JMIR Publications

ABSTRACT

Background:

An assessment tool is needed to measure clinical severity of nursing home residents to improve prediction of outcomes and provide guidance in treatment planning.

Objective:

To describe development of a Nursing Home Severity Index, a clinical severity measure targeted for nursing home residents with the potential to be individually tailored to different outcomes, such as pressure injury.

Methods:

A retrospective, non-experimental design was used to develop and validate a Nursing Home Severity Index using secondary data from nine nursing homes participating in the 12-month pre-intervention period of the TEAM-UP pragmatic clinical trial. Expert opinion and clinical literature were used to identify indicators which were grouped into severity dimensions. Index performance and validation to predict risk of pressure injury was accomplished using secondary data from nursing home electronic health records, Minimum Data Sets, and Risk Management Systems. Logistic regression models including a resident’s Worst-Braden score with/without severity dimensions generated propensity scores. Goodness-of-fit for overall models was assessed using C statistics; significance of improvement of fit after adding severity components to the model was determined using the Likelihood Ratio Chi Square test. Significance of each component was assessed with odds ratios. Validation based on randomly selected 65% training and 35% validation data sets was used to confirm reliability of the severity measure. Finally, discriminating ability of models was evaluated using propensity stratification to evaluate which model best discriminated between residents with/without pressure injury.

Results:

Data from 1015 residents without pressure injuries on admission were used for the Nursing Home Severity Index-Pressure Injury and included laboratory, weights/vitals/pain, underweight, and locomotion severity dimensions. Logistic regression C-statistics measuring predictive accuracy increased 19.3% (from 0.627 to 0.748, P<.001) when adding four severity dimensions to Worst-Braden scores. Significantly higher odds of developing pressure injury were associated with increasing dimension scores. Use of highest quintile propensity scores predicting greatest risk of pressure injury improved predictive accuracy by detecting 21 more residents who developed pressure injury (n=44, 49.8% vs n=23, 25.8%) when both severity dimensions and Worst-Braden score were included in prediction modeling.

Conclusions:

A clinical Nursing Home Severity Index-Pressure Injury was successfully developed and tested using the outcome of pressure injury. Overall predictive capacity was enhanced when using severity dimensions in combination with Worst-Braden scores. This Index has the potential to significantly impact quality of care decisions aimed at improving individual pressure injury prevention plans. Clinical Trial: ClinicalTrials.gov NCT123456.