Date: Tuesday, June 4, 2019
Session Time: 6:00pm-7:00pm
Presentation Time: 6:00pm-7:00pm
Location: Hall C & D
*Purpose: The Scientific Registry of Transplant Recipients (SRTR) fits risk-adjusted models for its semiannual program-specific reports (PSRs). Although SRTR adjusts for as many important risk factors as possible, some may not be available. Confounding could occur if unadjusted risk factors are associated with transplant programs. If data were available for the unadjusted risk factors, their effect on program evaluations could be directly measured; the effect is impossible to quantify without data.
*Methods: The E-Value is the minimum strength of association of a confounder with either the treatment or outcome needed to explain the apparent relationship between treatment and outcome (VanderWeele and Ding, 2017). Regarding PSRs, the E-value for a program’s hazard ratio (HR) is the minimum association of the confounder with either the program or the outcome. The E-value, therefore, provides context for interpreting the HR.
*Results: Figure 1 shows boxplots of E-values for each program’s adult graft survival tier in the October 2018 PSRs. For programs in tier 1, the smallest E-values are >2, so a confounder twice as likely to be found among the program’s recipients and associated with twice the risk of graft failure would not explain the program’s HR. Alternatively, the association between the confounder and the program could be weaker if the HR for the confounder were higher, or the HR for the confounder could be lower if the association between the confounder and the program were stronger, but at least one measure of association must be at least 2. The overall prevalence of the confounder matters. If a hypothesized confounder is sufficiently rare, then even a strong association with the program and a strong effect on graft survival could not provide an alternative explanation for the program’s HR.
*Conclusions: In general, the E-values for programs in tiers 1 and 5 suggest that only a fairly common confounder with a strong effect on outcomes that is also strongly associated with the program could completely explain the program HRs. For programs in tiers 2 and 4, the necessary strength of association is weaker but not trivial. VanderWeele TJ, Ding P. Sensitivity Analysis in Observational Research: Introducing the E-Value. Ann Intern Med. 167:268-274. doi: 10.7326/M16-2607
To cite this abstract in AMA style:Salkowski N, Wey A, Snyder J. Sensitivity of 5-Tier System to Unadjusted Confounding [abstract]. Am J Transplant. 2019; 19 (suppl 3). https://atcmeetingabstracts.com/abstract/sensitivity-of-5-tier-system-to-unadjusted-confounding/. Accessed January 25, 2020.
« Back to 2019 American Transplant Congress