Development of a Predictive Hospitalization Model for Skilled Nursing Facility Patients
Bibliographic record
Abstract
OBJECTIVES: Identifying skilled nursing facility (SNF) patients at risk for hospitalization or death is of interest to SNFs, patients, and patients' families because of quality measures, financial penalties, and limited clinical staffing. We aimed to develop a predictive model that identifies SNF patients likely to be hospitalized or die within the next 7 days and validate the model's performance against clinician judgment. DESIGN: Retrospective multivariate prognostic model development study. SETTING AND PARTICIPANTS: Patients in US SNFs that use the PointClickCare electronic health record (EHR) system. We used data from the first 100 days of skilled stays for 5,642,474 patients in 8440 SNFs, from January 1, 2019, through March 31, 2023. METHODS: We used data collected in the course of clinical care to develop a machine learning model to predict the likelihood of patient hospitalization or death within the next 7 days. The data included vital signs, diagnoses, laboratory results, food intake, and clinical notes. We also asked SNF nurses and hospital case managers to make their own predictions as a comparison. The EHR was used as the source of information on whether the patient died or was hospitalized. RESULTS: The model had sensitivity of 35%, specificity of 92%, positive predictive value (PPV) of 18%, and area under the receiver operator curve (AUC) of 0.75. A variation of the model in which we did not include progress notes and food intake achieved an AUC of 0.70. Nurse raters achieved a sensitivity of 61%, specificity of 73%, and PPV of 10%. CONCLUSIONS AND IMPLICATIONS: Machine learning models can accurately predict the likelihood of hospitalization or death within the next 7 days among SNF patients. These models do not require additional SNF staff time and may be useful in readmission reduction programs by targeting more frequent monitoring proactively to those at highest risk.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".