Economic evaluation of nurse led intermediate care versus standard car
http://www.100md.com
《英国医生杂志》
1 School of Nursing and Midwifery, University of Southampton, Southampton SO17 1BJ, 2 Department of Geriatric Medicine, University of Southampton, 3 Medical Statistics Group, Health Care Research Unit, Southampton General Hospital, Southampton SO16 6YD, 4 Elderly Care Research Unit, University Department of Geriatric Medicine, Southampton General Hospital
Correspondence to: B Walsh b.m.walsh@soton.ac.uk
We undertook an economic evaluation of nurse led intermediate care of medical patients recovering from acute illness based on a randomised controlled trial.7 The trial used a randomised consent design and was intended to detect moderate differences in length of stay, physical functioning, and destination after discharge. In total, 240 medical patients were recruited and allocated to nurse led inpatient care or standard care, although two refused consent for data collection for economic evaluation. Overall, 238 patients were followed up for six months. No differences were observed in any outcomes other than length of stay, which was significantly longer in the nurse led group.7 We compared inpatient costs and costs after discharge from randomisation to six months using an intention to treat analysis. Costs are in pounds sterling and were calculated for the financial year 1998-9. Costs were not discounted, owing to the short study period. We made cost valuations after literature review and after interviewing staff in relevant trust directorates and accounting departments. When attributed cost data could not be isolated, we obtained aggregated cost estimates from the trust.
Data collection
To allow time for accurate information on service use to have been recorded in trust databases, we collected data on utilisation retrospectively. The unit of analysis for length of stay was one day. We ed data on use of hospital resources (by location and by day) from the hospital's Patient Administration System database. Data on use of physiotherapy and radiology were collected directly from the databases within each of these departments. We identified participants by their hospital patient registration number, and accuracy of identification was confirmed by checking date of birth, name, and address. General practice staff collected the primary care data at the end of the six month study period using a standardised data extraction form. Interviews with patients provided information on changes in residence, specifically to an institutional setting, with dates verified by the care home. We selected a random sample of 10% of the study cohort to test for inter-rater reliability, which showed 100% agreement between two researchers for all data sources.
Cost specifications
Most cost estimates for the hospital came from the centralised contracts and costing department. Estimates of costs per occupied bed day by ward came from the relevant directorate finance department. The estimates comprise direct staff costs (calculated for each ward) and indirect costs, including use of pathology, occupational therapy, clerical support, and hotel and laundry services (apportioned between wards in a directorate). The estimates exclude outpatient attendances, for which we used as a source of valuation Unit Costs of Community Care.10
We measured hospital stay by ward for two time periods: the admission period (period 1), which extended from randomisation to first discharge home to any destination other than a hospital, and the readmissions period (period 2), which included all subsequent days spent in hospital during the six month study period. From the specialty database, we identified physiotherapy input for both time periods. We measured contacts by the day, and we assumed that they lasted 20 minutes unless otherwise specified in the records. Radiology input was also established on an individual patient basis for both periods. Other contacts with therapy services, hospital doctor and nurse time, and inputs from pathology were embedded in trust estimates of cost per occupied bed day in each ward. (Cost per occupied bed day for the nurse led unit was lower than for other specialist wards, but higher than that for general acute medical wards.) Several factors contributed to the higher average costs for the nurse led unit: the small size of the unit (10 beds compared with 20-30 on general wards); higher than expected costs for nursing staff owing to employment on average of a higher grade of nurse than the medical wards; and a location distant from the main hospital that complicated, and possibly artificially inflated, the attribution of overhead costs.
Episodes of short term care, such as attendances at outpatient clinics, day surgery, and visits to the accident and emergency department, were measured by attendance or procedure. Other NHS resources included community hospitals (both periods), contacts with surgery based general practitioners and community nurses, home visits by general practitioners and community nurses, and telephone contacts with community nurses and general practitioners. Contacts with general practitioners in surgery were assumed to last 10 minutes unless otherwise indicated. Other contacts and days spent in new institutional care were costed according to units provided in Unit Costs of Community Care (table 1).10
Table 1 Costs and data sources for resources in hospital and community settings
Output specifications
We found no significant differences in the primary outcomes of the randomised controlled trial other than length of stay.7 In that analysis of effectiveness, hospital days were taken as an outcome, whereas in the economic analysis here, hospital days are treated as an input and therefore no output specifications are required.
Analysis
Although our study was not planned as an equivalence trial, given the evidence of the clinical trial7 we have assumed no clinical gains from the nurse led unit. We therefore carried out a cost minimisation analysis. To calculate costs per category of resource use we multiplied utilisation data by unit costs, then aggregated the result to produce total costs for period 1, period 2, and total costs for the six month study period. We produced summary statistics for each cost variable (means, medians, and standard deviations). Two sample t tests were used to compare mean costs between the nurse led group and standard care group, with 95% confidence intervals. Groups were also compared using a regression model controlling for referring ward and sex, as in the analysis of the primary outcomes.7 Results of the two analyses were virtually identical. Regression analyses are not reported here.
For our sensitivity analyses we focused on inpatient and total costs—that is, from the perspective of secondary care—because this was the only area in which costs differed between the groups. Comparisons were drawn in the same way as with other cost estimates. The sensitivity analyses were carried out to test the effect of varying the cost per occupied bed day for the nurse led unit, as this was the cost with the highest leverage and also the estimate most vulnerable to questions about accuracy. We recalculated inpatient costs during period 1 and period 2 and total costs for the study period, according to the following assumptions, which are based on reasonable variability observed between wards within the directorate: nurse led unit cost per occupied bed day 15% lower (￡213.08); nurse led unit cost per occupied bed day 20% lower (￡200.54); and nurse led unit cost per occupied bed day 25% lower (￡188.01). These values are higher than the mean for acute general medical wards (￡146.19, or 42% lower than the nurse led unit cost per occupied bed day), but this is consistent with the higher grade of nursing staff used by the nurse led unit. Cost reductions of 15%, 20%, and 25% equate to reductions in length of stay of 5, 6.4, and 8 days, respectively.
The fourth assumption, that nurse led unit cost per occupied bed day was equivalent to the cost of a general practitioner led community hospital (60% lower, ￡100.50), allowed comparison with the least expensive feasible alternative to care in a nurse led unit. If the direction of results does not change under these assumptions, then the findings can be taken as robust.
Results
Both inpatient costs and total costs were significantly higher for nurse led inpatient care compared with standard care of medical patients on an acute ward. This finding holds true whether the acute trust perspective or wider NHS perspective is taken and in a range of cost estimate situations. The only cost saving was a reduction in the costs associated with using a community hospital. Conventionally, when a trial produces no significant differences between treatment and control outcomes, the option with the lower cost should be the preferred choice. In this case, the nurse led model of care should not be pursued.
As with other evaluations of this type, it was difficult for us to obtain detailed cost breakdowns for hospital resources inputs.11 Differences in estimates from different sources within the trust raised concerns over the accuracy of costs attributed to the nurse led unit and to the acute wards, which seemed higher than anticipated in both cases, although not unreasonable. The evaluation was strengthened by the inclusion of costs for community care and residential care. Prospective, bottom-up data collection would have been preferable, but would have severely restricted the feasible sample size, reducing power without necessarily removing the need to make large assumptions in cost estimates. Instead, sensitivity analyses helped overcome the potential problem of inaccurate estimations or unusual costs; the analyses presented include values similar to the lower costs reported in the only other study that reports on costs for this model of care.9
Sensitivity analyses showed a clear trend for higher costs with nurse led care, although the differences were not significant; at no point did the nurse led option become significantly less expensive than standard care. Even assuming that the nurse led unit could reduce costs by as much as 25%, the equivalent of reducing length of stay by eight days on average, treatment costs would not be lower than those from standard care. Cost implications are not, however, the only guide to practice in the NHS. The continued growth of this model of care suggests that acute trusts are willing to pay more to maintain an intermediate care option under their own management control, perhaps especially during periods of high demand for beds. When making decisions on the development of nurse led intermediate care it is necessary to consider whether costs for indirect care are apportioned fairly; where the medical directorate fits in the larger context of secondary care; what the opportunity costs are of the resources (for example, space or staff) used in the model's implementation; and the effects of economies of scale: as these units increase in size (as they have done locally since this evaluation), costs may decrease. Given that patient outcomes are satisfactory, and such units reduce pressure on acute beds, a certain overall increase in cost may be acceptable.
What is already known on this topic
Nurse intermediate led care for post-acute medical patients is becoming increasingly popular
Such care does not seem to improve patient outcomes, raising the question of whether the model of care is cost effective
What this study adds
Costs for acute trusts and total costs are higher with nurse led intermediate care, even under generous assumptions about cost reductions
Investment in intermediate care in community hospital rather than acute hospital settings may be more cost effective.
It is possible that the costs of the nurse led model could be reduced, not only by increasing bed numbers but by setting boundaries on maximum length of stay. Setting boundaries on hospital stay seems to be the one favoured by the UK government in its approach to intermediate care.12 However, costs or length of stay would have to be reduced substantially for nurse led inpatient care to be less expensive than standard care, and the changes we outline could have a negative effect on patient outcomes. Given that, in this evaluation at least, a part of the stay on a nurse led unit seems to be substituting for a period of stay in a community hospital,7 investment in intermediate care in the community hospital setting may be a more appropriate way forward for some trusts. Finally, increased efficiency might be possible through education of staff on the ideal model of care delivery. Currently a high grade (more expensive) mix of skill seems to substitute for such education, but without improving quality of nursing above that in standard care settings.13 14 Training would require additional investment, but could prove cost effective if the mix of skills could be altered or outcomes improved in the nurse led unit. The decision, however, must also take into consideration the wider context of intermediate care; other models of care may be both feasible and more cost effective.11
We thank for their cooperation the Southampton University Hospitals Trust; the participating general practitioners; the managers, clinicians, and patients involved with the nurse led unit; the finance, accounting, and information staff in the trust; and the support of the other members of the Southampton Nurse Led Unit Evaluation Team: J Bray, J Brooking, D Coulson, P Lees, J Pearce, K Postle, L Sheron, J Warr, and R Wiles.
Contributors: BW conceived and led the original inpatient trial and, with RMP, defined the trial protocol. AS obtained funding and conceived and led the comprehensive follow-up and economic evaluation. RMP contributed as statistician to the design and analysis of the study. JWB collected data and, with BW, AS, and RMP, assisted in analysis and interpretation. BW led the writing of the paper and is its guarantor. All authors were involved in critical revision.
Funding: Grant D/10/11.97 from the NHS Executive Research and Development Directorate South and West Region.
Competing interests: None declared.
Ethical approval: South and west local research ethics committee.
References
Department of Health. Shaping the future NHS: long term planning for hospitals and related services. London: DoH, 2000.
Department of Health. National service framework for older people. London: DoH, 2001.
Department of Health. The NHS plan. London: DoH, 2000.
Steiner A. Intermediate care: more than a nursing thing? Age Ageing 2001;30: 433-5.
Pearson A, Punton S, Durant I. Nursing beds: an evaluation of the effects of therapeutic nursing. Harrow: Scutari, 1992.
Griffiths P, Evans A. Evaluating a nurse-led inpatient service: an interim report. London: King's Fund, 1995.
Steiner A, Walsh B, Pickering RM, Wiles R, Ward J, Brooking JI, et al. Therapeutic nursing or unblocking beds? A randomised controlled trial of a post-acute intermediate care service. BMJ 2001;322: 453-60.
Griffiths P, Wilson-Barnett J, Richardson G, Spilsbury K, Miller F, Harris R. The effectiveness of intermediate care in a nursing-led in-patient unit. Int J Nurs Stud 2000;37: 153-61.
Griffiths P, Harris R, Richardson G, Hallett N, Heard S, Wilson-Barnett. Substitution of a nursing-led inpatient unit for acute services: randomized controlled trial of outcomes and cost of nursing-led intermediate care. Age Ageing 2001;30: 483-8.
Netten A, Dennett J. Unit costs of community care, 3rd ed. Kent: Personal Social Services Research Unit, 1999.
Coast J, Richards S, Peters T, Gunnell D, Darlow M, Pounsford J. Hospital at home or acute hospital care? A cost-minimisation analysis. BMJ 1998;316: 1802-6.
Department of Health. Health service circular 2001/01: intermediate care. London: DoH, 2001.
Wiles R, Postle K, Steiner S, Walsh B. Nurse-led intermediate care: an opportunity to develop enhanced roles for nurses? J Adv Nurs 2001;34: 813-21.
Walsh B, Steiner A, Warr J, Sheron L, Pickering R. Nurse-led inpatient care: opening the `black box.' Int J Nurs Stud 2003;40: 307-19.(Bronagh Walsh, lecturer1, Andrea Steiner)
Correspondence to: B Walsh b.m.walsh@soton.ac.uk
We undertook an economic evaluation of nurse led intermediate care of medical patients recovering from acute illness based on a randomised controlled trial.7 The trial used a randomised consent design and was intended to detect moderate differences in length of stay, physical functioning, and destination after discharge. In total, 240 medical patients were recruited and allocated to nurse led inpatient care or standard care, although two refused consent for data collection for economic evaluation. Overall, 238 patients were followed up for six months. No differences were observed in any outcomes other than length of stay, which was significantly longer in the nurse led group.7 We compared inpatient costs and costs after discharge from randomisation to six months using an intention to treat analysis. Costs are in pounds sterling and were calculated for the financial year 1998-9. Costs were not discounted, owing to the short study period. We made cost valuations after literature review and after interviewing staff in relevant trust directorates and accounting departments. When attributed cost data could not be isolated, we obtained aggregated cost estimates from the trust.
Data collection
To allow time for accurate information on service use to have been recorded in trust databases, we collected data on utilisation retrospectively. The unit of analysis for length of stay was one day. We ed data on use of hospital resources (by location and by day) from the hospital's Patient Administration System database. Data on use of physiotherapy and radiology were collected directly from the databases within each of these departments. We identified participants by their hospital patient registration number, and accuracy of identification was confirmed by checking date of birth, name, and address. General practice staff collected the primary care data at the end of the six month study period using a standardised data extraction form. Interviews with patients provided information on changes in residence, specifically to an institutional setting, with dates verified by the care home. We selected a random sample of 10% of the study cohort to test for inter-rater reliability, which showed 100% agreement between two researchers for all data sources.
Cost specifications
Most cost estimates for the hospital came from the centralised contracts and costing department. Estimates of costs per occupied bed day by ward came from the relevant directorate finance department. The estimates comprise direct staff costs (calculated for each ward) and indirect costs, including use of pathology, occupational therapy, clerical support, and hotel and laundry services (apportioned between wards in a directorate). The estimates exclude outpatient attendances, for which we used as a source of valuation Unit Costs of Community Care.10
We measured hospital stay by ward for two time periods: the admission period (period 1), which extended from randomisation to first discharge home to any destination other than a hospital, and the readmissions period (period 2), which included all subsequent days spent in hospital during the six month study period. From the specialty database, we identified physiotherapy input for both time periods. We measured contacts by the day, and we assumed that they lasted 20 minutes unless otherwise specified in the records. Radiology input was also established on an individual patient basis for both periods. Other contacts with therapy services, hospital doctor and nurse time, and inputs from pathology were embedded in trust estimates of cost per occupied bed day in each ward. (Cost per occupied bed day for the nurse led unit was lower than for other specialist wards, but higher than that for general acute medical wards.) Several factors contributed to the higher average costs for the nurse led unit: the small size of the unit (10 beds compared with 20-30 on general wards); higher than expected costs for nursing staff owing to employment on average of a higher grade of nurse than the medical wards; and a location distant from the main hospital that complicated, and possibly artificially inflated, the attribution of overhead costs.
Episodes of short term care, such as attendances at outpatient clinics, day surgery, and visits to the accident and emergency department, were measured by attendance or procedure. Other NHS resources included community hospitals (both periods), contacts with surgery based general practitioners and community nurses, home visits by general practitioners and community nurses, and telephone contacts with community nurses and general practitioners. Contacts with general practitioners in surgery were assumed to last 10 minutes unless otherwise indicated. Other contacts and days spent in new institutional care were costed according to units provided in Unit Costs of Community Care (table 1).10
Table 1 Costs and data sources for resources in hospital and community settings
Output specifications
We found no significant differences in the primary outcomes of the randomised controlled trial other than length of stay.7 In that analysis of effectiveness, hospital days were taken as an outcome, whereas in the economic analysis here, hospital days are treated as an input and therefore no output specifications are required.
Analysis
Although our study was not planned as an equivalence trial, given the evidence of the clinical trial7 we have assumed no clinical gains from the nurse led unit. We therefore carried out a cost minimisation analysis. To calculate costs per category of resource use we multiplied utilisation data by unit costs, then aggregated the result to produce total costs for period 1, period 2, and total costs for the six month study period. We produced summary statistics for each cost variable (means, medians, and standard deviations). Two sample t tests were used to compare mean costs between the nurse led group and standard care group, with 95% confidence intervals. Groups were also compared using a regression model controlling for referring ward and sex, as in the analysis of the primary outcomes.7 Results of the two analyses were virtually identical. Regression analyses are not reported here.
For our sensitivity analyses we focused on inpatient and total costs—that is, from the perspective of secondary care—because this was the only area in which costs differed between the groups. Comparisons were drawn in the same way as with other cost estimates. The sensitivity analyses were carried out to test the effect of varying the cost per occupied bed day for the nurse led unit, as this was the cost with the highest leverage and also the estimate most vulnerable to questions about accuracy. We recalculated inpatient costs during period 1 and period 2 and total costs for the study period, according to the following assumptions, which are based on reasonable variability observed between wards within the directorate: nurse led unit cost per occupied bed day 15% lower (￡213.08); nurse led unit cost per occupied bed day 20% lower (￡200.54); and nurse led unit cost per occupied bed day 25% lower (￡188.01). These values are higher than the mean for acute general medical wards (￡146.19, or 42% lower than the nurse led unit cost per occupied bed day), but this is consistent with the higher grade of nursing staff used by the nurse led unit. Cost reductions of 15%, 20%, and 25% equate to reductions in length of stay of 5, 6.4, and 8 days, respectively.
The fourth assumption, that nurse led unit cost per occupied bed day was equivalent to the cost of a general practitioner led community hospital (60% lower, ￡100.50), allowed comparison with the least expensive feasible alternative to care in a nurse led unit. If the direction of results does not change under these assumptions, then the findings can be taken as robust.
Results
Both inpatient costs and total costs were significantly higher for nurse led inpatient care compared with standard care of medical patients on an acute ward. This finding holds true whether the acute trust perspective or wider NHS perspective is taken and in a range of cost estimate situations. The only cost saving was a reduction in the costs associated with using a community hospital. Conventionally, when a trial produces no significant differences between treatment and control outcomes, the option with the lower cost should be the preferred choice. In this case, the nurse led model of care should not be pursued.
As with other evaluations of this type, it was difficult for us to obtain detailed cost breakdowns for hospital resources inputs.11 Differences in estimates from different sources within the trust raised concerns over the accuracy of costs attributed to the nurse led unit and to the acute wards, which seemed higher than anticipated in both cases, although not unreasonable. The evaluation was strengthened by the inclusion of costs for community care and residential care. Prospective, bottom-up data collection would have been preferable, but would have severely restricted the feasible sample size, reducing power without necessarily removing the need to make large assumptions in cost estimates. Instead, sensitivity analyses helped overcome the potential problem of inaccurate estimations or unusual costs; the analyses presented include values similar to the lower costs reported in the only other study that reports on costs for this model of care.9
Sensitivity analyses showed a clear trend for higher costs with nurse led care, although the differences were not significant; at no point did the nurse led option become significantly less expensive than standard care. Even assuming that the nurse led unit could reduce costs by as much as 25%, the equivalent of reducing length of stay by eight days on average, treatment costs would not be lower than those from standard care. Cost implications are not, however, the only guide to practice in the NHS. The continued growth of this model of care suggests that acute trusts are willing to pay more to maintain an intermediate care option under their own management control, perhaps especially during periods of high demand for beds. When making decisions on the development of nurse led intermediate care it is necessary to consider whether costs for indirect care are apportioned fairly; where the medical directorate fits in the larger context of secondary care; what the opportunity costs are of the resources (for example, space or staff) used in the model's implementation; and the effects of economies of scale: as these units increase in size (as they have done locally since this evaluation), costs may decrease. Given that patient outcomes are satisfactory, and such units reduce pressure on acute beds, a certain overall increase in cost may be acceptable.
What is already known on this topic
Nurse intermediate led care for post-acute medical patients is becoming increasingly popular
Such care does not seem to improve patient outcomes, raising the question of whether the model of care is cost effective
What this study adds
Costs for acute trusts and total costs are higher with nurse led intermediate care, even under generous assumptions about cost reductions
Investment in intermediate care in community hospital rather than acute hospital settings may be more cost effective.
It is possible that the costs of the nurse led model could be reduced, not only by increasing bed numbers but by setting boundaries on maximum length of stay. Setting boundaries on hospital stay seems to be the one favoured by the UK government in its approach to intermediate care.12 However, costs or length of stay would have to be reduced substantially for nurse led inpatient care to be less expensive than standard care, and the changes we outline could have a negative effect on patient outcomes. Given that, in this evaluation at least, a part of the stay on a nurse led unit seems to be substituting for a period of stay in a community hospital,7 investment in intermediate care in the community hospital setting may be a more appropriate way forward for some trusts. Finally, increased efficiency might be possible through education of staff on the ideal model of care delivery. Currently a high grade (more expensive) mix of skill seems to substitute for such education, but without improving quality of nursing above that in standard care settings.13 14 Training would require additional investment, but could prove cost effective if the mix of skills could be altered or outcomes improved in the nurse led unit. The decision, however, must also take into consideration the wider context of intermediate care; other models of care may be both feasible and more cost effective.11
We thank for their cooperation the Southampton University Hospitals Trust; the participating general practitioners; the managers, clinicians, and patients involved with the nurse led unit; the finance, accounting, and information staff in the trust; and the support of the other members of the Southampton Nurse Led Unit Evaluation Team: J Bray, J Brooking, D Coulson, P Lees, J Pearce, K Postle, L Sheron, J Warr, and R Wiles.
Contributors: BW conceived and led the original inpatient trial and, with RMP, defined the trial protocol. AS obtained funding and conceived and led the comprehensive follow-up and economic evaluation. RMP contributed as statistician to the design and analysis of the study. JWB collected data and, with BW, AS, and RMP, assisted in analysis and interpretation. BW led the writing of the paper and is its guarantor. All authors were involved in critical revision.
Funding: Grant D/10/11.97 from the NHS Executive Research and Development Directorate South and West Region.
Competing interests: None declared.
Ethical approval: South and west local research ethics committee.
References
Department of Health. Shaping the future NHS: long term planning for hospitals and related services. London: DoH, 2000.
Department of Health. National service framework for older people. London: DoH, 2001.
Department of Health. The NHS plan. London: DoH, 2000.
Steiner A. Intermediate care: more than a nursing thing? Age Ageing 2001;30: 433-5.
Pearson A, Punton S, Durant I. Nursing beds: an evaluation of the effects of therapeutic nursing. Harrow: Scutari, 1992.
Griffiths P, Evans A. Evaluating a nurse-led inpatient service: an interim report. London: King's Fund, 1995.
Steiner A, Walsh B, Pickering RM, Wiles R, Ward J, Brooking JI, et al. Therapeutic nursing or unblocking beds? A randomised controlled trial of a post-acute intermediate care service. BMJ 2001;322: 453-60.
Griffiths P, Wilson-Barnett J, Richardson G, Spilsbury K, Miller F, Harris R. The effectiveness of intermediate care in a nursing-led in-patient unit. Int J Nurs Stud 2000;37: 153-61.
Griffiths P, Harris R, Richardson G, Hallett N, Heard S, Wilson-Barnett. Substitution of a nursing-led inpatient unit for acute services: randomized controlled trial of outcomes and cost of nursing-led intermediate care. Age Ageing 2001;30: 483-8.
Netten A, Dennett J. Unit costs of community care, 3rd ed. Kent: Personal Social Services Research Unit, 1999.
Coast J, Richards S, Peters T, Gunnell D, Darlow M, Pounsford J. Hospital at home or acute hospital care? A cost-minimisation analysis. BMJ 1998;316: 1802-6.
Department of Health. Health service circular 2001/01: intermediate care. London: DoH, 2001.
Wiles R, Postle K, Steiner S, Walsh B. Nurse-led intermediate care: an opportunity to develop enhanced roles for nurses? J Adv Nurs 2001;34: 813-21.
Walsh B, Steiner A, Warr J, Sheron L, Pickering R. Nurse-led inpatient care: opening the `black box.' Int J Nurs Stud 2003;40: 307-19.(Bronagh Walsh, lecturer1, Andrea Steiner)