One of the studies most commonly cited in support of anesthesiologist assistant expansion is the 2018 Anesthesiology article, Anesthesia Care Team Composition and Surgical Outcomes, by Sun and colleagues. It gets used in legislative testimony, talking points, advocacy materials, and hallway debates as if it settles a much larger question: whether AAs and CRNAs produce equivalent anesthesia outcomes.
That is not what the study proves.
The paper is worth reading carefully because the problem is not necessarily the study itself. The problem is what people keep trying to make it say. The article examined a very specific question in a very specific setting: whether inpatient Medicare outcomes differed when a physician anesthesiologist supervised an AA versus when a physician anesthesiologist supervised a CRNA. That is a supervised anesthesia care team study. It is not a direct comparison of CRNA practice and AA practice across modern anesthesia delivery.
That design choice changes what the paper can honestly tell us. When both comparison groups include physician anesthesiologist supervision, the study cannot isolate the independent clinical contribution of either the CRNA or the AA. The physician anesthesiologist is built into both sides of the comparison. So when this paper is used as proof that AAs and CRNAs are interchangeable, the claim has already gone beyond the study design.
Calling this a CRNA-versus-AA outcomes study is misleading. It compared supervised ACT billing categories and measured three broad hospital outcomes that may have little, or nothing, to do with the anesthetic.
The study evaluated 443,098 Medicare inpatient surgical cases from 2004 through 2011. The population was elderly Medicare patients, ages 65 to 89, undergoing inpatient surgery in states where AAs were permitted to practice during the study period.
The comparison was physician anesthesiologist plus AA versus physician anesthesiologist plus CRNA. Those are two physician anesthesiologist-supervised care team configurations. The study did not compare independent CRNA care to AA care. It did not compare CRNA-only models to AA models. It did not compare rural anesthesia models, collaborative models, obstetric models, outpatient anesthesia, pediatric anesthesia, or modern staffing arrangements.
The outcomes were also limited. The authors measured inpatient mortality, inpatient length of stay, and inpatient spending. These are broad hospital outcomes. They are not anesthesia-specific outcomes. The paper did not measure airway events, aspiration, awareness, rescue calls, medication errors, PACU complications, ICU escalation, hemodynamic instability, failure-to-rescue, or whether the anesthesia professional recognized and managed a developing crisis.
Those details are exactly where anesthesia care often lives. A patient can have an anesthesia-related complication that is rescued successfully and never appears as inpatient mortality. A difficult airway can be managed well or poorly without changing length of stay. A near-miss can be clinically important and completely invisible in claims data. A study built around mortality, length of stay, and spending is not designed to answer whether two anesthesia professionals have equivalent clinical training, judgment, independence, or capability.
No statistical difference in three broad inpatient outcomes does not prove two anesthesia professions are interchangeable.
The study found no statistically significant differences in the three measured inpatient outcomes. Adjusted inpatient mortality was 1.7% in the CRNA-team group and 1.6% in the AA-team group. Adjusted length of stay was essentially the same. Adjusted inpatient spending differed by about $56, which was also not statistically significant.
A careful summary would be: in this claims-based study of elderly Medicare inpatients receiving care in physician anesthesiologist-supervised teams, the specific supervised team member was not associated with statistically significant differences in inpatient mortality, length of stay, or inpatient spending.
That is a reasonable conclusion. It is also much narrower than the way this article often gets used publicly. The study did not show that AA care and CRNA care are equivalent. It showed that two supervised care team configurations did not differ on three broad inpatient outcomes in one retrospective claims-data analysis.
Weaponized social media summaries intentionally blur that distinction.
The central limitation of using this article for AA-versus-CRNA advocacy is that both study groups included physician anesthesiologist supervision. The authors were not comparing independently practicing CRNAs to AAs. They were not comparing CRNA-only models to AA models. They were not comparing rural CRNA practice to urban ACT practice. They were not evaluating what happens when no physician anesthesiologist is present.
They compared two versions of the physician anesthesiologist-supervised ACT model.
That makes the article more relevant to the internal structure of the ACT model than to the broader question of CRNA and AA equivalence. If ASA wants to argue that, under physician anesthesiologist supervision, the measured hospital outcomes in this dataset were similar, that is a fair reading. If the argument becomes “this proves AAs and CRNAs are equivalent,” the study is being stretched past its actual findings.
The authors themselves were careful on this point. They noted that team structures and physician supervision may mitigate any systematic differences between AAs and CRNAs, if such differences exist. That is very different from proving there are no meaningful differences between the two professions. A supervised system can reduce variation because the system itself is designed to buffer differences in training, experience, scope, and judgment.
That is supervision. It is not proof of provider equivalence.
The study tells us who was listed in the supervised care team. It does not tell us who managed the anesthetic, who needed rescue, or whether any outcome was anesthesia-related.
One of the biggest problems with using this article as a CRNA-versus-AA comparison is that a physician anesthesiologist was involved in every case group by design. That structure may be useful if the question is limited to supervised ACT composition, but it creates a major blind spot if the claim is professional equivalence.
The study does not tell us how much physician anesthesiologist involvement occurred in either group. It does not tell us whether the physician anesthesiologist spent more time in AA rooms than CRNA rooms. It does not tell us whether AAs required more direction, more correction, more rescue, or more real-time intervention than CRNAs.
That missing information changes what can honestly be claimed from the paper. If physician anesthesiologists were more involved in AA cases, then the study would not be measuring equivalent AA and CRNA performance. It would be measuring the outcome of a system where physician anesthesiologist involvement may have compensated for differences in training, scope, experience, or independence. The authors’ own framing leaves room for that possibility when they discuss supervision and team structure as mechanisms that may mitigate differences.
The same issue runs in the other direction. If physician anesthesiologists were minimally involved in both groups, the study still does not show what happened clinically inside the anesthetic. Claims data cannot tell us who recognized the problem, who made the key decision, who managed the airway, who treated the hypotension, who called for help, or who rescued the patient before a bad outcome occurred.
The “physician anesthesiologist was involved” point cuts both ways. It prevents the study from isolating the independent contribution of the CRNA or the AA, and it also prevents the study from proving the independent value of physician anesthesiologist involvement itself. The physician anesthesiologist is present in the billing structure, but the actual degree of clinical involvement is not measured.
A study that cannot account for physician anesthesiologist involvement cannot be used to claim that AAs and CRNAs perform the same clinically. At most, it can say that in this dataset, two physician anesthesiologist-supervised team configurations had similar broad inpatient outcomes.
Without measuring physician anesthesiologist involvement, the study cannot tell whether similar outcomes reflected provider equivalence or different levels of supervision.
The study used inpatient mortality as one of its three primary outcomes. That sounds powerful because mortality is obviously important, but inpatient mortality is a broad hospital endpoint. It captures whether the patient died at any point during the hospitalization. It does not tell us whether the death was caused by anesthesia care, surgical care, postoperative management, medical disease, trauma burden, or progression of the condition that brought the patient to surgery.
In practical terms, the mortality window extends from the operating room through the rest of the inpatient stay. A patient could die days later from sepsis, hemorrhage, pulmonary embolism, myocardial infarction, stroke, ICU complications, withdrawal of care, malignancy, frailty, or the severity of the original injury. Those deaths may have little or nothing to do with the anesthesia professional in the room.
That is especially relevant in an elderly Medicare inpatient population. These patients are often sicker, more fragile, and more likely to have complicated postoperative courses. Their risk of death is influenced by surgical urgency, baseline comorbidities, hospital resources, ICU care, nursing care, source control, postoperative complications, and goals-of-care decisions. None of that can be separated cleanly from the anesthesia claim used to classify the care team.
The article also does not tell us whether anesthesia-related mortality differed between groups. If one group had more airway events, aspiration events, hemodynamic collapses, medication errors, or rescue failures that contributed to death, we would not know from the primary endpoint. The study reports inpatient mortality, not anesthesia-attributable mortality.
The same problem applies to length of stay and spending. A longer admission may reflect surgical complications, infection, discharge barriers, rehabilitation needs, ICU bed availability, social issues, or severity of injury. Higher spending may reflect the same things. Without knowing why the patient stayed longer or why the hospitalization cost more, those endpoints cannot be used to make claims about anesthesia provider equivalence.
For this workforce debate, the more relevant outcomes would be anesthesia-specific complications, rescue calls, airway events, aspiration, awareness, PACU respiratory events, unplanned postoperative ventilation related to anesthesia, intraoperative hemodynamic instability, medication errors, regional anesthesia complications, and documented physician anesthesiologist intervention. The Sun study did not measure those outcomes.
Mortality, length of stay, and spending may be reasonable endpoints for a broad health services research question. They are not enough to prove that AAs and CRNAs are clinically equivalent anesthesia professionals.
Inpatient mortality tells us whether the patient died before discharge. It does not tell us whether anesthesia caused, prevented, or had anything to do with that death.
When hospital leaders, legislators, and anesthesia professionals talk about safety, they are usually not limiting the conversation to inpatient mortality, length of stay, and spending. They are talking about whether the person providing anesthesia can manage the case in front of them, anticipate deterioration, rescue complications, and make appropriate decisions when things change quickly.
This study does not tell us whether an AA and a CRNA are equivalent in airway management, crisis recognition, independent judgment, preoperative assessment, emergence planning, regional anesthesia, obstetrics, pediatrics, outpatient anesthesia, office-based anesthesia, trauma, rural coverage, or critical access hospital practice.
It also does not identify whether an adverse event was related to anesthesia, surgery, medical disease, nursing care, hospital resources, postoperative management, or discharge planning. A claims-based inpatient mortality endpoint cannot answer those questions.
That does not make the study useless. It makes the study limited. The problem is pretending those limits do not exist.
A near-miss can be clinically important and completely invisible in claims data.
The final sample included 421,230 CRNA-team cases and 21,868 AA-team cases. AA-team cases represented about 4.9% of the final analytic sample.
That point should not be exaggerated. The AA group was still large enough to study in the context of the authors’ model. But it is relevant when the paper is presented as a sweeping national statement about anesthesia workforce policy. The study was based on a relatively small AA-team share, drawn from AA-permitted states, during a historical period that ended in 2011, using traditional Medicare inpatient claims.
That is a narrow slice of anesthesia practice. It does not automatically generalize to today’s outpatient-heavy surgical environment, obstetric units, pediatric anesthesia, GI centers, office-based anesthesia, rural hospitals, private-payer populations, or modern staffing economics.
The authors used risk adjustment, hospital fixed effects, and an instrumental-variable approach to reduce confounding. That is better than a simple unadjusted comparison. Still, the paper acknowledges the possibility of residual confounding and specifically notes that unobservable differences in patient complexity and case assignment could remain.
That limitation is important because anesthesia staffing is not random in real life. Sicker patients, more complex procedures, difficult surgeons, unstable trauma cases, complex vascular cases, major abdominal procedures, and high-risk emergencies may not be distributed evenly between provider groups. A department may assign certain rooms, surgeons, procedures, or patient types differently based on staffing, experience, trust, supervision needs, or institutional habit.
If CRNAs were assigned more difficult cases and AAs were assigned less complex cases, the study may understate CRNA performance. If AAs were assigned more difficult cases but received more direct physician anesthesiologist involvement, the study may reflect physician-supported team performance more than AA performance. The claims data cannot resolve that.
The authors tried to address measurable differences. That is appropriate. But measurable differences are not the same as all clinically meaningful differences. In anesthesia, the variables that matter most are often not fully captured in administrative claims: airway difficulty, induction instability, bleeding trajectory, urgency, surgeon behavior, room assignment, supervision intensity, real-time rescue, and who actually made the critical decisions.
The study also could not control for provider experience. That is a major limitation in anesthesia because the label attached to the billing claim does not tell us whether the person in the room had one year of experience or twenty. It does not tell us whether the provider routinely handled high-acuity vascular cases, obstetrics, pediatrics, trauma, independent call, regional anesthesia, or complex airways. It also does not tell us whether a less experienced provider received more direct physician anesthesiologist involvement than a more experienced provider.
That matters clinically because experience changes the entire supervision equation. A new provider may need more direction, more backup, more prompting, and more rescue than a provider who has managed thousands of anesthetics across multiple practice settings. If one group had more experienced providers and the other group had less experienced providers but more physician anesthesiologist support, the study would not be measuring true provider equivalence. It would be measuring a mixture of provider type, experience level, case assignment, and unmeasured physician anesthesiologist involvement.
Claims data cannot separate those variables.
A billing claim tells us the provider type. It does not tell us whether the person in the room had one year of experience or twenty.
One of the most important practical issues in ACT staffing is supervision ratio. A physician anesthesiologist supervising one room is not the same operational model as a physician anesthesiologist medically directing four rooms. Availability, response time, case complexity, induction and emergence overlap, and competing demands all change as the number of concurrent rooms changes.
The Sun study could not measure supervision ratios. The authors acknowledged that limitation directly, noting that their data did not allow them to adjust for differences in supervision ratios between AA and CRNA care teams.
That limitation is important even within Medicare’s medical direction framework. Medicare billing can identify medical direction categories, but the billing structure does not give the clinical detail needed to know whether AA and CRNA cases were distributed similarly across 1:1, 1:2, 1:3, or 1:4 coverage. A medically directed AA case and a medically directed CRNA case may both appear as supervised team care, but the study does not tell us whether one group received closer physician anesthesiologist coverage than the other.
That leaves a major unanswered question. If AA cases were more often supervised 1:1 or 1:2 while CRNA cases were more often managed at 1:3 or 1:4, similar broad hospital outcomes would not prove equivalent provider performance. They could reflect different levels of physician anesthesiologist involvement. The same problem applies if one provider group required more rescue, more direction, or more real-time correction during the case. The study did not measure that.
If the claimed benefit of the AA model is physician anesthesiologist supervision, then the intensity and availability of that supervision cannot be treated as a background detail. It is central to the model. The Sun article did not control for it.
A 1:1 supervised room is not the same operational model as four rooms competing for one physician anesthesiologist’s attention.
This is one of the biggest overextensions of the article. The study has nothing to say about independent CRNA practice.
The authors excluded certain opt-out state cases after opt-out because the purpose was to compare physician anesthesiologist-supervised care teams. That design choice makes sense for the study they were conducting. But it also means the article cannot be cited as evidence against independent CRNA practice, CRNA-only models, collaborative practice, or rural models where CRNAs provide the anesthesia coverage.
CRNAs practice in every state and in many different care models. AAs do not. AAs are legally dependent on physician anesthesiologist supervision. That structural difference is central to the workforce debate, but the 2018 article was not designed to evaluate it.
So when someone uses this paper to imply that AA expansion is equivalent to CRNA practice across the healthcare system, they are skipping over the most important operational difference between the two professions.
AA expansion is often sold as a solution to anesthesia access problems. This article does not prove that claim.
The authors themselves identified access and competition as areas for future research. That means the study did not show that introducing AAs improves access to surgical care, improves rural coverage, reduces cancelled cases, lowers anesthesia costs, stabilizes call coverage, or expands services in underserved communities.
Those are separate policy questions. They require different data.
A study showing no statistically significant difference in inpatient mortality, length of stay, and spending inside supervised ACT models does not tell us whether adding a legally dependent anesthesia provider improves access. In many settings, especially rural and underserved areas, mandatory physician anesthesiologist supervision may be the limiting factor rather than the solution.
For CRNAs, the access argument has always been tied to flexibility. CRNAs can work in physician anesthesiologist-led teams, collaborative models, CRNA-only groups, rural hospitals, critical access facilities, obstetric units, and independent practice environments. AAs cannot provide that same flexibility because their practice is legally tethered to physician anesthesiologist supervision.
The Sun article does not resolve that workforce issue.
The article disclosed ASA funding, and one of the authors was employed by ASA. The paper also states that the Stanford investigators retained final control over the study design, manuscript, and publication decisions.
That disclosure should be handled fairly. Funding does not automatically invalidate a study. Industry and professional organizations fund research all the time. The question is not whether the article should be dismissed because ASA funded it. The better question is whether the study’s conclusions are being used within the limits of the study design.
In this case, the major issue is not the funding disclosure. The issue is the advocacy leap that followed. A narrow claims-based study using broad inpatient hospital metrics became a sweeping claim about professional equivalence. The study measured mortality, length of stay, and spending. It did not measure anesthesia-specific anything. That is not a small limitation. That is the difference between what the paper studied and what ASA/AAAA advocates keep claiming it proved.
Funding disclosure matters, but the bigger problem is the advocacy leap that followed.
If someone wants to cite the article accurately, the language should look something like this:
The 2018 Sun et al. study found no statistically significant difference in inpatient mortality, inpatient length of stay, or inpatient spending between physician anesthesiologist-supervised teams involving AAs and physician anesthesiologist-supervised teams involving CRNAs among elderly Medicare inpatient surgical patients from 2004 through 2011.
That is accurate.
Here is what the article should not be used to claim:
The study does not prove that AAs and CRNAs are equivalent anesthesia professionals. It does not evaluate independent CRNA practice. It does not evaluate CRNA-only models. It does not evaluate AA practice without physician anesthesiologist supervision. It does not measure anesthesia-specific complications. It does not measure supervision ratios. It does not prove AA expansion improves access. It does not prove AA expansion lowers total anesthesia costs. It does not address whether AAs can replace CRNAs across the full range of anesthesia practice.
That is the difference between reading the article and weaponizing the headline.
If the article did not measure it, do not let anyone claim the article proves it.
When this article gets cited as proof that AAs and CRNAs have equivalent outcomes, the response does not need to be complicated:
The study compared two physician anesthesiologist-supervised ACT models in elderly Medicare inpatient cases. It measured inpatient mortality, length of stay, and spending. It did not compare independent CRNA practice to AA practice, did not measure anesthesia-specific complications, did not evaluate supervision ratios, did not control for provider experience, did not measure physician anesthesiologist rescue or intervention, and did not answer whether AAs and CRNAs are equivalent across anesthesia practice.
That is the cleanest way to bring the discussion back to what the paper actually says.
The 2018 Anesthesiology article is a supervised ACT composition study. The study only evaluated general inpatient mortality, length of stay, and spending - that happened to use medical direction models - it had no anesthesia or surgical specific claims.
That finding should not be inflated into a claim that AAs and CRNAs are clinically equivalent, interchangeable, or equally capable across all anesthesia practice models.
The study does not answer that question.
And when ASA or AA advocates use it that way, they are not citing the study accurately. They are using a narrow claims-data article to support a broader policy claim the article was never designed to prove.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.