External Validation Of The Skilled Nursing Facility Prognosis Score For Predicting Mortality, Hospital Readmission, And Community Discharge In Veterans

Abstract [from journal]

Background/objectives: Prognostic tools are needed to identify patients at high risk for adverse outcomes receiving post-acute care in skilled nursing facilities (SNFs) and provide high-value care. The SNF Prognosis Score was developed in a Medicare sample to predict a composite of long-term SNF stay, hospital readmission, or death during the SNF stay. Our goal was to evaluate the score's performance in an external validation cohort.

Design: Retrospective observational analysis.

Setting: We used a Veterans Administration (VA) Residential History File that concatenates VA, Medicare, and Medicaid claims to identify care trajectories across settings and payers for individual veterans.

Participants: Previously community-dwelling veterans receiving post-acute care in a SNF after hospitalization from January 1, 2012, to December 31, 2014. Both VA and non-VA hospitals and SNFs were included.

Measurements: We calculated the five-item SNF Prognosis Score for all eligible veterans in our sample and determined its discrimination (using a receiver operating characteristic curve) and calibration (plotting observed and expected events).

Results: The 386,483 veterans in our sample had worse physical function, more comorbidities, and were more likely to be treated for heart failure, but they had shorter index hospital lengths of stay and fewer catheters than the original Medicare cohort. The SNF Prognosis Score had similar discrimination (C-statistic = .70; .75 in the derivation cohort) and calibration at low to moderate levels of risk; at high levels, calibration was poorer with the score overestimating risks of adverse events.

Conclusion: The SNF Prognosis Score has reasonable discrimination and calibration, and it is simple to calculate using an admission SNF assessment and a nomogram. Future work embedding the score into practice is needed to determine real-world feasibility, acceptability, and effectiveness.