MANY emergency departments (EDs) have

Transcription

1 54 SOCIAL WORK Gordon COST BENEFIT OF SOCIAL WORK SERVICES Cost Benefit Analysis of Social Work Services in the Emergency Department: A Conceptual Model JAMES A. GORDON, MD, MPA Abstract. Objective: To describe an economic model for formal cost benefit analysis of emergency department (ED)-based social services. Methods: The varied monetary costs and benefits associated with EDbased social work services were projected for three hypothetical levels of ED volume (30,000, 60,000, and 90,000 patients/year). Primary benefits included the prevention of return ED visits, the prevention of social hospital admissions, and the protection of doctor and nurse time. The primary cost was salary support for full-time social work staffing. Sensitivity analysis was performed to account for varying estimates. Results: For a small-volume ED, total benefits to offset costs ranged from $43,869 to $81,504, yielding a net cost of $99,936 up to $137,571 for full-time social work coverage. For a moderate-size ED, total benefits ranged from $87,660 to $162,930, yielding a net cost of only $18,510 on the high end of the sensitivity analysis, and $87,668 on the low end. For a largevolume ED, total benefits ranged from $131,529 to $247,434, yielding a net cost of $49,911 on the low end of the sensitivity analysis, but a net benefit of $65,994 on the high end. Conclusions: Dedicated social work staffing of EDs may yield net economic benefits, especially in large urban centers. Moderate-size EDs may almost break even in economic terms, and small EDs may realize a net cost, but in either case, the cost of social services can be significantly offset by decreased utilization of hospital and ED services, and by more efficient use of medical staff time. Key words: economic model; cost benefit; emergency department; costs; social services. ACADEMIC EMER- GENCY MEDICINE 2001; 8:54 60 MANY emergency departments (EDs) have begun to provide dedicated social work services for their patients. 1,2 The cost of such services, however, has become a major issue at many medical centers, hampering efforts to more fully deploy integrated support systems. 3 5 Although there have been some efforts to develop a billing system for ED social services in order to directly recoup costs, such a system is unlikely to become common practice. 6 This paper is intended to construct a formal model for the cost benefit analysis of uncompensated social work services in the ED. METHODS Study Design. A model was developed to approximate projected costs and benefits associated with the provision of social work services at three hy- From the Department of Emergency Medicine, Massachusetts General Hospital, Institute for Health Policy, Massachusetts General Hospital/Partners HealthCare System, Division of Emergency Medicine, Harvard Medical School, Boston, MA (JAG). Received May 11, 2000; revision received August 14, 2000; accepted September 12, Supported by the Robert Wood Johnson Clinical Scholars Program and the Department of Veterans Affairs. Address for correspondence and reprints: James A. Gordon, MD, MPA, Department of Emergency Medicine, Massachusetts General Hospital, 55 Fruit Street, Clinics 115, Boston, MA Fax: ; jgordon3@partners.org pothetical levels of ED volume: 30,000 patients per year (small/rural); 60,000 patients per year (moderate/suburban); and 90,000 patients per year (large/inner-city). The model was based primarily on data from five studies (Wrenn and Rice, 5 Keehn et al., 7 Boyack and Bucknum, 8 Williams, 9 and Hansen and Zwanziger 10 ). Presented below are the key variables and baseline estimates used in the model, and a description of the cost benefit factors used in the analysis. Key Variables and Baseline Estimates. The percentage of any given ED population who would see a social worker if such a service existed. This is a measure of the potential client base of a social work service in an ED, and was estimated to be 10% of the ED population, based on concordance from Wrenn and Rice, 5 Keehn et al., 7 and Boyack and Bucknum. 8 The number of return visits to the ED that may be prevented by social work intervention. This was estimated to be 4% of the social work client population (as above) based on data from Keehn et al. 7 A sensitivity analysis was provided for 2% and 6%. The marginal cost of a return visit to the ED. This was set at $88, based on work by Williams 9 in Michigan hospitals ( ). The number of hospital admissions that may be prevented by social work intervention. This was estimated to be 0.02% of the total ED volume, with a sensitivity analysis for 0.01% and 0.04%. This is

2 ACADEMIC EMERGENCY MEDICINE January 2001, Volume 8, Number 1 55 based on estimates ranging from 9 per year in a 60,000-visit ED (0.015%, Wrenn and Rice 5 )to30 per year in a 40,000-visit ED (0.075%, extrapolated from Boyack and Bucknum 8 ). Preventable admissions are called social admissions because there is no specific medical need for hospitalization, but circumstances such as lack of caretaker or appropriate home support system preclude discharge, and hospitalization provides additional time for appropriate home, rehabilitation, or skilled nursing facility arrangements to be made. The argument is that dedicated ED social work or case management can make these arrangements for the patient in the ED, enabling safe discharge and appropriate care without the need for hospitalization. The marginal cost of a day of hospitalization. This was set at $475 per day based on 1985 data from New York reported by Hansen and Zwanziger 10 in 1996 (reported to be $375 and corrected for 1993 dollars to $475, the common monetary year of analysis for this study). This amount is comparable to a 1993 State of Michigan estimate for hospital charge ($950/day), corrected to reflect marginal cost. 11 The number of days a patient with a potential social admission might spend in the hospital. This was set at 2.5 days on average, with a sensitivity analysis for 1 and 5 days. The literature provided an average length of stay at 7.5 (Wrenn and Rice 5 ) and 8.4 (Boyack and Bucknum 8 ), but this was believed to be a gross overestimation in the current health care climate, and so the adjustment was made downward. Medicare currently requires a three-day hospitalization before authorizing a stay in a skilled nursing facility. The average amount of time spent by a social worker with a client. This was set at 40 minutes, including paperwork. Data from Wrenn and Rice 5 show the average time to be 64 minutes (median 50 minutes, IQR 30 85), while Boyack and Bucknum 8 recorded minutes. The value chosen seemed a reasonable estimate based on the two sources and the author s experience, intended to produce a conservative estimate of work productivity. [Assuming a lower (30 minutes) or higher (50 minutes) estimate would change the cost benefit tabulations only very slightly (about $3,000 for a small ED; $9,000 for a large ED).] If one patient were seen every 40 minutes then 13,140 patients could be seen maximally in one year (number of hours in one year/ hours per patient). This seems a reasonable estimate for the maximum number of clients who can be handled in any ED. (The hypothetical 90,000-visit ED would handle about 9,000 social work clients, providing some cross-validation for 10% as a figure to approximate the fraction of an ED population potentially seen by a social work service.) This figure does not take into account hourly peak volumes that may overwhelm the social work service. The fraction of potential social work clients who, in the absence of a social worker, receive some attention to social needs from a doctor or nurse. This number was set at 75%, and implies that if social work services are absent, 25% of the potential social worker clients would receive no social support services at all due to time constraints on the medical staff. This is of course an artificial distinction made for purposes of the model. In reality the medical staff, without a social worker, would endeavor to meet some of the social needs of most patients and all of the needs of a few patients, with very few patients getting no attention at all. This estimate is without primary data support, but seems a reasonable representation of the spectrum of service described above, allowing the model to at least recognize this type of time cost trade-off. 6 The amount of time spent by medical staff on social concerns in the absence of a social worker. This was set at 20 minutes average, assuming that the medial staff spends half as much time on social issues as would a social worker. This was suggested as a reasonable conservative estimate by Ponto and Berg. 6 It was assumed that such limited intervention was for a bare minimum of social care, and had negligible preventive effects, either on return visits or on social hospital admissions. Of the time spent by medical staff on social concerns in the absence of a social worker, the percentage that is nursing vs physician time. It was assumed that 75% of this care is provided by nursing staff, and 25% by physicians. This estimate is also without primary data support, but seems a conservative split between physicians and nurses in terms of attention to social need. The hourly wage rate of nurses and physicians. This was taken from the locally accepted standard in the early 1990s, at $25/hour for nurses and $90/ hour for physicians, with allowance for an additional 20% fringe benefit calculation. The full-time equivalent (FTE) salary for a social worker (SW). This was estimated to be about $36,000 in the early 1990s. For 24-hour coverage in the ED, 4.2 FTEs are required for a total cost of $181,440 (including 20% fringe). 12 Cost Benefit Tabulations. For each of the three hypothetical EDs, three types of potential benefit from the availability of social work services were calculated: Benefit 1: The prevention of return visits to the ED by appropriate social service intervention. The value of this benefit (for one year) was calculated by: [no. return visits prevented] [average marginal cost of ED visit], where [no. return visits pre-

3 56 SOCIAL WORK Gordon COST BENEFIT OF SOCIAL WORK SERVICES vented] = [(base ED volume) (% seen by SW)] [% of SW population whose return visit is prevented] Benefit 2: The prevention of social hospital admissions. The value of this benefit (for one year) was calculated by: [no. of admissions prevented] [average marginal cost/day] [no. days spent in hospital] Benefit 3: The salvage of nurse/physician time spent doing social work activity. The value of this benefit (for one year) was calculated by: [(hr/yr spent by nurse) (wage/hr nurse 20% fringe)] [(hr/yr spent by physician) (wage/hr physician 20% fringe)] where [hr/yr spent by nurse] = [total hr spent by nurse/physician] [fraction of total accounted by nurse], [hr/yr spent by physician] = [total hr spent by nurse/physician] [fraction of total accounted by physician]; and [total hr spent by nurse/physician = [no. patients whose social needs addressed by nurse/physician in absence of SW] [hr spent by nurse/physician per patient]; and [no. patients whose needs addressed by nurse/physician in absence of SW] = [no. patients who would be seen by SW] [fraction who have needs addressed by nurse/ physician in absence of SW] Cost: Salary support for full-time social work services. The value of this cost (one social worker in ED at all times, for one year) was calculated by: [no. FTEs required for full staffing] [yearly salary/ FTE 20% fringe], where [no. FTEs required for full staffing] = [total hr in year]/[hr one person works/yr assuming 40-hr week] Study Setting and Population. Description of the primary data source populations: Wrenn and Rice s data, 5 contributing to estimates for the social work client base (8.7% of total ED volume), preventable social admissions (9), and time spent by social worker (64 minutes), were accrued during a six-week period in 1992 at the University of Rochester/Strong Memorial Hospital ED, in urban Rochester, New York. Annual ED volume stated 60,000. [Staffing was only at 60% (no weekday overnights, no weekend evenings/overnights). Wrenn and Rice s estimate of 5% of ED patients seen by social work was extrapolated upward to account for the partial staffing hours and patient pool).] Keehn et al. s data, 7 contributing to estimates for social work client base (12.5%, accounting for staffing levels/times) and prevention of return visits (4% of social work client base), were accrued during a 12-month period in , at the University of Michigan Medical Center ED in Ann Arbor, Michigan. Annual volume at that time was 40,000, now up to 60,000. Boyack and Bucknum s data, 8 contributing to estimates for social work client base (11.6%), preventable social admissions (extrapolated low estimate of 30), and average time spent by a social worker (20 45 minutes), were collected during a four-month period in late 1988 at a community hospital ED in Calgary, Alberta, Canada. Annual volume extrapolated to 40,000. Williams data, 9 contributing to the estimate for average marginal cost of an ED visit ($88), were based on analysis of six community hospitals in Michigan between 1991 and Hansen and Zwanziger s data, 10 contributing to the estimate for marginal cost of a day of hospitalization, were based on analysis of New York hospitals in 1985, reported in This amount ($375) was corrected upward to 1993 dollars ($475) using a standard accounting formula. Measurements and Outcome Variables. The main outcome measure for this model was aggregate cost benefit in dollars for one year, calculated by: [benefit 1] [benefit 2] [benefit 3] [cost of SW staff] = net benefit (or cost) This outcome variable was calculated for each of the three hypothetical EDs based on individual volume levels. Costs and benefits are assumed in this model to accrue to one entity, such as an integrated health system that owns and manages all components of its system, including its own insurance product (as is common among moderate to large teaching hospital systems). In this kind of system, the organization benefits when its insured members use the home hospital or ED less frequently, and when the ED doctors and nurses can make more efficient use of their time; the organization also bears the cost of providing additional social work services. This model is, by design, based primarily on measurable economic costs and benefits described in the literature. Although some of the primary benefits of enhanced social services may be invaluable, and may even constitute independent justification for providing such services (by helping avoid psychological crises, improve quality of life, and foster more efficient resource utilization) to the best of the author s knowledge, they have yet to be formally quantified in the ED literature, and are not included in this formal economic model.

4 ACADEMIC EMERGENCY MEDICINE January 2001, Volume 8, Number 1 57 Data Analysis. A basic cost benefit tabulation was performed for a single year. All dollar figures are assumed to be in 1993 terms, the last year in which much of the primary data were collected (1985 hospitalization costs were corrected upward to 1993). The model was constructed for full-time social work services (7 days a week, 24 hours a day). Sensitivity analyses were performed for the following variables: 1. percentage of return visits that may be prevented by social work intervention (2%, 4%, 6% of base social work client population), 2. annual number of admissions prevented by social work intervention (0.01%, 0.02%, 0.04% of yearly ED volume), and 3. number of days spent in the hospital for a social admission (1, 2.5, 5). High and low end of sensitivity analysis refers to the extreme limit of the estimate for a given variable, given the range of variables used for the estimate. For monetary benefit, the low end of the sensitivity analysis refers to the lowest amount of money saved, while the high end refers to the highest amount saved. For this analysis, costs (a negative monetary value) are fixed. No sensitivity analysis is needed. The low end of the sensitivity analysis for net figures (summary benefit minus cost) is the lowest number; if negative, it represents a cost, and if positive, it represents a net benefit. The high end of the sensitivity analysis for net values represents the highest number. For example, the more positive, the higher the benefit and lower the cost. RESULTS The primary cost benefit tabulation for one year of full-time ED social work services is provided in Table 1. The uniform cost for full-time social work coverage by a single social worker is $181,440. For a small-volume ED (30,000 visits/year), total benefits to offset costs range from $43,869 to $81,504, yielding a net cost of $99,936 up to $137,571 for full-time social work coverage. For a moderate-size ED (60,000 visits/year), total benefits range from $87,660 to $162,930, yielding a net cost of only $18,510 on the high end of the sensitivity analysis, and $93,780 on the low end. For a large-volume ED (90,000 visits/year), total benefits range from $131,529 to 247,434, yielding a net cost of $49,911 on the low end of the sensitivity analysis, but a net benefit of $65,994 on the high end. DISCUSSION Although social and environmental concerns are vital to health, 13,14 the integration of social services and health care remain subject to the economic realities of the marketplace. This paper was intended to provide a framework for analyzing the pure economic benefits of full-time social work services at three levels of ED volume, independent of compelling but currently unquantified measures (alleviation of suffering, enhanced well-being). While the model is driven by assumptions that have not been prospectively validated, the construction of a data-driven model, in and of itself, provides a basis for more informed discussion. Only when an integrated hospital system continues to see a large out-of-network fee-for-service population might the costs and benefits of ED social work services accrue to different entities. In this case, the hospital generates more revenue when patients use more services. Provision of preventive social work services is thus a cost to the hospital without benefit, and a benefit to the payor without cost. However, if the hospital system serves a largely uninsured population (as do many urban and rural centers), then all the benefit from preventive social services shifts back to the hospital, because less hospital utilization means less charity (i.e., uncompensated) care. Based on the assumptions in the model, large urban EDs may operate full-time social work services at a net benefit. Moderate-size EDs seem more likely to almost break even or incur moderate cost by implementing full social work services, while small or moderate-size EDs may experience net loss. In each case, the cost of services is largely offset by the benefits. Although this model was based on full-time social work services, the monetary benefit of parttime social work service may be approximated in a moderate-size to large ED by borrowing the benefit figures from the small ED analysis (cost would be based on personnel costs for part-time social workers). Some portion of the benefits cited here may also be achievable by a lesser paid case manager, mental health specialist with public health training, or even perhaps pastoral care staff, especially in lower-volume EDs where the cost benefit ratio for full-time social work staff may be less favorable. Moreover, nonintegrated hospitals (such as a community facility that contracts for an ED provider group) can also use this analysis to inform their practice: both costs and benefits as described here would still apply, but would have to be applied to appropriately different entities. LIMITATIONS AND FUTURE QUESTIONS The main limitations of this analysis lie in the assumptions and baseline estimates. Limited primary data have been collected on the main varia-

5 TABLE 1. Cost Benefit Tabulation for One Year of Full-time Social Work Services in an Emergency Department For complete reference citations, see the reference list. ED = emergency department; SW = social worker; RN/MD = nurse/physician; FTE = full-time equivalent. 58 SOCIAL WORK Gordon COST BENEFIT OF SOCIAL WORK SERVICES

6 ACADEMIC EMERGENCY MEDICINE January 2001, Volume 8, Number 1 59 bles in the model, with only one to three primary data sets contributing to some key variables. Corroboration of values between studies was sought wherever possible. Data collected over months were extrapolated over a year, assuming seasonal uniformity, which may not, in fact, exist. Data collected during parts of the day were extrapolated to estimate data for a full day, also assuming a diurnal uniformity. A correction factor for diminishing benefit with increased patient flow was not applied. Calculation of the hourly maximum ED volume that would begin to overwhelm the capacity of a single social worker to provide good service was not available from the data, although most primary data were collected during peak volume times. Conservative estimates, favoring cost over benefit, were used as a baseline when data were unclear, extrapolated, or believed to be inapplicable in the current health care climate. While sensitivity analysis was provided for some variables, it was not for others, based on what was thought to be the strength and reasonableness of the estimates. Calculation of the base social work client base, set at 10% of the overall ED population, was reasonably corroborated by three data sets, but may in fact be an overestimate in many centers. The decrease in return visits, set at 4% of social work client base, is supported by data in only one study (Keehn et al.) a sensitivity analysis was provided for half of this number. While the sensitivity analysis goes from 6% to 2%, the true effect may be 0%. Even in this worst-case analysis of the data, the net benefit would decrease by only about $5,000 to $15,000 beyond the current sensitivity estimate. Hospital stay and cost data vary from institution to institution. The marginal cost of a hospital day used here is a conservative estimate based on a 1993 correction of Hansen and Zwanziger s 1985 analysis for New York hospitals, which was corroborated by Michigan hospital charges during the same time period, corrected for marginal cost; however, Hansen and Zwanziger also studied California, where marginal costs were twice as high. Average marginal cost data for an ED visit were used based on Williams Michigan data, although he did additionally stratify visits as non-urgent, semi-urgent, or urgent. Perhaps the most labile part of the model revolves around the nurse and physician time saved by social work presence in the ED. In the absence of social work services, it was assumed in the model that the medical staff would intervene in 75% of the social need cases, and would spend only about half the time with these patients on social matters as would a social worker. Time spent by medical staff in these cases was split 75% to the nursing staff and 25% to physicians. These estimates are without primary data support, but seem reasonable estimates based on the author s experience intended only as one representation of the varied range and intensity of social services that could be provided primarily by the medical staff. Finally, while estimated salaries and compensation were based on a local market, they are reasonably representative of national averages during the time frame of analysis. Fringe benefits were included for all personnel, although nurses and social workers (as hospital-paid employees) are more likely to receive uniform benefit packages; emergency physicians have more varied arrangements based on their practice model. Even if no fringe benefits were assumed for physicians, the overall impact on reported calculations would be relatively small ($3,000 [small ED]; $7,000 [mid-size ED]; $10,000 [large ED]). Future studies would involve large-scale, multicenter prospective, randomized, controlled studies to assess the validity of the costs and benefits estimated here; case control studies would also be a robust, less costly alternative. Measuring the impact of a potential increase in demand for community social services as a result of vigorous ED social work would add breadth to the model. Factoring in a measure of personal health and wellbeing would also be useful. Such measurement may be possible by piloting dedicated social service units in EDs, serving not only to screen for social need in the medical population, but also to centrally coordinate health-related social services in a community. Such a center might be known as a social triage unit. 14 CONCLUSIONS Regardless of specific estimates, the cost benefit model outlined here can serve to inform the discussion about the economic impact of social work services in the ED. Dedicated 24-hour social work staffing of EDs may yield net economic benefits to a hospital system, especially in large urban centers. Moderate-size EDs may almost break even in economic terms, or small to moderate-size EDs may realize a net cost, but in either case, the cost of social services can be significantly offset by decreased utilization of hospital and ED services, and by more efficient use of medical staff time. The author thanks Kenneth Warner, PhD, University of Michigan School of Public Health, Ann Arbor, MI, for his assistance in completing this work. References 1. Gordon JA. The hospital emergency department as a social welfare institution. Ann Emerg Med. 1999; 33:321 5.

7 60 SOCIAL WORK Gordon COST BENEFIT OF SOCIAL WORK SERVICES 2. Gordon JA. Emergency care as a safety net [letter]. Health Aff. 2000; 19: Gordon JA, Goldfrank LR, Andrulis DP, D Alessandri RM, Kellermann AL. Emergency department initiatives to improve the public health [including Further thoughts from the reviewers ]. Acad Emerg Med. 1998; 5: Birnbaum A, Calderon Y, Gennis P, Rama R, Gallagher EJ. Domestic violence: diurnal mismatch between need and availability of services. Acad Emerg Med. 1996; 3: Wrenn K, Rice N. Social-work services in an emergency department: an integral part of the health care safety net [including Further thoughts from the reviewers ] Acad Emerg Med. 1994; 1: Ponto J, Berg W. Social-work services in the emergency department: a cost benefit analysis of an extended coverage program. Health Soc Work. 1992; 17(Feb): Keehn DS, Roglitz C, Bowden ML. Impact of social work on recidivism and non-medical complaints in the emergency department. Soc Work Health Care. 1994; 20(1): Boyack V, Bucknum AE. The quick response team: a pilot project. Soc Work Health Care. 1991; 16(2): Williams R. The costs of visits to emergency departments. N Engl J Med. 1996; 334: Hansen KK, Zwanziger J. Marginal costs in general acute care hospitals: a comparison among California, New York, and Canada. Health Econ. 1996; 5: Smith DG, Wheeler JR, Cameron AE. Benefits of hospital capacity reduction: estimates from a simulation model. Health Serv Manage Res. 1996; 9: Personal communication, Department of Social Work, University of Michigan Health System, Ann Arbor, MI, Bunker JP, Gomby DS, Kehrer BH (eds). Pathways to Health: The Role of Social Factors. Menlo Park, CA: Henry J. Kaiser Family Foundation, Gordon JA, Chudnofsky CR, Hayward RA. Where health and welfare meet: social deprivation among patients in the emergency department [abstract]. Acad Emerg Med. 1999; 6: 516. REFLECTIONS Half My Life Is Stat Half my life is stat. The other half: mounds of unread books, Leaning towers of unread journals, unpaid Bills amass on the kitchen counter. The stat life gulps time down In seconds or minutes, In quest of a fast turnaround. The other ponders, heavy, in sheer Befuddlement. Days to years to decades Pass without explication. (No tulips again this spring Without the well-planned bulb.) When half your life is stat, the rest Lengthens, slows, in no hurry To hurry. BONNIE SALOMON, MD Deerfield, Illinois