Article Text

Download PDFPDF

Psychological distress and risk of long-term disability: population-based longitudinal study
  1. Dheeraj Rai1,
  2. Kyriaki Kosidou2,
  3. Michael Lundberg2,
  4. Ricardo Araya1,
  5. Glyn Lewis1,
  6. Cecilia Magnusson2
  1. 1Academic Unit of Psychiatry, School of Social and Community Medicine, University of Bristol, Clifton, Bristol, UK
  2. 2Division of Public Health Epidemiology, Karolinska Institute, Norrbacka, Stockholm, Sweden
  1. Correspondence to Dr Dheeraj Rai, Academic Unit of Psychiatry, School of Social and Community Medicine, University of Bristol, Oakfield House, Oakfield Grove, Clifton, Bristol BS8 2BN, UK; dheeraj.rai{at}


Background Common mental disorders are known to cause long-term disability, although not much is known about long-term consequences of milder forms of psychological distress.

Objective To investigate the association between increasing levels of psychological distress and 5-year risk of long-term disability pensions awarded for somatic or psychiatric conditions.

Methods In this longitudinal population-based study, a cohort of 17 205 individuals, aged 18–64 years, recruited in 2002 in Stockholm County was prospectively followed up for new disability pension awards. The 12-item General Health Questionnaire (GHQ-12) was used to measure baseline psychological distress, and participants were categorised as having no, mild, moderate or severe psychological distress (GHQ-12 scores of 0; 1–2; 3–7 and 8–12, respectively). Details of new disability pension awards were obtained through record linkage with the Swedish National Insurance register. Comprehensive information on a range of sociodemographic, lifestyle and health characteristics was available.

Results Increasing levels of psychological distress at baseline were associated with an increased likelihood of obtaining a disability pension later in life. Even mild psychological distress was independently associated with the award of a disability pension for both somatic (HR=1.7; 95% CI 1.3 to 2.2) and psychiatric diagnoses (2.2; 1.4 to 3.6). Over a quarter of disability pensions awarded for a somatic diagnosis, and almost two-thirds awarded for a psychiatric diagnosis, could be attributed to psychological distress.

Conclusions Mild psychological distress may be associated with more long-term disability than previously acknowledged and its public health importance may be underestimated.

  • Depression
  • disability
  • disabling disease
  • psychological distress
  • psychosocial epidemiology

Statistics from


Psychological distress is a common experience in modern life and is known to occur on a continuum of severity with the depression and anxiety-related common mental disorders. Although relatively little empirical evidence exists on the consequences of mild distress, the common mental disorders are known to be associated with significant human and societal costs, including poor health outcomes,1 2 greater mortality3 and disability.4–9 The costliest societal outcomes of these disorders include loss of productivity and disability benefit disbursements, which have greatly increased in many developed countries over recent years.6 Sweden has also experienced large rises in people with permanent disability and seen the share of psychiatric disorders as underlying reason for disability pensions rise from 14.3% to 41.3% in women and from 17.6% to 39.9% in men between the years 1990 and 2007.10 Despite vigorous return-to-work incentives and policies implemented in many countries, <1% of people on disability pensions return to work each year.11

Only one population-based study has specifically investigated the longitudinal association of common mental disorders and long-term disability and unsurprisingly found them to be strongly associated with future disability pension awards.6 Studies in primary care,12 farmers,13 workforce cohorts14 and among those on sick leave15 have also reported similar findings. However, since milder forms of distress, below the threshold of conventional caseness, have not been considered in these studies, the societal and public health importance of psychological distress may have been underestimated. Medical literature on non-specific and milder forms of psychological distress has often attracted controversy, one important argument being that a clinical view towards mild psychological distress may represent unnecessary medicalisation of normal human suffering.16 However, a few longitudinal studies4 8 9 into some milder conditions have found them to be associated with short-term disability, although robust data on longer-term outcomes are lacking. Finally, whether psychological distress leads to disability related to somatic conditions has rarely been studied.

Using a unique link between a large population-based health survey and national registries in Sweden, we aimed to study if increasing levels of psychological distress at baseline were associated with long-term disability (measured by the award of a disability pension) over a 5-year period. We also examined the relationship between psychological distress and disability pensions awarded for somatic and psychiatric conditions and aimed to quantify the relative contribution of varying levels of distress to the burden of long-term disability in the population.


Study population

This was a prospective cohort study combining information from self-administered questionnaires and a range of Swedish health and administrative registries using unique national identification numbers. A baseline questionnaire survey was sent in 2002 to a random sample of 50 000 residents of Stockholm county, aged 18–84 years. Responders (n=31 182, 62.4%) were asked to participate in a follow-up study in 2007. Participants in both surveys (n=24 655, corresponding to a 79% retention rate) constitute the Stockholm Public Health Cohort, the dataset used for this analysis. In comparison with Stockholm County census data, a higher proportion of men, those under the age of 45, those born outside Sweden, either single or separated and those unemployed or on the lowest quartile of income were likely to be non-responders to the surveys. Individuals over the age of 64 (n=4818), those who were in receipt of disability pension at baseline (n=962) and those with missing data (n=1670) were excluded leaving 17 205 individuals that comprise the study population with complete data. All respondents consented to use of their self-reported data and record linkage with national registries for research. Ethical approval for the study was granted by the Stockholm regional ethical review board.


Psychological distress

The 12-item General Health Questionnaire (GHQ-12),17 included in the baseline survey, was used to identify levels of psychological distress. The GHQ-12 is a widely used self-reported questionnaire measuring psychological distress and screening for common mental disorders. It has been validated for use in the Swedish population18 and a score of ≥3 (using the recommended standard 0-0-1-1 scoring) is used in Sweden to denote significant psychological distress.19 However, in order to study the consequences of distress below this threshold we, a priori, divided our sample into four groups of increasing psychological distress—those with a score 0 (depicting no reported distress, reference category for comparisons); scores of 1 or 2 (depicting mild distress, subthreshold for caseness), 3–7 (moderate distress) and 8–12 (severe distress).

Disability pension

The primary outcome measure was the award of a new disability pension at any time between 1 January 2003 and 31 December 2007. These data were derived from the Swedish social insurance register, which has national coverage. Disability pensions are monetary benefits awarded to people who have an underlying disabling medical condition that leads to an inability to work full-time for at least 1 year using an assessment process guided by medical recommendation. This process is similar to that of other developed countries, although Scandinavian countries invest a greater proportion of their gross development product on social benefits.6 Two types of disability pension are awarded in Sweden: (a) sickness compensation (for people aged between 30 and 64 years) and (b) activity compensation (for people aged between 19 and 29 years). The criteria for award of these are identical but these two age-dependent categories exist in order to target more intensive interventions towards the rehabilitation of younger people. We combined these categories to form a single disability pension variable covering everyone aged 19–64 years. The primary diagnosis for reason of receipt of disability pension is recorded in the register according to the International Classification of Diseases, 10th revision (ICD-10). We coded these into mental health and somatic diagnosis for analysis. Details of the registers and disability pension system are available from the Swedish social insurance system website (

Potential confounders

Sociodemographic variables

Participant's age, gender and details of current employment were obtained by self-report. An occupational socioeconomic class variable with seven categories was defined according to Swedish guidelines.20 These included skilled and unskilled manual workers; low-, intermediate- and high-level non-manual employees; and self-employed. Those with no reported occupation (including housewives, students, retired or unemployed) were classified as ‘economically inactive’. The highest educational qualification of respondents in year 2002 in three categories (compulsory education, upper secondary education and higher or university education) was obtained by record linkage with the Integrated Database for Labour Market Research, a central database with national coverage held by Statistics Sweden ( comprising family and individual data on education, income and other demographic parameters. This register was also used to classify respondents into quartiles based on their total annual household income.

Lifestyle factors

Respondents were asked to estimate the average time each week they spent walking, cycling or getting any other form of exercise.21 Those reporting <2 h of exercise a week were classified as having a sedentary lifestyle. Baseline daily tobacco smoking and past 12-month cannabis use were ascertained by self-report and used as binary variables for analysis. Alcohol use was determined by asking the respondents to fill in a typical week's timetable of the type and amount of alcohol used. These responses were used to estimate the amount of 100% alcohol in grams consumed per week and respondents were classified into abstainers, low alcohol users, moderate alcohol users and heavy alcohol users according to WHO guidelines.22 Self-reported height and weight were used to calculate the body mass index in kg/m2 categorised into <20, 20 to <25, 25 to <30 and ≥30.

Musculoskeletal pain

The presence of daily pain in the neck, shoulder, arms or lower back was ascertained by self-report using the Nordic Musculoskeletal Questionnaire23 at baseline. We handled this variable separately from the somatic morbidity variable to ensure a fuller coverage of musculoskeletal pain conditions that are responsible for a large proportion of disability pension claims.10

Somatic morbidity

Somatic morbidity was assessed using a combination of self-reported and register data. Respondents were asked a series of questions about the presence of somatic illnesses diagnosed by a healthcare professional that included diabetes, angina pectoris, hypertension, myocardial infarction, heart failure, stroke and asthma. Further, we used data from the Swedish National Hospital discharge register to include everyone with a baseline secondary-care diagnosis of any somatic illness. We combined these data to construct a somatic illness variable with three categories: (a) those with no self-reported or registered diagnosis of somatic illness; (b) self-reported but not registered diagnosis and (c) register-recorded diagnosis of somatic illness.

Diagnosed psychiatric disorders

Diagnosed psychiatric disorders were identified by record linkage with a centralised database of all public mental health service usage (including outpatient and inpatient service use) in Stockholm County. Each person in current contact with the mental health services has a primary diagnosis listed according to the ICD-10. We divided these diagnoses into two main groups—the common mental disorders (ICD-10 F32.0-F48.9) and all other psychiatric diagnoses (ICD-10 FXX.XX except F32.0–F48.9).

Statistical analyses

Descriptive analyses were conducted to study characteristics of the sample according to the baseline levels of psychological distress. The association between psychological distress (independent variable) and onset of disability pension (dependent variable) was studied using multivariable Cox proportional hazards regression models. Time at risk (in years) for onset of disability pension was calculated. People who died (n=119) during the follow-up period were censored from analyses at time of death. The assumption of proportionality was tested using the log-rank test. After estimating age- and gender-adjusted associations (model 1), we adjusted for sociodemographic and lifestyle variables, including employment status, highest educational qualification, social class, born in or outside Sweden, cohabiting or living alone, quartile of gross household income, presence of sedentary lifestyle, daily smoking and past 12-month cannabis use (model 2). In our final model we further adjusted for pain, somatic illness and secondary care diagnosis of mental illness (model 3). This model, which included diagnosed mental illness at baseline, although probably overadjusted, was included to ensure any associations between mild distress and disability were not observed owing to a small number of people with diagnosed mental illnesses in remission at baseline. The alcohol use variable had >9% missing data and was excluded from these models, but was studied in separate sensitivity analyses to exclude its effect as a confounder. Hazard ratios (HRs) and 95% CIs from regression models depicting risk of being awarded a disability pension within 5 years are presented.

The above analysis was repeated after categorising the outcome by documented somatic or psychiatric diagnosis in order to assess if psychological distress was associated with increased risk in either or both. A previously described method24 was used to estimate population attributable fractions from the final model in order to study the public health impact of different levels of psychological distress on disability pension. These depict the proportion of disability pension in the population that might be prevented if the corresponding exposure (categories of psychological distress in our study) were completely eliminated; assuming the associations were causal and all confounding had been accounted for. All analyses were conducted using SAS version 9.1 (SAS Institute Inc).


Characteristics of the sample according to baseline psychological distress are presented in table 1. The sample comprised 17 205 individuals and over half of them reported no psychological distress (GHQ score=0) at baseline. Six hundred and forty-nine people started receiving disability pension during the follow-up period, of whom 203 were granted a pension for a psychiatric illness and 446 for a somatic illness.

Table 1

Baseline characteristics and details of follow-up of the sample according to psychological distress at recruitment (n=17205)

Before and after adjusting for potential confounders a graded positive association between levels of psychological distress and receipt of disability pension (p<0.001 for all comparisons, table 2) was found. Even those reporting mild distress (GHQ-12 score of 1 or 2) had an increased risk of disability (HR=1.7, 95% CI 1.4 to 2.2). More than a third of the disability pension in the population (population attributable fraction 38.1%, 95% CI 26.2% to 48.8%) in our final model could be attributed to psychological distress after accounting for all other factors, including diagnosed mental and somatic illness. Figure 1 shows a Kaplan–Meier curve plot demonstrating the cumulative probability of being awarded a disability pension in 5 years according to varying levels of baseline psychological distress.

Table 2

Twelve-item General Health Questionnaire (GHQ-12) scores at baseline and 5-year risk of disability pension—Cox proportional hazards regression models

Figure 1

Kaplan–Meier curve depicting probability of receiving a new disability pension within 5 years of baseline measures of the 12-item General Health Questionnaire (GHQ).

Results of repeating these models after categorising the outcome into disability pension for somatic versus psychiatric diagnoses are presented in table 3. Similar to the overall analysis, a strong and increasing risk was observed for receipt of disability pension for both somatic and psychiatric diagnoses with increasing levels of psychological distress (p<0.001 for all comparisons, table 3). As expected, risk at any level of psychological distress was higher for those awarded disability pension for a psychiatric than for somatic diagnosis. Population attributable fractions for these models suggest that more than a quarter of the disability pensions due to somatic diagnoses (26.6%, 95% CI 12.6% to 39.5%) and nearly two-thirds due to psychiatric diagnoses (64.2%, 95% CI 46.4% to 77.0%) could be attributed to psychological distress.

Table 3

Twelve-item General Health Questionnaire (GHQ-12) scores at baseline and risk of disability pension for somatic or psychiatric diagnosis—Cox proportional hazard models


We found a strong graded relationship between increasing levels of psychological distress and the likelihood of being awarded a new disability pension within 5 years in this large Swedish population-based cohort. Even mild distress, as assessed at baseline by the General Health Questionnaire was associated with an increased risk of disability pension awarded both for psychiatric and somatic diagnoses. Our findings were robust even after accounting for a large number of sociodemographic, lifestyle and health characteristics (including self-reported pain, and diagnosed somatic and psychiatric illnesses).

Strengths and limitations

The large population-based sample, longitudinal design and access to comprehensive information on potential confounders are the major strengths of this study. The register data we used are complete with national coverage and likely to have reduced any systematic bias in ascertainment of disability pension and other variables. Diagnoses recorded in these administrative registers have been shown to have high clinical validity in previous research.25 However, using only primary diagnoses for receipt of disability pensions may have led us to underestimate the role of certain comorbid conditions such as personality disorders, which may be important ‘hidden’ drivers for disability pension awards as they may often remain undocumented or labelled as secondary diagnoses. The use of standardised validated self-reported measures17 21–23 should have minimised the possibility of measurement and observer bias. We used a single GHQ-12 measurement at baseline and did not ascertain the reason for psychological distress or study its course before and over the follow-up period. Our results are unlikely, however, to have resulted by chance or type 1 error as the associations consistently followed a graded pattern in the expected direction and were robust in multiple models. Also, the participants had no prior knowledge of our research hypotheses; therefore we expect minimal reporting bias for secondary gain—for example, to gain access to disability pensions. Some associations during the early follow-up period may have been related to reverse causality—for example, people anxiously awaiting results of their disability pension application at baseline may have had elevated GHQ-12 scores, but these are unlikely to have substantially biased our results over the 5-year period. Owing to differential non-response to the surveys, our study might have some non-response bias; we tried to minimise this by controlling for the most important characteristics of non-responders in regression models. Although we had access to information about a large number of confounders, the possibility of residual confounding can never be excluded.

Comparison with previous literature

The only other population-based study6 testing a similar hypothesis to ours used established cut-off points (≥8 on the Hospital Anxiety and Depression Scale) to ascertain anxiety or depressive disorders and hence did not report on the impact of symptoms below these thresholds. Using a score of 0 on the GHQ as reference, our study adds to the literature by studying the full spectrum of psychological distress and adjusting for a larger set of potential confounders over a longer follow-up period. Our results are also consistent with the few longitudinal population-based studies that report on the short-term disability associated with subthreshold forms of depression4 8 9 and add to their findings by using a more robust real-life measure of long-term disability, related to great and measurable societal costs.

Our results are readily generalisable to Stockholm County, an urbanised, secular, Western European society, although caution must be exercised before interpreting them for other areas. In the absence of similar studies in other regions of the world, this study presents the most robust evidence of the effect of mild psychological distress on long-term disability.

Mild psychological distress and disability

An important finding of our study was that even mild psychological distress was associated with later onset of long-term disability. It should be noted that the people that we categorised as mildly distressed (with GHQ scores of 1–2) are considered as having ‘good mental wellbeing’ in Swedish health surveys that form the basis of public mental health policy.19 Also remarkable is that the distress was measured only once using the GHQ-12, which may be sensitive to a large number of stressors. Mild and subthreshold disorders are known to be risk factors for future development of clinically overt psychiatric illness26 and one mechanism for our finding may be that people with mild distress at baseline may later be awarded disability pensions because their symptoms worsen into an appreciable psychiatric disorder. Another possibility is that people with mild symptoms at baseline who receive disability pensions later may represent chronic cases8 of psychiatric disorders in partial remission not known to secondary care.

It is also important to consider these results in the context of modern labour market and societal characteristics. A body of economics literature documents the increasing demands placed on employees, including greater knowledge, social, technical and intellectual skills and greater flexibility to succeed in similar jobs over the past decades.27 28 Factors outside the workplace, such as fewer close relationships, smaller social networks, reduced perceived social support, participation and trust, may also be associated with psychological distress and potentially contribute to disability.29 30 Is it possible that in a modern, education-based, post-industrial society like Sweden, which provides universal access to benefits, people with mild psychosocial deficits find it easier to access benefits than meet complex labour market demands? Are the strains and demands of modern society commonly exceeding human ability? Whether our results reflect societal or labour market problems beyond a public health issue is an important subject to be explored in future research, especially through international comparisons, as the Swedish population is known to enjoy a better work–family balance and fewer work pressures than people in other European countries.28 31

Psychological distress and risk of somatic morbidity

It is expected that psychological distress may be associated with a diagnosis of psychiatric illness and related disability later in life. However, more surprisingly, over a quarter of the total burden of disability pensions granted for somatic diagnoses in our study could be attributed to psychological distress. Mykletun et al6 reported similar findings for cases of common mental disorders, but the pathways of psychological distress leading to pension awards due to physical illness are unclear. There may be several possible explanations. First, comorbid psychiatric symptoms, especially depressive, have been shown to lead to poor outcomes in a range of somatic disorders32 33 by several mechanisms, including non-compliance with treatment and multiple adverse health behaviours.1 34 Second, self and societal stigma35 towards psychological problems may prompt doctors and patients to prefer somatic labels.36 Consultations where psychological distress is expressed through physical symptoms are common in primary care and are known to lead to non-recognition of primary mental disorder.37 Furthermore, almost all medical specialties have their own ‘functional’ syndromes presenting with medically unexplained somatic symptoms that are often associated with psychological disorders.38 Finally, the search for biological explanations linking mental and somatic disorders has generated some exciting results—for example, the role of proinflammatory cytokines as a common pathway.39 Further development of integrated mind–body research may advance the understanding of these pathways.


We found that mild psychological distress may be associated with more long-term disability than previously known. From a public health perspective, our results confirm that the importance of poor mental health on society may be underestimated when assessing common mental disorders using strict categorical diagnostic criteria. A dichotomous view towards common mental disorders may be useful for standardising clinical and treatment decisions but limits the understanding of the continua in which symptoms naturally exist. Our results support the argument for using dimensional measures40 to account for people with symptoms below established thresholds.

People with psychological distress often present to primary care doctors but for many, especially those with mild and non-specific presentations, this goes undetected.41 42 It must be remembered that psychological distress has many psychosocial explanations apart from biological and should not be ‘overly medicalised’.16 However, our results suggest that presentations of psychological distress at the least merit an assessment before being considered unimportant. The use of simple and quick self-reported screening questionnaires may aid this process. A related problem is whether improved recognition can be translated into better health or functioning. A small but growing evidence base documents better health and societal outcomes associated with active measures to reduce psychological stress in a range of non-clinical43 44 and clinical populations.42 45–48 However, the use of standard cut-off scores for measurement of distress in the above studies excludes people with milder psychological distress who we highlight.

Evidence relating to the possibility and means of reducing milder forms of psychological distress in the population is extremely limited and we hope our study will fuel this line of empirical enquiry. The development of cost-effective psychosocial or medical interventions for people with mild psychological distress may be a priority worth pursuing as it may reduce the disability burden, improve health outcomes and generate significant societal savings.

What is already known on this subject

  • Anxiety and depressive disorders are associated with long-term disability but not much is known of the consequences of milder psychological distress.

  • Anxiety and depressive disorders may contribute to disability related to somatic conditions.

What this study adds

  • A single point estimate of psychological distress as measured by the General Health Questionnaire may provide information on long-term disability.

  • Even mild psychological distress, below the threshold of conventional caseness was associated with future long-term disability related to both somatic and psychiatric conditions, even after accounting for a large number of sociodemographic, lifestyle and health characteristics.

  • Over a quarter of the disability pension burden associated with somatic conditions, and almost two-thirds associated with psychiatric diagnosis in the population could be attributed to psychological distress.


We thank Dr Christina Dalman for her support in data linkage and useful comments on the design of this study.



  • Funding Supported by a grant from Stockholm County Council, Swedish Council for Working Life and Social Research (DNR 2007-2064).

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Stockholm Regional Ethical Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.