Intended for healthcare professionals

General Practice

Population based study of fatigue and psychological distress

BMJ 1994; 308 doi: (Published 19 March 1994) Cite this as: BMJ 1994;308:763
  1. T Pawlikowska,
  2. T Chalder,
  3. S R Hirsch,
  4. P Wallace,
  5. D J M Wright,
  6. S C Wessely
  1. Department of General Practice, St Mary's Hospital Medical School, London
  2. Department of Psychological Medicine, King's College School of Medicine and Dentistry, London SE5 9RS
  3. Departments of Medical Microbiology and Psychiatry, Charing Cross Hospital Medical School, London
  4. Department of General Practice, Royal Free Hospital School of Medicine, London NW3
  1. Correspondence to Dr Wessely.


    Abstract Objectives: To determine the prevalence of fatigue in the general population and the factors associated with fatigue.

    Design: Postal survey.

    Setting: Six general practices in southern England.

    Subjects: 31 651 men and women aged 18-45 years registered with the practices.

    Main outcome measures: Responses to the 12 item general health questionnaire and a fatigue questionnaire which included self reported measures of duration, severity, and causes of fatigue. Results - 15 283 valid questionnaires were returned, giving a response rate of 48.3%, (64% after adjustment for inaccuracies in the practice registers). 2798 (18.3%) of respondents reported substantial fatigue lasting six months or longer. Fatigue and psychological morbidity were moderately correlated (r=0.62). Women were more likely to complain of fatigue than men, even after adjustment for psychological distress. The commonest cited reasons for fatigue were psychosocial (40% of patients). Of 2798 patients with excessive tiredness, only 38 (1.4%) attributed this to the chronic fatigue syndrome.

    Conclusion: Fatigue is distributed as a continuous variable in the community and is closely associated with psychological morbidity.

    Practice implications

    • Practice implications

    • Fatigue is a common symptom in the community

    • In this community survey symptoms of fatigue fitted a normal distribution

    • Most patients with excessive fatigue blamed psychosocial factors for their symptoms. The chronic fatigue syndrome was mentioned by only 1%

    • Fatigue and psychological morbidity were related and the association increased in those with prolonged or severe fatigue or with other symptoms such as muscle pain


    Community surveys have shown that fatigue is common in developed countries. A British survey of those attending general practice found that 10.2% of men and 10.6% of women had had substantial fatigue for over a month.1 Despite its prevalence, however, surprisingly few data exist on the demographic and social associations of fatigue.2

    In recent years fatigue has attracted renewed attention, largely because of the prominence given to the chronic fatigue syndrome, also known as postviral fatigue or myalgic encephalomyelitis. However, controversy exists concerning not only the relative contributions of infective and psychological risk factors but the classification of the syndrome itself. These disputes cannot be resolved without sound community studies.3

    Nearly all studies of chronic fatigue syndrome have been conducted in tertiary referral settings, where the typical patient is a young adult, usually a woman, with a professional job. Most patients trace their illness to an infective trigger but also fulfil criteria for a psychiatric disorder.*RF 4-7* The infective and psychiatric characteristics may, however, be the result of referral patterns and illness behaviour and not intrinsically related to the chronic fatigue syndrome.

    We conducted a large community survey to determine the level of fatigue and psychiatric morbidity in a population, to explore the association between fatigue and psychological distress, and to identify what fatigue is commonly attributed to.

    Subjects and methods

    We sent questionnaires to all patients aged 18 to 45 years registered with selected practices. We recruited six practices to the study, three from London and three from rural or semirural settings, two of which belonged to the Medical Research Council general practice research framework.

    Two practices (subsequently referred to as practice 1) were working from the same health centre in south London in a mixed urban community with a large amount of temporary accommodation. Practice 2 was an inner city London practice where most patients were socially deprived. The third practice was located on the Surrey-Hampshire border with patients predominantly in socioeconomic classes II and III. Practice 4 was in an urban area of a south coast port, and the last practice was in a Somerset village, with a static close knit community and many stable families. The total number of patients aged 18-45 registered with the practices was 31 651 (15 222 men, 16 429 women).

    Fatigue was assessed with a self reported questionnaire (the fatigue questionnaire), which was developed for a hospital study of the chronic fatigue syndrome7,8 and was refined during a validation study conducted in primary care.9 It consists of 11 items covering the physical and mental aspects of fatigue. Additional questions concern the duration of fatigue, the percentage of time during the day the respondent felt tired, and two questions on muscle pain at rest and after exercise. The response to the questions on muscle pain were summed to give a total score. Respondents were also asked “why do you think you are feeling tired?” We also included the 12 item general health questionnaire.10 The general health questionnaire is a well validated measure of psychological morbidity; those scoring above the predefined cut off have an increased chance of suffering from a psychiatric disorder. Both questionnaires use similar responses, usually of the form not at all, same as usual, more than usual, and much more than usual.

    Two systems of scoring were used for both questionnaires, traditional (0,0,1,1) and Likert (0,1,2,3). Responses to the presence or absence of muscle pain at rest and on exercise were scored 0 or 1.

    We posted the questionnaires in November 1991. Non-responders were sent a second questionnaire four weeks later. In four practices a random sample of 100 persistent non-responders was followed up by telephone or personal visits and the reasons for non-response determined. Addresses were checked with the electoral register, and telephone numbers were taken from notes within the practice or telephone directories. In one of the four practices it was considered inadvisable for the research nurse to make home visits. The last practice did not allow any follow up.

    Statistical analysis

    Statistical analysis was performed with the SPSS program. We applied parametric statistics since with large numbers the distribution of responses to both questionnaires follows a normal distribution if Likert scoring is used.


    We received 15 283 completed replies, 6532 from men and 8757 from women. This was a response rate of 48.3%. Table I shows the response rate for each practice. The lowest rates were obtained in the two inner London practices (43% and 37.5% respectively). A total of 387 incomplete questionnaires were returned and were discarded from the analysis. As others have found in similar surveys,11 non- responders were more likely to be men (53%) and were slightly younger than responders (mean age 30.8 (95% confidence interval 30.7 to 30.9) years for non-responders, 32.4 (32.3 to 32.5) years for responders; F=436.66, df=1, P<0.001).


    Response rates in participating practices

    View this table:

    The main reason for non-response was change of address. In the four practices surveyed 19%, 39%, 46%, and 69% of non-responders did not receive the questionnaire because they have moved house. The differences between the practices reflect the greater social mobility in the inner city practices. Adjustment for those with wrong addresses or who had moved and extrapolation of this data to the entire sample (arbitrarily assuming an error rate of 19% for the fifth practice in which no data could be obtained) gave an adjusted response rate of 64%.

    Findings of questionnaires

    Figures 1 and 2 show the frequency distribution of fatigue questionnaire and general health questionnaire scores. In all, 5799 (38%) patients had a score above the cut off established in a primary care validation study for substantial fatigue,9 and 5621 (36.7%) scored above 3/4 (traditional scoring system) in the general health questionnaire, which is the cut off for psychological disorder. A total of 2798 (18.3%) were substantially fatigued for six months or longer. Scores in the fatigue and general health questionnaires were moderately correlated (r=0.62, 95% confidence interval 0.61 to 0.63) with a linear trend between fatigue score and mean general health questionnaire score (fig 3). Mental and physical fatigue scores were also significantly correlated: (men r=0.54, 0.52 to 0.56; women r=0.54, 0.53 to 0.55).

    FIG 1
    FIG 1

    Frequency distribution of general health questionnaire score (Likert scale)

    FIG 2
    FIG 2

    Frequency distribution of total fatigue score; FIG 3 - Mean and 95% confidence interval score for general health questionnaire according to fatigue score.

    Age was very weakly correlated with scores on both the fatigue (r=0.08, P<0.01) and the general health questionnaires (r=0.03; P<0.01). Fatigue scores in women increased with age but only slightly. The mean fatigue score was 24.1 (95% confidence interval 24.0 to 24.2) for men and 25.2 (95% 25.1 to 25.3) for women the difference in means being 1.1 (0.96 to 1.24). The mean general health questionnaire score was 24.7 (24.6 to 24.9) for men and 26.2 (26.1 to 26.4) for women, the difference in means being 1.5 (1.3 to 1.7). Because of sample size these sex differences were highly significant (independent t test P<0.0001) but are probably not clinically important.

    We explored the relation between sex, fatigue, and general health questionnaire score with analysis of variance. Women had significantly more fatigue than men (F=45.41, df=1, P<0.001). Psychiatric morbidity was a confounding variable since community surveys usually report that women have higher psychological morbidity than men, as confirmed in this study. Stratifying by psychological morbidity on the basis of general health questionnaire caseness did not remove the excess of fatigue in women. This was confirmed by an analysis of variance using the general health questionnaire score as a covariable. The effect of sex remained significant (F=45.51 P=0.001).

    We asked all patients the reasons why they were tired (table II). Most patients gave one reason but where there were two or more the first was taken to be the most important. The commonest attribution was psychosocial, such as work, family, and lifestyle. The next most popular category was psychological distress (usually anxiety or depression), followed by physical causes (for example, surgery, anemia). The last major category consisted of “other” causes, which were mainly environmental (including the weather and pollution). Thirty eight people described themselves as having the chronic fatigue syndrome. More detailed follow up of these patients will be reported elsewhere. All belonged to social classes I, II or IIIa and 30 were women. Patients with a self diagnosis of the chronic fatigue syndrome had the highest mean fatigue scores, followed by those who attributed their fatigue to pregnancy, physical, and psychological causes respectively. Psychosocial attributions were associated with the lowest mean fatigue scores.


    Causes that fatigue was attributed to by patients

    View this table:

    A total of 2107 (13.8%) subjects complained of muscle pain at rest, and 3286 (21.5%) complained of muscle pain after exercise. There was a linear relation between general health questionnaire score and proportion complaining of either muscle pain at rest or after exercise (X2 test for trend, P<0.001). With a cut off of 4 for the general health questionnaire 19% (2387) of non cases complained of muscle pain at rest or after exercise compared with 40% (1088) of cases (odds ratio=2.8; 95% confidence interval 2.7 to 2.9).

    Table III shows the effect of using some of the current criteria for the chronic fatigue syndrome12 on the prevalence of fatigue and its relation to psychological morbidity. The criteria include the effect of duration, proportion of time fatigued, and the presence of myalgia. Each restriction progressively decreased the prevalence of arbitrarily defined fatigue states but increased both the proportion of cases on the general health questionnaire and the female:male ratio.


    Effect of different criteria for chronic fatigue syndrome on prevalence of fatigue syndrome and its relation to psychological morbidity

    View this table:


    Studies of the chronic fatigue syndrome have been based in either general practice or tertiary referral centres. We conducted a large community survey of fatigue in which the participating practices were chosen to reflect a wide spectrum of social and geographicalsettings.

    The overall response rate of 48.3% was disappointing, and was largely due to the low rates obtained in the inner London practices. However, others have shown, again in south London, that the family health services authority registers are only 75% accurate because of the young and relatively mobile population in the inner city.13 In our study 109 out of the 200 non-responders randomly followed up in the London practices had moved or had non-existent addresses. Hence only three quarters of the questionnaires for the London sample reached their destination, reducing the true sample size. Non-responders who received the questionnaire almost certainly differed in both fatigue and psychological distress, introducing a possible response bias, but a non- response bias due to change of address seems less likely. Overall the adjusted response rate, although far from ideal, is comparable with that in other community postal surveys.11,14

    The results support the concept of a continuum of fatiguability, as already suggested by data from a primary care study in the United Kingdom and a community survey in the United States.12 The chronic fatigue syndrome, defined only by symptoms, disability, and duration, may represent a morbid excess of fatigue rather than a discrete entity, just as high blood pressure and alcohol consumption are morbid ends of normal spectrums. Hence the definition may have arisen as a result of referral patterns to specialists.

    In our survey 18% of respondents complained of excessive fatigue for six months or longer, confirming the importance of fatigue as a symptom in the community. The demographic associations found in this study agree with previous data.2 Women were more fatigued than men. Women also had more psychological disorder than men, but this finding did not explain the excess of fatigue in women. The relative risk of fatigue in women compared with men in our study was 1.3, identical to the relative risk found in a single general practice in this country,1 and in a single general practice in Canada.15 The failure of both groups to find significant sex differences may reflect lack of power.

    Association with psychological morbidity

    We found an association between the number and severity of fatigue symptoms and psychological morbidity (fig 3). Two primary care studies have reported a close association between the symptoms of fatigue and either psychological morbidity16 or psychiatric diagnoses.17 Numerous tertiary care studies have found associations between chronic fatigue syndrome and psychiatric diagnoses.18 Muscle pain was also related to psychological morbidity, confirming the associations noted in a recent general practice study from Lancashire.11 We conclude not that fatigue is caused by psychological morbidity, but that the two conditions overlap. The concordance between fatigue and psychological morbidity, and between the chronic fatigue syndrome and psychiatric disorder, is inevitable given the similarities of the criteria and measures used to define them.

    Although this is not a study of the chronic fatigue syndrome, it does give some information about it. Firstly, self diagnosis of the chronic fatigue syndrome is extremely uncommon even among those with fatigue, accounting for only 1% of those with significant fatigue and 0.2% of the entire sample. Self diagnosis of the syndrome was associated with being female, high social class, and high levels of both fatigue and psychological morbidity.

    Secondly, we measured two factors included in the current consensus criteria for chronic fatigue syndrome (duration of fatigue and percentage of time fatigued),11 as well as the presence of myalgia, which is thought to be characteristic of the syndrome in Britain. Application of each of these criteria reduced the prevalence of a fatigue syndrome but increased the association with psychological morbidity. Thus 38% of the sample complained of excessive fatigue, but in only 1% had this lasted for more than six months, been experienced all the time, and been associated with myalgia. Among this group 90% were probable cases of psychiatric disorder, and women were over represented. This pattern confirms that increasing the number of restrictions in the definition of chronic fatigue syndrome does not reduce the association with psychological morbidity but has the opposite effect.19 Future studies will look at the development of post infectious fatigue in this cohort.

    We thank the study nurses Melanie Brown, Julie Dennison, Sue Fox, Brenda Hinton, Vivien Hobbs, Margerie Peers, and Dawn Baker and the general practitioners and staff of the participating health centres Jenner Health Centre, London SE23; Hurley Clinic, London SE5; Yateley Medical Centre, Yateley, Surrey: Rosemary Medical Centre, Poole, Dorset, and Mendip Country Practice, Somerset. The last two practices are members of the Medical Research Council general practice research framework. We are also grateful for the help of Peter Turk, Charing Cross Hospital Medical School, and David Blackett with computing. The study was supported by Linbury Trust, with additional support from Private Patients Plan and British United Provident Association.