Evaluation of hospital readmissions in surgical patients: do administrative data tell the real story?

IMPORTANCE: The Centers for Medicare & Medicaid Services has developed an all-cause readmission measure that uses administrative data to measure readmission rates and financially penalize hospitals with higher-than-expected readmission rates.
OBJECTIVES: To examine the accuracy of administrative codes in determining the cause of readmission as determined by medical record review, to evaluate the readmission measure's ability to accurately identify a readmission as planned, and to document the frequency of readmissions for reasons clinically unrelated to the original hospital stay. DESIGN, SETTING, AND PARTICIPANTS: Retrospective review of all consecutive patients discharged from general surgery services at a tertiary care, university-affiliated teaching hospital during 8 consecutive quarters (quarter 4 [October through December] of 2009 through quarter 3 [July through September] of 2011). Clinical readmission diagnosis determined from direct medical record review was compared with the administrative diagnosis recorded in a claims database. The number of planned hospital readmissions defined by the readmission measure was compared with the number identified using clinical data. Readmissions unrelated to the original hospital stay were identified using clinical data. MAIN OUTCOMES AND MEASURES: Discordance rate between administrative and clinical diagnoses for all hospital readmissions, discrepancy between planned readmissions defined by the readmission measure and identified by clinical medical record review, and fraction of hospital readmissions unrelated to the original hospital stay.
RESULTS: Of the 315 hospital readmissions, the readmission diagnosis listed in the administrative claims data differed from the clinical diagnosis in 97 readmissions (30.8%). The readmission measure identified 15 readmissions (4.8%) as planned, whereas clinical data identified 43 readmissions (13.7%) as planned. Unrelated readmissions comprised 70 of the 258 unplanned readmissions (27.1%). CONCLUSIONS AND RELEVANCE: Administrative billing data, as used by the readmission measure, do not reliably describe the reason for readmission. The readmission measure accounts for less than half of the planned readmissions and does not account for the nearly one-third of readmissions unrelated to the original hospital stay. Implementation of this readmission measure may result in unwarranted financial penalties for hospitals.