MULTI-LEVEL PREDICTORS OF DISCHARGES AGAINST MEDICAL ADVICE- DECOMPOSING VARIATION USING AN ALL-PAYER DATABASE.
Author(s)
Nagarajan M1, Onukwugha E1, Offurum AI2, Gulati M2, Alfandre D3
1University of Maryland School of Pharmacy, Baltimore, MD, USA, 2University of Maryland Medical School of Medicine, Baltimore, MD, USA, 3New York University School of Medicine, New York, NY, USA
OBJECTIVES : METHODS : We used the National Inpatient Sample (NIS) 2014, an all-payer healthcare database that provides a stratified sample of 20% of all discharges from US hospitals. We included patients >18 years, in the general medical group, with known discharge status, and who were not transferred out or did not die in hospital. With our final sample of 2,687,430 discharges, we grouped variables from our data, and ran incremental mixed-effects logit models, with grouping at the level of the discharge, the hospital, and the census region. We obtained the intraclass correlation coefficients (ICC), and evaluated the percentage change in ICC. RESULTS : Our preliminary analysis showed associations with DAMA in line with previous studies: younger age, male gender, African-American race, residence in a large metropolis. Of interest, however, is our finding that of the overall variation in DAMA outcomes, 12.8% is associated with the hospital the discharge occurred from, and 1.2% of the variation with the census division the hospital is located in. This decreased with the addition of variables to the models, and the final, fully-adjusted model has 7.3% of variation in DAMA associated with the hospital-level, with the greatest percentage reductions occurring due to the addition of patient demographics. CONCLUSIONS : Our study is the first to explore the percentage in variation in DAMA due to patient, hospital and census-division characteristics. We find that even after adjusting for patient-level characteristics, there is a contribution of non-patient-level factors to DAMA outcomes.
Conference/Value in Health Info
2018-05, ISPOR 2018, Baltimore, MD, USA
Value in Health, Vol. 21, S1 (May 2018)
Code
PHP75
Topic
Epidemiology & Public Health, Health Policy & Regulatory, Health Service Delivery & Process of Care
Topic Subcategory
Health Disparities & Equity, Public Health, Quality of Care Measurement
Disease
Multiple Diseases