Date of Award

January 2020

Document Type


Degree Name

Medical Doctor (MD)



First Advisor

Robert B. Schonberger

Second Advisor

Paul Heerdt


Variability in healthcare practice patterns is widespread, and the causes of such variability are numerous. Within the context of anesthetic care, many variations of care – both in terms of the selection of specific aesthetic agents and the chosen dosing of such agents – rely on individual clinician choices that cannot be explained purely by patient or procedural specifics. Provider preferences and habits have been suggested as primary drivers of much of this variability. While variation due purely to provider preferences may appear undesirable from a traditional operations research perspective, in the absence of evidence that one or another provider habit is better than the other, the natural variability in practice that results may be a positive development in allowing the opportunity for outcomes researchers to study associations between drugs and outcomes that may approach the causal implications of prospective studies. Put another way, many choices of drug and dose - when viewed across a large population sample and in the absence of evidence that one or another should be preferred - can be considered “almost as good as random.”

In the setting of such natural variability, the accumulation of numerous seemingly random healthcare choices across a population allows for the possibility of high-quality outcomes research using purely observational data sources. In that context, I first describe some background literature on the importance and possible sources of variation in healthcare practice patterns. I then proceed to describe the background, findings, and conclusions of two separate studies of different contexts of anesthetic variability.

In the first of two retrospective observational studies, my colleagues and I sought to describe variability in choices of anesthetic agents used during endovascular aortic valve replacement done under conscious sedation. The primary hypothesis of this study was descriptive – seeking to identify, describe, and quantify the variability in practice within a single institution during conscious sedation for essentially identical procedures. The descriptive hypothesis was that cases would be performed under conscious sedation using dexmedetomidine, propofol, and/or fentanyl as the primary sedation agents with no a priori prejudice as to the degree of preference shown among these three agents. Moreover, the study sought to determine, in an exploratory fashion, whether any of the observed drug combinations was associated with the need to convert from conscious sedation to general anesthesia. While the literature on doing these procedures under conscious sedation as opposed to general endotracheal anesthesia has received significant attention, almost all such studies have treated “conscious sedation” as a monolith without subdividing exactly what type of sedation was used. The study herein described is among the first to delve specifically into conscious sedation to explore whether the somewhat arbitrary provider preferences for conscious sedation might have demonstrated a signal of better or worse outcomes as has been suggested by proceduralists relying on anecdotal impressions.

After the above study, I turned my focus to another reasonably homogenous population – females 65 years of age and older undergoing gynecologic-oncologic surgery requiring general endotracheal anesthesia. As the goal of this second study was to investigate variability in drug dosing rather than drug selection, we limited our analysis to a group of patients who all received general anesthetic induction doses of propofol, the most used general anesthetic induction agent. Within this cohort, we sought to describe variations in provider dosing choices during induction. As with the first study, one of the primary hypotheses was descriptive in nature. Namely, we sought to describe the degree of dose variability, and specifically how such dose distribution related to FDA guidance on the appropriate induction dose of propofol in this population. We hypothesized that the maximum FDA recommended dose would be exceeded in at least 30% of cases. Furthermore, in secondary hypotheses, we explored whether deviations above the FDA guidelines were associated with particular demographic characteristics of patients (age, race, ethnicity, body mass-index), and then whether such overdose was associated with post-induction hypotension or acute kidney injury.

Of note, both studies were conducted in accordance with our a priori analytic protocols as registered on (NCT02786264 and NCT03699696), and both studies will have resulted in peer reviewed PubMed indexed publications.


This thesis is restricted to Yale network users only. This thesis is permanently embargoed from public release.