Productivity at Work and QoL in Patients With RA
Productivity at Work and QoL in Patients With RA
This is a cross-sectional study of the baseline measurements of a randomised controlled trial (RCT) evaluating an intervention program to support at-work productivity for workers with RA. The RCT design is described in detail elsewhere. Patients with RA were recruited from a specialized rheumatology clinic (Reade, formerly the Jan van Breemen Institute), regional hospitals (outposts of Reade), and an academic hospital (the VU University Medical Center, department of rheumatology), Amsterdam, the Netherlands.
Inclusion criteria for this study were: (1) a diagnosis of RA, (2) ages 18–64 years, (3) having a paid job (either paid- employment or self-employed), (4) working at least eight hours per week, (5) and experiencing at least minor difficulties in functioning at work. We asked patients to indicate on a scale of one to five the extent that RA interfered with their work functioning. Patients who indicated one on this scale, had a severe comorbidity, were unable to read or understand Dutch, or had had more than three months of sick leave at the time of inclusion were excluded. The study was approved by the medical ethics committees of the participating centres (Slotervaart Hospital, Reade, and the VU University Medical Center), and all patients gave written informed consent.
Patients completed a questionnaire and their clinical characteristics were obtained from their medical records. All measured variables were categorized into either internal (personal or disease-related variables) or external factors related to the external environment.
Outcome Measures. At-work productivity loss was measured by hours lost from work due to presenteeism. Presenteeism is defined as being present at the job, but not being able to function optimally. Presenteeism was measured using the 25-item Work Limitations Questionnaire (WLQ). The WLQ includes four dimensions; physical demands, time management, mental-interpersonal demands, and output demands. Based on these four dimensions, a score was calculated to represent the percentage of productivity loss. This score was multiplied by the number of work hours in two weeks, leading to an estimation of the number of hours that a patient had not been fully productive at work over the last two weeks. The internal consistency is high for the WLQ, with a Cronbach's alpha of 0.88. Several studies showed that the WLQ is a reliable and valid questionnaire for assessing productivity loss in workers with RA.
Quality of life was measured using the RAND 36. All nine subscales of the RAND 36 were included in the questionnaire, and used for further analyses including mental health (Cronbach's α 0.85), pain (α 0.88), physical role limitations (α 0.90), physical functioning (α 0.92), social functioning (α 0.71), vitality (α 0.82), emotional role limitations (α 0.86), general health perception (α 0.81), and perceived health change (α not applicable). The subscales of the RAND 36 are transformed into a scale score ranging from 0–100. A higher score indicates better health. The subscales of the RAND 36 are included as internal factors in the analysis.
Internal Factors. The Disease Activity Score of 28 joints (DAS28) was assessed as part of usual care and was collected from patient records. The DAS28 score was based on the number of tender and swollen joints out of 28 joints, the erythrocyte sedimentation rate (ESR) and the patient's general health measured on a visual analogue scale (VAS) of 100 mm. We retrieved the use of disease-modifying antirheumatic drugs (DMARD) and the use of biological therapeutics. Serological measures (anti-cyclic citrullinated peptide antibodies [aCCP] and IgM rheumatoid factor [IgM-RF]) were recorded as positive or negative.
Daily functioning was measured with the Health Assessment Questionnaire (HAQ), a reliable and valid questionnaire widely used in RA research. The HAQ score ranges from 0–3; a higher score indicates increased disability.
Pain and fatigue were measured with single items using VAS scales. VAS scales range from 0–10, with 0 meaning no pain/fatigue at all, and 10 meaning severe pain/very tired.
The presence of comorbidity (yes/no) was investigated using a list with 15 common comorbidities including cardiovascular diseases, diabetes mellitus, and psychological disorders. Disease duration was determined by one open-ended question about the year of the patient's RA diagnosis, and the duration of complaints due to RA (answer categories were 0–2 years, 3–5 years, 6–10 years, >10 years). We included one question about the highest attained educational level. Low education was operationalized as primary school, middle education or basic vocational education. Middle education was operationalized as secondary vocational education or intermediate vocational education. High education was operationalized as higher vocational education or a university degree. The patients' gender and age were collected from patient medical records.
External Factors. We measured supervisor and co-worker support, psychological and physical job demands with subscales from the Job Content Questionnaire (JCQ). Items of the JCQ such as 'My colleagues are friendly' are answered on a scale of 1 (totally disagree) to 4 (completely agree). The score of each subscale ranges from 1–4 with a higher score indicating more supervisor support, co-worker support, etc.
Work instability was measured with the RA Work Instability Scale (RA WIS). The RA WIS contains 23 items (yes/no), which are summed for a total score ranging between 0–23. A score of 0–9 indicates low, 10–17 moderate, and 18–23 high work instability.
Sick leave was measured with a single item of the Productivity and Disease Questionnaire (Prodisq). Patients were asked to count the number of days they had been absent from work during the last 12 months.
Three single item questions about being a supervisor (yes/no), shift work (yes/no), and type of job contract (permanent contract/self-employed) were included in the questionnaire worded as The Netherlands Working Conditions Survey.
For the first aim, we determined which combination of factors was best associated with the outcome of at-work productivity loss using linear regression models. These models were built on a dataset of 100 randomly selected participants. The remaining 50 patients were used to cross-validate our model.
In step 1, we performed univariate linear regression analyses for each independent variable with at-work productivity loss. We did not include the number of working hours as an independent variable, since the number of work hours is already incorporated in the WLQ score for at-work productivity loss. Independent variables with a p-value <0.15 were selected for further analyses. In step 2, we built a multivariate model for the nine subscales of the RAND 36, using a backward stepwise procedure, because the subscales are highly correlated. Similarly we built a backward multivariate model for the DAS 28 score and its four components that were highly correlated as well. We retained the components with a p-value <0.15 for further analyses.
In step 3, we built two multivariate models; one multivariate model for the selected internal variables and one model for the selected external variables. All independent variables selected in step 1 and 2 were entered in the models, using a backward stepwise procedure. Variables with a p-value <0.15 were retained in the multivariate models, resulting in an internal and external model. In step 4, one final model was built in which all variables retained in the internal and external model in step 3 were entered again using a backward procedure. Variables with a p-value <0.05 were retained in the final model. R was used to calculate the proportion of explained variance.
The final model was cross-validated in the remaining sample of n = 50 by calculating the R of the final model applied to the sample of n = 50. R was calculated by determining the correlation between the actual outcome measured with the questionnaire, and the outcome calculated based on the regression coefficients in the final multivariate model of step 4 (i.e. outcome calculated = constant + B1 * var1, etc.). The R of both samples (n = 100 and n = 50) were compared.
For the second aim of this study, univariate linear regression analyses were performed to gain insight into the crude associations between at-work productivity loss and quality of life. The independent variables were each subscale of the RAND 36 retained in the backward procedure from step 2, and the dependent variable was at-work productivity loss. Second, we investigated potential confounders for each univariate model. All variables with a p-value <0.15 with at-work productivity loss were considered as potential confounders. Confounders were selected based on a stepwise forward strategy. Each potential confounder was entered separately into the model. The strongest confounder based on the change in the regression coefficient was retained in the model, and subsequently, all remaining potential confounders were entered one by one. The strongest confounder was retained from each round, until the regression coefficient did not change significantly by adding the next confounder. A change in the regression coefficient of more than 10% was considered significant.
All analyses were performed in SPSS version 20.
Methods
Design
This is a cross-sectional study of the baseline measurements of a randomised controlled trial (RCT) evaluating an intervention program to support at-work productivity for workers with RA. The RCT design is described in detail elsewhere. Patients with RA were recruited from a specialized rheumatology clinic (Reade, formerly the Jan van Breemen Institute), regional hospitals (outposts of Reade), and an academic hospital (the VU University Medical Center, department of rheumatology), Amsterdam, the Netherlands.
Inclusion criteria for this study were: (1) a diagnosis of RA, (2) ages 18–64 years, (3) having a paid job (either paid- employment or self-employed), (4) working at least eight hours per week, (5) and experiencing at least minor difficulties in functioning at work. We asked patients to indicate on a scale of one to five the extent that RA interfered with their work functioning. Patients who indicated one on this scale, had a severe comorbidity, were unable to read or understand Dutch, or had had more than three months of sick leave at the time of inclusion were excluded. The study was approved by the medical ethics committees of the participating centres (Slotervaart Hospital, Reade, and the VU University Medical Center), and all patients gave written informed consent.
Measurements
Patients completed a questionnaire and their clinical characteristics were obtained from their medical records. All measured variables were categorized into either internal (personal or disease-related variables) or external factors related to the external environment.
Outcome Measures. At-work productivity loss was measured by hours lost from work due to presenteeism. Presenteeism is defined as being present at the job, but not being able to function optimally. Presenteeism was measured using the 25-item Work Limitations Questionnaire (WLQ). The WLQ includes four dimensions; physical demands, time management, mental-interpersonal demands, and output demands. Based on these four dimensions, a score was calculated to represent the percentage of productivity loss. This score was multiplied by the number of work hours in two weeks, leading to an estimation of the number of hours that a patient had not been fully productive at work over the last two weeks. The internal consistency is high for the WLQ, with a Cronbach's alpha of 0.88. Several studies showed that the WLQ is a reliable and valid questionnaire for assessing productivity loss in workers with RA.
Quality of life was measured using the RAND 36. All nine subscales of the RAND 36 were included in the questionnaire, and used for further analyses including mental health (Cronbach's α 0.85), pain (α 0.88), physical role limitations (α 0.90), physical functioning (α 0.92), social functioning (α 0.71), vitality (α 0.82), emotional role limitations (α 0.86), general health perception (α 0.81), and perceived health change (α not applicable). The subscales of the RAND 36 are transformed into a scale score ranging from 0–100. A higher score indicates better health. The subscales of the RAND 36 are included as internal factors in the analysis.
Internal Factors. The Disease Activity Score of 28 joints (DAS28) was assessed as part of usual care and was collected from patient records. The DAS28 score was based on the number of tender and swollen joints out of 28 joints, the erythrocyte sedimentation rate (ESR) and the patient's general health measured on a visual analogue scale (VAS) of 100 mm. We retrieved the use of disease-modifying antirheumatic drugs (DMARD) and the use of biological therapeutics. Serological measures (anti-cyclic citrullinated peptide antibodies [aCCP] and IgM rheumatoid factor [IgM-RF]) were recorded as positive or negative.
Daily functioning was measured with the Health Assessment Questionnaire (HAQ), a reliable and valid questionnaire widely used in RA research. The HAQ score ranges from 0–3; a higher score indicates increased disability.
Pain and fatigue were measured with single items using VAS scales. VAS scales range from 0–10, with 0 meaning no pain/fatigue at all, and 10 meaning severe pain/very tired.
The presence of comorbidity (yes/no) was investigated using a list with 15 common comorbidities including cardiovascular diseases, diabetes mellitus, and psychological disorders. Disease duration was determined by one open-ended question about the year of the patient's RA diagnosis, and the duration of complaints due to RA (answer categories were 0–2 years, 3–5 years, 6–10 years, >10 years). We included one question about the highest attained educational level. Low education was operationalized as primary school, middle education or basic vocational education. Middle education was operationalized as secondary vocational education or intermediate vocational education. High education was operationalized as higher vocational education or a university degree. The patients' gender and age were collected from patient medical records.
External Factors. We measured supervisor and co-worker support, psychological and physical job demands with subscales from the Job Content Questionnaire (JCQ). Items of the JCQ such as 'My colleagues are friendly' are answered on a scale of 1 (totally disagree) to 4 (completely agree). The score of each subscale ranges from 1–4 with a higher score indicating more supervisor support, co-worker support, etc.
Work instability was measured with the RA Work Instability Scale (RA WIS). The RA WIS contains 23 items (yes/no), which are summed for a total score ranging between 0–23. A score of 0–9 indicates low, 10–17 moderate, and 18–23 high work instability.
Sick leave was measured with a single item of the Productivity and Disease Questionnaire (Prodisq). Patients were asked to count the number of days they had been absent from work during the last 12 months.
Three single item questions about being a supervisor (yes/no), shift work (yes/no), and type of job contract (permanent contract/self-employed) were included in the questionnaire worded as The Netherlands Working Conditions Survey.
Statistical Analyses
For the first aim, we determined which combination of factors was best associated with the outcome of at-work productivity loss using linear regression models. These models were built on a dataset of 100 randomly selected participants. The remaining 50 patients were used to cross-validate our model.
In step 1, we performed univariate linear regression analyses for each independent variable with at-work productivity loss. We did not include the number of working hours as an independent variable, since the number of work hours is already incorporated in the WLQ score for at-work productivity loss. Independent variables with a p-value <0.15 were selected for further analyses. In step 2, we built a multivariate model for the nine subscales of the RAND 36, using a backward stepwise procedure, because the subscales are highly correlated. Similarly we built a backward multivariate model for the DAS 28 score and its four components that were highly correlated as well. We retained the components with a p-value <0.15 for further analyses.
In step 3, we built two multivariate models; one multivariate model for the selected internal variables and one model for the selected external variables. All independent variables selected in step 1 and 2 were entered in the models, using a backward stepwise procedure. Variables with a p-value <0.15 were retained in the multivariate models, resulting in an internal and external model. In step 4, one final model was built in which all variables retained in the internal and external model in step 3 were entered again using a backward procedure. Variables with a p-value <0.05 were retained in the final model. R was used to calculate the proportion of explained variance.
The final model was cross-validated in the remaining sample of n = 50 by calculating the R of the final model applied to the sample of n = 50. R was calculated by determining the correlation between the actual outcome measured with the questionnaire, and the outcome calculated based on the regression coefficients in the final multivariate model of step 4 (i.e. outcome calculated = constant + B1 * var1, etc.). The R of both samples (n = 100 and n = 50) were compared.
For the second aim of this study, univariate linear regression analyses were performed to gain insight into the crude associations between at-work productivity loss and quality of life. The independent variables were each subscale of the RAND 36 retained in the backward procedure from step 2, and the dependent variable was at-work productivity loss. Second, we investigated potential confounders for each univariate model. All variables with a p-value <0.15 with at-work productivity loss were considered as potential confounders. Confounders were selected based on a stepwise forward strategy. Each potential confounder was entered separately into the model. The strongest confounder based on the change in the regression coefficient was retained in the model, and subsequently, all remaining potential confounders were entered one by one. The strongest confounder was retained from each round, until the regression coefficient did not change significantly by adding the next confounder. A change in the regression coefficient of more than 10% was considered significant.
All analyses were performed in SPSS version 20.
Source...