Sage Journals: Discover world-class research

Abstract

Background:

Value-based care aims to improve quality and reduce costs, yet racial and ethnic disparities in health outcomes persist, including within Medicare Shared Savings Program accountable care organizations (ACOs). This study introduces the Health Disparities Prevention Quality Index (HDPQI), a novel measure designed to evaluate disparities in preventable hospital admissions, using heart failure (HF) as a case study.

Methods:

We analyzed administrative claims for Medicare Fee-for-Service beneficiaries (2018–2019) attributed to 476 ACOs. The HDPQI stratifies preventable HF admissions by race/ethnicity, sex, and dual eligibility, producing subgroup observed-to-expected (O:E) ratios, which are aggregated to generate ACO-level scores. Higher scores indicate more preventable admissions than expected. We used t-tests to compare HDPQI scores and related metrics between higher- and lower-performing ACOs (top and bottom 50%) and assessed sensitivity to varying HF prevalence rates.

Results:

Higher-performing ACOs (bottom 50%) had an average HDPQI score of 1.13, compared with 1.40 for lower-performing ACOs (top 50%). Lower-performing ACOs exhibited significantly higher total inpatient admissions (1.79 vs. 1.70 per beneficiary, p < 0.001) and greater proportions of HF patients with inpatient admissions (1.46 vs. 1.12, p < 0.001). Subgroup analysis revealed dual-eligible Black females had the highest disparities (O:E ratio = 1.9), while Asian subgroups consistently exhibited lower scores (e.g., Asian males O:E ratio = 0.4). Lower-performing ACOs also showed greater variability in subgroup metrics and higher mean subgroup scores (1.51 vs. 1.26, p < 0.001).

Conclusion:

The HDPQI provides a granular tool to quantify disparities in preventable HF admissions and identify performance gaps. These results highlight the importance of subgroup-specific strategies to advance equity within ACOs and lay the foundation for validating the HDPQI for broader applications in value-based care.

Introduction

Value-based payment models, like the Medicare Shared Savings Program (MSSP), have had limited success in reducing health disparities, and performance measures to evaluate progress toward health equity are limited.^1,2 The scarcity of performance measures for health equity is partly due to insufficient data collection and reporting standards, and there remains no consensus on the best method to measure health care equity.^3,4 Incorporating health equity performance measures into value-based purchasing programs and quality reporting could incentivize health care organizations to target quality improvement efforts toward historically marginalized groups.⁵ A health equity measurement approach can be described as a method for assessing how well an organization’s health care quality efforts address and reduce disparities in health and care outcomes at the population level, particularly by improving the care and health of patients who face greater social risks.⁶

An approach for promoting health equity in the Medicare population is leveraging Medicare’s value-based purchasing programs, quality reporting initiatives, and confidential feedback reports to promote quality improvements. Recognizing this need, tools such as the Centers for Medicare and Medicaid Services (CMS) Health Equity Index (HEI) and Health Equity Summary Score (HESS) have been developed to advance equity in Medicare Advantage plans.^7,8 The HEI focuses on aggregate contract-level performance among Medicare beneficiaries with specified social risk factors (e.g., dual eligibility, low-income subsidy, disability status) and assigns broad scores based on top-third, middle-third, and bottom-third rankings of these measures. While useful for tracking high-level equity trends, it lacks subgroup granularity and does not identify specific populations experiencing disparities within health care organizations. The measure is to be included in the CMS star ratings program in 2027.⁹

Similarly, the HESS aggregates scores for Medicare Advantage plans, incorporating both cross-sectional quality metrics and improvements for specific social risk factors (e.g., race/ethnicity, dual eligibility).⁷ However, it combines these measures into a single blended score, potentially masking disparities within individual subgroups beyond a single dimension of social risk. It also focuses on relative rather than absolute disparities, potentially overlooking universally poor outcomes across all groups or progress at the expense of groups with lower or different social risks.¹⁰

These indices can fail to provide actionable data for targeted interventions, making it challenging for health care organizations to pinpoint and ameliorate disparities affecting particular demographic groups. The objective of this study was to demonstrate an approach—the Health Disparities Prevention Quality Index (HDPQI)—for measuring disparities that could support performance evaluation and quality improvement efforts among Medicare accountable care organizations (ACOs). The HDPQI provides subgroup-specific metrics that highlight disparities at the within-organization level, allowing ACOs to pinpoint inequities and develop targeted interventions. By disaggregating performance by subgroups (e.g., race, ethnicity, dual eligibility), the HDPQI enables more precise identification of disparities and supports tailored strategies to address them.

This article explores the utility of the HDPQI in identifying subgroup-level inequities and comparing performance across ACOs.¹¹ The measure aims to enable health care organizations to pinpoint areas of underperformance and develop targeted strategies to address disparities. We use preventable heart failure (HF) admissions as a use case. We first outline how the measure is calculated, then apply it to Medicare beneficiaries attributed to MSSP ACOs, and analyze its score distribution across Medicare ACOs. Finally, we discuss the potential implications of measures of this kind for public reporting and payment programs.

Methods

Data sources

We used administrative claims from the 100% Medicare Fee-for-Service (FFS; Parts A, B, and D) databases between 2018 and 2019. The claims contain information on enrollee demographics (e.g., race and ethnicity), diagnosis and procedure codes, and site-of-service codes. The Medicare Beneficiary Summary Files (MBSF) were used to identify race and ethnicity and whether an enrollee was dually eligible for Medicare and Medicaid. Race and ethnicity classifications were derived from the research triangle institute (RTI) race/ethnicity variable, which is validated for claims-based research and accounts for variations in reporting in Medicare enrollment files.¹²

Health Disparities Prevention Quality Index

We developed a prototype performance measure—HDPQI—an outcome measure designed for ACOs. The HDPQI segments preventable HF admissions by racial and ethnic groups—specifically, non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic Asian populations. It is further stratified by sex (male or female) and dual eligibility for Medicare and Medicaid, resulting in 16 distinct subgroups. While these indicators of social status are limited, this approach represents an effort to move beyond single-dimensional analyses.

The metric calculates the excess number of preventable HF admissions for beneficiaries attributed to ACOs, benchmarking this against expected values for each demographic subgroup within the measurement period. This expected rate is adjusted for the subgroup’s population size within the ACO, the national HF diagnosis rate among ACOs, inpatient service utilization propensity, and inpatient service utilization frequency. The HDPQI aggregates weighted disparities in preventable HF admissions across all subgroups to provide both overall ACO-level scores and actionable subgroup-level metrics, emphasizing granular insights into equity performance.

The measure calculation is based on a quality improvement tool developed and implemented in the Sutter Health System of Northern California.¹¹ Definitions and calculations for the metrics used in the HDPQI, including subgroup population size within the ACO, the national HF diagnosis claims rates, inpatient service utilization propensity, and inpatient service utilization frequency, are provided in Supplementary Appendix. Table 1 includes a summary of the measure calculation.

Table 1.

Steps for Calculating the Health Disparities Prevention Quality Index

Step	Description
Numerator: observed admissions	The actual number of preventable HF admissions within each subgroup, identified using ICD-10-CM codes. Excludes transfers from skilled nursing facilities and cardiac procedures to focus on preventable admissions.
Denominator: expected admissions	Expected number of preventable HF admissions, calculated using: $E_{i} = {Subgroup Population Size}_{i} \times National H F Diagnosis Rate \times {Utilization Propensity}_{i} \times {Utilization Frequency}_{i}$ Where: Subgroup population size: Total number of individuals in subgroup iii attributed to the ACO. National HF diagnosis rate: Baseline HF proxy prevalence in the Medicare Fee-for-Service (FFS) population. Utilization propensity: Proportion of subgroup members with ≥1 inpatient admission during the study period. Utilization frequency: Average number of HF admissions per user in subgroup iii.
Observed-to-expected (O:E) ratio	The ratio observed-to-expected admissions for each subgroup: O:E ratio = Observed admissions/Expected admissions. Interpretation: Ratio = 1.0: Observed matches expected (parity). Ratio > 1.0: Excess admissions, indicating disparities. Ratio < 1.0: Fewer than expected admissions, suggesting better-than-average performance.
Weighted disparity score	Calculated as: $Weighted score = (\frac{O_{i}}{E_{i}} - 1) \times O_{i}$ Weights disparities by observed admissions, emphasizing larger populations or higher disparities.
Aggregate HDPQI score	Sum of weighted subgroup disparities normalized by total observed admissions: $HDPQI = \frac{\sum_{i = 1}^{16} (\frac{O_{i}}{E_{i}}) - 1) (O i)}{(\sum_{i = 1}^{16} O_{i})}$ Creates comparability across ACOs by accounting for subgroup size and population differences.

Preventable HF admissions are identified using ICD-10-CM codes, excluding transfers and cardiac procedures. Expected admissions adjust for subgroup population size, national HF diagnosis rate, utilization propensity, and utilization frequency. The observed-to-expected (O:E) ratio compares observed admissions to expected values (1.0 = parity). The HDPQI aggregates weighted subgroup disparities, normalized by total observed admissions, for comparability across ACOs.

ACO, accountable care organizations; HDPQI, Health Disparities Prevention Quality Index; HF, heart failure.

Inclusion and exclusion criteria

Medicare beneficiaries were included if they met the following criteria: (1) were 65 years or older as of January 1, 2019; (2) had Medicare Part A and B FFS coverage (i.e., not in managed care); (3) had 12 months of FFS enrollment during the calendar year; and (4) were attributed to a single ACO for all four quarters of the year. Beneficiaries were considered to have had an HF admission if they had at least one admission in the year with a principal diagnosis for HF. Beneficiaries were excluded if they had HF admissions resulting from specific circumstances, such as transfers from skilled nursing facilities or admissions for cardiac procedures, to ensure a focus on preventable admissions.¹³ The MBSF was used to verify coverage details and to set demographic flags for race/ethnicity (Asian, Black, Hispanic, and White), sex (male, female), and dual eligibility. A beneficiary is considered dually eligible if this status is maintained for all months of enrollment during the calendar year.

Measure calculation

The HDPQI includes two primary components: subgroup scores and the aggregated overall ACO score. Subgroup scores are calculated using the observed-to-expected (O:E) ratios of preventable HF admissions, adjusted by each subgroup’s excess over expected. These subgroups are categorized based on sex, race/ethnicity, and dual eligibility, comprising 16 distinct categories. We used the same approach for all demographic groups, including potentially advantaged groups (e.g., White, male, nondual), to present all available data and link improvement to any particular group. This approach recognizes that beneficiaries’ diverse attributes confer advantages and disadvantages, which performance measures may not fully capture. The ACO-level summary scores provide an aggregate measure of subgroup scores.

Numerator calculation

Preventable admissions were defined as inpatient hospitalizations for HF, identified using ICD-10-CM codes. For consistency, the term “admissions” is used throughout to describe these hospitalizations. The observed number of preventable HF admissions is determined using specific ICD-10-CM diagnosis codes for HF, as outlined by the Agency for Health Care Research and Quality PQI 8 guidelines (Supplementary Appendix). This count includes new HF admissions, excluding any transfers from other facilities or patients who have undergone certain cardiac procedures, to maintain accuracy in tracking new admissions.

Denominator calculation for the HDPQI

The expected number of preventable HF admissions for each ACO was calculated using a standardized approach that incorporated subgroup-specific characteristics, health care utilization patterns, and population size. For each subgroup, the expected number of HF admissions was determined by multiplying the subgroup population size, the national HF diagnosis rate, the utilization propensity, and the utilization frequency. Subgroup population size represents the total number of individuals within a subgroup attributed to an ACO. This value is used to calculate expected admissions for the subgroup, ensuring that expected rates account for the subgroup’s size relative to the total ACO population. The national HF diagnosis rate among the ACO-attributed population is used as a proxy for the baseline prevalence of HF in the population. Utilization propensity is defined as the proportion of individuals within a subgroup who had at least one inpatient admission during the study period, capturing variability in health care-seeking behavior. Utilization frequency represents the average number of inpatient admissions per user within the subgroup, accounting for differences in health care use intensity. The expected admissions for subgroup i were calculated using the formula: $E_{i} = {Subgroup Population Size}_{i} \times National H F Diagnosis Rate \times {Utilization Propensity}_{i} \times {Utilization Frequency}_{i},$ where $E_{i}$ is the expected number of preventable HF admissions for subgroup i. The O:E ratio for each subgroup compares observed admissions ( $O_{i}$ ) to expected admissions $E_{i}$ , where a ratio of 1.0 indicates parity, and values greater than 1.0 reflect excess admissions. The ratio was calculated for each subgroup and weighted based on observed admissions to derive the HDPQI.

Calculation of the HDPQI

To compute the HDPQI, we calculated a weighted excess of admissions across all subgroups. By summing the weighted excess of preventable HF admissions across subgroups—for example, non-Hispanic Black men eligible for both Medicare and Medicaid—a summary score is calculated for each ACO. 1.

Calculate O:E ratios: For each subgroup, the ratio ( $\frac{O_{i}}{E_{i}}$ ) represents the observed admissions relative to the expected admissions. A value of 1.0 indicates parity (observed matches expected), while values greater than 1.0 indicate excess admissions.

Weight by observed admissions: The ratio ( $\frac{O_{i}}{E_{i}}$ ) is multiplied by $O_{i}$ , giving greater emphasis to disparities in subgroups with larger populations or higher numbers of admissions. The measure reflects both the magnitude of disparity and the size of the affected subgroup.

Aggregate across subgroups: The weighted disparities are summed across all subgroups within an ACO to create a total weighted disparity score.

Normalize by total admissions: Dividing by the total number of observed admissions ( $\sum_{i} O_{i}$ ) ensures the HDPQI score is comparable across ACOs of different sizes, reflecting the proportionate impact of disparities within each ACO. To compute the HDPQI, we used the following formula:

HDPQI = \frac{\sum_{i = 1}^{16} (\frac{O_{i}}{E_{i}}) - 1) (Oi)}{(\sum_{i = 1}^{16} O_{i})}

Here, $O_{i}$ is the observed number of preventable HF admissions for subgroup i within an ACO. $E_{i}$ is the expected number of preventable HF admissions for subgroup i, adjusted for (A) the subgroup’s population size within the ACO, (B) national HF diagnosis rates (reflecting the distribution of HF diagnoses across the ACO-attributed population), (C) inpatient service utilization propensity (the proportion of individuals with at least one inpatient admission), and (D) inpatient service utilization frequency (the average number of admissions among users). The HDPQI reflects both the magnitude of disparity and the size of the affected subgroups, providing a proportional measure of inequities in preventable HF admissions across ACOs. Scores above 1.0 indicate a disparity, showing more admissions than expected and suggesting possible areas of overutilization or higher HF diagnosis among specific subgroups. For example, an index score of 1.41 indicates that an ACO has 41% more preventable HF admissions than expected.

Analysis

ACOs were categorized based on HDPQI scores, where the bottom 50th percentile represents higher performance (less disparity) and the top 50th percentile represents lower performance (greater disparity). This approach allows for balanced group sizes and avoids arbitrary thresholds for performance classification, aligning with methods commonly used in similar evaluations of ACO performance.¹⁴ T-tests compared key characteristics (e.g., HDPQI scores, subgroup-specific metrics) between high- and low-performing ACOs, assessing disparities and patterns of equity performance. Subgroup-specific metrics included O:E ratios and adjusted admission rates. These comparisons aim to highlight disparities and assess patterns of performance within and across ACOs. Sensitivity analyses evaluated the impact of different HF diagnosis rates (e.g., subgroup-specific vs. national rates) (Supplementary Appendix). All analyses were completed in Stata 18.0.

Results

A total of 476 ACOs were eligible for analysis, with an average of 2,711 attributed beneficiaries per ACO (Table 2). Subgroup analysis (Table 3) revealed significant disparities in preventable hospital admissions across racial/ethnic and dual eligibility groups. Dual-eligible beneficiaries consistently exhibited higher HDPQI subgroup scores, indicating excess preventable admissions. For example, dual-eligible Black females had an HDPQI score of 1.9 compared with 0.8 for nondual Black females. Similarly, dual-eligible White males had a score of 1.9, whereas nondual White males scored 1.0. Asian subgroups consistently showed lower HDPQI scores, such as 0.4 for Asian males (nondual), suggesting fewer-than-expected preventable admissions compared with other groups.

Table 2.

Summary Characteristics of Medicare Shared Savings Accountable Care Organizations (n = 476)

Variable	Mean	SD	Min	Max
Total beneficiaries attributed to the ACO	2,711	26,867	223	586,951
HF admissions	225	2,009	10	43,867
Beneficiaries with an HF admission	153	1,274	8	27,809
Total admissions	909	8,453	101	184,696
Total beneficiaries with an admission	588	5,081	62	111,005
National HF diagnosis rate among ACO-attributed beneficiaries	0.05	0.00	0.05	0.05
Total admissions per beneficiary = (Total admissions [utilization]/Total attributed beneficiaries)	2	0	1	3
Expected number of beneficiaries with HF based on total attributed beneficiaries and diagnosis rate = (National HF diagnosis rate * Total beneficiaries attributed to the ACO)	113	1,222	9	26,677
Proportion of people with HF who sought care = (Total beneficiaries with an HF admission/Expected number of beneficiaries with HF based on total attributed beneficiaries and HF diagnosis)	1	0	0	3
Expected HF admissions	200	2,095	17	45,753
HDPQI ratio O:E	0.93	0.33	0.14	2.14
Weighted sum of excess	0.02	0.01	0.00	0.09
HDPQI subgroup scores	1.38	0.29	1.00	3.22
HDPQI ACO scores	1.27	0.19	1.00	2.50

Data source: Centers for Medicare and Medicaid Services, January 2018–2019 Financial and Quality Data for Accountable Care Organizations participating in the Medicare Shared Savings Program.

Notes: Total admissions per beneficiary is calculated as the total number of admissions (utilization) divided by the total number of attributed beneficiaries for each ACO. Expected number of beneficiaries with HF is calculated as the national HF diagnosis rate multiplied by the total attributed beneficiaries for the ACO. The proportion of people with HF who sought care is calculated as the total number of beneficiaries with an HF admission divided by the expected number of beneficiaries with HF.

ACO, accountable care organizations; HDPQI, Health Disparities Prevention Quality Index; HF, heart failure; SD, standard deviation.

Table 3.

Subgroup-Specific HDPQI Metrics for Medicare Beneficiaries Attributed to ACOs

Subgroup	Attributed beneficiaries	Inpatient admissions per beneficiary	Expected beneficiaries with HF	Proportion with HF and >1 admissions	Expected HF admissions	HDPQI subgroup scores (O:E ratio)
White male (ND)	7,201 (7,446)	1.633 (0.08)	327.3 (338.4)	1.0 (0.2)	533.2 (547.8)	1.0 (0.2)
White male (D)	472.9 (499)	1.9 (0.2)	21.5 (22.7)	2.2 (0.5)	41.3 (42.1)	1.9 (0.5)
White female (ND)	9,404 (9,805)	1.6 (0.08)	427.4 (445.6)	0.8 (0.2)	669.5 (692.2)	0.8 (0.2)
White female (D)	1,057 (1,072)	1.8 (0.1)	48.0 (48.7)	2.0 (0.4)	86.7 (86.4)	1.8 (0.4)
Black male (ND)	410.1 (504.5)	1.8 (0.3)	17.9 (22.7)	1.2 (0.5)	31.9 (40.6)	1.1 (0.5)
Black male (D)	124.5 (168.8)	2.05 (0.4)	4.9 (7.2)	2.4 (1.3)	10.4 (15.1)	1.8 (1.1)
Black female (ND)	635.8 (822.3)	1.7 (0.2)	27.6 (37.0)	1.0 (0.5)	46.4 (61.8)	0.8 (0.4)
Black female (D)	273.3 (375.1)	1.9 (0.3)	11.4 (16.4)	2.3 (1.1)	22.5 (31.3)	1.9 (0.9)
Asian male (ND)	73.3 (112.3)	1.6 (0.5)	2.6 (4.6)	0.6 (0.4)	4.4 (7.7)	0.4 (0.4)
Asian male (D)	71.5 (168.7)	1.7 (0.5)	2.0 (6.2)	1.2 (1.5)	3.8 (10.7)	0.6 (0.6)
Asian female (ND)	98.5 (166.6)	1.5 (0.4)	4.1 (7.4)	0.4 (0.4)	6.4 (11.2)	0.3 (0.3)
Asian female (D)	102.1 (270.8)	1.6 (0.5)	3.5 (10.8)	0.9 (1.0)	5.9 (17.2)	0.6 (0.6)
Hispanic male (ND)	47.2 (54.0)	1.6 (0.5)	1.4 (2.2)	0.9 (1.2)	2.4 (3.6)	0.4 (0.6)
Hispanic male (D)	72.6 (99.0)	1.8 (0.5)	2.1 (3.9)	1.5 (1.6)	4.2 (7.4)	0.8 (0.8)
Hispanic female (ND)	54.1 (66.7)	1.5 (0.5)	1.8 (2.7)	0.5 (0.6)	3.0 (4.4)	0.3 (0.4)
Hispanic female (D)	116.4 (186.4)	1.7 (0.6)	3.8 (7.5)	1.3 (1.4)	7.1 (13.9)	0.7 (0.7)

Data source: Administrative claims from the 2018 to 2020 100% Innovator Research Database, including Parts A, B, and D for all Medicare Fee-for-Service beneficiaries attributed to ACOs participating in the Medicare Shared Savings Program.

HDPQI subgroup scores: O:E, ratio comparing observed with expected HF admissions. Scores >1.0 indicate excess preventable admissions, signaling potential inequities.

ACO, accountable care organizations; D, dually eligible for Medicare and Medicaid; HF, heart failure; ND, not dually eligible for Medicare and Medicaid.

Figure 1 illustrates the distribution of HDPQI scores among ACOs. Higher-performing ACOs (bottom 50%) cluster tightly around a score of 1.0, indicating more consistent performance. In contrast, lower-performing ACOs (top 50%) demonstrate greater variability, with scores ranging from 1.0 to over 2.0. This variability highlights the broader range of quality and equity challenges among lower-performing ACOs.

FIG. 1.

Distribution of HDPQI scores among Medicare ACOs. Data source: Centers for Medicare and Medicaid Services, January 2018–2019 Financial and Quality Data for Accountable Care Organizations participating in the Medicare Shared Savings Program. Notes: The kernel density plot illustrates the distribution of HDPQI ACO scores for the bottom 50% (higher-performing ACOs) and top 50% (lower-performing ACOs). Lower-performing ACOs show greater variability in scores, while higher-performing ACOs cluster around a score of 1.0, indicating more consistent performance. ACO, accountable care organizations; HDPQI, Health Disparities Prevention Quality Index.

Table 4 compares key characteristics between the highest- and lowest-performing ACOs based on HDPQI scores. Lower-performing ACOs had significantly higher inpatient utilization, with 1.79 admissions per beneficiary compared with 1.70 in higher-performing ACOs (p < 0.001). Similarly, the proportion of beneficiaries with HF who sought care was higher in lower-performing ACOs (1.46 vs. 1.12, p < 0.001), as were O:E HF admissions ratios (1.08 vs. 0.79, p < 0.001). Subgroup-specific HDPQI scores were also significantly higher for lower-performing ACOs (1.51 vs. 1.26, p < 0.001), indicating more pronounced disparities in preventable admissions.

Table 4.

Bivariate Comparisons Between Highest and Lowest Performing ACOs Using the HDPQI for HF

	Top 50th percentile (n = 239)					Bottom 50th percentile (n = 237)
	Mean	SE	SD	CI low	CI high	Mean	SE	SD	CI low	CI high	p value
Total beneficiaries attributed to the ACO	4,133.84	2,450.74	37,887.58	−694.08	8,961.76	1,276.36	66.91	1,030.09	1,144.54	1,408.18	0.25
HF inpatient claims	319.45	183.20	2,832.27	−41.46	680.36	130.31	7.66	117.89	115.22	145.40	0.30
Beneficiaries with an HF admission	213.79	116.14	1,795.46	−15.00	442.59	91.09	5.84	89.91	79.58	102.59	0.29
Total inpatient claims	1,340.38	771.09	11,920.74	−178.65	2,859.41	473.06	23.86	367.37	426.05	520.07	0.26
Total beneficiaries with an inpatient claims	858.21	463.33	7,162.84	−54.53	1,770.95	315.49	16.74	257.77	282.50	348.48	0.24
Total admissions per beneficiary	1.70	0.01	0.19	1.68	1.73	1.79	0.01	0.17	1.77	1.81	0.00
Expected number of beneficiaries with HF based on total attributed beneficiaries and HF diagnosis rate	176.23	111.44	1,722.79	−43.30	395.77	48.29	3.02	46.43	42.35	54.23	0.25
Proportion of people with HF who sought care	1.12	0.02	0.38	1.07	1.17	1.46	0.03	0.42	1.41	1.51	0.00
Expected HF admissions	307.46	191.09	2,954.22	−68.99	683.91	91.17	5.21	80.14	80.91	101.42	0.26
HDPQI ratio O:E	0.79	0.02	0.29	0.76	0.83	1.08	0.02	0.30	1.04	1.12	0.00
Weighted sum of excess (HEI)	0.01	0.00	0.00	0.01	0.01	0.03	0.00	0.01	0.02	0.03	0.00
Subgroup scores	1.26	0.01	0.22	1.23	1.29	1.51	0.02	0.30	1.47	1.55	0.00
ACO scores	1.13	0.00	0.06	1.12	1.14	1.40	0.01	0.17	1.38	1.43	0.00

Data source: Centers for Medicare and Medicaid Services, January 2018–2019 Financial and Quality Data for Accountable Care Organizations participating in the Medicare Shared Savings Program.

The table presents t-tests comparing higher-performing ACOs (bottom 50% of HDPQI scores) and lower-performing ACOs (top 50%). Total inpatient admissions per beneficiary = (Total admissions [utilization]/Total attributed beneficiaries); Expected number of beneficiaries with HF based on total attributed beneficiaries and HF diagnosis rate = (National HF diagnosis rate * Total beneficiaries attributed to the ACO) and Proportion of people with HF who sought care = (Total beneficiaries with an HF admission/Expected number of beneficiaries with HF based on total attributed beneficiaries and HF diagnosis rate).

ACO, accountable care organizations; CI, confidence interval; HDPQI, Health Disparities Prevention Quality Index; HEI, Health Equity Index; HF, heart failure; SD, standard deviation; SE, standard error.

Discussion

We found variation in HDPQI HF scores between ACOs, which suggests a performance gap (i.e., demonstrating the opportunity for improvement). The finding is expected given the high rates of preventable HF admissions reported among ACOs and evidence that suggests lower performance is associated with racial and ethnic disparities masked within overall performance scores.^15,16 These disparities in utilization patterns underscore the equity challenges facing lower-performing ACOs. By providing both ACO-level summary scores and detailed subgroup scores (used to calculate the summary scores), the HDPQI facilitates peer comparisons. Summary scores allow for quick benchmarking, while the subgroup scores offer the insights necessary for targeted improvements and shared learning, ultimately driving more effective efforts to reduce health disparities.

The HDPQI leverages readily available data and can be integrated into quality dashboards to enable ACOs to identify improvement areas and direct interventions more effectively.¹⁷ Its transparent format enhances usability for administrators, offering an alternative to complex models.¹⁸ The HDPQI can also be adapted to various data sources and contexts, enabling its application beyond ACOs to other entities like health departments and health plans. It can be used to facilitate the assessment of disparities across multiple types of demographic strata (e.g., LGBTQ+), thereby widening its scope of use and impact.

Limitations

Our study, demonstrating the feasibility of calculating the HDPQI, has some limitations and areas for future research. While race/ethnicity misclassification is consistent across claims data, geographic variation in ACO patient composition and language access may result in differential misclassification, particularly for Hispanic and Asian subgroups. This variability could influence subgroup-specific HDPQI scores and should be considered in the interpretation of results. Our approach uses an existing measure to simplify interpretation and allow detailed analysis of subgroups at a fixed time point, consistent with current practices. We also employ a within-group comparison with ensure that ACO scores proportionally represent subgroup data, and a case-mix adjustment similar to the MSSP to account for variations in HF severity.

Conclusion

We developed a health equity performance measure for shared savings ACOs in this cross-sectional study. We found evidence of a performance gap across ACOs using the HDPQI. From a practical standpoint, the recently announced CMS ACO Realizing Equity, Access, and Community Health model could use the HDPQI or similar methods to evaluate improvements in reducing racial and ethnic disparities over time.^19,20 This study focused on the initial development and application of the HDPQI as a tool to measure disparities in preventable hospital admissions. Future research will evaluate the measure’s reliability and validity to further establish its utility.

Footnotes

Authors’ Contributions

A.A.: Secured the funding for this work, involved in the methodology, formal analysis, and writing of the original draft of the article. B.W., K.C., A.T.D., and M.A.: Instrumental in the conceptualization and methodology of the performance measure, contributed to the writing of the original draft of the article. N.M.: Played a key role in the data curation and formal analysis.

Author Disclosure Statement

All the authors declare no financial or personal interests to disclose.

Funding Information

This work was funded by the Preparedness and Treatment Equity Coalition.

Supplementary Material

Abbreviations Used

References

Liao

, Navathe

. What comes next in prioritizing equity in payment? The ACO REACH model. Health Aff Forefront, 2022; doi: 10.1377/forefront.20220404.728371/full/

Hoyer

, Dee

, O’Leary

, et al. How do we define and measure health equity? The state of current practice and tools to advance health equity. J Public Health Manag Pract, 2022; 28(5):570–577; doi: 10.1097/PHH.0000000000001413

Harrington

, Washington

, Paliani

, et al. A new effort to address racial and ethnic disparities in care through quality measurement. Health Aff Forefront, 2021; doi: 10.1377/forefront.20210907.568444/full/

Keppel

, Pamuk

, Lynch

, et al. Methodological issues in measuring health disparities. Vital Health Stat 2, 2005(141):1–16.

Anderson

, O’Rourke

, Chin

, et al. Promoting health equity and eliminating disparities through performance measurement and payment. Health Aff (Millwood), 2018; 37(3):371–377; doi: 10.1377/hlthaff.2017.1301

Martino

, Ahluwalia

, Harrison

, et al. Developing Health Equity Measures. Office of the Assistant Secretary for Planning and Evaluation: Washington, DC; May 21, 2021. Available from: https://www.rand.org/pubs/external_publications/EP68644.html

Agniel

, Martino

, Burkhart

, et al. Incentivizing excellent care to at-risk groups with a health equity summary score. J Gen Intern Med, 2020; 35(5):1572–1579; doi: 10.1007/s11606-019-05332-6

Jacobs

, Schreiber

, Seshamani

. The CMS strategy to promote equity in quality and value programs. JAMA Health Forum, 2023; 4(10):e233557; doi: 10.1001/jamahealthforum.2023.3557

McNamara

, Rudy

. A signal for health equity: Differences in whole health and social needs among medicare advantage subpopulations. Innov Aging, 2023; 7(Supplement_1):981–981; doi: 10.1093/geroni/igad104.3826

10.

Adams

. Charting the course toward more equitable healthcare systems. Med Care, 2023; 61(1):1–2; doi: 10.1097/MLR.0000000000001770

11.

Pressman

, Lockhart

, Petersen

, et al. Measuring health equity for ambulatory care sensitive conditions in a large integrated health care system: The development of an index. Health Equity, 2019; 3(1):92–98; doi: 10.1089/heq.2018.0057

12.

Jarrín

, Nyandege

, Grafova

, et al. Validity of race and ethnicity codes in Medicare administrative data compared with gold-standard self-reported race collected during routine home health care visits. Med Care, 2020; 58(1):e1–e8; doi: 10.1097/MLR.0000000000001216

13.

Agency for Healthcare Research and Quality. AHRQ PQI Technical Documentation, Version v2024. AHRQ: Rockville, MD; [cited 2024 Dec 8]. Available from: https://qualityindicators.ahrq.gov/measures/pqi_resources

14.

Bartels

, Gill

, Naslund

. The affordable care act, accountable care organizations, and mental health care for older adults: Implications and opportunities. Harv Rev Psychiatry, 2015; 23(5):304–319; doi: 10.1097/HRP.0000000000000093

15.

Lewis

, Fraze

, Fisher

, et al. ACOs serving high proportions of racial and ethnic minorities lag in quality performance. Health Aff (Millwood), 2017; 36(1):57–66; doi: 10.1377/hlthaff.2016.0626

16.

Benchetrit

, Zimmerman

, Bao

, et al. Admission diagnoses among patients with heart failure: Variation by ACO performance on a measure of risk-standardized acute admission rates. Am Heart J, 2019; 207:19–26; doi: 10.1016/j.ahj.2018.09.007

17.

Randell

, Alvarado

, McVey

, et al. Requirements for a quality dashboard: Lessons from national clinical audits. AMIA Annu Symp Proc, 2019; 2019:735–744.

18.

Kansagara

, Englander

, Salanitro

, et al. Risk prediction models for hospital readmission: A systematic review. JAMA, 2011; 306(15):1688–1698; doi: 10.1001/jama.2011.1515

19.

Bleser

, Tchuisseu

, Shen

, et al. ACO REACH and advancing equity through value-based payment, Part 1. Health Aff Forefront, 2022; doi: 10.1377/forefront.20220513.630666/full/

20.

Bleser

, Pokam

, Shen

, et al. ACO REACH and advancing equity through value-based payment, Part 2. Health Aff Forefront, 2022; doi: 10.1377/forefront.20220517.755520/full/

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB

Development of the Health Disparities Prevention Quality Index for Medicare Accountable Care Organizations

Abstract

Background:

Methods:

Results:

Conclusion:

Introduction

Methods

Data sources

Health Disparities Prevention Quality Index

Inclusion and exclusion criteria

Measure calculation

Numerator calculation

Denominator calculation for the HDPQI

Calculation of the HDPQI

Analysis

Results

Discussion

Limitations

Conclusion

Footnotes

Authors’ Contributions

Author Disclosure Statement

Funding Information

Supplementary Material

Abbreviations Used

References

Supplementary Material