Disparities in Anesthesia Care

This issue of Anesthesia & Analgesia features an article by Andreae et al1 which should disturb and embarrass all perioperative physicians. Using data from the National Anesthesia Clinical Outcomes Registry (NACOR), the authors establish—through multiple statistical methods both simple and refined—that lower socioeconomic status is associated with inferior treatment during anesthesia, as measured by the administration of antiemetic medications. This is not a good message for our profession. Before rushing to judgment, however, we must ask: Should we believe the results? Readers must make their own judgments, but I believe the answer is “yes, the results are believable.” The authors studied a large cohort of cases selected from NACOR—more than 440,000 in the primary research dataset—in which intraoperative medications were recorded and the patient’s socioeconomic status could be inferred from their ZIP code. Although this set of data represents substantial attrition from the 12 million cases in NACOR at the time of the analysis, it is still a far larger population than has ever been studied on this topic. Only a small percentage of anesthesia departments participating in NACOR have succeeded in linking and routinely contributing data from their Anesthesia Information Management System, but even this small cohort can generate a large volume of data.2 This is especially true when the question asked is a broad one, applicable to many types of cases and patients. The 440,000 cases include data from 6 large institutions in the northeast and south of the United States. Should we be concerned that a selection bias contaminates the results? Probably not. Why should practice in other facilities or other parts of the country be different? While not identifiable in the data, these are obviously large facilities covering an array of procedures and patients. Viewed as a multi-institutional clinical trial, participation of 6 large sites and collection of more than 400,000 cases would be considered a significant achievement. Use of ZIP code to infer socioeconomic status is a common technique in health policy research, especially in retrospective reviews where this information has not been directly captured. Median family income for each block of codes is obtained from the US Census Bureau. While the specific income of any single patient is not known, the median for their geography is a reasonable surrogate, and valid when large numbers from each area are included. For the other economic indicator used by the authors—health care payer—NACOR does directly capture the data for each case. Although the authors suffered from missing records in this area of the study, the numbers remaining are still large and the relative proportions of cases covered by Medicare, Medicaid, commercial, and no insurance are consistent with other national observations. Prophylactic administration of antiemetics may seem a relatively trivial marker for the quality of anesthesia; patients seldom die from postoperative nausea and vomiting (PONV). As the authors note, however, use of antiemetic prophylaxis is an important patient-centered outcome. The American Society of Anesthesiologists has promoted appropriate administration of antiemetic prophylaxis as a federal measure in the Quality Payment Program and the Society for Ambulatory Anesthesia has published statements describing best practice which have been widely distributed and applied.3 Appropriate prophylaxis against PONV has been in the teaching curriculum of anesthesia residencies and the testing curriculum of the American Board of Anesthesiologists for several decades. Further, antiemetic prophylaxis is not strongly influenced by patient comorbidities, is not hugely expensive, generally occurs in the intraoperative period, and is usually the sole responsibility of the anesthesia clinician. It is likely that every anesthesiologist recognizes common risk factors for PONV, understands the need for prophylaxis, and knows what must be done. So this is actually a very strong marker for the quality of care. Is it possible that patients in higher socioeconomic categories were actually being overserved? The authors measured administration of medication, a surrogate, rather than actual occurrence of PONV, an outcome. Maybe all patients had a similar rate of PONV, and the additional prophylaxis given to the richer cohorts was merely wasteful. About this point we cannot be certain, barring complete understanding of every patient’s risk factors, but this is a relatively small quibble considering the magnitude of the observed differences in practice and the low cost and risk of prophylaxis. Most patients—most of us—put a high value on avoidance of PONV, and would rather have unneeded prophylaxis than nausea and vomiting.4 Finally, it is worth commenting on the statistical methods. This article introduces some highly sophisticated methodology, including Bayesian inference, to demonstrate the association between socioeconomic status and the use of antiemetics. The value and precision of these methods may be beyond most readers—including this one—and into the realm of advanced statistical science. What is reassuring, however, is that results of the study were strongly significant across multiple methodologies ranging from simple and familiar to highly complex. Risk factors identified for the use of prophylactic medications were consistent with national recommendations and previous studies. The numbers were large, as was the statistical confidence in the results in each analysis. The authors were not describing an edge-effect or lucky observation, but rather a strong and consistent correlation. If we believe that an association exists between the patient’s socioeconomic status and the use of antiemetic medications, the next question is: Why does this bias occur? One possibility is that anesthesiologists consciously omit prophylaxis in poor patients, but I do not believe this to be the case. It is unlikely that any of us set out to provide inferior care to any group of patients, and many anesthesiologists—including those at large teaching hospitals—would not know the patient’s insurance carrier in the first place. I believe the cause of this association is more subtle. The authors observed significant variability by site, suggesting that some of the facilities studied had more conscientious clinicians than others, or—more likely—had more robust mechanisms for ensuring PONV prophylaxis. Why? Because systematic improvement trumps individual effort every day. However, systematic quality improvement costs time, energy and money, and all of these things are in short supply in hospitals in lower socioeconomic neighborhoods. Not only are resources scarce but production pressure might be higher, language barriers more common, and staff less well trained and more transient. A plausible argument can be made for any of these factors increasing the barrier to establishment of a forcing function to encourage appropriate assessment of patient risk and administration of PONV prophylaxis. Some of these effects can be adjusted away by sophisticated statistics, but not completely, and facility-level differences are certainly a contributor to the observed disparity in care. While lack of resources to improve quality is a deplorable problem, it is at least recognizable. More concerning would be the idea that providers really do treat poor patients less compulsively. Even without direct information, it is certain that the clinician is forming a subconscious opinion about the patient’s socioeconomic status, based on hundreds of social cues including age, gender, race, clothing, language, behavior, questions asked, and even the appearance of family and friends. Does recognition of a poor patient make the clinician less compulsive? This is a risk that we should acknowledge is possible, but seek to inoculate ourselves against. How might we make improvements? Publication of this article is one way, through increased awareness of the problem. Continued measurement will certainly help, by creating a yardstick for performance and visibility of results. Better preoperative patient education will create a demand for good care, and increase the likelihood that patients may themselves remind the clinician of the need for antiemetics. Residency programs and professional societies could develop teaching materials to help anesthesia departments improve their quality systems, showing them how to implement proven best practices and Anesthesia Information Management System vendors themselves could build in functions and features—in a nationally standardized way—to promote appropriate risk assessment and prophylaxis. It is our responsibility as professionals to provide the best possible care to our patients, regardless of their situation in life. Recognizing our own blind spots in this area might be the first best step toward this goal. DISCLOSURES Name: Richard P. Dutton, MD, MBA, FASA. Contribution: This author reviewed the original paper and wrote this editorial. This manuscript was handled by: Nancy Borkowski, DBA, CPA, FACHE, FHFMA.