Randomised controlled trial of effect of intervention by psychogeriatric team on depression in frail elderly people at home

SUBJECTS

The study took place in Lewisham, an inner city London borough. Its protocol was approved by the local ethics committees. All those who were aged 65 or more and were receiving home care but were not currently under psychiatric care were eligible for inclusion. Power calculations before starting the study (based on 30% of the control group and 60% in the intervention group recovering from depression, with a 5% significance level and a power of 80%) indicated that 30 people would be needed in each of the two groups (intervention and control).

RECRUITMENT

Elderly people eligible for entry to the study were screened by means of the Selfcare(d) questionnaire, which has 12 items and is a self rating depression questionnaire developed for use in elderly populations.14 Those scoring 8 or more were visited at home and asked for their consent to an assessment that included demographic data, mental state, physical illness, limitations to activity, and social support.15 The ratings of mental state included the geriatric mental state examination, a semistructured psychiatric diagnostic interview,16 17 and the Montgomery Asberg depression rating scale, an observer rated depression scale sensitive to change.18

DEFINITION OF DEPRESSION

The geriatric mental state/AGECAT system was selected for case definition because it is the most comprehensive and best validated research assessment of mental state in elderly people. (AGECAT stands for automatic geriatric examination for computer assisted taxonomy.) Psychiatric diagnoses are generated by the AGECAT computer algorithm. These diagnoses have been extensively validated against clinical diagnoses and the criteria of the Diagnostic and Statistical Manual of Mental Disorders, third edition (DSM-III)19 with good levels of agreement.17 20 21

RANDOMISATION

Subjects were randomised to the intervention or control group on a 1:1 basis by means of a computer generated three digit random number (odd numbers to the intervention group).

MANAGEMENT OF INTERVENTION GROUP

The intervention was designed to be as naturalistic as possible to maximise generalisability. Members of the intervention group were treated in the same way as anyone referred to the psychogeriatric team in the catchment area. Each case was presented at a multidisciplinary team meeting which included community psychiatric nurses, occupational therapists, senior and junior medical staff, a social worker, and a psychologist. A management plan was formulated by the team for each subject on an individual basis, as for any referral to the team.22 This could include any combination of physical interventions—for example, prescription of antidepressants, physical review—psychological interventions—for example, bereavement counselling, family work—and social interventions—for example, referral to a day centre, benefit check.

One of us (SB) acted as each person's keyworker. All members of the team, whether medically qualified or not, may be assigned any case referred. The study population differed in their management only by their all being assigned a doctor. Subjects were visited in their own homes and progress regularly reviewed by the team.

MANAGEMENT OF CONTROL GROUP

The control group received general practitioner care. Letters were sent to general practitioners to say that their patient was participing in the study as a control but that this should not affect their management of him or her. The letter also reiterated that patients could be referred as needed and that they would be accepted by the psychogeriatric team as normal if they were referred.

FOLLOW UP

Six months after recruitment subjects were reassessed in their own homes by one of us (KS), blind to the group to which the subject had been randomly allocated. The ratings completed at the initial interview were repeated. Any member of the control group who remained depressed at follow up was referred directly to the psychogeriatric team.

STATISTICAL ANALYSIS

We carried out intention to treat analyses. We calculated the proportion of people recovering from depression in each group. Recovery from depression was defined as changing from an AGECAT case of depression at recruitment to a non-case at follow up. We calculated differences in the change in the AGECAT depression ratings and score on the Montgomery Asberg depression rating scale during the study. Depression was deemed to have improved if the AGECAT depression rating was lower at follow up than recruitment, to be the same if the rating was unchanged, and to be worse if it increased. We calculated 95% confidence intervals for the differences in proportions or change in score.23

To evaluate the effect of intervention and allow for possible confounding, we carried out logistic regression modelling with EGRET.24 Only subjects completing the study were included, the outcome being recovery from depression. This approach allowed the independent effect of each variable in the model on recovery to be estimated and controlled for the effect of all others in the model. Odds ratios with 95% confidence intervals were calculated. To estimate the effect of antidepressant prescription, the model was then extended to include their consumption at follow up.

RECRUITMENT AND RANDOMISATION

A total of 441 subjects were eligible for the study, and 317 (72%) completed the Selfcare(d) screening. Twenty four subjects had died or moved and 100 were unwilling or unable to complete it. Overall, 180 subjects scored above 8; 154 of them were interviewed, nine having died or moved and 17 having refused to give informed consent. Sixty nine subjects with AGECAT cases of depression were identified and entered into the study. Thirty six were randomly allocated to the control group and 33 to the intervention group. Baseline characteristics are presented in table 1. There were no significant differences in the distribution of these variables between the two groups.

CLINICAL OUTCOME

At six month follow up four of the intervention group and three of the control group had died, but none had committed suicide or died of any cause attributable to depression or psychiatric treatment. One of the control group declined follow up. table 2 presents the six month outcome of depression. Significant beneficial effects of randomisation to the intervention group were apparent using each of the three measures of change in depression.

The logistic regression analysis is summarised in table 3. No significant interactions were identified. After the effects of all variables in the model were allowed for, the powerful treatment effect persisted (odds ratio 9.0 (95% confidence interval 2.0 to 41.5). Increasing depression severity and first episode of depression predicted worse outcome at six months. When antidepressant use at follow up was added to the model there was no significant effect (likelihood ratio statistic 1.86, df = 1, P = 0.173). The multiply adjusted odds ratio for the effect of antidepressants was 0.3 (0.0 to 1.9).

MANAGEMENT OF DEPRESSION

Three (9%) of the control group and four (14%) of the intervention group were receiving antidepressants at recruitment compared with five (16%) and 20 (69%) respectively at follow up. During the study two (6%) of the control group were referred to the psychiatric team and three of the intervention and one of the control group were admitted to psychiatric units.

table 4 presents a summary of the management plans formulated for the intervention group and the proportion of these interventions completed. When more than one intervention was proposed in a category, the intervention was regarded as completed if any were successfully introduced. Counselling or psychotherapy includes all specific psychological interventions but not the supportive exchanges which all subjects received. Reviews of physical state were carried out by primary or secondary care services as appropriate. In all but one case the prescription of antidepressants was carried out by the patient's general practitioner after advice from the team.

POSSIBLE LIMITATIONS

There is the possibility of non-response bias; if the non-responders differed from the responders in terms of depression or prognostic factors, then the generalisability of the study findings might be compromised. The non-responders to the first stage screening included people who were unable to complete the questionnaire because of severe physical disorder and dementia. Ten of the 17 who withdrew at the second stage were excluded because they were cognitively impaired to the extent that they could not give informed consent. However, this is unlikely to have introduced bias because AGECAT operates a hierarchy whereby people with dementia and depression are assigned an organic rather than a depressive diagnosis. Such cases would therefore not have met the entry criteria for the intervention study. Nevertheless, the study findings may not be generalisable to subjects with both depression and dementia.

The study was completed in subjects receiving home care rather than the wider disabled elderly community of which the home care population is a part. Subjects receiving home care make up a large and important group in terms of numbers and use of services, but the results of this study may not be directly generalisable to those with equivalent disability who are maintained in the community without home care.

INTERVENTION

The study has methodological strengths: the definitions of depression were stringent and based on clinical criteria; follow up was blind to group membership; the size of the treatment effect indicates a clinically significant benefit; and the persistence of the positive finding after logistic regression to control for confounding adds to its validity. The treatment effect observed in this study was greater than that reported for intervention by community nurses in the general elderly population11 and for elderly patients attending their general practitioner.25 This may be because of tighter case definition and greater success in implementing management plans. Although the intervention group all had a doctor as their keyworker, we think that this is the most naturalistic controlled trial of the treatment of depression in elderly people to date and the first, so far as we know, to focus on a disabled population.

The natural course of depression in this group seems bleak: only a quarter of the control group recovered in six months. The finding that the most severe disorders were least likely to recover is consistent with the results of other studies.9 26 However, the fairly poor prognosis for the first episode of depression, which persisted even after controlling for its duration (data not shown), was unexpected. This population may have a particular excess of irreversible physical or social causal factors precipitating or perpetuating depressive disorder, so limiting recovery from the first episode of depression. The effect observed seem not to have been solely a function of antidepressant prescription and compliance. However, the study was designed as an evaluation of team intervention as a whole rather than as a drug trial, and thererfore any consideration of the impact of individual elements of the intervention is limited.

IMPLICATIONS

The study has practical implications. The intervention approximated to a naturalistic process, being formulated and implemented through a psychogeriatric team in the same way as for anyone referred to that team. Although the keyworker for the intervention group was always a doctor, he implemented a plan formulated by the team and made no more visits than other team members. Our results were therefore achieved in what might be considered optimum conditions. This design was adopted to investigate whether anything could be done for depression in this group. Further work is needed to confirm that similar benefits can be achieved by other psychogeriatric services working entirely normally. In addition, the degree to which these people with depression can be successfully managed by their primary health care team needs to be investigated.

No formal economic evaluation was carried out, but the only additional resource required by the team to achieve the effect was an extra half time doctor. Further work is needed to investigate which are the most important components of the intervention. An important development of this work may be the formulation and evaluation of management protocols and training packages for members of primary and secondary health care teams and social services to aid the recognition, assessment, and treatment of depression in this housebound group.

We thank the staff of the Lewisham East Home Care Service and all the members of the Community Team for Mental Health in the Elderly (B) at Hither Green Hospital for their exceptional cooperation with the study, and Dr Graham Dunn for statistical advice.