Financial Incentives, Not Behavioral Nudges, Led to Optimized HIV Testing among Pregnant Women in a High-Burden Urban Population in Ecuador

ABSTRACT.

Behavioral economic principles are increasingly being used in the fight against HIV, including improving voluntary testing in sub-Saharan Africa and South America. However, behavioral nudges have not been widely tested as a strategy to optimize HIV testing in pregnant women. Here, we assessed whether behavioral nudges or financial incentives were effective in optimizing HIV testing among pregnant women in a high-HIV burden setting. A randomized clinical trial was conducted between May 21 and Oct 5, 2018, to allocate pregnant women in Ecuador into three study arms: information only, soft commitment (i.e., a behavioral nudge), and financial incentives. All participants received an informational flyer, including the address of a testing location. Participants in the soft-commitment arm signed and kept a form on which they committed to get tested for HIV. Those in the financial incentive arm received a $10 incentive when tested for HIV. A stepwise logistic regression analysis estimated the effect of the study arms on HIV testing rate. Participants in the financial-incentive arm had higher odds of getting an HIV test (adjusted odds ratio 17.06, P < 0.001) as compared with information-only participants. Soft-commitment had the opposite effect (adjusted odds ratio 0.14, P = 0.014). Financial incentives might be useful in improving HIV testing among pregnant women, especially among those who might be at higher risk but who have not completed an HIV test.

INTRODUCTION

According to the Joint United Nations Program on HIV/AIDS (UNAIDS) Fast Track Targets, 95% of all people living with HIV (PWH) should know their HIV status by 2030.¹ Similarly, early HIV diagnosis is the first strategy of the “Ending the HIV Epidemic: A Plan for America.”² The National Public Health Strategy for HIV/AIDS-STI of Ecuador also emphasized testing.³ Optimizing HIV testing rates is a cornerstone in the strategies to end the HIV/AIDS epidemic.

The purpose of this article is to report on behavioral economics interventions to optimize voluntary HIV testing among pregnant women in a high-burden urban setting in Ecuador. Pregnant women represent a vulnerable population group requiring concerted efforts to deter HIV propagation. According to the Ministry of Health of Ecuador, HIV diagnosis among pregnant women is the first and most critical step to prevent and eliminate mother-to-child HIV transmission (MTCT).^4,5

Barriers for HIV testing may include lack of access to health care services, stigma and discrimination, perception of low transmission risk, lack of transportation or financial resources to visit an HIV testing facility, fear of a positive result and its socioeconomic implications, confidentiality issues, clinic setting, lower educational status, fear of not having access to effective treatment, and the lack of trained personnel in clinical encounters, among others.^6–9 Despite these barriers, improving HIV testing rates remains a priority because, according to the CDC, 40% of new HIV diagnoses can be linked to individuals unaware of their HIV status.¹⁰

Incentives and insights from behavioral economics have been used to motivate health-affirming behaviors. For instance, behavioral nudges and social or financial incentives have been studied to improve tuberculosis testing rates in India,¹¹ blood donations in Argentina,¹² and cord blood donations in Italy.¹³ Previous research shows that behavioral nudges and financial incentives may positively affect HIV testing rates in the general population.¹⁴ Hence, we hypothesize that behavioral nudges and financial incentives can positively affect HIV testing rates among pregnant women.

METHODS

Study population.

Ecuador suffers from an HIV epidemic that is unevenly distributed throughout the country. Between 1984 and 2011, the majority of PWH in Ecuador received antiretroviral treatment in the following urban centers where the epidemic concentrates: Guayaquil (63.1%), Quito (14.5%), and Esmeraldas (4.3%).¹⁵ In this city, specific population subgroups, such as men who have sex with men and female sex workers, have vastly differential disease prevalence of 10%¹⁶ and 0.7%,¹⁷ respectively.

Study design.

A randomized field clinical trial allocated pregnant women into three study arms: 1) information only, 2) soft commitment (i.e., the behavioral nudge), and 3) financial incentive. To ensure broad socioeconomic representation, pregnant women were recruited from five high-traffic areas in Esmeraldas: 1) the Municipal Fresh Market, catering to low to middle-income individuals; 2) Bolivar Avenue, one of the city’s main streets; 3) the shopping mall, catering to higher-income individuals; 4) the bus terminal for interprovincial travelers; and 5) the Malecon or ocean promenade. A randomized cohort design, described elsewhere,¹⁴ was used to allocate individuals into study arms while reducing cross-contamination. A recruitment and randomization schedule¹⁴ randomly allocated participants into the study arms in each location. Recruitment was done similarly to our previous research.¹⁴ Recruitment stands with a sign reading “Your health is important: be informed” were set up in the aforementioned locations. Upon interest, participants were approached by trained, nonclinical enumerators who provided information about the research study. Participants who agreed to participate were read a script that was specific to the study arm randomly allocated to the participant. Afterward, participants completed a basic sociodemographic survey that included age, education attainment, occupation, and ethnocultural self-identification. Participants were recruited between May 21 and October 5, 2018.

Women allocated to the information-only arm received a flyer with general information about HIV/AIDS testing and prevention, along with information on the nearest HIV testing facility. Afterward, they were encouraged to get tested. Those in the soft-commitment arm were asked to date and initial a paper form, which they retained, on which they promised to get tested in the next 15 days. No copy of the form remained with the study team. These participants received the same flyer with HIV/AIDS information, including contact and location information of the HIV testing facility, and were encouraged to get tested. Finally, pregnant women allocated to the financial incentive arm were offered a $10 incentive to get tested. The amount of the incentive was calibrated after consultation with community members to provide a meaningful incentive without generating undue pressure on the poorest participants. Although it corresponded to about half of the daily minimum wage, the purchasing power of $10 is not high (a 2018 Purchasing Power Parity conversion factor of 0.532 has been reported for Ecuador¹⁸). The incentive amount was meant to cover the cost of time and transportation to the testing site. These participants received the same flyer with information on HIV/AIDS prevention and the testing site and were encouraged to get tested. Those who subsequently completed the test received the incentive at the test site.

In summary, all participants were encouraged to get tested within 15 days, regardless of study arm allocation; the study team waited for up to 60 days to verify if participants were tested. The CONSORT flow diagram enrollment to data analysis is shown in Supplemental Figure 1.

Eligibility criteria.

Women 18 years or older were invited to participate. Inclusion criteria included being pregnant and living in Esmeraldas, Ecuador. Exclusion criteria included those unable or unwilling to complete the sociodemographic survey.

Data management and analysis.

Survey, HIV test completion, and HIV test result data were entered into a database by study team members. Univariate statistics were obtained for data cleaning and preparation for data analysis. Sociodemographic variables were recoded to reduce the number of categories. Afterward, bivariate analyses were conducted to identify the unadjusted effects of the study arms as well as potential confounders in the empirical model. A backward stepwise logistic regression model was fitted to control for sociodemographic and experimental characteristics. Additional backward stepwise logistic regression models were fitted to assess the following: 1) a randomly generated, age- and history of HIV testing–matched subsample, to account for study arm allocation distribution and to reduce possible selection bias; 2) a subset of participants with histories of HIV testing, to account for the effect previous HIV test might have on the relationship between study arm and HIV test completion; and c) a subset of participants with a history of previous HIV testing and who belonged to younger age groups, to account for the effect that previous HIV testing and age have on the relationship to study arm and HIV test completion.

Ethical considerations.

This study was approved by the institutional review board at (blinded for review).

RESULTS

Baseline characteristics.

A total of 425 participants were recruited. Most participants were younger than 28 years old (N = 288, 67.8%; Table 1). The proportion of older participants was highest among those allocated to the incentive arm (P = 0.013).

Table 1.

Baseline characteristics of the study population

Characteristic	Study arm			Total N = 425§	P value‖
	Information only N = 110* n (%)	Self-commitment N = 138† n (%)	Incentive N = 177‡ n (%)
Sociodemographic
Age					0.013
Younger	80 (72.7)	102 (73.9)	106 (59.9)	288 (67.8)
Older	30 (27.3)	36 (26.1)	71 (40.1)	137 (32.2)
Education					0.546
Non-university	89 (80.9)	104 (75.4)	135 (76.3)	328 (77.2)
University-level	21 (19.1)	34 (24.6)	42 (23.7)	97 (22.8)
Occupation#					< 0.001
Employee	23 (20.9)	33 (23.9)	74 (41.8)	130 (30.6)
Unemployed	6 (5.5)	4 (2.9)	10 (5.6)	20 (4.7)
Other occupation	81 (73.6)	101 (73.2)	93 (52.5)	275 (64.7)
Ethnocultural**					0.132
White	4 (3.6)	11 (8.0)	18 (10.2)	33 (7.8)
Others	106 (96.4)	127 (92.0)	159 (89.8)	392 (92.2)
Personal history
Previous HIV test					< 0.0001
Yes	106 (96.4)	132 (95.7)	144 (81.4)	382 (89.9)
No	4 (3.6)	6 (4.3)	33 (18.6)	43 (10.1)
Experimental
Recruitment location					0.066
Children’s park	32 (29.1)	27 (19.6)	30 (16.9)	89 (20.9)
Bolivar Avenue	21 (19.1)	38 (27.5)	50 (28.2)	109 (25.6)
Shopping mall	9 (8.2)	10 (7.2)	25 (14.1)	44 (10.4)
Fresh market	28 (25.5)	43 (31.2)	52 (29.4)	123 (28.9)
South	20 (18.2)	20 (14.5)	20 (11.3)	60 (14.1)
Recruitment process					0.003
Walking by	109 (99.1)	134 (97.1)	161 (91.0)	404 (95.1)
Referral	1 (0.9)	4 (2.9)	16 (9.0)	21 (4.9)
Interviewer					0.828
1	28 (25.5)	31 (22.5)	35 (19.8)	94 (22.1)
2	18 (16.4)	25 (18.1)	37 (20.9)	80 (18.8)
3	15 (13.6)	25 (18.1)	35 (19.8)	75 (17.6)
4	28 (25.5)	31 (22.5)	35 (19.8)	94 (22.1)
5	21 (19.1)	26 (18.8)	35 (19.8)	82 (19.3)
Experimental outcomes
HIV test completion					< 0.0001
Yes	10 (9.1)	2 (1.4)	118 (66.7)	130 (30.6)
No	100 (90.9)	136 (98.6)	59 (33.3)	295 (69.4)
HIV test result					0.823
Reactive	0 (0)	0 (0)	2 (1.7)	2 (1.5)
Nonreactive	10 (100.0)	2 (100.0)	116 (98.3)	128 (98.5)

^*Except for HIV test results, where N = 10.

^†Except for HIV test results, where N = 2.

^‡Except for HIV test results, where N = 118.

^§Except for HIV test results, where N = 128.

^‖Reported P values are for χ² tests, except for HIV test results where P value for Fisher’s exact test is reported.

^¶Age: Younger = 18–27 years old; older = 28–47 years.

# Occupation: Employee includes participants working either for governmental or private offices/businesses. Other occupation includes participants who are primarily engaged as students, housewives, or independent workers (either businessowners or artisans/technicians).

** Ethnocultural: Others includes participants self-identifying as African descendants, mestizos, or Montubio.

In terms of ethnocultural background, 7.8% of participants self-identified as White (N = 33), and 92.2% self-identified in other categories (N = 392). Other ethnocultural categories included Mestizos (N = 214; 50.4%), African descendants (N = 177; 41.6%), and Montubio (N = 1; 0.2%). Most participants (59.3%) completed secondary school (i.e., equivalent to high school in the United States). Other participants (N = 97; 22.8%) completed some level of university education, whereas 71 (16.7%) received only primary education (either complete or incomplete). There were no major ethnocultural (P = 0.132) or educational attainment (P = 0.546) differences across the different study arms.

Regarding occupation, most participants were housekeepers (37.4%), followed by employees (30.6%) and students (19.1%). Other occupational categories included business owner (8.2%) and unemployed (4.7%). The proportion of formally used participants (i.e., employees for either government or private companies) was higher in the incentive arm (P < 0.001).

For analytical purposes, all sociodemographic variables were recoded to reduce the number of categories and facilitate/clarify data analysis and interpretation of results, as shown in Table 1. Frequency distributions for all categories are shown in Supplemental Table 1.

As shown in Table 1, most participants (N = 382, 89.9%) reported having previously been tested for HIV, with 100 of them completing an HIV test during this study. Of those who completed an HIV test in this study, 57% (N = 57) stated that they were being retested because of pregnancy and/or for “control,” 13% explaining that they were being retested because of the incentive offered, one participant reported being retested both because of their pregnancy and because of the incentive, and one stated that she wanted to confirm previous results (38% did not provide a reason for retesting, Supplemental Table 2).

In terms of experimental characteristics, the most frequent recruitment site was the Municipal Fresh Market (N = 123, 28.9%) followed by Bolivar Avenue with 109 participants (25.6%). More participants were allocated to the incentive arm (N = 177; 41.6%) than to the soft-commitment (N = 138; 32.5%) or the information-only arms (N = 110; 25.9%). Most participants were recruited while walking by our recruitment site (N = 404; 95.1%).

There were two experimental outcomes: HIV test completion and HIV test result. Overall, approximately 30% of participants (N = 130) completed an HIV test. Of those, 98.5% (N = 128) returned a nonreactive or negative result.

Effect of behavioral nudges and financial incentives on HIV testing rate.

A stepwise backward logistic regression was fitted to control for potential confounders (Table 2). Participants in the financial incentive study arm completed more HIV tests, after controlling for recruitment location, age, ethnocultural self-identification, educational attainment, occupation, previous HIV test completion, recruitment process, and interviewer. Individuals who received the financial incentive had 17 times higher odds of getting tested than those in the information-only arm (P < 0.001). Those in the soft-commitment arm were 7 times less likely to get tested compared with the information-only arm (P = 0.014).

Table 2.

Predictors of HIV test completion among pregnant women

Variable	Initial model				Final model
	OR	95% CI		P value	aOR	95% CI		P value
		Lower	Upper			Lower	Upper
Study arm
Information-only (ref)	–				–
Soft-commitment (1)	0.15	0.031	0.705	0.017	0.14	0.031	0.679	0.014
Incentive (2)	20.56	9.114	46.374	< 0.001	17.06	8.138	35.770	< 0.001
Age*
Younger (ref)	–				–
Older	1.98	1.029	3.809	0.041	2.55	1.396	4.657	0.002
Previous HIV Test
No (ref)	–				–
Yes	0.27	0.099	0.760	0.013	0.20	0.076	0.510	0. 001
Location
Children’s park (ref)	–
Boulevard (1)	0.29	0.113	0.758	0.011
Shopping mall (2)	0.79	0.245	2.525	0.687
Fresh Market (3)	0.54	0.219	1.308	0.170
Supermarket (4)	0.47	0.157	1.390	0.171
Education
Nonuniversity level (ref)	–
University level	1.16	0.546	2.460	0.701
Occupation**
Other occupation (ref)
Employee (1)	1.39	0.713	2.701	0.334
Unemployed (2)	1.13	0.282	4.529	0.863
Ethnocultural***
White (ref)	–
Other races	2.23	0.607	8.203	0.227
Recruitment
Walking by (ref)	–
Referred by someone	1.81	0.387	8.437	0.452
Interviewer
1 (ref)	–
2	0.64	0.247	1.644	0.352
3	0.62	0.237	1.613	0.326
4	0.56	0.219	1.430	0.225
5	1.09	0.427	2.758	0.864

aOR = adjusted odds ratio; CI = confidence interval; ref = reference.

^*Age: younger = 18–27 years; older = 28–47 years.

^†Occupation: Employee includes participants working either for governmental or private offices/businesses. Other occupation includes participants who are primarily engaged as students, housewives, or independent workers (either businessowners or artisans/technicians).

^‡Ethnocultural: Others includes participants self-identifying as African descendants, mestizos, or Montubio.

Other variables that affected the HIV testing rate included age and history of previous HIV test. Older participants (i.e., those aged 28–47 years) were more than 2.5 times more likely to be tested than those 18 to 27 years old (P = 0.002). There was a negative association with having been previously tested for HIV; participants with HIV test history were 8 times less likely to be tested compared with those without previous tests (P = 0.001).

As noted earlier, the distribution of participants across study arms was uneven, with more participants allocated to the incentive arm. This resulted in differences in certain variables (age and history of previous HIV test) at baseline that were retained in the final model. Hence, we controlled for the effects of these variables in the final estimates by including them in the multivariate models. Moreover, to control for the uneven distribution of participants across study arms and to account for possible selection bias, we generated a random subsample of 30 participants in the incentive study arm and matched it, based on age and history of previous HIV testing, to 30 participants in the other two arms. Subsequently, we compared the baseline differences (Supplemental Table 3) and evaluated the impact of study arms on HIV test completion rates (Supplemental Table 4) in the randomly generated, matched subsample of pregnant women.

As expected, randomizing and matching removed the baseline differences in the sociodemographic, personal history with HIV testing and study characteristic (Supplemental Table 3). Multivariable models showed that women allocated to the financial incentive arm had higher odds (i.e., adjusted odds ratio = 9.75) of getting tested than those in the information-only arm, after controlling for the effects of potential confounders (Supplemental Table 4). Because no participant in the soft-commitment arm completed an HIV test, we compared the financial incentive arm and a new control group formed by combining the information-only and soft-commitment participants. This step increased model stability. Results from a backward stepwise logistic regression model indicate that women in the financial incentive arm were 21 times more likely to complete an HIV test when compared with women in this new control group (P < 0.001; Supplemental Table 5).

Because we were able to confirm the effect sizes and direction, we proceeded to conduct further analyses on the full dataset.

Effect on HIV seroreactivity.

Because only two participants had a reactive HIV test, it was not possible to ascertain whether the study arm had any association with HIV reactivity. No multivariable analyses were performed.

Financial incentives have differential effects according to participants’ history of HIV testing.

Because having been previously tested for HIV reduced the odds of HIV test completion by a factor of 5 (P = 0.001), we decided to assess further how a history of HIV testing affected the relationship between the study arm and HIV test completion rate. Two subsample analyses were performed on the participants with histories of HIV testing. First, we tested the hypothesis that pregnant women in the financial incentive study arm were more likely to get tested.

Among participants with a history of previous HIV testing, the proportion of pregnant women who completed an HIV test was higher in the financial incentive than in the soft-commitment and information-only arms (Table 3).

Table 3.

HIV test completion analysis restricted by history of previous HIV test

Characteristic	Study arm			Total N = 382	P value∞
	Information only N = 106 n (%)	Soft commitment N = 132 n (%)	Incentive N = 144 n (%)
HIV test completion					< 0.001
Yes	10 (9.4)	2 (1.5)	88 (61.1)	282 (73.8)
No	96 (90.6)	130 (98.5)	56 (38.9)	100 (26.2)
χ2 tests∞

A backward stepwise logistic regression analysis (Table 4) demonstrated that, after controlling for sociodemographic and study design characteristics, financial incentives increased the odds of HIV test completion by 14 times (P < 0.001). In contrast, soft-commitment decreased those odds to 0.15 (P = 0.014) compared with the control arm.

Table 4.

Predictors of HIV test completion among pregnant women with a history of previous HIV testing

Variable	Initial Model				Final Model
	OR	95% CI		P value	aOR	95% CI		P value
		Lower	Upper			Lower	Upper
Study arm
Information only (ref)	–				–
Soft commitment (1)	0.15	0.032	0. 719	0.018	0.15	0.031	0.681	0.014
Incentive (2)	16.71	7.423	37.623	< 0.001	14.13	6.728	29.693	< 0.001
Age*
Younger (ref)	–				–
Older	2.00	1.031	3.867	0.040	2.55	1.388	4.665	0.003
Location
Children’s park (ref)	–
Boulevard (1)	0.32	0.122	0.831	0.019
Shopping mall (2)	0.67	0.198	2.256	0.516
Fresh Market (3)	0.56	0.222	1.393	0.211
Supermarket (4)	0.41	0.130	1.256	0.117
Education
Non-university level (ref)	–
University level	1.02	0.465	2.236	0.961
Occupation†
Other occupation (ref)
Employee (1)	1.38	0.688	2.747	0.367
Unemployed (2)	1.00	0.230	4.361	0.998
Ethnocultural‡
White (ref)	–
Other races	0.61	0.145	2.593	0.506
Recruitment
Walking by (ref)	–
Referred by someone	1.66	0.336	8.158	0.535
Interviewer
1 (ref)	–
2	0.68	0.256	1.812	0.441
3	0.65	0.237	1.756	0.391
4	0.70	0.264	1.880	0.484
5	1.10	0.428	2.849	0.838

aOR = adjusted odds ratio; CI = confidence interval; ref = reference.

^*Age: younger = 18–27 years; older = 28–47 years.

^†Occupation: Employee includes participants working either for governmental or private offices/businesses. Other occupation includes participants who are primarily engaged as students, housewives, or independent workers (either businessowners or artisans/technicians).

^‡Ethnocultural: Others includes participants self-identifying as African descendants, mestizos, or Montubio.

Among pregnant women without a history of previous HIV testing (N = 43), only those in the financial incentive arm (N = 33) completed an HIV test (N = 30), whereas none of the women in the soft-commitment (N = 6) or information-only (N = 4) arms were tested (P < 0.001; Supplemental Table 6). Of the tested women, one was diagnosed with HIV (3.7%). The distribution of experimental outcomes in this subset of the sample precluded us from conducting multivariable analyses.

Financial incentives are effective among younger pregnant women.

Previous analyses showed that the odds of completing HIV tests among older women were 2.5 times higher than younger women. We hypothesized that financial incentives had an impact on HIV test completion rates among younger women. We tested this hypothesis by further restricting our analysis to women between 18 and 27 years.

As shown in Supplemental Table 7, younger pregnant women with a history of previous HIV testing in the financial incentive arm were 12.77 times more likely to get tested than the younger women in the information-only arm, after controlling for socio-demographic and study design characteristics.

DISCUSSION

The prevalence of HIV among pregnant women is a critical indicator that informs national HIV prevention strategies.⁵ Previous research has shown that the national estimated prevalence of HIV among pregnant women in Ecuador is 0.60%, with higher levels reported in the coastal region (1.13%) where Esmeraldas is located.⁵ However, accurately measuring HIV prevalence requires adequate HIV testing (i.e., detection critically depends on diagnostic efforts). According to UNAIDS, only 81% of PWH in Ecuador know their status,¹⁹ well below the 95% target.¹ This figure is likely different for pregnant women because national guidelines require early HIV testing during antenatal care.⁴ Accordingly, and consistent with observations in previous studies,⁵ 89.9% (N = 382) of pregnant women in our sample reported having previously been tested for HIV.

Improving HIV testing among pregnant women is critical because it is the first step in the strategy to prevent mother-to-child HIV transmission (PMTCT).⁴ Ecuador has recently made progress on PMTCT. For instance, the antiretroviral therapy (ART) coverage among pregnant women was estimated to be 88%, whereas the number of HIV-exposed children who are uninfected was increased from 3,900 in 2018 to 6,500 in 2019.¹⁹ Nonetheless, only 59% of expected pregnant women were tested for HIV in 2016,¹⁵ and there are still approximately 1,000 pregnant women who need to be on ART.¹⁹

In Esmeraldas, previous studies have found that up to 83% of pregnant women have completed at least one antenatal care visit.²⁰ In 2018, 57 women were diagnosed with HIV in Esmeraldas during antenatal care, representing 13% of all pregnant women diagnosed that year in Ecuador,²¹ compared with 46 (8%) in 2019.²² A total of 64 newborns were exposed to HIV during the perinatal period in Esmeraldas in 2018, which corresponds to near to 15% of all exposed newborns in the country,²¹ compared with 62 (13%) in 2019.²² This highlights the need to address PMTCT in Esmeraldas. Optimizing HIV testing among pregnant women is an essential tool in this fight.

In this study, we only found two participants with reactive HIV results in our sample, thus making it difficult for us to estimate the proportion of pregnant women PWH who knew their status in our target population. Improving HIV testing could be an appropriate approach to continually improve these statistics.

Broad societal inequities, which still affect Ecuadorians,²³ are likely to have an impact on healthcare access, HIV testing rates, and HIV testing preferences (i.e., choosing whether to test for HIV or not). Most of the published literature on health disparities in Ecuador focuses on indigenous populations, rural populations, low socioeconomic populations, or a combination thereof.^23–26 In Esmeraldas, higher-than-national-average poverty, lower-than-national-average number of physicians (per 1,000 population), and HIV-related stigma compounded, by higher-than-national-average proportion of African descendants (43.9% versus 7.2%), likely translate into HIV-related health disparities, including access to HIV testing. For instance, African descendants have lower rates of access to professional assistance during delivery than other nonindigenous races and lower contraceptive coverage than other races.²⁷

In our study, financial incentives were effective in increasing HIV testing, and thus leading to a path of optimization of HIV testing rates, among pregnant women (in the whole sample and subsets who were younger and/or had a history of previous HIV tests; see Supplemental Figure 2). Specifically, financial incentives increased the odds of getting tested by a factor of 17 (P < 0.001). On the other hand, soft commitment had an opposite effect, although at a smaller magnitude (odds were decreased by a factor of 7; P = 0.014). These results are similar to findings from our previous research conducted among the general population of Esmeraldas.¹⁴

Similarly, financial incentives increased the odds of completing a test by 14 times among women with a history of previous HIV testing (P < 0.001). In our sample, age was unevenly distributed at baseline. We accounted for that in a subsample of younger women. We found that financial incentives increased the odds of completing a test by 12 times among women 18 to 27 years old with a history of previous HIV testing compared with the control arm (Supplemental Figure 2).

There is now a growing literature on financial incentives in various forms, such as conditional fixed incentives, prizes (e.g., lotteries), prize-linked savings accounts, cash transfer programs, conditional economic incentives, or some combination of these.^{7,11,13,28–31} HIV-related outcomes improved by financial incentives include HIV testing, HIV retesting, voluntary medical male circumcision (VMMC), and HIV prevention and treatment.^28–36 Our study contributes to this growing literature by examining the effect of financial incentives on HIV testing among pregnant women.

Financial incentives can effectively address structural, interpersonal, and individual-level barriers, such as testing costs, anticipated stigma/fear of disclosure (i.e., they may facilitate social support if provided in public spaces). Incentives may act as a last-line motivator for people who perceive themselves as susceptible to HIV and consider HIV/AIDS disease as severe but still otherwise hesitate to be tested.³² Our data suggest that the latter mechanism may have played a role among pregnant women in urban Ecuador because HIV seroreactivity was detected only among women who were allocated to the financial incentive arm (Table 1).

We also noticed an interesting pattern in the proportion of participants being recruited by referral. It is higher in the financial incentive arm (9.0%, N = 16) compared with those in the soft-commitment (2.9%, N = 4) or information-only (0.9%, N = 1; P = 0.003) arms. Reasons may include altruism (“I would like my relative/friend/neighbor to also receive this benefit”), profit (“I tell another about this benefit and request that I get a share”), or social norms (“I am expected to tell others how I received this benefit”).

We should also note that as effective and acceptable as financial incentives might be, scaling up such interventions might be problematic. For instance, Kenyan providers have expressed concerns over the possibility that financial incentives create expectations of financial support for further care.³⁷

Similar to other studies of this type, the limited number of participants means that we should be cautious in interpreting the results. The distance between study results and policy formulation is still considerable. In particular, more large-scale field experiments should be conducted. Finally, our study tested only two interventions. Other approaches exist, and these should be tested in future studies. For instance, opt-out strategies may be more cost-effective than financial incentives by improving the number of new HIV diagnoses at a lower cost per diagnosis.³⁶ That said, the appropriate approach may plausibly vary across cultural, economic, sociodemographic, and geographic contexts. We should continue to refine and find the best ways to achieve the UNAIDS 95–95–95 Fast Track Targets.

Public health implications.

The use of financial incentives to improve HIV testing among pregnant women would be more cost-effective if focused on those who are likely to be living with HIV while being unaware of their status. Such narrowing of the target population might improve feasibility and sustainability, as has been suggested by other researchers.³⁷ In our sample, approximately 10% of participants had not been previously tested for HIV (N = 43), even though HIV testing is free to pregnant women. Of those we sampled, one tested positive for HIV (Supplemental Table 6), implying that 50% of the reactive results came from women who had not previously been tested and were in the financial incentive arm. None of the women who had not been previously tested for HIV and who were in the other arms completed an HIV test (Supplemental Table 6). We surmise that financial incentives may act as the final push for high-risk women to get an HIV test. Financial incentives can be used to diagnose previously undetected pregnant women by targeting those who have not previously been tested. This may suggest a policy direction. In addition, only approximately 26% of pregnant women with previous HIV test completed an additional HIV test in this study (i.e., completed retesting). Of those who retested, only one participant stated that she wanted to confirm results while most (at least 57%) stated that were re-testing because of pregnancy and/or control purposes. These findings suggest that pregnant women with a history of previous HIV test were less likely to respond to our interventions and, when they did retest, it was for preventive reasons. Therefore, the effects of incentives herewith observed might be addressing unmet HIV testing needs even among pregnant women with a history of previous HIV testing, at least in part.

Of note, the behavioral nudge that we tested was not effective in increasing HIV testing in this population of pregnant women. Altogether, the findings from financial incentives and behavioral nudges could help countries strategize their approaches to optimize HIV testing among pregnant women. Financial incentives can lower the last barrier to achieving zero MCHT in countries with epidemiological profiles similar to Ecuador.

Finally, pregnancy is a vulnerable time, and PWH who are pregnant commonly report depressive symptoms.³⁸ Learning a new HIV diagnosis may add stressors, such as disclosure-related worries. For instance, pregnant women with HIV in PMTCT programs have expressed worry about nondisclosure to the father of the child.³⁹ Thus, financial incentives could potentially add HIV disclosure burden to pregnant women. Yet HIV disclosure during pregnancy may reduce MTCT.⁴⁰ Policy decisions should carefully consider the reproductive rights of women living with HIV⁴¹ as well as the population health benefits of interventions.

Note: Supplemental tables and figures appear at www.ajtmh.org.