Comparing Access to Allogeneic Hematopoietic Cell Transplant across the United States: Physician Density, Minority Health Social Vulnerability Index, Transplant Unmet Need and Utilization

Background: Access to allogeneic hematopoietic cell transplantation (alloHCT) is limited by many factors, with only 1 in 3 patients who may need an alloHCT receiving one. Physician density (PD) has been associated with access to care but is largely unexplored in alloHCT. This study aimed to examine associations among county-level PD, Minority Health Social Vulnerability Index (MHSVI), and alloHCT unmet need/utilization across the U.S.

Methods: A retrospective analysis aggregated data across all 3142 U.S. counties, including: MHSVI (Centers for Disease Control and Prevention); Physician Compare Data (Centers for Medicare and Medicaid); and 2019-23 alloHCT unmet need and utilization data (NMDP/CIBMTR). MHSVI is a multivariable index (0-1) across 6 themes (socioeconomic status, SES; household characteristics, HH; race/ethnicity, R/E; housing/transportation, H/T; healthcare access and infrastructure, HCA; medical vulnerability, MV), where higher values indicate greater vulnerability. PD was defined as hematology-oncology (hem-onc) and HCT physicians per 10,000 population. Unmet need is based on estimated need (Surveillance, Epidemiology, and End Results data, population characteristics, and current alloHCT eligibility criteria) and actual alloHCT utilization per 100,000 population. Variables were collapsed into low (≤50%) and high (>50%) percentile categories for interpretability. National data were compared with a subset of 7 states previously found to be at risk of access challenges due to high unmet and vulnerability and low PD: Texas, Louisiana, Georgia, Florida, North Carolina, South Carolina, and Nevada. Univariable logistic regression analysis and chi-square was used to determine initial associations and variables to include in multivariable (MVA) logistic regression models. Total MHSVI and the 6 themes were tested separately. HCA and PD were also tested separately due to collinearity.

Results: At-risk states had higher MHSVI across all themes, higher unmet need, lower alloHCT utilization, and a greater proportion of counties with no hem-onc or HCT physicians (low PD) compared to national data. National univariable analysis found all MHSVI themes to be significantly associated with alloHCT utilization and unmet need; all but MV were associated with PD. In at-risk states, all MHSVI themes were associated with utilization, unmet need, and PD, including MV. In national MVA models, SES (odds ratio, OR=0.59, p<0.001) and MV (OR=0.72, p<0.001) were associated with lower odds of living in an area with high alloHCT utilization, while HH (OR=1.29, p=0.002) and R/E (OR=1.29, p=0.003) were associated with greater odds of high alloHCT utilization; when controlling for PD (OR=1.78, p<0.001), only SES and MV remained significant. In at-risk states, HH (OR=1.42, p=0.045) and R/E (OR=1.46, p=0.031) were associated with higher odds of living in an area with high alloHCT utilization, while H/T (OR=0.59, p=0.002) and HCA (OR=0.75, p=0.076) were associated with lower odds of high alloHCT utilization; when controlling for PD (OR=1.79, p<0.001), only HH (OR=1.43, p=0.039) and H/T (OR=0.60, p=0.003) remained significant. Similar relationships were true for models of unmet need, except that SES [OR=1.60, p<0.001) was highly associated with higher unmet need at the national level. PD was significantly associated with higher unmet need for both national data (OR=0.52, p<0.001) and in at-risk states (OR=0.50, p<0.001).

Conclusions: Meaningful associations seem to exist among MHSVI, PD, and alloHCT utilization/unmet need. Specifically, study data suggest that PD may be critical for alloHCT access and utilization both nationally and in at-risk states with higher MHSVI and unmet need. Therefore, targeted initiatives and resources for counties with low PD and high MHSVI could enable more at-risk patients to receive alloHCT. R/E and HH variables both capture characteristics of a diverse patient population, so positive associations with areas of increased alloHCT utilization could be skewed by diverse metropolitan areas. This suggests specialized strategies to increase access may need to vary in metropolitan vs. rural areas. Data may not directly relate to individual access, but these variables merit further exploration to inform access-related initiatives and resources at a state and national level.

No relevant conflicts of interest to declare.