METHODS

Our review method was based on that used by the NHS Centre for Reviews and Dissemination.⁶ This method involves a systematic examination of selected databases using a variety of strategies, including keywords and subject headings. It allows the integration of quantitative data across studies, where they have similar outcome measures, and the summary of findings where methods used are diverse.

Identification of papers

Four electronic databases were used: Medline (1987–99), PsychInfo (1987–99), Embase (1991–99), and the Cochrane Controlled Trials Register (1987–99). Relevant papers up to and including 1997 were selected from a larger study.¹ The search strategy in the larger study was of MeSH key words and text words in each of three categories: work factors; staff; and ill health/absenteeism/economic consequences. The search included all types of employment and all developed countries but was limited to abstracts in English. Secondary references were chosen from the primary paper references and by contacting academics researching this area. Psychological ill health included measures of anxiety, depression, emotional exhaustion, and psychological distress (“stress” was excluded since it is a mediating hypothetical construct rather than an outcome measure of psychological ill health). For the purpose of this review, papers from 1998 and 1999 were identified using the same search strategy, but excluding physical ill health and economic consequences.

Selection criteria

Abstracts were selected for retrieval of the paper if they were judged to include data about both work factors and psychological ill health or absenteeism. Dissertations were excluded, as were studies of very specific staff groups or settings, work patterns (for example, shift working), or events (for example, violence). All abstracts were selected independently by two researchers (three researchers were involved in this activity). The percentage of abstracts for which two researchers agreed about inclusion and exclusion varied between 80% and 90%. Disagreements were resolved by discussion.

Information extraction

Information from papers was extracted and coded within the following categories: study aim, study design, type of study population (for example, occupational group), sampling strategy, sample size and response rate, demographic characteristics, type of intervention, type of study measure, main outcomes, and summary of results.

Further selection criteria

Coded papers excluded from the review were studies with: volunteer or inadequately described sample; response rate of less than 60%; no standardised measures of psychological outcome.

RESULTS

Of the studies identified as part of the larger study,¹ 40 were selected for this study (34 associations and six interventions). A further nine studies meeting the above selection criteria were identified in the period 1998–99, all of associations. No studies were found in the Cochrane Controlled Trials Register. The results are summarised in tables 1–4.

Because these studies were diverse in terms of outcomes and measures used to assess these outcomes, a meta-analysis was not appropriate.

Associations with work

The results are presented in three groups: health care workers in the UK, health care workers in other developed countries, and non-health care workers. This enabled an assessment of whether associations between work factors and psychological ill health are similar across sector and country.

Health care

In the UK, factors associated with psychological ill health in doctors, from junior to senior grades, are long hours worked,⁹ high workload and pressure of work,^7,16,11 and lack of role clarity¹² (table 1). Pressure of work has also been found to be associated with poor mental health in dentists.¹⁰ In family doctors, the issues were interruptions during and outside surgery hours and patient demands.¹⁶

Among UK nurses, the most frequently reported source of psychological ill health was workload pressures.¹⁷ Distress in student nurses has been caused by low involvement in decision making and use of skills, and low social support at work.¹³ In a study of health care workers across job type, bullying was found to be prevalent, carried out mainly by managers and associated with both anxiety and depression.¹⁴ Of the two studies addressing sickness absence, one found a negative association with job demands,¹³ while the other found no association with control over work.¹⁵

Similar factors are associated with psychological ill health in health care workers in the rest of Europe, the USA, and Australia (table 2). The one study of doctors found an association between work control and social support and psychological distress.²² Among nurses, lack of co-worker support,^24,27 job influence,²⁶ and organisational climate and role ambiguity²⁸ were associated with psychological distress. Among other hospital workers, work overload and pressure, role ambiguity, lack of control over work, and lack of participation in decision making were all found to be associated with distress.^18,20,25

Sickness absence was associated with work pressures and lack of training,²³ unsupportive management style,²¹ role ambiguity, tolerance of absenteeism, and low pay.¹⁹

Beyond health care

The picture among non-health care workers in Europe and the USA was similar to that of health care workers (table 3). The key work factors associated with psychological ill health were: work overload and pressure^{31,34,36,39–40,41,47,48}; conflicting demands⁴⁷; lack of control over work and lack of participation in decision making^{34,36,37,39,40,46–48}; poor social support at work^{31,33,35,38,39,41,47,48}; unclear management and work role^{29,30,41,34,38}; interpersonal conflict^42,46; and conflict between work and family demands.⁴⁶ Long hours were found to be associated with depression in women, but not in men.⁴⁴

Sickness absence was negatively associated with high job demand,⁴⁵ and positively associated with monotonous work, not learning new skills and low control over work,^36,37,45 and non-participation at work.⁴³

Evaluated interventions

Six intervention studies met our methodological criteria (table 4). Three were randomised controlled trials,^49,50,52 three were conducted in the USA,^50,51,54 one in the UK,⁵³ and two in Scandinavia.^49,52 Three were of health care workers.^50,52,54 Five were training programmes offered as part of the working day and one was an organisational intervention.⁵³

Skills to mobilise support at work and to participate in problem solving and decision making were taught to care staff of people with mental ill health or mental disability in a randomised controlled trial.⁵⁰ Groups of 20 had six sessions of 4–5 hours training over two months, and were trained to train those in their workplace. Compared to those in the control group, the intervention group reported more supportive feedback, more ability to cope, and better work team functioning and climate. Among those most at risk of leaving, those undergoing the training reported reduced depression. The second randomised controlled trial compared receiving support, advice, and feedback from a psychologist with having the passive presence of the same psychologist at staff meetings in a geriatric hospital facing organisational change.⁵² Staff were taught skills of stress management, and how to participate in, and control, their work. The intervention was an hour a fortnight during the 10 weeks before, and the 10 weeks after, the organisational change. There was a significant difference between groups, with a decrease of stress hormone levels in the intervention group.

Staff of a psychiatric hospital were taught verbal and non-verbal communication and empathy skills.⁵⁴ Groups of 6–8 had eight hour weekly sessions for four weeks involving information, videos, modelling, and role playing. Compared to a matched control group, the intervention group showed reduced staff resignations and sick leave, although no statistical tests are reported.

Among physically inactive employees of an insurance company, a randomised controlled trial found stress management training and aerobic exercise interventions had mixed effects.⁴⁹ After three sessions a week for 10 weeks, stress management training resulted in improved perceived coping ability but no change in physical or psychological health. Aerobic exercise resulted in improved feelings of wellbeing and decreased complaints of muscle pain.

Employees of a fire department underwent one of seven training programmes emphasising one or more aspect of stress management: physiological processes, coping with people, or interpersonal awareness processes.⁵¹ Weekly sessions for 8–10 people were run over 42 weeks. There was no control group. Compared to baseline, there were reductions in depression, anxiety, psychological strain, and emotional exhaustion immediately after the programme. There was a further reduction in psychological strain and emotional exhaustion at 9–16 months follow up.

A structural intervention for local authority staff on long term sickness absence was effective in reducing sickness absence. Referral to occupational health services was triggered after two or three months absence, rather than at six months which was the practice before the intervention. The average duration of sickness absence reduced from 40 to 25 weeks before resumption of work and from 72 to 53 weeks for those staff who left employment for medical reasons. The authors describe large financial savings but no statistical tests are reported.⁵³

DISCUSSION

This systematic review of a large number of studies covers a wide range of employment sectors in the developed world and summarises those studies that use rigorous methods. The studies show that, while levels of psychological ill health are higher in health care than in non-health care workers,⁵ the associations between work factors and psychological ill health are similar. They are also similar across continents. This suggests that a generic approach to reducing work related psychological ill health may be appropriate.

The most common work factors associated with psychological ill health were work demand (long hours, workload, and pressure), lack of control over work, and poor support from managers. These were also associated with sickness absence. The findings of this review, summarised in tables 1–4, are consistent with the demand-control model of job strain.³⁶ Interventions aimed at changing these workplace factors reduced psychological ill health.

This review highlights limitations in the research identified. The studies that have been carried out are limited in the questions addressed and in the study designs used. Since most studies are cross sectional, causal relations cannot be shown. It may be that the associations found reflect a tendency for more vulnerable people to choose work in caring roles or other types of job which are well represented in published research studies. The question of what aspects of work lead to ill health and sickness absence can only be addressed by longitudinal studies that are able to investigate the causal relations between work factors and health outcomes and by randomised controlled trials of interventions. A longitudinal study that directly addressed the nature of the relation found a causal relation between psychological stress and psychosomatic complaints.⁵⁵

There are several practical implications suggested by the studies of association in this review, for both employment practices and management style. Intervention studies, however, have focused mainly on staff training. There is a need for future studies to evaluate interventions based on employment practices and management style. This would represent primary prevention, reducing sources of psychological ill health, rather than secondary prevention, training individuals who are already experiencing work related stress, to be more robust in the face of such pressures. Only one of the intervention studies included an economic evaluation: such evaluations are important in facilitating employers to make decisions about whether or not to implement interventions.

Future research should adhere to minimum scientific standards absent in many of the studies reviewed, such as adequate design, sufficiently large samples, and valid outcome measures. Lessons that are learnt from rigorously evaluated interventions can then be applied more generally.