Why Has Mortality Related to Pancreatectomy Declined?

SAN FRANCISCO—Nationally, the rate of in-hospital mortality associated with pancreatectomy for pancreatic cancer and other neoplasms has decreased in recent years, researchers reported at the 2006 Gastrointestinal Cancers Symposium (abstract 94). Although a greater proportion of these procedures are now being done in higher-volume hospitals, this shift only partly explains the drop in mortality. Investigators at the University of Massachusetts Medical Center led by James T. McPhee, MD, used a nationally representative sample—the National Inpatient Sample—to study trends in outcomes of pancreatectomy for neoplasms between 1998 and 2003.

The overall rate of in-hospital mortality for the entire 6-year period was 5.8%, Dr. McPhee said. When the data were stratified by year, the rate decreased significantly, from 7.7% in 1998 to 4.4% in 2003. A regression analysis further confirmed the drop in mortality, indicating that this outcome fell by 0.42% per year.

The investigators next stratified hospitals by volume, in a way that maintained equal numbers of patients in each group. By this process, low-volume hospitals were those performing fewer than four pancreatectomies for neoplasms each year; medium-volume ones, 4 to 13 per year; and high-volume ones, more than 13 per year. For the entire 6-year period, the in-hospital mortality rate was 9.6%, 5.5%, and 2.1% in low-, medium-, and high-volume hospitals, respectively. As found among hospitals overall, the rate fell within each group during the study period: from 14% to 8.3% in the low-volume facilities, from 4.3% to 3.5% in the medium-volume ones, and from 2.8% to 2.0% in the high-volume ones.

In a multivariate analysis including seven factors (age, sex, hospital surgical volume, hospital teaching status, year of resection, payer status, and comorbid conditions), compared with patients operated on in high-volume hospitals, those operated on in medium-volume ones had more than double the risk of death (odds ratio, 2.4) and those operated on in low-volume hospitals had more than quadruple the risk (odds ratio, 4.1).

"Interestingly, while in the univariate analysis, hospital teaching status was correlated with operative mortality, with urban teaching hospitals having superior outcomes, when adjusted for the other variables, specifically hospital surgical volume, hospital teaching status then became insignificant," Dr. McPhee noted.

Paradigm Shift?

The observed temporal pattern of in-hospital mortality could not be explained by disproportionate sampling of hospitals of various volumes over time or by differing patient case mixes at hospitals with differing volumes, according to Dr. McPhee. "This prompted us to ask this question: Could the decrease in mortality over time reflect a paradigm shift whereby a higher percentage of pancreatic resections are being performed at the higher-volume surgical centers?" he said.

Further analysis supported at least some role for this phenomenon: From 1998 to 2003, the proportion of all resections done annually in high-volume hospitals increased from 32% to 40%, and for medium- or high-volume hospitals, from 60% to 69%.

In a final analysis, Dr. McPhee and his team calculated the contribution of this shift to the observed reduction in in-hospital mortality over time. This analysis showed that the shift accounted for a 0.8% drop in mortality during the 6-year period—only a modest part of the 3.3% drop actually observed.