Skip to main content

Table 10 Random Forest Analysis Results

From: Improving evidence-based grouping of transitional care strategies in hospital implementation using statistical tools and expert review

Covariate Number of Splits G2 Relative Importance
Comorbidity 30,784 114.293 0.152
Hierarchical Condition Category (HCC) Cohort (e.g. Medical, Surgical, etc.) 29,459 106.811 0.142
Helpful Healthcare Contact 23,132 54.112 0.072
Promote Trust at Home 20,174 41.156 0.055
Symptom Management 21,791 38.865 0.052
Plain Language Communication at Home 19,470 38.479 0.051
Promote Trust at Hospital 17,908 36.913 0.049
Plain Language at Hospital 14,670 35.181 0.047
Integrated health system affiliation 6611 24.507 0.033
Referral to Community Services 11,620 19.863 0.026
Post Discharge Care Consultation 11,162 18.050 0.024
Patient and Family Caregiver TC Needs Assessment 10,775 17.161 0.023
Follow-up Appointment 6720 15.894 0.021
Home Visit 8808 15.822 0.021
TC Group: Patient/Family Caregiver Assessment and Information Exchange 9882 15.776 0.021
No TC Group 9707 15.496 0.021
TC Group: Hospital-based Trust, Plain Language, and Coordination 8540 15.482 0.021
TC Group: Patient Communication and Care Management 10,122 15.446 0.021
Teach Back for Information and Skills 9164 15.239 0.020
Timely Exchange of Critical Patient Information among Providers 7947 14.422 0.019
TC Group: Home-based Trust, Plain Language and Coordination 9593 14.227 0.019
Patient Goal/Preference Assessment 8617 13.815 0.018
Identify High-Risk Patients and Intervene 7320 12.338 0.016
Language Assessment 6329 11.422 0.015
Medication Reconciliation 5783 9.977 0.013
TC Group: Assessment and Teach Back 4770 8.429 0.011
Transition Team 4750 7.687 0.010
Transition Summary for Patients and Family Caregivers 3713 5.653 0.008
  1. Notes:G2 = Likelihood Ratio Test Statistic, which is twice the [natural log] entropy, or twice the change in the entropy. Entropy is Σ -log(p) for each observation, where p is the probability attributed to the response that occurred. The G2 for a particular tree is the sum of the G2 values for each node that splits on that predictor. For a bootstrap forest model, the G2 values are added across the trees to give the G2 for that variable
  2. HCC cohorts include: Medical, Surgical, Neurology, Cardiorespiratory, Cardiovascular