Choice architecture-based prescribing tool for TB preventive therapy: a pilot study in South Africa

Abstract

Background:

All people with HIV who screen negative for active tuberculosis (TB) should receive isoniazid preventive therapy (IPT). IPT implementation remains substantially below the 90% WHO target. This study sought to further understanding of IPT prescription by piloting a simplified prescribing approach.

Setting:

Primary care clinics in Matlosana, South Africa.

Design:

This was a mixed-methods implementation study.

Methods:

Nine providers were recruited and underwent training on 2018 WHO guidelines. A simplified prescribing tool containing antiretroviral therapy (ART) and IPT prescriptions was introduced into the workflow for 2 weeks. Prescription data were collected from file review. Interviews were conducted with prescribers.

Results:

During the study period, 41 patients were evaluated for ART initiation; 34 (83%) files used the simplified prescribing tool. Thirty-seven (90%) patients were eligible for same-day ART and IPT initiation, of whom 36 (97%) received IPT prescription. Qualitative interviews identified the following barriers to IPT prescription: cognitive burden, extensive documentation, limited management support, paucity of training, stock-outs, and patient-related factors. Provider acceptability of the tool was favorable, with unanimous recommendation to colleagues on the basis of streamlining documentation and reminding to prescribe.

Conclusions:

This simplified prescribing device for IPT was feasible to implement. Streamlining documentation and reminding providers to prescribe can reduce work-flow barriers to IPT provision.

TB remains a leading cause of morbidity and mortality in South Africa, with 73 per 100,000 HIV-related deaths attributed to TB.¹Isoniazid preventive therapy (IPT) reduces the risk of active TB in people with HIV (PWH) by as much as 62%.² South African national guidelines specify that PWH without active TB should be provided IPT.^3,4 Economic analyses have shown that IPT in PWH is cost-effective: for every dollar spent on TB prophylaxis, four dollars are saved in subsequent medical care.⁵ As part of the Progress towards End TB Strategy, the UN has set a milestone of providing IPT to at least 30 million people worldwide by 2022, a task that is currently at less than 10% of target.⁶ IPT provision among newly diagnosed PWH in South Africa is presently estimated at 65%, below the WHO and South Africa Department of Health targets of ⩾90%.⁶

Qualitative research has identified barriers to IPT integration into HIV care in sub-Saharan Africa, including provider factors such as insufficient knowledge, as well as burden of time and documentation,^7,8 operational barriers both perceived (drug resistance, low screening specificity) and actual (referral availability, long wait times, staff shortage),^9,10 stigma regarding TB,¹¹ and patient factors such as understanding, compliance, socio-economic access to care, and pill burden.^12–14 Facilitators of implementation include understanding of TB and IPT, diagnosis acceptance, social support, and trust in the healthcare system.¹⁵ Multiple barriers to effective delivery must be overcome if IPT goals are to be met.

Behavioral economics is a field that seeks to explain human behavior through the lens of cognitive biases and to harness this understanding to affect behavior. This approach has gained traction as a way to improve service uptake or delivery, such as increasing medication adherence and reducing unnecessary laboratory testing.^16–18 Choice architecture is one such approach that makes use of two biases to decision-making —availability bias and cognitive load—to guide a decision maker into selecting the generally preferred choice. Applying choice architecture in low-resource clinical settings could prove a powerful tool in mitigating the multifaceted barriers to delivery of care.

In this study, a simplified prescribing approach was designed based on principles of choice architecture in order to reduce cognitive load and documentation effort faced by providers. Among the barriers to IPT provision, we hypothesized that these were key barriers. Appropriate identification and management of the key barriers can unlock substantial improvements, even in complex systems.¹⁹ This simplified prescribing approach aims to facilitate clinician decision-making by streamlining a decision-intensive process in order to increase IPT prescribing. We piloted this approach among providers prescribing IPT for newly diagnosed PWH and followed up the intervention with in-depth interviews to gain a nuanced understanding of the prescription process.

METHODS

Design and setting

This prospective, mixed-methods, proof-of-concept pilot study was conducted in Matlosana Municipality, North West Province, South Africa. The site was selected based on longstanding relationships between the investigators and regional clinics. The target population of ART prescribers and their patients receiving ART was conveniently sampled from three primary care clinics. Two of the clinics were located in the community, while one was affiliated with and adjacent to the main regional tertiary care hospital. Every clinic was staffed with 5–15 providers, each with an average daily volume of 60–80 patients. In all study clinics, providers had a supply of the most commonly prescribed medications in the consulting room for direct dispensing to patients. Providers were recruited by study team members through in-person clinic visits. Inclusion criterion for providers was a minimum experience of 3 months in ART prescribing. Selected providers were subsequently invited to participate in semi-structured interviews.

Intervention

Study team members issued participating providers a brief training on the 2018 WHO IPT guidelines and a tool to streamline prescribing.³ The tool was an ink stamp and set of stickers for application in the HIV care book to link ART and IPT prescribing (Figure 1). Participating providers were asked to use their choice of stamp or sticker in the file, in addition to regular documentation for all their patients initiating ART during a 2-week period. Prescribers were advised to place the tool in a location on the clinical forms where a handwritten prescription would have been made. The purpose of the tool was to guide decision-making and to normalize combined prescribing of IPT and ART at initiation, specify the few indications not to use IPT, and reduce handwritten documentation.

FIGURE 1.

Prescribing ink-stamp/sticker. FDC = fixed-dose combination; TDF = tenofovir; 3TC = lamivudine; EFV = efavirenz; PO = per os; FTC = emtricitabine; TB = tuberculosis; WHO = World Health Organization; INH = isoniazid.

Data collection

To assess current practice and the potential value in approaches to increase IPT prescribing, the local electronic patient management system for HIV care (tier.net) was reviewed at participating clinics to abstract prescribing of IPT and ART during the 6 months preceding the study (January to June 2019).

During the study period, records of all patients initiating ART at the three clinics were reviewed to abstract demographics, prescriptions, and use of the prescribing tool. Local tier.net databases were queried for new ART initiations to ensure complete file capture.

Semi-structured hour-long interviews were conducted with participating providers using a study specific guide that aimed to assess provider experiences with IPT prescription and responses to the intervention. Interviews were conducted in English at the study sites before or after the clinic day. Of specific interest were provider-related challenges that influence IPT prescription, such as management, training, supply, and cognitive burden.

Data analysis

Quantitative analysis was performed using Stata v14.0 (StataCorp, College Station, TX, USA). Results were expressed as proportions or medians with interquartile ranges. Qualitative data were audio-recorded, transcribed verbatim, de-identified, and processed using Atlas.ti v8.4 (Technical University of Berlin, Berlin, Germany). Inductive thematic analysis was applied to the interview transcripts, which were coded by two independent analysts working in parallel. Developed codes were matched to ensure consistency and integrity, after which a codebook was co-developed. From the codebook, broad themes and subthemes were identified and reviewed to ensure appropriateness for interpretation.

Ethics

International good clinical practice and local human subject research guidelines were adopted; approvals were granted by ethics review boards of the University of the Witwatersrand Human Research Ethics Committee (Johannesburg, South Africa), North West Department of Health (Mahikeng, South Africa), Kenneth Kaunda Department of Health (Klerksdorp, South Africa), and the Johns Hopkins University School of Medicine (Baltimore, MD, USA). Written informed consent was provided by all participants.

RESULTS

IPT prescribing

During the 6 months prior to implementation, IPT was prescribed for 68% (522/773) of new ART initiations across the three study clinics. During the 2-week study period, the nine participating providers (three per clinic) evaluated 41 newly diagnosed PWH for ART initiation. Demographic characteristics of participating providers and their patients are presented in Table 1. Of the 41 newly diagnosed PWH, 37 (90%) were eligible for IPT and 36 (88%) were prescribed IPT. Providers used the tool for 83% (34/41). The four ineligible patients were either being investigated or treated for active TB disease.

TABLE 1.

Characteristics of participating providers, demographics of patients seen during the 2-week study period, and IPT prescriptions issued during the intervention

	n (%)
Providers	(n = 9)
Female	6 (67)
Age, years, median [IQR]	42 [34–45]
Occupation
Professional nurse	8 (89)
Clinical officer	1 (11)
At least 4 years’ experience in HIV and TB care	8 (89)
Patients	(n = 41)
Female	25 (61)
Age, years, median [IQR]	33 [28–44]
Age groups, years
15–24	3 (7)
25–34	20 (49)
35–44	9 (22)
45–55	9 (22)
Prescription outcomes
Tool used	34 (83)
Eligible for IPT	37 (90)
Prescribed IPT	36 (88)

IPT = isoniazid preventive therapy; IQR = interquartile range; HIV = human immuno-deficiency virus; TB = tuberculosis.

Perceived barriers to IPT provision

Provider perceptions on the challenges of IPT provision as well as the impact of the simplified prescribing intervention are shown in Figure 2. Detailed thematic analysis with relevant quotes on IPT prescribing and assessment of the intervention can be found in Tables 2 and 3, respectively.

FIGURE 2.

Conceptual framework of barriers to IPT provision and effects of simplified prescribing tool. ART = antiretroviral therapy; TPT = tuberculosis preventive therapy; IPT = isoniazid preventive therapy.

TABLE 2.

Major themes and subthemes related to the challenges associated with IPT

Theme	Sample quote
Work load
Patient burden  Redundant tasks	You see maybe when you get to your 67th patient today you’re drained Sometimes I feel like there’s no need [to document prescription], it’s already there why should I have to repeat it again. We have so many patients, other patients to see …reduce the documentation. I think it will also reduce the workload, maybe then we will [prescribe IPT] without thinking of a lot of documents to write up
Burnout	…you know honestly, sometimes it [burnout] affects even my punctuality here at work. Sometimes I feel like I can just take days off…
Mental errors	We are working very hard, it’s too much that we end up making mistakes, not deliberately
Management
Policy dissemination	[Policy changes relating to IPT] comes from the national department of health…I don’t know why [the national department of health] gives information first to both [Aurum] and/or our top management. Then our top management delays the information…I think maybe the problem is there
Guideline adherence	Management is not around most of the time to enforce guidelines…managers are always out on meetings. We have to make time to know the guidelines because if we don’t, guidelines will never be followed
Support	I must say I’ve considered working somewhere else that would provide me with everything I need to work. As it stands now, almost everything I do is a struggle I don’t get support from my facility to give IPT and as a result don’t give [IPT]
Stockouts
Provider morale	…now it’s you [providers] who are disappointing us, we want to take medication but we come here and you tell us there’s no medication Shortage...kills practitioner morale now
Treatment interruption	[with] stock-outs you won’t be able to monitor, because those patients have to be on for 12 months. So you skip, then it brings a lot of confusion
Patient lost to follow-up	[if] they come for follow-ups [and] there’s no IPT you end up interrupting the process and maybe you end up losing the patient and they don’t finish their IPT

IPT = isoniazid preventive therapy.

TABLE 3.

Major themes and subthemes related to assessment of the simplified prescribing intervention

Theme	Sample quote
Tool design
Usability	It was simple and straightforward It’s clear
Preference	I think the stamp needs to be modified. On some paper it smudges, but it’s ok…I think for cost effectiveness the stickers. If we were to continue with the stamps I think I would prefer the stickers for cost effectiveness. Because stamps the ink is expensive and stuff I will say the stamp is more easy to use than the sticker. Because that sticker is too big, takes much of the space
Proposed changes	…write the lot number, that’s the only thing that could be missing. They want the lot number as well as the quantity and all those. And if we can identify all the stickers a little bit smaller And you should also have with the start date and end date
Documentation
Tracking patients(rename)	...[helps] monitor the patient and track the time when the patient was initiated It’s going to be simple for you to see if the patient was initiated on IPT or not, and because the date is there at least it will be able to track this patient. When the patient was initiated and when [IPT prescription] should stop
Workload	…it reduces the writing that one has to do
Scalability
Potential target	With a new colleague who doesn’t know guidelines it’s a must they must use it
Workflow impact
Memory aid	I think it’s [stamp/sticker] going to help us do justice to the patients… Sometimes you are so tired that you forget important things… We are working very hard, it’s too much that we end up making mistakes, not deliberately…So some of the things like this [stamp/sticker] really they [help] you…remember ‘oh I must do this[prescribe IPT]
Guidelines	So when a sister in maternity sees I have to give IPT here maybe they’ve forgotten how to give IPT. But it gives you guidelines

Work load

All providers expressed that demands of clinical work created cognitive burdens due to the combination of a high patient volume and task complexity and redundancy. Clinical complexity was seen as a particular barrier to IPT prescribing due to the perception among some providers that additional evaluation was required to determine patient eligibility for IPT. Providers further described that documentation of IPT in the patient file added to the workload. Providers stated that occasionally they made unintentional errors because their minds were preoccupied with the extensive needs of their patients.

…remember if this client is on ART, now you are going to start the IPT and it’s still a lot of medications, unlike the TB. The TB is now – our client is complicated that way. Now, is there [rifampin], is there cotrimoxazole, is there pyridoxine, is there vitamin B, now those are complications. (Interview 6)

But then comes the problem this [chart], with this one there is no reminder at all… [it doesn’t] remind you of any drugs. Meaning you must remember yourself the drugs that you must give and then prescribe them with your head. (Interview 6)

Stock-outs

All providers noted that supply shortages of IPT interfered with effective implementation of IPT due to a concern that medication may not be on hand during all 6 months of therapy. The concern for stock-outs, even if rare, added to the cognitive load of considering how to manage the possibility.

…[if] they come for follow-ups [and] there’s no IPT you end up interrupting the process and maybe you end up losing the patient and they don’t finish their IPT. (Interview 5)

Prescription guidelines

Six of eight providers identified delayed or insufficient communication on guideline updates, especially between providers and management, as a barrier to prescribing. Changes in guidelines in recent years, such as consideration of tuberculin skin test and varying prescription durations, can lead to uncertainty in prescribing and subsequent prioritization of ART over associated management.

Management is not around most of the time to enforce guidelines…managers are always out on meetings. We have to make time to know the guidelines because if we don’t, guidelines will never be followed. (Interview 3)

Intervention impact

Providers unanimously expressed satisfaction with the tool and intention to recommend it to colleagues on the basis of reduced cognitive load and decreased burden of documentation. They often cited additional benefits such as clarification of IPT start dates and informing future re-evaluation via documentation of reasons for lack of IPT at initial visit.

Cognitive load

All providers reported that the tool served as a memory aid, assisted in overcoming cognitive load, and reinforced the guidelines, serving to enhance the implementation of IPT.

I think it’s going to help us do justice to the patients… Sometimes you are so tired that you forget important things… We are working very hard, it’s too much that we end up making mistakes, not deliberately…So some of the things like this [stamp/sticker] really they [help] you…remember ‘oh I must do this [prescribe IPT]. (Interview 2)

Documentation

According to five of eight providers, time spent documenting medications and contraindications was reduced with the tool. There was also a prevailing belief that this documentation would better enable providers to follow the patient over the course of treatment.

It’s going to be simple for you to see if the patient was initiated on IPT or not, and because the date is there at least it will be able to track this patient. When the patient was initiated and when [IPT prescription] should stop. (Interview 5)

DISCUSSION

This prospective, mixed-methods, proof-of-concept study demonstrated the feasibility and acceptability of guiding clinician decision-making for IPT prescribing. During the time period preceding the use of the tool, IPT prescribing at participating clinics was lower than national targets (68%), which indicates the importance of finding ways to improve prescribing. The choice architecture tool was used during the observation period for 83% of new ART initiations. Interviews with providers suggested that the tool worked as planned by simplifying prescribing, guiding prescribing choice, and linking IPT prescribing to the priority activity in the clinic of ART management. This is the principle of choice architecture: to make the desired choice the easiest choice.

Our findings suggest acceptability and, potentially, an advantage over the prior system. The stamp was used for all newly diagnosed HIV patients in two clinics, while another clinic had 70% usage. The clinic with the lower tool usage had a higher daily rotation of providers and more frequently absent management, which may have contributed to lower use.

Our findings of barriers and facilitators to IPT prescribing help confirm results of prior studies. This was anticipated and desired, as we sought to learn from prior knowledge in developing our implementation strategy. Previously reported barriers such as inadequate training, long visit time, staff shortages, and stock-outs continue to be challenges to IPT provision in this setting.^8–10,12,14

Providers most often attributed their acceptance of the tool to its usefulness in generating prescription reminders and in streamlining the documentation process. Like a standardized checklist, the tool centralizes all HIV medications, including IPT, prompting providers to review all possibly indicated prescriptions in the complex care of PWH. Choice architecture forms the conceptual basis of this approach.^20,21 It has been applied effectively in other clinical and public health settings, such as to increase influenza vaccination, improve opioid prescribing, and increase consent to deceased donor organ donation.^22–25 In this pilot study, reducing the cost of time and effort in the prescription process encouraged providers to prescribe IPT.

Limitations and strengths

This study implemented a prescribing tool specifically designed to reduce clinician cognitive load and make the usually preferred option the routine choice during routine care in South African public sector clinics. As such, the study is novel in its theoretical basis, but not without limitations. The intervention period was brief, and consequently, the number of patients for whom providers could use the tool was modest. In addition, the study was not designed to test whether use of the tool changed practice or increased IPT prescribing. We believe the tool may have contributed to increased prescribing based on findings from in-depth interviews. The qualitative component also had limitations: interviewer effects, such as social desirability bias, may have affected participant responses. Moreover, the small number of clinics and providers included preclude generalization, which was not the aim of the present study.

To our knowledge, this is the first study to implement an IPT prescribing approach that is grounded in choice architecture. Our findings suggest the feasibility of using this approach to guide optimal prescribing while reducing, rather than increasing, provider work and cognitive load.

CONCLUSION

Choice architecture provides potential for subtle changes in choice presentation to make major changes in behavior when appropriately presented for the local context. This overall approach to IPT prescribing has the potential to work in settings with similar HIV and TB management needs, given appropriate adaptation of the actual implementation package, and is currently being tested in a larger cluster randomized trial.