Critical care medicine use and cost among Medicare beneficiaries 1995-2000: Major discrepancies between two United States federal Medicare databases

Objective. A comparison of federal Medicare databases to identify critical care medicine (CCM) use, cost discrepancies, and their possible causes. Design: A 6-yr (1995-2000) retrospective analysis of Medicare hospital and CCM use and cost, comparing the Hospital Cost Report Information System (HCRIS) with Medicare Provider Analysis and Review File (MedPAR) supplemented when necessary by Health Care Information System (HCIS) (identified herein as Med-PAR/HCIS). Setting. All nonfederal U.S. hospitals. Subjects: None. Interventions. None. Measurements and Main Results., Data are presented as days (M = million) and costs ($; B = Billion) for both hospitals and CCM. Between 1995 and 2000, the number of hospital days decreased in both databases: HCRIS (- 13.2%; 78M to 67.7M) and MedPAR/HCIS (- 14.1 %; 82.8M to 71.1 M). CCM days decreased in HCRIS (-4.6%; 8.3M to 7.9M). In contrast, CCM days increased in MedPAR/HCIS (7.2%; 13.9M to 14.9M). The discrepancy in CCM days between HCRIS and MedPAR/HCIS increased from 40% (5.6M days) in 1995 to 47% (7M days) in 2000. Two CCM billing codes (intensive care unit and coronary care unit post/intermediate) used in MedPAR/HCIS were responsible for 73% on average per year, over the study period, for this CCM discrepancy. The use of these two codes progressively increased (44%; 3.9M to 5.6M days) by the end of the study. The cumulative 6-yr discrepancy in CCM days between HCRIS and MedPAR/HCIS (37.3M days) had a calculated cost of $92.3B. Conclusions. We have identified major, and progressively increasing, discrepancies between two U.S. federal databases tabulating hospital and CCM use and cost for Medicare beneficiaries. Two CCM post/intermediate billing codes in MedPAR/HCIS were predominantly responsible for the CCM discrepancy. To accurately assess Medicare CCM use and cost, either HCRIS, or Med-PAR/HCIS without the post/intermediate codes, should be used.