Risk Factors for Suboptimal Treatment in Patients with Newly Diagnosed Multiple Myeloma across US Healthcare Systems

Introduction Multiple myeloma (MM) is a plasma cell malignancy disproportionately affecting underrepresented populations. Despite advances in MM treatment, racial and ethnic minorities have poorer outcomes likely due to disparities in treatment initiation and access to autologous stem cell transplantation and novel therapies. Literature suggests that social determinants of health (SDoH), in part, contribute to treatment disparities. Thus, MM risk models incorporating the impact of SDoH factors are necessary to develop evidence-based guidelines and targeted interventions to combat treatment disparities. For this reason, we sought to determine demographic, SDoH, and disease-related risk factors for suboptimal MM care in a racially, ethnically, and geographically diverse cohort of MM patients from three US health care systems. Methods We retrospectively evaluated the electronic health records of treatment-naive MM patients from the Veterans Affairs (VA; 1999 - 2022), MedStar Health (2004 - 2023), and Boston Medical Center (2007 - 2023) representing three healthcare settings with different infrastructures suited to provide healthcare regardless of ability to pay. Demographics and data regarding distance from home to the hospital, area deprivation index, and income levels (VA only); MM disease characteristics including international staging system (ISS) stage and cytogenetics; and overall health status via the Charlson Comorbidity index. Descriptive statistics were used to examine the distribution of categorical and continuous variables. We evaluated two outcomes representing optimal treatment: 1) time to initiation of induction therapy, and 2) type of first-line MM therapy, specifically triplet/ quadruplet therapy, or consolidation therapy with high dose chemotherapy, and autologous stem cell transplant. Logistic regression models were calculated to determine demographic, SDoH, and MM-related factors associated with each of the outcomes. Results At the time of this report, we analyzed data from 8,458 patients from the VA; 805 patients from MedStar Health; and 171 patients from Boston Medical Center. The median age was 69, 66, and 65 years (interquartile range [IQR] 63 - 76; [IQR] 58 - 73; [IQR] 57 - 73) for the VA, MedStar, and BMC cohorts, respectively. More than 1/3 of patients in each cohort were non-White or Hispanic 35% (n = 2942), 66% (n = 530), and 60% (n =102) in the VA, MedStar, and BMC cohorts, respectively. There were no significant differences in disease characteristics across the cohorts. At the VA, a national healthcare setting with no individual or insurance payor, patients living a further distance from the nearest VA hospital were significantly less likely to receive treatment within 14 days (< 25 kilometers vs. > 100 kilometers; odds ratio [OR] 0.72, 0.52 - 0.99) while patients having a higher income level were significantly more likely to receive triplet/ quadruplet therapy or an autologous stem cell transplant (> $100k vs. < 25k per year; [OR] 1.7, 1.3 - 2.23). At MedStar Health (a healthcare system with numerous hospitals within the metropolitan area) and Boston Medical Center (the largest safety net hospital in New England), uninsured patients were significantly less likely to receive triplet/ quadruplet therapy or an autologous stem cell transplant ([OR] 0.23, 0.06 - 0.79 and 0.12, 0.01 - 0.8, respectively). There were no significant differences in receipt of optimal treatment by race or ethnicity across the three cohorts. Conclusion Our findings suggest that SDoH can present significant barriers to optimal care across US healthcare systems. Specifically, living a further distance from the hospital may be associated with later treatment initiation while lack of insurance and a lower income level may limit access to novel therapies or autologous stem cell transplant. This informs next steps to develop a prediction algorithm incorporating these SDoH factors that identifies patients at risk for suboptimal treatment. Ultimately, this model will prompt automatic referral to guideline driven supportive care resources, reducing barriers to optimal MM care.