How to identify potentially preventable readmissions by classifying them using a national administrative database.

Area of Science:

• Healthcare Quality
• Health Services Research
• Medical Informatics

Background:

• Hospital readmissions are increasingly used as a quality of care indicator.
• Accurate identification of potentially preventable readmissions is challenging.
• Administrative data offers a scalable approach to classifying readmissions.

Purpose of the Study:

• To evaluate the accuracy of administrative data in identifying potentially preventable hospital readmissions.
• To compare classification of readmissions using administrative data versus medical record review.
• To assess the agreement, sensitivity, and specificity of administrative data for this purpose.

Main Methods:

• A random sample of 455 readmissions from a Dutch university hospital in 2014 were classified.
• Classifications were made using both administrative data and medical record review.
• Statistical analysis included Cohen's kappa for agreement and calculation of sensitivity and specificity.

Main Results:

• Medical record review identified 28.5% of readmissions as potentially preventable.
• Administrative data identified 44.1% as potentially preventable.
• Agreement between methods was slight (ƙ=0.08), with sensitivity at 63.1% and specificity at 63.5% for administrative data.

Conclusions:

• Significant differences exist between readmission classifications from administrative data and medical record review.
• Administrative data overestimates potentially preventable readmissions, limiting its use for hospital penalties.
• This classification may serve as a preliminary screening tool for identifying potentially preventable readmissions more efficiently.