Using AI, Socioeconomic Factors to Predict Outcomes After Allo-SCT for AML

AI and socioeconomic factors enhance risk stratification for patients with AML post transplant, aiming to reduce hospital readmissions and disease progression.

Image credit: Prostock-studio - stock.adobe.com

Targeting certain socioeconomic and clinical factors, as well as using artificial intelligence (AI) to improve risk stratification, can help identify patients with acute myeloid leukemia (AML) who may be readmitted to the hospital or experience disease progression after allogeneic stem cell transplantation (allo-SCT).

These findings from 2 abstracts were presented at the annual meeting of the American Society of Clinical Oncology.

How Socioeconomic and Mental Health Factors Into Readmissions

An analysis of a national database identified 15,757 patients with AML who were readmitted to the hospital after allo-SCT from 2016 to 2021.1 These patients were stratified into 3 age groups: younger than 45 years (n = 3743), 45 to 65 years (n = 8169), and older than 65 years (n = 3844).

The overall readmission rate within 30 days was comparable among the 3 groups, with 27% for the under-45 and 45-to-65 age groups and 30% for the 65-plus age group (P = .16). Mortality during readmission was also similar (5% for < 45 years vs 7% for 45-65 years and > 65 years; P = .42).

The most common reasons for readmission were infections (34%), gastrointestinal/hepatobiliary complications (10%), active AML (5%), and kidney dysfunction (5%). The mean length of stay was 29 days for patients older than 65 years and 32 days for patients younger than 45 years.

Factors linked to higher readmission risk were being in the lower-income quartile (vs being in the wealthiest quartile) and being on Medicare (vs being on private insurance). Other factors that elevated risk were depression, graft-vs-host disease (GVHD), chronic kidney disease, and acute respiratory failure. However, receiving home health care lowered the risk of readmission.

“Targeted interventions, such as optimizing post-discharge care and providing psychosocial support, may help reduce the readmission burden in high-risk patients,” the authors concluded.

AI Tool Pinpoints Risk of Disease Progression in AML

Advanced computational models can analyze complex patient data to provide more precise predictions of disease progression, paving the way for personalized posttransplant care in AML.2 Identifying the risk factors is important because disease progression is the leading cause of all-SCT failure in AML.

The analysis included 64 patients at a single center who underwent their first all-SCT for AML between January 2017 and June 2024. The median age was 55 years, and 23% of patients had adverse-risk disease per the European Leukemia Network 2022 (ELN2022) guidelines.

Nearly all (97%) were in complete remission at the time of transplant, and most had a human leukocyte antigen–matched unrelated donor. More than half (53.1%) received reduced-intensity conditioning, and 28% received posttransplant cyclophosphamide to prevent GVHD.

An Elastic Net machine learning algorithm was used to sift through clinical and genetic markers, and the analysis revealed several significant predictors of disease progression.

High-risk factors for disease progression were mixed chimerism at day 100, not receiving posttransplant cyclophosphamide, a Hematopoietic Cell Transplantation–specific Comorbidity Index score of 2 or greater, the presence of a secondary AML, DNMT3A or FLT3-ITD mutations, and an adverse ELN2022 risk.

Conversely, factors that were favorable for disease progression were having an NPM1 mutation and developing low-grade acute GVHD.

“AI models improve risk stratification for disease progression after allo-SCT by analyzing multiple variables simultaneously,” concluded the researchers. “This approach enables personalized post-transplant strategies to enhance outcomes.”

References

1. Sharma A, Singh V. Socioeconomic and clinical predictors of 30-day readmissions in AML patients undergoing allogeneic stem cell transplantation. Presented at: ASCO 2025; May 30-June 3, 2025; Chicago, IL. Abstract 6548.

2. Taj H, Adatorwovor R, Troper N, et al. AI-driven risk stratification for disease progression after allo-SCT in AML. Presented at: ASCO 2025; May 30-June 3, 2025; Chicago, IL. Abstract e18562.