From: Improving evidence-based grouping of transitional care strategies in hospital implementation using statistical tools and expert review
Purpose of the Procedure
Provides an indication of how TC strategies naturally group (i.e., are implemented) together at hospitals.
Resultant groups overlap, with TC strategies able to be included in multiple groups. While overlap in groups is reflective of natural practice patterns, too much overlap among groups can reduce their contrast, rendering comparative effectiveness evaluation more difficult to interpret.
Latent Class Analysis
Resultant groups are mutually exclusive, with no overlap in group membership for TC strategies.
In addition, latent, or unmeasured characteristics are revealed among the resultant classes or groups
The resultant groups from LCA provides the strength of mutual exclusivity of TC strategy membership (i.e., strategies are only grouped into one class). However, this does not reflect natural practice patterns, in which some strategies may be ubiquitous and therefore “grouped” in multiple combinations.
Finite Mixture Model
This step models the probability of individual patients belonging to each unobserved group of TC strategies and classifies individuals into the groups.
In addition to grouping strategies based on observed traits, it draws inferences about how each group performs relative to an outcome (e.g., hospital readmissions). Thus, the total variance of each group’s contribution to readmission outcomes is also provided.
Our research team, with experience and expertise in TC practice, research, and hospital care delivery reviewed results of the above analyses. Based on the criteria below, they determined definitions for final TC strategy groups based on these criteria.
1. Maintain consistency with TC strategy groups from the study’s retrospective analysis (Appendix 3)
2. Follow signals suggested by the analytic procedures described above
3. Ensure the conceptual relevance of TC strategies grouped together
4. Reduce overlap in TC strategies across groups to enable comparative evaluation