Pine, Michael

Numerical laboratory data at admission have been proposed for enhancement of inpatient predictive modeling from administrative claims. In this study, predictive models for inpatient/30-day postdischarge mortality and for risk-adjusted prolonged length of stay, as a surrogate for severe inpatient complications of care, were designed with administrative data only and with administrative data plus numerical laboratory variables. A comparison of resulting inpatient models for acute myocardial infarction, congestive heart failure, coronary artery bypass grafting, and percutaneous cardiac interventions demonstrated improved discrimination and calibration with administrative data plus laboratory values compared to administrative data only for both mortality and prolonged length of stay. Improved goodness of fit was most apparent in acute myocardial infarction and percutaneous cardiac intervention. The emergence of electronic medical records should make the addition of laboratory variables to administrative data an efficient and practical method to clinically enhance predictive modeling of inpatient outcomes of care.

Predictive modeling for postdischarge outcomes of inpatient care has been suboptimal. This study evaluated whether admission numerical laboratory data added to administrative models from New York and Minnesota hospitals would enhance the prediction accuracy for 90-day postdischarge deaths without readmission (PD-90) and 90-day readmissions (RA-90) following inpatient care for cardiac patients. Risk-adjustment models for the prediction of PD-90 and RA-90 were designed for acute myocardial infarction, percutaneous cardiac intervention, coronary artery bypass grafting, and congestive heart failure. Models were derived from hospital claims data and were then enhanced with admission laboratory predictive results. Case-level discrimination, goodness of fit, and calibration were used to compare administrative models (ADM) and laboratory predictive models (LAB). LAB models for the prediction of PD-90 were modestly enhanced over ADM, but negligible benefit was seen for RA-90. A consistent predictor of PD-90 and RA-90 was prolonged length of stay outliers from the index hospitalization.

ObjectiveAssess algorithms for linking patients across de‐identified databases without compromising confidentiality.Data Sources/Study SettingHospital discharges from 11 Mayo Clinic hospitals during January 2008–September 2012 (assessment and validation data). Minnesota death certificates and hospital discharges from 2009 to 2012 for entire state (application data).Study DesignCross‐sectional assessment of sensitivity and positive predictive value (PPV) for four linking algorithms tested by identifying readmissions and posthospital mortality on the assessment data with application to statewide data.Data Collection/Extraction MethodsDe‐identified claims included patient gender, birthdate, and zip code. Assessment records were matched with institutional sources containing unique identifiers and the last four digits of Social Security number (SSNL4).Principal FindingsGender, birthdate, and five‐digit zip code identified readmissions with a sensitivity of 98.0 percent and a PPV of 97.7 percent and identified postdischarge mortality with 84.4 percent sensitivity and 98.9 percent PPV. Inclusion of SSNL4 produced nearly perfect identification of readmissions and deaths. When applied statewide, regions bordering states with unavailable hospital discharge data had lower rates.ConclusionAddition of SSNL4 to administrative data, accompanied by appropriate data use and data release policies, can enable trusted repositories to link data with nearly perfect accuracy without compromising patient confidentiality. States maintaining centralized de‐identified databases should add SSNL4 to data specifications.