Systematic review of day hospital care for elderly people

Inclusion criteria

We set out to identify all relevant controlled clinical trials of geriatric day hospital care. A day hospital was defined as an outpatient facility where older patients attend for a full or near full day and receive multidisciplinary rehabilitation in a health care setting.3 Trials evaluating social day centres, other types of day hospitals such as those for patients with dementia or psychiatric conditions, and single condition day hospitals were excluded.

Search strategy

We searched for relevant published and unpublished papers up to January 1997. Our search strategy included Medline, SIGLE (System for Information on Grey Literature in Europe), BIDS (Bath Information and Data Services), CINAHL (Cumulative Index to Nursing and Allied Health), and the Cochrane Library. Key search terms included day hospital, day care, ambulatory care, and outpatient clinic. We also searched Index Medicus, British and international dissertation abstracts, conference abstracts (Society for Research in Rehabilitation, British Geriatrics Society), bibliographies of known trials, and other relevant articles and books. We publicised our work through presentations at geriatric symposia and by contacting authors of previous articles on day hospital care.

Study appraisal and extraction of data

We selected outcomes to reflect a previous definition of the purpose of day hospital care: “to facilitate and prolong independent living for the elderly in the community.”4 Thus death, the need for institutional care, disability, hospital use, and resource use were selected as quantitative outcomes. In anticipation of incomplete data we also defined a global “poor outcome” comprising death or one of the following (in order of preference): resident in institutional care, severe disability at end of follow up, or deterioration in physical function during follow up. The research reports were independently assessed by three reviewers to establish eligibility, agree subcategories for the trials based on the treatment comparison, and to extract data.

Statistical methods

When possible, studies were analysed on an intention to treat basis. Surviving authors were contacted and asked to supply additional information when necessary. Patients who were lost to follow up or for whom outcome data were not available were excluded from the initial analysis. However, they were included in “best case” (all missing data in favour of day hospital care), “neutral” (all missing data recorded at the average event rate for the trial), and “worst case” (all missing data in favour of alternative care) sensitivity analyses.

We calculated odds ratios (with 95% confidence intervals) for the dichotomous outcomes using standard methods. A fixed effects approach5 was used initially and, if substantial heterogeneity was present, we confirmed this with a random effects approach.6 Outcomes were recorded at the end of scheduled follow up.

Patient characteristics

One study was run by the Department of Veterans Affairs in the United States and was exceptional in that the sample was 96% male.15 In all trials the mean age of participants was at least 65 years and usually over 70 years. Three trials 14 16 17 recruited only stroke patients, but the rest had a mixed population. Most patients had significant disability on a variety of scales.

Outcome data

Details on deaths were published, or provided on request, for all trials. Information about institutionalisation was published in only four studies 7 9 13 16; additional data were provided by the authors for a further six studies. 11 12 14 15 17 18 Data were also available concerning some aspect of disability (11 trials) and use of hospital beds (12 trials).

No significant differences in numbers of deaths occurred between patients attending day hospitals and those receiving any of the alternative services (fig 1). The pooled odds ratio for death was 1.02 (95% confidence interval 0.82 to1.26). There was no significant heterogeneity between the results of individual trials or categories of trials.

Figure 2 shows that overall there was no significant difference between day hospital and alternative forms of care in preventing death or a “poor” outcome at final follow up (odds ratio 0.90, 0.71 to 1.14; random effects model). However, patients attending day hospitals were less likely to have poor outcome than those receiving no comprehensive care (0.72, 0.53 to 0.99; P<0.05).

Fig 2

Odds of death or poor outcome in patients receiving day hospital care or alternative services (random effects model). Heterogeneity between trials is presented as χ²

• Download figure
• Open in new tab
• Download powerpoint

Sensitivity analyses

Sensitivity analyses indicated that excluding trials with uncertain or insecure randomisation procedures would not substantially affect our conclusions. 8 10 12 Data were missing on death for 155 (6%) patients and on death or poor outcome for 174 (7%). Best case and worst case sensitivity analyses include significant benefit (P<0.001) or harm (P<0.05) from day hospital care. The most plausible (neutral) sensitivity analysis suggested no effect on death (odds ratio 1.02, P>0.05) or death and poor outcome (0.92, P>0.2).

Disability

Eleven of the trials included a standardised measurement of activities of daily living,8^–18 but seven different measures were used which could not be combined in a statistical summary. Seven trials 8 11 12 14 16^–18 provided data on 905 survivors (data missing for 65, 7%) whose disability scores were monitored during follow up (fig 3). Overall, day hospital patients had a similar chance as controls of suffering deteriorating function (1.11, 0.68 to 1.80; random effects model) but tended to have better outcomes than those receiving no comprehensive care (0.61, 0.38 to 0.97; P<0.05).

Fig 3

Deterioration in activities of daily living among survivors allocated to day hospital care or alternative services (random effects model). Heterogeneity between trials is presented as χ²

• Download figure
• Open in new tab
• Download powerpoint

Use of resources

Institutional care —Eleven trials provided information about the number of patients requiring institutional care at the end of follow up. Overall, there was a trend towards fewer day hospital patients requiring long term (median follow up 12 months) institutional care than those receiving alternative services (table) and the difference was significant in comparison with no comprehensive care.

Use of hospital beds —Average hospital bed use per patient recruited could be calculated for all the trials, but a measure of variance could not be obtained and therefore we cannot report confidence intervals. The table shows a small reduction in bed use by day hospital patients across all trials (15.0 v 16.4 days).

Costs —Eight trials compared treatment costs, only two of which included costs of nursing home care. 10 15 Six reported that day hospital care was more expensive than the comparison treatment, 8 9 15 18 22 23 and two trials reported that the costs were similar. 10 17

Limitations of the review

The main limitations of our review lie in the 30 years over which the trials were done, during which health and social care policies will inevitably have changed, and in the multinational location of the studies. Nevertheless, our data provide the best evidence available on which to judge the effectiveness of day hospitals.

A further problem is the limitations of the trials themselves, in particular the lack of statistical power resulting from small, heterogeneous trials. This raises the possibility of false positive and false negative results. Furthermore, the amount of missing outcome data adds uncertainty to most of our conclusions.

Publication bias remains a possibility, but our search strategy was extensive and included contacting authors of papers on day hospital care around the world. Many of the authors were able to provide unpublished additional information. A funnel plot analysis24 did not show any major evidence of missing data.

Findings

A big problem was that most trials compared day hospital care with an alternative “active” control. Only three trials 7 8 17 used a comparison group of patients who received neither comprehensive care nor domiciliary rehabilitation. Compared with patients receiving no comprehensive care, patients attending day hospitals had less functional deterioration and institutional care and a small reduction in average hospital bed use. However, there was considerable loss to follow up in these three trials.

When day hospital services were compared with comprehensive elderly care (integrated inpatient, outpatient, and domiciliary services), the outcomes were largely similar. One interpretation is that day hospitals are comparable with an alternative service for which there is considerable evidence of effectiveness.25 However, the question arises whether day hospitals provide additional benefits when other comprehensive elderly care services are already in place. Our systematic review does not directly address this question.

We focused on a few specified outcomes, but many other important outcome domains, including patient preference and instrumental daily living activities, could not be examined as these outcomes were not widely available. It was also difficult to determine a summary statistic for disability because different measurement instruments were used and analysed in different ways. Moreover, commonly used measures of disability may be insensitive to change in the outpatient setting of a day hospital.26 This may explain why six trials reported no significant difference in disability outcome between day hospital and comparison services. Future trials should incorporate measures of intrumental activities of daily living as a more relevant and sensitive outcome.

Our analysis suggests that attending day hospital may reduce patients' use of hospital and institutional care resources. However, costing studies need to balance any possible savings with the direct costs of day hospital care. Several studies have drawn attention to the expense of day hospital services. 22 23 For six of the eight trials reporting cost information, day hospital care was more expensive than the alternative treatment.

On the whole, the trials we included took a pragmatic design approach and attempted to address a broad question of overall efficacy of day hospitals. Given the diversity of patients attending day hospitals and the corresponding diversity of interventions used, future trials need to be large multicentre trials or should examine more focused questions.