Mortality Among Solid Organ Waitlist Candidates During COVID-19 in the United States

Jonathan Miller; Andrew Wey; Donald Musgrove; Yoon Son Ahn; Allyson Hart; Bertram L. Kasiske; Ryutaro Hirose; Ajay K. Israni; Jon J. Snyder


American Journal of Transplantation. 2021;21(6):2262-2268. 

In This Article


Population and Data

This study used SRTR data, which includes data on all donors, waitlisted candidates, and transplant recipients in the United States, submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The Health Resources and Services Administration, US Department of Health and Human Services, provides oversight of the activities of the OPTN and SRTR contractors.

This analysis used SRTR Standard Analysis File (SAF) candidate datasets from August 2020, which represent all patients in the United States who are, or have been, registered on the waiting list for a solid organ transplant since October 1, 1987. The SRTR SAFs have been described in great detail previously.[6,7] Because recording at least 95% of patient deaths on the waiting list can take 2 or more months due to a lag in reporting, candidates were included if they were prevalent on the organ transplant waiting list between March 13, 2019, and May 31, 2020. Therefore, this study presents early findings from the first 10 weeks after the declaration of the COVID-19 national emergency. Candidates with no reported listing date or age at listing or who were younger than 18 years at listing were excluded. Candidate follow-up was censored at transplant, recovery without a transplant, or transfer to another center. The analyses were performed separately for kidney, pancreas, liver, lung, and heart candidates. Waiting time and outcomes were included for patients regardless of whether they were listed as active or inactive.


Outcomes. The primary outcome was waitlist mortality, specifically the cause-specific hazard of waitlist mortality, which does not mathematically depend on the transplant rate.[8] Thus, any differences in the cause-specific hazard of waitlist mortality before and after COVID-19 are not inherently attributable to lower transplant rates after COVID-19.

Predictors and Covariates. The main predictor of interest was the COVID-19 pandemic, as defined by time before or after the declaration of a national emergency in the United States on March 13, 2020. These analyses track temporal trends in mortality before and after COVID-19, because individual-level incidence status and cause of death are not available in the SRTR SAF.

Covariates modeled for all solid organ types were age in years, sex, ethnicity, race, urban or rural residence, miles between candidate and program, blood type, body mass index (BMI), primary diagnosis, insurance type, previous transplants, and waiting time. In addition, these covariates were modeled for these types of transplants:

  • Kidney: Calculated panel reactive antibodies (cPRA), dialysis duration, and whether the patient was also listed for pancreas transplant

  • Pancreas: Listing for pancreas-only, pancreas-after-kidney, or simultaneous kidney–pancreas transplant

  • Liver: End-stage liver disease (MELD) scores and hepatocellular carcinoma (HCC) status

  • Heart and lung: Height at listing

  • Lung: Lung allocation score (LAS)

  • Heart candidates: Ventricular assist device (VAD) status at listing

Time-varying Covariates. Candidate characteristics with time-varying values were updated at the beginning of each month before and after March 13, 2020. For example, the LAS constantly changes as patients become more or less sick. Thus, a patient's LAS value at the beginning of each month was the value used for analyses for that entire month. The last available value was used for follow-up after removal from the waiting list.

Statistical Analysis

Modeling Framework. Given the interest in the time-varying effect of COVID-19 on waitlist outcomes, piecewise exponential models (PEMs) were used to estimate the rate of waitlist mortality after the COVID-19 declaration on March 13, 2020, versus before. PEMs are proportional hazards models with a constant baseline hazard in a priori defined intervals. The models included two intervals for the baseline hazard: before and after COVID-19. The time scale for these models was calendar time. To ensure sufficient precision, each analysis required a minimum number of events after March 13, 2020, detailed below in each subsection.

Overall Effect of COVID-19. The overall effect of COVID-19 was the difference between the intervals before and after March 13, 2020. For this analysis, the models assumed each covariate had the same effect before and after COVID-19. These models were estimated only when the cohort had more than 10 deaths both before and after COVID-19. A post hoc sensitivity analysis additionally adjusted the models for patient inactive status as a time-varying covariate.

To assess whether early trends in waitlist mortality rates attenuated in later months, a post hoc preliminary analysis using the December 2020 SAF modeled time trends in waitlist mortality hazard by month before and after COVID-19 using a PEM with a random effects for each month and adjusted for covariates.

Geographic Variability in the Effect of COVID-19. Generalized linear mixed models (GLMMs) estimated the DSA-level variability in waitlist mortality rates before and after COVID-19. Specifically, the model included two DSA-level random effects: one for before March 13, 2020, and one for after. The empirical Bayes estimates of the individual DSAs estimated difference of each DSA from the national average before and after COVID-19. Therefore, the kidney model included, for example, 58 DSA-specific pre-COVID effects and 58 DSA specific post-COVID effects. The GLMMs allowed a correlation between the random effects. The difference between the pre- and post-COVID effects identified the relative difference in waitlist mortality rates after, compared with before, COVID-19. The GLMMs included an offset equal to the linear predictors from the PEMs for the overall effect of COVID-19, which accounted for candidate risk factors. These models were estimated only when the number of deaths in the post-COVID-19 period was more than twice the number of DSAs.

Subgroup-specific Effects for COVID-19. Separate models estimated the subgroup-specific effects of COVID-19. Specifically, the model included an interaction between each candidate risk factor and the overall effect of COVID-19, allowing the effect of COVID-19 to differ across, for example, candidate age groups. Models were only estimated for each of the covariates listed above when deaths in the post-COVID-19 period were at least 10 plus 2 times the number of variables in the model.