Stress in UK intensive care unit doctors

S. Coomber1,6, C. Todd*,2,7, G. Park1, P. Baxter3, J. Firth-Cozens4,8 and S. Shore5,9

1 Addenbrooke’s Hospital NHS Trust, Hills Road, Cambridge CB2 2QQ, UK. 2 Health Services Research Group, Department of Public Health and Primary Care, University of Cambridge, Institute of Public Health, Cambridge CB2 2SR, UK. 3 University of Cambridge Occupational Health Service, Fenner’s, Gresham Road, Cambridge CB1 2ES, UK. 4 Department of Psychology, University of Leeds, Leeds, UK. 5 Centre for Applied Medical Statistics, Institute of Public Health, Robinson Way, Cambridge CB2 2SR, UK 6 Present address: Ipswich Hospital NHS Trust, Heath Road, Ipswich IP4 5PD, UK. 7 Present address: School of Nursing, Midwifery and Health Visiting, Coupland III Building, University of Manchester, Oxford Road, Manchester M13 9PL, UK. 8 Present address: London Deanery, 20 Guilford Street, London WC1N 2BZ, UK. 9 Present address: MRC Environmental Epidemiology Unit, University of Southampton, Southampton General Hospital, Southampton SO16 6YD, UK

Accepted for publication: July 19, 2002


    Abstract
 Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 References
 
Background. Doctors have long been considered at risk of occupational stress.

Methods. A postal survey of all members of the Intensive Care Society using validated instruments.

Results. Eight-five per cent of members returned questionnaires and 70% were eligible for the study. Twenty-nine per cent were suffering General Health Questionnaire-12 (GHQ-12) identified distress and 12% Symptom Checklist-Depression (SCL-D) defined depression. There were no significant age or sex differences between staff suffering distress or depression and those who did not. Dissatisfaction with career correlated highly with both distress and depression (P<0.01). Twenty doctors (3%) were bothered by suicidal thoughts. The most stressful aspects of work were bed allocation, being over-stretched, effect of hours of work and stress on personal/family life, and compromising standards when resources are short. Logistic regression revealed mental health problems were predicted by five stressors: ‘lack of recognition of one’s own contribution by others’; ‘too much responsibility at times’; ‘effect of stress on personal/family life’; ‘keeping up to date with knowledge’; and ‘making the right decision alone’.

Conclusions. Nearly one in three ICU doctors appeared distressed (GHQ), and one in 10 depressed (SCL-D); this is no greater than that reported in other specialities. Perceived stressors reveal some key areas of concern for the employer and the specialty.

Br J Anaesth 2002; 89: 873–81

Keywords: intensive care, depression; intensive care, occupational stress


    Introduction
 Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 References
 
Occupational stress is a recognized problem in health care workers, and doctors have long been considered to be at special risk.14 Relatively few studies have been undertaken of UK hospital specialists, with six recent studies of consultants showing a range between 22 and 46% of respondents exhibiting clinically important levels of psychiatric morbidity, as assessed by standard psychiatric measures.511 Figure 1 presents data from studies conducted by one of us that have used the General Health Questionnaire (GHQ) to assess psychiatric morbidity. It is clear that one in five to one in three doctors suffer from mental distress at any one time, depending on the threshold for morbidity accepted and the specialty in which they work. To study job stress within a single speciality, we undertook a national survey of intensive care specialists using the membership list of the Intensive Care Society (ICS) as a sampling frame. Intensivists routinely work in a demanding, highly technical environment where death and dying are common events,12 and errors can be dangerous.13 Limited resources affecting allocation of intensive care unit (ICU) beds can be a particular source of stress for ICU doctors.14 In addition, many ICU consultants are trained or still practice in anaesthesia, a specialty suspected of being at a higher risk of stress-related ill-health,1518 and suicide19 20 compared with other hospital doctors. The aim of the survey was to identify ‘distressed’ doctors as defined by the GHQ,21 and relate this to repeated and long-term exposure to job stressors.



View larger version (10K):
[in this window]
[in a new window]
 
Fig 1 Percentage (95% CIs) of sample with GHQ scores indicative of mental distress in previous studies (morbidity thresholds of >3 and >4).

 

    Methods
 Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 References
 
A complete list of ICS members working in ICUs in the UK was obtained. We sent a letter to each member explaining the study and asking him or her to take part. To maximize the response, a letter from the President of the UK ICS was enclosed, which emphasized the importance and confidentiality of the survey. Non-responders were followed-up with a maximum of two further mailings. The main questionnaire used was based on one validated in previous studies by one of us,2224 which included the 12-item GHQ,21 and the Symptom Checklist Depression Sub-scale (SCL-D).25 Information was also requested about workload, job satisfaction, and personal characteristics. Thus, all data in this paper are from validated screening instruments or based on self-report, using questions from previous studies. Questions were for the most part closed in nature.26 For example, respondents were requested to indicate in a box how many hours they had worked during the previous week in total and then how many hours were engaged in NHS work, specific ICU work, on call time, teaching, etc. The frequency and severity of a list of ICU-specific stressors were rated using Likert type scales. Individuals indicated how frequently (scaled 0–2=never/occasional/often) and how stressful (scaled 0–4=not at all/slightly/moderately/very/extremely) they found each of the stressors. This list of stressors was based on those identified previously in surveys of junior doctors27 and anaesthetists,24 and adapted for the present study with the help of staff from Addenbrooke’s Hospital NHS Trust, Cambridge.

The GHQ is a well-validated, self-administered questionnaire commonly used to detect psychiatric disorders in the community and other settings.7 28 It focuses on two symptom categories: the inability to carry out normal functions and the appearance of new phenomena of a distressing nature.21 It is known to detect both transient and sustained symptoms of distress. GHQ scores can be interpreted as an indicator of morbidity, and a threshold differentiates psychiatric ‘cases’ from ‘non-cases’. We used two threshold scores; the ‘standard’ threshold of >3 and the ‘high’ threshold of >4, so that our data would be comparable with previous work.2224 The SCL-D was also scored in the standard way and a threshold of >1.5 used to distinguish depression ‘cases’.25

As part of our study we asked about alcohol and drug use in a second separate questionnaire. Because of their sensitivity and so as to maximize response, these responses were anonymous and returned separately from the main questionnaire. Thus, we cannot link alcohol and drug use to specific responders.

Statistics
Data were double entered onto the computer. Parametric and non-parametric statistical analyses were undertaken using SPSS V.6.1. Throughout this paper, data reported are based on numbers of valid responses to items for each group or subgroup. Thus, numerators and denominators may not always total to the full sample size. To avoid products of zero, we added a constant of one to frequency and stressfulness stressor scores when calculating products. Univariate analysis of personal characteristics and professional status was performed with the GHQ defined groups (using t-tests for those with two categories and ANOVA for three or more). Linear regression was performed with GHQ and SCL-D scores as dependent variables and age, sex, marital status, and each product of the individual stressors as independent variables. Forwards selection procedures were used to fit a multivariate model with a reduced number of variables and to identify predictors of GHQ and SCL-D defined morbidity. In the literature review, confidence intervals for estimates from other reports (Fig. 1) were calculated using Confidence Interval Analysis (CIA) Version 1.2.29 We also used CIA in analysis of our anonymous alcohol and drugs questionnaire.


    Results
 Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 References
 
Response rate
A total of 896 sets of questionnaires were sent out and 85% were returned (758/896). However, the ICS membership list used as the sampling frame included 108 (12%) doctors who replied that they were not currently working in ICU in the UK, and so we excluded them. Twenty-three questionnaires returned were inadequately completed to be included (i.e. indicated they did not wish to take part, or returned form blank). Thus, there were 627 eligible participants who had adequately completed the main questionnaire, giving a response rate of 70 (627/896), or 80% of potentially eligible responders (627/788). In addition, we had 643 anonymous alcohol and drugs questionnaires (643/788, 82% response) all of which were analysed.

Personal characteristics and professional status
Most respondents (533, 85%) were ICU directors and/or consultants (Table 1). Males predominated (523, 83%) in both senior and junior posts. The mean age of the participants was 41.8 yr, the youngest director being 34 yr, and the youngest consultant 30 yr. The majority (543, 87%) were married or cohabiting, about one in four with another doctor. Two thirds of females were married/cohabiting: half (37/68) with another doctor. Almost all the respondents (587, 94%) had trained in anaesthesia; the others were from general medicine, paediatrics, accident and emergency medicine, or surgery. Most worked in a general ICU (534, 85%), the remainder in cardiological or cardiothoracic ICU (36, 5.7%), neurology ICU (25, 4.0%), paediatric ICU (25, 4.0%), and a few in other specialized units. About half (331, 53%) worked in district general hospitals, and half in teaching hospitals (289, 46%), with the remainder in military or private hospitals (3, 0.5%). Four individuals did not state the type of hospital in which they worked.


View this table:
[in this window]
[in a new window]
 
Table 1 Characteristics of survey responders (self report). Note in this and subsequent tables percentages are calculated on basis of valid responses to the item for each group or subgroup: two people did not report their sex
 
Workload assessment
The respondents’ jobs comprised a mean of 3.1 weekly ICU sessions (Table 2). A mean of 68.5 (SD 30.3) h were worked reportedly in the previous week, with a mean of 49.1 (16.9) h per week in the hospital. At home, the respondents reportedly worked a mean of 15.5 (22.2) NHS hours and 7.7 (14.7) non-NHS hours. Directors reported working a mean total of 64.3 (25.4) h and consultants 68.4 (32.5) h. Juniors reported the longest working week for the previous week, a mean of 77.8 (26.0) h.


View this table:
[in this window]
[in a new window]
 
Table 2 Workload indicators reported by responders. *Six not stated whether full or part time
 
The majority of respondents (505, 81%) were in full-time posts. Significantly more females (93/102, 91.2%) than males (412/517, 79.7%) worked in full-time posts ({chi}2=7.48, df=1, P<0.01). Full-timers reported working a mean of 70.5 (30.9) h in total. Those in part-time posts reported working 61.0 (27.3) h which, although significantly less than full-timers (t=2.64, df=435, two-tailed P=0.009), was still high. Those who reportedly worked more than 30 ‘non-NHS hours’ (23/627, 3.7%) were mostly full-time teaching hospital consultants, each responsible for five ICU sessions and 3.5–22 ICU beds. We interpret these ‘non-NHS hours’ to represent academic and/or private work, but this was not clear from the questionnaire. The respondents were responsible for a mean of 7.3 (3.8) ICU beds and 6.3 (7.7) non-ICU beds—mostly high dependency unit beds.

Psychiatric morbidity
For the GHQ, 174/610 (28.5%, 95% CI 24.9–32.1) survey respondents scored above the threshold score of three indicative of psychiatric morbidity (Fig. 2), and 136/610 (22%, 95% CI 19.0–25.6) scored above the more conservative threshold of four. On the SCL-D, 78 (12%, 95% CI 9.9–15.0) scored above the threshold score for depression of 1.5. There were no significant differences by post or sex for either scale (P>0.05, one-way ANOVA or two-tailed t-test) (Table 3). The most common symptoms of depression (highest mean scores for each symptom in the SCL-D) were ‘feeling low in energy’ (mean score 1.76) and ‘worrying too much’ (mean 1.50). Twenty doctors (3.2%, 95% CI 2.0–4.9) reported suicidal ideas (‘thoughts of ending your life’) ‘moderately’ to ‘extremely’ often in the past month. There was no significant difference in the frequency of these suicidal ideas between males and females (P>0.05). There was a moderate correlation between GHQ score and the ‘thoughts of ending your life’ SCL symptom score, particularly for males (males r=0.47, females r=0.43; P<0.001). No significant correlations existed between GHQ score and age, total hours worked in past week, or number of ICU beds. There was for males (but not for females) a weak correlation (r=0.14; P<0.01) between fear of making mistakes and ‘thoughts of ending your life’.



View larger version (7K):
[in this window]
[in a new window]
 
Fig 2 Percentage (95% CIs) of sample with GHQ scores indicative of mental distress in national sample of ICU doctors (morbidity thresholds of >3 and >4).

 

View this table:
[in this window]
[in a new window]
 
Table 3 GHQ-12 and SCL-D score by post and sex (two consultants did not state their sex). Figures in italics are proportions stressed or depressed—expressed as percentage of the number of valid cases in each group or subgroup
 
Job satisfaction
Most respondents were reasonably satisfied with their choice of ICU as a career with 189 (30.4%) extremely satisfied. However, five respondents (0.8%) were extremely dissatisfied. Level of career satisfaction correlated with GHQ score (r=0.30, P<0.001) and depression (r=0.30, P<0.001).

Analysis of individual stressors
An impression of the relative day-to-day perceived importance of each stressor can be represented by the mean product of frequency (f) times stressfulness (s) (i.e. mean (fxs)). But to avoid zero products we calculate (f+1)(s+1) (Table 4). The stressor ‘bed allocation when ICU is full’ scores highly: it is the most stressful factor of all (s=2.40), and reportedly occurs often (f=1.59). At the other end of the spectrum, this group of predominantly male doctors report that sexual harassment is both infrequent (0.08) and not stressful when it occurs (0.17). Talking to distressed relatives’ is the most frequent stressor (f=1.60) but only considered slightly-to-moderately stressful (s=1.46). Similarly, ‘dealing with death’ is perceived as a frequent event (f=1.58) but is not especially stressful (s=1.25). ‘Being over-stretched at times’ is also frequently reported (f=1.46), with the second highest score for stressfulness (s=2.27). In contrast, ‘feeling under-utilized’ is relatively rare (f=0.44) and has low reported stressfulness (s=0.57).


View this table:
[in this window]
[in a new window]
 
Table 4 Mean scores for all perceived stressors
 
As a stressor, ‘making the right decision alone occurs as often as ‘making the right decision as a team (f=1.53 and 1.51, respectively), but it appears to be more stressful to do so alone (s=1.32 compared with 0.93). ‘Fear of making mistakes’, ‘difficult relations with senior colleagues’, ‘sleep deprivation’, and ‘overzealous/inappropriate treatment’ are each perceived to be stressful when they occur (respectively, s=1.71, 1.78, 1.70, and 1.69), but are only occasional events (f=1.09, 1.04, 1.16, and 1.06).

‘Effect of hours of work on personal/family life’ appears to be slightly greater than the ‘effect of stress on family/personal life’, and occurs more frequently (f=1.52 and 1.19), although both are stressful (s=2.02 and 1.93).

Predictors of morbidity
In the multivariate model, three variables (Table 5) were highly significant in predicting GHQ score; ‘effects of stress on personal/family life’, ‘too much responsibility at times’, and ‘lack of recognition of one’s own contribution by others’. All of them were important stressors, but not those expected from Table 4. Similarly, linear regression analysis with SCL-D depression score as the dependent variable (Table 6) showed the same three variables as highly significant plus ‘making the right decision alone’ and ‘keeping up to date with knowledge’.


View this table:
[in this window]
[in a new window]
 
Table 5 Main stressors for distressed doctors (linear regression model) (GHQ score)
 

View this table:
[in this window]
[in a new window]
 
Table 6 Main stressors for depressed doctors (linear regression model) (SCL-D score)
 
Use of alcohol and drugs
Nearly one in five (18%, 95% CI 14.8–20.7) (114/643) reported ‘many irregular’ or ‘many regular’ drinks, but most (53%, 95% CI 48.9–56.6) (339/643) were ‘few regular’ drinkers and just 24 responders (4%, 95% CI 2.4–5.5) were abstinent. Some 14% (95% CI 11.0–16.3) (88/643) smoked and 5% (95% CI 3.2–6.6) (30/643) reportedly smoked and drank heavily. Twelve per cent reported taking drugs to help them sleep (95% CI 9.0–14.0) (74/643), 4% took antidepressants (95% CI 2.9–6.2) (28/643), and 4% anxiolytics (95% CI 2.3–5.3) (23/643). Whilst about half of drugs used to aid sleep were self-prescribed (55%, 95% CI 43.4–67.0) (41/74), most antidepressants were prescribed by another doctor (75%, 95% CI 55.1–89.3) (21/28). In addition, 5% (95% CI 3.3–6.8) (31/643) reported use or past use of drugs such as cannabis for relaxation or recreation.


    Discussion
 Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 References
 
The satisfactory response rate probably reflects both the suitability of the study design, and the doctors’ interest in the topic. It compares favourably with previous studies79 11 30 31 which had response rates between 72 and 80%. The high response rate in this national group gives us confidence that the data are representative. There are, however, two possible sources of selection bias. First, ICS membership may not be entirely representative of the target population of UK intensive care doctors, but we could not identify any other realistically usable source of data to check on this. Second, the responders may not represent the whole ICS membership in terms of psychiatric morbidity, but we had no way of confirming the state of mental health of the non-responders. Nonetheless, we would argue that the level of morbidity we have detected is likely to be a conservative estimate, as it can be argued that non-response is associated with ‘burn out’ and ‘stress’. However, we do recognize that such conjecture needs further investigation, as non-response may be simple disinterest rather than overwork and lack of time.

Our main findings in this national survey of intensive care specialists were that nearly one in three (29%) showed evidence of psychiatric morbidity as identified by the GHQ-12. Some 12% showed clinically important levels of depression identified by the SCL-D. Some 3% reported having suicidal thoughts moderately or extremely often in the previous month, and suicidal ideation correlated with GHQ scores. The doctors’ GHQ and SCL scores showed no relation with age, reported hours worked in the previous week, nor the number of ICU beds for which they were responsible. This lack of association with hours of work has been noted in other studies.23 24 30 31 Although we asked about hours worked, not hours slept in the previous week, sleep deprivation was perceived as an occasional and important stressor. Nonetheless, these doctors reportedly worked long hours, clearly in excess of the subsequently introduced Working Time Regulations,32 an issue in itself of some concern.

Our anonymous alcohol and drugs questionnaire revealed quite high levels of alcohol (mis)use and indication of other substance (mis)use. However, we cannot make a great deal of these data as numbers are relatively small and for methodological and confidentiality reasons we cannot link them to the data from our main datasets. They do, however, suggest that there may be number of maladaptive ‘coping’ strategies in use by these doctors.

Job satisfaction has been suggested previously to be protective against burnout,8 and we found that dissatisfaction with choice of ICU career correlated significantly with both GHQ and SCL scores. The five most frequent and most stressful factors reported were: bed allocation when ICU is full; being over-stretched at times; effect of hours of work on personal/family life; effect of stress on personal/family life; and compromising standards when resources are short. At first sight, all of these stressors seem to be integral to the work of an ICU doctor, and are related to levels of resource for an ICU. Use of paired stressors in the list show that ‘personal/family life’ is reported to be more affected by ‘hours of work’ than by ‘stress’ per se, although both are seen as important. Further down the list, ‘making the right decision alone’ is reported to be more stressful than ‘making the decision as a team. The benefits of teamwork on stress levels are well reported elsewhere.3 However, further statistical analysis showed that a different range of stressors predicted psychiatric morbidity (i.e. GHQ and SCL scores): lack of recognition of one’s own contribution by others; too much responsibility at times; effects of stress on personal/family life; keeping up to date with knowledge; and making the right decision alone. This list is not so much a reflection of resource issues, but may be more related to the responsibilities of ICU doctors and the perceived reputation of the specialty. If these are the stressors of the distressed and depressed doctors, then addressing them is likely to have a greater impact on doctors’ well being than purely focusing on allocation of resources.

The SCL depression sub-scale has also been used by Firth-Cozens, in studies of a cohort of doctors followed since medical school.22 23 Ten years after qualifying, some 18% of those doctors scored above the clinical threshold for depression,30 but only 12% of ICU doctors were depressed. The confidence intervals for the proportion of ICU doctors expressing suicidal ideas (3.2%: 95% CI 2.0–4.9) overlap with other studies; of junior doctors23 (5.9%, 95% CI 2.9–10.6) and of anaesthetists11 (12.2%, 95% CI 4.1–26.2), suggesting they do not differ. Thus, the levels of stress reported here are not very different from those observed in previous studies (Figs 1 and 2), and certainly less than the 44% recently reported for accident and emergency doctors.33 Our findings compare with findings amongst anaesthetists, whose major stressors were effects of work on personal life, overwork, making mistakes, and making decisions.24 Apart from worry about making mistakes, the ICU doctors (who are mostly from an anaesthetic background) rated similar stressors highly. The findings that number of hours worked and number of beds the doctor is responsible for relate to morbidity have been noted before.3 34 The lack of sex difference in GHQ scores has also been reported elsewhere.23 We cannot, of course, infer causality in any relationship between occupational stressors and mental health. The more depressed or distressed doctor may well give greater weighting to his or her stressors, and feel more dissatisfied with choice of career. Previous surveys of medical students and hospital doctors using the GHQ-125 8 22 23 have also shown high percentages (26–30%) of doctors scoring above a GHQ-12 threshold of 3 (Fig. 1). Surveys using GHQ-287 11 have shown 21–46% scoring above the GHQ-28 threshold of >5 for psychiatric morbidity.

Doctors, it seems, have more problems than the average British worker (17.8% for the 1995 Household Panel Survey; www.irc.essex.ac.uk/bhps).35 However, some authors dispute that stress and mental health problems in health care workers are any higher than the general population, and ascribe raised stress levels to questionnaire bias, especially in doctors.36 37 Such an explanation, however, does not easily account for differences within the medical profession as a whole.

Whilst anecdotally, ICU is a high-stress speciality, it seems that measuring distress alone may be missing the point, as our regression modelling revealed. The level of mental health problems we found is not apparently related to the ICU doctors’ characteristics, nor to the type of unit, nor long hours of work. Feeling that the responsibility is too great, lack of peer recognition, worry about compromising standards, effects of stress on home life, distressed relatives, and the pressure to keep up to date are clearly issues in need of resolution. We may need to develop teamwork to modify these stressors,3 but we must recognize that this also implies a need for resources to develop teams.

At an average age of 42 yr these doctors have another 20 yr or so of professional life ahead. Over a prolonged period, stress could lead to increased risk of health problems, both mental and physical. Consideration needs to be given to what aspects of the occupational stressors are amenable to change, and what practical steps, if any, can be taken to identify and support vulnerable individuals at an early stage. It is logical to address factors which may contribute to attrition rates of consultants, as they are expensive to train (£232 000) and each lost working year of a consultant costs the NHS £30 000 in terms of annuitized training costs over his or her expected working life.38 A rational strategy for the NHS would be to provide employment environments in which doctors can practice effectively without compromising their own health.

Strategies to reduce harmful effects on doctors’ health should include reduction of frequency and intensity of identified stressors, early detection of problems and maladaptive coping (e.g. alcohol or drug use), and effective medical treatment and rehabilitation of the sick doctor. The principles of management of stress in healthcare workers in general apply to ICU. However, reducing stressors in ICU may in practice be difficult. In the meantime, individual distressed doctors need to have improved access to services, such as the BMA Stress Counselling Service,39 and the National Counselling Service for Sick Doctors (www. ncssd.org.uk). Prevention, in the form of stress management has begun in some trusts already,40 and is worthy of future development.


    Acknowledgements
 
We thank the ICS for their help with supporting the study throughout. Our thanks to the doctors who gave their time to complete our questionnaire, to Julian Lipscombe for his help with data preparation, to Brian Thom who assisted Sarah Shore with the regression analysis and to Malcolm Campbell for further data checking and analysis. Thanks are also due to Prof Sydney Brandon for his encouragement and helpful comments on this paper prior to his death. Funding: the Intensive Care Society, the John Farman Intensive Care Unit and the Occupational Health Department, Addenbrooke’s Hospital NHS Trust all helped to fund the survey. Dr Todd’s post was funded by NHS Eastern Region R&D.


    References
 Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 References
 
1 Robins S, Cooper CL, Mayers W. Stress and the Medical Profession. London: BMA, 1992

2 Smith R. All doctors are problem doctors. Br Med J 1997; 314: 841–2[Free Full Text]

3 Firth-Cozens J. Hours, sleep, teamwork and stress. Br Med J 1998; 317: 1335–6[Free Full Text]

4 Berger A. Surviving (and even enjoying) medicine. Br Med J Classified, 29 April 2000; 2–3

5 Borrill CS, Wall TD, West MA, et al. Mental Health of the Workforce in NHS Trusts. Phase 1. Final report. March, 1996

6 Kapur N, Borrill C, Stride C. Psychological morbidity and job satisfaction in hospital consultants and junior doctors: a multicentre cross-sectional survey. Br Med J 1999; 317: 511–12[Free Full Text]

7 Caplan RP. Stress, anxiety and depression in hospital consultants, general practitioners and senior health service managers. Br Med J 1994; 309: 1261–3[Abstract/Free Full Text]

8 Ramirez AJ, Graham J, Richards MA, Cull A, Gregory WM. Mental health of hospital consultants: the effects of stress and satisfaction at work. Lancet 1996; 347: 724–8[ISI][Medline]

9 Heyworth J, Whitley TW, Allison EJ jr, Revicki DA. Correlates of work-related stress among consultants and senior registrars in accident and emergency medicine. Arch Emerg Med 1993; 10: 271–8[ISI][Medline]

10 Agius RM, Blenkin H, Deary IJ, Zealley HE, Wood RA. Survey of perceived stress and work demands of consultant doctors. Occup Environ Med 1996; 53: 217–24[Abstract]

11 Blenkin H, Deary IJ, Wood RA, Zeally HE, Agius RM. Stress in NHS consultants. Br Med J 1995; 310: 534[Free Full Text]

12 Ryan DW. Providing intensive care. Br Med J 1996; 312: 654[Free Full Text]

13 Abramson NS, Wald KS, Grenvik ANA, Robinson D, Snyder JV. Adverse occurrences in Intensive Care Units. JAMA 0000; 244: 1582–4[Abstract]

14 Carnall D. UK reviews intensive care and emergency services. Br Med J 1996; 312: 655[Free Full Text]

15 McNamee R, Keen RI, Corkhill CM. Morbidity and early retirement among anaesthetists and other specialists. Anaesthesia 1987; 42: 133–40[ISI][Medline]

16 Redfern N. Morbidity among anaesthetists. Br J Hosp Med 1990; 43: 377–81[ISI][Medline]

17 Milner QJW, Zeigler ESM. Early death amongst anaesthetists. Anaesthesia 1997; 52: 797–8[ISI][Medline]

18 Seeley HF. The practice of anaesthesia—a stressor for the middle-aged? Anaesthesia 1996; 51: 571–4[ISI][Medline]

19 Neil HA, Fairer JG, Coleman MP, Thurston A, Vessey MP. Mortality among male anaesthetists in the UK. Br Med J 1987; 295: 360–2[ISI][Medline]

20 Helliwell PJ. Suicide amongst anaesthetists-in-training. Anaesthesia 1983; 38: 1097[ISI][Medline]

21 Goldberg D, Williams P. A User’s Guide to the General Health Questionnaire, 1st Edn. Windsor: NFER-NELSON Publishing, 1988

22 Firth-Cozens J. Levels and sources of stress in medical students. Br Med J 1986; 292: 1177–80[ISI][Medline]

23 Firth-Cozens J. Emotional distress in junior house officers. Br Med J 1987; 295: 533–6[ISI][Medline]

24 Firth-Cozens J. Levels and Sources of Stress in Anaesthetists. Report to Association of Anaesthesists of Great Britain and Ireland. London: Association of Anaesthetists of Great Britain and Ireland, 1989.

25 Derogatis LR, Lipman RS, Covi MD. SCL-90: an outpatient psychiatric scale—preliminary report. Psychopharm Bull 1973; 9: 13–20[Medline]

26 Fowler FJ. Survey Research Methods. London: Sage, 1993

27 Firth-Cozens J. Morrison LA, Sources of stress and ways of coping in junior house officers. Stress Med 1989; 5: 121–6[ISI]

28 Banks MH, Clegg CW, Jackson PR, Kemp JN, Stafford EM, Wall TD. The use of the general health questionnaire as an indicator of mental health in occupational settings. J Occup Psychol 1980; 53: 187–94.[ISI]

29 Gardner MJ, Gardner SB, Winter PD. Confidence Interval Analysis Microcomputer Program Version 1.2. BMJ Publishing Group, 1992.

30 Firth-Cozens J. Predicting stress in general practitioners: 10-year follow-up postal survey. Br Med J 1997; 315: 34–5[Free Full Text]

31 Firth-Cozens J. Individual and organisational predictors of depression in general practitioners. Br J Gen Pract 1998; 48: 1647–51[ISI][Medline]

32 Byers S, The Secretary of State for Trade and Industry. Working Time Regulations 1999: Statutory Instrument No. 3372. London: HMSO Publications, 1999

33 Burbeck R, Coomber S, Robinson S, Todd C. Occupational stress in consultants in accident and emergency medicine: a national survey of levels of stress at work. Emerg Med J 2002; 19: 234–8[Abstract/Free Full Text]

34 Goodfellow A, Varnham R, Rees D, Shelley MP. Staff stress on the intensive care unit: a comparison of doctors and nurses. Anaesthesia 1997; 52: 1037–41[ISI][Medline]

35 British Household Panel Survey 1993–4. British Household Panel Survey, UK Data Archive. Colchester: Institute for Social and Economic Research, 2001

36 Weinberg A, Creed F. Stress and psychiatric disorder in healthcare professionals and hospital staff. Lancet 2000; 355: 533–7[ISI][Medline]

37 McManus IC, Winder BC, Gordon D. Are UK doctors particularly stressed? Lancet 1999; 354: 1358–9[ISI][Medline]

38 Netten A, Knight J, Dennett J, Cooley R, Slight A. A ‘Ready Reckoner’ for Staff Costs in the NHS. Canterbury: Personal Social Services Research Unit, Vol. 1, 2000

39 BMA Stress Counselling Service (analysis of nearly 800 calls made between 9 April launch and 21 June). BMJ News 1996; 6 July; 313

40 Davies RH. Junior doctors and stress. Lancet 1998; 352: 1780