Background Antiretroviral therapy (ART) has been shown to reduce HIV-related morbidity and mortality amongst those living with HIV and reduce transmission of the virus to those who are yet to be infected. However, these outcomes depend on maximum ART adherence, and HIV programs around the world make efforts to ensure optimal adherence. Predictors of ART non-adherence vary considerably across populations and settings with respect to demographic, psychological, behavioral and economic factors. The objective of this study is to investigate risk factors that predict non-adherence to antiretroviral treatment among HIV-infected individuals in northern Tanzania. Methods At Kilimanjaro Christian Medical Centre (KCMC), a tertiary and referral hospital in northern Tanzania, we used an existing ART database to randomly select HIV-infected patients above 18 years of age who have been on triple ART for at least two years. We used interviewer administered structured questionnaires to cross-sectionally determine predictors of ART non-adherence. We determined non-adherence through retrospective review of pharmacy drug refill (PDR) records of the interviewed participants using a pharmacy database. Results Non-adherence was defined as collecting less than 95% of expected monthly refills in the previous 2 years. Multivariable logistic regression model was used to determine the predictors of non-adherence. Of the 256 patients enrolled mean age was 44 years (SD ± 11) and median CD4 count was 499 cells per microliter (IQR 332-690). Median PDR adherence was 71% (IQR 58%-75%). Non-adherence was associated with younger age and unemployment. Conclusion In this setting, adherence strategies could be adapted to address issues facing young adults, and those with household challenges such as unemployment. Further research is required to better understand the potential roles of these factors in suboptimal adherence.