Construction and evaluation of a prognostic risk prediction model for community-acquired pneumonia in older adults based on Comprehensive Geriatric Assessment
SHEN Xiaoqian
FU Yimin
WANG Xiaojun
Abstract：Objective To investigate the predictive value of models constructed based on the Comprehensive Geriat‐ric Assessment(CGA)for short-term prognosis in older adults with community-acquired pneumonia(CAP).Methods A prospective cohort study was conducted,enrolling 100 older adults with CAP.CGA indicators were collected,and follow-up was performed on the length of hospital stay,unplanned outpatient/emergency visits,and unplanned readmissions with‐in 3 months after discharge.Using univariable analyses(Spearman rank correlation,Mann-Whitney U test,and univari‐able binary logistic regression)for screening and multivariable analyses(multiple linear regression and multivariable bina‐ry logistic regression)for modeling,prediction models were constructed and evaluated via internal validation for discrimi‐nation,calibration,and robustness.Results The prediction model for length of hospital stay incorporated comorbidi‐ties,frailty,and Activities of Daily Living(ADL),explaining 50.7% of the variance.The risk prediction model for un‐planned emergency/outpatient visits within 3 months after discharge included frailty and depression score(AUC=0.78,95%CI:0.68-0.88).The risk prediction model for unplanned readmission within 3 months after discharge included de‐pression score,ADL,and number of medications(AUC=0.839,95%CI:0.662-1.000).All models showed stable perfor‐mance in internal validation.Conclusion The CGA-based prediction models can effectively identify the risk of short-term adverse outcomes in older adults with CAP,providing a reference for clinical risk assessment and intervention.
Keywords：comprehensive geriatric assessmentcommunity-acquired pneumoniaagedprediction modellength of stayreadmissionprognosis
Publication Date：2026-02-25
Online Publishing Date：2026-03-19(First online date of this platform, not the publication date of the document)
Pages：5( 1-5 )