AUTHORS
Musa Yilmaz, MD, Alex Bataller, MD, PhD2, Mary-Alma Welch, MSc, PA-C3, Nicholas Short, MD4, Lianchun Xiao5, Jing Ning6, Mahesh Swaminathan, MD7, Alessandra Ferrajoli III, MD8, Tapan Kadia I, MD9, Sherry Pierce, BSN, RN10, John Villarreal11, Christine Lacke12, Katie Gaw13, Derek Cheung14, Kaycee Wilson15, Annie Toms16, Rachel Gershenson17, Marina George18, Kerin Adelson19, Christopher R. Flowers Jr, MD, MS20, Guillermo Garcia-Manero, MD21, William G. Wierda, MD, PhD22
AFFILIATIONS
1 UT MD Anderson Cancer Center, Houston, Texas, United States
2 UT MD Anderson Cancer Center, Houston, Texas, United States
3 UT MD Anderson Cancer Center, Houston, Texas, United States
4 UT MD Anderson Cancer Center, Houston, Texas, United States
5 UT MD Anderson Cancer Center, Houston, Texas, United States
6 UT MD Anderson Cancer Center, Houston, Texas, United States
7 UT MD Anderson Cancer Center, Houston, Texas, United States
8 UT MD Anderson Cancer Center, Houston, Texas, United States
9 UT MD Anderson Cancer Center, Houston, Texas, United States
10 UT MD Anderson Cancer Center, Houston, Texas, United States
11 UT MD Anderson Cancer Center, Houston, Texas, United States
12 UT MD Anderson Cancer Center, Houston, Texas, United States
13 UT MD Anderson Cancer Center, Houston, Texas, United States
14 UT MD Anderson Cancer Center, Houston, Texas, United States
15 UT MD Anderson Cancer Center, Houston, Texas, United States
16 UT MD Anderson Cancer Center, Houston, Texas, United States
17 UT MD Anderson Cancer Center, Houston, Texas, United States
18 UT MD Anderson Cancer Center, Houston, Texas, United States
19 UT MD Anderson Cancer Center, Houston, Texas, United States
20 UT MD Anderson Cancer Center, Houston, Texas, United States
21 UT MD Anderson Cancer Center, Houston, Texas, United States
22 UT MD Anderson Cancer Center, Houston, Texas, United States
ABSTRACT
Background
Unplanned hospital readmissions are increasingly recognized as an important healthcare quality metric and are associated with increased healthcare utilization, patient morbidity, and cost. Patients with leukemia are particularly vulnerable to readmission due to intensive treatment regimens, immunosuppression, infectious complications, and high disease acuity. We aimed to develop and internally validate a simple, clinically applicable risk score using routinely available clinical variables to predict unplanned 30-day hospital readmission of leukemia patients.
Methods
We retrospectively analyzed all consecutive admissions to a tertiary leukemia service over a 3-month period (June–August 2023). A total of 641 index admissions were identified and each followed for 30 days after discharge; 206 (32%) had subsequent unplanned readmission whereas 435 (68%) did not . Unplanned readmission was defined as any non-elective hospitalization within 30 days of discharge. Patient demographics, disease-related, treatment-related, and laboratory variables obtained on the first day of the index admission were compared between index admissions with and without a subsequent 30-day unplanned readmission. Univariable logistic regression was performed for each variable, and those with p<0.20 were entered into a multivariable logistic regression model to identify independent predictors of readmission. To develop and internally validate the clinical risk score, index admissions were randomly assigned in a 70:30 ratio to development and validation cohorts.
Results
The final multivariate model identified nine routinely available clinical variables: type of leukemia diagnosis, age <60 years, treatment status (frontline vs. salvage), and laboratory parameters using clinically relevant risk thresholds (hemoglobin <9 g/dL, platelet count ≤15×10⁹/L, white blood cell count ≤50×10⁹/L, absolute neutrophil count <1,000/µL, albumin <3.5 g/dL, and presence of circulating blasts). The greatest contributions to the score were the type of leukemia diagnosis (AML/ALL/MDS/MPN vs. CML/CLL/Others) and the presence of circulating blasts (6.6 points each), followed by receiving a salvage therapy (5.1 points), platelet count ≤15×10⁹/L (4.4 points), hemoglobin <9 g/dL and ANC <1,000/µL (3.8 points each), age <60 years (3.4 points), white blood cell count ≤50×10⁹/L (3.2 points), and albumin <3.5 g/dL (3.1 points).
These variables were combined into a weighted point-based clinical score, with the cumulative score used to stratify patients into four clinically meaningful risk groups: low (0-8.3), intermediate (8.4-18.3), high (18.4—29.7), and very high risk (>29.7). In the development cohort, unplanned 30-day readmission rates increased progressively from 9% in the low-risk group to 19%, 40%, and 70% in the intermediate-, high-, and very-high-risk groups, respectively. This gradient was preserved in the validation cohort, with corresponding readmission rates of 8%, 23%, 32%, and 53%. Model discrimination was acceptable, with an AUC of 0.73 in the development cohort and 0.61 in the validation cohort.
Conclusions
We developed and internally validated a simple, point-based clinical risk score that stratifies patients admitted to a leukemia service into low-, intermediate-, high-, and very-high-risk groups for unplanned 30-day readmission using routinely available clinical variables. Because the score relies exclusively on data routinely captured within the electronic health record, it can be automatically calculated early during hospitalization, allowing identification of patients at increased risk before discharge planning begins. Integration of this tool into electronic health record (EHR) systems may facilitate risk-adapted discharge planning, targeted transitional care interventions, closer outpatient follow-up, and more efficient allocation of healthcare resources. External validation and prospective evaluation of EHR-integrated implementation are warranted.
DISCLOSURES
Musa Yilmaz, MD: Curio Science - Honoraria; Daiichi Sankyo - Research Funding; Cardinal Health - Consultancy (Includes expert testimony); Daiichi Sankyo - Consultancy (Includes expert testimony)