Optimizing heart failure care: A machine learning-based prediction of hospital length of stay for heart failure patients.
Arthur De Souza, Ray Opoku
PMID 42176646WHAT IT FOUND
A model predicted short, medium, and long stays for heart failure patients, but it missed many long stays.
It is not ready to guide discharge planning.
Key findings
01LightGBM, trained on 2,008 heart failure records, was the best of 13 models, with a micro-average ROC AUC of 0.78 and a macro-average ROC AUC of 0.68.
02When the model predicted a long stay it was usually correct, with precision of 0.67, but it missed many true long stays, with recall of 0.148.
03The top variables were brain natriuretic peptide, creatine kinase, and high density lipoprotein cholesterol.
STILL TO COME
How it was doneWhat they foundWhat it means for RNs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The records came from one hospital in China, so the model may not work for other populations. The study did not test the model in real clinical care, so it does not show that using it changes length of stay, discharge, or patient outcomes. The long-stay group was small and imbalanced, and the model missed many true long stays. The model performed less well when each stay category was considered equally, and it had difficulty separating short and medium stays. Patients and the public were not involved in the design, conduct, reporting or dissemination plans.
Declared interests
The supplied article text does not give a conflicts or funding declaration; the publication types list NIH extramural research support.
The easy way to misread this
Do not treat this as a validated tool for discharge or staffing decisions. It was built on retrospective records from one hospital and missed many long stays, so a predicted short stay could be wrong.