VA Beats Private Sector Hospitals for Post-Op Surgical Mortality

Veterans frustrated with the VA health system who seek surgical care at private hospitals -- as they can now do, with the government footing the bill -- might actually face worse outcomes, according to a new study.

After adjusting for differences in frailty and other risk factors for postoperative complications, surgical patients in Veterans Health Administration (VHA) hospitals were markedly less likely to die within 30 days compared with those treated in private sector facilities (relative risk [RR] 0.59, 95% CI 0.47-0.75), reported Elizabeth George, MD, MSc, of Stanford University in California, and colleagues.

A similar difference was seen among patients experiencing complications: the risk of dying from them was also nearly halved in the VA system compared with private sector hospitals (adjusted RR 0.55, 95% CI 0.44-0.68), the researchers' report in JAMA Surgery indicated.

The findings came from an analysis of about 740,000 noncardiac procedures performed at VHA centers during 2015 to 2018 and recorded in the VA's Surgical Quality Improvement Program (VASQIP), along with some 3.2 million performed in the private sector over the same period and included in the National Surgical Quality Improvement Program (NSQIP), which was modeled on the VA's system.

In an invited commentary, however, three specialists at Massachusetts General Hospital in Boston cautioned that the study isn't quite definitive proof that VA care is superior to that provided elsewhere.

Aside from the familiar limitations that come with database analyses such as this one, the trio, led by David C. Chang, PhD, MPH, MBA, noted that "failure to rescue" (that is, dying after experiencing postoperative complications) may not be a very revealing outcome measure.

"A hospital that causes many iatrogenic complications and injuries would inflate their denominator and paradoxically lower their [failure to rescue rate]," the commentators wrote. "Even if that hospital could ultimately rescue its patients from death, most patients would probably not consider it to be a good hospital."

But overall, Chang and colleagues said the study was good and useful, for two reasons. One, it "demonstrate[s] the important distinction between quality of care and timeliness of access," the latter being the private sector's chief purported advantage.

And two, the commentators applauded George and colleagues for making use of a relatively new statistical tool called the E-value that provides a numerical cap on the potential influence of unmeasured confounders -- the Achilles' heel of retrospective and observational studies. The E-value's developers, introducing it in a 2017 Annals of Internal Medicine paper, explained that it indicates "the minimum strength of association, on the risk ratio scale, that an unmeasured confounder would need to have with both the treatment and the outcome to fully explain away a specific treatment–outcome association, conditional on the measured covariates."

For the current study's main finding of a relative risk of 0.59 for 30-day mortality, George and colleagues calculated an E-value of 2.78 -- meaning that an unmeasured confounder must nearly triple the risk of 30-day mortality in VHA hospitals for it to erase the advantage shown in the authors' data. That seems vanishingly unlikely.

George's group concluded that, based on their findings, allowing and even encouraging veterans to obtain care outside the VA may be misguided.

"Veterans are a unique patient population that benefit[s] from the tailored care processes the VHA has developed, and it could be difficult to replicate this in the private sector," they wrote.

The authors also argued that the VA's better care could be jeopardized if "private sector diversion continue[s] to reduce VHA utilization and, by extension, funding."

Other Study Highlights

Patients in the two databases differed substantially, as one might expect. More than 90% of those in the VA system were men versus about 47% in the NSQIP data (hence, gynecologic procedures were excluded from the study).

Moreover, VA patients were considerably older (mean age 64 vs 58) and, most importantly, far more likely to be frail as rated by the Risk Analysis Index (RAI; developed by some of the current study's authors). Fully 60% of VASQIP patients were rated as frail or very frail, compared with 21% of the NSQIP cohort.

The authors' adjusted statistical models included the RAI, measures of each procedure's estimated physiologic stress, surgical urgency, and complications.

NSQIP patients appeared to fare better prior to these adjustments: 30-day mortality was one-third greater in the VA cohort (1.1% vs 0.8%) and complication rates were nearly doubled (17.1% vs 9.5%). But taking account of VA patients' greater frailty led to the finding of reduced mortality risk. And rates of failure to rescue were lower at VA facilities even in the raw data (4.7% vs 6.7%).

Comment