Background: Clinical guidelines and policies often fail to achieve high levels of delivery of intended clinical interventions. The difference in what the investigators know works and what is actually delivered at the clinic-level to patients, is known as the "science-to-service gap." In the realm of tuberculosis (TB) prevention, this gap is reflected in <20% of TB preventive therapy (TPT) -eligible persons living with HIV (PWH) being offered or initiated on isoniazid preventive therapy (IPT) in many settings. Recent innovation in TPT have brought new pharmacological options allowing for shorter courses, intermittent dosing, or both.

The overarching goal of this study is to identify a generalizable approach to overcome current barriers to delivery of TPT in order to achieve high levels of TPT delivery during routine care in public clinics. Multiple approaches are in standard use to change prescribing behavior including in service training, audit and feedback, clinical mentoring, the use of clinical decision aids, and "academic detailing." However, the overall change is generally modest. To achieve a substantial increase in TPT delivery (from current approximately 20% to 60-80%) will require a fundamental change in the approach to selecting patients for TPT - a redesign of the choice architecture of TPT prescribing.

Methods: The investigators are proposing a choice architecture that makes prescribing TPT the "default" or standard option and that for TPT not to be prescribed will require a choice by a clinician to "opt-out" of TPT for a specific patient.

The investigators are proposing a cluster randomized design to test the choice architecture approach to increasing delivery of TPT. Clinics will be randomized to one of two strategies: (1) standard implementation and (2) choice architecture default TPT. Because of the clinic-level nature of the implementation strategies, all PWH receiving care at a clinic will be exposed to the standard implementation or TPT routinization implementation. Clinical process data will be used to assess the effectiveness of each strategy to determine the proportion of PWH (1) screened for TPT, (2) eligible for TPT, and (3) prescribed TPT.

Significance: TB is the leading cause of death among PWH in South Africa and elsewhere on the continent. TPT is a proven intervention to reduce mortality among PWH but is not widely prescribed. This study seeks to identify an implementation strategy to reach optimal TPT prescribing.

BACKGROUND

HIV associated mortality and TB disease in LMIC:

TB is the leading cause of death due to an infectious agent worldwide. Among PWH in low and middle income countries (LMIC), TB represents the leading cause of mortality among PWH. South Africa is particularly affected by the syndemic of HIV and TB with an estimated HIV prevalence in 2017 of 7.2 million and TB incidence of 567/100,000 population with over 60% of TB cases among PWH. There were an estimated 78,000 deaths from TB in South Africa in 2017; 56,000 of the deaths among PWH. Findings from the investigators' group suggest this is likely an underestimate of true TB disease and TB related mortality.

TB preventive therapy reduces TB disease and mortality:

For the past 30 years isoniazid based TPT has been known to reduce the risk of TB among PWH. Multiple studies have subsequently confirmed and strengthened this finding. Following cohort studies conducted in South Africa and Brazil by the investigators' team that showed marked reductions in TB incidence for PWH, clinical trials of isoniazid preventive therapy have reported up to a 70% reduction in TB incidence among PWH.

Current delivery of TPT to PWH is anemic:

Despite the evidence and guidelines supporting TPT, current delivery of TPT is anemic. Reports from South Africa (and preliminary data by members of this investigative team) have observed between 14% and 35% of eligible patients are prescribed TPT. Other countries in sub-Saharan Africa have similarly low initiation. In 2017, TPT initiation, as reported in the 2018 World Health Organization (WHO) Global TB Report ranged from 1 to 53% for high burden African countries. South Africa reports higher rates of TPT initiation compared to the rest of the continent, but progress has plateaued and current initiation levels are insufficient for impact.

Current barriers to TPT delivery:

Key factors leading to low TPT initiation are (1) lack of provider confidence in definitively ruling-out active TB, (2) concerns about increasing drug resistant TB, (3) understanding appropriate screening for TB disease, (4) concerns for drug-induced liver injury (DILI), (5) the complexity of prescribing guidelines, (6) misperceptions regarding which patients may benefit, (7) not understanding the benefit of TPT, (8) assumption that patients will not adhere to TPT, and (9) the increased work-load of TPT initiation. All of these factors contribute to the cognitive load of TPT initiation, some also add to the workload. Drug stock-outs in some settings have also posed a barrier to initiation.

Choice architecture and TPT delivery:

Choice architecture is a behavioral economics approach that is used to improve decision making. Choice architecture involves deliberate consideration of how options are presented and what follows from each option, including what happens if no active decision is made (sometimes referred to as passive selection or the "default" setting). Choice architecture can reduce time and complexity required for decision making leading to a reduction in the cognitive load. Lower cognitive load can improve decisions; high cognitive load leads to poorer decisions.

Currently TPT initiation requires an active process to select TPT and a passive (neglecting to consider TPT) process for not initiating. The current active process required for TPT initiation includes at least four steps: (1) consider TPT, (2) evaluate reasons for not initiating, this includes potential concerns for undiagnosed TB and the presence of liver disease, (3) weigh the risks and benefits of TPT, (4) consider patient adherence, and (5) write the TPT prescription. These steps require consideration (cognitive load) and time. Both cognitive load and time compete with other priorities of a complex clinical encounter. As a result, clinicians may "defer" consideration to the "next" visit or turn to rules of thumb to justify deferring or not initiating TPT. Given that the vast majority of ART patients in South Africa would benefit from TPT, the use of choice architecture making TPT the "default" could improve provider decision making and safely increase TPT initiation.

STUDY OBJECTIVES

The overarching goals of this proposal are to test a low-cost and context appropriate approach to translating TPT policy into on-the-ground TPT delivery. The investigators are proposing a cluster-randomized trial (CRT) to measure the effectiveness of a choice architecture-based TPT initiation strategy versus usual prescribing. The investigators anticipate a substantial increase in TPT initiating, potentially increasing from approximately 35% to 60-80% of ART initiators. This can be achieved through choice architecture and a decrease in cognitive load. The investigators will assess the effectiveness, implementation measures, and patient-level delivery and acceptability to providers.

The primary objectives of this study are:

1. To compare the proportion of patients newly starting ART also initiating TPT between the choice architecture and usual prescribing arms

2. To characterize the processes of TPT implementation by study arm including:

   • Provider adoption
   • Fidelity
   • Provider acceptability
   • Intervention maintenance
   • Provider cognitive load of TPT prescribing
   • Clinic workflow integration

3. To describe patient-level characteristics associated with initiating TPT and adhering to TPT

The secondary objectives of this study are:

1. The proportion of patients already on ART initiating TPT
2. The proportion of eligible patients newly starting ART also initiating TPT
3. The proportion of eligible patients already on ART initiating TPT
4. The proportion of patients who initiate TPT and discontinue TPT prior to completion

SIGNIFICANCE

TB is the leading cause of death among people with HIV in South Africa and much of the world. TPT is a proven approach to reduce mortality. This study has the potential of identifying an approach to markedly increase TPT prescribing.

METHODS

Study design:

This is a clinic-level cluster randomized trial of a strategy to increase the delivery of evidence-based care. Clinics will be randomized to the novel choice architecture strategy arm or usual prescribing arm with analysis based on the clinic-level proportion of TPT initiated. This study will further assess implementation outcomes, the underlying theory of change (the effect of choice architecture on cognitive load), provider experiences with the novel strategy, and patient experiences with TPT by study arm.

Study setting:

The research will be conducted by the Perinatal HIV Research Unit (PHRU) in two districts: the Kenneth Kaunda district in the North West Province of South Africa and Mangaung District in the Free State Province. Kenneth Kaunda district has a population of 742,000 and is served by 36 public-sector clinics (primary care clinics and community health centers) providing HIV care (ART, TPT, TB treatment). Mangaung District has a population of 747,431 and is served by 45 public clinics. Both districts have urban, peri-urban, and large rural areas. Both districts are also settings with high HIV and TB prevalence with which the team is very familiar, has a working relationship with the local Department of Health, and has completed prior research on TB and TPT delivery.

The proposed study will be conducted in 36 public clinics that provide antiretroviral therapy (ART). Clinics will be located in rural and peri-urban areas in Kenneth Kaunda and urban and rural areas of Mangaung, especially around Botshabelo. The investigators selected these two areas based on a long-standing working relationship between the team and the Department of Health in these locations and established PHRU research infrastructure in both locations.

Study population:

The study is a cluster-randomized trial with the randomization at the clinic level and the primary outcome at the clinic level. The strategy arm seeks to improve delivery of guideline-mandated services to the clinic population. Thus the strategy clinics may have greater prescribing of TPT and anticipated improved health outcomes, but clinic patients will not perceive any differences in care nor will any individual randomization or patient-level consent occur for the primary outcome of the proportion of patients who receive TPT.

Clinic selection:

Baseline data collection will be completed for a stratified randomization. Key clinic characteristics used for stratification will be measured during a pre-study initiation baseline assessment in the study clinics prior to randomization and will include clinic-level characteristics of patient volume, staffing, and monthly ART initiation.

Prior to randomization, data on TPT prescribing, ART initiation, ART maintenance, and clinic staffing will be obtained for the prior 12 months. These data will be abstracted as monthly aggregates from electric and paper clinic reporting systems. No patient level data or identifiers will be abstracted.

Randomization:

The investigators will conduct a covariate-constrained randomization to balance, if an imbalance is detected, ensuring validity of the randomization process. Implementation will be staggered over time with each of two clinics from the intervention arms matched to control clinics for the purpose of contemporaneous study initiation. Randomization will occur either electronically or through drawing names with Department of Health representatives.

Standard of care arm:

The standard TPT implementation is for a clinician to screen for TB and to consider TPT for those who do not have "presumptive TB". Clinicians in the study district (and most districts in South Africa) have received training and job aids to assist in appropriate application of the TPT initiation algorithm. Prescribing for TPT and ART is done by writing, by hand, the prescription in the patient's paper file. As part of this study, all study clinic providers will have access to standard Department of Health printed material and clinical training.

Implementation arm:

In the choice architecture implementation strategy, all opt-out clinic providers and pharmacists will be trained on the approach. The fundamental tenant of this approach is that TPT will be prescribed with any ART initiation and any ART re-prescribing for 3-12 months of TPT (adherent to current guidelines) if TPT has not been previously prescribed. This will be facilitated by co-prescribing ART and TPT. That is when ART is being prescribed TPT is meant to be prescribed at the same time of the clinic visit.

The simultaneous prescribing will be facilitated through the introduction of an ink stamp or pre-printed sticker to use for quick entry of the ART prescription along with TPT and cotrimoxazole. This stamp/sticker will be available in the consulting rooms and will be pre-stamped/applied by file clerks to files when retrieved. The stamp/sticker for ART prescription, the prescription for TPT and for cotrimoxazole will be "automatically" included. Active canceling of these prescriptions (and indicating the reasons) will be needed to not have TPT dispensed.

Inclusion Criteria:

• Adult (≥18 years old) patients initiating ART
• Adult (≥18 years old) patients already on ART and coming for ART re-prescribing

Exclusion Criteria:

• None

De-identified trial data will be made available one year after completion of all study activities.

One year after completion of study activities

Contact PI

The unit of randomization was the "clinic" - 18 clinics were allocated to the intervention and 18 clinics were allocated to the control arm. We analyzed data from all 50,798 patients who received care (had 1+ HIV care visit) at the each of the 18 clinics during the 12-month implementation period.

From enrollment up to 12 months

All events and all-cause mortality were identified through file abstraction and clinic reporting.

The data identified by clinic allocation will be blinded to the PI and co-investigators and study statistician until completion of comparative analysis. There will be no other blinding.