We performed a retrospective cohort study using data from the United States Renal Data System (USRDS), which is the national registry of all patients treated with dialysis or kidney transplantation in the US. Children with KF aged 0 to 17 years old who started kidney replacement therapy (KRT) from January 1, 2000, to December 31, 2019, were included in the study. Children with missing covariates and living in US territories were excluded.

Demographic characteristics were extracted from Patients file and the Centers for Medicare and Medicaid Services Medical Evidence 2728 (MEDEVID) form at the time of KF onset. Race/ethnicity was based on provider attestation in the Patients file. Race/ethnicity was categorized as Hispanic, Black, non-Hispanic White, and Other.

This study was reviewed by the University of California San Francisco Institutional Review Board and considered to be exempt human subjects research.

Rurality of the patient residence was determined using the rural–urban commuting area (RUCA) codes as defined by the United States Department of Agriculture. Zip codes of residence were matched to RUCA codes which ranged from 1.0 (most urban) to 10.3 (most rural) based on population size and commuting flow [13]. We categorized each patient’s residence at the start of KRT as metropolitan (1.0–3.9, corresponding to urbanized areas with ≥ 50,000 population); micropolitan (4.0–6.0, corresponding to urban clusters of 10,000–49,999 population); or small town/rural areas (7.0–10.3, towns with population of lower than 10,000 inhabitants and outside urban areas and urban clusters) [14].

The primary outcome was time to kidney transplantation starting from the date of dialysis initiation. If patients received preemptive transplantation, time to kidney transplantation was set at 0.5 days. We restricted our analyses to only the first kidney transplant event. We then examined outcomes separately by whether the donor source was living or deceased.

Our secondary outcome was time to waitlist registration starting from the date of dialysis initiation. If a patient was preemptively waitlisted, time to waitlist registration was set at 0.5 days.

The association between rurality of the patient residence and time to transplantation or waitlisting was examined using separate Fine–Gray subhazard models for each outcome and treating death as a competing risk. Patients were censored administratively on December 31, 2019. The model was adjusted for age, sex, race/ethnicity, primary cause of KF, region of the US (Northeast, South, Midwest, and West), calendar year of onset of KF (grouped in 5-year categories) health insurance status (Medicare/Medicaid, private, or none), and income as the neighborhood median income by zip code of patient’s residence [15]. We did not adjust for comorbidities as the prevalence of comorbidities (e.g., heart failure) in children is low [16]. The proportional hazards assumption was tested with Schoenfeld residuals and log–log plot.

The association between rurality of patient residence and time to deceased donor kidney transplantation was examined using Fine–Gray models treating death and living donor kidney transplantation as competing risks. When living donor kidney transplantation was considered the outcome of interest, death and deceased donor transplantation were considered competing risks.

We assessed for interactions between rurality of patient residence and race/ethnicity, neighborhood median income, calendar year, and region of US as defined a priori. Interactions were considered statistically significant if the p value was < 0.05. All analyses were conducted using Stata 17 (StataCorp, College Station, TX).