Zachary Klaassen: Good afternoon. Thanks for joining us today for another UroToday Journal Club. I'm joined by Dr. Raj Satkunasivam, who is an Assistant Professor in the Department of Urology at Houston Methodist Hospital and Dr. Chris Wallis, who is a Fellow in urologic oncology at Vanderbilt University Medical Center, and I am Zach Klaassen, Assistant Professor at the Medical College of Georgia in Augusta, Georgia. Today we're going to be discussing the relationship between surgeon age and postoperative outcomes for this Journal Club.

This paper was published this last week in the Canadian Medical Association Journal entitled "Relationship Between Surgeon Age and Postoperative Outcomes, A Population-Based Cohort Study". As you can see, Raj, myself, and Chris were all authors on this paper.

So really, it's important to discuss quality outcomes in surgery. There are several items that are required for high-quality outcomes in surgery. One is certainly knowledge. The surgeon has to have a good knowledge basis for the pathophysiology of the disease, for the steps of the procedure, but certainly, other outcomes that are important may include the technical proficiency. Certainly, all of us go through a rigorous residency program, some of us may do additional years of fellowship, and then even into being an attending and a junior attending, learning the technical proficiency of complex operations.

And the third requirement for high-quality outcomes in surgery is communication. Communication not just with each other, but with the patients, with our staff around us, and having a good flow in the operating room.

The fourth thing, and which cannot be understated, is good clinical judgment. Oftentimes the joke is made that we can teach people to operate, but we can't teach them clinical judgment. That sometimes is understated knowing when to take somebody to the operating room is as important as when not to take somebody to the operating room.

And the final thing in terms of what's required for high-quality outcomes in surgery is sort of this context of volume and it's team system. So how many times do we do an operation, how often we do an operation over the course of a year or so can impact the quality of outcomes after surgery.

Why is surgeon age and impact on surgical outcomes worth studying? Well, the surgical workforce is aging, specifically for urologists, depending on which statistics you read, we're somewhere between the second and third oldest operating group amongst surgeons. I think in terms of this study being impactful for urologists in general, this is pertinent to us in terms of our aging workforce. We also know that aging impacts cognitive, visual, and motor functions and that this may have a negative impact on surgical ability. Certainly, there are developments of tremors as we get older and certainly how well we react to difficult situations may be impacted by age.

However, older surgeons may have the potential to offset age-related decline. This is where the argument comes into play where they may have more experience, their age and their experience may improve clinical acumen and which ultimately may refine clinical judgment. So taken together, there's a lot of factors that go into aging and whether there are negative or positive factors. It's probably a combination of both.

So over the last several years, there have been several studies looking at data on surgeon age and outcomes, and really there's been conflicting results. So this is still somewhat of a controversial area in outcomes research. Looking back at this paper from 2006 from the University of Michigan led by John Birkmeyer and Justin Dimick. They looked at surgeon age and increased operative mortality among Medicare beneficiaries in the late '90s. So what they did for this study, they looked at eight different procedures. You can see four of them listed here, and you can see that with increasing age, there was a higher adjusted operative mortality, specifically for carotid endarterectomy as well as coronary artery bypass grafting.

The complementary to figure this is the remaining set four procedures and you can see here, lung resection, esophagectomy, cystectomy, and pancreatectomy. Once again, pancreatectomy with increasing surgeon age had increased adjusted operative mortality. And this was primarily seen in patients that were older as well as low volume operative surgeons.

In the second study published in BMJ in 2018, this group looked at surgeon age and operative mortality once again among Medicare beneficiaries from 2011 to 2014, and interestingly we see the opposite. So as surgeons get older, they have decreased odds of having postoperative mortality. You can see here, greater than 60 compared to the reference less than 40 as well as 50 to 59 decreased operative mortality. As you can see just from these two studies alone, there are different outcomes based on surgeon age and outcomes.

However, specifically for this study, this study was looking at just nonelective and emergent procedures, and as we've seen both these Medicare studies, these are restricted to patients 65 years and older. ***So really this lack of generalizability is paramount because this is sort of contraindicated... not contraindicated, it's... we'll have to delete this part.*** As you can see in conclusion, this data fails to inform the safety of older patients performing the majority of surgeries, which are elective cases and maybe not these nonelective emergent procedures.

The aim of this study was to determine whether postoperative outcomes for patients undergoing common surgical procedures, both emergent and elective, differ by the age of the operating surgeon using a large population-based cohort.

Raj Satkunasivam: Thanks, Zach. Raj here, I'm going to continue from here. I think it's important to really understand the context in the healthcare structure in Ontario from which we derived this cohort for study, and what's important in Ontario is that every single person in Ontario has a unique identifying number referred to as the OHIP number, which allows them access to a single government payer as part of the Ontario Health Insurance Program.

And what this is is very powerful in terms of what we can do with the health services research. Using Ontario data, it allows us to use that unique identifier and link to a variety of different data sources in order to identify important covariates in a study of this nature.

So for example, we're able to pull in demographic information, information regarding relationships to providers that provide surgeon level information. The OHIP database includes claims paid for physician billings and labs that really tell us about the types of procedures that a patient underwent, and then we can also get information on their hospitalization and what happened during the hospitalization based on procedural codes. And lastly, visits outside of hospitals are important to the emergency departments as well so that we can track readmissions.

So when we looked at this broadly and we're looking at the question of surgical procedures, we wanted to define what procedures to study, and it was important that we ensured generalizability when asking this question of relationship between surgeon age and outcomes. And so we picked intentionally procedures that were very common and we picked procedures that would represent a wide variety of specialties when we think about surgery in general.

Furthermore, we wanted to pick procedures that had events and really procedures that would have complications and allow us to really have enough events so we can answer some of these questions robustly. We also wanted to pick procedures and stay focused on procedures that were performed by surgeons and not those performed by interventionalists such as radiologists or cardiologists who perform procedures. And we did this by really seeking out a multidisciplinary consultation across a broad range of specialties and really trying to identify the types of specialties that we would pick.

It's also allowed us to identify not only the types of specialties, which was 10 different specialties, but the procedures that we would identify, and across this time period we identified a total of about 1.2 million patients that were undergoing unique procedures for the first time.

So variables of interest that we identify that were important for this study were whether a procedure was an emergency or an elective procedure, and you'll understand later why this is important. Certain characteristics as related to age of the surgeon, the sex, the years in practice, and volume which we talked about is critical in terms of relationship to outcome. We wanted to know about patient characteristics, which also impact outcomes such as sex, age, comorbidity, socioeconomic status, and rurality. Lastly, we know that this happens in the context of different institutions which are highly variable and we needed to know the institutional identifier in order to adjust for this.

When we came to looking at surgeon age, we knew very well that we need to operationalize surgeon age in a variety of ways. We operationalized it as a continuous variable, and you'll see how we also scaled it for interpretation as well. We also operationalized surgeon age as increments of decades. And lastly, while a little bit crude, we also wanted to look at surgeon age as a dichotomous variable. So 65 is in society considered the standard age for retirement in the general workforce. It has certain societal implications. For example, someone might become eligible for Medicare. But there's some data suggesting that surgeons really inherently might think about this age cutoff as the time to potentially retire. So it made sense to also look at age in this respect.

The outcomes that we wanted to study included death, complications, and readmissions and we took our primary outcome as a combination of any of these occurring within the first 30 days. Secondary outcomes were each of these individually.

I want to just spend a little bit of time here on complications, and really we define this as major morbidity and includes going back to the OR. These were ascertained from codes that are uniformly collected across all hospitals and on all patients in Ontario, and really this has been very well validated in a couple of studies that have been previously published. So this is very sound methodology that has been published before in other health services projects, and these are a couple of examples and really have shown that these are validated methods to identify surgical complications.

So just in terms of the details of our specific methods, and I'll go through this quickly, is really first to analyze the graphical relationship both by a couple of ways of fitting curves between surgeon age and outcome as the dependent variable. We obviously used regression methods and specifically one that accounts for clustering because we're looking at things that there is inherent order, many patients are operated on by the same surgeon, and the same surgeons cluster within hospitals. So we took into account some of these issues. We looked at the length of stay in a different way than the traditional type of estimate equation that's listed above. So we used poisson regression there.

And lastly, we performed a series of sensitivity analyses that are important, stratified analyses according to a variety of physician, patient, and hospital-level characteristics and we had a couple of preplanned analyses that are really important. We know that emergency procedures are important in that patients are not given the ability to select their surgeon and we wanted to control for that as best as possible and look at whether there was heterogeneity of effect by looking at emergent versus elective procedures. And lastly, we performed a couple of post hoc analyses at the end looking at case complexity and era of surgery. I'm going to turn it over to Dr. Wallis now.

Christopher Wallis: Thanks. And so this is the easy job as both Dr. Klaassen and Dr. Satkunasivam have set this up really nicely. Dr. Satkunasivam alluded to the fact that we have 4.2 million patients included in this cohort, and so you can see that among those patients, just over 3,300 surgeons operated, ranging in age between 27 years and 81 years of age at the time of the operation. You can see a graphical distribution of surgeon age here. As you may expect, the majority of surgeons are performing the bulk of surgery in the middle of their careers in their 40s and in their 50s.

And as you might expect, there are some differences between older surgeons and younger surgeons. This is all analyzed at the patient level, but certainly, when we examine older surgeons, classified as those over the age of 65 years, these surgeons were more likely to be high volume and that put them in the highest quartile of surgical volume. They're also more likely to practice in an academic rather than a community hospital. And then there were differences in the specialties that older and younger surgeons practice and so older surgeons were more likely to be neurosurgical and younger surgeons were more likely to practice in general surgery or thoracic surgery.

And now we move on to the primary outcome having discussed those background characteristics. This is [inaudible 00:15:37] fitted curve and so you see a smooth curve with a 95% confidence interval here around the composite primary outcome looking at death, complication, and readmission ranging the spectrum from surgeon age of 29 up to 78, and you can see that I told you the range was 27 to 81 before and we truncated the very ends of this curve because there were so few surgeons and so few operations as to make the estimates a little unreliable at the extremes. But overall I think it's relatively easy to see that there's a linear relationship here between surgeon age and the adjusted rate of these complications, such that patients who have older surgeons are statistically less likely to experience a complication.

We observed this when we modeled the data slightly differently using restricted cubic spline analyses. Here we put five knots in the curve, and we see a similar trend. The effect is larger when you look at the crude estimates, that's the line with the red dots, then the adjusted with the green dots indicating that some of the differences between younger and older surgeons are accounted for on the basis of differences in surgical volume or case complexity due to patient characteristics. But there remains a significant effect even after adjustment. When we look at our secondary outcomes, that's mortality, readmissions, and complications, we see a very similar trend with decreasing rates of these adverse events over time.

And so in our regression analyses, we first operationalized the age continuously, and so what you see is for each 10-year increase in surgeon age, there's a 5% decreased likelihood of the composite outcome. In death this is similar. Readmissions, the effect estimate is somewhat smaller but in the same direction. And complications, again, very similar results. So this was not driven by changes in any one specific component of the composite primary outcome, but rather a consistent effect across each of these aspects of the outcome.

When we dichotomized age using the 65-year cutoff that Dr. Satkunasivam explained, we see absolute differences ranging between 30 per 1,000 patients for the composite endpoint with a relative effect of about 7% likelihood change. This again was driven primarily in this case by the higher event rates in the complications such that the relative effect was stable, but the absolute difference was larger in complication rates whereas 24 events per thousand patients and somewhat smaller absolute differences in readmissions and deaths.

We then moved on to our stratified analyses, and this is looking for effect modification on the basis of surgeon, patient, and hospital factors. And so when we look at surgeon factors overall there's no evidence of statistical heterogeneity in these subgroups, but you can see that interestingly urology along with obstetrics and gynecology are the only two specialties in which there's any indication that patients might actually have lower event rates when treated by a younger surgeon. Interestingly, surgeon sex did not really modify this relationship, and surgeon volume minimally affected it.

We then moved on to hospital and patient factors, while there is a little bit of variability here, in general, we see consistent results without any statistically significant heterogeneity on the basis of any of these factors. We then moved on to sensitivity analyses, and these are important because patient and physician decision making is more likely to influence the relationship and the selection of surgeons in elective cases than in emergent cases. Interestingly we see very stable results in these two subgroups, suggesting that this is not a significant modifier of the relationship between surgeon age and postoperative outcomes.

Then on the basis of some really good suggestions from reviewers as this manuscript wound its way through the peer-review process, we performed two post hoc stratified analyses and the first is looking at case complexity. So this is in terms of high-level questions of case complexity. We were not able to assess the nuances of complexity at the individual case. For example, like is an appendectomy a difficult or an easy appendectomy, but rather we looked at low complexity procedures as those which are commonly performed in a wide variety of settings without significant variability between patients, and high complexity procedures being procedures like Whipple pancreaticoduodenectomy for example where we see the centralization of care and much more complex surgical planning, both intraoperatively as well as preoperatively. We found differences in both of these groups with a slightly bigger effect in the high complexity procedures, although this was not statistically significant.

We then moved to the era of surgery looking to see whether there were changes over time, and we found very comparable results whether patients were treated early in our cohort inception or later on.

There are a variety of hypothesized explanations. Obviously an epidemiologic study of this sort does not give causation or direct ideologic explanation, but there are underlying principles that we may draw on, and the first is self-selection. So obviously surgeons who are still operating in their late 70s are a relatively rare breed and that reflects the fact that many surgeons have already retired. As a result, there is a self-selection process through a combination of self-evaluation, peer review, and/or regulatory based evaluation and those who have declining capabilities are either spontaneously or with some encouragement retiring, such that the remaining cohort is a select group with well-tested and proven skills. This is notable because only 5% of surgeons in our cohort were over the age of 65.

In addition, and perhaps underpinning the good outcomes of those self-selected surgeons, is preoperative case selection. And so we obviously can't assess individual surgeon technical skill or case selection in a study of this sort, but differences in outcomes may be driven by the experience derived from many years of practice, which allows one to optimize patients preoperatively to select those likely to do well.

A few limitations need to be discussed to appropriately put a study of this sort into context, and the first is residual confounding. I've alluded to this a little bit, but the important thing to remember is that we have relatively high-level data that were not designed to be used to answer questions of this sort. So differences on the basis of case complexity or other factors are unable to be captured in a study of this sort. Additionally, we focused on 30-day outcomes, and these may or may not translated into the longer-term outcomes. I'd argue that the 30-day outcomes themselves are of importance even if they don't transfer, but we didn't assess in this study. It may be noteworthy to just mention here that the seminal work by Justin Dimick and Birkmeyer looking at peer-reviewed video assessment of surgeon skill showed much larger differences in short-term outcomes than in long-term outcomes suggesting that surgeon factors are more likely to contribute to those short-term outcomes, whereas the underlying disease process and patient characteristics may drive longer-term outcomes to a greater extent.

Finally, with respect to the procedures captured within this specific study in questions of generalized ability, the Ontario healthcare system, and Canadian healthcare system, in general, is somewhat slower in the adoption of new technologies, especially expensive ones, and so approaches such as robotic-assisted surgery were not widely disseminated in the cohort during the time that we performed this study. However, there's not a strong underlying basis to believe that this would drive differences, except for potential late adopters of these technologies.

To offset potentially some of those limitations we just mentioned, there are a few strengths that are worth particularly mentioning. Number one, is that this is the largest study to date assessing this question and with numbers come strength, but in addition by capturing a large number of patients, we are also able to capture greater variability and that translates into generalizability because we've included a broad range of surgical specialties as well as patients throughout the range of an adult life. This contrast to previous work that Dr. Klaassen alluded to in which Medicare beneficiaries were the study cohort and such, all patients in those studies were over the age of 65. Additionally, we've captured both elective and emergent procedures. Again, this contrasts with prior studies that have provided a subset analysis of these issues and may or may not inform the majority of surgical care that is provided.

Finally, the Ontario healthcare system is a single-payer healthcare system which allows capture of all patients undergoing the procedures of interest and the administrative data derived from a single-payer healthcare system allows us comprehensive identification of all outcomes for these patients, whether they present at the same hospital as their initial surgery or elsewhere.

So finally, just in conclusion and to put the study in a little bit of context, we found that patients undergoing common surgeries in Ontario, Canada, had lower rates of readmission, death, and complications as the age of their surgeon increased with approximately 5% decreased odds of this composite outcome per decade of surgeon age.

Now, the underlying etiology of this relationship is a little bit unclear. However, physician age may be a surrogate or complex series of behaviors that relate to risk-taking, communication, and teamwork. Better understanding this relationship between behaviors such as risk-taking, communication, and teamwork and the outcomes experienced by patients offers an opportunity for all surgeons to improve their care and thus improve outcomes for patients undergoing surgery regardless of the age of the surgeon.

I want to take a moment here just to thank you for your time and for joining us in this UroToday Journal Club. I hope it's been both interesting and informative for you.

As we move to how we apply these data, there are a few important implications. Number one, we argue against a policy of mandatory retirement age given that self-selection processes that we observed here demonstrate that surgeons who are both interested and able to continue to practice beyond traditional retirement ages continue to provide good outcomes for their patients. Aging of the population will both increase the need for surgical care as well as aging of the surgical workforce and meeting these demands requires thoughtful consideration of how we can best make use of the existing surgical workforce. Thank you for your time.