Objective: Nonadherence to the prescribed antibiotic treatment is a widespread problem, affecting not only individual patients, but also public health. As measures to predict it are still lacking, we aimed to design an instrument to identify nonadherence risk in short-term antibiotic therapies.
Methods: Patient adherence was assessed by pill count and an extensive set of standardized questions, newly developed according to the five-dimension adherence model proposed by the World Health Organization. It combined previously validated questionnaires and original questions adapted from our exploratory survey. The extensive set was analyzed prospectively in adult patients treated with antibiotics in outpatient settings. The number of items was reduced stepwise keeping as many of five factor subgroups as possible. The items were dichotomized as positive (indicating nonadherence risk) or negative.
Results: One third (32%) of the respondents (N = 122) did not adhere to the prescribed antibiotic regimen. The extensive set of questions was reduced to a short questionnaire. It consisted of four items independently associated with nonadherence (P < 0.05) and covered three of the five analyzed factor subgroups: healthcare-related factor (consultations about medicines), therapy-related factor (complexity of treatment), and patient-related factors (forgetfulness, health behavior). The sum of 0-1 positive answers indicated an adherent patient, and 2-4 a risky patient. Such a differential approach led to relatively good sensitivity, specificity, and total predictive value (80%, 82%, and 81%, respectively) in our sample.
Conclusion: The short measure representing the multiple factor subgroups was developed to identify nonadherence to antibiotic treatment with acceptable levels of sensitivity, specificity, and predictive values. It could be valuable for healthcare professionals, especially in primary healthcare. Considering limitations of the study (medium sample size, justification of patient's responses, no re-test), future studies could continue investigations on the validity and elaboration of this instrument in larger samples.