Introduction: Acute myeloid leukemia (AML) is an aggressive hematologic malignancy with poor prognosis, particularly in older adults. As evidence grows linking social determinants of health (SDOH) to cancer outcomes, neighborhood-level socioeconomic measures have gained importance in oncology research. Two widely used indices, Social Vulnerability Index (SVI) and Area Deprivation Index (ADI), aim to quantify area-level disadvantage. SVI, developed by the Centers for Disease Control and Prevention (CDC), includes 15 variables across four domains: socioeconomic status, household composition, minority status/language, and housing/transportation. It is calculated at the county level, with higher SVI indicating greater social vulnerability. ADI, developed by the University of Wisconsin, comprises 17 census-based indicators focused on income, education, employment, and housing quality. It is measured at the census block group level, with higher values indicating greater deprivation. This study evaluated and compared SVI and ADI in predicting survival, treatment receipt, and end-of-life care among older AML patients.

Methods: We used SEER-Medicare claims data (2009–2020) to identify patients ≥66 years old diagnosed with AML or acute monocytic leukemia from 2010–2019. Eligible patients had continuous Medicare A/B enrollment for 12 months before and Medicare A/B/D 3 months after diagnosis. SVI and ADI were linked to patient ZIP code-based geographies and dichotomized at the median. Kaplan-Meier and Cox proportional hazards models were used to assess survival. Chemotherapy, bone marrow transplant (BMT), ICU admissions, and hospice use were analyzed during a 3- and 12-month period post-diagnosis and near end-of-life.

Results: The cohort included 11,547 patients (mean age 77.5 years; 47.2% female). Those residing in high SVI or high ADI areas were more likely to have lower income, less education, and more comorbidities. In unadjusted analysis, high ADI was associated with significantly worse overall survival based on the log-rank test (p<0.001), with median survival of 2 months in the high ADI group vs. 4 months in the low ADI group. In contrast, survival did not differ significantly by SVI (log-rank p=0.16), with a median survival of 3 months in both groups. In adjusted Cox models, high ADI remained a significant predictor of worse survival (HR 1.09; 95% CI 1.04–1.14), while SVI was not (HR 0.97; 95% CI 0.94–1.01).

Among the 6,152 patients who survived ≥3 months, chemotherapy use was lower in high ADI (48.3% vs 51.5%; p=0.012) and high SVI (48.2% vs 51.7%; p=0.007) groups. In the 12-month post-diagnosis cohort (n=2,927), BMT was less common among those in high ADI (11.5% vs 15.2%; p=0.003) and high SVI (12.0% vs 15.1%; p=0.015) areas. ICU use was higher in high ADI patients (mean 0.28 vs 0.23; p=0.049), while high SVI was associated with longer time to BMT (p=0.016). High ADI was also linked to greater hospice use early in the disease course.

Among 5,189 decedents, high ADI was associated with increased hospice use in the final 3 months of life (p<0.001). In a subset with ≥12 months of pre-death enrollment (n=1,970), high SVI was linked to more hospitalizations (mean 4.1 vs 3.6; p<0.001), and high ADI was associated with higher hospice initiation (p <0.001).

Conclusions: In this national cohort of older adults with AML, ADI was a robust and independent predictor of survival, treatment disparities, and end-of-life care, whereas SVI was not. ADI's stronger performance likely reflects its focus on core socioeconomic factors and finer geographic granularity. In contrast, SVI's broader domains and county-level resolution may obscure within-county disparities. Recent work (Rollings et al. 2023; Staloff et al. 2025) similarly found low concordance between SVI and ADI and support ADI's superiority in capturing cancer-relevant structural inequities. Our findings reinforce the utility of ADI in future health equity research and interventions. Future studies should examine ADI's relevance in the era of venetoclax and other targeted therapies approved after 2020, which were not captured in this cohort.

This content is only available as a PDF.
2025
Sign in via your Institution