With the new Medicare penalties for excessive readmission rates, hospitals are more attuned to identifying and preventing avoidable readmissions. One approach is a manual review of discharged patients, with more intensive follow-up on the likely readmission candidates. Some vendors also sell software that purports to identify and rank potential readmissions. The Kaiser system studied over 450 thirty-day all-cause readmissions for potential preventability. All the cases were reviewed both manually and through the use of 3M software commercially available. (Readmission Study) Past research has suggested that a little over 20% of readmissions are avoidable. A few studies on automated classification methods found that these over-identified potential readmissions. The Kaiser group conducted a manual review that involved chart review, physician interviews and patient and caregiver interviews, with a final judgment being made by a nurse reviewer and physician. The software uses algorithms based on diagnoses and severity information, and other data, to make a judgment about whether readmission could have been prevented.
The software identified 78% of readmissions as preventable, while manual review found 49% to be so classified. The methods agreed on 56% of the cases. The software agreed with 85% of the manual findings of preventability, but only with 28% of findings of non-preventability. This means the software would suggest intervention in far more cases than manual review, in fact it appears to think almost all readmissions are avoidable. The researchers concluded that the software would not be an adequate replacement for manual review. The software vendor suggests that it be used as a tool to focus manual reviews, but that would add to expense and the software doesn’t exclude many readmissions from manual review. While readmissions may be avoidable and costly to the system, taking steps to reduce them also is expensive. The less targeted the effort is, the more it costs, and realistically, the less incentive hospitals have to make the effort. Ideally, software would automatically scan medical records and other data to identify readmission candidates and allow focused interventions. Software should be a less expensive method for this step than human review. But this study suggests that at least one piece of commonly used software doesn’t do a particularly job of guiding readmission prevention programs.