A qualitative exploration of health-related present bias among HIV-positive adults in Uganda.

Decision-making errors such as present bias (PB) can have important consequences for health behaviors, but have been largely studied in the financial domain. We conducted a mixed-method study on PB in the context of ART adherence among clinic-enrolled adults in Uganda (n = 39). Specifically, we quantified PB by asking about preferences between medication available sooner to minimize headaches versus available later to cure headaches. We describe demographic similarities among PB participants and qualitatively explored how participants reflected on their PB (or absence thereof) in the context of health. Participants reporting PB were predominantly male, single/unmarried, older, had higher levels of education and income and more advanced HIV progression. Three common reasons for more present-biased choices provided were: (1) wanting to avoid pain, (2) wanting to return to work, and (3) fear of one's health worsening if s/he did not address their illness immediately. While PB in the financial domain often suggests that poorer individuals are more likely to prefer immediate rewards over their wealthier counterparts, our results suggest poor health is potentially a driving factor of PB. Further research is needed to build on these finding and inform how programs can frame key messages regarding ART adherence to patients displaying PB. Trial registration: ClinicalTrials.gov identifier: NCT03494777.

Keywords: HIV/AIDS; Uganda; qualitative analysis; adherence; present bias

In Uganda, despite free access to antiretroviral therapy (ART) for people living with HIV (PLWH), suboptimal adherence, defined as missing multiple doses in a given time period, continues to be widely reported and contributes to the failure to achieve viral suppression (Inzaule et al., [11]; Linnemayr et al., [13]; MacCarthy et al., [18]; Nabukeera-Barungi et al., [19]). Achieving and sustaining viral suppression through ART adherence is important for PLWH because it can also slow or prevent HIV from advancing to more severe stages of disease, such as chronic Stage 2 HIV infection or Stage 3 AIDS (as defined by the WHO), and ultimately reduce associated morbidity and mortality. Even so, studies suggest that between 21% to 44% of PLWH who have initiated ART have suboptimal adherence over their lifetime (Boender et al., [6]; Ortego et al., [22]).

Several individual (e.g., medication toxicity concerns), programmatic (e.g., scheduling conflicts) and structural (e.g., stigma, medical and transportation costs) barriers to ART adherence have been identified in the HIV literature (Adejumo et al., [1]; Ammon et al., [2]; Bikaako-Kajura et al., [5]; Hudelson & Cluver, [10]; MacCarthy et al., [17]; MacCarthy et al., [18]; Nabukeera-Barungi et al., [20]). There is growing evidence that decision-making biases studied by behavioral economics further contribute to a range of suboptimal behaviors (Bor & Thirumurthy, [7]; Linnemayr & Stecher, [14]; Thirumurthy et al., [25]). However, the role of behavioral economics in examining how personal biases influence adherence-related decision-making is less well studied in ART participants. Decision-making errors or behavioral "biases" as highlighted by behavioral economics (BE) are important to health because they may cause individuals to engage in harmful practices that are known to be unhealthy and for which they are highly motivated to avoid, such as smoking or overeating (Linnemayr & Thomas, [16]). One such BE bias is present bias, defined as a form of impatience in which individuals forgo long-term benefits from optimal decision-making in lieu of short-term considerations or suboptimal decisions they may later regret (Linnemayr, [12]; White & Dow, [27]). In the context of antiretroviral adherence, present bias may manifest in the form of skipping a medication dose to enjoy an alcoholic beverage or pausing medications to avoid being suspected of being HIV-positive by the family or a partner.

Existing literature on present bias is largely quantitative in nature (DellaVigna, [8]; White & Dow, [27]) and often focuses on a range of financial decisions such as spending or debt acquisition, or delayed investment in preventative health goods (Ashraf et al., [3]; Duflo et al., [9]). There have been few qualitative studies investigating individual preferences and rationale supporting suboptimal and biased adherence choices. This has hindered the development of more behaviorally-informed adherence programs to address potential present bias in long-term HIV treatment outcomes. In addition, while some present bias studies have examined influences on health-related decisions (Wang & Sloan, [26]), investigations of present bias on ART adherence (Linnemayr & Stecher, [14]; Stecher, [24]; Wang & Sloan, [26]) are scarce. The few available studies have suggested that individuals are often unaware of their present bias and tendencies to make short-term decisions that are disadvantageous for them in the long run (O'Donoghue & Rabin, [21]; Wang & Sloan, [26]).

The aim of this study was to examine present bias in the context of ART adherence in clinic-enrolled adults in Kampala, Uganda. Specifically, we aimed to quantify the prevalence of present bias among participants with recent ART adherence problems and describe potentially relevant demographic factors associated with displaying present bias. We then qualitatively explore whether and how ART participants reflected on their present bias (or absence thereof) relative to decision-making patterns underlying their suboptimal ART medication adherence. Ultimately, this study aimed to identify potential decision-making patterns that drive present bias in the context of HIV care and treatment and inform how future programs can frame key messages regarding ART adherence to patients experiencing this form of bias.

As part of a larger randomized clinical trial (RCT) (Linnemayr et al., [15]), we conducted individual semi-structured interviews (n = 39) in October and November 2018 with a subset of study participants, i.e., Clinic-enrolled adult ART patients. The study was conducted at Mildmay Uganda, an HIV clinic in Kampala, Uganda that specializes in the provision of comprehensive HIV/AIDS prevention, care, and treatment services.

Eligibility criteria for study participation included being age 18 and older, having been on ART at Mildmay Uganda clinic for two or more years, and having demonstrated adherence problems within the past 6 months (e.g., Defined as lack of viral suppression, being referred to adherence counseling, or showing disease stage 2 or 3 as per WHO guidelines). Participants were excluded if they were not mentally fit to provide informed consent, did not speak either English or Luganda (the local language), participated in any other adherence-related studies, were using a third-line treatment, and/or were unable to regularly use the Medication Event Monitoring System (MEMS) caps that monitored adherence during the parent RCT study visits.

We used Mildmay Uganda's clinic records to identify eligible ART participants. Once identified, we then used the date of subsequent appointment to identify the next opportunity for recruitment. After recruitment began, the study coordinators identified the patients due for a visit that day and approached all eligible participants regarding whether they were interested in study enrollment. Participants that initially agreed were taken to a separate study room to verify eligibility and administered informed consent, while maintaining their spot in the queue for seeing a provider. Following consent, all study-enrolled participants were administered an individual interview which lasted approximately 40 min. Participants were paid 20,000 USh (equivalent to about $5 USD) for their time.

Data were collected using a semi-structured interview guide to explore present bias in ART participants. The interview guide began with close-ended, quantitative questions to determine participants' stated preferences for immediate vs. longer-term benefits of different treatment options for a hypothetical ailment, followed by open-ended, qualitative questions inviting participants to articulate the reasoning behind their choices and to provide context on the range of factors influencing their decision-making process. A hypothetical drug to "cure a headache" was used to enable participants to evaluate their choice independent of prior ART medication habits or scenarios, and the responses to these questions were used to categorize participants as present-biased or not present-biased Figure 1.

Graph: Figure 1. Participant Categorization of Present Bias.

The subsequent qualitative questions asked participants, "When you answered the questions I just asked you, what did you think about when you were trying to decide between Drug A and Drug B?" Participants were encouraged to describe how they came to prefer one drug over another, and what led to them changing their minds (or not) when the characteristics of the drug were modified. We also asked participants to describe actual personal experiences to further illuminate their preference for immediate gratification versus delayed or long-term benefits with respect to their health.

All interviewers had extensive training and prior experience conducting qualitative interviews focused on ART adherence in Uganda. Additional training on the interview guide was led by the qualitative research lead (SM) and included mock interviews along with didactic instruction that focused on the concept of present bias and its representation in questions related to decision-making for ART adherence. The team continued to meet every two weeks to receive further feedback on interviews, discuss issues that came up (e.g., Differing interpretations regarding the health issue raised), and discussed whether or not certain typologies emerged among participants (described in Table 1) and whether or not there were similarities in the sociodemographic characteristics and/or reasons that appeared to drive their decision-making. These conversations informed the initial analysis of the qualitative data. Sociodemographic information was also collected as part of the larger RCT regarding participant's sex, age, income, education, and health status (e.g., Viral load levels and WHO HIV stage).

Table 1. Participant demographic and HIV-related characteristics by present bias assignment and overall (n = 39).

1 [a] Individuals in Uganda typically obtain vocational education after primary or secondary school education as post-primary or post-secondary training, but always prior to any university training.

2 [b] USD estimates calculated based on exchange rate of 1 USD=3700 Ugandan Shillings in December 2018.

All interviews were audio recorded and transcribed verbatim and translated from Luganda into English. The analysis involved two primary phases: quantitative descriptive statistics and thematic qualitative coding. In the first phase, we started by assigning ART participants into one of two groups: displaying present bias or not displaying present bias (referred to as "patient"). We then descriptively examined the potential differences in socio-demographic factors among individuals with present bias compared to those who did not experience this type of bias. Given the small sample size and power, no statistical inferences were made. Rather, this step provided preliminary insights regarding potential shared characteristics of ART participants with and without present bias.

In the second phase of the analysis, all de-identified qualitative data were entered into Dedoose software (version 8.1). We then created a list of the qualitative reasons given for their decision-making and grouped the list into core themes. The core themes informed a preliminary structured codebook that included descriptions, inclusion and exclusion criteria, and example quotations. The final codebook included 5 primary themes and 17 sub-themes. Two coders then jointly coded 10 interviews, and to establish inter-rater reliability using the Cohen's Kappa coefficient, they also separately coded a set of 25 transcript excerpts. Cohen's Kappa was 89.4% across coded excerpts, indicating "good agreement" between reviewers. After attainment of reliability, one coder coded the remainder of the interviews. The two coders met biweekly to identify emergent themes, codes, resolve any issues, and draft summary results which were discussed with the local research team and revisited from additional angles as necessary. Quotations were extracted from these excerpts based on the themes identified beforehand, and each quote was labeled with the participant's sex and WHO HIV stage (based on CD4 count).

The study was approved by the institutional review boards of the RAND Corporation (HSPC study number 2016–0956), Mildmay Uganda Research Ethics Committee (MUREC) (02013–2018), and the Uganda National Council for Science and Technology (HS 2394).

Thirty-nine ART participants (n = 39) were administered a semi-structured interview (Table 1). Over half of the sample was male (51%) and 39% were married. Most participants (72%) were aged 18–39 compared to 28% who were aged 40 or older. Highest level of education attained varied substantially with some participants having no education (3%) and some having vocational and/or university training (25%). The mean monthly income of participants was $34.92 USD (note the income estimation is based on the sample after excluding one outlier due to their disproportionate income – nearly 12 times – the reported monthly income of the sample which was likely due to data entry error). Most participants (80%) had a Stage 1 HIV classification (with CD4 > 350 cells/μL) compared to 13% and 5% with Stage 2 and 3 classifications (with CD4 < 350 cells/μL), respectively. Two-thirds (66.7%) were virally suppressed at last clinic visit (< 200 copies/mL).

A majority of participants in the sample (56%, n = 22) were categorized as present biased, resulting in 44% (n = 17) categorized as not showing present bias (Table 1). Although beyond the statistical power of this study, several demographic characteristics may warrant further assessment of association with present bias. ART participants displaying present bias in their drug choices were predominantly male (64% versus 35% in ART participants without present bias), unmarried (82% versus 35% in ART participants without present bias), older (36% versus 18% aged > 40), had higher post-secondary education (18% versus 6%), and had a higher mean income ($40.66 USD) versus ($27.03 USD) among non-present biased ART participants (Table 1). ART participants with present bias may also have higher prevalence of severe Stage 3 HIV infection (9%) compared to non-present biased participants (0%) (Table 1).

Table 2 presents the qualitative reasons that ART participants reported underpinning their present-biased choice for a currently-available drug (rather than future-available) as well as for a currently-available drug with a longer cure duration. Three reasons were commonly provided for more present-biased drug choices: (1) wanting to avoid pain, (2) wanting to get back to work, and (3) fear of one's health worsening if s/he did not address her/his illness immediately. Annotated, exemplary quotations for each reason are listed in Table 2.

Table 2. Reported reasons and exemplary quotations from ART participants for present bias selection of a currently-available drug rather than a future-available drug (n = 39).

Table 3 presents reported qualitative reasons from ART participants explaining why some people can wait for the future-available drug (rather than currently-available) with a longer cure duration. Two common reasons were provided to explain their willingness to wait: (1) wanting to feel better in the long-term; and (2) preferring to be cured rather than have recurring illness in the short-term. Annotated, exemplary quotations for each reason are listed in Table 3.

Table 3. Reported reasons and exemplary quotations from ART participants for non-present biased selection ("willingness to wait") of a future-available drug rather than a currently-available drug (n = 39)

This study aimed to explore the prevalence of present bias in the health domain among participants with recent ART adherence problems. It also describes sociodemographic factors potentially associated with present bias, as well as participants' qualitative reasons underlying their present- and non-present biased choices. Our findings suggest that potentially important demographic factors related to present bias were being male, single/unmarried, older, having higher levels of education and income and more advanced HIV progression. ART participants reporting present bias were predominantly male, single/unmarried, and older. They also commonly had higher levels of education and more advanced HIV progression yet showed higher rates of viral suppression. While a larger, sufficiently powered sample is needed to ascertain statistically significant differences, these findings suggest that ART participants' preferences for immediate or delayed rewards may be influenced by demographic characteristics (such as being young or having familial responsibilities). Further research is also needed to ascertain whether the present bias we measured also translates to ART drug-taking choices (such as whether to skip or maintain a dosage to avoid pain or worsening health or whether to skip or maintain dosages in order to return to work or feel better in the long-term). It is also plausible that older ART participants who have experienced being sicker as a result of HIV, and ART participants at a more advanced WHO HIV stage provided reasons for present bias that revolved around feeling better sooner and avoiding further pain. Conversely, non-present biased ART participants may not have experienced as much ill-health due to HIV and therefore were more focused on feeling better in the longer-term and potentially waiting for the possibility of a medication to better manage their HIV.

An additional important finding from this study relates to the relatively new assessment of present bias among ART patients in a sub-Saharan African clinical setting in the health domain. If confirmed in future analyses, our results suggest that present bias may manifest differently in a health-related context such as ART adherence compared to present bias in wealth-related contexts that is the focus of most existing studies on this topic, such as choosing between monetary rewards rather than choosing between health-related rewards (e.g., cure, treatment of headache and pain). For example, in the context of wealth, present bias studies often suggest that poorer individuals are more likely to prefer immediate hypothetical rewards over their wealthier counterparts, who in turn are more likely to wait for long-term gains (Banerjee & Mullainathan, [4]; Reuben et al., [23]). Our study found that poor health is potentially a driving factor of present biasedness in contrast to poor economic status, which may have been a greater driver of patience among ART participants who valued treatment of the hypothetical illness we asked about, given their limited economic resources to counter sustained effects of poor health. Additional studies are needed to elucidate and confirm these findings both in the context of qualitative inquiry and statistical analyses. Understanding whether and how present biasedness manifests in the health domain, particularly in the context of antiretroviral adherence, may be critical to the effectiveness of adherence strategies. For example, ART adherence counselors working with participants reporting present bias may be more effective by framing the importance of taking ART to quickly reduce the pain and negative symptoms resulting from their HIV infection.

Finally, there may be important sociocultural factors contributing to present biased decision-making among participants. For example, in Uganda, individuals may experience a greater urgency to take what is available now versus later due to reported mistrust of health care institutions, or an expectation that a promised future option may not come to fruition. A common saying in Luganda (the local language in Uganda) is that "when a mad person offers you something now, take it ... " suggesting that individuals are culturally advised to take what is offered without considering future (but uncertain) pay-offs. Future research on present bias in Uganda and similar settings are likely to benefit from mixed-methods assessments that take into account local perceptions of immediate and delayed rewards, and inform incentivized ART adherence strategies in this population.

There are several limitations to consider in this study. First, past ART medication and drug experiences may influence how participants responded to the drug choice questions regarding present bias. Some participants may also have based their response on their previous experience with a headache. This could be conceived as a strength in that responses reflected explanations consistent with prior behaviors. Conversely, this could be seen as a limitation if responses reflected past behavior more than underlying rationale. Second, while efforts were made to ensure adequate and culturally-appropriate translation of key terms relating to time (e.g., Present versus future), drugs, choice, headache, cure, health, and/or illness, it is possible that some intended meanings were not well conveyed or understood. Future studies may be enhanced by validated questions on present bias specific to a health context – which, to our knowledge - are not currently available. We observed that more literate participants and those who were more proficient in English often understood interview questions more easily than less literate and non-English speaking participants who required more time and explanation to respond to quantitative and qualitative questions. Importantly, the strength of our work is that it builds on existing literature on present bias (though typically in the financial domain) in a relatively new context of medication choice and adherence in sub-Saharan Africa. It also introduces the use of qualitative methods which are less commonly applied in behavioral economics and present bias research.

The role of behavioral economics in examining how personal biases influence drug and adherence-related decision-making is less well studied in ART participants. This study provides preliminary evidence on how present biasedness may be assessed, both qualitatively and quantitatively, in a sub-Saharan African clinical setting. More research is needed to inform how messages regarding ART adherence may be framed to address differential preferences of present and future gains in this population.

We would like to thank the Mildmay staff for allowing this study to take place in their clinic and for providing administrative support. We are particularly grateful to the study participants for sharing their experiences, and to our excellent study team Peter Wabukala, Lillian Lunkuse, Stewart Walukaga, Pius Kimuli, and Philip Aroda.

No potential conflict of interest was reported by the author(s).

The data that support the findings of this study are qualitative transcripts and are therefore not publicly available due to their containing information that could compromise research participant privacy.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

INTERVIEWER INSTRUCTIONS – READ THE FOLLOWING:.

I'd like to know a bit more about how people make decisions having to do with money and also their health. Thank you for talking to me about how you yourself make such decisions. Our talk will last about 30 min. I will be recording the conversation. Whatever you say to me will be kept private. Do you have any questions before we begin?

INTERVIEWER INSTRUCTIONS – Start Recording.

INTERVIEWER INSTRUCTIONS: To make sure the participants understands, ask them "When will drug A be effective and for how long? Good, now what about drug B?"

READ OUTLOUD: Now I'd like to understand how you make decisions about your health.

• Suppose you suffer from a bad headache (that prevents you from doing the things you usually do, such as going to work or meeting friends), and two drugs are available to treat it. Both drugs are the same price but only one of them can be used, and once you take any of these two drugs you cannot take another drug for a month due to possible drug interactions. So that taking another drug afterwards may be really bad for your health.
• Drug A can be taken tomorrow and will cure your headache for one week, after which a headache may or may not return. Drug B will not be available for 3 days, but will cure your headache for sure for at least 3 weeks, after which a headache may or may not return. Which would you choose? (Choose one)
• Drug A (available tomorrow and effective for one week, after which the headache may return but you cannot take another drug for another three weeks)
• Drug B (available only in 3 days, but then works for 3 weeks)

IF participant chooses 'Drug A':

• What if instead Drug B was able to cure your headache for at least 4 weeks? Which would you choose? (Choose one)
• Drug A (available tomorrow and effective for 1 week)
• Drug B (available only in 3 days, but then works for 4 weeks)

IF participant chooses 'Drug B':

• What if instead Drug A was able to cure your headache for at least 2 weeks? Which would you choose? (Choose one)
• Drug A (available tomorrow and effective for 2 weeks)
• Drug B (available only in 3 days, but then works for 3 weeks)
• When you answered the questions I just asked you:
• What did you think about when you were trying to decide between Drug A and Drug B?
• Tell us how come you:
• o (If participant made the same type of choices): Preferred Drug A | Drug B?
• o (If participant made different types of choices): Changed your mind?

Thank you for taking your time to participate in our study. We value your input and appreciate your support.

INTERVIEWER INSTRUCTIONS – Finish recording; add note if person left early or if there are any issues that we should take into consideration.