• Garrett W Esper, Ariana T Meltzer-Bruhn, Abhishek Ganta, Kenneth A Egol, and Sanjit R Konda.
• Division of Orthopedic Trauma Surgery, Department of Orthopedic Surgery, NYU Langone Health, NYU Langone Orthopedic Hospital, New York, NY, United States.
• Injury. 2023 Feb 1; 54 (2): 630635630-635.

IntroductionThe purpose of this study was twofold: 1. To assess how adaptive modeling, accounting for development of inpatient complications, affects the predictive capacity of the risk tool to predict inpatient mortality for a cohort of geriatric hip fracture patients. 2. To compare how risk triaging of secondary outcomes is affected by adaptive modeling. We hypothesize that adaptive modeling will improve the predictive capacity of the model and improve the ability to risk triage secondary outcomes.MethodsBetween October 2014-August 2021, 2421 patients >55 years old treated for hip fracture obtained through low-energy mechanisms were analyzed for demographics, injury details and hospital quality measures. The baseline Score for Trauma Triage in the Geriatric and Middle-Aged (STTGMA) tool for hip fractures (STTGMAHIP) was calculated in the emergency department setting. A new mortality risk score (STTGMAHIP_ADPTV) was created including inpatient complications. Each models' predictive ability was compared using DeLong's test. Patients were grouped into quartiles based on their respective STTGMAHIP_ADPTV and comparative analyses were conducted.ResultsAUROC comparison demonstrated STTGMAHIP_ADPTV significantly improved the predictive capacity for inpatient mortality compared to STTGMAHIP (p < 0.01). STTGMAHIP_ADPTV correctly triaged 80% and 64% of high-risk patients with inpatient and 30-day mortality compared to 64% and 57% for STTGMAHIP. STTGMAHIP_ADPTV quartile stratification demonstrated that the highest risk cohort had the worst mortality outcomes and hospital quality measures. Patients whose risk classification changed from minimal risk using STTGMAHIP to high risk using STTGMAHIP_ADPTV experienced the highest rate of mortality, readmission, ICU admission, with longer lengths of stay and higher hospital costs.DiscussionAdaptive modeling accounting for inpatient complications improves the predictive capacity and risk triaging of the STTGMAHIP tool. Real-time modulation of a patient's mortality risk profile can inform their requisite level of medical management to improve the quality and value of care as patients progress through their index hospitalization. STTGMAHIP_ADPTV can better identify patients at risk for developing complications whose mortality and readmission risk profile increase significantly, allowing their new risk classification to inform higher levels of care. While this may increase length of stay and total costs, it may improve outcomes in both the short and long-term.Level Of EvidenceIII.Copyright © 2022. Published by Elsevier Ltd.

13 keywords...

hide…