Prioritizing Retention Efforts using Data Intelligence and Cohort Targeting (PREDICT), Mpumalanga province in South Africa

Partner organizations: Right to Care (RTC), Health Economics and Epidemiology Research Office (HE2RO) and Palindrome.



Study objectives: To evaluate fidelity, feasibility, acceptability, and effectiveness of adherence scorecards and a treatment referral plan on clinic visit attendance of people living with HIV (PLHIV) eligible for and on ART. 



Intervention description: The PREDICT project utilizes machine learning and recipients of care medical record data to build a predictive model that identifies PLHIV who may be at risk of poor treatment outcomes and generates a risk level score. This scoring system has been developed into prototype tools – adherence scorecards – that alerts the clinician to the risk profile of the care recipient and allows them to tailor care.


The intervention includes the following:

  •  Adherence scorecards: Paper-based and digital scorecards that help providers identify recipients of care who are at risk for missing a scheduled visit. 
  • Treatment referral plan: This risk score is then converted to a treatment referral plan that helps recipients of care to plan for their next visit. 


The following behavioural economics principles will be incorporated into the design of the adherence scorecards and treatment referral plan: present bias, choice overload, ostriching, optimism bias and ambiguity aversion.



Evaluation description:

This quasi-experimental design will compare clinic visit attendance and provider usage between intervention arms and the standard of care, as outlined below: 

  • Comparison group: Standard of care
  •  Intervention arm 1: Paper-based adherence scorecard
  • Intervention arm 2: Digital adherence scorecard
  • Intervention arm 3: Paper-based adherence scorecard + referral plan
  • Intervention arm 4: Digital-based adherence scorecard + referral plan



Study outcomes:

  • Fidelity: Quantitative measure of provider usage of the adherence scorecards and treatment referral plan. 
  • Feasibility: Qualitative assessment of provider experience implementing adherence scorecards and treatment referral plans with regards to feasibility. 
  • Acceptability: Qualitative assessment of provider experience implementing adherence scorecards and treatment referral plans with regards to acceptability.
  • Effectiveness: Proportion of care recipients who attended their scheduled clinic visit within 28 days of the scheduled date during the pre and post implementation periods as observed on electronic patient records.



Study site:

Right to Care will enrol four intervention clinics and four comparison (standard of care) clinics in Ehlanzeni district, Mpumalanga Province. Each enrolled intervention clinic will carry out only one of the four approaches. 



Target population:

  • Patients: The target population will be all recipients of care who access HIV care. There will be no age restrictions. No randomization will be done.  
  • Providers: Study providers must currently work as a doctor, nurse, counsellor, or other HIV clinic staff at a selected study facility.  



Sample size:

  • Patients: Between 1,400-2,400 recipients of care will be enrolled over approximately an 8-week period. 
  • Providers: Between 1 to 3 health care providers from each intervention clinic will be enrolled in the study.



Duration: 12 months