Patient Education and Counseling 94 (2014) 342–350

Assessment

Patient communication competence: Development of a Germanquestionnaire and correlates of competent patient behavior

Erik Farin *, Erika Schmidt, Lukas Gramm

Department of Quality Management and Social Medicine, University Freiburg – Medical Center, Freiburg, Germany

A R T I C L E I N F O

Article history:

Received 17 April 2013

Received in revised form 25 October 2013

Accepted 19 November 2013

Keywords:

Patient–provider communication

Patient communication competence

Questionnaire development

Chronic back pain

Chronic ischemic heart disease

Breast cancer

A B S T R A C T

Objective: The aim of our study was to design and psychometrically test a patient questionnaire to

capture patient communication competence in the context of patient–provider interaction (CoCo

questionnaire). We also aimed to determine patient characteristics associated with competent patient

behavior.

Methods: To assure content validity, we initially conducted 17 focus groups (n = 97) made up of patients

and providers. In the main study n = 1.264 patients with chronic back pain, chronic-ischemic heart

disease or breast cancer who underwent inpatient rehabilitation were surveyed at the end of

rehabilitation.

Results: The CoCo questionnaire contains four scales (patient adherence in communication, critical and

participative communication, communication about personal circumstances, active disease-related

communication) and 28 items addressing competent patient behavior. We provide evidence of

unidimensionality, local independence, reliability, a Rasch-Model fit, the absence of differential item

functioning, and signs of construct validity. The most important correlates of communication

competence are health literacy and communication self-efficacy.

Conclusion: The CoCo questionnaire has good psychometric properties in German. Future research

should examine CoCo’s responsiveness and analyze criterion validity by means of observation data.

Practice implications: The CoCo questionnaire can be recommended for use in evaluating patient

communication training programs.

� 2013 Elsevier Ireland Ltd. All rights reserved.

Contents lists available at ScienceDirect

Patient Education and Counseling

jo ur n al h o mep ag e: w ww .e lsev ier . co m / loc ate /p ated u co u

1. Introduction

Research over the last several years has increasingly focused onpatient–provider communication, revealing that it is a dyadiccommunication situation involving both the treatment provider’scommunication skills as well as the patient’s competencies ([1–6]).Such approaches tend to be theoretically based on communicationtheories that regard the division of power in unequal relationships(such as the patient–provider relationship – i.e., the dyadic powertheory of Dunbar [7]), as well as specific theoretical models such asCegala’s Medical Communication Alignment Theory [2].

A key argument favoring the significance of patient communi-cation competencies is that the idea of active patient–providerinteraction corresponds closely with the aims of patient empow-erment [8] and self-management [9]. The patient that is capable of

* Corresponding author at: University Freiburg – Medical Center Department of

Quality Management and Social Medicine, Engelbergerstr. 21, D-79106 Freiburg,

Germany. Tel.: +49 761 270 74430; fax: +49 761 270 7331.

E-mail address: erik.farin@uniklinik-freiburg.de (E. Farin).

0738-3991/$ – see front matter � 2013 Elsevier Ireland Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.pec.2013.11.005

managing symptoms, treatment, and the consequences of achronic condition should also play an active role in shapingpatient–provider communication. Furthermore, various studieshave identified that patient communication skills exert a positiveinfluence on physicians’ communicative behavior [2,3,10,11].

There are two main methods by which patients’ communica-tion skills can be measured in conjunction with their interactionwith treatment providers: for one, observation methods involvingrecording the interaction with audio tapes or video technology thatare then analyzed for competent and active patient behavior([1,11]), another is via questionnaires in which patients or theirproviders report after an interaction during which communica-tively competent behavior had been exhibited ([12,13]).

The drawback of questionnaires is that they capture theperception of a behavior that may be less objective than theobservation made by an unbiased researcher. On the other hand,they are inexpensive to administer and allow larger cohorts to beexamined for the same expenditure. We find this to be a keyadvantage, as an important application of such an instrument is toevaluate communication skill training, and we can count on effectsthat are not that strong ([14,15]). So, high sample sizes are required

E. Farin et al. / Patient Education and Counseling 94 (2014) 342–350 343

to ensure adequate study power. Bylund et al. ([5], p. 301) came tothe conclusion that ‘‘. . .a second deficit of the literature in this area isthe lack of a valid, reliable patient self-report measure aboutcommunication behavior.’’

Our primary goal was to design and psychometrically test apatient questionnaire measuring communication competence inthe context of patient–provider interaction (the CoCo question-naire). The instrument should thoroughly reflect relevant patientcommunication skills, seen as important by providers andchronically ill patients, it should be economical to administer,and meet stringent methodological standards. To minimize theinfluence of social desirability, the CoCo questionnaire shouldcapture the competent communication behavior that was actuallydemonstrated (‘‘I posed questions . . .’’, behavioral competence),and not be simply a self-assessment of competence (‘‘I am ableto pose questions . . .’’, perceived competence). We assume thatthis operationalization enables us to capture communicationcompetence.

We are unaware of any questionnaire that fulfills the abovementioned demands. Bylund et al. [5,16] reported on developing aPatient Report of Communication Behavior (PRCB), which istheoretically oriented toward the PACE system [17]. However,each competence area is covered by only two items, there are nosubscales, and only data on internal consistency are presented.Roter et al. [12] address self-reported patient communicationbehavior with an 18-item self-report questionnaire designed fortheir study. With the exception of internal consistency, no furtherpsychometric properties are given. The reliability values rangeonly between 0.61 and 0.88. Ashton et al. [18] report on thedevelopment of a patient self-assessment tool to measurecommunication behaviors during doctor visits, but we know ofno study that has presented psychometric properties. Olderinstruments such as the Medical Communication CompetenceScale by Cegala et al. [19] or the Perceived Involvement in CareScale by Lerman et al. [20] were either not subjected to thoroughtesting, or they just capture partial aspects of communicationcompetence. To summarize: the psychometric characteristics ofthe instruments currently available have not been thoroughlyresearched. Moreover, they fail to address aspects such ascommunication about personal circumstances. This last factorappears quite important, as certain patient groups (i.e. the elderly,c.f. [21]) value more personal, face-to-face communication.

A second aim of our study was to determine patientcharacteristics potentially associated with patient communicationcompetence. We tested two hypotheses in this context: (1) that thesignificance of correlates varies somewhat according to thecompetence area under consideration, (2) that, in addition toother basic sociodemographic characteristics, communicationhealth literacy and communication self-efficacy play a key rolein predicting communication competence. The confirmation ofhypothesis 2 can be considered proof of the CoCo questionnaire’sconstruct validity, while the confirmation of hypothesis 1 wouldprovide further proof that it makes sense to develop an instrumentthat differentiates various facets of patient communicationcompetence.

2. Methods

2.1. Instrument development – focus groups and cognitive interviews

In determining the content of the CoCo questionnaire, weoriented ourselves primarily on that which patients and experi-enced providers consider to be useful and important patientcommunication behavior. While most of the research efforts onthis topic have investigated providers’ opinions only [6,22,23])(one exception being the study by Cegala et al. [24]), we wanted to

integrate in parallel the views of both the patients and providers ingenerating items.

To that end, we conducted 17 focus groups (9 with patients, 8with providers) from rehabilitation centers. In the focus groupswere patients diagnosed with chronic back pain (CBP, n = 22),chronic-ischemic heart disease (CIH, n = 18) and breast cancer (BC,n = 9), as well their providers (physicians, nursing staff andtherapists totaling n = 48). The key question posed to allparticipants was ‘How should a patient act when speaking tohis or her physician so that the conversation proceeds in a usefuland beneficial manner for the patient?’. In case those beingquestioned failed to bring up any of the four areas of communica-tion proven to be important to patients in our previous work [21](patient participation and patient orientation, effective and opencommunication, emotionally supportive communication, commu-nication about personal circumstances), we addressed those areasspecifically regarding whether a patient should behave in thatmanner when consulting with their physician. The patientinterviews lasted between 40 and 75 min, the focus groups withproviders between 35 and 45 min.

The group discussions were recorded and transcribed. Thecontents were analyzed by two coders with the aid of Atlas.tisoftware [25]. A coding system was developed in several steps. Asthe feedback from the patients and providers did not differfundamentally, we were able to devise a uniform coding system. Inthe final version, this contained 12 supercategories (i.e., ‘‘providingfactual information’’) and a total of 109 subcategories (e.g.,‘‘provide specific information about symptoms’’). One hundredand forty-four items were generated from the subcategories (CoCoVersion 1), while we made sure that the contents of the mostfrequently coded subcategories were also adequately representedin the items. We devised statements having five responsecategories: strongly disagree, disagree, somewhat disagree,somewhat agree, agree, strongly agree. Two scientists wereinvolved in this process. Version 1 was examined by a thirdscientist for redundancies and comprehensibility, and the item setwas reduced to 81 Items (Version 2). These items were presentedto 10 patients in a cognitive interview [26] in which thinking-aloudand verbal-probing techniques were used. The patients’ remarkswere used to revise and if necessary, omit items. After the cognitiveinterviews, 77 items remained (version 3).

2.2. Sample

To test the CoCo questionnaire psychometrically, n = 1.264patients with CBP, CIH or BC undergoing inpatient rehabilitationwere surveyed at the end of rehabilitation. Only those patientswere enrolled in this study who could comprehend a German-language questionnaire, but not all of them were native speakers.34 Rehabilitation centers participated in the survey. The study wasapproved by the ethics committee of the University of Freiburg(approval number 149/11). The mean percentage of patients thatdid not fill out the questionnaire was 35.8% from all centers. Themost important reason for non-inclusion was refusal to participate(57.2%) followed by cognitive or physical limitations (15.5%) andlanguage difficulties (10.3%). Table 1 provides information on thepatients in the study.

2.3. Instruments

We administered instruments other than the CoCo question-naire to test construct validity (c.f. the hypotheses in Section 2.4.2),and to determine predictors of communication competence (c.f.Section 2.4.3). We employed the KOVA questionnaire [27] (whichcaptures physicians’ communication behavior as perceived bypatients) and a scale to evaluate communication with the

Table 1Respondent characteristics and comparison between diagnoses.

Chronic low back

pain (N = 611)

Chronic ischemic

heart disease

(N = 311)

Breast cancer

(N = 342)

Significance of

differences between

diagnoses

Sociodemographic variables

Age (Mean/SD) 49.2 (10.5) 60.2 (10.7) 58.3 (10.3) p < 0.001

Sex (% female) 52.3 24.3 100.0 p < 0.001

Level of education (highest level completed)

% Elementary school 26.9 36.0 25.9 p < 0.001

% Secondary school 46.1 28.1 39.2

% University-entrance diploma or technical college qualification 22.6 32.9 31.7

Employment

% Employed 81.7 52.5 48.5 p < 0.001

Native German speaker (%) 94.3 94.9 93.0 p = 0.573

Health literacy (Mean/SD)

Comprehension of medical information 2.32 (0.82) 2.26 (0.77) 2.37 (0.86) p = 0.266

Applying medical information 1.91 (0.69) 1.88 (0.65) 1.92 (0.72) p = 0.718

Communicative competence 1.94 (0.74) 1.82 (0.71) 1.98 (0.80) p = 0.015

Medical variables

Duration of disease (%)

<1 year 17.1 58.6 57.8 p < 0.001

1–2 years 14.4 6.8 25.5

2–5 years 21.1 9.6 8.0

5–10 years 16.7 12.0 4.6

>10 years 30.8 12.9 4.0

Health status (Mean/SD)

SF-12 Physical 34.6 (8.5) 39.4 (10.4) 39.6 (8.9) p < 0.001

SF-12 Mental 48.0 (11.6) 48.5 (10.9) 46.3 (10.7) p = 0.042

Psychological variables (Mean/SD)

HADS anxiety 7.14 (3.89) 6.49 (4.01) 7.52 (4.00) p = 0.004

HADS depression 5.60 (3.64) 5.02 (3.57) 5.12 (3.89) p = 0.040

General self-efficacy 2.96 (0.46) 3.04 (0.46) 2.89 (0.52) p < 0.001

Communication self-efficacy 6.92 (1.97) 7.37 (1.74) 7.02 (2.05) p = 0.004

To assess the significance of differences between diagnoses, we applied the ANOVAs or Chi2-tests.

Health literacy: Scale range is 1–5, lower values indicate higher health education literacy (1 = no difficulty, 2 = slight difficulty, 3 = moderate difficulty, 4 = serious difficulty,

5 = extreme difficulty). SF-12: Scale range is 0–100, higher values indicate better health status. HADS: Scale range is 0–21, higher values indicate higher anxiety/depression.

General self-efficacy: Scale range is 1–4, higher values indicate higher self-efficacy. Communication self-efficacy: Scale range is 0–10, higher values indicate higher self-

efficacy.

E. Farin et al. / Patient Education and Counseling 94 (2014) 342–350344

physician (the ECP scale). That scale was newly developed for thestudy; it contains 5 Items (i.e., ‘‘I accomplished what wasimportant to me during the consultation’’) reflecting a summariz-ing assessment of the utility of talks with the physician. That scalereveals good psychometric quality: it is unidimensional accordingto the criteria listed in the section entitled ‘‘Analyses’’ and veryreliable (Cronbachs Alpha = 0.95); it does not exhibit differentialitem functioning.

To assess correlates of communication preferences, we used therecently developed HELP questionnaire to capture health literacy[28] because it possesses such good psychometric quality: Weobtained Cronbach’s Alpha of 0.94, 0.88 und 0.95 in the subscalescomprehension of medical information (6 items), applying medicalinformation (5 items) and communicative competence (7 items);unidimensionality and Rasch model fit were established. The SF-12[29], HADS (Hospital Anxiety and Depression Scale [30]), PEPPI(Perceived Efficacy in Patient–physician Interactions, [31]) and ascale measuring general self-efficacy [32] were also applied in ourstudy.

2.4. Analyses

2.4.1. Psychometric analyses

The sequence of psychometric analyses was based on theprocedure described by Reeve et al. [33] and Rose et al. [34]. For ouranalyses we applied the IBM SPSS Statistics (Version 20), IBM SPSSAMOS (Version 20) and WINSTEPS (Version 3.68) softwareprograms.

(a) Response frequency and ceiling/floor effects

An item was removed if one of these conditions was fulfilled: (a)more than 5% missing values, or (b) ceiling or floor effects (morethan 50% of values in the extreme categories).

(b) Exploratory factor analysis

To determine the number of factors to be extracted, thefollowing criteria were used: (a) scree test, (b) interpretability, and(c) explained variance. An item was removed if it did not loadunambiguously on the extracted factors (factor loading � 0.50 onexactly one factor; factor loading � 0.40 on all other factors).

(c) Unidimensionality and local independence

Every scale was tested separately to check whether it measuresa latent dimension. Single-factor confirmatory factor analyseswere carried out. Model fit was evaluated using the ComparativeFit Index (CFI) [35], the Tucker–Lewis index (TLI) [36], the rootmean square error of approximation (RMSEA) and the standard-ized root mean square residual (SRMR). CFI and TLI values >0.90indicate an acceptable fit, values >0.95 are good. RMSEAvalues < 0.10 suggest moderate fit; values < 0.05 are a good fit.The SRMR value should be under 0.08 [35]. Unidimensionality isassumed whenever at least three of four parameters produce goodvalues. Covariances between error terms were set free only iftheoretically justifiable. The analysis of local independence

E. Farin et al. / Patient Education and Counseling 94 (2014) 342–350 345

examines the residual correlation matrix produced by the single-factor CFA. The absolute values of the residual correlations shouldall fall under 0.20, and the proportion of residual correlationsunder 0.10 should be as small as possible (see [33,37]).

(d) IRT analyses

The 1-parameter IRT model (Rasch model [38]) was used. Infitand outfit mean square statistics (Infit MNSQ, Outfit MNSQ) wereapplied as goodness-of-fit statistics. Poor item fit was defined asinfit or outfit <0.6 or >1.4 ([38], p. 179). Items with a poor infit oroutfit were eliminated.

(e) Reliability

To determine reliability, Cronbach’s Alpha and the personseparation index (PSEP) were calculated. PSEP describes thenumber of performance levels the test measures in a particularsample; it can be converted to a reliability value [39].

(f) Differential item functioning

Differential item functioning (DIF) was tested in reference toage, gender, education and diagnosis (LBP, CIH, BC). We usedordinal logistic regression models to evaluate DIF [40]. Todetermine the size of the DIF, we determined the increase inNagelkerkes R2 after including the DIF variable and DIF variable-sum score interaction. As in Rose et al. [34], a value greater than0.03 was considered a criterion for noticeable DIF.

2.4.2. Construct validity

To evaluate construct validity, we tested the hypotheses below.

1. As communication is based on the participation of at least twoindividuals (in this case, the patient and physician), theperception of a certain patient’s behavior according to the CoCoshould be associated with that of his or her physician. In otherwords, we would expect to observe positive and substantial(r > 0.30) relationships between the KOVA and CoCo ques-tionnaires. The more similar the scales’ contents are, the closerthese relationships should be.

Table 2Psychometric properties of the CoCo scales.

Adherence in

communication

ADH (9 items)

Distribution properties

Mean percentage of missing values (%) 0.8

Scale mean (sum score, 0–100) (standard deviation) 71.5 (7.7)

Unidimensionality

Fit values CFI = 0.951

TLI = 0.935

RMSEA = 0.078

SRMR = 0.039

Number of covariances between error terms that were set free 0

Local independence

Percentage of residual correlations >0.10 (%) 19.4

Percentage of residual correlations >0.20 (%) 0

IRT analyses

Number of items with poor item fit (infit or outfit <0.6 or >1.4) 0

Reliability

Cronbach’s Alpha 0.87

Person Separation Index 0.79

Differential item functioning

Number of items with DIF 0

2. As various studies have demonstrated that active patientcommunication exerts a positive influence (see the introduc-tion), we suppose that all of the CoCo scales correlate positiveand substantial (r > 0.30) with the evaluation of the talks (EPCscale), namely that a patient that obviously acts in acommunicatively competent manner will rate these consulta-tions more positively.

2.4.3. Correlates of communication competence

Hierarchical regression analyses were performed in which first,sociodemographic variables (age, sex, education, German as nativelanguage) were included. In the second block, health literacyvalues were added, and in the third step medical variables (healthstatus assessed by SF-12, duration of disease). Finally, in the fourthstep, psychological variables (anxiety, depression, self-efficacy)were included. The backward method of variable selection wasemployed within the blocks (PIN = .05 POUT = .10). Separateanalyses were conducted for each of the three diagnoses.

2.4.4. Further analyses

We compared the CoCo scale means between the threediagnoses for descriptive purposes by carrying out one-wayanalyses of variance with post hoc tests (Scheffe) and Bonferronicorrections. To determine diagnosis differences on the individual-item level, we performed Kruskal–Wallis tests.

3. Results

3.1. Psychometric analyses

We eliminated two items for having displayed ceiling effects.The subsequent exploratory factor analysis revealed a four-factorsolution that can be interpreted as follows:

Factor 1: Patient adherence in communication (ADH)Factor 2: Critical and participative communication (CRI)Factor 3: Communication about personal circumstances (PER)Factor 4: Active disease-related communication (ACT)

After the exploratory factor analysis, 39 items remained. Toensure sufficient unidimensionality, five items in the PER scale were

Critical and participative

communication

(CRI) (9 items)

Communication about

personal circumstances

(PER) (5 items)

Active disease-related

communication

(ACT) (5 items)

2.4 1.2 1.9

52.3 (14.4) 47.2 (18.9) 42.6 (17.3)

CFI = 0.929 CFI = 0.973 CFI = 0.979

TLI = 0.905 TLI = 0.933 TLI = 0.958

RMSEA = 0.085 RMSEA = 0.120 RMSEA = 0.076

SRMR = 0. 044 SRMR = 0.028 SRMR = 0.027

0 1 0

27.8 60.0 60.0

0 0 0

0 0 0

0.85 0.85 0.80

0.83 0.82 0.78

0 0 0

Table 3CoCo scale means (with standard deviation) and differences between diagnoses.

Adherence in

communication

Critical and participative

communication

Communication about

personal circumstances

Active disease-related

communication

Chronic low back pain (LBP) 72.6 (7.3) 54.0 (14.0) 45.3 (18.9) 41.5 (17.2)

Chronic ischemic heart disease (CID) 70.4 (7.8) 53.0 (14.5) 45.8 (18.2) 44.0 (16.7)

Breast cancer (BC) 70.6 (7.8) 52.0 (14.4) 52.1 (18.9) 43.3 (18.1)

Significant post hoc tests (Scheffe with

Bonferroni correction: p < 0.00417)

LBP vs. CID

LBP vs. BC

– BC vs. LBP

BC vs. CID

–

Table 4CoCo questionnaire items (in order of mean values within the scales).

Item Mean (Standard deviation) Corrected item-scale

correlation

Diagnoses

differences

Adherence in communication (9 items)

I displayed obvious interest in my therapy. 4.42 (0.65) 0.61 LBP > BC > CID

I always paid close attention to what the doctor was saying. 4.39 (0.64) 0.68 LBP > BC > CID

I was very friendly to the doctor. 4.38 (0.61) 0.59 –

I was completely attentive during the consultation. 4.37 (0.61) 0.72 LBP > BC > CID

I listened very carefully to what the doctor was explaining. 4.36 (0.60) 0.67 LBP > BC > CID

I did not interrupt the doctor when he/she was speaking. 4.23 (0.74) 0.53 –

I was entirely open to the doctor’s suggestions. 4.21 (0.70) 0.54 LBP > BC > CID

I kept the nature of our conversation factual. 4.18 (0.71) 0.55 –

I stuck to the subject during our talk. 4.08 (0.73) 0.56 LBP > CID > BC

Critical and participative communication (9 items)

I posed questions to the doctor at the appropriate moment. 3.87 (0.98) 0.50 –

I made a self-confident impression. 3.73 (0.94) 0.43 LBP > CID > BC

Whenever my opinion differed from the doctor’s, I said so clearly. 3.30 (1.31) 0.65 –

I expressed my own opinions of the doctor’s suggestions. 3.20 (1.30) 0.60 –

When I was in doubt, I let my doctor know. 3.15 (1.30) 0.59 –

In case of disagreements with my doctor, I spoke about them clearly. 3.10 (1.49) 0.57 –

I expressed my opinion of my therapy clearly to the doctor. 3.02 (1.34) 0.58 LBP > BC > CID

Sometimes I challenged the treatment. 2.91 (1.45) 0.60 –

I was critical at times when speaking to the doctor. 2.42 (1.46) 0.56 –

Communication about personal circumstances (5 items)

My conversation with the doctor was casual. 3.39 (1.29) 0.57 –

Sometimes the doctor and I laughed together. 3.13 (1.39) 0.60 BC > LBP > CID

I occasionally exchanged private information with the doctor. 2.73 (1.55) 0.76 BC > CID > LBP

I also told the doctor about things that have nothing to do with my illness. 2.51 (1.41) 0.60 BC > CID = LBP

I talked with the doctor about things that have nothing to do with my sickness. 2.42 (1.57) 0.74 BC > CID > LBP

Active disease-related communication (5 items)

I posed questions regarding the treatment’s goals. 3.00 (1.37) 0.62 –

I expressed my fears and apprehensions about my therapy. 2.61 (1.52) 0.57 –

To better understand the doctor’s explanations, I asked many additional questions. 2.49 (1.27) 0.51 –

I posed many questions regarding my treatment in general. 2.34 (1.30) 0.60 –

I questioned my doctor about any side effects from my therapy. 2.28 (1.51) 0.62 CID > BC > LBP

Scale range is 0–5 for all items, higher values indicate higher perceived competence. Response categories are: strongly disagree, disagree, somewhat disagree, somewhat

agree, agree, strongly agree.

The column ‘‘Diagnoses differences’’ shows the sequence of mean values for three diagnoses (chronic low back pain = LBP, chronic ischemic heart disease = CID, breast

cancer = BC), if the Kruskal–Wallis test was significant (with Bonferroni correction, p < .0018).

E. Farin et al. / Patient Education and Counseling 94 (2014) 342–350346

eliminated. No items needed to be eliminated from the other threescales. To ensure the fit to the Rasch model, four items in the PER scaleand one item in the ADH and ACT scales (each) had to be eliminated.The final version of the CoCo questionnaire thus contained 28 items(ADH: 9 items, CRI: 9 items, PER: 5 items, ACT: 5 items). The scale

Table 5Correlation between CoCo scales and other concepts.

Adherence in

communication

ADH

KOVA: patient participation and patient orientation PPO 0.28

KOVA: effective and open communication EOC 0.33

KOVA: emotionally supportive communication ESC 0.38

KOVA: communication about personal circum stances CPC 0.10*

Evaluation of communication with physician 0.39

Pearson correlations, Nmin = 1178, p < .001 for all correlations (exception: * p = 0.001).

KOVA: Perceived communication behavior of physician.

values lie between 0 and 100, with the higher values correspondingto more pronounced behavior. Table 2 summarily illustrates all theCoCo’s psychometric properties. The instrument displays gooddistribution properties, all the scales are unidimensional, and there isonly minor local dependence. It fits the Rasch model, and the

Critical and participative

communication CRI

Communication about

personal circumstances

PER

Active disease-related

communication ACT

0.32 0.42 0.42

0.32 0.43 0.43

0.24 0.50 0.28

0.26 0.73 0.27

0.29 0.45 0.31

Table 6Hierarchical regression analyses to predict communication competence.

Adherence in communication ADH Critical and participative communication

CRI

Communication about personal circum-

stances PER

Active disease-related communication

ACT

LBP CID BC LBP CID BC LBP CID BC LBP CID BC

Block 1

Age �.147

(p< .001)

�.191

(p = .003)

�.142

(p = .001)

Education: elementary school �.121

(p = .007)

Education: university-entrance

diploma

�.130

(p = .034)

.133

(p = .001)

.203

(p = .001)

.139

(p = .012)

�.125

(p = .048)

Native language: German .086

(p = .028)

Block 2

HL: Comprehension of medical

information

.210

(p = .004)

HL: Applying medical information .303

(p< .001)

�.132

(p = .025)

�.173

(p = .019)

�.180

(p = .004)

�.282

(p = .001)

HL: Communicative competence .263

(p< .001)

.233

(p = .001)

.344

(p = .001)

.595

(p< .001)

.174

(p = .012)

.210

(p = .003)

.398

(p< .001)

Block 3

SF-12 physical scale �.147

(p = .023)

.153

(p< .001)

�.085

(p = .041)

SF-12 mental scale .183

(p = .012)

�.151

(p = .020)

Duration of disease <1year �.193

(p = .004)

Duration of disease 1–5 years �.159

(p = .006)

�.099

(p = .016)

�.159

(p = .010)

Duration of disease>5 years �.161

(p = .005)

.165

(p = .009)

�.108

(p = .011)

Block 4

HADS Anxiety .239

(p< .001)

.132

(p = .003)

.176

(p = .008)

HADS depression �.159

(p = .008)

General self-efficay .154

(p = .022)

.173

(p = .014)

Communication self-efficacy .186

(p< .001)

.148

(p = .030)

.218

(p< .001)

.125

(p = .047)

.147

(p = .034)

.290

(p< .001)

R2 .328 .331 .372 .498 .425 .537 .377 .214 .304 .378 .326 .307

Standardized regression coefficients, only significant regression coefficients (p< .05) are shown.

LBP = chronic low back pain, CID = chronic ischemic heart disease, BC = breast cancer.

E.

Farin

et a

l. /

Pa

tient

Ed

uca

tion

an

d C

ou

nselin

g 9

4 (2

01

4)

34

2–

35

0

34

7

E. Farin et al. / Patient Education and Counseling 94 (2014) 342–350348

reliability values are good (with Cronbach’s Alpha ranging from 0.80and 0.87). None of the 28 items reveals differential item functioning.

3.2. Descriptive results

Tables 3 and 4 show descriptive results on the scale and itemlevel. The patients show the behavior as captured in thecommunication adherence scale most clearly, followed by criticaland participative communication, communication about personalcircumstances and finally active disease-related communication.The differences among the various diagnoses were small (Table 3).The correlations among the four CoCo scales were all positive andranged from 0.17 (ADH-ACT) to 0.53 (CRI-ACT), with a 0.31 median.

3.3. Construct validity

As Table 5 illustrates, we confirmed hypothesis 1 for the mostpart: 10 of the 16 intercorrelations between the CoCo and KOVAscales lie above 0.30, the remaining (with one exception) between0.24 and 0.30. Where the scales’ contents are similar (CoCo-PERand KOVA-CPC, CoCo-PER und KOVA-ESC, CoCo-ACT und KOVA-EOC), the correlations are over 0.40. An exception to this is found inthe association between the CoCo-CRI and KOVA-PPO. As boththose scales capture the aspect of patient participation, we hadexpected a correlation even higher than 0.32. Hypothesis 2 wasconfirmed: the correlations between the CoCo and the assessmentof the usefulness of talks with the physicians vary according to thescale between 0.29 and 0.45.

3.4. Correlates of communication competencies

Depending on the CoCo scale and diagnosis, the proportion ofexplained variance ranges between 21.4% and 53.7%, while mostlies between 30% and 40% (see Table 6).

Regarding the hypothesis that the significance of correlatesdepends in part on the area of competence (see Section 1), there issome evidence:

� Education plays a different role for critical/participatorycommunication and for other areas of competence: highereducation is associated with a higher CRI scale value in all thediagnoses we examined. Concerning the other areas of compe-tence, we observed no correlation, or, as in the case of the BCpatients, a negative association.� Anxiety and depression have an influence in some patients on

those CoCo scales that capture emotional aspects (the PER andACT scales, which refer to communicating fears and anxiety), butthey exert no effect on the CoCo scales that measure factualcommunication aspects (ADH and CRI).

The most important correlates of communication competenceare (which also confirms our second hypothesis as described in theintroduction) health literacy (as related to communicative skills)and communication self-efficacy. General self-efficacy is lessimportant than communication self-efficacy. The significances ofhealth status and duration of disease are heterogeneous and varydepending on the competence area and diagnosis.

4. Discussion and conclusion

4.1. Discussion

4.1.1. Psychometric properties, conception, and the CoCo

questionnaire’s applicability

The CoCo questionnaire’s psychometric properties are good. Weprovide evidence of unidimensionality, local independence,

reliability, Rasch-model fit and the absence of differential itemfunctioning. We also present indications of construct validity. Thelow intercorrelations among the CoCo subscales seem to makeevident that we captured distinct facets of communicationcompetence.

Content validity has been guaranteed by the patient skillsidentified by the patients and providers in the focus groups. In thefollowing, we would like to examine more closely the extent towhich the contents of the CoCo’s final version display thecomponents of existing models for active patient communication.Street [41] and Cegala et al. [3] distinguish four components (as inthe PACE model): information seeking and verifying, informationprovision, assertive utterances, and expressions of affect. Informa-tion seeking and verifying are incorporated in the CoCo in the ACTand ADH scales, while the ACT captures the concrete aspect ofposing questions and the ADH the communication style requiredwhen seeking information (being attentive and listening). Infor-mation provision is covered in the CoCo primarily by the ADH scale,which contains the communication style necessary in that regard(sticking to the facts and keeping to the subject at hand).Concerning criticism and doubt, the CRI scale also capturesinformation provision. Assertive utterances are also measured inthe CRI. Personal expressions of affect are captured in the PER scale,disease-related affect in the ACT, and relationship-dependentaffect in the CRI.

Bylund et al. [5] emphasize as a further component theexpression of preferences which the CoCo captures in the CRI scale.The AGENDA model [22] contains components (in addition tothose already mentioned above), namely agenda and goal setting.Agenda setting is not captured in CoCo, as that behavior is foundmore in the run-up to, and not as a key part of patient–providerinteraction; goal setting (as an expression of one’s own opinion andexpectations) is captured in the CRI scale.

Had the final version of our questionnaire covered all theessential contents identified by the focus groups in spite of theitem selection, we would have been able to present an additionalindication of content validity. The comparison of our original 81items with the 28 remaining items demonstrates that this does, infact, usually apply. But two aspects were covered only by theomitted items: positive patient feedback to the physician andconcrete techniques to ensure the efficacy of consultations (i.e.taking notes during the discussion, and taking systematic steps toprepare for speaking with the physician).

In summary, we maintain that the CoCo incorporates all theessential patient communication behaviors that are specified inthe models currently in use and which are relevant to directinteraction between patients and their providers. Item selectionled to an only slight reduction in content validity. By capturingvarious competence components, the CoCo is well-suited for bothevaluating specific interventions (i.e., questions prompts [42]), aswell as for assessing patient communication training programsthat have a broader focus and in which it may not necessarily beobvious upon which patient behaviors the interventions have hadan effect.

We consider patient communication competence to be animportant element of health literacy (see ‘‘interactive healthliteracy’’, [28,43,44]). Patient communication skills correlateclosely with other forms of health literacy [28], thus skill trainingprograms that reflect this factor (that is, in addition tocommunicative skills, that also address the comprehension ofmedical information) may be more efficacious. An instrument thatassesses communication behavior as communication competenceaccounts for this link. The difference between communicativehealth literacy (as captured here with the HELP instrument) andCoCo is that the first captures general cognitive and social skillsthat a person possesses regardless of the actual encounter, while

E. Farin et al. / Patient Education and Counseling 94 (2014) 342–350 349

CoCo measures the communication competencies that a chroni-cally ill patient demonstrates in a particular interactive situation.

Can a questionnaire that asks a patient to describe his or herperception of their behavior really capture patient competence?This would be challenged (a) if the behavior sample considered bythe patient when completing the CoCo questionnaire is too small or(b) when the patient is capable of showing a behavior but fails to doso during the interval of behavior currently under examination, i.e.,because he or she finds it unnecessary or inappropriate to the givensituation. CoCo fulfills the competence-measuring requirementprovided the patient is instructed to report on his behavior over along period and when one can assume that all the patient’scommunication skills were in fact required. This was the case inour study, since we asked our patients to consider the entire three-week duration of their rehabilitation when filling out thequestionnaire. If a patient fails to display certain behavior despitea long time of reference, we interpret the CoCo questionnaire’sresults conservatively, that is, by assuming that that particularcompetence is absent. This makes good sense, as mistakenlyassuming that a skill is absent which the patient in fact possesses isa less serious error in practice than the reverse.

We could have included an additional ‘‘not relevant’’ answeringoption to distinguish a patient who is incapable of a certainbehavior from one who finds such a behavior inappropriate. Wedecided against adding such an answer in the CoCo questionnairebecause we maintain that the advantage of doing so fails tooutweigh a key disadvantage, namely that such a neutral optionwould have made it more likely that indecisive or poorly motivatedresponders would evade the question and use the ‘‘not relevant’’option.

4.1.2. Level and correlates of patient communication competencies

Numerous studies have shown that there is room forimprovement in patients’ communication skills [45]. Our datareinforce that finding, especially in terms of active disease-relatedcommunication (Table 3). Every sixth patient admits to not havingdemonstrated a single one of the behaviors captured in the ACTscale, and only 37% reveal a scale mean, which proves that thebehavior that is captured here was clearly shown (mean value inthe upper half of the scale’s range; data not presented in the resultssection). This is remarkable as the posing of questions captured inthe ACT scale is among those patient behaviors whose efficiencyhas been most definitively proven [46].

As in the literature, in our results regarding correlates ofcommunication competence, we also found ([1,47]) that partici-patory communication behavior is more pronounced amongyounger patients. However, this was not the case with our BCpatients. The fact that patient gender has no influence is also in linewith other study findings [48,49]. The convincing associationsbetween communication self-efficacy and CoCo provide furtherevidence of the frequently proven relevance of self-efficacy as apredictor of disease management [50,51]. The function ofcommunicative HL indicates the aforementioned propinquitybetween these concepts. As there are many other relevantpredictors in addition to communicative HL, and the bivariatecorrelations between communicative HL and communicationcompetence only range between 0.06 and 0.38 (data not shown),it is apparent that clearly different constructs are being capturedhere. HL in terms of applying medical information is occasionallyassociated with lower values in the CoCo, which may be due tosome patients sensing less of a need for consultation due to havinglearned from previous experience.

4.1.3. Strengths and limitations

The strengths of our study are that content validity has beenguaranteed in qualitative pre-tests, that we undertook thorough

psychometric testing, and that we had access to a large patientcohort involving three chronic illnesses. Beyond the problemsalready mentioned in Section 4.1.1, another limitation is our highpercentage of non-responders. Another limitation is that weretrieved all our data from one particular healthcare setting(rehabilitation in Germany), meaning that further testing is neededto ensure our findings’ generalizability. Our results are alsosomewhat weakened by the fact that we considered only patientcharacteristics and no context factors in seeking the correlates ofcommunication competencies (see [49]).

4.2. Conclusion

With its four scales and 28 items, the CoCo questionnairethoroughly captures patient communication competence. As itspsychometric properties are good, it can be recommended for use.Future research efforts will address the CoCo‘s responsiveness inparticular and seek to prove its criterion validity by means ofobservation data.

4.3. Practice implications

Our study demonstrates that some patients rate their commu-nication competence in patient–physician interaction as not veryhigh. Patient-communication training programs to improvepatients’ skills would be worthwhile. The CoCo questionnairecan facilitate the evaluation of such training programs.

Conflict of interest

None declared.

Acknowledgments

The study was conducted in the project ‘‘Communicationcompetencies of the chronically ill with regard to the interactionwith treatment providers’’ which is funded in Germany by theFederal Ministry of Education and Research (Grant No.: 01GX1042)as part of the funding priority for ‘‘Chronic Illnesses and PatientOrientation’’ (http://www.forschung-patientenorientierung.de/).We wish to thank the cooperating rehabilitation centers for theirsupport in data collection: Ambulante Reha-Klinik Passau SANARISin Passau, Ambulantes Rehazentrum Wohrderwiese in Nurnberg,Buckeberg-Klinik in Bad Eilsen, Fachklinikum Brandis in Brandis,Gesundheitszentrum Chiemgau in Traunstein, HabichtswaldKlinik in Kassel-Wilhelmshohe, Herz- und Gefaßzentrum BadBevensen in Bad Bevensen, Herzklinik Seebruck in Seebruck, KlausMiehlke Klinik in Wiesbaden, Klinik Graal-Muritz im Ostseeheil-bad Graal-Muritz, Klinik St. Marien in Bad Soden-Salmunster,Kurklinik Maximilianbad in Bad Waldsee, MEDIAN Adelsberg-Klinik in Bad Berka, MEDIAN Klinik Wismar in Wismar, MEDIANWeserklinik in Bad Oeynhausen, MEDICA Klinik fur ambulanteRehabilitation in Leipzig, Mediclin Fachklinik Rhein-Ruhr in Essen,medicos.Osnabruck in Osnabruck, Paracelsus Klinik am See in BadGandersheim, REGIO-Reha in Freiburg, Rehabilitations- undTherapiezentrum in Bautzen, Rehabilitationsklinik Lautergundder DRV Berlin-Brandenburg in Bad Staffelstein, Rehabilitationsk-linik Markische Schweiz in Buckow, Reha-Klinik Dahlener Heide inDahlen-Schmannewitz, REHA-Klinik Lehmrade in Lehmrade,REHA-Tagesklinik im Forum Pankow in Berlin, Reha-ZentrumUckeritz in Uckeritz, Rehazentrum Bayerisch Gmain in BayerischGmain, Rehazentrum Oberharz in Clausthal-Zellerfeld, Rheingau-Taunus-Klinik in Bad Schwalbach, Rheintal-Klinik in Bad Krozin-gen, St. Georg Vorsorge- und Rehabilitationskliniken inHochenschwand, Strandklinik Ostseebad Boltenhagen in Bolten-hagen, Vinzenz Klinik in Bad Ditzenbach, Zentrum fur ambulante

E. Farin et al. / Patient Education and Counseling 94 (2014) 342–350350

Rehabilitation in Rostock, Zentrum fur ambulante Rehabilitation inStuttgart.

We also thank Mr. Matthias Gustke for his support during dataacquisition.

References

[1] Street J, Slee C, Kalauokalani DK, Dean DE, Tancredi DJ, Kravitz RL. Improvingphysician–patient communication about cancer pain with a tailored educa-tion-coaching intervention. Patient Educ Couns 2010;80:42–7.

[2] Cegala DJ, Post DM. The impact of patients’ participation on physicians’patient-centered communication. Patient Educ Couns 2009;77:202–8.

[3] Cegala DJ, Street RL, Clinch CR. The impact of patient participation on physi-cians’ information provision during a primary care medical interview. HealthCommun 2007;21:177–85.

[4] Middleton JF, McKinley RK, Gillies CL. Effect of patient completed agenda formsand doctors’ education about the agenda on the outcome of consultations:randomised controlled trial. Brit Med J 2006;332:1238–42.

[5] Bylund CL, D’Agostino TA, Ho EY, Chewning BA. Improving clinical communi-cation and promoting health through concordance-based patient education.Commun Educ 2010;59:294–311.

[6] Rao J, Anderson L, Sukumar B, Beauchesne D, Stein T, Frankel R. Engagingcommunication experts in a Delphi process to identify patient behaviors thatcould enhance communication in medical encounters. BMC Health Serv Res2010;10:97.

[7] Dunbar NE, Burgoon JK. Perceptions of power and interactional dominance ininterpersonal relationships. J Soc Pers Relat 2005;22:207–33.

[8] Tengland PA. Empowerment: a conceptual discussion. Health Care Anal2008;16:77–96.

[9] Newman S, Steed L, Mulligan K. Self-management interventions for chronicillness. Lancet 2004;364:1523–37.

[10] Street RL, Krupat E, Bell RA, Kravitz RL, Haidet P. Beliefs about control in thephysician–patient relationship: effect on communication in medical encoun-ters. J Gen Intern Med 2003;18:609–16.

[11] Cegala DJ, Chisolm DJ, Nwomeh BC. Further examination of the impact ofpatient participation on physicians’ communication style. Patient Educ Couns2012;89:25–30.

[12] Roter DL, Wexler R, Naragon P, et al. The impact of patient and physiciancomputer mediated communication skill training on reported communicationand patient satisfaction. Patient Educ Couns 2012;88:406–13.

[13] Butow P, Devine R, Boyer M, Pendlebury S, Jackson M, Tattersall MHN. Cancerconsultation preparation package: changing patients but not physicians is notenough. J Clin Oncol 2004;22:4401–9.

[14] Cegala DJ. Patient communication skills training: a review with implicationsfor cancer patients. Patient Educ Couns 2003;50:91–4.

[15] Harrington J, Noble LM, Newman SP. Improving patients’ communication withdoctors: a systematic review of intervention studies. Patient Educ Couns2004;52:7–16.

[16] Bylund CL, Goytia EJ, D’Agostino TA, Bulone L, Horner J, Li Y, et al. Evaluation ofa pilot communication skills training intervention for minority cancerpatients. J Psychosoc Oncol 2011;29:347–58.

[17] Cegala DJ, Post DM. On addressing racial and ethnic health disparities: thepotential role of patient communication skills interventions. Am Behav Sci2006;49:853–67.

[18] Ashton CM, Holt CL, Wray NP. A patient self-assessment tool to measurecommunication behaviors during doctor visits about hypertension. PatientEduc Couns 2010;81:275–314.

[19] Cegala DJ, Coleman MT, Turner JW. The development and partial assessment ofthe medical communication competence scale. Health Commun1998;10:261–88.

[20] Lerman CE, Brody DS, Caputo GC, Smith DG, Lazaro CG, Wolfson HG. Patients’perceived involvement in care scale: relationship to attitudes about illness andmedical care. J Gen Intern Med 1990;5:29–33.

[21] Farin E, Gramm L, Schmidt E. Predictors of communication preferences inpatients with chronic low back pain. Patient Prefer Adherence 2013;7:1117–27.

[22] Arnold CL, Coran JJ, Hagen MG. Revisiting patient communication training: anupdated needs assessment and the AGENDA model. Patient Educ Couns2012;88:399–405.

[23] Talen MR, Grampp K, Tucker A, Schultz J. What physicians want from theirpatients: identifying what makes good patient communication. Fam SystHealth 2008;26:58–66.

[24] Cegala DJ, Gade C, Broz SL, McClure L. Physicians’ and patients’ perceptions ofpatients’ communication competence in a primary care medical interview.Health Commun 2004;16:289–304.

[25] ATLAS.ti (Version 5.2.0). Berlin: Scientific Software Development GmbH; 2007.[26] Collins D. Pretesting survey instruments: an overview of cognitive methods.

Qual Life Res 2003;12:229–38.[27] Farin E, Gramm L, Schmidt E. Taking into account patients’ communication

preferences: Instrument development and results in chronic back painpatients. Patient Educ Couns 2012;86:41–8.

[28] Farin E, Ullrich A, Nagl M. Health education literacy in patients with chronicmusculoskeletal diseases: development of a new questionnaire and socio-demographic predictors. Health Educ Res 2013;28:1080–91.

[29] Ware JE, Kosinski M, Keller SD. A 12-item short-form health survey: construc-tion of scales and preliminary tests of reliability and validity. Med Care1996;34:220–33.

[30] Zigmond AS, Snaith RP. The hospital anxiety and depression scale. ActaPsychiatr Scand 1983;67:361–70.

[31] Maly RC, Frank JC, Marhsall GN, DiMatteo MR, Reuben DB. Perceived efficacy inpatient–physician interactions (PEPPI): validation of an instrument in olderpersons. J Am Geriatr Soc 1998;46:889–94.

[32] Schwarzer R, Jerusalem M. Generalized self-efficacy scale. In: Weinman J,Wright S, Johnston M, editors. Measures in health psychology: A user’sportfolio. Causal and control beliefs. Windsor, UK: NFER-NELSON; 1995. p.35–7.

[33] Reeve BB, Hays RD, Bjorner JB, Cook KF, Crane PK, Teresi JA, et al. Psychometricevaluation and calibration of health-related quality of life banks Plans for thePatient-Reported Outcome Measurement Information System (PROMIS). MedCare 2007;45:S22–31.

[34] Rose M, Bjorner JB, Becker J, Fries JF, Ware JE. Evaluation of a preliminaryphysical function item bank supported the expected advantages of the Pa-tient-Reported Outcomes Measurement Information System (PROMIS). J ClinEpidemiol 2008;61:17–33.

[35] Hu LT, Bentler P. Cutoff criteria for fit indices in covariance structure analysis:conventional criteria versus new alternatives. Struct Equation Model1999;6:1–55.

[36] Tucker L, Lewis C. A reliability coefficient for maximum likelihood factoranalysis. Psychometrika 1973;38:1–10.

[37] Cook K, Teal C, Bjorner J, Cella D, Chang CH, Crane PK, et al. IRT health outcomesdata analysis project: an overview and summary. Qual Life Res 2007;16:121–32.

[38] Bond TG, Fox CM. Applying the Rasch model: fundamental measurement inthe human sciences. Mahwah, NJ: Lawrence Erlbaum Associates; 2001.

[39] Prieto L, Alonso J, Lamarca R. Classical test theory versus Rasch analysis forquality of life questionnaire reduction. Health Qual Life Outcomes2003;1:27.

[40] Teresi JA, Fleishman JA. Differential item functioning and health assessment.Qual Life Res 2007;16:33–42.

[41] Street RL. Active patients as powerful communicators: the linguisticfoundation of participation in health care. In: Robinson WPG, editor. Hand-book of language and social psychology. Chicester: John Wiley; 2001 . p. 541–60.

[42] Dimoska A, Butow PN, Lynch J, Hovey E, Agar M, Beale P, et al. Implementingpatient question-prompt lists into routine cancer care. Patient Educ Couns2012;86:252–8.

[43] Rubin DL, Parmer J, Freimuth V, Kaley T, Okundaye M. Associations betweenolder adults’ spoken interactive health literacy and selected health care andhealth communication outcomes. J Health Commun 2011;16:191–204.

[44] Wynia MK, Osborn CY. Health literacy and communication quality in healthcare organizations. J Health Commun 2010;15:102–15.

[45] Cegala DJ, McClure L, Marinelli TM, Post DM. The effects of communicationskills training on patients’ participation during medical interviews. PatientEduc Couns 2000;41:209–22.

[46] Shepherd HL, Barratt A, Trevena LJ, et al. Three questions that patients can askto improve the quality of information physicians give about treatmentoptions: A cross-over trial. Patient Educ Couns 2011;84:379–85.

[47] Levinson W, Kao A, Kuby A, Thisted RA. Not all patients want to participate indecision making. A national study of public preferences. J Gen Intern Med2005;20:531–5.

[48] Street RLJ, Gordon HSM, Ward MMM, Krupat EP, Kravitz RLM. Patient partici-pation in medical consultations: why some patients are more involved thanothers. Med Care 2005;43:960–9.

[49] Cegala DJ. An exploration of factors promoting patient participation in primarycare medical interviews. Health Commun 2011;26:427–36.

[50] Clark NM, Dodge JA. Exploring self-efficacy as a predictor of disease manage-ment. Health Educ Behav 1999;26:72–89.

[51] ten Klooster PM, Oostveen JCM, Zandbelt LC, Taal E, Drossaert CHC, HarmsenEJ, et al. Further validation of the 5-item Perceived Efficacy in Patient–Physi-cian Interactions (PEPPI-5) scale in patients with osteoarthritis. Patient EducCouns 2012;87:125–30.