Background The standard for selecting unrelated umbilical cord blood units for transplantation for non-malignant diseases relies on antigen-level (lower resolution) HLA typing for HLA-A and HLA-B, and allele-level for HLA-DRB1. We aimed to study the effects of allele-level matching at a higher resolution—HLA-A, HLA-B, HLA-C, and HLA-DRB1, which is the standard used for adult unrelated volunteer donor transplantation for non-malignant diseases—for umbilical cord blood transplantation. Methods We retrospectively studied 1199 paediatric donor-recipient pairs with allele-level HLA matching who received a single unit umbilical cord blood transplantation for non-malignant diseases reported to the Center for International Blood and Marrow Transplant Research or Eurocord and European Group for Blood and Marrow Transplant. Transplantations occurred between Jan 1, 2000, and Dec 31, 2012. The primary outcome was overall survival. The effect of HLA matching on survival was studied using a Cox regression model. Findings Compared with HLA-matched transplantations, mortality was higher with transplantations mismatched at two (hazard ratio [HR] 1·55, 95% CI 1·08–2·21, p=0·018), three (2·04, 1·44–2·89, p=0·0001), and four or more alleles (3·15, 2·16–4·58, p<0·0001). There were no significant differences in mortality between transplantations that were matched and mismatched at one allele (HR 1·18, 95% CI 0·80–1·72, p=0·39). Other factors associated with higher mortality included recipient cytomegalovirus seropositivity (HR 1·40, 95% CI 1·13–1·74, p=0·0020), reduced intensity compared with myeloablative conditioning regimens (HR 1·36, 1·10–1·68, p=0·0041), transplantation of units with total nucleated cell dose of more than 21 × 107 cells per kg compared with 21 × 107 cells per kg or less (HR 1·47, 1·11–1·95, p=0·0076), and transplantations done in 2000–05 compared with those done in 2006–12 (HR 1·64, 1·31–2·04, p<0·0001). The 5-year overall survival adjusted for recipient cytomegalovirus serostatus, conditioning regimen intensity, total nucleated cell dose, and transplantation period was 79% (95% CI 74–85) after HLA matched, 76% (71–81) after one allele mismatched, 70% (65–75) after two alleles mismatched, 62% (57–68) after three alleles mismatched, and 49% (41–57) after four or more alleles mismatched transplantations. Graft failure was the predominant cause of mortality. Interpretation These data support a change from current practice in that selection of unrelated umbilical cord blood units for transplantation for non-malignant diseases should consider allele-level HLA matching at HLA-A, HLA-B, HLA-C, and HLA-DRB1. Funding National Cancer Institute; National Heart, Lung, and Blood Institute; National Institute for Allergy and Infectious Diseases; US Department of Health and Human Services—Health Resources and Services Administration; and US Department of Navy.