To Treat or Not to Treat: Factors Associated with Receipt of Allogeneic Hematopoietic Cell Transplant

Background: Treatments for older adults with acute myeloid leukemia (AML) are often underutilized, with older adults often receiving no active treatment. Little is also currently known about patient, physician and hospital factors that may play a role in the treatment older adults receive. Data linkages can allow for a wider availability of variables to assess these factors. The objective of this study was to examine the association of patient, physician, and hospital factors with receipt of alloHCT in the first year of diagnosis, comparing patients who received an alloHCT to patients who did not receive any active treatment of alloHCT or chemotherapy (no treatment).

Methods: Administrative claims data from Centers for Medicare and Medicaid (CMS) Medicare from 2010 to 2018 were linked to publicly available CMS and National Provider Identifier (NPI) physician and hospital datasets. To be eligible for inclusion, patients needed to be diagnosed with AML between 2010 and 2017, enrolled in fee-for-service (Part A and B) Medicare for at least one year after diagnosis or until death, and be age 65-74. AlloHCT was identified using diagnosis and procedure codes; no treatment was identified by the absence of alloHCT or chemotherapy diagnosis and procedure codes. Multivariable hierarchical logistic regression was used to identify factors associated with treatment receipt, with hospital as a random effect. Fixed effects included patient (diagnosis year, sex, age, Elixhauser comorbidity index [ECI], race/ethnicity, United States census region), hospital (hospital ownership, rural, chemotherapy provision, medical school affiliation, hospital ranking) and physician (years since graduation, specialty, sex) characteristics identified at the time of diagnosis. Individuals with a racial/ethnic background of Black/African-American, Asian/Pacific Islander, Hispanic, American Indian/Alaska Native, and Other were grouped due to small numbers. Statistical Analysis Software (SAS^®) Enterprise Guide 7.1 was used for analysis.

Results: A total of 868 patients met eligibility criteria: 486 (56%) received alloHCT; 382 (44%) received no treatment. The majority of patients in both groups were Non-Hispanic White; more patients in the alloHCT group were age 65-69. Characteristics significantly associated with decreased odds of receiving alloHCT were: Elixhauser comorbidity index (≥3 compared to no comorbidities; odds ratio [OR]: 0.29; 95% confidence interval [CI]: 0.13-0.65), age (71-75 compared to 65-69 years: OR: 0.23, 95% CI: 0.16-0.34); medical school affiliation (no compared to yes: OR: 0.59; 95% CI: 0.40-0.87); and physician specialty (all other compared to hematologist-oncologist: OR: 0.60, 95% CI: 0.42-0.86). Living in an area with a higher population percentage with a college education was associated with an increased likelihood of receiving alloHCT (OR: 1.03, 95% CI: 1.02-1.04). Other factors were not statistically significant.

Conclusions: This analysis assessed the association of patient, hospital, and physician characteristics at the time of diagnosis with receipt of alloHCT in the first year after diagnosis, among those who received alloHCT and those who received no treatment. Notably, ECI, age, neighborhood education, hospital association with a medical school, and physician specialty at the time of diagnosis were associated with receipt of alloHCT. Comorbidities and age may be proxies for disease severity. Use of linked datasets provides the opportunity to study research questions that could not be evaluated with each dataset alone. CMS and NPI data are not collected for research, and thus, there is a lack of information on disease severity and cytogenetics, which may affect treatment. This study looked at patient, hospital, and physician factors at diagnosis. It is important to ensure patients have timely treatment. Study results will be used to help inform strategies to help improve access to treatment for patients diagnosed with AML.