Physician Staffing for the VA: Volume I (1991)

In this final chapter of the report, the committee presents its conclusions and recommendations on how the Department of Veterans Affairs (VA) should determine physician requirements. Also summarized are the committee's views on two important related topics, VA-medical school affiliations and nonphysician practitioners.

Most of the points below, and the committee's reasoning behind them, have already appeared in the course of the first 10 chapters. However, a question not addressed earlier concerns the steps the VA should take to ensure that the physician requirements methodology is further refined in the near term, then maintained and improved over time. This issue is discussed near the end of the chapter.

As a prelude, the main elements of the committee's charge are reiterated.

The central purpose of this study has been to develop a methodology to assist the VA in answering a basic, but extraordinarily complex, question: To accomplish its principal mission-related responsibilities of patient care, education, and research, how many physicians does the VA require?

Specifically, the VA asked the Institute of Medicine (IOM) to develop "a sound methodology for estimating the number of physicians, by specialty groupings, required for the efficient delivery of high quality physician services" (Institute of Medicine, 1987) in all programs and facilities operated by the Veterans Health Administration (VHA), which has responsibility for all VA physician-related activities.

The VA designated as the primary objective for the study the development of a "mathematical/statistical methodology, incorporating both empirically derived and professional judgment based values in the methodology's algorithms, which translates quantitative measures of ... mission related workload demands . . . into numerical estimates of physician staff requirements" (Institute of Medicine, 1987). Data for these analyses would be derived from three sources:

• The VA's own information systems, reflecting what may be characterized as "internal" performance norms;

• "External" (to the VA) physician performance norms, as gathered directly or else inferred from other public-and private-sector health care organizations; and

• Expert panels, which would evaluate the models, the data used in them, and external norms—and, in light of these assessments, recommend modifications to either the models or their physician staffing recommendations.

The committee determined that the overall methodology should be capable of assessing:

• The number of physicians required, at the present time, to meet the current patient care workload at VA medical centers (VAMCs). These assessments would be conditional on assumptions about the scope and case acuity of the workload; number and type of residents; availability of nonphysician resources (nurses, support staff, and other productivity-influencing factors); and the commitment of the VAMCs to teaching, research, continuing education, and other activities beyond direct patient care.

• Future VA physician requirements, taking explicit account of possible changes in the volume, mix, and case acuity of service demands resulting from the aging of the veteran population. The methodology should likewise be flexible enough to incorporate (in the future) projected changes in other factors influencing utilization, such as the distribution of veterans across eligibility-for-care categories and the proportion of females in the eligible population.

• The net effect on VA staff physician requirements of possible changes in the number, type, and intensity of VA-medical school affiliation relationships. In addition, there should be analyses of the potential effect of such changes on the VA's ability to accomplish the physician education component of its mission, both now and in the future.

Over the years, the VA has published staffing guidelines for most health care provider categories, but not for physicians. This omission reflects the genuine complexities—clinical, economic, administrative, political—that abound in attempting to estimate just how many doctors are required to meet the VA's mission.

In the majority of VAMCs, that mission is multipurpose: patient care, education, and research. In most of these activities, the VA staff physician is not a solo performer but works with a number of others—residents, non-VA consulting physicians, nurses, nonphysician practitioners, and other support staff. Hence, the number of staff physicians required, in any specific context, will be influenced by the availability and productivity of these other providers, who may function as complements to or substitutes for staff physicians. Nonpersonnel

factors, such as the availability of critical capital equipment or floor space, may also be important. The amounts of time set aside for research, classroom instruction, continuing education, administrative activities, and professional development all should figure directly into the computation of total physician requirements.

The central issue facing the VA decision maker is determining which methodological approach(es) should be adopted. Three general approaches were investigated:

1. The analyses in chapter 4 demonstrated how physician requirements can be derived from statistical models estimated from existing VA data. Specifically, the committee developed the Empirically Based Physician Staffing Models (EBPSM) with two, complementary variants—the production function (PF) and the inverse production function (IPF).

2. In chapter 5, two alternative formulations of an expert judgment model for physician staffing were introduced—one based on the Detailed Staffing Exercise (DSE) and the other on the Staffing Algorithm Development Instrument (SADI).

3. Another approach also discussed in that chapter would involve using non-VA physician staffing criteria, or external norms, for guiding the decision about physician requirements in the VA.

(A fourth approach is to adopt no new methodology. The committee rejects this option—and all others not based on operating principles that are clearly specified, logically defensible, and appropriate for policy making by some reasonable criteria.)

Over the final months of the study, the committee examined four alternative decision strategies for using these staffing approaches to derive the total physician Full-Time-Equivalent Employees (FTEE), by specialty, required for a given VAMC. The strategies called, in turn, for the VA decision maker to

1. Adopt one dominant approach for each specialty (e.g., medicine) or clinical program area (e.g., ambulatory care). For example, the core of the methodology could be an empirically based model, but expert panels could be appointed to evaluate its staffing recommendations—and the model itself. Whether or not all specialties and program areas would be guided by the same dominant approach would be a separate decision.

2. Use two or more approaches in conjunction to derive a range of physician staffing estimates There would be no formal model or algorithm for either justifying or reconciling differences among the approaches. Instead, the VA decision maker would have a menu of physician staffing estimates, each defensibly derived, from which to choose.

3. Use two or more approaches in conjunction to derive a range of physician staffing estimates sensitive to assumptions about budgetary and other constraints. This strategy differs from strategy B only in its advocacy of sensitivity analysis, optimization models, and related techniques to help the VA decision maker investigate important "what if" questions. These techniques were discussed and illustrated in chapter 7.

4. Through some integrative process (e.g., mathematical weighting scheme), combine physician staffing results from two or more approaches to produce either a single FTEE estimate or a range of estimates.

In the committee's terminology, this integration could be accomplished holistically (Strategy D. 1) to produce, in a single weighted-average calculation, an overall FTEE total for each specialty or program area. Or it could be implemented in a disaggregated format (Strategy D.2), which allows for different weights to be applied to the different component parts of physician FTEE; the total required FTEE in a VA specialty or a program area at the VAMC would be the sum of these weighted components.

The sensitivity analyses noted above could be conducted as well under either variant of this strategy.

As an overall framework for determining VA physician requirements (given workload and other factors), the committee endorses Strategy D.2, the disaggregated weighted-average variant of D. In chapter 6, this was termed the Reconciliation Strategy. The formulation of the strategy presented there is reproduced below using (for illustration) internal medicine, the PF variant of the EBPSM, and the SADI variant of the expert judgment models:

where

By varying the parameters b, c, and d jointly across their ranges (the unit interval in each case), a corresponding range of physician FTEE estimates is generated.

Regarding the interpretation of the Reconciliation Strategy, the committee emphasizes the following:

• The formula for deriving FTEE in each of the three components of Equation 6.1 consists of two terms, the Empirically Driven Baseline and the Modifier. Thus, for patient care, resident education, administration, and leaves of absence, the Empirically Driven Baseline is X₁, and the Modifier is b(X₂-X₁).

  This configuration of the Reconciliation Strategy conveys a particular policy perspective: In determining physician requirements for each specialty or program area, the first step is to derive FTEE estimates from a variant of the EBPSM—either the IPF or the PF. The second step is to investigate whether the Baseline FTEE estimate should be modified in light of factors threatening the validity of the empirically based model. As discussed in chapters 3 and 5, these factors fall into two broad groups of data-related problems—simple measurement and recording errors, and "clinically inappropriate" observations relating physician FTEE and workload (i.e., input-output relationships skewed by current VA resource constraints and other factors). To the degree that the validity of the Baseline estimate is threatened, one applies the Modifier. At the extremes, the Modifier can dominate entirely or have no influence at all.

This articulation of the Reconciliation Strategy reflects the committee's view that there are clear advantages—organizational and methodological—to building a physician requirements methodology around an empirically based model—if the important statistical and clinical assumptions are met. If they are not, then modification of the empirically driven estimates, whether through expert judgment staffing assessments or the application of external norms, is in order.

• Operationalizing the Reconciliation Strategy would require two types of policy choices from the VA decision maker. For each FTEE component (i.e., X, R, and C), which empirically based approach should be selected? Likewise, what expert judgment approach (SADI or DSE) should be used in calculating the Modifier? Given these, what are the most appropriate values for the weighting parameters b, c, and d?. Once these parameters are set, the "compromise" between the Baseline and the Modifier is effectively accomplished. Parameter values between the 0-1 endpoints would reflect the VA decision maker's view that "due weight" should be accorded to both the Empirically Driven Baseline and the Modifier.

• Hence, the Reconciliation Strategy offers considerable flexibility in determining physician requirements across specialties and program areas. For specialty A, the "X" component of FTEE might be computed as a weighted average of results from the PF model and the DSE. For specialty B, the "core" approaches to staffing reflected in the weighted average might be the IPF and the SADI. Even assuming that the VA decision maker were to select the same core approaches for both specialties, the weighting parameters b, c, and d could vary between the two.

  Some observers might point out, with concern, that this framework is so flexible that it fails to constrain the VA decision maker—that virtually any FTEE level could be selected. There are two responses to this. First, to determine physician requirements according to the version of the Reconciliation Strategy is to work within the FTEE boundaries established by the empirically based and expert judgment models, specifically the PF, IPF, SADI, and DSE approaches. Not every physician allocation is compatible with the Reconciliation Strategy. Second, the VA decision maker already possesses the authority to establish physician staffing levels. The relevant issue for the committee was how data, from a variety of sources, might best be analyzed and evaluated to derive physician FTEE levels that are "most appropriate" by criteria that are well defined and openly acknowledged.

  It follows that the Reconciliation Strategy should not be viewed as a preset staffing formula, but as a framework for choosing FTEE requirements.

• It frequently would not be practical for a VAMC to realize instantaneously the new "target" level of physician staffing in a given specialty that emerges from application of the Reconciliation Strategy. To achieve and

then accommodate any significant increase in physician FTEE, additional physicians must be recruited; and some adjustments would likely be required in support personnel, equipment, or space.

Hence, where there is a significant difference between the current staffing level and the target derived through the Reconciliation Strategy, the committee recommends that the VA consider phasing in the target by establishing an intermediate target.

An intermediate target should not be viewed as a vehicle for making merely cosmetic or symbolic changes in staffing; rather, it is intended to be a level as close to the target as material considerations permit. The implication is that a VAMC should proceed toward its staffing targets as rapidly as possible, subject to resource and organizational constraints.

These increments (or decrements) in staffing would provide the VA with natural experiments for analyzing prospectively and rigorously whether the new physician FTEE levels lead to the hypothesized changes in access to care, indicators of the quality of care, and other measures of system performance.

Within the "umbrella" of the Reconciliation Strategy, how exactly (by specialty and program area) should VA physician FTEE levels be calculated? On this, the committee sought and received advice from its six specialty and two clinical program panels. Their detailed recommendations are included in Volume II, Supplementary Papers, and summarized in the appendix to chapter 6 of this report. The study has benefited greatly from the panels' analyses and recommendations. But the committee underscores that the conclusions reported below are entirely its own, reflecting, it is hoped, a balanced and multidisciplinary perspective. The discussion below is organized around the three FTEE components delineated in Equation 6.1.

The eight panels have demonstrated on a small scale the types of analyses that the VA decision maker should undertake to determine physician requirements across the system for this dominant component of FTEE. For each of three actual VAMCs studied in depth (four, in the case of psychiatry), the current physician staffing level (including physician FTEE not in the Cost Distribution Report) was noted; the PF and the IPF variants of the EBPSM were applied (as appropriate); and the DSE and the SADI expert judgment models

were brought to bear. Only after considering the current FTEE level and the empirically based estimates and the expert judgment-based estimates did each panel reach a conclusion about appropriate staffing methodology.

Although the panels' conclusions varied (see the appendix to chapter 6), all conducted their analyses within the framework of the Reconciliation Strategy; so should the VA.

The committee's main charge was to develop a methodology, not implement it. The panels' main charge was to test and refine the methodology. Only after the Reconciliation Strategy has been applied to a significantly larger sample of VAMCs will there exist the breadth of empirical information required to reach a generalizable conclusion about whether the PF, IPF, DSE, SADI, or some weighted combination of these is preferred for a given specialty or program area.

On the basis of the analyses summarized in chapters 4 through 7, the committee reached the following conclusions regarding empirically based and expert judgment approaches to analyzing this major component of physician FTEE:

The PF and the IPF are potentially complementary variants of the EBPSM, and either is a viable candidate for helping generate the Baseline estimates for this component of physician FTEE.

The PF allows physician FTEE to be derived by PCA within the VAMC, while taking account of the productive contributions of residents, nonphysician providers, and other factors. The degree to which these substitute for VA staff physicians can be examined.

However, an acceptable PF cannot be estimated for specialties lacking a well-defined PCA. Hence, for laboratory medicine, diagnostic radiology, nuclear medicine, radiation oncology, and anesthesiology, no PF model was presented.

In addition, physician FTEE will be acknowledged in the PF model only to the extent that it is associated statistically with the production of workload. If a given specialty renders care on a given PCA but is not shown statistically to contribute to patient throughput, that specialty's FTEE variable will not be included in the PCA's PF; when total required FTEE for that specialty is subsequently derived for the facility, none will be shown associated with that PCA.

As specified in this study, the IPF generates a direct estimation of physician requirements at the facility level; because of this higher level of aggregation, it is less vulnerable than the PF to measurement errors due to misclassification of FTEE within the VAMC Cost Distribution Report (CDR). The IPF permits statements about statistical confidence to be constructed around physician FTEE predictions (in contrast to the PF, which permits confidence statements about the workload expected from a given set of physician and nonphysician inputs).

However, no acceptable IPF model can be estimated for VA program areas that are multidisciplinary. Thus, there is no IPF presented for either ambulatory care or long-term care.

In contrast to the PF, the IPF acknowledges all FTEE recorded in a given specialty at the VAMC regardless of the degree to which it is associated with the production of workload.

The IPF permits examination of actual versus predicted physician FTEE, by specialty, at a given VAMC, whereas the PF permits analysis of actual versus predicted workload, by PCA, at that same VAMC. Hence, the IPF and the PF can provide complementary insights into the relationship between workload and the physician staffing required to meet it (see chapter 7).

To derive expert judgment FTEE estimates for use in the Modifier term in Equation 6.1, the most promising approach is a methodology built around the SADI.

The specialty and clinical program panel analyses indicate, in sum, that it is feasible to develop SADIs for all specialties and VA program areas. Task time estimates were derived exhibiting strong face validity and yielding physician requirements for selected VAMCs that were generally plausible and acceptable to panel participants (see chapter 5 and the appendix to chapter 6).

Because the SADI focuses on the time required by physicians to perform specific tasks and functions, it is particularly suitable for the procedure-oriented specialties and compatible with all specialties.

Like the DSE, the SADI permits the derivation of physician FTEE requirements for VA programs, services, or procedures that are either in the planning stage or sufficiently new that valid empirical data are not available.

Because DSEs would have to be individually crafted for each VAMC assessed, applying this instrument across the system would be cumbersome and labor intensive.

Hence, the committee recommends the following: the VA, without delay, should apply the SADIs either across the board or to a representative sample of VAMCs; analyze the results; revise the instruments on the basis of what is learned; reapply the SADIs to VAMCs across the system; and, finally, integrate the resulting FTEE estimates into a Reconciliation Strategy-based assessment of physician requirements via Equation 6.1.

Regarding parameter b, denoting the relative weight accorded the Empirically Driven Baseline versus the Modifier in the Reconciliation Strategy, the committee recommends that it be determined on a facility-specific or facility-group basis. This contrasts with a policy of establishing, for each specialty, one value of b (or one range of values) to be applied to all VAMCs.

At any point in time, facilities will differ substantially in how well staffed they are relative to the system norm, in the accuracy of the CDR data allocating physician FTEE to activities and PCAs, and in factors affecting staffing that may

not be captured in any data base. Allowing b to vary gives the VA decision maker the flexibility to translate knowledge of such local factors into the overall determination about the relative emphasis accorded the Baseline and Modifier terms in Equation 6.1.

The proposed methodology is intended to help the VA determine the quantity of physicians, measured in FTEE, required to meet the mission-related demands of the VAMC. But, the committee does recognize that staff physicians in full-time administrative positions in Central Office and at other sites external to the VAMC have contributed significantly to patient care, education, and research. However, the committee regards the determination of FTEE for these purposes as traditionally a matter of administrative discretion and, in any event, beyond its technical competence. Similarly, although determining physician FTEE for full-time administrative positions at the VAMC does fall within the committee's defined purview, that, too, is better calculated on a site-by-site basis rather than through the application of formal staffing models.¹

The amount of research FTEE built into overall physician requirements should be related to measurable indicators of research productivity and excellence. Not all VAMCs should be accorded the same level of research FTEE.

Possible indicators—all potentially computable at the facility level and also by specialty—include the amount of VA and non-VA research funding obtained, the quantity of peer-reviewed papers published in scholarly journals, or (most simply) the amount of FTEE currently allocated by the specialty to research in the VA CDR.

Adopting any such empirically driven approach to determining research FTEE in the Reconciliation Strategy implies that c = 0 in Equation 6.1.

In principle, the committee's preferred approach is to link research FTEE earned to dollars of research support raised. This could be accomplished through specialty-specific statistical analyses taking the following general form: R₁ = f(VA Research Dollars Raised, Non-VA Research Dollars Raised, Specialty-Specific Characteristics, Facility-Specific Characteristics). Once estimated, the model could be used to derive the expected amount of research FTEE, , for a given specialty at a given VAMC as a function of right-hand-side variable values specific to that specialty and VAMC.

A significant limitation, however, is that data presently available systemwide can link research dollars (by funding source) to facility, but not to specialty or program area within the facility. If funding data were collected annually for each VAMC by cost center, specialty-specific models could be estimated directly. (Multidisciplinary research would have to be analyzed in a somewhat more elaborate model that accommodates two or more specialties simultaneously.)

Until the appropriate data emerge, the committee recommends an interim approach in which the VA decision maker allocates research FTEE by specialty on the basis of the specialty's currently reported research FTEE level.

Continuing education for staff physicians should be an important component of any VA quality assurance program. The committee therefore recommends that a certain minimum amount of continuing education FTEE be expected for all specialties at all VAMCs.

The committee proposes that the minimum commitment for any VA physician be no less than 60 hours per year—the time equivalent of what the American Medical Association requires for qualification for its Physician Recognition Award for Continuing Medical Education (American Medical Association, 1986). This translates into about 0.03 FTEE per full-time physician.

The committee regards this as a bare minimum, however, and believes that a higher floor allocation—for example, 80 hours per year—is both defensible and feasible. This would translate into about 0.04 FTEE per full-time physician.

If these minimums are regarded as based on expert judgment, then it is as if d = 1 in Equation 6.1.

Without exception, the specialty and clinical program panels concluded that the non-VA staffing criteria developed in the study were of limited usefulness in determining VA physician requirements. After reviewing these external norm analyses, the committee concurs.

Most analyses involved the application of simple staffing ratios—for example, patient days/physician FTEE (for inpatient and long-term care) and patient visits/physician FTEE (for ambulatory care)—to determine the implied level of appropriate physician staffing at the illustrative VAMCs examined in this study. These ratios were either published or directly computable from published data (e.g., Department of Defense criteria) or else were inferred from observed staffing patterns at selected non-VA treatment sites.

In most instances, applying these simple ratios to derive VA FTEE levels was technically straightforward. But across specialties and program areas, there were recurring concerns: comparability of patients between non-VA and VA sites; variations across sites in the definition of an FTEE; the effects of for-profit incentives on the number and workload production of physicians in certain private-sector sites; and skepticism about whether non-VA staffing patterns have, in fact, emerged from a conscious consideration of their effect on the quality of care.

These difficulties notwithstanding, the committee believes that useful external norms can be developed. A necessary (though not sufficient) condition is that physician staffing ratios be ''conditional'' constructs, computed as a function

of case mix and acuity, the availability of nonphysician personnel, and other factors affecting total requirements. Such ratios could be used to generate implied physician staffing at VAMCs, conditional on these factors.

In other words, external norm criteria should be applied at the level of detail and specificity already characterizing the expert judgment staffing exercises and the empirically based models.

To accomplish this, a detailed examination of physician staffing levels in relationship to workload and other factors affecting physician productivity would need to be undertaken at each non-VA facility selected for analysis. The committee recommends that the VA pursue these more detailed external norm analyses.

The primary purpose of the study has been to develop a physician staffing methodology. Physician requirements have been computed selectively for specific specialties and sites, but this was always for demonstrating or testing a method or model. Thus, there are inherent limitations in the committee's ability to address the question, by specialty and program area, of whether current physician staffing in the VA is adequate overall.

Although the estimated empirically based models were used to derive physician requirements for all VAMCs (see chapter 4), such was not the case here for the expert judgment models. Both the SADI and the final versions of the DSE were applied only to the three (for psychiatry, four) VAMCs chosen as test sites for developing and refining these approaches (see chapter 5). Only for these facilities were estimates made of physician requirements, via the panel analyses, by all applicable empirically based and expert judgment approaches (see the appendix to chapter 6).

Without exception, the panels declined to render a quantitative recommendation about whether the VA system was adequately staffed with physicians. Most panels did reach qualitative judgments about staffing adequacy, however, based on the test-site analyses and the general observations of individual panel members—VA as well as non-VA—with years of experience working in VAMCs.

The committee's own conclusions are as follows:

• Relying solely on analyses performed in this study, it is not possible to reach sound quantitative conclusions on whether current VA physician staffing levels are adequate in the aggregate. Though an important question, it is not one the committee was asked to address.

• However, the approach selected for determining physician FTEE for patient care, resident education, administration, and leaves (component X in Equation 6.1) does bear some logical connection to the qualitative judgment about whether staffing is adequate. To adopt an empirically based model—with its reliance on workload and FTEE data from the current system—for a given specialty or program area at a given point in time is logically consistent with the following qualitative judgment: Although individual VAMCs may have too many or too few of these physicians relative to VA systemwide productivity norms, the specialty or program is, in the aggregate, neither significantly understaffed nor overstaffed.

Adopting either of the expert judgment approaches for a specialty or program area at a point in time is logically compatible with either of two conclusions:

1. While a given VAMC may have too many or too few physicians, the VA system as a whole is inappropriately staffed at that point in time. Hence, the SADI or the DSE becomes the means to help move the system away from the status quo; or

2. The empirically based models are either conceptually inadequate or estimated with flawed data, so that expert judgment approaches are preferred on technical grounds. Note that the committee is not selecting among approaches here, but merely pointing out the logical implications of the relevant choices.

• A major difficulty in drawing valid inferences about VA staffing adequacy is the absence of data relating physician FTEE (in any specialty or program) to measures of patient access and quality of care. Recent efforts by the VA Office of Quality Management to develop quality indices are noted in chapter 2. Analytical models for deriving staffing levels that meet or exceed such quality standards are presented in chapter 7. Until these linkages can be analyzed, inferences about the relationship between physician staffing intensity and patient outcomes will have to be derived by expert judgment, informed by the relevant available data.

• A close reading of the panels' final reports (see Volume II, Supplementary Papers) and their meeting transcripts (unpublished) reveals a recurring theme, enunciated in qualitative terms: In most specialties and program areas, the VA currently has too few physicians in aggregate; in no case does it have too many.

In keeping with the report's focus on methodology rather than the adequacy of specific staffing levels, the committee acknowledges the panels' views, but takes no formal position on their specific conclusions about the adequacy of

current staffing. But, these panel conclusions, emerging after months of careful deliberation by the panels, bear sufficient policy significance to warrant immediate investigation by the VA.

The proposed physician requirements methodology provides the means to do this. Specifically, the following should be undertaken:

After the SADI has been further tested and refined (see chapter 5), the Reconciliation Strategy should be applied across the system to determine which specialties or programs at which VAMCs are significantly understaffed. At a selected sample of these, the VA should provide the additional resources to bring physician staffing up to the recommended target levels (or intermediate target levels, as the local situation dictates).

The effect of improved physician staffing on indicators of access and quality should then be formally evaluated.

In chapter 7, important issues related to the implementation and policy applications of the methodology were discussed. The committee's views can be summarized as follows:

By its very structure and logic, the Reconciliation Strategy implies that the allocation of physician FTEE across the system would become more centrally directed; at present, each VAMC has broad discretion to establish physician FTEE levels, subject only to constraints involving its total budget and total assigned personnel ceiling. Under the Reconciliation Strategy, all facilities would be judged by the same criteria within each specialty or program area. There is the presumption that facilities facing similar mission-related demands would be prescribed similar physician FTEE levels.

The committee was not asked to consider the budgetary costs of meeting VA physician requirements or how, if at all, the methodology could or should be linked to the budget process. However, the committee can envision a resource management policy in which that portion of the VAMC budget allocated to staff physicians is established in accordance with the FTEE targets (and intermediate targets) derived through applications of the Reconciliation Strategy.

The committee does believe that the likelihood of the physician staffing methodology influencing VA physician staffing is substantially greater if the methodology is made an integral part of the budget process at the facility level. Therefore, the committee recommends that the VA take steps to achieve this integration concurrently with the implementation of the methodology.

For the Reconciliation Strategy to be successfully implemented and to improve over time, there must be strong channels of communication between VA

Central Office and each VAMC. And the dialogue must be an active, two-way interchange. There are two reasons why this is crucial.

First, the acceptability of specific physician staffing levels—and of the methodology that produced them—is likely to be greater if they emerge from a process that genuinely engages the local facility.

Second, good two-way communication will enhance the quality of the decision process itself, increasing the likelihood that the physician staffing levels adopted are indeed appropriate. Any broadly applicable methodology for determining VA physician requirements will necessarily use models that are simplifications of reality. No true model will incorporate every factor that could influence the number of physicians required at every VAMC. In addition, certain important variables may have to be omitted simply because the data are missing or inadequate. Measurement errors may occur for some variables in ways known to the local VAMC but not apparent to decision makers in Central Office. Moreover, the greater the flow of good information between Central Office and the VAMCs, the less likely it is that individual participants will be able to ''game the system'' successfully.

How might this dialogue work in practice? Applying the Reconciliation Strategy, VA Central Office would derive, for all physician specialties and program areas at a given VAMC, FTEE targets and (as needed) intermediate targets (see chapter 6). Whatever differences exist between actual and targeted staffing would be communicated to the facility, along with information describing how the targets were computed.

The facility would be expected to respond. If it agreed with the recommendations, there would be little more to debate (except perhaps where the funds would be obtained to meet proposed staffing increases). If the facility takes exception to the targets, it might wish to introduce supporting evidence not generally available at Central Office or discuss further the interpretation of existing data. Thus, the final determination of appropriate targets and intermediate targets for physician FTEE would be as informed as possible.

Also, occasions may arise when the facility would request new (typically additional) physician staffing levels in a specialty or program area as part of a proposed expansion of services. The Reconciliation Strategy could be applied to generate evidence either supporting, or failing to support, the facility's request.

Quite intentionally, the proposed dialogue between Central Office and the VAMCs—and the Reconciliation Strategy itself—is oriented around the interpretation and evaluation of formal models for staffing.

Such models allow all parties in the decision process to analyze a range of "what if" questions important to the interpretation, policy application, and validation of the methodology. In chapter 7, it is shown how management techniques such as sensitivity analysis, outlier analysis, and linear programming can be used to enrich the information base available to VA Central Office and

the VAMCs. These analyses would be facilitated if the physician requirements methodology were incorporated into a larger VA "decision support system" that promotes a comprehensive integration of resource planning and budgeting.

Estimates of future physician requirements hinge crucially on estimates of future patient workload. The models adopted by the committee for projecting inpatient, ambulatory, and LTC workload have several noteworthy features (see chapter 8).

• In their structure and logic, all three represent adaptations of existing workload projection models used presently in VA strategic planning. The major difference in each case is that workload is expressed here in the form of a weighted-work-unit (WWU) index (see chapter 4) rather than in terms of patient days or visits, as in the VA models.

• These models produce workload projections that are adjusted for anticipated changes in the age structure of the veteran population over time. They could be adapted readily to adjust also for changes in the distribution of the veteran population by gender or eligibility-for-care categories.

• The workload projections from these models can be input directly into both the PF and the IPF variants of the empirically based models to derive future physician requirements, by VAMC and PCA within each VAMC. Although these projections are not directly applicable to the expert judgment staffing models, it is shown in chapter 8 how they can be used to obtain indirect estimates of workload at the level of detail required by the SADI and the DSE.

The committee's views about VA-medical school affiliation relationships, presented at length in chapter 9, are summarized below.

The committee has reached a very firm conclusion that the overall effect of affiliations on the VA health care system is strongly positive. These benefits include:

• An improved ability to attract and retain well-qualified physicians and other health professionals;

• A wide spectrum of services provided by a pool of highly qualified physicians, both those on the VA staff and those whose services are made available to the VA through other relationships with the medical schools;

• Access to state-of-the-art tertiary care;

• Participation in the education of physicians, which is a mandated part of the VA mission and which cannot realistically take place currently in the absence of affiliations;

• Participation in medical and health services research, resulting in contributions to medical knowledge and improved health services, that benefits the general population as well as veterans.

Underlying all of the above factors is the assumption, and some inferential indications, that affiliations contribute significantly to improving the quality of patient care.

In other parts of this report, the committee has urged the VA to continue its work, being led by the VA Office of Quality Management, to develop quality-of-care indicators. These indicators will be critical not only for the full development of the physician requirements methodology, but also for a more definitive evaluation of the effect of affiliations.

These analyses should focus not only on structure and process indicators of quality, but on outcome indicators that include, but go beyond, mortality measures. In these efforts, the VA should track closely the extensive efforts and developmental work being done by many health services researchers and health care organizations on outcome-related quality measures.

The committee recommends that the VA explore strategies for developing and expanding affiliations to include facilities that currently are not affiliated. This recommendation follows logically from the previous conclusion that affiliations bring benefits to the VA health care system. Given that conclusion, the committee believes that there is no logical reason not to provide at least some of these benefits to veterans cared for in all VA facilities. Such an expansion of affiliations would assist the recruitment and retention of high-quality staff and promote achievement of the other benefits outlined above.

The committee further recommends that while maintaining and nurturing the current model of affiliations between VAMCs and medical schools, with its emphasis on tertiary care, the VA should work to develop innovative models of affiliation targeted specifically to the chronically ill, including those requiring psychiatric care and rehabilitation services. These innovative models would, in general, be oriented around and give emphasis to ambulatory and long-term care.

The nature of the VA patient population presents special opportunities, and needs, for the development of new models. The Geriatric Research, Education, and Clinical Centers (GRECCs) serve to illustrate a successful model already developed by the VA to meet the particular needs of the population being served. Other opportunities—emphasizing a broad array of research related to these patient care needs, including health services research and research on health outcomes—could continue to make the VA health care system a resource for the benefit of the entire nation, as well as for veterans. Similarly, training opportunities focused in innovative ways on these particular patient care needs could make a major contribution to veterans and to the general population. The VA is in a logical position to support its own purposes and the purposes of the broader society.

In developing these innovative affiliation approaches, the VA should explore the establishment of relationships with other medical institutions in addition to medical schools. The VA already has created the beginnings of a new model of affiliations with a recent program connecting VAMCs to community-based health care institutions not primarily related to medical schools.

The committee believes that this type of extension represents an exciting opportunity that could help the VA meet its physician requirements, especially for primary care, in the years ahead.

Early in the study the committee hypothesized that VA physician requirements—at present, but especially in the future—may be influenced by the availability of certain nonphysician practitioners (NPPs).

The committee's interest in NPPs was spurred by two considerations: (1) a substantial literature indicating that these practitioners can enhance physician productivity while maintaining the quality of care; and (2) the changing demographic structure of the VA patient population, which will increase the demand for ambulatory care and long-term care—arenas in which NPPs may be particularly productive.

In chapter 10 the committee presented recommendations on the present and future role of four types of NPPs: physician assistants, nurse practitioners, certified registered nurse anesthetists, and clinical nurse specialists.

The committee believes that the degree to which these four types of NPPs are utilized either in complementary or substitutive roles has a direct effect on physician requirements. Therefore, the committee urges the VA to account

more precisely for the influence of these NPPs, in both the empirically based and the expert judgment approaches to physician staffing, by incorporating the following:

• For the empirically based models, the VA should establish CDR cost centers for each of these NPPs. At present, the total FTEE of each type of NPP is available at the facility level, but not allocated across PCAs. If each NPP there were given a designated CDR cost center—as is the case presently for physicians (by specialty), nurses, psychologists, and social workers—it would be possible to analyze them explicitly in the PF and the IPF variants of the empirically based models. At present, these NPPs are reflected in the PF and IPF equations only through their inclusion in the SUPPORT/MD and NURSE/MD variables (see chapter 10).

• For the expert judgment models, NPPs are already explicitly recognized (see Figures 5.1 and 5.2). However, in subsequent versions of the SADI and the DSE that the VA may choose to create, these NPPs should be incorporated with greater specificity than at present. In particular, the assumed number of each of the four types in each relevant PCA should be built into these staffing instruments.

From the NPP panel survey data and commentary, the committee concludes that the utilization of NPPs is more dependent on the particular attitudes and knowledge bases of individual physicians than on the training and clinical skill level of the NPP. Before the VA can begin to utilize NPPs in an efficient manner consistent with quality care, ongoing education programs for VA physicians must be established. The committee recommends that this physician education effort be initiated on an ongoing basis with a centralized program for senior VA management staff, and that, over time, programs be established at every VAMC.

To support this recommendation of continuing education for physicians on the role and utilization of NPPs, the committee recommends that the VA pursue and establish, wherever possible, academic affiliations with these NPP training programs. The VA should also require and actively support the participation of NPPs in continuing education related to their roles and functions.

National guidelines on the use of NPPs should be strengthened where they exist, established where they do not, and updated on a regular basis over time. They should be orchestrated in a way that allows the VAMC adequate flexibility for innovation and quality control.

As knowledge about the appropriate and effective use of NPPs continues to grow, the VA should periodically and thoroughly review its national policies on the use of these practitioners. For PAs and CRNAs, such policies already serve to establish the boundaries of practice, by listing specific permissible functions in various arenas of activity; the subset of these activities that may be delegated to the individual NPP has traditionally been determined entirely at the facility level. Because physician attitudes and knowledge bases regarding NPPs vary greatly across (and even within) VAMCs, there is wide variation in the activities actually performed by NPPs. This view was underscored in survey results presented to the committee by its NPP panel.

As analyses emerge indicating that specified functions can be performed efficiently by NPPs with no anticipated loss in quality, the VA should give priority designation to these functions in its guidelines. This information should be communicated promptly to chiefs of staff, service chiefs, and clinicians (to the latter via the continuing education programs recommended above). Similarly, when the weight of evidence indicates that the NPP's performance of a function does not promote efficiency or quality, the function should be removed from the guidelines (if it was there); and this action also should be communicated promptly. The outcome of these studies should influence not only the specific functions that NPPs perform, but their overall roles vis à vis physicians in the patient care process.

At present, explicit national guidelines on the utilization of NPs and CNSs do not exist. The committee urges the VA to develop such guidelines through a careful evaluation of existing evidence on their efficient and appropriate utilization.

To promote the development and diffusion of new information about the appropriate use of NPPs, the VA should support research projects that examine the range of activities now performed by these practitioners across the system. The focus should be on innovative uses of NPPs that hold promise for increasing access to care while not compromising quality.

At present, there are numerous opportunities to observe NPPs in a variety of settings, in different specialties, and for various functions, inside as well as

outside the VA. The VA should take advantage of these ''natural experiments'' to evaluate the appropriate (and inappropriate) uses of NPPs across a range of practice conditions.

The committee concludes that the task of developing a methodology to determine the number of physicians required by the VA is best pursued in an "evolutionary" fashion. The methodology recommended in this report should be regarded not only as a concrete beginning, but as a springboard to further experimentation and analyses. These would serve to test the validity of the statistical and expert judgment models as well as the overall appropriateness of staffing recommendations from clinical and economic perspectives. In the course of this report, the committee has presented a number of proposals for testing, refining, and extending the current methodology.

In what follows, specific steps that the VA should take to launch this evolution are discussed.

The VA should test, evaluate, and revise (as needed) the EBPSM on an ongoing basis. With the demands on the VA health care system in dynamic transition, the EBPSM should not be treated as a static construct.

The VA should consider several options for strengthening the empirical foundation of these models:

• Each VAMC is now required to have a data validation committee. These committees should be actively encouraged to work aggressively at quality control.

• Positive incentives should be instituted for individual physicians and administrators to fill out CDR worksheets accurately—or penalties should be assessed for evident errors.

• For short, concentrated periods, physicians and administrators should be required to track how physician time is being allocated across activities; the results could be compared with the corresponding FTEE allocation in the CDR.

The scope of the VA national data system should be broadened to permit the refinement of existing variables and the construction of potentially important new variables for use in the EBPSM:

• At present, it is not possible to distinguish full-time (FT) and various levels of part-time (PT) physician FTEE in the national CDR accounts, though the data are available in the VA payroll system. This information should be integrated into the CDR accounts to yield specialty-specific observations on the amount of physician FTEE, by FT or PT category, allocated to each PCA at all VAMCs.

• The CDR should be amended so that physician FTEE for resident education, research, and administration not occurring in the PCAs can be clearly distinguished.

• It is also not possible at present to distinguish physicians by subspecialty in the national CDR accounts. Investigations exploring the merits of including subspecialty FTEE in the PF equations and of producing IPFs specific to subspecialty should be undertaken, and they will require this more detailed FTEE data.

• The VA should strongly consider focused, time-limited surveys to collect information, by specialty, on the amount of FTEE contributed at VAMCs by C&A and WOC physicians.

• Data on the type, amount, and vintage of capital equipment affecting the efficient delivery of high-quality care should be made available for each PCA in all VAMCs.

• The CDR should be amended so that it is possible to obtain direct observations on the allocation of residency time, by postgraduate year (PGY), to all PCAs in the VAMC.

• As noted in the previous section, the VA should modify the CDR national accounts so that model-relevant FTEE data for four types of NPPs are available at the PCA level. This would require establishing distinct CDR cost centers for PAs, NPs, CRNAs, and CNSs.

• The strong statistical performance of most PF and IPF equations provides prima facie evidence supporting the validity of the workload measures used. But the VA should consider further analyses testing whether there are other output variables, derivable from existing VA data, that are more sensitively related to physician time requirements.

• Studies of the relationship between the intensity of physician staffing and indicators of the quality of care should be pursued, as indicated earlier.

The committee recommends that the VA periodically review the selection of variables and functional form for each PF and IPF equation.

Over time, a number of factors affecting the PF and the IPF equations can be expected to change, at varying rates: the mix and acuity level of cases presenting at VAMCs, medical technology, practice patterns, the range of services offered by the VA, and the quality and scope of data from the CDR and other sources.

Consequently, it is important that all equations be reestimated periodically to test whether these various secular changes indicate that the models should be modified—either in their mathematical form or in the variables that make them up. The present data systems would permit reanalysis of these equations on an annual basis.

Moreover, as multiple years (and, hence, samples) of data accumulate, it will become possible to undertake certain innovative, split-sample methods of internal model validation, such as bootstrapping.

The committee has recommended that the expert judgment component of the physician requirements methodology be built around application of the SADI. However, the committee does regard the SADIs developed in this study as first-generation instruments, requiring additional evaluation and refinement.

Because the SADI approach emerged late in the study, it was not feasible to use a modified Delphi method, the committee's preferred approach, to derive physician activity-time estimates. The SADI estimates reported in chapter 5 and in the appendix to chapter 6 are based, instead, on staffing judgments elicited through one mail survey of all panel members; in a sense, they can be viewed as the results from the initial iteration of a modified Delphi process.

To build upon this first-generation model, the committee recommends the following:

1. The VA should proceed immediately to apply these SADIs to all VAMCs, or at least a large representative sample. For the four VAMCs analyzed by the specialty and clinical program panels, staff members were able to obtain the required facility-specific workload and related data by phone and mail (on a voluntary basis) in a matter of days.

2. Following an evaluation of these applications, each SADI should be considered for revision. The focus initially would be on:

• Appropriate designation of activity time categories, with special attention to new programs and services (e.g., hospital-based home care);

• Appropriate specification of the type and the range of workload for each category, with special attention to whether case acuity is sufficiently differentiated; and

• Adequate delineation of factors influencing physician productivity, such as residents (by specialty and PGY), NPPs (by type), nursing and support staff, and certain items of capital equipment.

The challenge is to construct a SADI with enough detail to capture significant distinctions while omitting factors that have little influence on physician time allocations.

In this vein, the committee notes that throughout the study there was persistent discussion about the importance of nonphysician personnel of all types (nurses, various support staff, and NPPs—including psychologists and clinical social workers) in promoting the quality and efficiency of VA health care. The potential influence of these various providers on workload productivity was formally acknowledged in both the SADI and the EBPSM (particularly the PF variant)—to the extent that current data permitted.

Subsequent versions of the SADI should be structured to examine more precisely the contributions of these nonphysician providers. This would require that physician activity times for each PCA be estimated as a function of the assumed mix of all nonphysician personnel (not just NPPs) judged to be pertinent to the appropriate operation of that PCA.

Over time, the VA should investigate several other issues important to the validity, reliability, and relevance of the SADI approach, including:

• The reliability and consistency of the expert judges' physician activity time estimates;

• The feasibility of deriving from experts not simply point estimates, but probability distributions for the physician time required to perform various activities in the SADI. As discussed in Volume II, Supplementary Papers, such a probabilistic treatment of the SADI would permit the VA to develop statistical confidence statements about each of the staffing recommendations emerging from application of the instrument.

• The availability and appropriateness of observational (empirical) data from which to derive alternative estimates of these physician activity times;

• The subsequent integration of expert judgment and empirically derived activity time estimates through Bayesian statistical analysis (see Volume II, Supplementary Papers).

To pursue the more detailed external norm analyses recommended by the committee, the VA should initiate a sequence of analyses as detailed below.

• Select a small number of clinical sites whose patient populations and scope of services are reasonably comparable to the VA's;

• From each site, collect data on workload, physician FTEE, nonphysician FTEE, and other descriptive information in sufficient detail that average physician time per unit of workload can be computed conditional upon patient characteristics (e.g., age, DRG classification); the availability of residents, nurses, NPPs, and support staff; and other productivity-influencing factors.

• These non-VA physician task times—which at this point would be at a level of detail comparable to those in the SADI—could be applied to the workload data from any given VAMC to derive an implied total quantity of physician FTEE required. This staffing estimate could then be compared with physician requirements for the facility as derived from the SADI and with the actual level of physician FTEE there.

The VA should explore this and other scenarios for applying norms to all specialties and program areas.

The precision and specificity, and thus the policy usefulness, of the workload projections required by the physician staffing models would be enhanced if veteran utilization of the VA system could be analyzed as a function of factors known to influence the demand for care. These include income, health insurance coverage, perceived quality of care, availability of alternative sources of care, and distance from the VAMC, as well as age, gender, and eligibility-for-care status.

In addition, projection models such as those used by the VA currently and adopted here (see chapter 8) do not exploit the total information embedded in a given data set as efficiently as standard statistically based demand models. In particular, to investigate the joint and possibly interactive influence on utilization of two or more explanatory variables is a much more cumbersome undertaking. Statements about statistical precision and confidence are simply not possible with projection models since they are not derived statistically in the first place.

The committee urges the VA to pursue the patient demand analyses described above.

There are a number of ways the VA could organize and support the analyses recommended above: a task force staffed principally from within the VA and coordinated through the office of the chief medical director; a targeted program of grants for which VA health services researchers, and possibly others, would be invited to compete; a program of contracts to perform specific analytical tasks related to testing or extending the methodology; or some combination of these approaches.

With respect to these alternative approaches, the committee makes no specific recommendation. But it does recommend that the analyses proceed according to a two-phase process, defined roughly as follows:

• Phase I—an intensive period of analysis to evaluate and refine the SADIs, producing second-generation instruments in each case; to begin constructing new data and variables, as recommended, for the EBPSM; to undertake a more intensive application of external norms; and to produce first-generation versions of demand-based workload estimation models for inpatient, ambulatory, and long-term care.

• Phase II—an ongoing operation in which the VA (through its designated analysts) periodically reevaluates and possibly revises the SADIs, the EBPSM, external norms, and the workload models. The implications of these revisions for the content and execution of the Reconciliation Strategy would be analyzed.

With the veteran population aging, with technology ever changing, with practice patterns evolving in the non-VA sector, it is crucial that the physician requirements methodology be reexamined on a regular basis.

The committee estimates that Phase I could be completed within 24 months; Phase II would represent an ongoing commitment by the VA to ensure the continuing quality of its physician staffing policies.

American Medical Association, Office of Physician Credentials and Qualifications. 1986. Information Booklet on the Physician Recognition Award. Chicago, Illinois.

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.