Multi-Dimensional Patients' Dataset for: Development of Online Assessment and Machine Learning Prediction of Medication Non-Adherence with Intervention System

Published: 10 January 2023| Version 1 | DOI: 10.17632/3t39fctrzv.1
Michael Julius,
Fergus Onu,
Kingsley Okorie,
Uzoma Alo,
Clementina Eze,


The prevalence of medication non-adherence among outpatients with chronic illnesses has continued to pose a challenge to the success of clinical treatment outcomes. Assessment of medication non-adherence of existing patients with the prediction of new patients who are unlikely to adhere to their prescribed medication is needed to aid the development of a patient-centered intervention system. We used a cross-sectional survey questionnaire to distribute to 609 outpatients with non-communicable chronic diseases across three (3) tertiary healthcare facilities in the southeast region of Nigeria. The facilities include Alex Ekwueme Federal University Teaching Hospital Abakaliki, Ebonyi State; Federal Medical Centre Umuahia, Abia State; and Federal Medical Centre Owerri, Imo State. The dataset contains data on demographics, medication regimens, use of USSD code, an agent's voice call with SMS to deliver persuasive adherence intervention messages, and reasons for missing medications. Also, generated data on attitude, medication belief, and disease knowledge of patients towards medication non-adherence were by a likert scale of 5-point for categorization. The collated data formed the multi-dimensional datasets. Leveraging patients’ multi-dimensional datasets could advance the course of improving patient’s medication adherence in the following ways: (a) developing predictive, prescriptive, prognostic, descriptive, and diagnostic applications tailored towards solving individual patients’ needs; (b) providing in-depth insight on why patients fail to adhere to their prescribed medications; (c) easy identification of clusters of patients who are likely to be non-adherent across the domain of chronic diseases; (d) improving administrative and operational efficiency (e) supporting clinical decision with patient monitoring and management.


Steps to reproduce

The following steps outlined how the data was gathered and how it can be reproduced especially for statistical analysis and machine learning applications: 1. Questionnaire was developed, validated for reliability consisting of factors influencing medication non-adherence according to World Health Organization and other literature. 2. Applications for ethical approval were written and approval given from three tertiary healthcare facilities in South East, Nigeria. 3. With Approval, informed consent form was distributed and explained to the prospective respondents (patients) along side with the questionnaire. 4. The categorization variables such as belief level, adherence level, behavioural pattern and knowledge level in the form of 5-likert scale can be converted into fuzzy variables. Mean and standard deviation of the values can also be used for machine learning algorithms for prediction, description, prescription based models.


Federal University Ndufu Alike, Ebonyi State University, Evangel University Akaeze


Public Health, Machine Learning Algorithm, Clinical Intervention, Medication Adherence, Mobile Health