Next, we'll have Dr. Catalino speaking to us on axis fractures in older adults. Good afternoon. I'll be presenting on our work with the Gillings School of Public Health and the Comparative Effectiveness Research Team at UNC on axis fractures in older adults. I have no disclosures. So we all know that management of axis fractures is challenging and there's unclear mortality benefit currently from surgery. Hadley et al presented data a couple of years ago in neurosurgery that there was level two evidence that we should consider surgery in these patients. We used propensity methods and standardized mortality to control for covariates in a 20% random sample of Medicare beneficiaries to assess between 2006 and 2013 primary outcomes of mortality. We found that surgery was associated with fewer deaths compared to non-surgical management at one year. So, looking into that, we included all patients with IC9805.02 coding, which is about 10,000 patients with one year continuous enrollment. We excluded patients with a outpatient diagnosis code and patients who had a previous fracture code of the axis within the last year. This included, this brought us to all inpatient diagnoses and that were new diagnoses and not prevalent fractures. So, it was important to distinguish between incident fractures and prevalent fractures because prevalent fractures in these databases can be counted twice and thus the patient characteristics can bias the outcomes because they're contributing multiple outcome measures to the same analysis. So, then we excluded patients with pathologic fracture, coma, severe traumatic brain injury, and combined C12 fractures as well as C2 and subaxial fractures. This eliminated a number of patients so we were able to focus on our code of interest, which was isolated acute traumatic axis fractures and the incident fracture hospital admission was the index admission used to begin data analysis. We used propensity weighting to control for covariates. Propensity weighting says that measured covariate distribution mimicking allows optimal estimation of treatment effect in the treated. This is becoming sort of a gold standard for analyzing large registry databases in a retrospective fashion to create pseudo cohorts that mimic randomization when randomization is unable to be performed. So, propensity weighting is a really robust method for controlling for covariates. It's the odds of estimated probability, in our case, the probability of receiving surgery or not receiving surgery. PSI is the estimated probability of receiving surgery based on the measured covariates for any individual patient enrollee. Then we use the odds, which is calculated as you can see there, and then the SMR weights, which are the standardized mortality weights that are weighed, this non-surgical cohort is weighed against the surgical cohort, which is unweighted in order to make a comparison that limits confounding from covariables. The measured covariates are listed here. Key ones are the Charleston Comorbidity Index, frailty score, the mechanism of injury, baseline health utilization, healthcare utilization, and also a number of medications. Almost every single factor here has been shown in literature to predict mortality in older patients. So, that's why we included them. We compared surgical and non-surgical cohorts and the weighted standardized difference between the two cohorts was less than 0.1, which means that our model was well-balanced across these covariates. So, results, we found that, as expected, the incidence of fracture over time was approximately stable. The rate of fracture in older adults was higher. The rate in 85 and older was significantly higher than those 75-84 and those 65-74 shown in the graph. The mortality, again, as expected, over a six-year follow-up, we're able to identify patients who were the highest risk of mortality, was associated with increasing age, and females and males both showed that same trend. So, the cohort description, we have about 10% of the patients in the cohort did receive surgery, 90% did not receive surgery. The CPT codes are listed there. Those are the most common codes and represented almost 99% of the patients. The CPT coding for the non-surgical cohort did include halo use in only about 3% of patients. So, our results, this table on the left, you see, this is one-year mortality representing the number of patients who died at one year per 100 patients. On the left, you see a stratified by age. We're comparing surgery versus no surgery and then risk reduction, which is presented essentially as a risk difference between the non-surgical patients and the surgical patients after adjusting for the covariates using our model. So, first, we see that in surgical patients compared to non-surgical patients overall, there was an association with improved survival. It was about seven fewer deaths per 100 patients at one year compared to the non-surgical group. Taking the age stratification into consideration, you look at the 65 to 74-year-old group, the difference was more pronounced. So, we see 7.7 patients per 100 died at one year in the surgical group and almost more than double died in the non-surgical group. After waiting and then the risk reduction was 11.2 with a 95% confidence interval presented. This was a statistically significant result. And then if you look at age stratification along the different age groups, we see that the benefit or the association with lower mortality decreases with age and that the 75 to 84 and 85 plus group, actually the risk difference or risk reduction actually crosses into negative values which means it crosses zero and actually means that there's, we're not confident that surgery doesn't harm patients. So, because it actually crosses over into almost a harm scenario. So, our conclusions, we know that axis fractures are associated with high mortality in older adults in the U.S. This is comparable to hip fractures in literature. Surgical treatment of patients age 65 to 74 was associated with significantly lower mortality than non-surgical treatment. We know that there's residual confounding because of unmeasured covariates just because this is not a randomized control trial. But our methods we feel are very robust in adjusting for confounders and we're confident in our results and the limitation or the limiting of confounding variables. Then the bottom line here for our takeaway is not necessarily that a causal relationship between surgery and better outcome. But the fact is that based on a retrospective review of this data, we can say that as surgeons over the past 10 years we've been better at determining which patients will benefit from surgery in the younger cohort of patients compared to the older patients and that's revealed by our analysis and we're very confident in that conclusion. Determining causality based on this data, we cannot make a statement on that. It's only association. But we do know retrospectively we've done a very good job in over the last 10 years choosing which patients will truly benefit from surgery. That's it, any questions? Thank you.

axis fractures

older adults

mortality outcomes

surgical management

non-surgical management