Background. Patient adherence to anti-tuberculosis therapy is a key issue in achieving effective treatment outcomes. Evaluation of treatment adherence is limited by regulatory documents to the term "chemotherapy interruption," which is characterized by failure to take medications for 2 months. The role of short interruptions in anti-tuberculosis drug intake and their contribution to the effectiveness of tuberculosis treatment remain unclear. The aim of the study was to develop and validate a new methodological approach to assessing treatment adherence in tuberculosis patients and to identify factors influencing patient adherence to predict its low level. Material and methods. A retrospective cohort study was conducted, including 332 patients with drug-resistant tuberculosis who started treatment in a tuberculosis hospital. Patients were divided into groups according to the number of missed doses of drugs during the intensive phase of treatment: Group 1 (n=86) - patients with 11% or more missed doses of drugs, Group 2 (n=246) - patients who missed 10% or less of drug doses. Study results. The identified groups were compared across 47 social, clinical, and organizational characteristics to identify risk factors for the development of a "critical level of adherence" to anti-tuberculosis therapy in patients. Based on the identified risk factors, a computer program, "Predicting the Level of Adherence in Patients with Tuberculosis," was developed. Conclusions. The outcome of chemotherapy depended on periodic missed doses of more than 11% of doses (OR = 4.8) and more than 20% of doses (OR = 18.4). Risk factors for low adherence included male gender (OR=2.2), low level of education (OR=2.2), imprisonment (OR=3.0), dependence on psychoactive substances (OR=7.7) and alcohol (OR=5.5), disseminated tuberculosis (OR=2.2), HIV infection (OR=5.1), no history of chemoprophylaxis (OR=3.3), decompensation of concomitant pathology (OR=2.5), satisfactory patient condition (OR=1.8), premature discharge (OR=2.6), adverse drug reactions (OR=2.2), referral to a narcologist two or more times during hospitalization (OR=6.7). A program for predicting the risk of "critical adherence" to treatment was developed, with a sensitivity and specificity of 86.1% and 81.7% for patients with HIV infection, and 84% and 81.7% for patients with unknown HIV status.