OBJECTIVES

This study compares healthcare resource utilization between cardiac surgical patient clusters determined by unsupervised machine learning using 2017 National Inpatient Sample (NIS) discharges with a primary, major therapeutic cardiac procedure.

METHODS

The 2017 NIS database from the Healthcare Cost and Utilization Project (HCUP) describes U.S. patient discharges with weights to generate national discharge and charge estimates. We examined patients ≥18 years of age, with a primary procedure listed as “Major Therapeutic” per HCUP Procedure Classes for ICD-10-PCS, with a complete discharge record, and with a primary cardiac procedure based on the Clinical Classification Software for ICD-10-PCS. Clustering was performed on patient and hospital characteristics, independent of outcomes, through K-Prototypes from the R package clustMixType.

RESULTS

There were 170,326 discharges identified after applying inclusion criteria. K-prototypes yielded 3 clusters exhibiting differences in average age (Cluster 1: 68.3; Cluster 2: 69.8; Cluster 3: 61.9), average Elixhauser comorbidity index (Cluster 1: 9.8; 2: 2.9; 3: 0.8), percentage of discharges with non-elective admissions (Cluster 1: 88.1%; 2: 15.6%; 3: 95.4%), and most common procedures (Cluster 1: percutaneous transluminal coronary angioplasty [PCTA] – 31.3% and coronary artery bypass graft [CABG] – 22.6%; 2: heart valve procedures – 30.1% and CABG – 20.5%; 3: PTCA - 73.7% and CABG – 8.6%). Estimated U.S. national total cardiac surgery discharges and associated hospital charges were extrapolated for each cluster using weights (Cluster 1: 17.8% discharges, 31.2% charges; 2: 33.4% discharges, 29.6% charges; 3: 34.4% discharges, 24.6% charges).

CONCLUSIONS

Cluster analysis of U.S. cardiac surgical discharges revealed that a cluster of patients with the greatest comorbidities comprised only 17.8% of cardiac surgery discharges, but made up 31.2% of cardiac surgery charges. The most common primary surgeries in Cluster 1, PCTA and CABG, totaled only 53.9% of procedures. Disproportionate charges for these more severely-ill patients warrant further analysis of their care and resource utilization.