Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 321
Physician Staffing for the VA: Volume I 8 PROJECTING FUTURE PATIENT WORKLOAD A critical element in any application of the methodology presented in chapters 4 through 6 to determine future VA physician requirements is an estimate of future VA patient utilization (workload). This chapter focuses on the derivation of these estimates. Consistent with the VA's original request to the Institute of Medicine, the committee determined that its methodology should be capable of taking into account projected changes over time in the volume and case mix of workload resulting from the aging of the veteran population (Institute of Medicine, 1987). The methodology also should be flexible enough, in the committee's view, to incorporate the influence of other factors possibly affecting workload, such as the proportion of females in the veteran population or the distribution of veterans across eligibility-for-care categories.1 A detailed account of how the methodology can be used to determine future VA physician requirements, by specialty or program area, is presented in chapter 4, where illustrative applications are shown for four actual VA medical centers (VAMCs) for FY 2000 and 2005. The important role that these workload projections play in management applications of the methodology, such as 1 In 1986, Congress established three categories of eligibility for VA health care. The VA is required to provide hospital care and may provide outpatient and nursing home care, free of direct charge, to veterans within category A, defined to include those with service-connected disabilities, low-income veterans without such disabilities, and certain ''exempt'' veterans, including (for example) former prisoners of war, those exposed to Agent Orange, recipients of VA pensions, and those eligible for Medicaid. Veterans not in category A are assigned to either category B or category C on the basis of current income and net worth; the VA may provide care to these veterans on a "resources available" basis. At any point in time, there are well-defined income limits establishing eligibility for category B. Veterans not eligible for category B on the basis of either income or net worth are placed in category C. In fiscal year 1989, more than 95 percent of the applications for health care at all VA facilities in the country were from veterans in category A (U.S. Department of Veterans Affairs, 1989).
OCR for page 322
Physician Staffing for the VA: Volume I sensitivity analysis, is demonstrated in chapter 7. However, those chapters and their applications deal only with the empirically based approaches built around the production-function (PF) and inverse-production-function (IPF) equations. From chapter 5 (and especially Figures 5.1 and 5.3), it should be clear how the expert judgment-based approaches, after further refinement, could be applied in an analogous fashion to determine future physician requirements. Each empirically based or expert judgment approach requires estimates of future workload, developed at the level of specificity and detail appropriate for the case at hand. In the PF equations, the required projections take the form of Wiji, the workload for patient care area (PCA) j at VAMC i for fiscal year t. In the IPF equations, projections are required for Wijki, the workload associated with physician specialty k in PCA j at VAMC i for FY t. (See Equations 4.9 and 4.10, and the subsequent discussion "Using VA Data to Assign Values to Variables" in chapter 4.) In the expert judgment approaches, future workload must be projected at a somewhat greater level of detail than for the Empirically Based Physician Staffing Models (EBPSM) applications. In applications of either the Staffing Algorithm Development Instrument (SADI) or the Detailed Staffing Exercise (DSE), inpatient and long-term care workload must be projected by ward within PCAs, while taking into account consultations across PCAs as well as the volume of specialized diagnostic and intervention procedures performed by physicians. Under either expert judgment approach, ambulatory workload must be projected by specialty clinic within ambulatory PCAs. The main purpose of this chapter is to describe and briefly illustrate how these workload projections can be derived. For the analyses, the committee adopted several working assumptions: • Following the recommendation of its data and methodology panel, the committee agreed that the workload projection methodology developed for this study would reflect an adaptation of existing VA approaches. This is in line with the statement of work, which notes that, "The development of a formal mathematical/statistical patient care effective demand model is beyond the scope of this study." The committee also was influenced by two other considerations. First, the workload projection approaches described in this chapter are being used to guide VA resource allocation decisions, especially those related to facility planning. All else equal, it is preferable that the workload projections driving decisions about the requirements for beds, physicians, and all other resources be mutually compatible and logically consistent. This consideration becomes all the more important if the allocation of these resources is analyzed interactively within a decision support system, as discussed in chapter 7. Second, the major alternative approach to deriving future workload— namely, a statistically estimated demand-for-care model based on economic
OCR for page 323
Physician Staffing for the VA: Volume I theory considerations—is a complex undertaking, sufficient in scope and scale to merit a separate study. The committee believes that it is a study worth undertaking, as indicated in its concluding recommendations in chapter 11. The precision and specificity—and thus the policy usefulness—of the workload projection methodology would be enhanced if predictions about veteran utilization of the VA system could be derived as a function of such demand-influencing factors as income, health insurance coverage, availability of alternative sources of care, perceived quality of VA care (as indexed by such variables as scope of services), and distance from the VAMC, as well as age, gender, and eligibility-for-care status. A veteran's employment status will be an underlying determinant of several of these factors. • The formulas presented in the next three sections of this chapter—and the resulting workload projections that were used in the chapter 4 illustrations—do adjust for age, but not for gender or eligibility-for-care status. Because only 4.6 percent of the present veteran population is female (U.S. Department of Veterans Affairs, 1991), the sample sizes would be exceedingly small for most of the female-specific population cells required for a gender-specific breakout of Wiji, and Wijki. Unstable estimates for the female cells would likely result. Not presently available are projections of the veteran population by eligibility-for-care status at the level of specificity required for splicing Wiji, and Wijki on this basis. However, it is straightforward to extend the present workload projection models to accommodate both gender and eligibility (and other factors), once the required data become available. • Over the next three sections it will become clear that existing VA workload models are readily adapted for the projections required by the EBPSM. This is less the case for the expert judgment approaches. In each section, procedures are proposed and illustrated for using projections of Wiji, and certain proportionality assumptions, to derive corresponding projections for the workload variables used in the SADI and the DSE. It is not difficult, in concept, to derive independent projections for the SADI or the DSE workload variables, but additional data collection and analyses would be required. • The sequence of steps to derive workload projections from any of the models below is similar. For example, age is assumed to be the only variable in the projection model. For each VAMC i and PCA j, the mean value of current workload is computed per veteran for each age cell, the size of the veteran population in each cell is estimated for the fiscal year of interest, the (current) mean workload per veteran of each cell is multiplied by its projected cell size, and cells are summed to derive total workload for that fiscal year.
OCR for page 324
Physician Staffing for the VA: Volume I Throughout, this estimate is termed the projected workload to denote its derivation via this particular, nonprobabilistic process. Workload estimates that might be derived from statistically based demand-for-care models are termed predictions. Statements about statistical confidence can be made about the latter but not about the projections derived by use of existing VA techniques. (A similar distinction in terminology was made in chapter 4 regarding the physician FTEE estimates emerging from the IPF versus the PF.) In the remaining sections, the models are presented in some detail because the committee believes that it is important for the reader to appreciate the way in which particular assumptions translate into particular workload projections. Only then can the strengths and weaknesses of this projection approach be assessed objectively. One should not infer from the formula-oriented presentation below that these models are to be applied mechanistically in the Reconciliation Strategy. In the dialogue envisioned between VA Central Office and the VAMCs, the validity of particular workload projections would be a prime topic for discussion. As noted in chapter 7, a precedent for such dialogue has been established in at least one aspect of the VA's strategic planning operations. In the Medical District Initiated Program Planning (MEDIPP) process that was active until 1990 (Veterans Health Services and Research Administration, 1989), planners in Central Office provided district planners with preliminary estimates of future VA hospital bed requirements. District planners would typically transmit these projections to decision makers at the individual VAMCs; discussions would ensue; and periodically the district would ask Central Office to modify the projections, marshaling data and qualitative arguments to make the case. With the abolition of districts as part of a VA reorganization in 1990, this process has been suspended temporarily (and replaced by the Resource Planning and Management [RPM] methodology). But the concept of the process is important, for it provides a practical means to carry out the committee's intent that the formulas guide, not govern. INPATIENT WORKLOAD This discussion focuses on the derivation of inpatient workload projections required for the EBPSM. A procedure for obtaining workload projections for the expert judgment models is presented subsequently. Projections for the EBPSM As noted in chapter 4, the inpatient care workload measure that, with one exception, performed best overall on statistical and clinical criteria was the
OCR for page 325
Physician Staffing for the VA: Volume I Weighted-Work-Unit WWU) score; the exception came in the inpatient psychiatry PF equation, where bed-days of care (BDOC) was preferable. Hence, the presentation here centers around the WWU. In applications involving the inpatient psychiatry PF equation in chapter 4, BDOC in the psychiatry inpatient PCA is assumed to grow over time in proportion to total WWUs there. Projection Model The basic equation underlying the inpatient workload projection model is The equation says that future WWUs will be calculated as the product of the projected number of WWUs per inpatient discharge, the number of discharges per veteran (Discharge Rate), and the size of the veteran population. Since the product of the latter two is simply the projected number of discharges, the equation calculates projected WWUs as the product of projected WWUs per discharge and projected discharges. The operational form of this equation (the version used to project workload in practice) is somewhat more complicated because it must accommodate several considerations: aging of the veteran population, differentiation of WWUs by physician specialty category (see "Using VA Data to Assign Values to the Variables" in chapter 4), and the breakout of the VAMC into PCAs. When these are acknowledged, Equation 8.1 becomes where WWUijk, 1989, a = total WWUs associated with specialty k generated by age group a in PCA j of VAMC i in FY 1989;
OCR for page 326
Physician Staffing for the VA: Volume I Dischargesij, 1989, a = the number of age group a discharges from PCA j of VAMC i in FY 1989; Discharge Rateijia = the projected number of age group a discharges from PCA j of VAMC i in FY t, divided by the projected age group a veteran population size in the Primary Service Area (PSA) associated with VAMC i in FY t; Vet Popita = the projected age group a veteran population for the PSA of VAMC i in FY t. The inpatient PCAs are medicine, surgery, psychiatry, neurology, rehabilitation medicine, and spinal cord injury (SCI). The age groups are 0-24, 25-34, 35-44, 45-54, 55-64, 65-74, and 75 +, so that A = 7. The PSA of VAMC i is defined as the set of contiguous counties such that each has a plurality of its medical and surgical VA inpatient discharges from VAMC i. Loosely, the PSA of VAMC i is the group of counties generally served by that facility. It is a concept long used in VA facility planning and serves to define the catchment area (and thus, roughly, the VA target population) for each VAMC. Workload projections for all PCAs in this study use this same definition of a PSA, except for SCI. PSAs are defined on a regional basis, but the principle of trying to capture the appropriate target population is the same. For the PF variant of the EBPSM, the required form of projected workload is where the five specialty-associated types of WWUs generated on any inpatient PCA j are (using the notation of chapter 4) MEDWWU, SURWWU, PSYWWU, NEUWWU, and RMSWWU. For the IPF variant, the required form of projected workload is where the sum is across the six inpatient PCAs, and k is now properly interpreted as one of the six physician specialties linked expressly to an inpatient PCA (medicine, surgery, psychiatry, neurology, rehabilitation medicine, and
OCR for page 327
Physician Staffing for the VA: Volume I SCI). For a specialty such as nuclear medicine, with no single PCA link, the inpatient workload variable is defined as the sum of some specified subset of the WWUiki, in Equation 8.3 (see "Estimated PF and IPF Equations" in chapter 4). Although these projection techniques are similar to those used in the VA hospital planning model, there are some notable differences. In particular, the hospital model does not have a specialty-specific capability at present. It uses BDOC/Discharges rather than the WWUs/Discharges found in Equation 8.1 and thus expresses workload in terms of patient days rather than WWUs. Both the VA model and the one proposed here produce workload projections based on the projected age-adjusted veteran population. Using VA Data to Assign Values to Variables The projected value of each component of Equation 8.1' is derived from data collected and analyzed by the VA: WWUs/Discharges—Values for both the numerator and the denominator are contained in the VA Patient Treatment File (PTF) (with WWUs appended), as discussed in chapter 4, and the Annual Patient Census. Since WWUs first became accessible nationally at the required (PCA-specific) level of detail in FY 1989, historical observations on this ratio were available for FY 1989 only (given the time frame of the analysis). Discharge Rate—For the numerator (Discharges), the required data are from the PTF and the Census, as just noted. For modeling purposes, a VA patient is said to be discharged if he/she is either (1) discharged from the facility or (2) transferred to another PCA in the facility. In addition, for the most recent fiscal year only, a "pseudo discharge" is generated for each patient occupying a bed in the facility at the end of a fiscal year. The veteran population data for the denominator are available, by age and PSA, from the VA official internal projections. The projected discharge rate used in Equation 8.1' is computed as a function of the three most recently available historical discharge rates, as follows: If the historical rate has risen continuously over three years, the projected discharge rate for FY t is derived by taking the most recent rate as the base and imparting to it a one-time percentage increase equal (in percentage terms) to the observed percentage increase over the three years, up to a maximum increase of 10 percent. If the historical rate has declined continuously over the three years, the projected rate is set at the most recent historical rate. If the historical rate fluctuates over the three years (i.e., was not monotonically increasing or decreasing), the projected rate is the overall average historical rate for the three-year period.
OCR for page 328
Physician Staffing for the VA: Volume I If there are no discharges in a given PCA-age cell in any of the three years, the projected discharge rate equals the rate observed in the most recent year. • Vet Pop—Historical as well as projected veteran population estimates, by age and PSA, are based on VA internal projections. Numerical Illustration Focusing on Equation 8.1', calculations are begun at the most micro level, then are aggregated as required to obtain illustrative estimates for Equations 8.2 and 8.3. The calculations pertain throughout to VAMC II, an actual VA facility used in analyses in chapters 4 through 6, and to FY 2000. The first problem is the projection of the MEDWWUs that will be generated in the inpatient medicine PCA for the oldest age group, 75 +. For FY 1989, MEDWWUs for the 75 + age group in the medicine inpatient PCA was 628.95, and there were 1,075 discharges. Thus, WWUs/Discharges = 628.95/1,075 = 0.59. For this age group, the discharge rates from the inpatient medicine PCA for the three most recent fiscal years of 1987, 1988, and 1989 are, respectively, 0.066793, 0.067805, and 0.063077. Since these rates do not continuously increase or decrease, the projected discharge rate is computed as the overall average of the three, which turns out to be 0.066109. The projected veteran population for the 75+ age group in the PSA associated with VAMC II for FY 2000 is 54,813. The projected workload that results when these components are combined—namely, WVAMC II, Inpatient Medicine, MEDWWUs, FY 2000—equals 0.59 × 0.066109 × 54,813 = 2,120 MEDWWUs. After similar calculations are completed for each of the other six age groups, total projected MEDWWUs on the medicine inpatient PCA are found to be (7 + 78 + 290 + 817 + 1,639 + 1,789 + 2,120) = 6,740, where the age-specific projections have been arrayed in ascending chronological order. Total projected WWUs for the inpatient medicine PCA—the key workload value required in applications of the inpatient medicine PF—appears in the notation of Equation 8.2 as WVAMC II, Inpatient Medicine, FY 2000 and is computed as the sum of the MEDWWUs, SURWWUs, PSYWWUs, NEUWWUs, and RMSWWUs generated on this PCA. The first of these has been computed as 6,740; the remaining four are 1,703, 168, 210, and 22, respectively, yielding total of 8,843 WWUs. Total projected MEDWWUs—the key inpatient workload variable in the medicine IPF—appears in the notation of Equation 8.3 as WVVAMC II, Medicine, FY 2000. It is the sum of the projected MEDWWUs across the six inpatient PCAs of medicine, surgery, psychiatry, neurology, rehabilitation medicine, and SCI. For
OCR for page 329
Physician Staffing for the VA: Volume I VAMC II and FY 2000, this is the sum of 6,740, 1,195, 317, 454, 98, and 56, respectively, yielding total projected MEDWWUs of 8,860. Similar computations can be performed for each inpatient PCA, for each type of WWU, for any VAMC, and for any future fiscal year. Projections for the Expert Judgment Approaches As noted in chapter 5, data for computing most of the workload variables used in the SADI and the DSE either now exist, or could readily be collected, at the individual VAMCs. But an automated, national data base containing this information does not now exist. Thus, the VA decision maker is in no position currently to apply analogues of the workload projection formulas shown above in "Projections for the EBPSM" to obtain direct estimates of such SADI or DSE inpatient workload variables as average daily census (ADC) by ward, admission rates by PCA, various physician-performed procedures, and consultations to other inpatient PCAs. However, what may be termed indirect estimates can be derived from the workload projections discussed under "Projections for the EBPSM" above, as follows: Average Daily Census For the inpatient medicine PCA at VAMC II, total projected WWUs for FY 2000 are 8,843. The corresponding total WWUs for FY 1989 were 7,484, implying a projected 18 percent increase in workload by FY 2000. Indirect estimates of the corresponding ADC workload variables, as required for both the SADI and the DSE, are obtained by assuming that ADC changes in proportion to total WWU between the two years. To be specific, Figure 5.1 indicates an ADC of 28 on Ward 1 (a general medicine unit) of the inpatient medicine PCA at VAMC II in FY 1989. If it is assumed that the ADC will increase in proportion to total WWUs for the PCA, an ADC on Ward 1 for FY 2000 of 28(1.18) = 33.04 can be projected. Note that this particular proportionality assumption, like others used below, is simple to implement as a first cut, but should not be adopted without close scrutiny. For example, if the projected aging of the veteran population leads to an increase in WWUs/Admissions, then ADC will not be proportional to WWUs. This suggests a larger point. There are case mix and case severity assumptions embedded in any SADI at a point in time. When projecting future workload for the instrument, one must be alert to possible changes in case mix or severity that would affect physician time requirements.
OCR for page 330
Physician Staffing for the VA: Volume I Admission Rates Two approaches to an indirect estimate of admissions to the PCA are available. First, one can assume the admission rate will change over time in proportion to total WWUs for the PCA. An alternative (given the population aging effect noted just above) is to assume that the admission rate moves in proportion to total PCA discharges (as defined earlier), the projection of which is an intermediate product in the workload formula Equation 8.1'. The daily admission rate for the inpatient medicine PCA at VAMC II in FY 1989 was 15.2. Under the first approach, the admission rate for FY 2000 is projected to be 15.2(1.18) = 17.94. Under the second approach, there were 9,868 discharges in FY 1989 from the inpatient medicine PCA of VAMC II. It can be shown that 11,848 discharges from this PCA are projected for FY 2000, a rate of increase from FY 1989 of 20 percent. When this value is applied to the FY 1989 daily admission rate in medicine at VAMC II, the resulting projected rate for FY 2000 is 15.2(1.20) = 18.24. Physician-Performed Procedures To obtain indirect estimates here, the rate at which a given diagnostic or interventional procedure is performed is assumed to change over time in proportion to total WWUs for the PCA most closely associated with the procedure. An example is the problem of projecting the number of endoscopies at VAMC II for FY 2000. This procedure is associated with the inpatient medicine PCA. In FY 1989, about 13 endoscopies were performed daily. Thus, the projected daily performance rate for FY 2000 is 13(1.18) = 15.34. Consultations in Other Inpatient PCAs To derive projections for these workload measures, which are explicitly required for the SADI, the following is assumed: The consultation rate by physicians in a given specialty to a given inpatient PCA will change over time in proportion to the projected change in that specialty's associated WWU total generated in this PCA. In the context of the present example, the problem of projecting consultations from the medicine service to the inpatient surgery PCA is considered. Records at VAMC II indicate that, in FY 1989, there were approximately 2,778 consultations from medicine to inpatient surgery, implying a daily rate of 10.7 (based on 260 consultation days/yr). In FY 1989, total MEDWWUs attributed to the inpatient surgery PCA were 1,102. For FY 2000, the corresponding projected MEDWWUs are 1,195, an increase of about 8.4
OCR for page 331
Physician Staffing for the VA: Volume I percent. Applying the proportionality assumption again, projected consultations per day from medicine to inpatient surgery for FY 2000 are computed as 10.7(1.084) = 11.6. If the VA pursues the concentrated testing and development of the SADI recommended in chapter 6, a national data base of observations on the workload variables required for the instrument would emerge naturally. The stage would be set for projecting their future values via techniques analogous to those summarized in Equation 8.1', and there would be no need to use indirect estimate methods. AMBULATORY CARE WORKLOAD The initial focus again is on workload projections for the EBPSM. Subsequently, a procedure for using these to derive projections for the expert judgment approaches is discussed. Projections for the EBPSM The ambulatory care workload variable that, with one exception, performed best overall on statistical and clinical criteria in the estimated equations of chapter 4 was the Capitation Weighted Work Unit (CAPWWU); the exception came in the PF equation for the ambulatory other physician services (OPS) PCA, where the use of clinic-stop visits was preferable. Hence, this presentation concentrates on the CAPWWU workload variable. Since the projection of future CAPWWUs requires a projection of future clinic stop visits, no additional steps are required to obtain this workload variable for the OPS equation. Projection Model The basic equation underlying the ambulatory care workload projection model is The equation says that future CAPWWUs will be calculated as the product of the projected CAPWWUs per clinic-stop visit, the projected number of clinic-stop visits per veteran, and the projected size of the veteran population. Since the product of the second two elements is projected clinic stops, the equation implies
OCR for page 332
Physician Staffing for the VA: Volume I that projected CAPWWUs is simply the product of projected CAPWWUs per clinic stop and projected clinic stops. The version of this equation used to project workload here must also accommodate two additional factors: the aging of the veteran population and the partitioning of the ambulatory care arena into six mutually exclusive and exhaustive PCAs. Hence, Equation 8.4 assumes the expanded form where CAPWWUij, 1989, a = total PCA j CAPWWUs generated by age group a at VAMC i in FY 1989; Clinic Stopsij, 1989, a = the number of age group a visits to clinic stops associated with PCA j at VAMC i in FY 1989; Clinic-Stop Rateijta = the projected number of PCA j clinic-stop visits generated by age group a at VAMC i in FY t, divided by the projected age group a veteran population size in the PSA associated with VAMC i in FY t; Vet Popita = the projected age group a veteran population for the PSA of VAMC i in FY t. The ambulatory PCAs are medicine, surgery, psychiatry, neurology, rehabilitation medicine, and other physician services. The age groups are the same seven defined previously for the inpatient workload model. Because there is only one (specialty-linked) category of CAPWWU associated with each ambulatory PCA (i.e., MEDCAPWWUs are generated only in the ambulatory medicine PCA), the left-hand side of Equation 8.4' is the workload variable used in the PF equation for that ambulatory PCA. For the IPF equations, the assignment of ambulatory workload variables is as follows: for medicine, the sum of MEDCAPWWUs and OPSWWUs (from the OPS PCA); for surgery, SURCAPWWUs; for psychiatry, PSYCAPWWUs; for neurology, NEUCAPWWUs; for rehabilitation medicine, RMSCAPWWUs; and for the other six physician specialties studied here, the sum of some subset
OCR for page 333
Physician Staffing for the VA: Volume I of the six PCA-specific projections. See ''Estimated PF and IPF Equations' in chapter 4 for a discussion of these PCA-to-specialty linkages required for the definition of ambulatory workload variables in the IPF equations. The model summarized in Equation 8.4' differs from the existing VA outpatient workload model in several respects. The latter generates workload projections in terms of ''patient visits" for the entire VAMC, disaggregated into several broad categories: compensation and pension examinations, applications for care, five distinct categories of mental health visits, and a residual category for "other" types of visits. On a single visit to the facility, however, a patient may generate several clinic-stop visits (see chapter 4). In contrast, the model above uses patient-specific information about the pattern of clinic-stop visits, and their corresponding direct costs, in order to generate workload measures (CAPWWUs) that are specific to the PCA (not just the facility) and that reflect some information about relative case severity. Using VA Data to Assign Values to the Variables The projected value of each component of Equation 8.4' is derived from data collected and analyzed by the VA: CAPWWUs/Clinic Stops—Values for both the numerator and the denominator are contained in the VA Staff Outpatient File (with CAPWWUs appended). Since observations on CAPWWUs assigned to the ambulatory PCAs defined for this study were first available for FY 1989, historical observations on this ratio are consequently limited to that year (given the time frame of the analysis). Clinic-Stop Rate—For the numerator (Clinic-Stop Visits), the required data are from the Staff Outpatient File, as noted. The veteran population data for the denominator are available, by age and PSA, from official VA projections. The projected clinic-stop rate in Equation 8.4' is computed from the three most recently available clinic-stop rates, as follows: If each rate is higher than the previous one, the projected clinic-stop rate for FY t is derived by taking the most recent rate as the base and imparting to it a "one-shot" percentage increase equal (in percentage terms) to the observed increase over the three years, up to a maximum increase of 20 percent. If the historical rate has declined continuously over the three years, the projected clinic-stop rate is set equal to the most recent rate. If the historical rate fluctuates over the three years, the projected clinic-stop rate is calculated as the overall average rate for the three years. If there are no clinic-stop visits in a given PCA-age cell in any of the three years, the projected rate equals the most recent rate.
OCR for page 334
Physician Staffing for the VA: Volume I • Vet Pop—Historical as well as projected veteran population estimates, by age and PSA, are based on VA internal projections. Numerical Illustration First, the calculation of one of the components of Equation 8.4' is considered in a particular case; then this result is combined with other similar ones (not derived here) to obtain a projected value for CAPWWUijt, the key ambulatory care workload variable. The calculations all pertain to the ambulatory medicine PCA at VAMC II and to FY 2000. Hence, the workload projection is for CAPWWUVAMC II, Ambulatory Medicine, FY 2000. Consider first the projection of MEDCAPWWUs for the oldest age group, 75+. In FY 1989, the workload total for this group was 758,453 MEDCAPWWUs, and total clinic-stop visits in the ambulatory medicine PCA were 8,266. Thus, MEDCAPWWUs/Clinic Stops were 91.76. For this age group, the clinic-stop rates in the ambulatory medicine PCA for the three (most recent) fiscal years of 1987, 1988, and 1989 are, respectively, 0.3029, 0.5002, and 0.4919. Since these rates exhibit neither an increasing nor a declining pattern, the projected clinic-stop rate is the overall average of the three, namely 0.4368. The projected veteran population for this group is 54,813. When these three components are combined, the projected workload for the 75+ age group at VAMC II for FY 2000 is 91.76 × 0.4368 × 54,813 = 2,196,947 MEDCAPWWUs. For the remaining six age groups, projected MEDCAPWWUs have been computed to be, in ascending chronological order, 46,931, 249,726, 610,430, 1,427,560, 2,354,481, and 3,330,144. Summing over all age groups, as required by Equation 8.4', yields an overall projection of 10,216,219 for CAPWWUVAMC II, Ambulatory Medicine, FY 2000. Similar computations can be performed for each ambulatory PCA in any VAMC for any fiscal year. Projections for the Expert Judgment Approaches In both the SADI (see Figure 5.2) and the DSE (see Figure 5.1), calculating the physician FTEE for ambulatory care in some future year requires projecting the visit rates to specified types of clinics. Moreover, these projected rates typically must distinguish between initial and followup visits and be conditional on whether either residents or nonphysician practitioners (or both) are available to work with staff physicians in the clinic.
OCR for page 335
Physician Staffing for the VA: Volume I As was the case with inpatient workload, there is no national data base containing observations on ambulatory care visit rates, so defined. (Again, this would change if the SADI or the DSE were implemented, even experimentally, across the VA system.) Hence, a projection model such as that shown in Equation 8.4' cannot be used at present to derive direct estimates of future ambulatory workload. However, indirect estimates of workload can be obtained by invoking proportionality assumptions similar to those used in the inpatient calculations. Suppose it is assumed that, within each ambulatory care PCA, the visit rate for a given clinic changes over time in proportion to that PCA's CAPWWU score. For a medicine clinic (e.g., pulmonary), the visit rate is made proportional to MEDCAPWWUs. For a surgery clinic (e.g., urology), the visit rate is proportional to SURCAPWWUs. For the emergency and admitting & screening areas, the rate is proportional to OTHCAPWWUs, the workload index associated with the ambulatory OPS PCA. To illustrate, total MEDCAPWWUs for VAMC II in FY 1989 was 9,705,108, and the projection for FY 2000 is 10,216,219. This represents a 5.3 percent increase between the two years. From Figure 5.1, the visit rate for the pulmonary clinic in FY 1989 was 53/week. Invoking the proportionality assumption, the projected pulmonary clinic visit rate for FY 2000 is 53(1.053) = 55.8. This is the workload projection for determining physician requirements at this clinic in FY 2000. (A caveat again is that the aging of the veteran population may lead the visit rate for the PCA not to be proportional to its total CAPWWU score; for instance, the latter may grow faster than the former.) LONG-TERM CARE WORKLOAD As before, the workload projection method for the EBPSM is examined first, then extensions of the analysis that yield (indirect) workload estimates for the expert judgment approaches are considered. Projections for the EBPSM The long-term care (LTC) workload variable that performed best overall on statistical and clinical criteria in the estimated equations of chapter 4 was the Resource Utilization Group Weighted Work Unit (RUGWWU). Hence, the presentation will concentrate entirely on the RUGWWU workload measure.
OCR for page 336
Physician Staffing for the VA: Volume I Projection Model The basic equation underlying the LTC workload projection model is The equation says that future RUGWWUs will be calculated as the product of projected RUGWWUs per discharge from an LTC unit, the projected number of discharges per veteran, and the projected size of the veteran population. For use in this study, the equation must be expanded to acknowledge three specific factors: the aging of the veteran population, the breakout of the long-term care arena into PCAs, and the differentiation of RUGWWUs by physician specialty category (see "Using VA Data to Assign Values for the Variables" in chapter 4). Hence, the LTC workload model becomes where RUGWWUijk, 1989, a = total RUGWWUs associated with specialty k generated by age group a in long-term care PCA j of VAMC i in FY 1989; Dischargesij, 1989, a = the number of age group a discharges in FY 1989 from PCA j of VAMC i; Discharge Rateijta = the projected number of age group a discharges from PCA j of VAMC i in FY t, divided by the projected age group a veteran population size in the PSA associated with VAMC i in FY t; Vet Popita = the projected age group a veteran population for the PSA of VAMC i in FY t. The LTC PCAs are the nursing home and intermediate care. The age groups are the same seven defined for the inpatient and the ambulatory care workload models.
OCR for page 337
Physician Staffing for the VA: Volume I For the PF variant of the EBPSM, the required form of projected workload is where the three specialty-associated types of weighted work units in LTC are (using the notation of chapter 4) MEDRUGWWUs, PSYRUGWWUs, and RMSRUGWWUs. Thus, each VA patient discharged from a nursing home unit or intermediate care ward will generate a certain number of RUGWWUs, which are labeled either as medicine, psychiatry, or rehabilitation medicine, depending on the primary diagnosis at discharge. In contrast to the inpatient projection model, there are no RUGWWUs specific to surgery or neurology. For the IPF, the required form of projected workload is where the sum is across the two PCAs of nursing home and intermediate care, and k is now properly interpreted as one of the three physician specialties (either medicine, psychiatry, or rehabilitation medicine) with a specific RUGWWU linkage to the LTC PCAs. Thus, in the IPF for medicine, the LTC workload variable is simply which can also be expressed (in the notation of chapter 4) as MEDRUGWWU it. The LTC workload variables for psychiatry and rehabilitation medicine are constructed similarly. For each of the remaining physician specialties (e.g., surgery, neurology, or diagnostic radiology), the LTC workload variable in its IPF is defined as the sum of some specified subset of the RUGWWUikt, in Equation 8.7 (see "Estimated PF and IPF Equations" in chapter 4). For example, in the diagnostic radiology IPF, RUGWWUi, Medicine, t is used as the surrogate measure of LTC workload. The approach to workload projection summarized in Equation 8.5' differs from current VA models in several respects. For intermediate care, the VA uses bed-days of care per discharge rather than RUGWWUs/Discharges, and
OCR for page 338
Physician Staffing for the VA: Volume I expresses workload in patient days rather than WWUs. To project nursing home workload, in patient days, the VA multiplies the age-specific veteran population in a PSA by a corresponding civilian male nursing home utilization rate, derived from the 1985 National Nursing Home Survey (National Center for Health Statistics, 1985). In the present study, nursing home workload projections are derived entirely from VA data. Using VA Data to Assign Values to the Variables The projected value of each component of Equation 8.5' is derived from data collected and analyzed by the VA: RUGWWUs/Discharges—Values for both the numerator and the denominator are taken from the VA Patient Treatment File, with RUGWWUs appended. For intermediate care, the source is the same PTF from which the inpatient PCA workload data were derived. For the nursing home, the source is the Extended Patient Treatment File. In keeping with the inpatient and ambulatory care models, this component of the LTC workload projection equation is calculated for FY 1989 only. Discharge Rate—For the numerator (Discharges), the required data are taken from the PTF (for intermediate care) or the Extended PTF (for the nursing home), as noted above. Similar to the algorithm established for inpatient care, a VA patient is classified as discharged if he/she is either (1) discharged from the facility, (2) transferred to another PCA within the facility, or (3) occupies a bed in the facility at the end of the fiscal year. The veteran population data for the denominator are available, by age and PSA, from VA internal projections. The projected discharge rate in Equation 8.5' is computed from the three most recently available historical rates via "trending rules" identical to those used for deriving the projected discharge rate for inpatient PCAs (see "Using VA Data to Assign Value to the Variables" under "Inpatient Workload," above). Recall that these rules serve to establish certain upper and lower bounds on the projected discharge rate, regardless of the observed rate of change over the three-year period. Vet Pop—Historical as well as projected veteran population estimates, by age and PSA, are based on VA internal projections. Numerical Illustration The objective now is to demonstrate how the LTC workload is derived, via Equations 8.6 and 8.7, for use in the PF and IPF equations. To do so, first a workload projection is performed at the most detailed level possible. Then, it
OCR for page 339
Physician Staffing for the VA: Volume I is shown how such results can be aggregated to produce LTC workload estimates in the desired form. The calculations again pertain to VAMC II and to FY 2000. First, the focus is on the task of projecting the number of medicine RUGWWUs (MEDRUGWWUs) generated by the 75 + age group in the nursing home PCA at VAMC II in FY 2000. For FY 1989, MEDRUGWWUs for the 75 + age group in the nursing home PCA were 12,690.7, and there were 62 discharges. Thus, MEDRUGWWUs/ Discharges = 12,690.7/62 = 204.69. For this age group, the discharge rates in the fiscal years of 1987, 1988, and 1989 are, respectively, 0.0028857, 0.0033123, and 0.0036898. Since these discharge rates are increasing, the projected discharge rate for FY 2000 is calculated by taking the FY 1989 rate as the base and applying to it a one-time percentage increase equal to the lesser of the actual rate of increase observed over these three years, or 10 percent. Since the FY 1989 discharge rate is about 28 percent greater than the FY 1987 rate, the projected discharge rate for FY 2000 is calculated here as 0.0036898(1.10) = 0.0040588. The projected veteran population for this group in the PSA associated with VAMC II for FY 2000 is 54,813. The projected nursing home workload for the age group 75+—namely RUGWWUVAMC II, Nursing Home, MEDRUGWWU, FY 2000—equals 204.69 × 0.0040588 × 54,813 = 45,538 RUGWWUs. To obtain projected medicine RUGWWUs for the nursing home PCA, all seven age-specific projections are added: (513 + 710 + 5,498 + 9,760 + 17,944 + 45,538) = 79,962. Total RUGWWUs projected for the nursing home PCA—the workload value required in applications of the nursing home PF—is the sum of all RUGWWUs associated with medicine, psychiatry, and rehabilitation medicine: (79,962 + 2,514 + 22,116) = 104,593 = RUGWWUVAMC II, Nursing Home, FY 2000. Total projected RUGWWUs in medicine (MEDRUGWWUs)—the LTC workload variable required for the medicine IPF—is the sum of the MEDRUGWWUs projected for the nursing home and intermediate care PCAs: (79,962 + 47,298) = 127,260 = RUGWWUVAMC II, Medicine, FY 2000. Projections for the Expert Judgment Approaches In applications of both the SADI and the DSE, patient workload projections are required for both the nursing home and intermediate care PCAs. Under either expert judgment approach, the same types of workload variables relevant to assessing physician requirements for the inpatient PCAs apply, as well, in the LTC PCAs: ADC, admission rates, physician-performed procedures (e.g., swan ganz catheter, spinal tap, nasogastric tubes), and consultations to other PCAs.
OCR for page 340
Physician Staffing for the VA: Volume I Procedures for using model-derived (WWU) workload projections to derive indirect estimates for these SADI/DSE workload variables have been described and illustrated under "Inpatient Workload" above. To reinforce that these procedures apply directly to LTC as well, one particular example is provided: the projection of ADC for the nursing home PCA at VAMC II for FY 2000. In FY 1989, the ADC in the nursing home PCA was 96, and 63,584 RUGWWUs were generated. For FY 2000, the LTC workload model projects 104,593 RUGWWUs, a 64 percent increase from FY 1989. Invoking the assumption that changes in ADC are proportional to changes in RUGWWUs, the indirect estimate for nursing home ADC for FY 2000 is 96(1.64) = 157.44. A Caveat With the average occupancy rate at 95 percent and a growing waiting list of veterans to be admitted, there is presently an "excess demand" for nursing home beds in the VA (Audrey Urquhart, Program Analyst, Office of the Assistant Chief Medical Director for Geriatrics and Extended Care, Department of Veterans Affairs, personal communication, 1991). Given the expected growth of the age 65 + veteran population, this excess demand is likely to persist for years unless the VA rapidly increases the number of nursing home beds. This issue is important to the interpretation of the LTC workload projection model. The discharge rate in Equation 8.5' is based on current VA nursing home utilization and thus reflects current supply constraints; if there were more nursing home beds available, this projected rate would undoubtedly be higher for most VAMCs. This raises the question of whether it is the appropriate rate to use for projecting LTC workload. The answer would appear to hinge on whether the VA will maintain, or change, its present policy assumptions about the provision of nursing home care to veterans. Specifically, VA strategic planners in recent years have projected nursing home bed requirements under the assumption that the VA will provide nursing home care to about 16 percent of the eligible veteran population. Within this total market share, planners have assumed further that 30 percent of admitted veterans would be in state nursing home beds, 40 percent in community beds, and 30 percent in VA beds. The VA would pay (as it does now) for all of this care, but thus provide directly only about 30 percent of this 16 percent market share. The observed nursing home discharge rate, critical to Equation 8.5', is a reflection of this market-share policy. If the policy does not change over time—so that the VA's share of total veteran nursing home care is stable—there will likely be a roughly stable relationship between current and future age-specific discharge rates, since discharges will tend to rise with the eligible
OCR for page 341
Physician Staffing for the VA: Volume I veteran population, all else equal. In that event, the case for using the model summarized in Equation 8.5' to project LTC workload is supported. On the other hand, if the VA significantly increases its supply of nursing home beds, the workload projection model would have to be modified accordingly. This could involve reassessing both the projected discharge rate and the projected RUGWWUs per discharge, since the (age-specific) severity mix of patients may change as the fraction of total market share treated increases. REFERENCES Institute of Medicine. 1987. Study Workplan (Statement of Work) for a Study to Develop Methods Useful to the Veterans Administration in Estimating its Physician Needs. Washington, D.C. Unpublished. National Center for Health Statistics. 1985. National Nursing Home Survey: 1985 Summary for the United States. GPO 017-022-01065. Washington, D.C.: Government Printing Office. U.S. Department of Veteran Affairs. 1989. Summary of Medical Programs, September 1989. Washington, D.C.: Department of Veterans Affairs. U.S. Department of Veteran Affairs. 1991. Annual Report of the Secretary of Veterans Affairs, Fiscal Year 1990. Washington, D.C.: Department of Veterans Affairs. Veterans Health Services and Research Administration. 1989. Manual M-9, "MEDIPP." Washington, D.C.: Department of Veterans Affairs.
OCR for page 342
Physician Staffing for the VA: Volume I This page in the original is blank.
Representative terms from entire chapter: