated that this is a problemmore, there is a serious effect on the national which affects particularly the countries of Westerneconomy since well over one million people are Europe.absent from work each day in Great Britain due to The degree of attention paid to sickness absencecertified incapacity, and the annual cost to the has usually depended upon economic conditions,

being greater when labour is scarce, as in wartime,LThis work was awarded the Occupational Health Prize of or costly, as it is today. Most published reports on

the British Medical Association, I967. sickness absence have been concerned with methodsReceived for publication October 31, I967. of measurement and the effects of age, job, civil

io6

on June 29, 2020 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.25.2.106 on 1 A

pril 1968. Dow

nloaded from

http://oem.bmj.com/

Personal Factors Associated with Sickness Absence

status, and sick pay schemes on absence rates.There have been few studies designed to measurethe importance of personal factors, although thisformed the basis of extensive studies in relation toaccidents by the Industrial Health Research Board(Farmer and Chambers, I929) and others such asSmiley (I955) and Cresswell and Froggatt (I963).

In recent years evidence has been presented(Sutherland and Whitwell, 1948; Hinkle andPlummer, I952; Taylor, I967a) that the pattern ofsickness absence within members of an industrialpopulation resembles that of an unequal chancedistribution (negative binomial) rather than one ofrandom distribution (Poisson). In this respectsickness absence resembles occupational accidentsfor which this pattern was first described by Green-wood and Woods (I9I9).The ability to identify, by epidemiological tech-

niques, groups of workers with extremes of sicknessabsence behaviour led to the present study whichdescribes an investigation designed to provideanswers to the following questions:i. Will a detailed examination of individual menwho have widely contrasting sickness absencerecords reveal other significant differences betweenthem ?2. If such differences exist, are they medical, social,occupational or economic, and are any of them likelyto be important causes of sickness absence?

Material and Selection of Groups for Study

A major problem in any study of sickness absence isthe considerable effect imposed by variables such as age,sex, occupation, status, sick pay arrangements, and thelike. For this reason, a brief outline of some of the morerelevant characteristics of the refinery population will beincluded before describing the criteria used to identifythe groups of men and the methods by which they werestudied.

Some Characteristics ofthe Refinery PopulationThe refinery, situated in a semi-rural area in south-eastEngland, operates a continuous process productionsystem. The reference year in this investigation wasI964, and the number of employees on the payroll at theend of that year was 2,076. Of these, 95 were female,598 were male staff on an annual salary, and 1,383 werehourly paid men. Since the investigation was concernedonly with the last group of men, no further referencewill be made to females or to staff. The industry isrelatively prosperous and has developed considerably inthe last 20 years. Almost half of the men are employedon three-cycle, continuous shift work; the remainderare on a five-day week. The working week has been 40hours since I962, and overtime among the men averagedfive hours per week in I964. All men retire at the age of60; and the mean age of the population at the end of

i964 was 40 0 years. An 'ex gratia' sick-pay schemeintroduced in I946 provides for payment of full wagesto all employees with over six months' service from thefirst day of absence. The duration ofpayment varies withthe length of service from an initial three weeks at fullpay followed by two weeks at half pay increasing to 27weeks at full pay and 24 weeks at half pay after 15 years'service. Chronic sick are kept on until they are fit toresume some sort of work or reach a pensionable age.Single-day sickness absences are eligible for sick pay,and at the time of the survey all sickness absences had tobe covered by medical certificates.

Medical Services and Records A medical depart-ment provides 24-hour cover with orderlies on shift aswell as a full-time doctor and qualified nursing staff onday work. A sickness absence record card is maintainedfor every employee upon which details of all absencesattributed to illness or injury are recorded. Thesedetails include the dates of commencement and return,the duration of each spell in calendar days, and thediagnosis. All employees absent for two weeks or moreare seen by the doctor when they return to work. Theserecords have been scrupulously maintained and arechecked against the wages returns for accuracy.

Selection of Men for the Survey There are twomain measures of sickness absence-the number ofspells experienced in a period (frequency), and thenumber of days of absence in a period (severity).The personal sickness absence record cards of all

hourly paid men employed throughout I964 wereinspected as soon as the records for the year werecomplete. The number and duration of the spells inthat year are shown in Table I.

TABLE ITOTAL NUMBER OF SPELLS OF SIcKNEss ABSENCE AND THETOTAL DURATION IN CALENDAR DAYS AMONG ALL 1,350

MEN EMPLOYED THROUGHOUT I964

Spells Days

Total recorded i.... .. ,65I I7,490Mean per man I.... .. 22 12-96Range .. .. .. .. O-II 0-245Standard deviation .. .. .. I-46 20-8I

Two of the groups for the survey were selected fromthis information, 'frequently sick' being men who hadfive or more spells of absence in the year, and 'long sick'men who had 6o or more calendar days of absence inthe year. These levels were chosen because they werethe nearest suitable numbers above the sum of the meansand twice the standard deviations, and they also provideda convenient number ofmen for study.The other extreme of sickness absence which it was

decided to investigate was a negative one, in other words,men who did not have any sickness absence. Sincemore than one man in three had no such spell in i964,

107


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/

io8 P. J. Taylor

the personal record cards were inspected to find out howmany men had had no sickness absence for several yearsand the results are shown in Table II.

TABLE IINUMBER AND PROPORTIONS OF HouRLY PAI

MEN HAVING NO SicKNEss ABSENCE FOR UP TO I5 YEARS'CONTINUOUS EMPLOYMENT UNTIL I964

No. of No. of No.Calendar Years at Men at without NeverYears Risk Risk Sickness Sick (%)

Absence

I964 only I I,350 531 39.4I963-64 2 1,294 294 22.7I962-64 3 I,236 I82 I4.7i96i-64 4 I,I62 133 I1I5I960-64 5 1,094 I02 9.3

I959-64 6 985 74 7.5i958-64 7 868 64 74I957-64 8 721 50 6-9I956-64 9 696 42 6-01955-64 10 623 3I 5.0

I954-64 II 604 26 4 3I953-64 I2 566 I8 3.2195244 I3 523 I5 2-91951-44 14 450 I2 2.7I950-64 1I5 371I II 3-O

It was decided that the 'never sick' group shouldconsist of all the 5o men who had not been awaysick for a minimum of eight years (I957-64). Thisperiod was chosen to obtain a large enough number ofmen for the clinical study.

Finally a control group was selected. Ideally, thisshould have been a large and representative cross-section of all the men. Unfortunately, since the threegroups already selected included 147 men, it was notpossible to have a control group of similar or larger sizeas the time available for the survey was limited and itwas also considereti important to see all the men as soonas possible after the end of I964, since any short-termfactors affecting absence would be more likely to beoperative or fresh in the memory.

Controls were therefore selected for each man in thefrequently sick group. Their names were obtained fromthe nominal roll of all employees after excluding the I47men already identified, and were matched within (a)identical occupation and department; (b) similar lengthof service, to less than two years; and (c) similar age, toless than three years. Where more than one man providedan equal match the one whose name came nearest inalphabetical order was selected.

Size of Groups and Reasons for Exclusion fromthe Survey The total number of men selected forthis survey was 207. Thirteen were not seen and thereasons for their exclusion are described below. Thenumbers of men seen in each group were as follows:

Frequently Sick Sixty men were identified but fourofthem left the Company early in I965 before they couldbe contacted. This left 56 men, all of whom were seen.

Controls Since the men were individually matched,the four men paired with those who had left in the firstgroup were not seen (one of them had also left theCompany). All the remaining 56 men were seen.

Long Sick Of the 37 men originally identified, onewith motor neurone disease never returned to work anddied in December I965, and one man refused to take partin the survey. Thus 35 men were seen.

Never Sick Three men out of the 50 identified(Table II) refused to take part in the survey and thus47 men were seen.A refusal rate of four out of the I98 men invited to take

part in the survey represents just over 2% and is notsufficiently large to invalidate the results. The fact thatthree of them came from the never sick group may wellbe significant in the light of some of the observationsmade on this group which will be discussed later.

Methods

Management agreed to allow the survey, and con-sultation with senior shop stewards produced assurancesof their full support. A questionnaire was designed andmodified by a trial on I2 staff employees.Each man was approached individually by the author

and invited to help in the survey. It was explained thatany information obtained would be treated in confidenceand would not be included in his ordinary medical notes,and that no action which could affect his career would betaken unless he requested it. Each interview took aboutone hour.The questionnaire was not changed at all throughout

the survey and, to minimize the effect of alteration inobserver bias, men from the four groups were seen inrotation. They could be seen only during their workinghours; this and the fact that many were on shift workmade a predetermined random order impossible.The information collected about each man came

partly from documents and records but mostly from theinterview and examination; it can be summarized underthe following headings: personnel records; medicalrecords; occupational history and factors related to thejob; social background and present home circumstances;family and personal past medical history; presentsymptoms; physical examination and special investiga-tions; and, supervisor's work performance rating.The questions about present symptoms included

those of the M.R.C. Short Questionnaire on RespiratorySymptoms (I960). The physical examination wasstandardized as far as possible to provide measurableinformation on function. Special investigations werelimited to tests simple to perform and acceptable to themen. For the latter reason routine blood samples werenot taken. Urine was tested for protein and glucose,respiratory function was measured by the Vitalograph


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/


direct spirometer, and a standard chest radiograph wastaken. It was decided to use some form of personalitytest and, after consideration of those in current use, thenew Eysenck Personality Inventory was chosen. Thisconsists of two question papers and is designed tomeasure extroversion and neuroticism.The supervisors' ratings of work performance were

obtained after all the men had been seen and wererequested verbally and in confidence. They were askedto rate their men on working performance and abilityon a four-point scale: exceptional, above average,average, and below average. Such ratings are open tocriticism, but it was felt that some indication of workperformance should be included in the analysis.

Results

A large amount of information was collectedduring this survey and over 200 facts about eachman were examined, separately or in combination,for their association with sickness absence. In-evitably many of them showed no real differencebetween the groups, but a few of these negativeobservations are of interest and will be mentionedin this section. Factors which were found to beassociated with sickness absence will be describedin more detail.

Statistical comparisons are difficult to make inthis sort of survey since, by definition, the groupsare not representative samples of the whole popula-tion from which they were selected. Most of theinformation collected is also not available for thewhole population. Furthermore, it has been shown(Taylor, I967a) that indices such as sicknessabsence and lateness as well as occupational injuriesare distributed in a negative binomial manner sothat, with the numbers involved in this study andwithout elaborate correction factors, distributionfree tests must be used. For much of the data,therefore, the only valid means of comparison is theuse of a relative rate having the characteristic of aproportion.

Personnel RecordsAge and Occupation Although a wide age range

was found in each group, the frequently sick tendedto be younger whilst the long and never sick wereolder, but for the latter group this was partly due tothe method of selection which required a minimumof eight years' continuous service. Shift workerswere rare among the frequently sick, commonamong the never sick, and made up almost half ofthe long sick group. As in all factories, workers canbe divided into three main functional groups-production, maintenance, and service. Thefrequently sick were mostly from the maintenancegroup whereas the never sick were mostly produc-

don workers. These findings are presented inTable III. The control group had by their selectionthe same age and occupational structure as thefrequently sick group.

TABLE IIIAGES AND OCCUPATIONS OFTH THREE MAnN GROUPS AND OF

THE WHOLE POPULATION FOR I964

Frequently Long Never All HourlySick Sick Sick Paid Men

No. of men 56 35 47 I,350Mean age (yrs) 32-8 45-4 48.I 40-0Age range (yrs) I7-55 20-59 25-59 I6-59

Production (%) I8 37 62 42Maintenance (%) 73 29 I3 40Services (%) 9 34 25 I8

Shift work (%) 9 49 72 49

The number of promotions achieved by eachman in his first three years of company servicewas noted and the means for the groups showed nosignificant difference.

Overtime The total hours of overtime worked byeach man in I964 were divided by the number ofweeks he actually worked, thus allowing for absenceon holiday and sickness. Each group showed a widerange; some men had done none, some a good deal.The mean figure for the whole population in I964was 7X6 hours per worked week, and that for eachgroup was:

Frequently sick I I *4 hours (range 0-23 4)Controls i 0o hours (range 0-24 4)Long sick 7-2 hours (range 0-23.5)Never sick 5 2 hours (range o-I8 i)

One important reason for this difference betweenthe groups is that overtime in the daywork mainten-ance force is always higher than in shift productionand service workers. There was no correlationbetween sickness spells and overtime because a fewmen with high rates did no overtime and others withno sickness spells did a lot. For instance, the manin the control group who had worked an average64-4-hour week in i964 had no sickness absencethat year. This problem will be referred to againbelow.

Lateness and Absenteeism Positive correlationswere found between sickness spells and lateness in'clocking on' to work (r = +±o4), and also withepisodes of absence with or without permission forreasons other than sickness, holidays or union

Iog


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/

meetings (r = + os5). Since all these indices aredistributed in a negative binomial pattern, standarderrors and the usual tests of significance cannot beapplied.However, a comparison of the proportions ofmen

with more than four episodes of lateness and alsothose with over four other absences in I964 (TableIV) allows the use of a statistical test. These levelswere chosen as they were the nearest whole numberabove the mean lateness and other absence episodesfor the whole population.

TABLE IVMEN IN EACH GROUP WITH MORE THAN FouR EPISODES OF

LATENESS AND ABSENTEEISM IN I964

Lateness AbsenteeismGroup

No. % No. %

Frequently sick .. .. 31 55 27 48Controls .. .. .. i8 32 1I3 23Long sick .. .. .. 5 I4 I2 34Never sick .. .. .. 0 0 I 2

Al men .. .. .. 54 28 53 27

The differences between the proportions of thefrequently sick and controls are significant (P<o os)for both measures, as are those between the long andnever sick groups. With these as well as withsickness spells, the frequently sick and never sickgroups are clearly at opposite ends of a spectrum.

Medical Records The scope and nature of thepre-employment medical examination in this factoryhas changed only in detail in the past 30 years. Asin many firms with a medical officer, it consists of aquestionnaire about past health, which the man isrequired to sign, followed by a routine physicalexamination.

These records were inspected for evidence of bothtrivial and also more clinically important conditions.The latter were those which could be expected toaffect the type of work the man could undertake andincluded, for example, chronic chest disease, provenpeptic ulcers, definite rheumatic heart disease, anddisabling orthopaedic conditions.

Neither the trivial nor the clinically significantconditions differed statistically in incidence betweenthe four groups of men studied. The former wererecorded in one quarter of the men, and the latter in5 per cent. One of the never sick group was on theRegister of Disabled Persons before he wasemployed in I945. He had then, and still has,

definite asthma but has never lost a day from workwith this or any other condition.Ten of the men were found to have concealed a

clinically significant fact in their history at thepre-employment examination, but these also wereequally divided between each group. Thus thepre-employment examination as it had beenrecorded was of no predictive value in thisinvestigation.

Sickness Absence before and after I964 Thecomprehensive records of every spell of sicknessabsence enabled a number of different methods ofanalysis to be used. One aspect studied was a searchfor evidence of a 'sickness absence habit', in otherwords, the degree of consistency in sickness be-haviour over the years as shown by the groups andalso by the men themselves.

Since i964 was the reference year, the meanannual spells per man was calculated for each groupfrom the first year of employment up to the end ofI963. Two years have elapsed since the referenceyear so this calculation has been done also for I965and I966 combined. The results are shown inTable V and suggest that each group had reasonablyconsistent mean rates before and after I964.

TABLE VGROUP MEAN ANNUAL SPELLS OF SICKNESS ABSENCE PER MAN

BEFORE, DURING, AND SINCE I964

Mean Spells/Man/Years

Frequently Long NeverSick Controls Sick Sick

From first year ofemployment to I963 3-46 171 I46 0-12In I964 .. .. 557 I55 2-09 0In I965 + I966 .. 439 I-98 I-88 o-I6

To find out whether the sickness absence recordin the first year of employment was significantlydifferent between the frequently sick, who had amean of 2-9I spells that year, and their matchedcontrols, who had had I 36 spells per man, the 't'test was applied for each of the 56 pairs. Thisdifference was highly significant (t = 36-47P <O*OOi).Although the 't' test compares individuals with

their matched controls, there remained to becompared each man's sickness spells in I964 withhis own record in other years. This has been doneby calculating coefficients of correlation in both thefrequently sick and the control groups. The resultsare shown in Table VI.

P. J. TaylorIIO


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/


TABLE VICOEFFICIENTS OF CORRELATION BETWEEN SICKNESS SPELLS IN

I964 AND EACH YEAR BETWEEN I960 AND I966

Year compared Frequentlywith 1964 Sick Controls

I960 +02 +02I96I +o*6 +0°3I962 +±-I +0.7I963 +0.3 +O*8I965 +0.3 +O*6I966 +0 4 +0°3

The observed correlations are variable and notparticularly high, even though they are all positive.The low values are due largely to a number of menin each group whose sickness spells differ widelyfrom one year to the next, although most men ineach group show reasonable consistency. Thus inI962, when the correlation in the frequently sickgroup was only + O-I, the group as a whole had3 -64 spells per man, whereas the control group hadI-75 spells.Three examples of annual sickness absence spells

over a io-year period of employment from men inthe frequently sick group illustrate this in Table VII;one (case 5) was consistently high, the second(case I9) was intermittently high, whereas the third(case I62) had frequent spells only after I960.

TABLE VIITHREE EXAMPLES FROM THE FREQUENTLY SICK GROUP WITH

CONSISTENT, INTERMITTENT, AND OCCASIONAL YEARSWITH MANY SICKNESS ABSENCE SPELLS

No. of Spells in Each YearYear

Case 5 Case I9 Case I62

I964 6 5 5I963 5 6 6I962 5 2 4I96I 6 8 41960 7 5 2

I959 5 I 2I958 4 5 3I957 6 2 21956 5 2 II955 4 4 2

One further point deserves mention. Although inany one year about one man in three in the wholepopulation has no spell of sickness, this wasextremely uncommon among men in the frequentlysick group, and although the correlations were low,only five of the 27 men with I0 years' continuous

service ever had a single year without a sickness spell.The long sick group, on the other hand, showed

no such consistency and, with a few exceptions,their sickness absence before and after I964resembled that of the control group and of the wholepopulation. It appears from their age distribution,occupations, and sickness absence that most of themcould be considered as a reasonable sample of thewhole population. Further evidence on this aspectof the long sick group will be presented in a laterpaper.

Diagnoses of the Sickness Spells in I964 Althoughthe men in the frequently sick group had nearlythree times as many spells of sickness absence astheir matched controls during the reference year,there was little difference in the types of illness.Upper respiratory tract diseases, including influenza,caused 20% of all spells in each group; gastro-enteritis was the next most common (I9%), andmusculo-skeletal conditions were third (i I %). Onlyneurotic conditions showed any significant differ-ence between the two groups, causing 5.5% of thetotal spells among the frequently sick and I-I% ofthose in the controls, a difference significant at the0o05 level.When the number of men having spells in each

disease category was analysed, however, thefrequently sick outnumbered the controls for everydiagnosis (Table VIII).

TABLE VIIIMEN IN FREQUENTLY SICK AND CONTROL GROUPS HAVING

SPELLS OF SICKNESS ABSENCE IN I964 IN BROAD DISEASEGROUPS

Frequently Sick ControlsDiagnosis . _

No. Propor- No. Propor-of tion of of tion ofMen 56(%) Men 56(%)

Upper respiratory tractdiseases, including 'flu 44 78.5 21 37.6

Gastro-enteritis . . . . 32 57*2 I4 25*0Musculo-skeletal disorders 23 41-2 7 I2-5Gastritis and dyspepsia. . 17 30°4 7 I2.5Neuroses .. .. 2 2I 4 I I *8Ill-defined .. .. 12 2I 4 4 7.2Non-occupational injuries io I79 2 3-6Bronchitis .. .. 8 I43 7 12.5Migraine .. .. 89 I i 8Sepsis .. .. .. 5 89 I I-8Occupational injuries .. 4 7-2 I I *8Other conditions .. I9 340 7 12.5

A comparison of the proportion of men in eachgroup who had spells of each type of disease

III


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/

shows that the differences were highly significant(P<0o003) for upper respiratory tract diseases,gastro-enteritis, musculo-skeletal disorders, andneuroses; and significant at the o o5 level forgastritis, non-occupational injuries, and ill-definedconditions. Only for bronchitis are the groupsclosely similar.Thus the frequently sick group had more spells

of all the usual diseases affecting the controls, and,with the single exception of neuroses, they were notespecially susceptible to any one type of sickness.The illnesses of the long sick group, however,

were of quite a different pattern. The commonestdiagnostic group was bronchitis and asthma, whichcaused 23% of all spells; musculo-skeletal conditionscaused I9% and upper respiratory tract diseasescame third with I2% of spells. The order ofduration of absence, however, was different again.Six spells from cardiovascular disease caused 780days away, I2 spells from musculo-skeletal con-ditions caused 723 days away, and the I7 spells frombronchitis and asthma came third with 394 days.

Occupational Injuries The number of suchinjuries reported by men in each group differedconsiderably both in i964 and in earlier years. Thefrequently sick had had more injuries than thecontrols, who in turn had reported more than thelong sick and never sick groups. In all the men whohad been employed for the five years i960-64, thecorrelation coefficient between total spells of sicknessand total number of injuries was + 0o4. It shouldbe emphasized that almost all these injuries wereminor and thus ascertainment depends largelyupon the tendency of the man to report.

Occupational History and Factors relatingto the Job The frequency of changing jobs beforejoining the company was compared but no sig-nificant differences were found between thefrequently sick and control groups. This is becausemany of the maintenance workers had previouslybeen employed by engineering contractors whousually take on labour for each separate contract.

Job Satisfaction The attitude ofthe man towardshis job proved to be of considerable importance.This was assessed both directly, by inviting hisopinion of the job as such, and indirectly, by askingwhether he wanted more responsibility and whetherhe would like a change of foreman. The results ofthese questions are presented in Tables IX and X.

In this direct assessment of job satisfaction,definite enjoyment of the job was very significantlyless common (P<o0oo3) in the frequently sick thanin the control group. The never sick were almost

TABLE IXOPINIONS EXPRESSED ABOUT THE JOB

Definitely Indifferent DislikeJobGroup Enjoy

No. % No. % No. %

Frequently sick.. 26 46 20 36 IO i8Controls.. .. 45 8o 9 I6 2 4Long sick .. 26 74 7 20 2 6Never sick .. 45 96 I 2 I 2

ARlmen.. 142 72 37 20 I5 8

TABLE XTHE DESIRE FOR MORE RESPONSIBILITY

Actively Would Do notGroup Want Accept Want

No. % No. % No. %

Frequently sick.. 35 62 2 4 I9 34Controls. . .. I7 30 I4 25 25 45Long sick .. 10 29 4 II 21 6oNever sick .. 5 II IO 21 32 68

Al men.. 67 35 30 15 97 50

unanimous in saying that they definitely enjoyedtheir job. At the other extreme of opinion, only thefrequently sick had a significant number of menadmitting that they disliked their job (P<o0o5).An active desire for more responsibility was found

in a very significantly greater proportion (P<o oo3)of the frequently sick than in the control group.Few of the never sick felt this way. The inter-mediate answer-that he would accept moreresponsibility if it was offered-tended to be lesscommon in both groups with high sickness.The other indirect measure of job satisfaction

was provided by a question about wanting a changeof foreman. The only group with a significantproportion (20%) wanting such a change was thefrequently sick.

The J7ourney to Work There was no significantdifference between the groups in the time taken totravel to work; three-quarters of each group tookless than half an hour. The method of transport,however, proved to be of some interest. Over halfthe men came to work in their own car, but only28% of the never sick did so. This was not becausethey came in other men's cars but because manymore of them (40%) cycled to work. When one

P.3'. TaylorII2


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/


recalls that these men were older and more highlypaid than those in the other groups, and most ofthem were shift workers, this observation becomeseven more remarkable.

Performance Ratings There was no firm relation-ship between the supervisor's assessment of workperformance and sickness behaviour. Furthermore,some men who said they liked their jobs were givenbad reports, and, of the IO frequently sick who saidthat they disliked their jobs, four were rated as'above average', five as 'average', and only oneas 'bad'.

Social Background and Present HomeCircumstancesFamily Background Men bom or brought up in

the neighbourhood of the factory with relativesliving in the same area formed rather more thanhalf the total and were evenly distributed betweenthe four groups. Social class, based on the father'soccupation, and also family size, in number ofsiblings, showed no difference between the groups.The attitudes of the men about their own child-

hood, however, revealed marked differences. Eachman was asked whether he now considered that hisown childhood had been very happy and secure,quite happy, or unhappy. Their answers are shownin Table XI.

TABLE XIGROUP ATTITUDES TowARDs THEiR OwN CHILDHOOD

Very Happy Quite UnhappyGroup Happy

No. % No. % No. %

Frequently sick.. 28 50 I5 27 13 23Controls.. .. 42 75 9 I6 5 9Long sick .. 24 69 3 8 8 23Never sick .. 4I 87 5 II I 2

AR men.. 135 70 32 i6 27 14

The frequently sick and controis differed in boththe 'very happy' and 'unhappy' answers to a sig-nificant degree (P<0oos), as did the long and neversick groups. An unhappy childhood was claimed byalmost one quarter of both the frequently and thelong sick groups. The incidence of a brokenparental marriage was low for all, but the long sick(I4%) and the frequently sick (7%) had higherrates than the never sick (2%) and controls (4%).If these two pairs of groups are combined, thedifference becomes significant at the o0os level.

Present Social Conditions Almost all the men inthe survey were married and only seven had beendivorced, three in the never sick group, two in thelong sick, and two in the control group. Housingconditions were similar for all with almost half themen living as owner occupiers. Illness of a wife orchild in i964 was uncommon and showed nodifference between the groups.The number of dependent relatives was counted

for each man, excluding wives or children in full-time employment. The mean number of suchdependants and the proportions of men with none,a few or several were very similar for each group.The mean number of sick spells in i964 for eachman with o, I, 2, 3, and 4 or more dependants wascalculated for all the men in the survey. Thisshowed that men with one dependant had only I.5spells in the year whereas men with none and morethan one had about 2.7 spells each.

Other Work Second jobs, either as self-employedor 'helping out friends', were admitted to by i6 menin all-io from the frequently sick group (i8%),four from the never sick group (9%), and two fromthe long sick group (6%), but none of the controls.This difference between the frequently sick andtheir matched controls is highly significant(P<o0oo3). The number of such men among thenever sick group could be misleading until onerecalls from results already presented that this groupdid little overtime. The men who did a mean weeklyovertime of at least i0 hours and also a second jobcame only from the frequently sick group.

Voluntary work, on the other hand, showed noassociation with sickness even when overtime wastaken into account.

Family Medical History One factor in whichthe groups differed was the loss of a parent by deathbefore the age of 6o years. These men, expressedas a proportion of those whose parents had, orcould have, reached that age, amounted to 42% ofthe frequently sick group and 20% of the controls,a difference significant at the o0os level, but theother two groups came between these two extremes.The incidence of parental bereavement in childhoodwas higher in the frequently sick group but did notdiffer significantly from that in the other-groups.

Past Medical History As with the pre-employ-ment medical records, most of the factors invest-igated in this section of the survey were morenotable for their similarity than for their differences.

Previous hospital admissions were divided intothose as a child (under i6 years), which did notdiffer between the four groups, and admission as an

II13


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/

adult. In this latter category only the long (86%)and never sick (26%) groups differed significantlysince both the frequently sick and control groupswere similar with 46 and 43% respectively.Three other points, however, showed significant

differences-the incidence ofa 'nervous breakdown',proven peptic ulceration, and one or more episodesof loss of working time due to back pain. Thenumbers and proportions ofmen in each group withthese facts in their past history are shown inTable XII.

TABLE XIIINCIDENCE IN PAST HISTORY OF 'NERVOUS BREAKDOWN',PEPTIC ULCERATION, AND INCAPACITATING BACK PAIN

Nervous Peptic Incapac-Group Breakdown Ulcer itating

Back Pain

No. % No. % No. %

Frequently sick.. II 20 3 5 32 57Controls.. .. 3 5 0 0 I8 32Long sick .. 6 I7 7 20 i8 53Never sick .. I 2 0 0 6 13

All men .. .. 2I II 10 5 74 38

All three conditions were more common(P<<oos) among both the frequently and long sickgroups than in the control or never sick groups.However, although a past history of a peptic ulcerconfirmed by barium meal was most common inthe long sick group (although only four of the 35men were absent with the diagnosis in I964), theother two conditions were most prevalent in thefrequently sick group.

Present Medical Symptoms Here, too, mostof the factors studied occurred in similar proportionsin each group. Despite the findings regarding pepticulceration and incapacitating back pain in the pasthistory, indigestion and muscle or joint pains wereremarkable for their closely similar prevalence-about 40% in each of the four groups.Only in the proportions of men who stated that

they had four or more upper respiratory tractinfections each year, that they had a cough all day(M.R.C. grade II), and that they were troubled byconstipation, did the groups differ to a significantdegree (Table XIII).The finding of more frequent upper respiratory

tract infections and persistent coughs in thefrequently sick and long sick groups might have beenexpected. However, the significant difference

TABLE XIIIINCIDENCE OF MEN ADMITTING TO AT LEAST FOUR EPISODES OF

U.R.I. ANNUALLY, ALL-DAY COUGH, AND TROUBLESOMECONSTIPATION

4+ U.R.I. All-Day Troublesomeper year Cough Constipation

GroupNo. % No. % No. %

Frequently sick.. II 20 7 I3 i6 29Controls 2 4 3 5 6 I ILong sick .. I 3 7 20 13 37Never sick .. 3 6 I 2 6 13

Al men.. .. 17 9 i8 9 41 2I

between the proportions affirming that they weretroubled by constipation in these two groups whencompared with the control and never sick groups(P<o0o) came as a surprise.Although the proportions with an all-day cough

differed significantly, those with sputum, althoughsuggestive, were not significant. The long sickgroup was most affected, and in I964 nine of the35 men had been absent with a diagnosis of asthmaor bronchitis.

It is relevant to note the smoking histories at thispoint. Similar proportions of each group werenon-smokers and ex-smokers. The only significantdifference found was that 35% of the frequentlysick and I4% of controls smoked more than I4 g.of cigarette tobacco daily.

Physical Examination The standardizedexamination did not show many significant differ-ences between the groups although limitation ofthe range of movements of both cervical and lumbarspines was found more commonly among the longand the frequently sick (P<o os).Poor dental hygiene, which was easily visible on

inspection of the mouth with a torch and spatulaonly, showed a suggestive trend, being found in29% of the frequently sick and 43% of controls,and in 24% of the long sick and 4I% of the neversick. It appears that men with more sickness absencetend to take more care of their teeth.

Finger nails were inspected for evidence of nailbiting. A number ofmen explained that since beingfitted with dentures they had been unable tocontinue the habit. The number in each groupshowing definite evidence of nail biting expressedas a proportion of those with their own front teethwas very similar in the frequently sick, control, andlong sick groups, amounting to I9% for the threegroups combined, but for the never sick the

P. 3' TaylorII4


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/


proportion was higher (28%), an interesting thoughnot statistically significant difference.Any idea that the never sick group might consist

of exceptionally fit men was soon disproved.Thirteen of them (28%) were found to haveabnormalities, a proportion similar to that found inthe frequently sick and control groups. Theconditions found among them included a case ofpreviously undiagnosed diabetes, two cases ofchronic bronchitis and one with long-standingasthma, two with gross kyphoscoliosis, and one case

of mild Parkinson's disease. Others in this group

included two men with cranial nerve palsies andone man each with a mixed parotid tumour, a

hydrocoele, a spermatocoele, and finally a man

with severe pes planus and hallux rigidus.

Special Investigations Lung function testsof vital capacity and forced expiratory volumeconfirmed the physical signs, and, as expected, thelong sick group had the highest proportion of menwith a significant reduction in function. Apartfrom this, no important differences were foundbetween the groups when age, height, and smokinghistory were taken into account. Regression curves

have been calculated for F.E.V.1.0 on age and heightfor the frequently sick and control groups butshowed no appreciable difference between them.Similar results were obtained from the chestradiographs.

The Eysenck Personality Inventory Both testpapers were used and the mean scores for each groupare shown in Table XIV. The normal values andtheir standard deviations (Eysenck and Eysenck,I964) are also included in the Table.These results suggest that the never sick are

characterized by introversion and stability, the longsick by introversion and neuroticism, and thefrequently sick by a degree of extroversion and a

higher neuroticism score than the controls. Thehigh lie score of the never sick is of some interestand will be discussed later.These results suggested that it might be of value

to estimate the relation of these scores to other

TABLE XIVRESULTS OF EYsENcK PERSONALITY INVENTORY

(PAPERS A PLUS B)

Group Extroversion Neuroticism Lie ScoreScore Score

Frequently sick .. 27-6 I9-I 5-3Controls .. .. 263 I54 5-ILong sick.. .. 23-4 22-5 5.INever sick .. 239 I22 7-0

Normal values .. 262±7-8 I9*6±9*o 3*6±2-5

factors already shown to be associated with sicknessabsence in this survey. To avoid errors due to a

single year's experience the figures for the I49 menout of those seen who had been continuouslyemployed for five years were used, and the correla-tion coefficients are presented in Table XV.

All the five variables correlate with extroversionand, except for lateness, all are over + 0-2. Thislevel, although low, is comparable with otherpublished results. The usual tests of significancecannot be applied since none of these variables isnormally distributed. Neuroticism correlates lesswell and is negligible with occupational injuriesand other absences. The lie score has a negativecorrelation with all variables in the same order ofmagnitude as the neuroticism score.

Discussion

This type of survey imposes a number of limita-tions which should be considered before attemptingto interpret the results.The risk of becoming ill can never be equal for

all men since, apart from variations in the externalinfluence or stimulus there are also differences inindividual susceptibility. The difficulty is increasedbecause this survey is not about sickness, but ratherabsence from work certified as due to medicalincapacity. Spells of absence arise because the manreports the condition and hence there are alwaysunknown errors in ascertainment. All doctors have

ILE XVCORRELATIONS BETWEEN PERSONALITY INDEX ScORES AND EACH MAN'S TOTAL SPELLS OF SICKNESS, OCCUPATIONAL

INJURIES, TREATMENT ROOM ATTENDANCES, LATENESS AND OTBER ABSENCE FOR I960-64 INCLUSIVE (149 MEN)

Personality Scale. Sickness Injury Treatment Lateness Other Mean ScoreAttendances Absences

Extroversion .. .. .. .. +0-22 +0-23 +o-26 +0°I5 +0-23 25-I2Neuroticism .. .. .. .. +O-I7 -o-o6 +0-II +0oI7 +0o07 I6*46Lie score .. .. .. .. -0-5 -0I4 -016 -O-I9 -000 5'83

I15


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/

experienced some patients who demand sicknesscertificates for trivial conditions whereas othersrefuse to take time off work for quite seriousillness.

Furthermore, any results of this survey can betaken as valid only for the refinery population fromwhich they were obtained, and without furthersimilar studies must not be inferred to otherworking populations. This is illustrated by theattempt of Castle (I956) to confirm an observationby Hill and Trist (I962) that able-bodied absentee-ism was an acceptable mode of behaviour. Theobservation had been made in a steel foundry inthe Midlands, and Castle found that this type ofbehaviour was not acceptable among workers in aphotographic factory near London where it wasreplaced by 'sickness'.

Sickness Absence Liability and ResistanceThe evidence from this survey suggests that thegroups did show a consistent tendency to continuewith high, average or low frequencies of sicknessspells. Among individual men, however, thecorrelations do not follow the criteria laid down for'accident proneness' by Cresswell and Froggatt(I963) as they are neither stable nor independent ofthe length of interval between the periods studied.Nevertheless, few of the frequently sick menever had a single year without a spell of sickness,although one third of the factory population had nospells each year.

It seems reasonable to conclude that personalliability to sickness absence does exist and that,although for most men this is only prominent fora few years at a time and can be dramatically affectedby a change of job (Taylor, I967a), it sometimesrecurs, and for a few it appears to be relativelystable. The phenomenon of sickness absenceresistance has seldom been investigated, although itwas observed in the work of Hinkle, Plummer, andWhitney (I96I) and also by Kessel and Shepherd(I965) in their survey of a London suburbanpractice in which 3% of the population had notconsulted a doctor for IO years. The evidencefrom the present survey suggests that the never sickgroup is more homogeneous than any of the otherthree groups.

Factors Related to Sickness Absence Mostof the factors found to be significantly associatedwith sickness absence can be considered asbehavioural rather than medical, environmental oreconomic. The strong associations with otherforms of absence and lateness are not surprisingand have already been observed by social scientistssuch as Behrend (I95I) in her study of industrial

absenteeism. The very low rate of such absencesamong the never sick provides confirmatoryevidence not previously reported.The failure to show a significant association

between sickness absence and overtime, despitesuggestive trends, was also noted by Lokander(I962). Nevertheless the combination of a highlevel of overtime and a second job was significant.

Before assessing the results of the interview andpersonality inventory one should consider thevalidity and reliability of the answers. Although itis certainly possible that many of the men may havedeliberately lied, the voluntary nature of the inter-view and the explanation given to each man,together with the assurance of confidentiality, makesuch a suggestion unlikely to be true.Vernon (I964) has considered the problem at

some length and stated, 'The ordinary man who isnot highly educated is stereotyped, intolerant, selfsatisfied, non-psychologically minded, and rejectsany imputation of psychological weaknesses: where-as the better educated man takes a more detachedview of himself and is more aware of his conflictsand anxieties.' Only 17 men in the survey had hadany secondary education and all of these had leftschool at the age of I6; thus the men seen could bedescribed as 'ordinary' in this context. It is essentialto appreciate that the answers should be taken as asample of the person's self-concept and memoryand not necessarily entirely true in an objectivesense.The reliability of the answers could have been

tested by repeating the questions after a lapse oftime, but this was impracticable. However, work onpersonality tests has shown a high degree ofreliability in retests (Vernon, I964; Eysenck andEysenck, I964) and it seems reasonable to assumethat this would be true for this survey.

In the results, the attitudes of the man towardshimself, his work, and his own health appeared tobe the most important factors of all. As Chiesman(I957) expressed it, 'Staying away from work orreturn to work does not depend upon the start andcessation of a particular disease process, but on thestandard of health which the patient sets himself or,in other words, on his ability to adjust himself tohis working environment.'

In the medical history the most importantassociations with sickness absence were a history ofprevious neurotic illness, peptic ulceration, andloss ofwork from back pain. The present symptomsof importance were frequent upper respiratorytract infections, a persistent cough, and beingtroubled by constipation. Alcoholism, found to beof great importance by Lokander (I962), was notsignificant in this survey, but the isolation of the

P. J. TaylorII6


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/


refinery and the strictly enforced rule that noalcohol is allowed inside the gates would ensurethat alcoholics selected themselves away from work-ing there.The never sick group tumed out to be the most

interesting one. Almost without exception theydenied that there was anything wrong with theirhealth, their home or their work. The three whorefused to take part in the survey did so politely butfirmly, and all stated that since there was nothingwrong with their health there could be no point inany medical examination. It is suggested that theymay have been afraid of what an examination mighthave revealed. It is interesting that Kessel andShepherd (I965), in their survey of a generalpractice, failed to achieve an interview with as manyas 30% of the Io-year non-attenders.The high lie score in the personality inventory

among the never sick should be considered. Gibson(I962) stated that a high score could arise fromeither deliberate faking, a response in terms of anideal self concept, or an honest lack of insight. Inview of the atmosphere of the interview and themen's apparently frank response to other questionsthe honest lack of insight or an ideal self conceptwould appear to be the reason for the high lie scorein this group.The higher extroversion score found in the

frequently sick group confirms several studies byindustrial psychologists who have shown that thistends to be associated with various forms of with-drawal from the work situation (Cooper and Payne,I967) and with delinquency (Eysenck, I964).Eysenck's hypothesis implies that extroverts con-dition poorly and therefore find it more difficult tosettle in regular routine jobs whereas introvertscondition more readily. The higher neuroticismscore in the long sick group may well be secondaryto the long-term sickness absence that these men hadexperienced the year before; prospective studiesshould be made to confirm or refute this suggestion.

Job Satisfaction-Association or CausationAlthough a number of factors were found to beassociated with sickness absence, few could fulfillthe nine criteria listed by Hill (I965) which shouldbe considered before they can be proposed ascausative. These criteria were strength, consistency,specificity, temporality, biological gradient, plaus-ibility, coherence, experiment, and analogy.

Job satisfaction, assessed directly or indirectly,does fit several of these tests. The association hasstrength and is reasonably consistent; specificityhas been suggested by other work in the study ofabsenteeism; temporality and a biological gradientcan be seen; it is plausible and coherent. Experiment

in the sense of change in supervision or change injob has been achieved in isolated cases (Taylor,I967a).Both age and occupation might be associated with

job satisfaction as well as with sickness behaviourand could thus influence the results. Since thenumber ofmen involved is small it was only possibleto measure the effect of age by comparing menunder and over the age of 40 years. The proportionswith a desire for more responsibility, liking the job,and wanting a change of foreman in the frequentlysick and never sick groups were still significantlydifferent in both younger (P<o0os) and older(P<o0oo3) men. The effect of occupation wasallowed for in the selection of the control group, buta second control group would have been advisablefor the never sick had time and resources allowed.

Clearly more work is required to clarify thisproblem. Do the men whose sickness recordimproves dramatically after a change continue togive a high extroversion score ? The observationthat three-cycle, continuous shift workers havesignificantly less sickness absence than day workersof similar age in similar jobs (Taylor, I967b)suggests that the self-disciplined job is more satisfy-ing than the job in which discipline is imposed.The findings in this paper also have important

implications for industrial management. A numberof the men seen admitted that they would like tochange their job but the non-transferable pensionand the 'good money' they receive made themunwilling to move. Older men with dependantscould not afford to move. The increasing develop-ment of automation will inevitably impose changeon the present pattern of jobs in this country.Enforced alteration of jobs should not be made atthe cost of increased absenteeism attributed tosickness.

Most of the material in this paper has been approvedby the University of London for the award of the M.D.degree. Further data may be obtained on request fromthe author.

REFERENCES

Behrend, H. (i95i). Absence under Full Employment. TheUniversity, Birmingham.

Castle, P. F. C. (I956). Accidents, absence and withdrawalfrom the work situation. Hum. Relat., 9, 223-233.

Chiesman, W. E. (I957). Clinical aspects of absenteeism.Roy Soc. Hlth ., 77, 68i-686.

Cooper, R., and Payne, R. (I967). Extraversion and someaspects of work behaviour. Personnel Psychol., 20,45-57-

Cresswell, W. L., and Froggatt, P. (I963). The Causation ofBus Drivers' Accidents. Oxford University Press,London.

II7


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/

ii8 P. J. Taylor

Enterline, P. E. (I964). Sick absence in certain Westerncountries. Industr. Med. Surg., 33, 738-741.

Eysenck, H. J. (I964). Crime and Personality. Routledge andKegan Paul, London.

-, and Eysenck, S. B. G. (I964). Manual of the EysenckPersonality Inventory. University of London Press.

Farmer, E., and Chambers, E. G. (1929). A Study of PersonalQualities in Accident Proneness and Proficiency.Rep. Industr. Hilth. Res. Bd., No. 55, H.M.S.O., London.

Gibson, H. B. (I962). The lie scale of the M.P.I. Actapsychol. (Amst.), 20, I8-23.

Greenwood, M., and Woods, H. M. (I9I9). The Incidenceof Industrial Accidents upon Individuals with specialreference to Multiple Accidents. Rep. Industr. fat.Res. Bd., No. 4. H.M.S.O., London.

Hill, A. Bradford (I965). The environment and disease:association or causation? Proc. roy. Soc. Med., 58,295-300.

Hill, J. M., and Trist, E. L. (I962). Industrial Accidents,Sickness and Other Absences. Tavistock PamphletNo. 4. Tavistock Publications.

Hinkle, L. E., and Plummer, N. (1952). Life stress and in-dustrial absenteeism. Industr. Med. Surg., 2I, 363-375.

-, - , and Whitney, L. H. (I96I). The continuity ofpatterns of illness and the prediction of future health.J. occup. Med., 3, 4I7-423.

Kessel, N., and Shepherd, M. (I965). The health andattitudes of people who seldom consult a doctor.Medical Care, 3, 6-io.

Lokander, S. (I962). Sick absence in a Swedish company.Acta Med. scand., 17I, Suppl. 377.

Morris, J. N. (I965). Capacity and incapacity for work. Somerecent history. Proc. roy Soc. Med., 58, 82I-824.

Smiley, J. A. (I955). A clinical study of a group of accident-prone workers. Brit. J3. industr. Med., I2, 263-278.

Sutherland, J., and Whitwell, G. P. B. (1948). Studies inoccupational morbidity. Ibid., 5, 77-87.

Taylor, P. J. (I967a). Individual variations in sickness absence.Ibid., 24, I69-I76.

(I967b). Shift and day work. A comparison of sicknessabsence, lateness and other absence behaviour at anoil refinery from I962 to I965. Ibid., 24, 93-IOI.

Vernon, P. E. (I964). Personality Assessment. A CriticalSurvey. Methuen, London.


http://oem.bm

j.com/

Br J Ind M


pril 1968. Dow

nloaded from

http://oem.bmj.com/

personal factors associated with sickness absence' · the length ofservice from aninitial...

Documents