In patients undergoing stem cell transplantation (SCT), nonadherence has potential for significant medical impact and potentially life-threatening complications. No study thus far has demonstrated an effective way to predict adherence in SCT recipients. A structured rating scale, the Stanford Integrated Psychosocial Assessment for Transplantation (SIPAT), has been shown to predict psychosocial outcomes and medical morbidity in solid organ transplant recipients. We assessed the SIPAT in SCT recipients. We hypothesized that the SIPAT rating would be associated with nonadherence to the post-SCT regimen. We retrospectively studied SCT recipients who had psychiatric evaluations with the SIPAT before SCT. The primary outcome was nonadherence, defined a priori as at least 1 life-threatening nonadherence event in the first 6 months post-transplant. Association of the SIPAT with outcomes was evaluated by logistic regression, and an optimal cutoff score was determined using a receiver operating characteristic curve. Of 85 patients (mean age 47 years; range, 18 to 74 years), 56 (66%) were male, and 43 (50.5%) received autologous SCT. Eighteen (21%) patients were nonadherent. The SIPAT rating, treated as a continuous variable and controlling for autologous versus allogeneic SCT, was significantly associated with nonadherence (per 1 point; odds ratio [OR], 1.162; P< .0001). Allogeneic SCT also conferred a significantly increased risk of nonadherence (OR, 14.184; P= .005). Multivariate analysis stratifying for allogeneic versus autologous transplantation and controlling for age, sex, and disease confirmed an independent association between the SIPAT score and nonadherence. A cutoff score of 18 provided optimal specificity (89.6%) and sensitivity (55.6%) for nonadherence. Nonadherence rates were 58.8% and 11.8% for subjects with SIPAT ratings of 18 and above or 17 and below, respectively (relative risk = 4.98, P < .0001). Psychosocial risk as quantified by the SIPAT correlated with SCT recipients' adherence to the post-transplant regimen, suggesting that this instrument can contribute to medical risk stratification models. Further study should evaluate long-term mortality data and the effects of intervention on psychosocial risks.