2026 Critical Care Congress Skip to main content
2026 Critical Care Congress Main banner
Final Program


Advance Program
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Session Icons
Thought Leader Session
Thought Leader Session
Guidelines Session
Guidelines Session
Aging Adult Track
Aging Adult Track
Additional Registration Fee
Additional Registration Fee

Research Snapshot Theater: Quality and Patient Safety, Adult II

Predischarge Decision Trees Analysis Predicts 30-Day Readmission of CHF, COPD, Pneumonia Patients

Tuesday, March 24, 2026
11:15 AM - 12:15 PM Central Time
Location: Connections Central - RST 10

  • Samuel Houser, N/A (he/him/his)

    New Jersey

    Disclosure(s): No relevant financial relationship(s) to disclose.

  • AK

    Ayman Khatib, N/A

    Medical Student
    Rowan-Virtua School of Osteopathic Medicine

    Disclosure information not submitted.

  • NS

    Nicole Stouffer, n/a

    Biostatistician
    Inspira Health Network

    Disclosure information not submitted.

  • GS

    Gus J. Slotman, N/A

    Director of Graduate Medical Education Research, Clinical Professor of Surgery
    Inspira health Network

    Disclosure information not submitted.

Introduction: CMS penalizes hospitals for patient re-admissions within 30 days (RA30) after hospital discharge with a primary admission diagnosis of CHF, COPD or pneumonia. Identifying CHF/COPD/pneumonia inpatients at highest risk for RA30 is key to formulating interventions for reducing re-admission. However, diagnosis-specific RA30 prognostic scoring is not standardized. Hypothesis: RA30 can be predicted from pre-discharge (PD) data by decision trees (DTA).


Methods: From deidentified data during admission on 4696 CHF, 2294 COPD, 3350 pneumonia admissions, DTA models were constructed separately using recursive partitioning to classify all cases in each dataset. Trees were generated through automatic training, applying the default settings in SAS Enterprise Miner for splitting and node division. These defaults guided the selection of splitting criteria and tree depth, ensuring a standardized and reproducible modeling approach across each dataset. DPD=days since previous discharge.


Results: CHF: 23% (node n=4696) RA30. RA30 increased with < 34.5 DPD (36%)(node n=1210) + >4 admissions last 12 months (45%)(node n=453) + lowest 96hr PD anion gap >14.5 (68%)(node n=62).

COPD: 21% (node n=2294) RA30. RA30 increased >4 admissions/12 months (43%)(node n=311) + 96hr PD serum creatinine < 0.85 (59%)(node n=111); RA30 >37.5 DPD + 48hr PD %PMN >92.5 (57%)(node n=21).

Pneumonia: 15% (node n=3350) RA30. RA30 increased < 57 DPD (33%)(node n=657) + 96hr PD PMN >80% (45%)(node n=145) + 96hr PD creatinine >1.85 (61%)(node n=41). RA30 increased >57 DPD + 48hr PD PMN >89% (45%)(node n=62).


Conclusions: DTA models predicted RA30 risk up to 4× baseline after sentinel CHF, COPD, or pneumonia admissions, but in nodes comprising just 1 to 20% of patients in each diagnosis group. In CHF, lower DPD, increased prior 12 months admissions, and higher anion gap were linked to %RA30. In COPD, increased prior 12 months admissions, lower serum creatinine, higher DPD, and elevated %PMN predicted %RA30. In pneumonia, lower DPD, higher %PMN, and higher serum creatinine identified %RA30. These findings support DTA-based risk stratification and highlight diagnosis-specific predictors. Deeper DTA penetration in RA30 populations may yield clinically useful prognostic models.