Twenty beds at Søbstad Nursing Home were re-assigned in late 2002 to be a community hospital performing intermediate care, which included increased numbers of trained nurses from 12.5 to 16.7 man-labours per week and doctors' hours, performed by three general practitioners, from 7 hours to 37.5 hours per week. All employees underwent a training programme provided by the general hospital. The department was also upgraded with laboratory facilities including intravenous pumps, equipment for continuously monitoring of oxygen-saturation in blood, laboratory equipment to measure infectious variables, hemoglobin and glucose in blood. Other blood tests could be delivered each day to the main laboratory at the general hospital with answers provided within the same working day.
The city general hospital in Trondheim, St. Olavs University Hospital, is both a general hospital for the municipality of Trondheim and a university hospital for the three counties in Mid-Norway. In this trial the function as a general hospital was included.
Intermediate care intervention
The experimental intervention was based on individualised intermediate care including evaluation and treatment ("care" and "cure") of each patient's diseases . However, the main focus was to improve the patients' ability to manage daily activities when returning home.
On admission to the community hospital the physicians performed a medical examination of the patients and a careful evaluation of available earlier health records from the admitting general practitioner, the general hospital physicians and the community home care services. The communication with each patient and his family focusing on physical and mental challenges was also essential to understand the needs and level of care.
The care at the different departments at the general hospital and the communication with primary health care followed the standard routines through the formal organisation.
Intermediate care at the community hospital was compared to conventional care in general hospital beds at medical, surgical and orthopedic departments.
Before the trial started participating physicians and nurses at the general hospital together with general practitioners and community nurses developed inclusion criteria through a Delphi technique . One of the authors (HG) facilitated requests for proposals and organised the proposals received, and was responsible for communication between the participants. Eventually, there were four inclusions criteria as eligible participants should be; 1) patients aged 60 years or more admitted the general hospital due to an acute illness or an acute exacerbation of a known chronic disease, 2) probably be in need of inward care for more than three to four days, 3) admitted from their own homes and 4) expected to return home when inward care was finished. Exclusion criteria were severe dementia or a psychiatric disorders needing specialised care 24 hours a day.
When an eligible patient was identified and accepted for inclusion, a blinded randomisation was performed by the Clinical Research Department at the Faculty of Medicine using random number tables in blocks to ensure balanced groups.
The number of deaths was monitored continuously during the whole trail as it was decided prior to the study that an increase in number of deaths at the community hospital should terminate the study.
Outcome variables were number of readmissions for the same disease, need of community home care and need of long-term nursing home. Readmissions for the same disease, according to the national definition, are defined as acute, non-planned admissions within 60 days for the same disease. Number of days in institutions after randomisation, number of deceased patients and days before death were assessed as well. All data were collected by one of the authors, (HG), according to prepared schemes from patients' medical records at the hospitals and at primary health services. The assessments of days in institution, readmissions and cause-specific deaths were monitored through the patient administrative systems, independent of treatment groups.
Two specially trained nurses monitored physical functioning (ADL) on 72 items with scores from one to four in each item, both at the intermediate department and at the general hospital, by a national system, Gerix . With an average ADL of one the patient is functioning perfectly in all areas, whereas an average score of four indicates a need of excessive help and care in all aspects of daily living. ADL was assessed for all patients prior to the inclusion to the trial, and the ADL was used as covariate or confounder in the multivariate analysis. General hospital doctors set the diagnosis at all patients prior to randomisation.
The Regional Committee for Medical Research Ethics for Central Norway approved the study, the patient information and the consent schemes. The study was granted license by the Norwegian Data Inspectorate to process personal health data. Each participating patient signed a written informed consent formula at the general hospital prior to the inclusion to the study.
The sample size was estimated to detect a difference of 25 per cent in the number of readmissions for the same disease, as an assessment of morbidity, between the groups with alpha 0.05 and power of 0.80. To achieve this we needed 65 patients in each group, altogether 130 patients.
All data are presented an analysed according to the CONSORT checklist (see Additional file 1). The comparisons between the intervention and control group were analysed as intention-to-treat analyses according to the CONSORT instructions. Some results from treatment analyses, dependent on where the patient received his treatment, are also presented.
We undertook all analyses using SPSS version 14.0. for Windows. Survival curves were estimated by Kaplan-Meier. The distribution of continuous variables was tested by comparing means and medians and by normality plots. Differences in number of patients with readmissions for the same disease and need of home care services or nursing homes between groups were tested by chi square tests, and differences in mean number of days in institution were tested both by paired t-test and by Wilcoxon signed rank test. Differences in readmissions and need of home care or nursing home were also analysed in logistic models adjusted for gender, age, ADL score and diagnosis. Hosmer and Lemeshows goodness of fit test tested the fit of the logistic models. The number of days in institution was compared between groups using covariance analyses with age, gender, ADL scores and diagnoses as covariates. The level of significance was set to p = 0.05.