Do individual anesthesiologists directly affect patient outcomes? For years, anesthesiologists have attempted to answer this question empirically. Now, a recent study has concluded that anesthesiologists do affect outcomes, and at a rate similar to that of surgeons.
“Despite everything that’s happened with surgical improvement over the years, we still see troubling rates of complications and mortality, even in the near term,” said Sachin Kheterpal, MD, MBA, associate professor of anesthesiology at the University of Michigan Medical School, in Ann Arbor. “There’s been overwhelming focus on surgeons’ effects on outcomes. In fact, it seems like everything has been looked at, other than the anesthesiologist. This is despite the fact that we know there is a wide variation in the process of care happening on the other side of the drape.”
Although previous research on the topic has been conducted, it was limited to a single state, and the original article (Anesth Analg 2015;120:526-533) was retracted because of faulty statistical methodology.
So there are no robust analyses of national data spanning a range of procedures that evaluate the influence of the anesthesiology provider on surgical outcomes. This did not stop Dr. Kheterpal and his colleagues from hypothesizing that a measurable proportion of outcome variation can be attributed to differences in anesthesiology practice and the provider. “Our impact may be nothing too large, depending on patient risk and procedure type,” he said.
To help tease out these potential effects, the investigators obtained data on more than 1 million vascular, cardiac and colorectal operations from the Medicare Provider Analysis and Review files, which contain all hospital discharges for Medicare recipients receiving acute care. International Classification of Diseases, Ninth Revision codes were used to identify all patients who underwent abdominal aortic aneurysm repair, coronary artery bypass graft surgery or colectomy, between 2010 and 2013. These data were then linked to Part B professional claims to identify the primary surgeon and anesthesiologist by codes. “We chose these three procedures because we believe they reflect a wide variety of risks and variations in care,” Dr. Kheterpal said.
The primary outcome of the analysis was a composite of 30-day all-cause mortality or anesthesia-related morbidity: pulmonary failure, pneumonia, myocardial infarction, venous thromboembolism or acute renal failure. The secondary outcome was the occurrence of an anesthesia-related morbidity.
“Let’s be honest that it’s impossible to completely isolate the anesthesiologist’s effect,” Dr. Kheterpal said. “It’s hard to figure out how much is one person’s fault and how much is someone else’s. But that doesn’t mean we can’t come up with a conceptual model.”
That took the form of a mixed-effects model, combining the anesthesiology provider as the random effect with fixed effects for a patient’s comorbidity burden, including such factors as age, male sex, emergency procedure, year of procedure, teaching hospital status, hospital bed size and surgeon procedure volume quartile. Each procedure was modeled separately, and the primary model sought to determine the amount of outcome variation that was attributable to the anesthesiologist. The study’s end point was intraclass correlation, which reflected the proportion of total outcome variation accounted for by the provider.
“We obviously focused on the anesthesiologist, but for a sensitivity check we also looked at the surgeon’s impact to see if we were totally off base,” Dr. Kheterpal said.
As reported at the 2016 annual meeting of the American Society of Anesthesiologists (abstract BOC01), the analysis involved 168,715 AAA repairs, 434,874 CABG surgeries and 437,846 colectomies. It was found that the 30-day combined mortality/morbidity by procedure was 22%, 44% and 14%, respectively. “When you include complications, a lot of bad things happen, despite all the advances we’ve made,” he said.
The intraclass correlation for the anesthesiologist ranged from 3.1% to 4.5% across the procedures (Table). Of note, the intraclass correlation for the surgeon varied from 4.2% to 5.2%. After adjusting for risk and reliability, absolute complications rates were found to vary threefold between low- and high-performing anesthesiologists.
Table. Overall Events and Morbidity Attributable to Anesthesiologists and Surgeons For Three Surgical Procedures
Surgery Total Patients, N Events, n (%) Outcome Variation Attributable to Anesthesiology Provider, % (range) Outcome Variation Attributable to Surgery Provider, % (range)
AAA 168,715 36,892 (22) 4.4 (3.5-5.7) 4.2 (3.5-5.1)
CABG 434,874 190,805 (44) 4.5 (4.2-4.8) 5.1 (4.7-5.5)
Colectomy 437,846 59,140 (14) 3.1 (1.8-5.3) 5.2 (4.0-6.7)
AAA, abdominal aortic aneurysm repair; CABG, coronary artery bypass graft
Despite the robustness of the analysis, Dr. Kheterpal was quick to point out its limitations. “We were only looking at in-hospital events, and anesthesiologists’ effects may take longer. Also, this is only Medicare data. Our population was largely patients over age 65, and we only looked at three relatively homogeneous operations.”
“So do we matter?” he asked. “The answer is complex. Either way, I think this reopens the debate regarding our value. In the future, we can study more procedure types, different patient types, and dive into the clinical variation underneath it all.”
As Franklin Dexter, MD, PhD, told Anesthesiology News, learning how best to partition the observed outcomes of patients among hospitals, surgeons and anesthesiologists is a challenging statistical problem. “It is great to see the efforts that Dr. Kheterpal and his colleagues are making in this regard,” he said, a professor of anesthesia at the University of Iowa’s Carver College of Medicine, in Iowa City. “Gaining statistical experience in this area is important, because the data are available.
“Regulations and accreditation are principally at the level of the hospital, and those regulations and accreditations require evaluation—credentialing—of the surgeons and anesthesiologists,” Dr. Dexter explained. “In the long term, there are opportunities to have greater statistical reliability at the hospital level. This is like financial accounting—it must be rigorous and of known statistical reliability. In contrast, for internal evaluation and decision making, managerial accounting can reasonably accept a lower standard, and it is not generalizable—i.e., science.”