Abstract 4019: Enhancing Administrative Data to Improve the Validity of Comparisons of Risk-Adjusted Inpatient Quality Indicators for Cardiovascular Care
The Agency for Healthcare Research and Quality’s Inpatient Quality Indicators (IQIs), which are based solely on administrative claims data (adm), are being used increasingly in published comparisons of the quality of cardiovascular (CV) care. Adm and clinical data (clin) abstracted from medical records from 188 Pennsylvania hospitals (hosp) from July 2001 through June 2004 were analyzed to assess the value of adding clin to adm to risk adjust IQIs for mortality of patients hospitalized for abdominal aortic aneurysm repair (n=5,309), myocardial infarction (n=104,110), coronary artery bypass graft surgery (n=58,879), congestive heart failure (n=200,506), and cerebrovascular accident (n=82,682). Risk factors were created using:
only adm with secondary diagnoses (sec) restricted to those that occurred only rarely as hospital-acquired complications,
All-Patient Refined Diagnosis Related Groups (APR-DRGs),
enhanced (enh) adm including sec documented as present on admission,
enh adm plus numerical laboratory results (lab),
enh adm plus lab and vital signs (VS), and
enh adm plus all available clin.
Risk-adjustment equations were compared using c-statistics, 5th to 95th percentile ranges of differences between predictions based on enh adm plus all clin and predictions based on less complete data, and the percentage of hospitals with more than 0.5 standard deviations (SD) of bias introduced by limiting data available for risk-adjustment. APR-DRGs had excellent predictive power but predictions correlated poorly with enh adm plus all clin because APR-DRGs improperly incorporated hospital-acquired complications into risk factors. Designating sec as present on admission and adding lab and VS to enh adm markedly improved predictive power and reduced the number of hospitals with unacceptable levels of systematic bias. Current IQIs are questionable indicators of the quality of CV care, but they can be improved greatly by modest additions of clin.