Despite improvements in most health indicators, the proportion of the population receiving disability benefits is increasing.1 However, there are few empirically based data on the health consequences following the receipt of disability benefit.1^–3

It is likely that the awarding of benefit has both negative and positive health effects. Models of stress and disease have previously focused on pathogenic wear and tear from stressor-provoked behaviour, ultimately leading to disease. The existence of persisting cognitive factors including negative anticipation, slow recovery from stressors and dwelling on past sources of stress has been recognised.4^–6 These factors appear to be associated with health problems by inducing changes in cardiovascular, endocrine and immune systems.5 7 8 An increased prevalence of such factors seems plausible in the face of job insecurity or altering work patterns,9 and by providing security through disability pensioning, stress and its related exposure may be reduced or removed.

Negative effects might also occur. Transitions from work into various forms of non-employment are associated with psychological distress.10 11 In one cross-sectional study, benefit recipients reported that being on benefits impaired their sleep, reduced their subjective well-being and led to negative psychosocial outcomes.12 Although most disability pensions are granted for non-fatal diseases, several longitudinal studies have found increased all-cause mortality rates among disability benefit recipients,13 even after adjustment for long-standing illnesses.14 15 Increased suicide rates are found among both unemployed and disability benefit recipients, suggesting that stigma, social isolation and behavioural factors associated with being outside the work force may contribute to mortality.16

Interventional studies in this field are unethical and cohort studies present logistical and financial difficulties. As a first empirical step to elucidate possible changes in health status over the process of disability pension award, we used existing data resources to compare health status across groups of recipients in the stages before, during and after disability pension award.

METHODS

Population, data and design

The Hordaland Health Study 1997–1999 (HUSK) was a joint epidemiological research project carried out by the National (Norwegian) Health Screening Service in collaboration with the University of Bergen. The base population included 29 400 individuals in Hordaland County in western Norway born in 1953–57 and aged 40–47 years at the time of data collection. Data were collected by questionnaires and clinical examinations. A total of 18 581 (8598 men and 9983 women) individuals answered the first questionnaire and attended clinical examinations, resulting in a general participation rate of 63% (57% for men and 70% for women).

Using personal identification numbers (issued to all Norwegians at birth), the health survey was linked to National Insurance Administration (NIA) records on disability pensions covering the period January 1992 to December 2004. The records are based on monthly payments, and the accuracy of the registries is well documented.17 Dates of entry, exit and any adjustments in diagnosis, degree of coverage or type of benefit are registered and allow for precise calculations of benefit spells.

For all disability pensioners, we calculated the time interval between their award and date of participation in the health survey, and categorised them according to the corresponding positive or negative time lags: group “7–3 before” had their health measured from 7 to 3 years before being awarded a pension. The groups “3–1 before” and “1–0 before” consisted of those measured from 3 to 1 and from 1 to 0 years before the award, while groups “0–1 after”, “1–3 after” and “3–7 after” had their health measured the corresponding number of years after being awarded a disability pension. Between these six non-overlapping groups, we compared estimated means on the health variables, assuming that recipients of disability pensions were comparable within the period 1992–2004.

Physical conditions

Questions concerning physical diagnoses were framed as follows: “Do you have or have you had (one of the following): coronary infarction, stroke, diabetes, asthma, multiple sclerosis, chronic bronchitis, osteoporosis or fibromyalgia?”. In addition, participants were asked if they had taken any medication the previous day, and if so, for which condition. From these responses, diagnoses according to the Anatomical Therapeutic Classifications system18 were produced and used in a continuous variable of current number of prescribed medications the participant was taking. Systolic blood pressure was measured three times on each participant and the mean of the two last measurements was used.

Mental and somatic symptoms

Anxiety and depression symptoms were assessed with the 14 item Hospital Anxiety and Depression Scale (HADS).19 Good psychometric properties of the HADS have been demonstrated in the Norwegian population.20 The HADS scores were used as separate continuous variables, reflecting anxiety and depression symptom load. Information on sleep problems in the last 30 days was self-reported on a four point Likert scale. Participants were also asked about the frequency of 17 common somatic symptoms in different organ systems in accordance with the ICD-10 research criteria for F45.0 Somatization disorder,21 with ordinal scale labels of “almost never”, “rarely”, “sometimes”, “often” and “almost always”. Responses of “almost always” were counted and accumulated in the variable somatic symptoms. In addition, participants were asked if they had been troubled with muscle pain and/or stiffness in muscles or joints continuously for over 3 months during the last year, and how many areas/joints were involved, reflecting pain distribution.

Statistical analysis and models

To ease comparison, all these variables were transferred to z scores; the mean equals 0, and standard deviation equals 1.

Mean values (95% CI) on the measured health variables were compared across the groups as identified by time between disability award and health survey participation (figs 1 and 2). Deviations from linearity between health and time intervals were tested by first predicting each variable in a linear regression model, using time between benefit and health survey participation as a covariate in model 1. In model 2, we added the time-by-time term and examined p values and changes in explained variance upon inclusion of this term.

Ethics

The study protocol was approved by the Regional Committee for Medical Research Ethics, Western Norway and by the Norwegian Data Inspectorate.

RESULTS

Comparison of disability pensioners to the rest of the population

From January 1992 to December 2004, 1087 (5.9%) of the participants in HUSK were awarded a disability pension at a mean age of 44.2 years. The six strata of individuals awarded a disability pension before and after the health screening were similar with regard to gender and educational level, but age at disability pension award varied systematically due to the design of our study (p<0.001, table 1). Numbers of participants and participation rate in the strata also varied (p<0.01).

Pattern of health status across the strata

The level of somatic and mental symptoms was higher in the groups closer to the time of disability pension award (fig 1). This non-linear inverse U-shaped trend observable across the strata was statistically significant for all symptoms (pain distribution, somatic symptoms, anxiety symptoms, depression symptoms and sleep problems, all p<0.01). Direct comparison of the strata 3–7 years after disability pension award with 3–7 years before (independent sample t test) showed that scores were similar for all symptoms (p>0.05) except for sleep problems which were reported to be 0.38 standard deviations higher in the group 3–7 years after (p<0.001).

No non-linear trends were found for physical diagnoses, prescribed medication or blood pressure (fig 2). There was a tendency to report more pathology after disability pension award (as above directly compared with the group 7–3 years before), but this was only significant for prescribed medication (p<0.05).

DISCUSSION

Main findings

Symptoms of depression, anxiety and sleep problems, pain distribution and somatic symptoms followed an inverse U-shaped trend when comparing the strata in chronological order across stages of the disability pension process. The highest levels of health problems were found at the time of award with better reported health status afterwards. For the more objective measures of health (blood pressure, physical diagnoses and prescribed medication), levels and prevalences were relatively similar across the strata defined by temporal proximity to the disability pension award. For most measures, the level of health problems was equal 3–7 years before compared to 3–7 years after the award, the exceptions being an increase in prescribed medication and more sleep problems after the award.

Strengths and limitations

The main strengths of the study arise from combining a population based health survey with registry data: health was measured in a context separate from disability pension award applications, which should reduce instrumental and incentive based reports of health problems. The information on disability pension award was collected from accurate registries, and misclassification is unlikely. During the study period, there was no major change in disability policy likely to alter the health profiles of disability pensioners. Changes in disability pension influx over this period have been marginal in the age group of interest in this study.22 The health information was gathered as part of a large population based study in an area covering both urban and rural locations. Finally, most variables were measured by employing inventories with well documented psychometric properties.

The limitations of the study mostly apply to design issues, in particular the possibility of selection bias, as we assume that disability pensioners were comparable over this era. Age at disability pension award diverged across the strata (as a result of the study design). However, the collinearity between age at award and the strata far exceeded the recommended tolerance statistics,23 advising against adjusting for this variable. The participants in the different strata did not differ significantly in terms of gender and education, and adjusting for these variables did not influence the results. The participation rate varied significantly across the strata, potentially challenging the presumption of strata comparability. Non-response was higher among those invited around the time of their disability pension award. We had no information on the non-responders, but these were possibly in frequent touch with health services for other reasons at the time, leaving less added gain from the “health check-up”. Health selection can however not be excluded, and generalisations of these observations should be done carefully. Both psychiatric morbidity24 and mortality25 are higher among non-participants in general, and we risk having sampled the healthier disability pensioners. To the degree that health selection was present, the higher proportion of non-responders around the time of disability pension award would lead to reduced reporting of health problems in these strata (1–0, 0–1). As a result, the inverse U-shape observed might be deflated producing an underestimate of the non-linear trends.

Interpretation

There are several possible explanations for the observed differences in health status across the strata, and the following explanations are not mutually exclusive.

(1) Are disability pensions awarded at the peak of health problems?

The increase in symptoms closer to the award may be caused by conditions with progressing severity, increasingly causing work-related disability, and the eventual pension award. As regards the reduction after the award, the point at which disability pensions are awarded should by definition coincide with poor health. If the award is given at a peak during the course of illness, the population’s experience of morbidity could lessen from this point, in line with our observations. Progressive disorders, worsening over time, might explain the repeated finding of raised mortality amongst disability pensioners.14 15 However, the raised mortality remains after adjusting for morbidity, suggesting a role for malign factors other than physical conditions in disability pensioners. Also, most conditions for which disability pension are awarded (eg, mental and musculo-skeletal disorders) are chronic and fluctuating, which might reduce the relevance of this explanation. Also our observation that physical diagnoses do not become more common contradicts this interpretation.

(2) Are the observations caused by detrimental work place factors?

Work place factors might cause or exacerbate health problems, and these might have an adverse effect on health with an increase in symptoms until relief in the form of a disability pension is given. If so, the award would then remove the individuals from this exposure and allow health to improve. However, in the period leading up to disability pension award, most are in rehabilitation and/or on sick leave spells and already removed from the workplace.26

(3) Could the disability pension process affect health?

The observations could be seen as compatible with a general beneficial effect of the disability benefit award on health. The trend of increased health problems towards the time of the award, and a subsequent lower reporting of health problems after the award, could be explained as an artefact of features brought on by the disability pensioning process. Models of stress and disease provide plausible mechanisms, given that prospects of labour market exit, financial insecurity and health problems are potentially highly stressful.

Several meta-studies have demonstrated that financial compensation for disease or injury and related impairment worsens outcomes or prognosis of treatment.27^–29 We cannot exclude the relevance of similar processes in relation to pending disability pension award. Thus, whether conscious or unconscious, symptom amplification to increase the likelihood of receiving a disability pension may explain part of the observed general increase towards the award, and the subsequent reduction following receipt of a pension. Floderus found largely negative effects of being awarded benefits, particularly in sleep quality and subjective well-being.12 In the present study, more sleep problems are reported several years after the award compared to a corresponding number of years before being awarded a disability pension.

Implications

The design limitations preclude any conclusions regarding causality from this study. Nevertheless, if the observations are a result of the disability pension process, they are relevant for policy. First, in terms of doctors’ efforts to apply certification criteria, it is hard to imagine how temporary changes possibly caused by the award process itself can be separated and subtracted from work impairment attributable to the underlying medical problem. Further, the potential positive effects of rehabilitation delivered up towards the time of the award might be countered by this negative trajectory.

Finally, there is currently a shift away from permanent benefits towards temporary benefits with scheduled evaluations at fixed intervals.30 It is possible that the uncertainty and stressful repeated examination might have unintended negative effects on subjects’ health. Evaluations of work capacity must be contextualised such that improvements in work capacity can be identified without the by-product of increasing health problems, whatever the underlying cause.