Previous studies have analyzed the association between late versus early nephrologist referral (LR, ER) and poor clinical outcomes in patients with end-stage renal disease. We sought to determine whether these poor outcomes were causally related to LR, or whether LR was a proxy for poorer access to health care in general. An inception cohort of incident dialysis patients enrolled in the New Jersey Medicare or Medicaid programs was identified. Using a large number of demographic, clinical, and health care utilization covariates, propensity scores (PS) were then calculated to predict whether a given patient had been seen by a nephrologist at 90 d before first dialysis. Cox proportional hazards models were then built to test the association between timing of nephrologist referral and mortality during the first year of dialysis, using PS adjustment and matching to determine whether this association was confounded by other measures of reduced healthcare utilization. Neither adjustment for PS (HR = 1.31; 95% CI, 1.17 to 1.47) nor matching (HR = 1.40; 95% CI, 1.23 to 1.59) materially changed the initial 36% excess mortality in LR compared with ER patients (HR = 1.36; 95% CI, 1.22 to 1.51). Excess mortality among LR was limited to the first 3 mo of dialysis (HR = 1.75; 95% CI, 1.48 to 2.08) but not present thereafter (HR = 1.03; 95% CI, 0.84 to 1.25). Late nephrologist referral is an independent risk factor for early death on dialysis, even after controlling for other indicators of healthcare utilization. Further research is needed to identify patients at particular risk so that interventions to prevent early deaths on dialysis in LR patients can be developed and tested.
