Effectiveness of case management and post-acute services in older people after hospital discharge

Participants were all patients aged 65 years and over who were discharged between August 1998 and October 1999 and met the eligibility criteria in Box 1.

Before randomisation, informed consent was obtained from patients, baseline demographic data were collected, and the Assessment of Quality of Life (AQoL) questionnaire was administered.13 This Australian questionnaire has five dimensions (Illness, Independent Living, Social Relationships, Physical Senses and Psychological Well-being). Patient carers were interviewed using the Caregiver Strain Index.14

Patients were randomly allocated to the intervention (PAC) or control group by computer-generated numbers. These were provided in sequentially numbered, sealed envelopes which were opened after baseline interviews.

• telephone follow-up as required;

• availability to patients in the event of a crisis;

• liaison with service providers (eg, local councils and nursing agencies);

• coordination of service provision; and

• ensuring adequate referral before discharge from the PAC program (eg, to councils or community health centres).

A budget was available to PAC coordinators to purchase community services directly for patients in the post-discharge period, enabling them to deliver a substantial patient-centred service plan unavailable to control patients.

Control patients received usual hospital discharge planning, provided by ward nursing staff and the social work department. Services were typically limited to several nursing visits per week, as well as community services, such as delivered meals and housekeeping support.

Statistical analysis was performed using SPSS.15 Analysis was by modified intention to treat. The study aimed to recruit 720 patients aged 65 years or over. Allowing for an attrition rate of 20%, this would give about 600 patients available for analysis (α = 0.05 and β = 0.2), which would allow the study to detect a 15% difference in readmission rates.

Continuous variables were compared between groups using t tests. Although health services data were skewed, it was felt that, because of the nature and size of the study, parametric tests were sufficiently robust to allow comparison between groups. For categorical variables, Fisher's exact test or Pearson's χ2 test was used. To examine interaction between variables, stepwise multiple linear regression was used where the dependent variable was continuous. Log rank analysis was then used to compare differences between the intervention and control groups.

A total of 654 patients were recruited, and 598 were included in the analysis. Patient allocation is shown in Box 2. There were no significant differences between the groups in demographic characteristics, diagnostic categories, comorbidities, number of medications used, other baseline admission characteristics and length of hospital stay during the index admission (Box 3).

There was also no difference between the patient groups in mortality, both in terms of proportion of patients who died during the six months of follow-up (6% in both groups; χ2, P = 0.92) and time to death (log rank analysis, P = 0.84).

All patients completed the AQoL at baseline, and 540 completed it at one-month follow-up. At baseline, there was no difference in quality-of-life scores between the two groups, except in independent living, where the PAC group had a lower mean baseline score than the control group (0.4 v 0.45 on a scale of 0–1, with a higher score representing better quality of life; P = 0.02). However, the groups differed significantly in changes in scores between baseline and one-month follow-up (Box 4). The PAC group had significantly greater improvements in independent living (P = 0.002) and overall quality-of-life scores (P = 0.02).

Caregiver burden was assessed at baseline and one-month follow-up for 263 patients (140 PAC and 123 control). There was no significant difference in scores for the Caregiver Strain Index between PAC and control patients (mean score, 3 for both groups, on a scale of 0–10).

In the six-month follow-up period, there were no significant differences between the PAC and control groups in unplanned readmissions to the index hospital or emergency department presentations (Box 5). However, hospital bed-day use was significantly lower in the PAC group than in the control group (mean, 3.0 days v 5.2 days; P = 0.01; mean difference, 2.2 days; 95% CI, 0.5–3.9).

Using multiple linear regression analysis, with hospital bed-day use as the dependent variable, the strongest predictor of this use was unplanned admissions in the 12 months before recruitment, followed by length of stay during the index admission. The PAC intervention was significantly negatively correlated with hospital use in the six-month follow-up period (P = 0.009).

Nursing and personal care services made up a significantly greater proportion of the community services used in the six months post-discharge by PAC participants than by control participants. In contrast, Meals on Wheels services made up a greater proportion of services used by control participants (Box 5).

There were no significant differences between groups in costs of community services used in the 12 months before the index admission and the six months after discharge (Box 6). However, hospital utilisation costs in the six months after discharge were significantly lower in the PAC group than in the control group (mean difference, $1770; 95% CI, $237–$3304). Total costs, including costs of the intervention ($292.40 per PAC client), hospitalisation and use of community services over the six months of follow-up, were also significantly lower in the PAC group (mean difference, $1545; 95% CI, $11–$3078).