American Journal of Epidemiology Copyright © 1992 by The Johns Hopkins University School of Hygtene and Public Health Al rights reserved
Vol 136, No. 12 Printed in U.S A.
Social Intervention and the Elderly: A Randomized Controlled Trial
Michael Clarke,1 Susan J. Clarke,2 and Carol Jagger1
A randomized controlled trial was set up in 1985 to test the effect of social intervention over 3 years among elderty people, aged 75 and above, living alone. The sampling frame was the age/sex register of a large group practice of 12 general practitioners serving the town of Melton Mowbray, Leicestershire, England, with a list size of approximately 32,000 patients. A total of 523 elderly people living alone in 1985 were identified, interviewed, and randomized into experimental and control groups. A lay worker offered the experimental group (n = 261) individual packages of support that aimed at enhanced social contacts. The outcome measures, approximately 3 years later in 1988, were mortality; changes in physical status; demand for medical, paramedical, social, and voluntary services; and changes in a number of subjective variables (morale, loneliness, and setf-perceived health). No significant differences were found for any of the variables with the exception of self-perceived health status, where the experimental group showed significantly greater improvements than did the control group. More importantly, half the elderly in this sample declined several offers of help. Am J Epidemiol 1992;136:1517-23. health status; morbidity; mortality; randomized controlled trials; social environment
Interest has been developing both nationally and internationally in recent years about the relations between social environment Received for publication October 17, 1991, and in final form July 13, 1992. 1 Department of Epidemiology and Public Health, University of Leicester, Leicester Royal Infirmary, Leicester, England. Center for Health Economics, University of York, York, England. Repnnt requests to Dr Carol Jagger, Department of Epidemiology and Public Health, University of Leicester, Clinical Sciences Building, Leicester Royal Infirmary, P.O. Box 65, Leicester LE2 7LX, England This work was supported by the Nuffield Provincial Hospitals Trust and Leicestershire Health Authority. The authors thank Dr. H. Hollis, Dr. P. W E. Johnston, Dr. P. J. C. Howe, Dr. M. N. G. Hatford, Dr. G. E. Martin, Dr. B. C. M. Williamson, Dr. R. J. Thew, Dr. B Kirkup, Dr. D. J. Corvin, Dr. T. D. W. Smith, Dr. D. A. Barrow, Dr. A. D. Firkin, Dr. D M. Lovett, Dr C. O'Shea, Dr B. E Holt, and J. Bishop for their cooperation. They also thank the social services and chiropody departments, the voluntary agencies and the community nursing services for their help, Linda Bingham for coordinating the care packages, John Woods for computer programming, and Elizabeth Box and Alison Hipkin for office management.
and health. These relations have been extensively reviewed by Broadhead et al. (1), who concluded that low social support leads to unfavorable health outcomes. These observations were originally based on studies across all age ranges and, with elderly populations, a clear relation between mortality, morbidity, and social support has generally become evident. Berkman and Syme (2) reported a significant relation between mortality and social support among elderly people from longitudinal population studies in Alameda County. Analysis of 7-year follow-up data revealed the importance of social networks in relation to mortality, having taken into account health status, both observed and subjective, and contact with health services. Seeman et al. (3) reported that social ties were still significant predictors of mortality among elderly people 17 years after the baseline study in Alameda County. Another study, in North Carolina, also showed that
1517
1518
Clarke et a).
mortality risks increased threefold among elderly people who reported impaired social support over a period of 2.5 years (4). Similar findings were confirmed by House et al. (5) in their analysis of the Tecumseh Community Health Study. In Europe, Orth-Gomer and Johnson (6) reported somewhat equivocal findings in a 6-year follow-up of a random sample of the Swedish population using a social contact score linked to mortality. In the oldest age groups (65-74 years), mortality rates were highest for men in the lowest social network tertile, but for women mortality rates were greatest in the highest social network tertile. An intervention study in Copenhagen (7) found that enhanced social intervention was significantly related to mortality and admission rates to hospitals and nursing homes. The study was a 3-year randomized controlled trial of 285 elderly people aged 75 years or more, the design of which was similar to the one described below. What is unclear from many of these studies is whether manipulation of the social environment is practicable and, if practicable, results in improved health outcomes. It is perhaps too readily assumed that the elderly living alone, with comparatively few social contacts, have social needs that can be met by statutory or voluntary agencies. The aim of this study therefore was to demonstrate the effect of social intervention in terms of mortality and morbidity of those aged 75 and over who lived alone. The method used was a randomized controlled trial with the null hypothesis that enhanced social intervention will not reduce mortality and morbidity among elderly people living alone. MATERIALS AND METHODS Identification of the sample
In 1981, a population survey of elderly people was undertaken that described the health and social status and the demand for services of the town of Melton Mowbray in central England, 23 miles northeast of the city of Leicester (8). The sample was drawn from the age/sex register of the 12-doctor
general practice that provides the only medical services for the 32,000 people within a defined radius of the town. This survey showed that 50 percent of the elderly people lived alone and that half had few social contacts. The population for the intervention study was drawn as follows. An address-sorted list of all patients on the age/sex register on September 30, 1985, was constructed. Those aged 75 years and above were marked, thus enabling those elderly persons living alone to be identified (n = 674) as having no one else at the same address. Subsequent interviewing by trained field-workers found 81 people to be ineligible for the study as they had either moved into institutions (n = 24) or warden-controlled accommodations (n = 57). A further 70 people refused to be interviewed, resulting in 523 elderly persons who formed the sample for intervention, a response rate of 88 percent. Information collected in the initial survey included 1) a measure of the Activities of Daily Living (9); 2) an information/orientation score (10), a measure of cognitive impairment and a simple screening tool for assessing dementia (11); 3) Wenger's scale for assessing loneliness (12); 4) Wenger's modification (12) of the Morris and Sherwood (13) 17-item version of the Philadelphia Geriatric Morale Scale (12); 5) perceived health status, i.e., the answer to the question, "Do you think yourself good/fair/ poor for your age?"; and 6) a social contact score developed by Tunstall (14), based on the number of personal contacts during the week and month prior to interview, including both nonfamily and family contacts. The sample of 523 people living alone was divided into deciles by social contact score. Each decile was then randomly allocated into control (262 people) and experimental (261 people) groups. Intervention was offered to all those in the experimental group, starting with those with the lowest social contact scores. Duration of intervention
The baseline interviews prior to randomization took place during 1985. At the initial
Social Intervention and the Elderly
approach in April 1986, the caseworker, who had not undertaken any of the initial interviews, introduced herself to the experimental group at a home visit. If intervention was declined, two further brief visits were made to give subjects further opportunity to request assistance. Thus, by December 1986, potential respondents in the experimental group had received three visits. The maximum period of social intervention extended to just under 2 years. The fieldwork for the postintervention assessments started in February 1988 and had been completed by May 1988. Respondents were interviewed in the same sequential order in which they were approached and had accepted/declined intervention in 1986. Thus, intervention extended from a period of 1.25 to nearly 2 years. Controls were reinterviewed in the same order as initially. Field-workers undertaking the second interview were unaware of the allocation of the groups. Nature of intervention
All interventions were initiated by the lay community worker, and many of them were undertaken by her. The type of assistance given varied but was tailored to each person's request for help. Some elderly people requested only one kind of assistance, others several. In some instances, the type and amount of assistance changed over the course of the intervention period; for example, an initial request for an introduction to a jigsaw puzzle circle may have evolved into organizing visits to the day center. Broadly speaking, the type of intervention carried out fell into one of five main categories, some examples of which are given in parentheses: social and social services (arranging visits to another elderly person or outings with voluntary organizations; Meals on Wheels; home help); financial (liaising with local administrative offices for rates, benefits, or collecting pensions); housing (installing safety chains and spy holes onto doors, arranging for volunteers to do gardening or decorating); nursing (referral for assessment for a bath nurse or requesting ad-
1519
vice from the continence nurse); and medical (assistance in making an appointment to see the family doctor or informal liaison/ discussion with the general practitioner). Statistical methods
Because the intervention was staggered according to the decile of the baseline social contact score to which the subject belonged, a variable (subsequently referred to as BATCH) was created for the control and experimental groups, denoting the decile to which they belonged. This variable was included in analyses as an independent factor to allow for the different lengths of time for which subjects had experienced intervention. Changes from 1985 to 1988 between control and experimental groups were assessed in two ways. For continuous variables, Friedman's two-way analysis of variance by ranks was used with the variable BATCH as a blocking variable. Categorical variables were analyzed using ordered polytomous logistic regression, the dependent variable being the change and the independent variables being the control/experimental group, baseline measure of the variable, and BATCH described above. Differences in survival were analyzed using Cox's proportional hazards regression model with survival time from initial interview as the dependent variable and the control/experimental group and BATCH as independent variables. RESULTS
The baseline characteristics of people allocated to control and experimental groups are shown in table 1. The greatest difference found when comparing the control and experimental groups was among those reporting themselves to be in good health, although this was not statistically significant. These data demonstrate the comparability of the experimental and control groups at the commencement of the intervention. Within the intervention group, the extent of acceptance of help was that, of 260 people (one missing) who were approached, 50 per-
1520
Clarke etal.
TABLE 1. Baseline characteristics of each of the control and experimental groups together with the subcategories of the experimental group (intervention, died prelntervention, and declined intervention): Melton Mowbray Intervention Study, 1985-1988 Control (n = 262)
Activities of Daily Living score, independent (%) Information/orientation score, 12/12 (%) Wenger scale, not lonely (%) Morale scale, high morale (%) Perceived health, good (%) Median social contact score*
Experimental (n - 261)
Intervention ( n - 101)
Died/removed (n-29)
DecSned (n = 130)
79
84
84
70
87
76 41 40 56
71 41 44 48
70 37 35 38
61 37 37 37
74 44 52 57
12.5
12.5
11.3
14.6
14.0
• Interquartile range: control, 8.1-18.4; expenmental, 8.0-19.8, Intervention, 7.4-17.1; died/removed, 8.8-18.5; dedned, 8.321.0.
cent (130 people) declined help, with a further 29 people having died or moved from the area prior to the intervention approach. The remaining 101 people (39 percent) accepted offers of assistance. As this study was designed as a pragmatic, randomized, controlled trial, analyses compared all subjects allocated to the experimental group with those allocated to the control group irrespective of whether or not they subsequently declined the offers of help. By the follow-up in 1988, 189 people in the control group remained in the community and were reinterviewed; 14 people were in institutions, and 57 had died or moved away. By 1988 in the intervention group, 190 people remained; eight people had been admitted to institutions, and 60 people had died or moved away. The median change in social contact score from 1985 to 1988 for the control group was zero (interquartile range, -5.2 to 4.6) and for the experimental group, 0.9 (interquartile range, -2.6 to 6.1). Although this difference was not statistically significant (p = 0.34), the median, minimum, and maximum social contact scores in the experimental group were greater than those in the control group. No statistically significant differences were found between pre- and postintervention scores for physical health status in the form of the Activities of Daily Living index (table 2). Overall, no change in physical health status was reported for 76 percent (146 of 191) of the control group and 74 percent (143 of 192) of the experimental
TABLE 2. Changes pre- and postintervention in Activities of Daily Living scores: Melton Mowbray Intervention Study, 1985-1988 Activities of Daly Living score
Experimentai
Control
No.
%
Less dependent No change More dependent
14 143
7 74
25 146
13 76
35
18
20
10
Total
192
99*
191
99*
Dead, institutionalized, or nonresponse
69
No.
71
• Because of rounding error.
TABLE 3. Changes pre- and postintervention in perceived health status: Melton Mowbray Intervention Study, 1985-1988 Experimental Perceived health status
Control
No.
%
No.
%
Improved No change Declined
38 123 30
20 64 16
21 128 40
11 68 21
Total
191
100
189
100
Dead, institutionalized, or nonresponse
70
73
group. More specifically, 88 percent of the control group who were independent in 1985 were still independent in 1988, compared with 80 percent of the experimental group. A similar pattern of no significant differences between the two groups was also found for perceived loneliness and morale. The findings in table 3 show a significant improvement in perceived health status in
Social Intervention and the Elderly
the experimental group. In all, 38 people reported an improvement in health status in the intervention group, compared with only 21 people in the control group. The largest contribution to this significance was made by 32 individuals (41 percent) whose health rating improved from "fair" in 1985 to "good" in 1988. There were also fewer people whose perceived health status had declined over the experimental period in the intervention group; 15 people (16 percent) reported a decline in their health status from "good" to "fair" compared with 25 people (23 percent) in the control group. Data were also collected on the use of health and social services. Approximately 17 percent of both groups had consulted their general practitioner within the month prior to either interview, while 12 percent had seen a district nurse (public health nurse) in the same time period. Table 4 shows those surviving at halfyearly points up to 3.5 years from interview. Using Cox's proportional hazards regression model with survival time from interview as the dependent variable and adjusting for time of entry into intervention, we found no significant difference in survival between the two groups. People who declined intervention
Because half of the sample in the experimental group declined intervention, it was felt important to describe the characteristics of this subgroup. Table 1 shows that, in terms of physical status, the group declining intervention were most independent. Of TABLE 4. Percentage of survival to various time points from Interview: Melton Mowbray Intervention Study, 1985-1988
0 6 months 1 year 1.5 years 2 years 2.5 years 3 years 3.5 years
Control (n = 262)
Experimental (n = 261)
100 96 92 90 87 85 81 77
100 96 95 90 87 81
78 73
1521
those who had died or moved from the area, a higher proportion were people with high dependency, a not unexpected finding since the majority of this group were terminally ill at the time of interview. Differences between the controls and other groups become apparent among the more subjective variables. Table 1 shows that the lowest proportions of those classified by Wenger's scale as not lonely were in the intervention group and the group who subsequently died or moved from the area. There were also differences between responders and nonresponders regarding perceived health status. Both groups had similarly low proportions of people rating their own health as good. Similar proportions perceived themselves to be in poor health among both those who had accepted assistance (16 percent) and those who had subsequently died or moved (17 percent). Those accepting assistance had a lower social contact score, on average, than did both those who subsequently died or moved and those who declined assistance (table 1). Although this may be explained by the greater levels of dependence in physical status and by the subsequent lower ability to go out of the home and socialize in those accepting intervention, the median social contact scores within each level of Activity of Daily Living score were also lower in the intervention group. DISCUSSION
A pragmatic design has been used in this trial with all those allocated to the experimental group being analyzed with the intention to treat under field conditions and to be compared with a control group. Thus, all subjects allocated to the experimental group in this trial have been compared with those of the control group irrespective of whether they subsequently declined the offer of intervention. However, because 50 percent of the people in this trial declined intervention, some of the characteristics of this group have been presented in an attempt to anticipate queries that may be raised by planners and providers of services.
1522
Clarke et al.
Differences relating to self reporting of loneliness and health were found within the experimental group. Over half of those who withdrew from the trial declared themselves to be never lonely and clearly felt that there were others more in need of help than themselves. These may have been the underlying reasons for declining offers of assistance. Among the survivors, more than twice the proportion of those who felt themselves to be in poor health took up the offer of assistance. In terms of age and social class distribution and of the standardized mortality ratio, this population is similar to that of Leicestershire and England and Wales. The first study in Melton Mowbray showed that the health and social services provision was adequate and compared well with that of England and Wales (8). Although there are no overall differences between the two groups in terms of contact with services, there are some differences in the levels of demand between 1985 and 1988. Contact with the general practitioner remained the same over the time period, but contact with the district nurse had increased. There was a small increase in the demand for home help and home delivery of meals. In times of greater financial constraints and cutbacks, these data again suggest that perhaps the more vulnerable groups are being targeted. It is against this background of an overall increase in contact with health and social services that this study was conducted and that may have been a contributory factor in the number of people declining intervention. No significant changes were found in physical status between the control and experimental groups as a result of social intervention. For the two groups, there appears to be a proportional difference between categories of high dependency before and after testing, but the numbers are almost identical. The increase in the number of social contacts was the main method by which we were able to judge whether or not the intervention had occurred. Although no statistically significant differences were found between the two groups in this variable, on average there was no change in the social
contact score during the study in the control group while, in contrast, the experimental group slightly improved their social contact score. The number of items of assistance arranged for those who accepted intervention ranged from one to 73 with a median of 3.5. Although no statistically significant dose-response relations were found between the number of items received and the main outcome variables, this may be as expected. For some, the solution of a single problem such as incontinence may be sufficient to enable the elderly person to reestablish a social network and, hence, produce quite large changes in social contact scores. For others with long-standing social isolation, many items of assistance may still produce, overall, small changes in the social contact score. It should be remembered that this method of measuring social contact is purely quantitative and that no attempt was made to measure the sometimes intense quality of the relationship between the lay worker and the elderly person. The lack of differences found in these variables raises two fundamental questions. First, it may be that the instruments used have not been sensitive enough to detect subtle changes in well-being. We have, however, attempted to anticipate this criticism by using a number of different assessments, ranging from the subjective such as "Do you ever feel lonely?" to more objective quantitative measures such as the social contact and loneliness scores. The second question raised by any lack of difference found between control and experimental groups may be a result of the time scale of the study which was rather short. The outcome measures may have been different if the intervention could have been extended over a longer period of time. Despite these problems, a significant difference to emerge from this trial has been that between control and experimental groups for perceived health status. In a previous report (15) we showed that elderly people rating their health as poor had over three times the risk of death as those who perceived their health as good. Therefore, the significant increase in perceived health status among
Social Intervention and the Elderly
those in the experimental group may lead ultimately to a decrease in mortality. This has not been demonstrated in this trial, but this may be a reflection of the short time period of the intervention and follow-up. From the first Melton survey (8), the survival rate over 2 years (the planned time of the trial) of those elderly living alone was 90 percent. With a sample size of 262 in each group, this trial has a 90 percent power of detecting as significant (at the 5 percent level) an increase in survival from 90 percent to 97 percent. In retrospect, this change in mortality from a baseline level of 90 percent would be impossible to achieve with the input of services that providers would think reasonable. In summary, therefore, there is insufficient evidence from this randomized controlled trial to reject the null hypothesis that enhanced social intervention will not reduce mortality and morbidity among elderly people living at home. With the exception of self-perceived health status, none of the outcome measures was significantly different between the two groups. That social intervention may have significantly improved perceived health status could be important since this characteristic ultimately affects service use and mortality. However, a finding from this trial of greater importance to service planners and providers is the high number of elderly people living alone who declined any help offered, declaring themselves to be quite content. The existence of this subgroup in the potentially high-risk group of those living alone means that scarce resources may have to be targeted even more carefully. Deaths in this population will continue to be monitored for a number of years.
1523
REFERENCES 1. Broadhead WE, Kaplan BH, James SA, et al. The epidemiologic evidence for a relationship between social support and health. Am J Epidemiol 1983; 117:521-37. 2. Berkman LF, Syme SL. Social networks, host resistance, and mortality: a nine-year follow-up study of Alameda County residents. Am J Epidemiol 1979; 109:186-204. 3. Seeman TE, Kaplan GA, Knudsen L, et al. Social network ties and mortality among the elderly in the Alameda County Study. Am J Epidemiol 1987; 126:714-23. 4. Blazer DG. Social support and mortality in an elderly community population. Am J Epidemiol 1982; 115:684-94. 5. House JS, Robbins C, Metzner HL. The association of socialrelationshipsand activities with mortality: prospective evidence from the Tecumseh Community Health Study. Am J Epidemiol 1982;116: 123-40. 6. Orth-Gomer K, Johnson JV. Social network interaction and mortality. J Chronic Dis 1987;40: 949-57. 7. Hendriksen C, Lund E, Stromgard E. Consequences of assessment and intervention among elderly people: a three year randomized controlled trial. BMJ 1984;289:1522-4. 8. Clarke M, Clarke S, Odell A, et al. The elderly at home: health and social status. Health Trends 1984; 16:3-7. 9. Jagger C, Clarke M, Davies RA. The elderly at home: indices of disability. J Epidemiol Community Health 1984;40:139-42. 10. Pattie A, Gilleard CJ. Manual of the Clifton assessment procedures for the elderly (CAPE). London: Hodder & Stoughton Educational, Ltd, 1979. 11. Clarke M, Jagger C, Anderson J, et al. The prevalence of dementia in a total population: a comparison of two screening instruments. Age Ageing 1991;20:396^»03. 12. Wenger GC. The supportive network. London: Allen & Unwin, Ltd, 1984. (National Institute, Social Services Library, no. 46). 13. Morris JN, Sherwood S. A retesting and modification of the Philadelphia Geriatric Center Morale Scale. J Gerontol 1975,30:77-84. 14. Tunstall J. Old and alone: a sociological study of old people. London: Routledge & Kegan Paul, Ltd, 1966. 15. Jagger C, Clarke M. Mortality risks in the elderly: five-year follow-up of a total population. Int J Epidemiol 1988; 17:111-14.
Vol 136, No. 12 Printed in U.S A.
Social Intervention and the Elderly: A Randomized Controlled Trial
Michael Clarke,1 Susan J. Clarke,2 and Carol Jagger1
A randomized controlled trial was set up in 1985 to test the effect of social intervention over 3 years among elderty people, aged 75 and above, living alone. The sampling frame was the age/sex register of a large group practice of 12 general practitioners serving the town of Melton Mowbray, Leicestershire, England, with a list size of approximately 32,000 patients. A total of 523 elderly people living alone in 1985 were identified, interviewed, and randomized into experimental and control groups. A lay worker offered the experimental group (n = 261) individual packages of support that aimed at enhanced social contacts. The outcome measures, approximately 3 years later in 1988, were mortality; changes in physical status; demand for medical, paramedical, social, and voluntary services; and changes in a number of subjective variables (morale, loneliness, and setf-perceived health). No significant differences were found for any of the variables with the exception of self-perceived health status, where the experimental group showed significantly greater improvements than did the control group. More importantly, half the elderly in this sample declined several offers of help. Am J Epidemiol 1992;136:1517-23. health status; morbidity; mortality; randomized controlled trials; social environment
Interest has been developing both nationally and internationally in recent years about the relations between social environment Received for publication October 17, 1991, and in final form July 13, 1992. 1 Department of Epidemiology and Public Health, University of Leicester, Leicester Royal Infirmary, Leicester, England. Center for Health Economics, University of York, York, England. Repnnt requests to Dr Carol Jagger, Department of Epidemiology and Public Health, University of Leicester, Clinical Sciences Building, Leicester Royal Infirmary, P.O. Box 65, Leicester LE2 7LX, England This work was supported by the Nuffield Provincial Hospitals Trust and Leicestershire Health Authority. The authors thank Dr. H. Hollis, Dr. P. W E. Johnston, Dr. P. J. C. Howe, Dr. M. N. G. Hatford, Dr. G. E. Martin, Dr. B. C. M. Williamson, Dr. R. J. Thew, Dr. B Kirkup, Dr. D. J. Corvin, Dr. T. D. W. Smith, Dr. D. A. Barrow, Dr. A. D. Firkin, Dr. D M. Lovett, Dr C. O'Shea, Dr B. E Holt, and J. Bishop for their cooperation. They also thank the social services and chiropody departments, the voluntary agencies and the community nursing services for their help, Linda Bingham for coordinating the care packages, John Woods for computer programming, and Elizabeth Box and Alison Hipkin for office management.
and health. These relations have been extensively reviewed by Broadhead et al. (1), who concluded that low social support leads to unfavorable health outcomes. These observations were originally based on studies across all age ranges and, with elderly populations, a clear relation between mortality, morbidity, and social support has generally become evident. Berkman and Syme (2) reported a significant relation between mortality and social support among elderly people from longitudinal population studies in Alameda County. Analysis of 7-year follow-up data revealed the importance of social networks in relation to mortality, having taken into account health status, both observed and subjective, and contact with health services. Seeman et al. (3) reported that social ties were still significant predictors of mortality among elderly people 17 years after the baseline study in Alameda County. Another study, in North Carolina, also showed that
1517
1518
Clarke et a).
mortality risks increased threefold among elderly people who reported impaired social support over a period of 2.5 years (4). Similar findings were confirmed by House et al. (5) in their analysis of the Tecumseh Community Health Study. In Europe, Orth-Gomer and Johnson (6) reported somewhat equivocal findings in a 6-year follow-up of a random sample of the Swedish population using a social contact score linked to mortality. In the oldest age groups (65-74 years), mortality rates were highest for men in the lowest social network tertile, but for women mortality rates were greatest in the highest social network tertile. An intervention study in Copenhagen (7) found that enhanced social intervention was significantly related to mortality and admission rates to hospitals and nursing homes. The study was a 3-year randomized controlled trial of 285 elderly people aged 75 years or more, the design of which was similar to the one described below. What is unclear from many of these studies is whether manipulation of the social environment is practicable and, if practicable, results in improved health outcomes. It is perhaps too readily assumed that the elderly living alone, with comparatively few social contacts, have social needs that can be met by statutory or voluntary agencies. The aim of this study therefore was to demonstrate the effect of social intervention in terms of mortality and morbidity of those aged 75 and over who lived alone. The method used was a randomized controlled trial with the null hypothesis that enhanced social intervention will not reduce mortality and morbidity among elderly people living alone. MATERIALS AND METHODS Identification of the sample
In 1981, a population survey of elderly people was undertaken that described the health and social status and the demand for services of the town of Melton Mowbray in central England, 23 miles northeast of the city of Leicester (8). The sample was drawn from the age/sex register of the 12-doctor
general practice that provides the only medical services for the 32,000 people within a defined radius of the town. This survey showed that 50 percent of the elderly people lived alone and that half had few social contacts. The population for the intervention study was drawn as follows. An address-sorted list of all patients on the age/sex register on September 30, 1985, was constructed. Those aged 75 years and above were marked, thus enabling those elderly persons living alone to be identified (n = 674) as having no one else at the same address. Subsequent interviewing by trained field-workers found 81 people to be ineligible for the study as they had either moved into institutions (n = 24) or warden-controlled accommodations (n = 57). A further 70 people refused to be interviewed, resulting in 523 elderly persons who formed the sample for intervention, a response rate of 88 percent. Information collected in the initial survey included 1) a measure of the Activities of Daily Living (9); 2) an information/orientation score (10), a measure of cognitive impairment and a simple screening tool for assessing dementia (11); 3) Wenger's scale for assessing loneliness (12); 4) Wenger's modification (12) of the Morris and Sherwood (13) 17-item version of the Philadelphia Geriatric Morale Scale (12); 5) perceived health status, i.e., the answer to the question, "Do you think yourself good/fair/ poor for your age?"; and 6) a social contact score developed by Tunstall (14), based on the number of personal contacts during the week and month prior to interview, including both nonfamily and family contacts. The sample of 523 people living alone was divided into deciles by social contact score. Each decile was then randomly allocated into control (262 people) and experimental (261 people) groups. Intervention was offered to all those in the experimental group, starting with those with the lowest social contact scores. Duration of intervention
The baseline interviews prior to randomization took place during 1985. At the initial
Social Intervention and the Elderly
approach in April 1986, the caseworker, who had not undertaken any of the initial interviews, introduced herself to the experimental group at a home visit. If intervention was declined, two further brief visits were made to give subjects further opportunity to request assistance. Thus, by December 1986, potential respondents in the experimental group had received three visits. The maximum period of social intervention extended to just under 2 years. The fieldwork for the postintervention assessments started in February 1988 and had been completed by May 1988. Respondents were interviewed in the same sequential order in which they were approached and had accepted/declined intervention in 1986. Thus, intervention extended from a period of 1.25 to nearly 2 years. Controls were reinterviewed in the same order as initially. Field-workers undertaking the second interview were unaware of the allocation of the groups. Nature of intervention
All interventions were initiated by the lay community worker, and many of them were undertaken by her. The type of assistance given varied but was tailored to each person's request for help. Some elderly people requested only one kind of assistance, others several. In some instances, the type and amount of assistance changed over the course of the intervention period; for example, an initial request for an introduction to a jigsaw puzzle circle may have evolved into organizing visits to the day center. Broadly speaking, the type of intervention carried out fell into one of five main categories, some examples of which are given in parentheses: social and social services (arranging visits to another elderly person or outings with voluntary organizations; Meals on Wheels; home help); financial (liaising with local administrative offices for rates, benefits, or collecting pensions); housing (installing safety chains and spy holes onto doors, arranging for volunteers to do gardening or decorating); nursing (referral for assessment for a bath nurse or requesting ad-
1519
vice from the continence nurse); and medical (assistance in making an appointment to see the family doctor or informal liaison/ discussion with the general practitioner). Statistical methods
Because the intervention was staggered according to the decile of the baseline social contact score to which the subject belonged, a variable (subsequently referred to as BATCH) was created for the control and experimental groups, denoting the decile to which they belonged. This variable was included in analyses as an independent factor to allow for the different lengths of time for which subjects had experienced intervention. Changes from 1985 to 1988 between control and experimental groups were assessed in two ways. For continuous variables, Friedman's two-way analysis of variance by ranks was used with the variable BATCH as a blocking variable. Categorical variables were analyzed using ordered polytomous logistic regression, the dependent variable being the change and the independent variables being the control/experimental group, baseline measure of the variable, and BATCH described above. Differences in survival were analyzed using Cox's proportional hazards regression model with survival time from initial interview as the dependent variable and the control/experimental group and BATCH as independent variables. RESULTS
The baseline characteristics of people allocated to control and experimental groups are shown in table 1. The greatest difference found when comparing the control and experimental groups was among those reporting themselves to be in good health, although this was not statistically significant. These data demonstrate the comparability of the experimental and control groups at the commencement of the intervention. Within the intervention group, the extent of acceptance of help was that, of 260 people (one missing) who were approached, 50 per-
1520
Clarke etal.
TABLE 1. Baseline characteristics of each of the control and experimental groups together with the subcategories of the experimental group (intervention, died prelntervention, and declined intervention): Melton Mowbray Intervention Study, 1985-1988 Control (n = 262)
Activities of Daily Living score, independent (%) Information/orientation score, 12/12 (%) Wenger scale, not lonely (%) Morale scale, high morale (%) Perceived health, good (%) Median social contact score*
Experimental (n - 261)
Intervention ( n - 101)
Died/removed (n-29)
DecSned (n = 130)
79
84
84
70
87
76 41 40 56
71 41 44 48
70 37 35 38
61 37 37 37
74 44 52 57
12.5
12.5
11.3
14.6
14.0
• Interquartile range: control, 8.1-18.4; expenmental, 8.0-19.8, Intervention, 7.4-17.1; died/removed, 8.8-18.5; dedned, 8.321.0.
cent (130 people) declined help, with a further 29 people having died or moved from the area prior to the intervention approach. The remaining 101 people (39 percent) accepted offers of assistance. As this study was designed as a pragmatic, randomized, controlled trial, analyses compared all subjects allocated to the experimental group with those allocated to the control group irrespective of whether or not they subsequently declined the offers of help. By the follow-up in 1988, 189 people in the control group remained in the community and were reinterviewed; 14 people were in institutions, and 57 had died or moved away. By 1988 in the intervention group, 190 people remained; eight people had been admitted to institutions, and 60 people had died or moved away. The median change in social contact score from 1985 to 1988 for the control group was zero (interquartile range, -5.2 to 4.6) and for the experimental group, 0.9 (interquartile range, -2.6 to 6.1). Although this difference was not statistically significant (p = 0.34), the median, minimum, and maximum social contact scores in the experimental group were greater than those in the control group. No statistically significant differences were found between pre- and postintervention scores for physical health status in the form of the Activities of Daily Living index (table 2). Overall, no change in physical health status was reported for 76 percent (146 of 191) of the control group and 74 percent (143 of 192) of the experimental
TABLE 2. Changes pre- and postintervention in Activities of Daily Living scores: Melton Mowbray Intervention Study, 1985-1988 Activities of Daly Living score
Experimentai
Control
No.
%
Less dependent No change More dependent
14 143
7 74
25 146
13 76
35
18
20
10
Total
192
99*
191
99*
Dead, institutionalized, or nonresponse
69
No.
71
• Because of rounding error.
TABLE 3. Changes pre- and postintervention in perceived health status: Melton Mowbray Intervention Study, 1985-1988 Experimental Perceived health status
Control
No.
%
No.
%
Improved No change Declined
38 123 30
20 64 16
21 128 40
11 68 21
Total
191
100
189
100
Dead, institutionalized, or nonresponse
70
73
group. More specifically, 88 percent of the control group who were independent in 1985 were still independent in 1988, compared with 80 percent of the experimental group. A similar pattern of no significant differences between the two groups was also found for perceived loneliness and morale. The findings in table 3 show a significant improvement in perceived health status in
Social Intervention and the Elderly
the experimental group. In all, 38 people reported an improvement in health status in the intervention group, compared with only 21 people in the control group. The largest contribution to this significance was made by 32 individuals (41 percent) whose health rating improved from "fair" in 1985 to "good" in 1988. There were also fewer people whose perceived health status had declined over the experimental period in the intervention group; 15 people (16 percent) reported a decline in their health status from "good" to "fair" compared with 25 people (23 percent) in the control group. Data were also collected on the use of health and social services. Approximately 17 percent of both groups had consulted their general practitioner within the month prior to either interview, while 12 percent had seen a district nurse (public health nurse) in the same time period. Table 4 shows those surviving at halfyearly points up to 3.5 years from interview. Using Cox's proportional hazards regression model with survival time from interview as the dependent variable and adjusting for time of entry into intervention, we found no significant difference in survival between the two groups. People who declined intervention
Because half of the sample in the experimental group declined intervention, it was felt important to describe the characteristics of this subgroup. Table 1 shows that, in terms of physical status, the group declining intervention were most independent. Of TABLE 4. Percentage of survival to various time points from Interview: Melton Mowbray Intervention Study, 1985-1988
0 6 months 1 year 1.5 years 2 years 2.5 years 3 years 3.5 years
Control (n = 262)
Experimental (n = 261)
100 96 92 90 87 85 81 77
100 96 95 90 87 81
78 73
1521
those who had died or moved from the area, a higher proportion were people with high dependency, a not unexpected finding since the majority of this group were terminally ill at the time of interview. Differences between the controls and other groups become apparent among the more subjective variables. Table 1 shows that the lowest proportions of those classified by Wenger's scale as not lonely were in the intervention group and the group who subsequently died or moved from the area. There were also differences between responders and nonresponders regarding perceived health status. Both groups had similarly low proportions of people rating their own health as good. Similar proportions perceived themselves to be in poor health among both those who had accepted assistance (16 percent) and those who had subsequently died or moved (17 percent). Those accepting assistance had a lower social contact score, on average, than did both those who subsequently died or moved and those who declined assistance (table 1). Although this may be explained by the greater levels of dependence in physical status and by the subsequent lower ability to go out of the home and socialize in those accepting intervention, the median social contact scores within each level of Activity of Daily Living score were also lower in the intervention group. DISCUSSION
A pragmatic design has been used in this trial with all those allocated to the experimental group being analyzed with the intention to treat under field conditions and to be compared with a control group. Thus, all subjects allocated to the experimental group in this trial have been compared with those of the control group irrespective of whether they subsequently declined the offer of intervention. However, because 50 percent of the people in this trial declined intervention, some of the characteristics of this group have been presented in an attempt to anticipate queries that may be raised by planners and providers of services.
1522
Clarke et al.
Differences relating to self reporting of loneliness and health were found within the experimental group. Over half of those who withdrew from the trial declared themselves to be never lonely and clearly felt that there were others more in need of help than themselves. These may have been the underlying reasons for declining offers of assistance. Among the survivors, more than twice the proportion of those who felt themselves to be in poor health took up the offer of assistance. In terms of age and social class distribution and of the standardized mortality ratio, this population is similar to that of Leicestershire and England and Wales. The first study in Melton Mowbray showed that the health and social services provision was adequate and compared well with that of England and Wales (8). Although there are no overall differences between the two groups in terms of contact with services, there are some differences in the levels of demand between 1985 and 1988. Contact with the general practitioner remained the same over the time period, but contact with the district nurse had increased. There was a small increase in the demand for home help and home delivery of meals. In times of greater financial constraints and cutbacks, these data again suggest that perhaps the more vulnerable groups are being targeted. It is against this background of an overall increase in contact with health and social services that this study was conducted and that may have been a contributory factor in the number of people declining intervention. No significant changes were found in physical status between the control and experimental groups as a result of social intervention. For the two groups, there appears to be a proportional difference between categories of high dependency before and after testing, but the numbers are almost identical. The increase in the number of social contacts was the main method by which we were able to judge whether or not the intervention had occurred. Although no statistically significant differences were found between the two groups in this variable, on average there was no change in the social
contact score during the study in the control group while, in contrast, the experimental group slightly improved their social contact score. The number of items of assistance arranged for those who accepted intervention ranged from one to 73 with a median of 3.5. Although no statistically significant dose-response relations were found between the number of items received and the main outcome variables, this may be as expected. For some, the solution of a single problem such as incontinence may be sufficient to enable the elderly person to reestablish a social network and, hence, produce quite large changes in social contact scores. For others with long-standing social isolation, many items of assistance may still produce, overall, small changes in the social contact score. It should be remembered that this method of measuring social contact is purely quantitative and that no attempt was made to measure the sometimes intense quality of the relationship between the lay worker and the elderly person. The lack of differences found in these variables raises two fundamental questions. First, it may be that the instruments used have not been sensitive enough to detect subtle changes in well-being. We have, however, attempted to anticipate this criticism by using a number of different assessments, ranging from the subjective such as "Do you ever feel lonely?" to more objective quantitative measures such as the social contact and loneliness scores. The second question raised by any lack of difference found between control and experimental groups may be a result of the time scale of the study which was rather short. The outcome measures may have been different if the intervention could have been extended over a longer period of time. Despite these problems, a significant difference to emerge from this trial has been that between control and experimental groups for perceived health status. In a previous report (15) we showed that elderly people rating their health as poor had over three times the risk of death as those who perceived their health as good. Therefore, the significant increase in perceived health status among
Social Intervention and the Elderly
those in the experimental group may lead ultimately to a decrease in mortality. This has not been demonstrated in this trial, but this may be a reflection of the short time period of the intervention and follow-up. From the first Melton survey (8), the survival rate over 2 years (the planned time of the trial) of those elderly living alone was 90 percent. With a sample size of 262 in each group, this trial has a 90 percent power of detecting as significant (at the 5 percent level) an increase in survival from 90 percent to 97 percent. In retrospect, this change in mortality from a baseline level of 90 percent would be impossible to achieve with the input of services that providers would think reasonable. In summary, therefore, there is insufficient evidence from this randomized controlled trial to reject the null hypothesis that enhanced social intervention will not reduce mortality and morbidity among elderly people living at home. With the exception of self-perceived health status, none of the outcome measures was significantly different between the two groups. That social intervention may have significantly improved perceived health status could be important since this characteristic ultimately affects service use and mortality. However, a finding from this trial of greater importance to service planners and providers is the high number of elderly people living alone who declined any help offered, declaring themselves to be quite content. The existence of this subgroup in the potentially high-risk group of those living alone means that scarce resources may have to be targeted even more carefully. Deaths in this population will continue to be monitored for a number of years.
1523
REFERENCES 1. Broadhead WE, Kaplan BH, James SA, et al. The epidemiologic evidence for a relationship between social support and health. Am J Epidemiol 1983; 117:521-37. 2. Berkman LF, Syme SL. Social networks, host resistance, and mortality: a nine-year follow-up study of Alameda County residents. Am J Epidemiol 1979; 109:186-204. 3. Seeman TE, Kaplan GA, Knudsen L, et al. Social network ties and mortality among the elderly in the Alameda County Study. Am J Epidemiol 1987; 126:714-23. 4. Blazer DG. Social support and mortality in an elderly community population. Am J Epidemiol 1982; 115:684-94. 5. House JS, Robbins C, Metzner HL. The association of socialrelationshipsand activities with mortality: prospective evidence from the Tecumseh Community Health Study. Am J Epidemiol 1982;116: 123-40. 6. Orth-Gomer K, Johnson JV. Social network interaction and mortality. J Chronic Dis 1987;40: 949-57. 7. Hendriksen C, Lund E, Stromgard E. Consequences of assessment and intervention among elderly people: a three year randomized controlled trial. BMJ 1984;289:1522-4. 8. Clarke M, Clarke S, Odell A, et al. The elderly at home: health and social status. Health Trends 1984; 16:3-7. 9. Jagger C, Clarke M, Davies RA. The elderly at home: indices of disability. J Epidemiol Community Health 1984;40:139-42. 10. Pattie A, Gilleard CJ. Manual of the Clifton assessment procedures for the elderly (CAPE). London: Hodder & Stoughton Educational, Ltd, 1979. 11. Clarke M, Jagger C, Anderson J, et al. The prevalence of dementia in a total population: a comparison of two screening instruments. Age Ageing 1991;20:396^»03. 12. Wenger GC. The supportive network. London: Allen & Unwin, Ltd, 1984. (National Institute, Social Services Library, no. 46). 13. Morris JN, Sherwood S. A retesting and modification of the Philadelphia Geriatric Center Morale Scale. J Gerontol 1975,30:77-84. 14. Tunstall J. Old and alone: a sociological study of old people. London: Routledge & Kegan Paul, Ltd, 1966. 15. Jagger C, Clarke M. Mortality risks in the elderly: five-year follow-up of a total population. Int J Epidemiol 1988; 17:111-14.
