Informal care and psychiatric morbidity

Journal of Public Health Medicine Vol. 20, No. 2, pp. 180-185 Printed in Great Britain Informal care and psychiatric morbidity Stephen Horsley, Steve Barrow, Nick Gent and John Astbury Abstract Background We aimed to examine and quantify the relationship between psychiatric morbidity and the provision of informal care in the community. Methods The study involved a comparison of carers and noncarers in a mixed urban and rural community (Morecambe Bay Health Authority). Data were collected by postal survey for 4550 adults; 10.9 per cent of respondents were identified as carers. Subjects were selected by quasi-random methods from the Family Hearth Services Authority (FHSA) registers. Potential psychiatric morbidity was defined as three or more symptoms on a standardized measure, the General Health Questionnaire (12-item version). Results The prevalence of morbidity was significantly higher in people who care for others in their own homes, even after adjustment for other known risk factors for psychiatric morbidity (odds ratio 1.51, 95 per cent confidence interval 1.11-2.05). In contrast, there was no significant relationship between morbidity and care outside the home in these data. Conclusion Health Authorities need to review support for carers and to consider ways to improve monitoring. Keywords: informal care, psychiatric morbidity Introduction In Great Britain it is estimated that six million people provide care on an informal basis for sick, disabled and elderly people in the community. 1 Moreover, these numbers appear to be increasing, in particular where carers support others who do not live with them. 2 The trend to informal caring has been encouraged by central government policies, although recent legislation has recognized that caring may be burdensome, not least in terms of mental distress. 3 Unfortunately, there has been little research on the stress associated with the provision of care; moreover, the evidence is conflicting. Early studies have indicated an increased prevalence of mental distress; however, these have used selected samples of carers from lists of voluntary organizations or health agencies. 4 " 3 More recent community-based studies have tended to suggest that caring in itself is not ariskfactor. However, these have sometimes involved restricted samples and have focused on care within one particular type of recipient. 6 " 8 In this paper we use a large community survey to examine two questions. Is caring in general in the community associated with an increased prevalence of psychiatric morbidity? Does any such association vary between care which is provided in the carer's own home and that which is provided elsewhere? Methods Sample In autumn 1995 we undertook a survey of health status and behaviours in Morecambe Bay Health District Questionnaires were posted to 6400 adults aged 18 and over; there was no upper age limit. The target of 6400 was chosen with a view to achieving 4200respondents;we had anticipated aresponserate of 66 per cent on the basis of an earlier survey in the district. 9 The total of 4200 responses was required because we wanted to be 95 per cent confident that any prevalence estimates from the survey would be within 1.5 percentage points of 'true' district values. In the event, a total of 4550 subjects was achieved after two postal reminders; this represents a response rate of 71 per cent (or 74 per cent if cases with invalid addresses are discounted). The project was approved by both district Ethics Committees. Potential respondents were selected by systematic sampling from a random startpoint in the Family Health Services Authority (FHSA) register. (We had to rely on existing FHSA sampling procedures which precluded access to a completely random sample.) Measures The questionnaire contained 50 items concerning health status, disability, risk behaviours, views of the health service, social support and demography. Most of the questions had been taken from health needs surveys elsewhere, such as the national Health and Lifestyle Survey. 10 Other items had been developed Public Health and Health Professional Development Unit, University of Lancaster LAI 4YX. Stephen Horsley, Professor North West Health Research Unit, Gateway House, Piccadilly South, Manchester M60 7LP. Steve Barrow, Epidemiologist Morecambe Bay Health Authority, Tenterfield, Brigsteer Road, Kendal LA9 5EA. Nick Gent, Director of Public Health John Astbury, Consultant in Public Health Address correspondence to Steve Barrow. Oxford University Press 1998

INFORMAL CARE AND PSYCHIATRIC MORBIDITY 181 locally and used in earlier district surveys. 9 Only the questions on caring itself were new. These were developed after consultation with local carer groups. All data were self-reported. Several items were of particular interest for the present study. First,respondentswere asked if they 'regularly cared for a person with a long-standing illness or disability'. Those defined as 'carers' on this basis were also asked for details of the number and location (home or elsewhere) of the people in their care. Second, respondents were assessed for (potential) psychiatric morbidity; the questionnaire contained the 12-item General Health Questionnaire (GHQ), which has been extensively validated and widely used for the screening of depression and anxiety in community samples. 11 ' 12 Eight other items from the survey were relevant here. These concerned social factors which have previously been identified as predictors of psychiatric morbidity. 13 " 15 Such factors needed to be considered in assessing the independent association of caring and morbidity. Five of the items were demographic, these being gender, age, employment status ('in work', 'not in work', 'student', 'retired'), living arrangements ('not alone', 'alone'), and housing tenure ('owner occupier', 'renting accommodation'). Two factors concerned social networks, respondents being asked (a) whether or not they had seen a friend 'to chat to recently', and (b) whether or not they had someone to 'discuss any problems with'. These were used as indicators respectively of social contact and perceived support. The final social factor concerned long-standing illness ('none', 'yes, but not limiting', 'limiting'). Analysis The standard 0/0/1/1 system was used to score the GHQ before conversion to a categoric indicator of 'potential psychiatric morbidity' (a score of three or more, the recommended threshold), and 'no morbidity' (two or less). All variables were categoric, except for age, which was converted to categories for exploration of associations. The categories were 18-39, 40-64, and, a grouping which had been used in earlier district surveys. 9 The data were analysed using SPSS 6.1 and GLIM 3.77. The characteristics of caring were first described in terms of basic statistics. These included tests of association in which a probability (/?) of less than 0.05 under the null hypothesis was taken to suggest a 'statistically significant' difference. A logistic regression analysis was then undertaken to examine the associations between caring and psychiatric morbidity. All variables entered were categoric. Associations were described in terms of odds ratios (ORs) along with their 95 per cent confidence intervals (CIs). Caring was initially considered on a univariate basis, the odds ratios being adjusted only for age and sex. The relation was then examined by a multivariate analysis in which caring was included along with the other eight predictors. This allowed the association between caring and morbidity to be seen after adjustment for all of the other predictors. It also allowed 'population attributable fractions' to be calculated for each significant predictor; these indicate the reduction in morbidity which would result if the predictor in question had a direct aetiological effect and was 'eliminated'. Of course, it is not possible to confirm such effects in a study of this sort; nevertheless, the attributable fraction gives an impression of the potential importance of particular factors in contributing to morbidity in the population. Before computing attributable fractions, we reweighted the prevalence figures in the respondent group to compensate for unevenness in the response by age and gender. The analysis at each stage included all cases for whom the relevant information was available, with consequent variations in numbers analysed. The univariate analysis above was thus based on 4269 cases; the number of respondents in the multivariate analysis was 3867. Results Structure of the respondent group As in most health surveys, women were slightly overrepresented, with 58 per cent of therespondentgroup compared with 52 per cent in the district population. The opposite situation pertained to men, who made up 42 per cent of the respondent group compared with 48 per cent of the population. Details of representativeness by age and sex are given in Table 1. Care and psychiatric morbidity amongst respondents The question on informal care was answered by 4385 respondents. Of these, 478 (10.9 per cent) identified themselves Table 1 Representativeness of the respondent group Women Total women Men Total men All Grand total Respondents 1995 n 869 1030 699 2598 579 796 512 1887 1448 1826 1211 4485 % 33.4 39.7 26.9 30.7 42.2 27.1 32.3 40.6 27.0 100 00 OPCS population estimate 1995 n 45180 46340 34440 125960 46510 45760 23080 115350 91690 92100 57 520 241310 % 35.9 36.8 27.3 40.3 39.7 20.0 38.0 38.2 28.8

182 JOURNAL OF PUBLIC HEALTH MEDICINE as carers. Just over a third (38.9 per cent) provided help outside men to provide help entirely outside their own households; 44.2 their homes; however, the majority (61.0 per cent) supported per cent of female carers gave this kind of support compared someone with whom they lived (including 4.0 per cent of carers with 30.6 per cent of males (p<0.01). who gave help both at home and elsewhere). Most carers in the Carers tended to be older than non-carers, with mean ages of sample (83.6 per cent) supported one person. Just over a tenth 54.9 years and 49.4 years, respectively (p<0.01, Kruskalof carers (10.8 per cent) provided help for two people, and 5.5 Wallis test). However, there was no difference in age between per cent supported three or more. carers 'at home' and 'elsewhere' (p = 0.22). Women were more likely to provide care than men, The relation between gender and care varied at different involvement being 11.9 per cent and 9.3 per cent, respectively ages (Fig. 1). Of people aged between 40 and 64, women were (p <0.01, x test). Women carers were also more likely than far more likely to provide care (p = 0.0002, x test). There was 03 1 18-39 40-64 65+ 17.2 Females D Males Figure 1 Percentage of males and females of each age band who provide informal care.

INFORMAL CARE AND PSYCHIATRIC MORBIDITY 183 no significant gender difference for people above and below that age banding (, p = 0.09;, p = 0.48). Just over a quarter (25.1 per cent) of non-carers were above the threshold of three or more symptoms on the GHQ. The prevalence figure for carers at home was 32.4 per cent; for carers elsewhere it was 27.7 per cent. Using logistic regression to adjust for age and sex, the equivalent odds ratio for carers at home was 1.52 (95 per cent CI 1.16-1.99). The odds ratio for caring elsewhere was 1.12 (95 per cent CI 0.80-1.59). On this 'univariate' basis, care at home is a significant predictor of increased morbidity; care elsewhere is not A sensitivity analysis was conducted using an alternative threshold of four or more symptoms on the GHQ. A similar picture emerged, the odds ratios for caring at home and caring Table 2 Odds* of psychiatric morbidity - caring and other factors Caring None Home Away from home Sex Female Male Age Employment In work Not in work Student Retired Living arrangements Not alone Alone Tenure Owner Renting Social contact Yes No Perceived support Yes No Long-standing illness No Yes, not limiting Yes, limiting Odds ratio 1.51 1.28 0.75 0.74 0.77 1.38 1.27 0.67 1.26 1.09 2.09 3.67 1.34 3.58 "Each predictor adjusted for all other predictors. 95% a Lower 1.11 0.87 0.64 0.61 0.53 1.12 0.86 0.48 0.89 1.59 2.81 1.07 2.95 Upper 2.05 1.90 0.88 0.90 1.14 1.70 1.89 0.95 1.57 1.33 2.75 4.80 1.68 4.33 elsewhere being 1.38 (95 per cent CI 1.03-1.84) and 1.05 (95 per cent CI 0.72-1.55), respectively. Table 2 gives odds ratios and confidence intervals for the different levels of care in a multivariate analysis of all our anticipated predictors of psychiatric morbidity. Caring at home continues to be an independent predictor. In fact, there appears to be little confounding between caring and the other predictors; both the univariate and the multivariate analysis give similar estimates (ORs of 1.52 and 1.51, respectively). Other significant predictors of increased morbidity were 'not being in work' (OR 1.38), living alone (1.26), (lack of) social contact (2.09), perceived (lack of) support (3.67), and long-standing illness (not limiting 1.34, limiting 3.58). Likelihood ratio tests were conducted for the possibility of interaction effects involving (1) caring and agegroup and (2) caring and gender. No significant effects were found. Scaling up to the district population It was important to consider the representativeness of the respondent group in assessing the likely impact of caring and our other predictors in the wider district population. Table 1 indicated that there had been some unevenness of response by age and gender groups. The categories of that table were therefore used to reweight the prevalence figures in the respondent group to reflect the age-sex distribution of the wider population. Table 3 compares salient results before and after this reweighting. Differences are virtually negligible. The reweighted figures were used in computing the population attributable fractions. These are shown in Table 4, which gives Table 3 Effects of reweighting results for respondents to reflect age and sex distribution of district population % caring Women 11.9 Men 9.3 All 10.9 % carers who care at home 61.0 % carers supporting 1 person 83.6 2 people 10.8 3+ people 5.5 Mean age (years) Non-carers 49.4 Carers 54.9 % above GHQ threshold Non-carers 25.1 Carers at home 32.4 Carers away 27.7 Respondent group Reweighted figures 11.7 8.6 10.2 62.0 83.4 11.2 5.4 47.7 53.9 25.1 33.1 27.2

184 JOURNAL OF PUBLIC HEALTH MEDICINE Table 4 Population attributable fractions - significant predictors of increased morbidity Risk factor Prevalence of factor (%) Attributable fraction Caring at home Not being in work Living alone Lack of social contact Perceived lack of support Long-standing illness - not limiting Long-standing illness - limiting 6.3 16.8 15.8 7.9 7.8 16.1 24.6 0.03 0.06 0.04 0.08 0.17 0.05 0.39 details for the factors which remained significant in the multivariate analysis. Discussion Methodological issues The present study investigates the prevalence of potential psychiatric morbidity amongst people who provide informal care in the community. It is based on a sample of 6400 adults in a health authority with a variety of urban and rural settings. This sample is larger and more geographically heterogeneous than those which have characterized recent studies of this topic. The response rate was high for a postal survey (74 per cent). Representativeness was acceptable in terms of broad age and sex categories, and areweightingby these categories made only minor differences to our prevalence estimates. Of course, there are other ways besides age and sex in which representativeness might be compromised. Not least of these was our reliance on the FHSA register, from which important groups are likely to be omitted (e.g. the homeless). The dependent variable was a score of three or more symptoms on the GHQ. This is a screening instrument and high scores do not in themselves confirm morbidity. Nevertheless, the instrument has been extensively validated as a measure of potential 'caseness'; it has been used before in prevalence surveys. Of course the GHQ is a self-report measure, as were all the items in the study. This creates inevitable limitations in the scope of the data. In particular, it was not possible to measure differences between carers in the amounts of support they provided. Nor was it possible to define differences in the types of people who received care, at least in a sufficiently robust way for inclusion here. Only current carers were identified, although some people (such as the recently bereaved) may have been providing informal support to others until just before the survey was undertaken. Caring and psychiatric morbidity In comparison with some recent studies, our results suggest that informal caring is a risk factor for (potential) psychiatric morbidity. People who provide care in the home are 51 per cent more likely to score three or over on the GHQ than those with no such responsibilities; moreover, this situation persists after adjustment for previously established predictors. In fact, caring at home is as statistically important an independent predictor of increased morbidity as 'not being in work' and 'living alone' (albeit a more modest one than lack of social contact, perceived lack of support, and limiting long-standing illness). In contrast, the provision of care outside the home is not a significant risk factor at all in these data, although its association is in the same direction of increased morbidity. Of course, our response to these results should be influenced not only by the existence of a significant risk factor but also by the extent to which it occurs in the population as a whole. Caring at home is less common than the other risk factors in our study, with an estimated 6.3 per cent of the adult population so engaged. In consequence, it has less of a potential effect in terms of population morbidity. Our attributable fractions suggest that the presence of limiting, long-standing illness is the major independent contributor. Lack of social support and social contact were also notable factors (although these may follow rather than precipitate any symptoms of psychiatric morbidity). Nevertheless, the results imply that informal caring is associated with problems which affect the psychological health of the carers; such problems may well reduce the ability of carers to continue their support and thereby increase the demand for formal care. Health Authorities should review the support they give to people who care for others in their own home. Some screening for psychological distress should be an element of that support, although this has notable resource implications. (Our figures imply, for example, that there are around 15 300 carers 'at home' in the Morecambe Bay area.) One way forward may be for carers' groups to be helped to develop their own screening. Of course, it would be better to prevent the development of distress in the first place; determining the best forms of support will require Health Authorities to collect further evidence in consultation with carers and their community representatives. References 1 Green H. Informal carers. A study carried out on behalf of the Department of Health and Social Security as part of the 1985 General Household Survey. London: HMSO, 1988.

INFORMAL CARE AND PSYCHIATRIC MORBIDITY 185 2 Office of Population Censuses and Surveys. General Household Survey: carers in 1990. OPCS Monitor SS 92/2. London: Government Statistical Service, 1992. 3 The Carers (Recognition and Services) Act 1995. London: HMSO, 1995. 4 George LK, Gwyther LP. Carer well-being. A multidimensional examination of family consequences of demented adults. Gerontologist 1986; 29: 253-259. 5 Bergmann K, Jackoby R. The limitations and possibilities of community care for the elderly demented. In: Elderly people in the community: their service needs. Department of Health and Social Services. London: HMSO, 1983. 6 Eagles JM, Craig A, Rawlinson F, et al. The psychological well-being of supporters of the demented elderly. Br J Psychiat 1987; 150: 293-298. 7 O'Connor DW, Pollit PA, Roth M, Brock CPB, Reiss BB. Problems reported by relatives in a community study of dementia. Br J Psychiat 1990; 156: 835-841. 8 Livingston G, Manela M, Katona C. Depression and other psychiatric morbidity in carers of elderly people living at home. Br MedJ 1996; 312: 153-156. 9 Barrow S, Fisher A, Roberts E. Baseline Health Needs Survey 1992: standard analysis. Manchester: North Western Regional Health Authority, 1992. 10 Cox B, Blaxter M, Buckle A, et al. The health and lifestyle survey. Cambridge: Health Promotion Research Trust, 1987. 11 Goldberg D, Hillier V. A scaled version of the General Health Questionnaire. Psychol Med 1979; 9: 139-145. 12 Goldberg D, Williams P. A users guide to the General Health Questionnaire. Windsor NFER-Nelson, 1988. 13 Lewis G, Booth M. Regional differences in mental health in Great Britain. J Epidemiol Commun Hlth 1992; 46: 608-611. 14 Office of Population Censuses and Surveys. The prevalence of psychiatric morbidity among adults aged 16-64 living in private households in Great Britain. OPCS Surveys of Psychiatric Morbidity in Great Britain: Bulletin 1. London: OPCS, 1994. 15 Williams AW, Ware JE, Donald CA. A model of mental health, life events and social supports applicable to general populations. J Hlth Social Behav 1981; 22: 324-336. Accepted on 23 December 1997