Patients

From November 1997 all patients referred to the nurse-led unit of Southampton University Hospitals Trust for post-acute care were considered for entry to the trial. We continued to recruit patients until we reached our target sample size (April 1999). Patients were referred from other general medicine wards of the hospital and assessed for admission to the unit. The admission criteria were that patients must be at least 16 years old, medically stable (that is, not requiring tests or investigations) for at least 24 hours, must consent to transfer to the unit, and must have the doctor's referral and patient's resuscitation status recorded in their medical notes. Eligible patients were randomised to care on the nurse-led unit or to usual post-acute care in the hospital with the stipulation that they should not be transferred to the nurse-led unit.

Patients who entered the trial or suitable proxies were approached for their written consent to be telephoned for a brief interview six months after randomisation and to agree that their general practitioner could, if willing, provide the research team with information about their use of services during the six months' follow up. Because funding for follow up was granted six months after the inpatient trial had begun, we sought consent for some patients by post, with telephone follow up. We approached all other patients in hospital after their assessment for transfer to the nurse-led unit. Patients were able to agree to one aspect of data collection and refuse consent for another.

The nurse-led unit

The unit opened in January 1997 as an alternative to usual post-acute care on general medical wards. It had 10 beds and 22 nursing staff (ratio of 3:2 for qualified to unqualified nurses, no special training required). It is part of a teaching hospital trust in Southampton (population about 300 000) and is located near the main hospital site in a smaller setting emphasising outpatient care. The main site has stroke rehabilitation and elderly care units. The area is also served by a community healthcare trust with nine small hospitals and by two local authorities providing social services.

The nurse-led unit has a policy of open visiting. Nurses dress informally to promote a non-clinical atmosphere. On a patient's referral from the medical directorate's acute wards, the unit's staff assess the patient, determine whether admission is suitable, plan treatments, deliver care, and arrange discharge. There is no formal medical involvement; if staff have concerns, they can request a medical review at any time. Doctors can be called to the ward in an emergency, or nurses can send a patient back to the main site for medical attention via the accident and emergency department. A dedicated physiotherapist visits three times a week, and other ancillary services are available on request.

Randomisation

A randomised consent design was used for the trial.18 We adopted this approach after the pilot analysis identified ethical and scientific difficulties with conventional methods of randomisation and consent, which were tried initially.17 The randomised consent design has been recommended in cases where new treatment is compared with the best available standard treatment, as in the control condition for this study, and where standard consent procedures for randomised controlled trials can lead to unnecessary distress and confusion.19 The nurse-led unit's patient population fitted this criterion, as patients were mostly elderly and still unwell. Moreover, a high proportion of recipients of the nurse-led service were cognitively impaired (17/48 (35%) of those in the pilot17). Their relatives or doctors were willing to decide on transfer, but many were reluctant to make proxy decisions about inclusion in a randomised controlled trial. To exclude such patients when they were major recipients of the service would have been unethical20 and could have seriously compromised validity by producing a non-representative study sample.

Allocations to the two treatment arms were computerised and stratified by referring ward to ensure equal access to the nurse-led unit. They were then placed in sealed opaque envelopes that were held by research nurses. When a bed became available on the nurse-led unit, a researcher assessed the next case on a waiting list of referred patients. If the patient was eligible, the next envelope for that ward was opened. If the allocation was to the nurse-led unit, the patient or carer was asked for written consent to transfer (this was standard practice before the trial started) and, at the request of the local research ethics committee, given the chance to refuse collection of trial data.

Outcomes

Sample size

We determined sample size for the primary outcomes of the inpatient trial and calculated that 200 patients in total would be required to have 80% power in a 5% test if the true difference in length of stay were 8 days (with standard deviation 20 days), that 210 patients would be needed to detect a 10 point difference in Barthel Index (SD 20 points) with 95% power, and that 220 patients be needed to detect a reduction from 55% to 35% in the rate of patients discharged to a higher level of support with 85%. We therefore chose a sample size of 240 in total to allow for refusal, dropout, and uncertainty about the true size of differences on which power calculations were based.

Statistical analysis

We compared binary outcome variables of the treatment and control arms of the trial in a logistic regression model, controlling for the stratifying variable, the hospital ward from which patients were referred. Results are presented as controlled odds ratios with 95% confidence intervals. We estimated differences in mean lengths of stay and Barthel scores, with their associated 95% confidence intervals, using regression models controlled for referring ward. In our analysis of Barthel scores we also controlled for functional status at entry to the trial. Median values are also presented for these quantitative variables. Where data were not normally distributed we performed uncontrolled Mann-Whitney U tests (results not shown because conclusions were unchanged).

We compared mean number of new complications, medical reviews, tests, changes to medication, and therapy sessions while in the hospital trust in Poisson regression models, controlling for referring ward and including length of stay as an exposure variable. These models excluded two subjects in each group who had zero length of stay. The Poisson models produced the ratio of mean numbers per patient per day between treated and control groups, with associated 95% confidence intervals, while we estimated ratios of means per patient from models that excluded length of stay.

We repeated analyses with all controlled models using patient's sex as a covariate (results not shown because conclusions were unaffected). Patients were included in their allocated group for analysis irrespective of any subsequent treatment received.

Statistical analyses of follow up data proceeded along similar lines. In our analyses of primary and community district nursing care we controlled for hospital length of stay.

Patients' characteristics

A total of 240 patients were randomised, and the final inpatient sample was 238 because medical records were missing for two patients (see figure). Of the 119 patients randomised to the nurse-led unit, six refused the offer of transfer and another six were not transferred because of unexpected deterioration or unexpectedly quick discharge from hospital. At follow up, we obtained data on use of secondary and primary care services for 238 and 179 patients respectively and interviewed 154 patients or carers. More of the patients allocated conventional care declined to participate in the follow up (23/121 (19%) v 4/117 (3.5%)), but we found no important differences between those who consented and those who refused.

Flow of patients through trial

• Download figure
• Open in new tab
• Download powerpoint

Table 1 shows the patients' baseline characteristics. Patients were generally elderly but not extremely old. Most lived alone in the community, and many were frail or cognitively impaired. Social needs were most commonly noted as the reason for referral to the nurse-led unit (65 cases), but no reason was given for 91 patients.

Length of stay

Table 2 gives details of patients' length of inpatient stay, including readmission days. Patients in the nurse-led unit stayed in the hospital trust more than 14 days longer, on average, than those managed conventionally. Six of the 107 patients in the nurse-led unit (6%) were transferred to community hospitals for post-acute convalescence compared with 30/113 (27%) patients in the control group (difference −21.0% (95% confidence interval −32.8% to −10.3%), P=0.000). Taking these transfers into account (that is, length of stay until first discharge home), we found the difference between groups in initial length of post-acute hospitalisation dropped to 4.5 days and was no longer significant. Over the entire study period, differences in total patient length of stay continued to diminish.

Patient outcomes

As tables 3 and 4 show, patients were neither better nor worse off if cared for in the nurse-led unit. Patients in the nurse-led unit were somewhat more likely than those in the control group to have been discharged to a higher level of support, especially when the discharge destinations for patients initially transferred to community hospital were included in the calculation, but these differences were not significant (table 3). After controlling for Barthel score at baseline, we found no difference between the groups' functional status at discharge or after six months (table 4). Mortality at six months' follow up was substantial in both groups and virtually identical, as was the patients' perceived quality of life. Patients in the nurse-led unit had noticeably, though not quite significantly, lower readmission rates than control patients (12% v 20% at 1 month, 32% v 41% at 6 months).

Processes of care

Table 5 compares processes of care in the two groups. As expected, patients in the nurse-led unit had significantly fewer minor medical reviews, both overall and per patient per day. They also had significantly fewer major medical reviews, tests and investigations, changes to medication, and number of new complications in hospital per patient per day. The patients in the nurse-led unit received significantly more physiotherapy and occupational therapy sessions than the control patients, but when we adjusted the analyses for the longer lengths of stay in the nurse-led unit the differences were no longer significant. Nurses transferred 22/103 patients (21% (95% confidence interval 14% to 31%)) from the nurse-led unit back to acute medical wards during their initial post-acute stay. After hospital discharge, we found no differences in the number or type of contacts between patients and general practitioners, practice nurses, or district nurses. We also found no difference between groups in their propensity to receive social services (70/86 (81%) of patients in the nurse-led unit who were interviewed v 56/68 (82%) of control patients).

Study limitations

At the level of nurse-led units, our study's sample size is one. However, in its staffing and funding, the unit we examined was more typical of normal NHS provision than nursing development units evaluated in the past. We would argue that all innovations in intermediate care must be locally sensitive and contingent on existing services as well as pressing needs. What is generalisable is the patient population, the pressures on general medical beds, and the low technology setting on a separate ward.

The use of a randomised consent design was somewhat unusual and might have resulted in a high refusal rate among the patient allocated to the nurse-led unit. This would have diluted the effects that could be observed. In the event, refusal rates were considerably lower than those reported elsewhere for such designs.22

There are other limitations to this evaluation. Firstly, we chose primary outcomes that are well validated and consistently used in studies of post-acute care, but these may not be optimal measures of effectiveness. Specifically, the Barthel index offers little in relation to patients' own goals for recovery but rather narrowly equates functional status with the need for formal support.

Secondly, we derived the Barthel scores at admission and discharge from patients' medical notes, but at six months we derived them from patients' or carers' reports. Although randomisation should ensure an unbiased comparison between groups, the judgments of capability might have differed between professional and lay assessors.

Thirdly, given that transfers to community hospital emerged as an important factor, it would have been useful if we had examined styles of service delivery, perhaps with the process indicators from the inpatient portion of the trial. However, this was beyond the scope of the trial.

Fourthly, not all localities feature community hospital alternatives to acute hospital care. In such areas, patterns of treatment, including the effectiveness of a nurse-led unit, may differ from what we observed.

Finally, a fully rounded view of the unit's effects can be gained only by integrating results from the economic, qualitative, and process studies with those presented here.

Implications of study

Our finding regarding length of stay has implications for the development of intermediate care. It suggests that if acute hospital trusts became responsible for post-acute transitional care, community hospital trusts would find resources released to develop strategies to avoid patient admissions. This may constitute an overall benefit for the NHS, but a whole-systems assessment of responsibilities along the full continuum of care is needed. For example, acute trusts may identify organisational benefits to maintaining an in-house transfer destination for those patients whose pre-admission status predicts complex discharge arrangements or slower than average recovery time. This could be especially true if, as in many inner cities, they lack the option of community beds. Clearly, however, there are also costs; hospital days are more expensive in acute than community trusts.

The clinical outcomes we observed suggest that concerns about patient safety in the absence of routine medical involvement are unfounded. Nor did we see any shifts in service demand towards primary care. However, our results cannot support claims made for the therapeutic benefits of a separate nurse-led unit above and beyond usual nursing care. There were some indications that readmission rates might have been lower for the patients discharged from the nurse-led unit, the value of which should not be minimised. If true, lower readmission rates would mean less disruption in the recovery process for a potentially large set of predominantly elderly patients, which would have related cost savings. However, this finding was only suggestive and should be treated with caution. The outcomes of functional status, dependency of living arrangement, and mortality were no better for patients cared for in the nurse-led unit than for control patients. Other randomised controlled evaluations of nurse-led units have also failed to show effectiveness in either organisational or patient outcomes. 6 12 13

Conclusions

At a time when the government has committed an additional £900m a year over a five year period to intermediate care, our central finding is that nurse-led post-acute care in hospital is no worse, but also no better, than usual care and follow up. To make the most of the new resources, the place of acute hospitals in the provision of intermediate care will need to be examined closely.