Economic evaluation of hospital at home versus hospital care: cost minimisation analysis of data from randomised controlled trialBMJ 1999; 319 doi: http://dx.doi.org/10.1136/bmj.319.7224.1547 (Published 11 December 1999) Cite this as: BMJ 1999;319:1547
- Jeremy Jones, lecturer in health economicsa,
- Andrew Wilson, senior lecturer ()b,
- Hilda Parker, research fellowb,
- Alison Wynn, research associateb,
- Carol Jagger, senior lecturerb,
- Nicky Spiers, fellow in health services researchc,
- Gillian Parker, professora
- a Nuffield Community Care Studies Unit, University of Leicester, Leicester LE1 6TP
- b Department of General Practice and Primary Health Care, University of Leicester, Leicester General Hospital, Leicester LE5 4PW
- c Department of Epidemiology and Public Health, University of Leicester, Leicester LE1 6TP
- Correspondence to: Andrew Wilson
- Accepted 15 November 1999
Objectives: To compare the costs of admission to a hospital at home scheme with those of acute hospital admission.
Design: Cost minimisation analysis within a pragmatic randomised controlled trial.
Setting: Hospital at home scheme in Leicester and the city's three acute hospitals.
Participants: 199 consecutive patients assessed as being suitable for admission to hospital at home for acute care during the 18 month trial period (median age 84 years).
Intervention: Hospital at home or hospital inpatient care.
Main outcome measures: Costs to NHS, social services, patients, and families during the initial episode of treatment and the three months after admission.
Results: Mean (median) costs per episode (including any transfer from hospital at home to hospital) were similar when analysed by intention to treat—hospital at home £2569 (£1655), hospital ward £2881 (£2031), bootstrap mean difference −305 (95% confidence interval −1112 to 448). When analysis was restricted to those who accepted their allocated place of care, hospital at home was significantly cheaper—hospital at home £2557 (£1710), hospital ward £3660 (£2903), bootstrap mean difference −1071 (−1843 to −246). At three months the cost differences were sustained. Costs with all cases included were hospital at home £3671 (£2491), hospital ward £3877 (£3405), bootstrap mean difference −210 (−1025 to 635). When only those accepting allocated care were included the costs were hospital at home £3698 (£2493), hospital ward £4761 (£3940), bootstrap mean difference −1063 (−2044 to −163); P=0.009. About 25% of the costs for episodes of hospital at home were incurred through transfer to hospital. Costs per day of care were higher in the hospital at home arm (mean £207 v £134 in the hospital arm, excluding refusers, P<0.001).
Conclusions: Hospital at home can deliver care at similar or lower cost than an equivalent admission to an acute hospital.
Managing the demand for hospital services by shifting activity elsewhere requires alternatives that can be justified on both clinical and economic grounds. Hospital at home is one such alternative, with a contested evidence base.1 In acute care, hospital at home can provide an alternative to inpatient care in two ways—early discharge of patients from hospital or avoidance of admission. The comparator adopted in most evaluations is the acute hospital, although this may not always be appropriate,2 and the use of average costs for inpatients has been challenged.3
Evaluations of early discharge of surgical patients to hospital at home care have suggested that it can save costs by reducing length of stay,4–6 although these savings may not always be achieved. More recent economic evaluations, conducted alongside randomised controlled trials, 7 8 failed to clarify the uncertainty, with one concluding that hospital at home provided care at lower cost than hospital9 whereas the other did not.10
No randomised controlled trials of hospital at home schemes to avoid acute admission have been published, despite demands for such evidence.11 Although the Kettering study included a small number of cases in which admission was avoided, these were not analysed separately.7
The processes of recruitment to the study, randomisation, and patient outcomes have been described elsewhere.12 The approach we adopted for the economic analysis was to calculate costs for the original episode and at three months from admission, following the convention regarding the intention to treat. Patients were costed as randomised, regardless of whether they accepted allocated care or were subsequently transferred to hospital. When patients transferred directly to nursing or residential care from hospital at home or inpatient care, this was included. We collected data using routine patient data for hospital at home and inputs from the community trust, additional encounter sheets (for inputs from general practitioners), and patients' questionnaires.
Hospital at home
We identified five main items in the use of resources for costing the stay of patients receiving hospital at home care. These were staff inputs, consumables, equipment (provided by the Red Cross on contract to the community trust), overhead costs (local scheme management and administration, car leasing and travel costs, the management and finance functions of the community trust), and capital costs associated with the scheme's health centre base.
Nurses' work study—We extracted hours of nursing and contact with therapists from patients' hospital at home notes and adjusted these in the analysis for staff time spent not in contact with patients, using information from a work study completed by nurses working on the scheme.
We based costing of patients' stay in hospital on the length of stay and the costs of specialty or ward For patients allocated to hospital who declined admission, any admission that occurred within seven days of the original referral was counted as an initial treatment episode.
Costs borne by patients and costs falling on family and friends
Concern has been expressed that hospital at home is made to seem cheaper than hospital care by shifting costs to patients and their families. The scope of analysis during the trial was limited to collecting descriptive information on who provided care during patients' stay in hospital and whether patients perceived home care as a burden in terms of increased heating and lighting of their homes, laundering, and other domestic arrangements.
Valuation of the use of resources
We used a combination of local and national sources to calculate unit costs. For nurses and therapists we adopted the methods outlined by the Personal Social Services Research Unit.13 We costed staff inputs at the midpoints of the appropriate salary scales, with employer costs of superannuation and national insurance added. Further additions were needed to take account of direct and indirect revenue costs. For hospital at home staff, we estimated the costs to cover the local management and administration of the scheme (for the manager and secretary, plus a divisional overhead) and the management and finance functions of the community trust. For physiotherapists and occupational therapists we used estimates from Unit Costs of Community Care.13
The cost of storing and delivering the equipment provided to patients in hospital at home care is subsumed in the non-pay budget heading—the value of the contract between the community trust and Red Cross has been included in full. The cost of equipment provided to patients' homes during the study has been calculated at replacement cost divided by the length of equipment's expected life (years), using a 6% discount rate.
We adopted the cost minimisation form of economic analysis as the clinical trial report showed no significant differences in outcome.12 We report the mean (SD) use of resources and standard deviations for initial admissions and the mean and median costs per case in each arm of the trial for the initial episode and at three month follow up Since cost data per patient (but not per day of care) are typically highly skewed, we used bootstrap estimation to derive a 95% confidence interval for average cost.14 15 We also used estimation in addition to a standard t test on the mean difference in cost between hospital at home and acute hospital care.
Altogether 199 patients were randomised, 102 to hospital at home and 97 to hospital. Median age was 84 years, and 71% were female. After randomisation six patients in the hospital at home arm and 23 in the hospital arm declined admission to their allocated place of care. These “refusers” were kept in the study, and any care they received in the three month follow up period was costed.
Use of resources
Table 1 shows the average use of resources in each arm of the trial. This includes days of care and number of journeys undertaken by patients in both arms and, for the group in hospital at home, a detailed breakdown of minutes of care by hospital at home staff and number of visits to the general practitioner. Length of stay, including any hospital transfer, was shorter in the hospital at home group.12 The average number of visits by the general practitioner to patients in hospital at home was 0.9.
Nurses' work study
During the nurses' work study (September-October 1996) a total of 12 patients were admitted to or being cared for in hospital at home. These patients had a total length of stay in hospital at home of 66 days (during the work study) and received 690 hours of nursing care. The ratio of time not in contact with patients to contact time varied from 1.8:1 for B grade staff to 8.0:1 for G grade staff. This was used to derive a cost per hour of contact for each staff grade as shown in table 2, which also includes a sensitivity analysis showing the effect of different ratios of contact time to non-contact time.
Costs of episodes
Table 3 shows average costs for patients in each arm of the trial. These were calculated in two ways—firstly, for all randomised patients, including those who refused their allocated place of care, and, secondly, with these refusers excluded. Hospital at home provided an episode of care more cheaply than an acute hospital, although this difference was marginal when patients who refused their allocated place of care were included.
Cost per day for hospital at home was significantly higher than for acute hospital care, reflecting a greater intensity of nursing input and a lack of the economies of scale attainable on a hospital ward Estimated cost per day for hospital at home includes hospital costs of those patients who refused hospital at home care and those who transferred directly to hospital. Hence it reflects the pattern of care provided to patients randomised to hospital at home and does not give an estimate of the cost per day purely of hospital at home.
Costs at three month follow up
Table 3 also shows costs at three months after randomisation. Patients allocated to and accepting hospital at home continued to have lower costs than those allocated to and accepting hospital treatment. Costs for the two arms were similar when refusers (all but three of whom received some care during the three months of follow up) were included in the analysis.
Contribution of different services to total costs
Table 4 shows the distribution of costs across services at the end of the initial episode of care and at three months' follow up. As expected, the nursing costs dominated the costs of hospital at home; the other main components were admissions to hospital and, to a lesser extent, nursing and residential care. Hospital costs dominated the costs of the initial episode of care for patients randomised to inpatient care, and costs of residential or nursing care constituted a similar proportion of total costs in this arm, as for hospital at home Community inputs for physiotherapy and occupational therapy constituted a comparatively small component of total cost, but they were more apparent in hospital at home than hospital care.
The costs reported here are sensitive to assumptions incorporated into the analysis. Several analyses tested the robustness of the study results to changes in certain important variables.
In the nurses' work study the comparatively high ratios of time not in contact with patients to contact time (table 2) show that the scheme was running under capacity during the trial Changing the balance of work within the scheme in favour of more direct nursing care reduced the mean cost to £2029.53 (£1130.59) and gave a cost per day for the initial episode of hospital at home that was similar to that for the hospital care when refusers were excluded (table 3). Repeating this analysis for costs at three months gave a mean cost of £3028.89 (£2023.28).
Payment for night work for hospital at home includes a weighting for unsociable hours. This was not included in the original analysis. The effect on cost of including this weighting was slight: it increased mean cost for the initial hospital at home episode to £2693.13 and median to £1654.57.
A reduction of 25% across the board in hospital costs reduced average costs for both arms, since some hospital at home patients transferred in. Hospital costs remained higher than for hospital at home when refusers were excluded.
Hospital at home had little input to the domestic care of patients—with the exception of washing the patient. This may point to hospital at home adopting a role oriented towards patients' personal care, in addition to their nursing needs, leaving domestic support to the usual carer, to patients themselves, or to their relatives.
Patients' perceptions of other costs associated with hospital at home varied: 44% reported using more lighting, 30% more laundry, 27% more heating, and 17% more hot water. In general the patients ascribed the use of extra utilities to the actions of night nurses, with patients themselves having little choice in the matter.
The economic analyses suggest that care can be provided in patients' homes using the model of hospital at home, to avoid hospital admission, at the same or lower cost than an equivalent admission to hospital. The sensitivity analyses on hospital at home nurses' contact time with patients compared with their non-contact time suggest that the level of recruitment to the trial had an impact on the estimated cost per case for hospital at home. Adopting a balance of nursing work that might be more plausible for an established hospital at home service reduced the estimated cost per day for hospital at home to a level similar to that estimated for hospital care.
What is already known on this topic
Economic evaluations of schemes for hospital at home care after early discharge have produced conflicting results. The cost of schemes to avoid admission compared with costs of hospital admission has not previously been assessed in a randomised trial.
What this paper adds
Patients allocated to hospital at home and hospital care incurred similar costs. Restricting analysis to patients accepting their allocated care showed that an episode in hospital at home was cheaper than hospital, and this cost difference was sustained over three months.
Hospital at home has the potential to provide care more cheaply than admission to hospital.
An examination of the contribution of different services to the costs of care of patients in the trial indicated an important role for acute hospital care in the management of patients in hospital at home Costs for acute hospital care represented 26% of the costs for the initial episode in patients randomised to hospital at home and 37% of costs at three months.
This study shows that hospital at home may provide a viable alternative to acute hospital when viewed in the long term. Currently the service runs as a complement to hospital care, but it may have a role in managing demand for hospital admission and can provide an acceptable form of care for patients who do not want admission to hospital.
This study would not have been possible without the cooperation of Fosse NHS Trust, Leicestershire Health, participating general practitioners, the acute hospitals, Leicestershire Bed Bureau, and, most crucially, the Hospital at Home Service itself.
Contributors: JJ was responsible for the design and collection of data on workload and health economics, was principal writer of the paper, and is its guarantor. AWi was responsible for the design and completion of the study. HP managed the trial, collected data, and assisted in analysis and interpretation AWy contributed to data collection, entry, and analysis. NS and CJ provided statistical advice for the protocol and undertook data analyses. GP contributed to study design and interpretation of results.
Funding National R&D Programme, Primary-Secondary Care Interface, NHS Executive, North Thames.
Competing interests None declared.