Design
The study was designed as a cross-sectional survey.
Selection of employees
Sick-listed employees were included in the study from April 2006 until December 2007 by 43 OHPs, in 5 group practices, derived from two large occupational health services. The exclusion criteria were:
1) Insufficient mastery of the Dutch language.
2) Consultation by telephone.
Selection of OHPs
We chose a mixture of group practices, providing services to large organisations (> 500 employees), medium-sized organisations (75-500 employees) and small organisations (< 75 employees) from different branches (Table 1).
Study size
In this study population we assumed an equal or higher prevalence, of at least 10%, than in the primary care population, with a worst acceptable rate being 7%. With a 95% confidence interval, this implies that the sample should consist of at least 384 employees.
We decided to include at least 40 OHPs, performing 6 sessions, of at least 4 consultations of employees fulfilling the inclusion criteria. This would result in 960 eligible employees. Assuming a maximum non-response of 50% from the employees and a maximum non-response of 15% from the OHPs, at least 408 employees were expected to be included in the study.
Data-collection
Over a period of six weeks the OHPs were asked to select a 4-hour consultation session every week on the same day (i.e. each Monday from 8-12 a.m.). The practice assistants in the administrative section of the occupational health service were instructed to invite all sick-listed employees, who had an appointment for this session to participate in the study. These employees received the research questionnaires one week before the actual consultation, or later if they received the invitation after that time. They were also requested to give informed consent. The OHPs were not involved in the selection of the patients.
The questionnaires were collected on the day of the consultation by the researcher (RH), just before the consultation with the OHP. After the consultation the OHP filled in the questionnaire about the presence of physical symptoms, the diagnosis, the employee's symptom attributions, and the OHP's own opinion about the causes of the symptoms.
Employees who had forgotten to bring their questionnaire were asked to send their questionnaire to the researcher within one week, otherwise they would be considered as non-responders, and no reminders were sent.
Measures
a) Questionnaires for the employee
The employees were asked to answer questions about their socio-demographic variables, MUPS, depression, anxiety, distress, health anxiety and functional impairment.
MUPS
The Patient Health Questionnaire (PHQ) [5, 12, 27] assesses MUPS and symptoms of depression, anxiety, distress, eating disorder and alcohol abuse. The PHQ-15 assesses MUPS and rates the extent to which the patient has been bothered during the past four weeks (score 0-2; not at all bothered to bothered a lot) by 15 common somatic symptoms (e.g. fatigue, dizziness, headache) that rarely have organic explanations.
The total PHQ-15 score range from 0-28 for men and 0-30 for women. For the diagnosis of a somatoform disorder a clinician's assessment is required, but high correlation has been reported between the PHQ-15 score and clinician-rated symptoms of somatoform disorder [28]. Kroenke indicated cut-off scores of 5, 10 and 15 for mild, moderate and severe MUPS. The cut-off point of 15 (PHQ-15 ≥ 15) is comparable with clinically representative samples of MUPS [5, 27]. Patients with a PHQ score < 15 are described as patients with non-severe MUPS, indicating they have moderate, mild or no MUPS. In this study we compare employees with severe MUPS with employees with non-severe MUPS. Kroenke found in a primary care population a prevalence of 9% and in a secondary care population a prevalence of 10% of severe MUPS [5].
The internal consistency of the PHQ-15 is satisfactory (Cronbach's a = 0.80) [12, 28]. The test-retest reliability in a high risk primary care population was moderate with 0.60 [12]. Although limited research has been done, these figures indicate a valid and moderately reliable questionnaire for detection of patients at risk for somatoform disorders [5, 12, 27, 28].
We also used the Four-Dimensional Symptom Questionnaire (4DSQ) to measure MUPS. This Dutch self-report questionnaire [14, 25] assesses the dimensions of distress, MUPS, anxiety and depression. The questionnaire is internally consistent, with Cronbach's alpha coefficients from .79 (anxiety), to .90 (distress) assessed in a working population [14], without sick-listed employees (personal information). The Cronbach's alpha coefficient was 0.80 for MUPS, assessed in a working population. Compared to the diagnosis of General Practitioners (GPs) of somatisation, the Area Under the Curve (AUC) was 0.62 [25]. We used the MUPS subscale of the 4DSQ, in addition to the PHQ-15, to allow comparison of our findings with other studies among employees, since to our knowledge there are no studies which used the PHQ-15 in a working population.
Depression
The PHQ-9 was used to assess symptoms of depression [29, 30]. The rating is comparable with the PHQ-15. Two questions (feeling tired and having trouble sleeping) in the PHQ-9 are also included in the PHQ-15. Although this makes the PHQ-9 score less independent of the PHQ-15 score, high construct validity and strong associations with clinical variables in the general population are found [31]. Compared to the Hospital Anxiety Depression Scale (HADS) the PHQ-9 categories a higher proportion with moderate or severe depression [32].
Total PHQ-9 score ranges from 0-27, with a cut-off point of 15 (PHQ-9 ≥ 15) for severe levels of depression. Algorithms are applied to indicate major depressive disorder or any depressive disorder (excluding the other diagnosis).
Anxiety disorders
The PHQ anxiety subscale contains a module for the assessment of the symptoms of panic disorder and a module for symptoms of other anxiety disorders [27]. Algorithms are applied to diagnose panic disorder and other anxiety disorders. The algorithm for the panic disorder module has been more validated [27, 33] than the algorithm for other anxiety disorders, and has a sensitivity of 75% and a specificity of 96% for diagnosing panic disorder [33].
Distress
We used the Four-Dimensional Symptom Questionnaire (4DSQ) to measure symptoms of distress. The distress subscale is associated with (job-) stressors and indicators of strain. The total score for the 16 distress symptoms range from 0-32, with a cut-off point of 20 for severe levels of distress [14, 25].
Health anxiety
The Whitely index (WI) was used to measure health anxiety. This 14-item self-report questionnaire with yes/no questions was designed to assess health anxiety [10, 34].
Functional impairment
The Dutch translation of the Short Form Health Survey (SF-36) was used [35] to measure levels of functioning, perceived disability and health related quality of life. The SF-36 has high validity [36], and measures eight aspects of health-related quality of life (physical functioning, role functioning physical, bodily pain, general health perceptions, vitality, social functioning, role functioning mental and mental health), with higher scores indicating higher levels of functioning and well-being.
b) Questionnaire for the OHP
The questionnaire for the OHP contained questions about:
-
the presence of physical complaints (yes/no), as reported by the employee
-
the symptom attribution as reported by the employee (somatic, mental, or both causes, physiological or not clear)
-
the OHP's diagnosis, with one classification according to the CAS classification (classification for occupational health and social insurance), derived from the IDH classification (international standard for diagnostic classification).
-
the opinion of the OHP about the causes of the symptoms (somatic disorder, distress, psychiatric disorder, hypochondriasis or somatisation). Example of a question is: 'Do you think that the physical symptoms of the employee are explained by distress?' The OHP was asked whether one or more explanations were present. The questions had a 4-point answering scale: completely, partly, (almost) not and unclear
The OHPs also filled in a questionnaire with regard to their personal socio-demographic status.
c) Registrations
Data on sick report and return to work (RTW) were collected from the computerized registration of the two participating occupational health services.
Data analyses and statistics
A non-response analysis was performed on age, gender, level of educational, ethnicity and duration of sick leave on the day of the consultation.
The cut-off point for the MUPS score (PHQ-15) was set at 15. The data were dichotomized revealing a PHQ ≥ 15 group (the PHQ 15+ group) and a PHQ <15 group (the PHQ 15- group). Chi-square tests were performed for categorical variables, and Fischer's Exact Tests were performed when more than 20% of the expected cell frequencies were less than 5. Chi-square tests for trend for ordinal variables were performed. Independent Students' t tests were performed for continuous and Mann-Whitney U tests for non-parametric distributions.
The SF-36 scores were compared for the PHQ 15+ and PHQ 15- group with respectively no psychiatric morbidity, one psychiatric disorder, mixed psychiatric morbidity (one depressive and one anxiety disorder) and three psychiatric disorders (major or other depressive disorder, panic disorder and other anxiety disorder). We performed a multivariate (MANOVA) and univariate analyses with the 8 SF-36 levels as dependent variables.
Psychiatric morbidity (0, 1, 2 and 3 disorders) and the PHQ score (15+ and 15- group) were the independent variables and we adjusted for gender, age and ethnicity. Also the interaction between psychiatric morbidity and the PHQ score was tested. If influence of the psychiatric morbidity was found we performed posthoc analyses. We also reported R2, which estimates the proportion of explained variance.
To identify the determinants of the PHQ score a logistic regression model was conducted. In this model with the PHQ score as the dependent variable, based on the literature [1–7]. As independent variables were chosen: gender, age, ethnicity, group practice, attribution of the employee, PHQ-9, WI, distress, panic disorder and anxiety disorder.
The Hosmer and Lemeshow Goodness-of-Fit test was applied, which divides subjects into ten equally sized groups, based on predicted probabilities and computes a Chi-square from observed and expected frequencies. If the p-value of the Hosmer and Lemeshow Goodness-of-Fit was .05 or less we rejected the zero hypothesis that there is no difference between the observed and the predicted values of the dependent variable. A high significance level implies a good fit of the model. We also report the Nagelkerke R2, which estimates the proportion of explained variance in a logistic regression model.
All analyses were performed in SPSS for Windows 15.0.
Ethical approval was obtained form the Medical Ethics Committee of the University Medical Center in Groningen, who informed us that ethical clearance was not required because only self-report questionnaires were used and the study reports at group level.