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 247
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act C Analyses of the Sensitivity of the Formula Allocations to Underlying Changes in Input Data This appendix provides detailed analyses of the sensitivity of allocation formulas to changes in the underlying data. The Committee originally intended to conduct extensive “what-if” policy simulations. That is, it intended to compare different factors in the funding formulas to examine the impact of including HIV cases on resource allocations across regions, and to compare the inclusion of HIV cases to other features, such as hold-harmless provisions, set-asides, minimum funding thresholds, and the potential addition of new Eligible Metropolitan Areas (EMAs) based on a more inclusive definition of HIV disease burden. The Committee could not examine the impact of including HIV data in the formula because data on HIV cases were not available from all states, including several key states with a high disease burden. Since data from those states could have a large influence on results, any analyses based on partial data could be very misleading. Nevertheless, the analyses and policy assessment in this appendix highlight the implications of current policies. These findings should allow policy makers to explore the implications of proposed changes in funding allocations (see Chapter 3 and 4). Given that constraint, the Committee chose instead to explore how assessments of the “fairness” of those awards might be influenced by including additional data regarding HIV prevalence. In particular, the Committee examined what the current allocation is, per unit of HIV disease burden, using two reasonable but distinct measures of that burden: the estimated number of living AIDS cases (ELCs) (the measure now used for Ryan White CARE Act [RWCA] resource allocation) and combined
OCR for page 248
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act estimates of HIV prevalence and AIDS prevalence provided by the Centers for Disease Control and Prevention (CDC).1 For states with mature name-based HIV reporting systems, estimates of HIV prevalence were based on data from their individual case-reporting systems. For code-based states or states without mature name-based reporting systems, CDC used modeling to produce the HIV prevalence estimates. California and Massachusetts declined to release CDC’s estimates of HIV prevalence, and thus no data were available for these two states. Given that California and Massachusetts did not permit CDC to share CDC’s modeled estimates of HIV prevalence in these states, we imputed the number of HIV cases for these states by assuming that the proportion of HIV to AIDS cases matched the reported proportion in New York. This is an important limitation. The Committee also employed multiple linear regression analysis to identify predictors of RWCA Title I and Title II funding. In its analyses, the Committee examined “dollar allocations per case” across jurisdictions as a point of departure. The Committee acknowledges that there are many reasons why an equitable system would depart from this standard, including unequal costs of care, unequal need, differences in the efficiency with which jurisdictions apply funds, differences in the quality and comprehensiveness of the existing resource base from one jurisdiction to another, and differences in economies of scale. In some instances, deviations from the “equal dollars per case” standard will highlight disparities to be corrected; in other instances, they will confirm the view that the system is applying appropriate flexibility to its standards to reflect legitimate differences in need from one jurisdiction to another. Viewed in this light, the Committee’s goal is not to hold up equal dollar allocation as an absolute standard, but rather to make explicit the consequences of allocation formulas that are the product of complex political negotiation, epidemiological evidence, and competing conceptions of fairness. Despite these limitations, analyses of current allocations are pertinent to stakeholders who wish to anticipate the distributive impact of changes to current formulas. If the current allocation appears unfavorable to states and EMAs that include a high proportion of reported HIV cases to ELCs, the move to a more inclusive definition of HIV burden may have large 1 The Committee also examined current allocations using estimated AIDS prevalence alone. The differences in allocations using estimated AIDS prevalence and ELCs were not informative, suggesting that any methodological differences between the calculation of ELCs and the calculation of AIDS prevalence is not important for the purposes of identifying allocation variations.
OCR for page 249
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act distributional effects. If, in contrast, the ranking of states and EMA per-case RWCA spending is similar for different measures of HIV burden, epidemiological factors may be less important than other features of RWCA funding allocations in shaping real or perceived funding disparities. The Committee began by examining total RWCA allocations in fiscal year (FY) 2001 for the 50 states, DC, and Puerto Rico, and the nonterritory EMAs. FY2001 was especially pertinent because it was the most recent year in which the Committee could match available surveillance data with RWCA funding allocations. In that year, Title I awards totaled $565,229,972 (HRSA, 2002a). Title II awards were significantly larger, totaling $873,424,373 (HRSA, 2002b). Combined Title I and Title II awards therefore totaled $1,438,654,345. These awards were determined based on the number of ELCs from the previous calendar year. In the year 2000, there were 280,759 ELCs, with 204,298 residing within EMAs (HRSA 2002a,b). If Title I and Title II funds were provided to funding units in strict proportion to the number of ELCs, the nationwide allocations per ELC would have been: Title I: $2,767 per ELC ($565,299,972 divided by 204,298) Title II: $3,111 per ELC ($873,424,373 divided by 280,759) Titles I and II: $5,124 per ELC ($1,438,654,345 divided by 280,759) These summary statistics provide one benchmark of equity with which to compare actual awards per ELC. Figure C-1 illustrates how the FY2001 Title I award per ELC varied across EMAs. With the exception of San Francisco, Title I awards per ELC spanned a narrow range across metropolitan areas (mean $2,754; standard deviation: $174). This uniform allocation reflects the strong role of ELCs in Title I funding. Thus, applying the standard of equal expenditures per ELC, the Title I formula allocations are highly equitable. This uniform pattern of per-ELC Title I expenditure is more open to question through other conceptions of equity, for example if one believes that formula allocations should vary in accordance with per-capita income, with the socioeconomic status of persons living with AIDS, or with other characteristics that systematically vary across EMAs. The large allocation to San Francisco reflects the influence of the hold-harmless provision. Removal of this provision would reduce San Francisco’s allocation to within the reported range for other EMAs. Under current allocation rules, however, removing the hold-harmless protection would have a small effect on other EMAs, which would observe a 2.6 percent increase in their allocation if San Francisco’s allocation were reduced. As noted by GAO (2000), the hold-harmless provision has a small overall effect on allocations to EMAs, yet a large effect on a single EMA.
OCR for page 250
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act FIGURE C-1 FY2001 Title 1 allocations to EMAs per ELC.
OCR for page 251
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act USING HIV DATA IN FORMULAS The Committee then examined other measures of HIV/AIDS burden. Not surprisingly, the number of ELCs was quite similar to estimated AIDS prevalence. Thus, moving from ELCs to AIDS prevalence appears to have a small impact on funding allocations. However, ELCs accounted for only about half of the total combined estimated HIV and AIDS prevalence. Combined estimated prevalence for the 50 states, DC, and Puerto Rico was 651,238 cases, with 449,898 of these cases associated with a specific EMA. If Title I and Title II funds were provided in strict proportion to the total estimated HIV and AIDS prevalence, the nationwide allocations per case would be: Title I: $1,256 ($565,229,972 divided by 449,898), Title II: $1,341 ($873,424,373 divided by 651,238), Titles I and II: $2,209 ($1,438,654,345 divided by 651,238) (Figure C-2) Again the EMA Title I award allocation is roughly proportional to the total number of HIV and AIDS cases (mean $1,198; standard deviation: $175), although using a combined HIV and AIDS measure accentuates the impact of the hold-harmless differential. This may be attributable to the relative maturity of the epidemic (and the resulting disproportion in the number of AIDS cases to HIV cases) in San Francisco compared with other areas (see discussion in Chapter 4). Although this analysis does not directly address alternative allocation rules, EMAs listed on the right-hand side of Figure C-2 are most likely to benefit from an allocation formula that includes HIV cases, because these EMAs include the highest proportion of HIV cases to ELCs. It is important to note that if there is any random fluctuation or measurement error in reported HIV cases, one would expect to observe the variability found in Figure C-2. Because funding allocations are determined by ELCs, localities that understate true HIV prevalence would appear toward the left of the diagram, and those which overstate true HIV prevalence would appear closer to the right. Also note that because all figures for HIV and AIDS cases in California and Massachusetts are hard-wired, there is some uncertainty about the reliability of the results on the left side of Figure C-2. This does not change the relative allocations in other places. Changing the unit of measurement to include HIV cases also appears to produce greater variance in funding allocation per unit of HIV burden. This suggests that states and localities vary in the ratio of HIV to AIDS cases. It is unclear whether such variation reflects variation in the matu-
OCR for page 252
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act FIGURE C-2 FY2001 Title I allocations per HIV/AIDS case.
OCR for page 253
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act rity of the HIV/AIDS epidemic, HIV treatment practices that delay the onset of AIDS, or the quality and aggressiveness of HIV surveillance case finding. Figures C-3 through C-6 illustrate the influence of the Title II formula on state-to-state allocation per unit of HIV disease burden. Non-EMA states receive somewhat greater per-ELC Title II funding than do EMA states ($3,985/ELC for non-EMA states, and $3,003/ELC for EMA states) (Figures C-3 and C-4). This difference is attributable to the 20 percent set-aside built into the Title II structure to favor non-EMA states (see Chapter 2). Title II awards per ELC are similar within each of the two groups: EMA and non-EMA states. Because every state receives at least $500,000 in Title II funding, some small non-EMA states receive high Title II awards per ELC. Kansas and New Hampshire have slightly lower allocations per ELC than other states, though the reasons for such variability are unclear. The Title II funding advantage to non-EMA states diminishes when one shifts the unit of HIV burden to all HIV and AIDS cases (Figures C-5 and C-6). Average awards for non-EMA states ($1,339 per case) are virtually identical to the awards for EMA states ($1,342). This suggests that non-EMA states account for a greater proportion of HIV cases than they do of ELCs. FIGURE C-3 FY2001 Title II allocations per ELC in states without an EMA.
OCR for page 254
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act FIGURE C-4 FY2001 Title II allocations per ELC in states with an EMA. FIGURE C-5 FY2001 Title II allocations per HIV/AIDS case in states without an EMA.
OCR for page 255
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act FIGURE C-6 FY 2001 Title II allocations per HIV/AIDS case in states with an EMA. The Committee limited its analysis to the subgroup of states with more than 1,000 ELCs to avoid “outlier” states that account for a small proportion of overall ELCs and RWCA Title I and II allocations. Within this subgroup, Title I and II combined awards average $5,124 per ELC (and range between $3,290 and $8,285) (Figure C-7). On a per HIV/AIDS case basis, the average is $2,209 (ranging between $1,003 and $2,931) (Figure C-8). Regional differences were evident in both figures, whether or not one uses reported HIV cases. In FY2001, southeastern states represented the four lowest allocations per HIV/AIDS case. The other critical disparity arises between EMA and non-EMA states. In FY2001, 11 out of the 12 states receiving the fewest dollars on a per-ELC basis lacked an EMA. States whose ELCs were highly concentrated within EMAs received more funding than did other states whose EMAs accounted for a smaller proportion of state ELCs. MULTIPLE REGRESSION ANALYSIS The Committee performed several cross-sectional state regressions to examine characteristics associated with FY2001 RWCA funding. These
OCR for page 256
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act FIGURE C-7 FY2001 Total Title I and II allocations per ELC in states with more than 1,000 ELCs. FIGURE C-8 FY2001 Total Title I and II allocations per HIV/AIDS in states with more than 1,000 ELCs.
OCR for page 257
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act analyses also allowed greater scrutiny of the qualitative patterns apparent in earlier figures. The resulting coefficients can then be evaluated in light of stakeholders’ broader judgments regarding RWCA funding allocation. Multiple regression analysis indicates the influence of a specific explanatory variable on formula allocation, presuming that other explanatory variables remained unchanged. Regression analysis is useful to distinguish the likely impact of one variable, holding other correlated variables constant. In conducting simulations and considering policy implications of these results, readers are cautioned that localities and jurisdictions differ across many variables included in these models. The analysis considered both Title I and Title II in examining RWCA resources available to states. Although Title II provides the primary funding to assist states, Title I funding is also pertinent, since these funds may augment or substitute for those provided by states. We examined funding per ELC and per HIV/AIDS case. In all four columns, regressions are weighted by state population size. Given the small sample size of RWCA funding jurisdictions, we chose a deliberately sparse regression specification. Many advocates and policy makers contacted by the Committee noted the possibility that southern states receive lower funding allocations per unit of HIV burden. We therefore included an indicator variable set to one for southeastern states, and set to zero for others. The role of EMAs is widely discussed in comparing state RWCA funding allocations. We therefore included a dummy variable to indicate whether a state included an EMA. To capture potential differences related to race or ethnicity, we also included the percentage of African American or Hispanic residents in each state. To examine whether RWCA funding allocations favor low-income or high-income states, we included the logarithm of state median income in our preferred specification. The coefficients in each column can be interpreted as a dollar change per unit of disease burden in RWCA funding associated with each unit change in the independent variable. Standard errors are shown in parentheses. For example, each percentage-point change in the fraction of the population that is Hispanic is associated with a −10.2 dollar change in Title II allocation per ELC. The standard error of 4.5 indicates that this coefficient is statistically significant at the 0.05 level. Results from African Americans are computed in an analogous fashion. As shown in Table C-1, the presence of an EMA is associated with a significant reduction in RWCA Title II funding, but is associated with a significant increase in overall (Title I plus Title II) funding. Few other variables yielded significant coefficients. Controlling for other factors, southern states received about $318 less per case of HIV/AIDS. The impact of the presence of an EMA is explored in greater detail in Table C-2. For each state that included an EMA, the Committee computed
OCR for page 258
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act TABLE C-1 Weighted Regression Analysis of RWCA Resources in FY2001 Dependent Variable $(Title II)/ELC (Std. err) $(Title I+Title II)/ELC (Std. err) $(Title II)/HIV+AIDS (Std. err) $(Title I+Title II)/HIV+AIDS (Std. err) State has EMA −751.4*** (123) 1096.6*** (187) −130.6+ (74) 560.4*** (123) Southern state 280.9* (134) −111.8 (204) −39.0 (81) −317.7* (134) Percent African American −6.8 (6.9) 1.5 (10.5) 3.3 (4.2) 13.0+ (6.9) Percent Hispanic −10.2* (4.5) 12.7+ (6.8) 5.5* (2.7) 23.1*** (4.5) Logarithm of median income −252.7 (501) −365.1 (761) 48.7 (302) 72.6 (500) R2 0.70 0.61 0.10 0.71 (+p<0.10, *p<0.05, **p<0.01, ***p<0.001) the proportion of ELCs who resided in an EMA. Table C-2 expands the analyses in Table C-1 to include this additional variable. Including the proportion of ELCs within an EMA increases the proportion of variance explained by our regression models. A state with 1,000 ELCs entirely concentrated within an EMA receives approximately $1,814,000 (1000x[2126-312]) more than an otherwise comparable non-EMA state with the same number of ELCs. The analysis of Table C-2 also accounts for the observed disparity between southern and nonsouthern states. The coefficient drops by approximately one-third of its base value, indicating that nonsouthern states with a high concentration of ELCs within EMAs contribute to the observed regional disparity. Per-capita income and race/ethnicity appear to play a small role in explaining RWCA formula awards. Although poorer states receive slightly more resources, this effect was small and statistically insignificant. A one-standard-deviation decline in state log (median income) (0.136 log points) is associated with a $50 (0.136 × 365) increase in per-ELC Title I and Title II resources. States with a high proportion of Hispanics also received greater resources. Although this result was statistically signifi-
OCR for page 259
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act TABLE C-2 Weighted Regression Analysis of RWCA Resources in FY2001 Dependent Variable $(Title II)/ELC (Std. err) $(Title I+Title II)/ELC (Std. err) $(Title II)/HIV+AIDS (Std. err) $(Title I+Title II)/HIV+AIDS (Std. err) State has EMA −241 (251) −312 (324) −66 (160) −236 (227) Proportion of ELCs in EMA −770* (334) 2,126*** (431) −97 (213) 1,202*** (303) Southern state 204 (132) 100 (171) −49 (84) −198+ (120) Percent African American −5.2 (6.6) −3.0 (8.6) 3.5 (4.2) 10.5+ (6.0) Percent Hispanic −4.8 (4.9) −2.1 (6.3) 6.2* (3.1) 14.8*** (4.4) Logarithm of median income −221 (4.8) −454 (618) 53 (3.0) 22.55 (434) R2 0.73 0.75 0.11 (not significantly different from 0) 0.79 (+p<0.10, *p<0.05, **p<0.01, ***p<0.001) cant, it was also quite small in all specifications. The Committee also examined residuals in the regression analyses in Table C-2. Most outliers (positive and negative) were small states that included small populations of individuals diagnosed with HIV and AIDS. Tables C-1 and C-2 should be interpreted in light of four principal limitations. First, HIV and AIDS cases provide an imperfect measure of local HIV/AIDS burden. These measures do not explicitly address distributional priorities or health service delivery concerns that matter to clinicians, policy makers, and RWCA beneficiaries. Differences in costs of care, the severity and complexity of patient needs, and local resources might merit departures from a funding allocation based solely on the number of reported HIV or AIDS cases.
OCR for page 260
Measuring What Matters: Allocation, Planning, and Quality Assessment for the Ryan White CARE Act Second, this analysis does not address the underlying quality of case reporting data. This analysis was conducted using the patchwork of reported and estimated HIV cases made available to the Committee. Results should therefore be interpreted with caution. Moreover, the Committee could not directly investigate whether the inclusion of existing HIV case-reporting data provides a superior measure of true known HIV/AIDS prevalence than the sole use of AIDS data. Despite these limitations, both the descriptive analysis and the multiple regression analysis highlight important features of RWCA funding allocation. Third, this analysis does not address regional medical practice patterns or regional disparities in the costs of AIDS care. As discussed elsewhere in this report, differences in the costs of important inputs might justify additional expenditures in high-cost cities and states. Fourth, this analysis does not address differences in program quality and outcomes across jurisdictions. Such differences might also justify additional expenditures in jurisdictions that contain the most effective or cost-effective programs and interventions. REFERENCES GAO (General Accounting Office). 2000. Ryan White CARE Act: Opportunities to Enhance Funding Equity. GAO/T-HEHS-00-150 Washington, DC: GAO. HRSA. 2002a. FY2002 Ryan White CARE Act Title I Emergency Relief Grants. (Email communication from Steven Young, HRSA, July 22, 2002). HRSA. 2002b. FY2002 Ryan White CARE Act Title II Emergency Relief Grants. (Email communication from Steven Young, HRSA, July 22, 2002).
Representative terms from entire chapter: