Absence of Evidence or Evidence of Absence?
What do we know of the impact unions have on residents?
The number of resident physician union drives has sharply increased since the COVID-19 pandemic. In response, health systems have engaged in intense union-busting campaigns, also known as Boss Campaigns, sowing doubts among residents, suggesting that self-determination and collective bargaining power may not yield significant benefits and may only create an adversarial relationship between residents and fellows. Most of the messages are not based on any evidence. That said, over and over, administrators rely on a 2021 JAMA Network Open study of resident labor union participation and surgical resident wellbeing. So let’s dive into this study.
1. Study Setting
This study was based on a national cross-sectional survey administered to residents in all non-military general surgery residents in January 2019 after their American Board of Surgery In-Training Examination (ABSITE), a high-stakes exam surgery residents take yearly to assess their progress.
The primary outcome was burnout, defined as experiencing any symptom of depersonalization or emotional exhaustion at least weekly. Secondary outcomes included suicidality, measures of job satisfaction, duty hour violations, mistreatment, educational environment, salary, and benefits.
2. What Did The Study Find
They found that there were no significant differences in suicidality, job satisfaction, duty hour violations, mistreatment, educational environment, salary, or benefits, except that unionized programs more frequently offered four weeks instead of two to three weeks of vacation and more frequently offered housing stipends.
3. Problems With the Study
First, I personally find it most concerning that the authors draw conclusions that imply causation (or lack thereof). That there were no major differences by unionization status does not mean that unions do not have an impact on resident burnout; absence of evidence does not equal evidence of absence. The appropriate design to answer this question would require comparing changes in either programs’ burnout rates or individuals’ burnout scores longitudinally, before and after unionization (what economists call a differences-in-differences analysis). Specific conditions tend to lead workers to form a union. It remains possible that burnout rates at unionized programs improved to the level of non-unionized programs, so finding no statistically significant difference at one point in time says nothing about the effect of unions on the outcome of interest essentially. Surgery residents are surveyed every year after ABSITE. Longitudinal burnout data is thus available, and one cannot help but wonder why seasoned, well-funded researchers would instead pick a cross-sectional design to answer a question that requires an event study.
Second, people self-select into programs and may be different enough that comparisons across programs may not be appropriate without using a tool like a propensity score matching approach to account for measurable confounders.
Third, measuring burnout at the end of a grueling exam residents have spent weeks studying toward, outside of 80-hour work weeks, may impact the validity of the measure. Unionized or not, it would not surprise me that most surgery residents feel emotionally exhausted for several weeks leading to their ABSITE.
3. Now, What?
When the study came out, I worried that anti-union leaders in medicine would weaponize it for years to come. Indeed, leaders at my organization have sent the study’s null findings in an email to the entire house staff as part of their boss campaign. A resident who thinks we should pause the union drive (a call to do nothing, otherwise put) cited it in a recent OpEd as part of his argument. There remains an opportunity to empirically study the impact that unions have on graduate medical education in the US. Burnout may not be the most useful measure, as unions cannot promise a change in culture within toxic organizations. They cannot force attendings to give residents more autonomy on rounds or in the operating room. We’re more likely to detect meaningful changes in the realm of compensation, benefits, workplace protection, and even a sense of agency. There could also be spillover effects: what changes happen at program B when a union is formed, and a contract is signed at program A?
There is a lot to learn about the impact of unions. Until more empirical work is published, people curious about what unions can do for residents can visit the CIR website to glance at various contracts, for example, or learn more about the lived experience of residents at unionized programs. Unions have a track record of helping workers across many professions. Why would they not help residents? We’re not unicorns.