Comparing Software Determination of Readmission Preventability With Chart Review, Provider, and Family Assessments

To explore the concordance between software, chart reviewer, provider, and parent perspectives when assessing whether readmissions are preventable or clinically related to the initial admission.

Providers and parents of patients readmitted within 3 days to a tertiary children's hospital were enrolled in this single-site observational study. 3M Potentially Preventable Readmissions Grouping Software, chart reviewers, discharge and readmission providers, and parents assessed if readmissions were clinically related to the index admission or potentially preventable. Agreement between perspectives was measured by using Cohen's κ values.

The software found 67 of 118 (57%) clinically related readmissions; the identical 67 of 118 cases (57%) were found to be potentially preventable. Chart reviewers found 107 of 125 (86%) clinically related and 60 of 125 (47%) preventable readmissions compared to 68 of 92 (74%) and 27 of 92 (28%) for discharge physicians and 69 of 93 (74%) and 33 of 93 (34%) for readmitting physicians. Parents reported 9 of 36 (25%) preventable readmissions. Cohen κ values revealed no to minimal agreement on clinical relatedness of readmissions between software and chart reviewer, discharge provider, and readmission provider (0.12-0.20), whereas chart reviewers and providers had weak to moderate agreement with each other (0.43-0.75). There was no to minimal agreement on preventability between software and the other perspectives (-0.04 to 0.21), whereas chart reviewers and providers had minimal to weak agreement (0.27-0.56).

Measurement of preventable readmissions remains problematic, and using financial penalties for readmissions on the basis of software determinations may be unwise given low levels of agreement. Chart review supplemented by information from providers and families offers a more inclusive way to identify potentially preventable readmissions.