In the period 1994–2009 average real risk-unadjusted health expenditures during the year following a heart attack, congestive heart failure, or hip fracture increased. For heart attack, expenditures increased 27 percent, from $39,300 to $49,900. For congestive heart failure, the increase was 43 percent, from $33,700 to $48,200; and for hip fracture, it was 39 percent, from $41,000 to $56,800. The risk-adjusted increases were $10,500, $12,600, and $14,400, respectively (–).
Medicare Cost Per Heart Attack Episode At 30 And 365 Days, By Type Of Care, 1994–2010
Per Episode Medicare Cost For Heart Attack, Congestive Heart Failure, And Hip Fracture At 365 Days, 1994 And 2009
Although spending grew for all major categories of care, it grew fastest for postacute services. As shows, average risk-adjusted spending on acute hospital care in the year following the index hospitalization was the category of spending that grew the slowest between 1994 and 2009 for all three conditions. In contrast, average spending for postacute care doubled for hip fracture patients, more than doubled for congestive heart failure patients, and more than tripled for heart attack patients during the same period.
For care between 30 and 365 days after an initial hospitalization, postacute spending growth was the main driver of increases in spending for the study period (results not shown). During this period postacute spending comprised 47 percent, 39 percent, and73 percent of the growth in expenditures for heart attack, congestive heart failure, and hip fracture, respectively.
In contrast, acute care contributed only 11 percent, 28 percent, and 7 percent of cost growth for these conditions during this period. See Appendix Exhibit 2 for cost growth during three periods lasting up to a year following an index event.13
To explore whether these increases in spending were associated with improvements in outcomes, we examined changes in mortality. Consistent with other studies on heart attack19
and congestive heart failure,20
we found that risk-adjusted mortality fell between 1993 and 2009. For patients with heart attacks during the study period, the thirty-day risk-adjusted mortality rate declined from 19.1 percent to 10.2 percent—a 47 percent reduction—and for patients with congestive heart failure, the rate fell from 10.6 percent to 7.4 percent—a31 percent reduction (see Appendix Exhibits 3–5).13
For patients with hip fractures, the thirty-day risk-adjusted mortality rate rose from 6.3 percent to 6.4 percent—a 1 percent increase during the study period. Mortality rates changed in two phases: Between 1994 and 2007 mortality gradually rose to a peak of 7.7 percent in 2002 before falling again. Our finding that mortality rates for hip fracture patients were subject to only small changes was consistent with a study of hip fracture mortality in the Medicare population, which showed that mortality decreased from 1986 through 1995 but stayed fairly constant after 1995.12
For patients with heart attacks and those with congestive heart failure, short-term survival gains persisted through one year after the index event. For patients with heart attacks, risk-adjusted 365-day mortality fell to 23.8 percent from 33.0 percent. The vast majority of this decline occurred in the first 30 days, with only small changes in the mortality rates between 30 and 365 days. For patients with congestive heart failure, risk-adjusted 365-day mortality fell to 31.7 percent from 36.9 percent, and risk-adjusted mortality between 30 and 365 days also fell between 1994 and 2009. Improvements in short-term survival were thus retained through the first year after the index event for patients with heart attacks and were improved on for patients with congestive heart failure.
For patients with hip fractures trends in 365-day mortality and 30-day mortality followed similar patterns. Risk-adjusted 365-day mortality for these patients rose from 23.2 percent in 1994 to a peak of 27.4 percent in 2002 and then declined to 23.8 percent in 2009. The peak in mortality rates during the interim years of the sample is partly a result of increased mortality between 30 and 365 days for patients with hip fractures during this period.
One concern with risk adjustment is that there is a general trend toward more thorough documentation of comorbidities on Medicare claims forms over time, which may make the study population appear to be getting sicker although it is not.21
To ensure that our risk-adjusted trends were not being driven by trends in coding practice, we calculated the same mortality trends while risk-adjusting only for age, race, and sex, and interaction terms between these variables.
During the study period, 365-day mortality rates adjusted only for age, race, and sex declined to 28.0 percent from 33.3 percent for patients with heart attacks—a 16 percent reduction. For patients with congestive heart failure the mortality rate declined to 36.4 percent from 37.4 percent—a 2.7 percent reduction. And for patients with hip fractures the mortality rate rose to 26.6 percent from 23.6 percent—a 12.7 percent increase.
Excluding comorbidities in the risk-adjustment formulas thus resulted in a much smaller drop in one-year risk-adjusted mortality, which implies that our original results were partially driven by increases in the reported disease burden. This may indicate that risk-adjustment was biased by increases in the coding of disease. However, it could also indicate that more recent cohorts were actually sicker, reflecting the increase in diabetes and obesity in the population.
Nevertheless, it was reassuring that the overall trends in mortality, and the relative mortality reductions for each cohort, were similar whether we fully adjusted for risk or adjusted only for age, race, and sex and their interactions.