Explaining the Education-Health Gradient in Preventing STIs in Andean Peru: Cognitive Executive Functioning, Awareness, and Health Knowledge

Abstract

Context:

Previous research finds that educational attainment is a central component in the social shaping of health disparities. But little is known about the underlying pathways between education and health because there are limited appropriate data available to identify them. The widely-held assumption that schooling leads to awareness about STIs and better awareness leads to preventive behavior appears insufficient to explain the mechanisms underlying the education-health gradient.

Methods:

Unique survey data were collected in 2010 from a sample of 247 adults aged 30–60 from a relatively isolated Andean district of Peru with extensive variation in schooling coping with elevated risks of STIs. Structural equation modeling estimates the degree to which schooling provides cognitive resources, STI awareness, and sexual health knowledge, and how these jointly lead to condom use.

Results:

The condom use prevalence is 32%. One additional year of schooling is associated with an increase in the probability of condom use by 2.7 percentage points, after controlling for sex, age, area, and wealth. The pathway from schooling to condom use was mediated by cognitive executive functioning (CEF) skills (0.26 standard deviations), STI awareness (0.09) and sexual health knowledge (0.10); CEF skills were associated with condom use both directly and indirectly, through STI awareness and sexual health knowledge, and accounted for two-thirds of the education–condom use gradient.

Conclusions:

Results confirm a more complex pathway behind the education gradient in STI prevention. Schooling’s influence on cognitive skills leads not only to STI awareness, but also to accurate health knowledge that likely supports condom use.

Net of wealth and other characteristics, formal education attainment is so frequently shown to have positive associations with better health and longevity that it is often referred to simply as the “education-health gradient”^1–6. Moreover, education plays a major role in a person’s health and is now considered a central social determinant of health disparities in line with the epidemiological “fundamental cause of disease theory”. This theory hypothesizes that social characteristics act to put individuals at more or less risk of proximate biological causes of many diseases and are themselves responsible for health disparities within populations^7,8 This is also true for social disparities in sexual and reproductive health. Greater educational attainment is associated with a lower risk of contracting STIs, as well as effective contraception practices, family planning, and better maternal and child health outcomes^9–12.

Evidence about the education-health gradient in general, and sexual health specifically, comes chiefly from analyses of large probability samples responding to demographic and epidemiological surveys¹³. This approach lends significant generalizability to the education-health gradient for many health outcomes among many nations, cultures, and sub-populations. However, because surveys include partial or little to no measurement of possible intervening factors, less is known about the underlying mechanisms behind the education-health gradient¹⁴ Given this gap in the literature, an important next step is to evaluate which resources education provides and how those resources yield preventive behavior^2,4,5,15

It is often assumed that the main pathway from formal education to health is through awareness about health risks and disease prevention attained through basic literacy and numeracy skills, along with learning health curricula^16,17 Emerging findings, however, indicate that while promoting gains in literacy, numeracy, and associated curricula, schooling also enhances generalizable cognitive functioning, known as cognitive executive functioning skills (hereafter, CEF)—mental resources for planning, organization, working memory, integration of experience, spatial reasoning, unique problem solving, decision making, numeracy, and skills for goal-directed behavior^18,19. At the same time, considerable psychological research shows that although awareness is necessary, it is not sufficient for prevention because adequate underlying cognitive functioning is required to turn awareness into deeper understanding—health knowledge—about the risk and disease that can guide accurate decision-making and enactment of effective behavior^20–23. Taken together, these findings lead to a new hypothesis that the pathway from education to health includes not only information resources about the risks associated with unsafe sexual practices, but also the essential resource of better cognitive skills that can yield health knowledge for enactment of effective preventive behavior. Included too is the supporting hypothesis that educationally- enhanced cognitive functioning is the central factor in this pathway as it is also required to acquire both awareness and accurate health knowledge. While the new hypothesis of a broader, more complex pathway involving education and general cognitive functioning has received some empirical support, it has never been applied to the education-STI gradient, nor has it been supported by data including awareness and health knowledge^15,24

To test this hypothesis, we analyze unique survey data on adults aged 30–60 from a relatively isolated Andean district of Peru. This sample is unique in that: 1) individuals live with a relatively new and increasing risk of STIs, along with some public health attempts to increase awareness of risks in local communities, and 2) there is extensive variation in educational attainment and low variation in post-school activities. These characteristics make the district of Carhuaz a suitable case to study an increasing health risk and its possible prevention ensuring enough variation for these indicators as well as reducing potential sources of endogeneity associated with educational attainment. Lengthy in-field interviews and cognitive testing with participants enabled the collection of measures of the components of the hypothesized pathway behind the education-STI gradient. Then, a series of structural equation models enabled us to examine the direct and indirect associations of education with condom use, through the mediation of three resources: cognitive functioning, STI awareness, and accurate sexual health knowledge. Lastly, an assessment of four possible sources of endogeneity and robustness tests of the results are completed.

Conceptual and Empirical Considerations

Education and Cognitive Executive Functioning

Substantial evidence now exists that exposure to formal schooling enhances generalizable cognitive functioning, which are widely hypothesized to be the foundation for reasoning ability applied in novel contexts and effective decision making^25–27. A review of approximately 30 empirical cognitive and neuro-developmental studies using various methods indicates that exposure to routine schooling activities has significant influence on enhancing these meta-cognitive skills and their supporting neurological infrastructure¹⁹. A subsequent study employing a neuroimaging experiment and a field study demonstrates that regional brain activation of the substrate responsible for CEF skills can be shaped by common learning activities in school, and variable exposure to schooling is associated with variation in these skills in adulthood¹⁸.

Studies report substantial effect sizes between various educationally-enhanced cognitive skills and overall health. For example, Cutler and Lleras-Muney²⁸ in a comprehensive analysis of multiple measures of health behaviors (e.g., smoking, drinking, overeating, drugs use, and automobile and household safety) in US and UK, estimate that cognitive ability and knowledge account for 30% of the education-health behavior gradient, while material and social networks resources account for 30% and 10%, respectively. Further, using data from the British Cohort Study, Conti, and Heckman show that education accounts for as high as 50% of the variance among self-reported poor health (scored as 1 if the individual reports “fair” or “poor” health), and the education gradient may be steeper among individuals with higher cognitive abilities^29,30. Also, after controlling for childhood cognitive abilities, compared to the bottom quartile those in the top quartile of academic performance experience a 0.07 point higher average score in general health on a scale from 1 to 5 and develop 0.9 fewer chronic conditions out of the seven possible (asthma, emphysema, cancer, diabetes, hypertension, heart disease, and stroke)¹⁵. Lastly, experimentally improved numeracy has a positive effect on future healthy behaviors among college students, and higher objective numeracy (i.e., the ability to understand and use mathematical concepts) and other CEF’s are associated with prevention of STIs^31–34

Education, Awareness, and Health Knowledge

As is well-established, gaining accurate awareness about prevalent health risks is the first step in forming effective strategies to prevent all kinds of risk taking and avoid negative consequences, and formal schooling increases the ability to do this^35–37. Then, as a new health risk becomes more prevalent in everyday environments, beyond just awareness, a deeper understanding of these risks is required to initiate effective prevention of diseases^24,31. For example, when HIV/AIDS in sub-Saharan African populations was widely evident and feared, young females who were aware of the virus did not accurately understand its transmission mechanism continued participating in high-risk sexual behaviors³⁸.

In addition to awareness, formal education, likely through its enhancement of CEF skills, may also affect an individual’s ability to reason with received basic information, forming more accurate and complex knowledge about the physical causes and prevention methods of new risks and diseases^39–41. When a new health risk appears, with growing awareness there is often a mix of incomplete, inconsistent, and inaccurate information, enhanced CEF skills leads to better evaluation of information and then reasoning about its implication for an individual’s specific circumstances. More educated individuals, often with only a few more years of education, are the first to comprehend and benefit from new and even incomplete information about health risks emerging from behavior once assumed to be safe¹⁰. For instance, among adolescents who varied in complexity of knowledge about the biology of HIV transmission, those holding more advanced, as opposed to more naïve, knowledge tended to reject myths and misunderstanding about the disease^42,43. Other research shows that individuals without schooling and less-enhanced CEF skills are more likely to fail to integrate otherwise correct partial information into an accurate understanding to guide their behavior; such as some sampled uneducated adults who stated that HIV could be transmitted through blood transfusion and separately that condom use was a prevention, but then indicated that the risk from transfusion was mitigated if a condom was used²⁴. Lastly, without estimation of an antecedent education effect, a study of youth finds an association between higher CEF skills and superior reasoning about prevention information such as anti-smoking messages⁴⁴

Examining STIs and Education in Andean Peru

Data from a sample of adults from Carhuaz, a relatively isolated Andean district of Peru suffering recent rising risk of STIs and uneven educational development, are well-suited to test the hypothesis. First, since the 1990’s, a mixture of traditional sexual practices (e.g. unprotected sex and multiple female sex partners⁴⁵) plus temporary in-migration of young male miners from urban areas with greater prevalence rates introduced an elevated risk of STIs (e.g., gonorrhea, syphilis, genital herpes, and HIV/AIDS) in the local adult population⁴⁶. In the Ancash region, the prevalence of STIs increased from 7.5% in 2002 to 11.6% in 2011⁴⁷ Also, before data collection, there had already been some introduction of awareness information about STIs and their possible prevention in this region through government and international organizations’ initiatives, which provided funds to strengthen school-based interventions and population AIDS programs^16,17,48. A prominent example is the 2004 National Antiretroviral Treatment program that, along with providing free access to antiretroviral therapy, promoted information campaigns to raise awareness about HIV transmission mechanisms and reduce stigma about HIV⁴⁹. These conditions ensure considerable variation among participants in health preventive behavior and accurate understandings of STI prevention.

Second, adults from these Andean communities have significantly different exposures to formal schooling, including individuals with no formal schooling. Access to formal schooling has increased in this region among recent generations, although it is far from universal across families and communities (primary enrollment rate around 85% in 2007⁴⁶). Also, as subjects were mostly subsistence-level farmers, there is relatively little variance in pre-school family environments, post-schooling work experiences, economic status, and access to health care, which all assist in reducing potential sources of endogeneity. Both the spreading STI risk and uneven educational development make this a prime region for the study of the pathways underneath the education-condom use gradient.

Furthermore, although STIs are a major health challenge in Peru as a whole, studies on their prevention have mostly been conducted in urban areas. Despite the prevalence of risky practices, such as unprotected sex, and a recent trend of growing risk of STIs in rural communities, particularly in the Andean region, research in rural settings has been limited^46,50,51. Thus, there is significant need for understanding sexual risk practices and prevention strategies for these communities.

Methods

Sample and Subjects

Data collection occurred in 2010 in the district of Carhuaz, in the highlands of the Peruvian Andes about 34 km from the city of Huaraz, the capital of the Ancash region. From the adult population of this agrarian district, a two-stage stratified sampling procedure was used. First, to construct a community-level sampling frame, Peru’s National Census 2007 was used to identify all small traditional agrarian communities from the Carhuaz district, out of which 14 communities were selected that had both: a) among the district’s highest within-community variation in exposure to schooling, measured as the standard deviation of years of attainment of residents; and b) at least 50% of its residents living and working on subsistence-level farms. Second, a door-to-door survey stratified by education attainment was conducted in each and every household to recruit participants aged 30 to 60 years old with no mental or physical disorder requiring regular or frequent medication, and no indications of past neurological trauma. We targeted participants aged between 30 and 60 years old to increase the variance in educational attainment. This procedure yielded a sample of 247 adults with a wide range of school attainment including unschooled individuals from communities of mostly farmers. Acceptance of participation invitations was near universal.

Instruments and Measures

During a pilot study, two focus groups of residents from similar communities and knowledgeable bilingual (Spanish and Quechua) local individuals were implemented to ensure that the language used in the survey questions was accurate and relevant. Through a process of discussion and consensus, all terms about STIs, health knowledge questions, and condoms in the local vernacular language were selected on both accuracy and sensitivity to local cultural attitudes and values. These local terms were back-translated to confirm accuracy with our research partners at Carhuaz, who had extensive prior field experience in the local area. For adaptation and validation, all CEF instruments were similarly translated from English to Spanish, then to the local Quechuan dialect. The instruments were then independently back-translated to confirm accuracy and pretested on individuals similar to the subjects of the main study with instructions assessed for clarity. Six local school teachers proficient in Spanish and Quechua with previous experience in fieldwork were trained over a week as interviewers and testers. They were given an in-depth explanation of the study and interviewing demonstrations followed by practice and debriefings, plus they had the opportunity to assist in the formulation of the final form of questions and instrument instructions. All interviews were conducted in the subjects’ homes or a local school during one-on-one sessions with an interviewer. Given the rhythms of subsistence-level farming and comparatively relaxed daily demands on subjects’ time, the considerable instrumentation was completed over the course of two 2–3 hour sessions. All participants were given a modest amount of household items and groceries for their time, and the study purchased new curricular materials for primary schools in each selected community. Variables for the analysis include:

• Prevention Behavior - Condom use: Participants were asked the yes-or-no question “Have you ever used a condom during sexual relations?”

• STI Awareness: Participants were asked if they had heard of any of the seven STIs proposed by the study (i.e., chlamydia, gonorrhea, herpes, HIV/AIDS, hepatitis B, syphilis, human papilloma). STI awareness is a binary variable coded as 1 if participants responded that they were aware of at least one of these diseases and coded as 0 if not.

• Sexual Health Knowledge: Participants were asked whether 10 behaviors could protect themselves against STIs. Total correct answers reflect sexual health knowledge. Some examples of the sexual health knowledge items included in the study were “Do not share food with people who have STIs,” “Use a condom whenever you have sex,” and “Avoid sexual relations with people that have many sexual partners.”

• Years of schooling: Measured as the maximum years of formal schooling attained by the participant.

• CEF skills: Participants were tested on seven widely used and known instruments measuring different components of CEF skills^18,20,24 1) Verbal Associative Fluency was measured using the Controlled Oral Word Association Test (COWAT) by which participants are required to produce as many words as possible belonging to a common category. The score is given by the total number of words produced⁵². 2) Working memory was measured with the Backward Digit Span task, by which participants repeated back progressively longer strings of digits in the reverse order of presentation. The score is given by the number of correct responses and the range is 0–10 points⁵³. 3) Delis-Kaplan Executive Function System Tower test measures abstract thinking, problem-solving, planning, impulse control, and concept formation. This test requires the participant to move disks varying in size across three pegs to match a displayed tower in the fewest possible valid moves. The score is given by the number of correct towers formed and the range is 0–9 points⁵⁴ 4) Raven’s Colored Progressive Matrices test measures both visual perception and reasoning ability. This test requires the participant to select the appropriate choice to match the missing part of an incomplete figure. The score is the result of the average of four sub-scores with 12 items each, resulting in score range of 0–12 points⁵⁵. 5) The first 18 items of the Woodcock-Johnson III Calculation test were used to measure numeracy through math calculations such as addition, subtraction, multiplication, and division. The score is given by the total number of correct responses and the range is 0–18 points⁵⁶. 6) The Peabody Picture of Vocabulary Test (PPVT) measures receptive vocabulary, by requiring participants to point the picture that best corresponds to a given word. The score is determined by adding the number of correct responses between the base and ceiling to the base score and the range is 0–125 points. 7) Decision-making abilities were assessed with the Stickman test, which requires participants to respond in which village it would be most likely that the first person they met would be HIV positive, after showing various diagrams of hypothetical scenarios. A scenario was composed of two villages. In each village, different combinations of red stickmen (HIV-infected individuals) and black stickmen (individuals without HIV) were shown. In 6 incongruent scenarios, the village with a large number of red stickmen correspond to a lower probability. Then, the score is determined by the number of correct responses in incongruent scenarios and the range is 0–6 points²⁴ Scoring of instrument’s responses were completed by interviewers trained and supervised by study staff who had prior training on each instrument scoring. For validation, tests scores obtained from the pilot study were analyzed to compare with distributions and intercorrelations from reported U.S. samples, and found to be consistent.

• Control variables: As is standard in past research on disease awareness, health knowledge and health protective behavior^57,58, control variables at the individual and family levels include: sex (coded as 0=male, 1=female); age in years; area of residence (coded as 1 if within village/town and 0 if outside); and wealth index (a standardized composite score with mean=1 and SD=0, based on materials used for housing construction, types of water access and sanitation facilities, and ownership of 11 assets, including fridge, computer, television, bicycle, etc.) See appendix 1 for correlations between all variables included in the models.

Analytical procedure

Four Structural Equation Models (SEM) were estimated using logistic regression analysis for binary outcomes to estimate the direct and indirect associations between schooling and condom use, through CEF skills, STI awareness, and sexual health knowledge. As a combination of factor analysis and path analysis, SEM is appropriate for testing the proposed models because it can simultaneously estimate a measurement model of the multiple CEF skills and a structural model to evaluate the mediation of the hypothesized pathway from years of schooling to condom use through CEF skills, awareness and health knowledge^59,60.

The first specification, Model A, estimates the existence of the frequently reported education gradient with the protective behavior (condom use), conditioned on the set of control variables. Second, Model B estimates the degree to which multiple indicators of a participant’s CEF skills mediate the relationship between education and condom use. Third, Model C examines the degree to which CEF skills, STI awareness, and sexual health knowledge simultaneously mediate the education-condom use gradient. Finally, Model D estimates the association between educationally-enhanced CEF skills and condom use through STI awareness and sexual health knowledge. All SEM analyses were performed with Mplus, version 6.1. Standard errors were adjusted to account for possible clustering of observations at the community level, which was the primary sampling unit in the study. It was not feasible to implement a community fixed effects specification due to the small number of observations.

Although all models control for several non-educational variables (i.e., gender, age, area of residence, wealth, and community clusters), education is not randomly assigned to subjects and other unobserved factors may simultaneously affect our variables of interest. Thus, we performed a series of robustness tests that address four particularly important endogeneity concerns that could bias our estimates to understate the strength and impact of educationally enhanced CEF skills and the acquisition of health risk knowledge on health protective behavior. First, individuals with more CEF skills may be more likely to attain greater levels of education. This is particularly important in agrarian communities since parents may balance decisions about educational investments with needs for their children to contribute to farming and household work. To assess this source of bias, we evaluated whether including participants’ reasons for attending or not attending school change significantly our results. Second, as prior literature has shown, the type and content of jobs may affect skills related to cognitive functioning⁶¹. Therefore, we included participants’ occupation as a control in our analysis. Third, other family characteristics such as parental education may also simultaneously affect participants’ educational attainment, CEF skills and health knowledge⁶². Then, we tested whether including parental education in the analysis change our main results. Finally, condom use may depend on marital status as married participants may be systematically less likely to use condoms during sexual intercourse with their partner. Therefore, we evaluated whether results differ after excluding married participants from the analysis.

Results

Table 1 shows descriptive statistics of all variables included in the analyses. Almost 60% of participants in the sample lived in villages or towns of the selected communities versus isolated farms; 49% were men; and their average age was 45.6 years. The average years of schooling was 7.3 (SD=5.03) and 15.8% were unschooled. Nearly one-third (31%) of the participants in our sample reported ever using a condom during sexual relations. Likewise, nearly two-thirds (61%) were aware of at least one of the seven STIs proposed by the study. The average correct response rate about health knowledge was 3.6 out of 10 items (SD=3.17). These results show considerable variation among participants in condom use and accurate understandings of STI prevention.

Table 1.

Characteristics of adults aged 30–60 in the Carhuaz sample.

Variable	Observations	Mean	Standard deviation	Minimum	Maximum
Condom use	247	0.31	0.46	0.00	1.00
STDs Awareness	247	0.61	0.49	0.00	1.00
Sexual Health Knowledge	247	3.62	3.17	0.00	10.00
Verbal fluency test score	247	16.58	4.79	6.00	31.00
Backwards digits test score	247	3.36	2.08	0.00	10.00
Tower test score	247	3.60	1.91	1.00	9.00
Raven test score	247	7.33	2.93	0.00	12.00
Numeracy test score	247	13.35	4.74	0	18
PPVT test score	247	87.77	23.00	11.00	125.00
Stickman test score	247	2.37	1.58	0.00	6.00
Years of education	247	7.30	5.03	0.00	18.00
Age (in years)	247	45.62	8.73	30.58	61.83
Gender (1=male)	247	0.49	0.50	0.00	1.00
Area of residence (1=village)	247	0.59	0.49	0.00	1.00
Wealth index (standardize)	247	0.00	1.00	−2.41	2.87

Structural Equation Modeling of the Education-Health Gradient

As shown in Figure 1, Model A estimates the pathway from years of schooling to condom use, before including the hypothesized mediating factors. Controlling for the set of individual and household variables, we find that an additional year of schooling is associated with an increase in the likelihood of condom use by a SEM coefficient of 0.38, interpretable as an effect size of 0.38 of the standard deviation (SD) in condom use. In other words, an additional year of schooling increases the probability of condom use by 2.7 percentage points.

Figure 1. Model A: Education and condom use.

Fit indexes: pseudo R-Square=0.30. Controlling by gender, age, area of residence, wealth index and community clusters. Coefficient is standardized. **p<.01, *p<.05, +p<.10.

Model B, as shown in Figure 2, estimates the measurement model of CEF skills and their partial mediation of the association between years of schooling and condom use. Although still significant, the estimated direct pathway from years of education to condom use decreases by 68%. In addition, a significant indirect association of 0.26 of a SD exists between years of education and condom use via CEF skills (0.74 × 0.35 = 0.26).

Figure 2. Model B: Direct and indirect pathways from education to condom use through cognition executive functioning (CEF) skills.

Fit indexes: RMSEA=0.042; CFI=0.920; TLI=0.891. Controlling by gender, age, area of residence, wealth index and community clusters. All coefficients are standardized. **p<.01, *p<.05, +p<.10.

As shown in Figure 3, along with CEF skills, Model C adds the potential mediators of STI awareness and sexual health knowledge. Here the estimated direct pathway between years of schooling and condom use becomes nonsignificant and thus zero, indicating that CEF skills, STI awareness, and sexual health knowledge mediate this association. The strongest mediation runs through CEF skills (0.26 of a SD) compared to STI awareness (0.09 of a SD) and sexual health knowledge (0.10 of a SD).

Figure 3. Model C: Direct and indirect pathways from education to condom use through cognition executive functioning (CEF) skills, STDs awareness, and sexual health knowledge.

Fit indexes: RMSEA=0.071; CFI=0.752; TLI=0.638. Controlling by gender, age, area of residence, wealth index and community clusters. All coefficients are standardized. **p<.01, *p<.05, +p<.10.

To examine the process in greater detail, Model D in Figure 4 examines the degree to which educationally-enhanced CEF skills facilitate the acquisition of STI awareness and accurate sexual health knowledge leading to the use of effective protective health behaviors. The associations between education and both awareness and health knowledge are partially mediated by CEF skills. Although still significant, the estimated direct pathways from years of education to STI awareness and to health knowledge decrease by 40% and 60%, respectively. And further, the direct association between educationally-enhanced CEF skills and condom use is partially mediated by awareness (0.03 of a SD) and health knowledge (0.04 of a SD), although the difference between these two mediation paths is not statistically different. In other words, educationally-enhanced CEF skills are not only directly associated with more condom use, these skills also are associated with the prevention behavior through their association with both greater awareness and more accurate knowledge about STIs.

Figure 4. Model D: Educationally-enhanced CEF skills, STD awareness, sexual health knowledge, and condom use.

Fit indexes: RMSEA=0.044; CFI=0.906; TLI=0.860. Controlling by gender, age, area of residence, wealth index and community clusters. All coefficients are standardized. **p<.01, *p<.05, +p<.10.

Robustness tests

The results suggest that education could influence health protective behavior through the enhancement of CEF skills directly and indirectly through its influence on developing STI awareness and sexual health knowledge. As mentioned before, one the main threats to the validity of inferences from this study is that education is not randomly assigned within this population, so along with health behavior, awareness and health knowledge, it may simultaneously depend on unobserved specific factors such as individual’s ability and family background characteristics. Therefore, four possible sources of endogeneity are analyzed.

First, it is possible that more cognitively talented individuals were selected by families to receive greater years of education. However, when participants in our sample were asked about the reasons for their schooling attendance or non-attendance, most reasons were related to non-cognitive factors such as access to schooling, economic problems at home, or parental health problems. To assess this source of bias, all these reasons for non-school attendance were coded and included as control variables. After re-estimating the model, our main results remained (see appendix 2 for coding and results). Furthermore, when participants were young, attendance to schooling was not mandatory in these communities; thus, social stigma for low or non-attendance was unlikely.

Second, it could also be the case that exposure to certain occupational activities after finishing school may have affected the development of CEF skills so that exposure to schooling was not the necessary component. However, this explanation was unlikely in the present sample because participants’ jobs varied very little in their cognitive demands (around 35% of the sample were farm workers, 25% small vendors, 20% administrative employees and 20% other occupations). After coding participants’ occupation and re-estimating the model controlling for these variables, our main results remained, although the coefficient for the association between years of schooling and CEF skills was slightly larger (see appendix 2 for coding and results).

Third, it is possible that certain family characteristics such as parental education may have an effect on our variables of interest. However, most participants’ parents were not exposed to formal schooling when they were school age (unschooled fathers and mothers accounted for 45% and 70%, respectively, in our sample). Thus, it is unlikely that parental influences played a major role in our analysis. Nonetheless, to assess this source of endogeneity, we coded participants’ parental education level and included it as a control variable in the analysis. After re-estimating the model, no significant change existed in the estimated coefficients, except for a smaller coefficient for the association between years of schooling and CEF skills (see appendix 2 for coding and results).

Fourth, another factor that may be correlated with the likelihood of condom use is marital status. However, in our sample, this was not the case as condom use during sexual intercourse was reported only slightly less among married participants (30% of married participants reported using a condom during sexual intercourse versus 32% of non-married participants). To confirm that marital status is not a factor that could bias our results, we excluded married participants, who accounted for 55% of the sample; after re-estimating the model, the main results remained (see appendix 2).

Discussion

Our results show that attainment of formal education is significantly associated with condom use in the context of a rising risk of STIs in rural communities with many subsistence-level farmers, net of individuals demographic, socioeconomic, and geographical factors. An increase of 2.7 percentage points in the probability of condom use for each additional year of schooling is consistent with prior findings about the education-STI gradient and gradients with other health risks in different settings^5,39,41. Our findings also demonstrate that along with awareness, the pathway from education to the prevention of STIs through using condoms includes cognition and accurate knowledge about risks and prevention. This is particularly the case for educationally-enhanced CEF skills, which account for (mediate) over two-thirds of the education-STI gradient in these data. Further, the findings indicate that education’s association with CEF skills centrally underlies the gradient as these skills are not only directly associated with prevention behavior, they also increase awareness and knowledge that are themselves associated with condom use.

Additionally, our results about specific parts of the hypothesized pathway are also consistent with prior research, and the unique data behind them add new findings to these literatures. Various past experiments support the hypothesis that the learning process in school enhances underlying general cognitive functioning. While aptitude for general intelligence includes a significant genetic component, environmental factors, even beyond early ages, schooling has significant influences on realized intelligence¹⁹. It is likely that exposure to schooling is an intense cognitive environment providing this general resource for more-schooled individuals to apply to new risks and prevention behaviors over the course of their lives. Similarly, the finding that educationally-enhanced CEF also assists in the acquisition of awareness of a recent elevation of the risk for STIs in subjects’ lives, and also its association with more accurate knowledge about risks and prevention, is in line with growing results from the study of decision-making and problem-solving behavior, as well as with accruing research on the dynamic relationship among education, cognitive functioning and health knowledge^3,23,35.

These new results are timely in several ways. First, there is mounting evidence of an education gradient in sexual and reproductive health that is in line with the increasing reporting of education gradients for many health risks and disease^63–65. Second, the relatively unfamiliar STIs were rapidly introduced into this region in part from an expanding mining industry’s employment of young males from the capital, where STIs were more prevalent, and sexual practices did not usually include the use of condoms, particularly as prevention of vaguely understood STIs^45,66. Therefore, the increased risk of STIs in this region and its substantial variation in exposure to formal schooling would make education a major contributor to health disparities in sexual and reproductive health similar to earlier cases of HIV/AIDS infection in some nations of sub-Saharan Africa^67,68.

In evaluating these results, certain limitations must be considered. First, the data is taken from one point in time, thus conclusions about causal direction may not be warranted. Despite the robustness tests and the evidence supporting the mechanism proposed to explain how years of schooling affect condom use, there may still be other pre-existing differences among participants that were not considered and that may predict both education and STI-preventing behavior. Second, the outcome variable used in this study is an indicator of ever using a condom during sexual relations. Thus, it may be possible, although unlikely, that participants used condoms before they completed their schooling or before they learned about the new rising risk of STIs in their communities. It is also possible that some use of condoms was for birth control although the survey did not inquire about birth control methods. Third, the participants’ age range is wide, from 30 to 60 years, and while age was a control, the modest sample size does not permit separate analysis by birth-cohorts.

While several community interventions, school-based programs, and efforts to include health risk information in the Peruvian school curriculum may have successfully increased the ability of individuals to remember basic facts and prevention strategies about STIs, our results support the importance of policies and budgets that spread access and attendance to quality schooling. Even though providing at least primary and some secondary schooling for all taught by professionally trained teachers with suitable curricular materials is an expensive undertaking, particularly in the relatively isolated communities of the Peruvian Andeans, its potential lasting impact on cognitive functioning and thus populations’ elevated abilities to understand and prevent STIs is a major benefit.

Finally, although the association among education, CEF skills, awareness, sexual health knowledge, and prevention of STIs is promising, additional investigation about specific causal mechanisms underneath schooling and cognitive processes will enrich this line of research on the education-health gradient^69,70. Thus, future investigation on the causal mechanisms of education should explore what it is about education-enhanced CEF skills that enables individuals to translate health knowledge into protective behaviors. In addition, future research should examine pathways across different stages of a population’s exposure to new risks of STIs. Finally, previous research in Peru and in similar contexts suggests that the association between education and health is not homogenous across regions and communities^7,71,72. Future research should consider how geographical, environmental and contextual differences may affect individual-level factors such as education, CEF skills and health related outcomes, thus bringing new insights on the effects of education and providing valuable implications for more effective and equitable public policy aimed at sexual and reproductive health.

Appendix 1: Correlations

Table A1.

Simple correlations among variables included in the models.

		Cognitive Executive Functioning (CEF) skills										Demographic controls
	Years of schooling	Verbal fluency	Backward digits	Tower	Raven	Numeracy	PPVT	Stickman	STDs awareness	Sexual Health knowledge	Condom use	Age	Gender (1=female)	Area (1=within village)
Verbal fluency	0.44*	1
Backward digits	0.58*	0.35*	1
Tower	0.25*	0.25*	0.38*	1
Raven	0.56*	0.37*	0.48*	0.30*	1
Numeracy	0.70*	0.34*	0.56*	0.31*	0.54*	1
PPVT	0.77*	0.46*	0.53*	0.30*	0.62*	0.64*	1
Stickman	0.18*	0.23*	0.18*	0.14*	0.21*	0.25*	0.22*	1
STDs awareness	0.65*	0.34*	0.43*	0.25*	0.41*	0.51*	0.61*	0.16*	1
Health knowledge	0.58*	0.31*	0.44*	0.28*	0.39*	0.48*	0.59*	0.09	0.85*	1
Condom use	0.46*	0.20*	0.39*	0.18*	0.37*	0.39*	0.41*	0.10	0.41*	0.42*	1
Age	−0.14*	−0.02	−0.17*	−0.18*	−0.19*	−0.09	−0.03	−0.07	−0.09	−0.11	−0.26*	1
Gender	0.05	0.02	0.04	0.05	0.14*	0.09	0.09	0.05	0.06	0.06	0.12	0.06	1
Area	0.49*	0.22*	0.33*	0.08	0.36*	0.36*	0.48*	0.09	0.53*	0.49*	0.37*	−0.00	0.01	1
Wealth index	0.62*	0.32*	0.36*	0.16*	0.42*	0.43*	0.59*	0.20*	0.49*	0.46*	0.37*	0.01	−0.01	0.65*

Note:

^*p<.05.

Appendix 2: Robustness tests

Table A2.

Structural Equation Models, Robustness tests.

	CEF skills	STDs awareness	Sexual health knowledge	Condom use
Model D (for comparison)
Years of schooling	0.74 (0.05)**	0.31 (0.15)*	0.18 (0.07)**	0.06 (0.09)
CEF skills		0.26 (0.06)**	0.37 (0.05)**	0.26 (0.05)**
STDs awareness				0.14 (0.09)
Sexual health knowledge				0.15 (0.07)*
RT1: non-school attendance reasons
Years of schooling	0.75 (0.04)**	0.30 (0.18)+	0.16 (0.09)+	0.10 (0.07)
CEF skills		0.28 (0.07)**	0.39 (0.06)**	0.29 (0.05)**
STDs awareness				0.11 (0.08)
Sexual health knowledge				0.12 (0.05)*
RT2: occupation categories
Years of schooling	0.80 (0.05)**	0.31 (0.14)*	0.31 (0.05)**	−0.01 (0.07)
CEF skills		0.30 (0.06)**	0.24 (0.04)**	0.27 (0.04)**
STDs awareness				0.14 (0.06)*
Sexual health knowledge				0.18 (0.06)**
RT3: parent’s education
Years of schooling	0.71 (0.06)**	0.32 (0.15)*	0.17(0.08)*	0.04 (0.10)
CEF skills		0.26 (0.07)**	0.38 (0.05)**	0.28 (0.04)**
STDs awareness				0.14 (0.09)
Sexual health knowledge				0.15 (0.08)+
RT4: only non-married persons
Years of schooling	0.71 (0.04)**	0.33 (0.16)*	0.22 (0.09)*	0.13 (0.16)
CEF skills		0.20 (0.11)+	0.36 (0.10)**	0.07 (0.16)
STDs awareness				0.17 (0.08)*
Sexual health knowledge				0.26 (0.04)**

In addition to the variable specified in each robustness test, all models control by gender, age, area of residence, wealth index and community clusters. Non-school attendance reasons: 1) No institution in the area, 2) need to work outside home, 3) need to work at home, 4) not interested, 5) economic problems, and 6) health problems. Occupation categories: 1) unemployed, 2) military/police, 3) farming/fishing, 4) services (non-qualified), 5) services (qualified), 6) industry/construction, 7) technician, and 8) professional. Parent’s education level: 1) none, 2) preschool, 3) primary, 4) secondary, 5) postsecondary. Coefficients are standardized. Standard errors are in parentheses

^**p<.01,

^*p<.05,

⁺p<.10.