Abstract
Starting in 2017, Ecuador gradually expanded its primary healthcare access program nationwide using mobile traveling healthcare teams through the Estrategia Médico del Barrio (EMB) [or Neighborhood Doctor Strategy]. EMB teams, composed of a primary care physician, a nurse, and a community health worker, made home visits in marginalized areas. We estimate the impact of the EMB on health and utilization outcomes using nationally representative household surveys for 2006 (N=55,666), 2012–13 (N=92,500) and 2018–19 (N=168,747). The treatment variable at the extensive margin is any exposure to EMB at the canton level. At the intensive margin, we use exposure in terms of weeks covered by EMB and the number and composition of EMB personnel per 1000 population. We identify outcomes of treated vs. non- or partially-treated cantons based on the random combination of the timing of the start of the program’s implementation and the timing of the survey interview, which varied across cantons. We use difference-in-difference (DD) and difference-in-difference-in-difference (DDD) frameworks, the latter for cantons with high indigenous concentration. We find significant effects on the reported health problem and preventive care, but mixed results in terms of curative healthcare. The DDD specification shows that EMB improved health problem diagnoses and preventive healthcare utilization, including in highly indigenous cantons, yet it seemed to have had mixed results in terms of curative care use in Ecuador. Various alternative specifications and robustness tests do not qualitatively alter the main findings.
Keywords: mobile traveling healthcare teams, primary care access, difference in differences, Ecuador
1. Introduction
Primary healthcare access is a significant issue in low- and middle-income countries (LMICs) (Bitton et al., 2019; Rule et al., 2014). These nations grapple with three key challenges: widening healthcare access, improving the quality of care, and ensuring an equitable distribution of healthcare resources and services (Fanelli et al., 2020; Ozawa et al., 2019). Ecuador serves as an example, with its physician density being relatively low at 2.2 physicians per 1,000 people as of 2017 (World Health Organization 2024 data.who.int, n.d.). Despite the abundance of research on healthcare systems in high-income countries (Braithwaite et al., 2019; Papanicolas et al., 2018), there is a noticeable gap in studies tailored to the distinct needs and circumstances of LMICs. This lack of targeted research hinders the development of effective strategies to address the multifaceted healthcare issues in countries like Ecuador.
In response to the research and implementation gaps in LMICs, various strategies have been explored, each targeting different aspects of the healthcare challenge. Initiatives like community-based health programs, which are promoted to reduce health inequities, have shown mixed effectiveness in reaching marginalized groups (Ahmed et al., 2022). Telemedicine, accelerated by the COVID-19 pandemic, has improved access to healthcare, though it faces infrastructure and regulatory challenges (Camacho-Leon et al., 2022; Mahmoud et al., 2022). Policies for equitable healthcare financing and universal coverage, while beneficial, often favor the richer population, highlighting a need for more pro-poor investment (Asante et al., 2016). Building on these strategies, mobile traveling healthcare teams, such as Ecuador’s Estrategia Médico del Barrio (EMB), represent a flexible approach to healthcare delivery in underserved areas.
Mobile traveling healthcare teams strategies like EMB offer a wide array of benefits that extend beyond the direct provision of medical care, affecting societal and individual well-being in multiple ways. One of the primary advantages is the enhanced access to healthcare services, enabling timely and more convenient care for patients (Solomon et al., 2020). These teams are instrumental in reducing transportation costs for patients by bringing healthcare services closer to their communities (Frijters et al., 2020). Furthermore, they support environmental sustainability through the promotion of less travel, indirectly contributing to health and carbon emission reduction benefits (Chapman et al., 2018). From a financial perspective, mobile healthcare services demonstrate significant return-on-investment, showcasing their economic value beyond healthcare outcomes (Morphew et al., 2013; Oriol et al., 2009; Roeper et al., 2018).The value proposition of mobile traveling healthcare teams is particularly pronounced in areas that are rural and peri-urban, where they bridge significant healthcare access gaps (Alexy & Elnitsky, 1998; Aung et al., 2015; Savaşer & Kara, 2022). The success of such interventions across regions underscores their wide relevance, yet detailed evaluations of their effectiveness in developing settings, especially formal assessments, remain particularly rare.
For instance, in Guatemala, the introduction of similar healthcare initiatives leads to measurable enhancements in healthcare utilization (Cristia et al., 2015). Brazil’s efforts, encapsulated by “Programa Mais Médicos” showed a reduction in hospitalizations for ambulatory care-sensitive conditions, particularly after the second year of the program (Fontes et al., 2018). Also, in Brazil, the Family Health Program has shown reductions in infant mortality, especially in the poorest municipalities (Rocha & Soares, 2010).
However, the current landscape of research presents a dichotomy. While large-scale studies spotlight the achievements of mobile healthcare traveling team programs, a sizeable segment of the literature consists of smaller, often localized studies. Many of these employ qualitative methodologies or are limited by their small sample sizes (Walker et al., 2018; Yu et al., 2017). For example, the study by Sritart et al. (2021) in Thailand provides valuable insights into the potential of mobile healthcare delivery, but its scope is limited, making broader generalizations challenging. Concurrently, Sims-Gould et al. (2017) and Sims-Gould & Martin-Matthews (2010) emphasize crucial operational facets like collaboration and resource coordination, which are pivotal for the success of mobile health teams.
In this study, we investigate the impact of the “Estrategia Médico del Barrio” (EMB) [Neighborhood Doctor Strategy], a mobile traveling primary healthcare initiative launched in Ecuador from 2017 to 2019. This program was designed to enhance healthcare accessibility by sending mobile healthcare teams to conduct home visits in underserved areas, thereby promoting health and fostering engagement in health activities within those communities. EMB teams, consisting of a primary care physician, a nurse, and a community health worker, aimed at not only providing direct medical services, but also supporting residents with health education and preventative care strategies. Initiated in February 2017, the EMB program gradually expanded nationwide by the end of 2019. The protocol involved EMB teams identifying at-risk individuals during their community visits, subsequently arranging for these individuals to receive further medical attention either through appointments at nearby health centers or via house calls, particularly for those unable to travel to a healthcare facility.
Our study identifies a natural experiment, arising from the progressive rollout of EMB across different cantons at varying times, allowing us to exploit regional and temporal variations in program exposure. We use difference-in-differences (DD) models to discern the program’s influence on reported health issues and utilization of healthcare services. By comparing health outcomes before and after the program’s implementation within the treated cantons, and against control cantons that had not yet received the intervention, we can obtain a more precise estimation of the EMB program’s effects in a way that reduces confounding that typically affects observational studies. Furthermore, we employ a triple-differences (DDD) approach to capture the differential impact of EMB across cantons with high indigenous concentration. Such a methodological design strengthens the causal inferences we can draw about the program’s effectiveness in improving healthcare accessibility and quality for Ecuador’s underserved populations.
This study employs comprehensive datasets gathered from both nationally-representative surveys and administrative records. We derived pre- and post-program healthcare data from the Encuesta Nacional de Salud y Nutrición (ENSANUT)––a cross sectional health and nutrition survey conducted nationwide in 2012–13 and 2018–19. Administrative information on EMB’s implementation, including the initiation dates and the total number of staff members within each canton, was sourced from official data-transparency requests summarizing health district records. Out of 107 health districts contacted, we received responses from 93, encompassing 92% of the cantons recorded in both ENSANUT rounds. To execute the DD analytical framework, we capitalize on the staggered nature of the EMB’s implementation and the ENSANUT data collection. This random variation in EMB exposure across cantons allow us to compare between individuals residing in cantons where the EMB program began before and those where it started after the ENSANUT survey dates.
Our contributions in this study are manifold. To the best of our knowledge, this is the first study to investigate the impacts of a large-scale mobile traveling primary healthcare initiative in a small, lower middle-income country, such as Ecuador, using a rigorous quasi-experimental approach. We also explore the program’s heterogeneous effects in regions with substantial indigenous populations, an aspect often neglected in economics research. Our primary administrative dataset, compiled through exhaustive interactions with each of the health districts in Ecuador, underpins our analysis, offering both depth and granularity.
Additionally, to ensure our findings are comprehensive, we consolidate data from various nationally-representative datasets, including both 2018 and 2012 ENSANUT surveys as well as the 2006 Encuesta de Condiciones de Vida (ECV), which we use to verify the parallel trend assumption. This effort is further strengthened by incorporating weather data from the National Oceanic and Atmospheric Administration (NOAA), enabling us to control for the potential impact of weather characteristics on health outcomes.
Our research pivots on a central hypothesis: mobile traveling primary healthcare initiatives, like the EMB program, have a substantial impact on health outcomes and healthcare service usage. Our findings confirm this hypothesis, albeit with nuanced results. The primary DD analysis indicates EMB had a significant reduction in the reporting of health issues in the past 30 days by 4.69 percentage points (from a baseline of 35.84%), yet no notable changes were detected in either curative or preventive care. The study further investigates the effect of having a greater number of EMB primary care physicians and mobile health team staff (nurses and community health workers). Cantons with more EMB primary care physicians experienced a 4.86 percentage point increase in reported health issues and a 5.06 percentage point decline in curative care in the past 30 days (from a baseline of 41.99%). In parallel, cantons with increased EMB team staff saw a 4.68 percentage point rise in reported health issues and a 1.52 percentage point increase in preventive care (from a baseline of 8.58%).
When executing the DDD model to assess effects on individuals living in cantons with a high indigenous concentration, we found that prolonged exposure to EMB significantly increased the reporting of health issues by 9.81 percentage points and the probability of receiving preventive care by 2.62 percentage points. The increased reporting of health issues is likely due to higher diagnosis by EMB teams as part of their community visits. Nonetheless, this did not translate to a significant rise in curative treatments within a 30-day period, which may be attributed to delays in accessing treatment. Meanwhile, the growth in preventive care services can be attributed to the capabilities of non-specialist providers, particularly community health workers, to deliver these services.
The rest of the paper proceeds as follows. Section 2 offers a detailed examination of Ecuador’s 2008 health reform and the EMB program. Section 3 presents the data, and section 4 introduces the identification strategy and econometric methods. Section 5 explains the results, mechanisms, and robustness checks. Section 6 presents the discussion of our findings and concludes.
2. Background
2.1. The Ecuadorian 2008 Health Reform and Model with a Family, Community, and Intercultural Approach
In 2008, Ecuador’s Constitution was updated to recognize health as a basic right (Constituent Assembly of Ecuador, 2008). In alignment with this constitutional mandate, a process of health reform was initiated with the primary objective of ensuring universal access to healthcare for all Ecuadorians. In 2011, the Comprehensive Healthcare Model with a Family, Community, and Intercultural Approach [Modelo de Atención Integral de Salud Familiar, Comunitario e Intercultural] was conceived and operationalized. The goal of this model is to strengthen primary healthcare by prioritizing health promotion and disease prevention. It endeavors to “transform the conventional curative approach, which has been disease and treatment-centric, into one that prioritizes health as an individual, familial, and community right” (Ministry of Public Health of Ecuador, 2012b).
Central to the model’s framework is its focus on community participation and interculturality. The model positions community involvement as a key mechanism for improving the population’s living conditions. In alignment with this strategy, the role of community health workers was introduced. These professionals, officially referred to as técnicos de atención primaria, play a pivotal role in enhancing community health through preventive and promotional efforts. In terms of interculturality, the model incorporates an intercultural approach into healthcare delivery. This integration includes resources from both traditional and complementary medicine practices across different healthcare levels, highlighting its approach to comprehensive and culturally-informed care (Ministry of Public Health of Ecuador, 2012b).
To support the objective of universal healthcare, the Ministry of Public Health provided health services at no cost to all users, leading to a 300% surge in demand. To accommodate this increase, the government restructured the territorial organization of the health network, decentralizing it into nine planning zones and 140 health districts. Between 2007 and 2016, over $16 billion US dollars were allocated to healthcare infrastructure, resulting in the construction of 47 hospitals and 74 health centers (Espinosa, Acuña, et al., 2017). Concurrently, significant investments were made in human capital. The number of healthcare workers tripled, salaries quadrupled, and scholarships were provided for training healthcare professionals abroad in specialized areas. From 2012 to 2015, approximately $30 million US dollars were allocated to these scholarships (Espinosa, Acuña, et al., 2017; Espinosa, De la Torre, et al., 2017).
Several studies have examined the role of Ecuador’s 2008 health reform on health equality and access to primary care. Flores Jimenez & San Sebastián (2021) reported that following the reform, there was a reduction in hospitalization rates for Ambulatory Care Sensitive Conditions (ACSC)—acute or chronic health conditions that can lead to potentially preventable hospitalizations when not addressed in an outpatient primary care setting. Similarly, Granda & Jimenez (2019) and Quizhpe et al. (2022) observed increased healthcare access for all socioeconomic groups post-reform, including low-income, indigenous populations, residents of rural areas, and those with limited education.
2.2. Overview and Brief History of Estrategia Médico del Barrio
“Estrategia Médico del Barrio” (EMB) [Neighborhood Doctor Strategy] was a national health program launched in 2017 as part of the aforementioned operational model. It consisted of home visits made by mobile traveling healthcare teams to promote primary healthcare services in marginalized areas. The objectives of EMB included: (a) to bring services closer to the community, guaranteeing equitable access to healthcare services with emphasis on vulnerable groups; (b) to reorganize the use of available resources at the different levels of healthcare services; (c) to promote active community involvement in health initiatives; (d) to optimize health data management for strategic decision-making within the Ministry of Public Health (Ministry of Public Health of Ecuador, 2017).
EMB teams, officially referred to as Equipos de Atención Integral de Salud, were composed of a primary care physician, a nurse, and a community health worker. Physicians, known as the “neighborhood doctors”, were healthcare professionals embedded in a local community or neighborhood. Their primary role was to advise, guide, and oversee the health activities of that population. Community health workers were in charge of health promotion and disease prevention in each of their communities, as well as strengthening social participation and co-responsibility (Ministry of Public Health of Ecuador, 2017). These practitioners are not required to have a professional health degree but need to complete a training program provided by the Ministry of Public Health (Ministry of Public Health of Ecuador, 2011).
EMB began in February 2017 and operated until the end of 2019, when the COVID-19 pandemic hit. It progressively expanded to cover the whole country. By the start of 2019, it had covered more than half of the territory; and by the end of 2019, it was operating in all the country. EMB was overseen by health districts. In Ecuador, health districts are responsible for the management of the implementation of health services in a specific geographic area (usually one or several cantons). However, in Ecuador’s two largest cantons, Guayaquil and Quito, this structure is inverted; there are several health districts within these cantons, each responsible for providing healthcare services to different parishes, which are sub-cantonal administrative divisions. These health districts then entrusted the delivery of EMB to their corresponding health centers–– small health facilities that provide a range of basic health services, including general medicine, obstetrics and gynecology, dentistry, vaccination, and laboratory.
At the health center level, EMB operated as follows. First, EMB teams would make home visits to identify vulnerable groups in their community and gather health data. For various communities served by EMB teams, getting there required traveling by foot or other unconventional means (e.g., boat) due to accessibility challenges. Once a patient was identified, they were referred to the closest health center for an appointment with a specialized family physician. In-home appointments were also made when a patient could not commute to the health center. Each specialized primary care physician was assigned a number of vulnerable patients and was responsible for their care and follow-up. After a geographic area had been covered, community surveillance was carried out to monitor new cases in that place (Ministry of Public Health of Ecuador, 2017).
2.3. Theoretical Framework
Our theoretical exploration of the Estrategia Médico del Barrio incorporates a multifaceted analysis to understand its impact on reported health problems, healthcare utilization, and engagement with preventive care, while also considering the unique dynamics within indigenous communities.
Central to our theoretical framework is the Andersen Healthcare Utilization Model, which proposes that peoplés interaction with healthcare is influenced by predisposing characteristics, enabling resources, and perceived needs (Andersen, 1995). In this context, the EMB program, with its diverse team composition, can act as a critical enabling resource, potentially leading to increased diagnostic activities by physicians and, consequently, a rise in reported health issues (Lee & Smith, 2012).
Preventive care highlights the importance of teaching about health and preventing illnesses to lower health issues. Teams with more nurses and community health workers tend to concentrate on these preventive actions, which could result in fewer health problems being reported as exposure to the program increases (Starfield et al., 2005). Evaluating both diagnostic and preventive strategies within the EMB program means we can accommodate potentially opposing results—either an increase in identified health issues due to more thorough diagnostics or a decrease due to more effective preventive measures.
The concept of task shifting enriches this discussion by suggesting that redistributing healthcare tasks among team members can optimize the delivery of healthcare services (McPake & Mensah, 2008). This strategy might enhance the delivery of preventive care, particularly in teams with more non-physician health workers, and influence healthcare utilization patterns towards more preventive and less acute curative care (Baylan et al., 2020; Callaghan et al., 2010; Hampshire et al., 2021).
The integration of indigenous community contexts into the EMB program framework highlights the significance of cultural competence. This program is designed with an understanding of the unique cultural narratives of these communities, emphasizing the necessity of aligning healthcare services with these narratives. The focus on cultural competence suggests that initiatives respectful and mindful of local traditions are more likely to be embraced (Govere & Govere, 2016). Such an approach may facilitate stronger, more sustainable relationships between healthcare providers and indigenous communities, potentially leading to increased engagement with preventive care measures over time (Witmer et al., 1995).
This theoretical background, incorporating the Andersen model–along with concepts related to preventive care, task shifting, and cultural competence–provides a more nuanced understanding of the EMB program’s potential impacts. It highlights the complex interplay between enhanced diagnostic capabilities and preventive care efforts in shaping health outcomes. The inclusion of the unique dynamics within indigenous communities further emphasizes the significance of culturally adapted health interventions in achieving long-term success.
3. Data
3.1. Data Sources
We analyze data from nationally representative surveys with multi-stage, stratified sample designs: the Encuesta de Salud y Nutrición (ENSANUT) and the Encuesta de Condiciones de Vida (ECV) (Freire et al., 2015). ENSANUT was conducted during 2012–13 and again in 2018–19. The first round includes data from 92,500 individuals under 60, capturing information on anthropometrics, blood and urine measurements, tobacco and alcohol use, physical activity, diet (via a 24-hour food recall diary), and healthcare usage and accessibility (Ministry of Public Health of Ecuador, 2012a). The second round involves 168,747 participants, with 82% surveyed between November and December 2018 and the remaining 18% from January to July 2019 (Ministerio de Salud Pública del Ecuador, 2019) (The staggered nature of the survey data collection is one of the components of our identification strategy). The ECV survey, conducted in 2006, collects data from 55,666 individuals. This comprehensive survey addresses various dimensions of household well-being, such as income, expenses, education, health, and access to public services, among other areas (National Institute of Statistics and Census of Ecuador, 2006). (We use the older survey data to test the parallel trends assumption).
Data on EMB’s implementation, including rollout dates as well as the number of primary care physicians and EMB staff involved in the program within each canton, are sourced by contacting health districts through QUIPUX (accessible at https://web.gestiondocumental.gob.ec/). The Document Management System QUIPUX is a web-based computer system that enables the registration, control, circulation, and organization of digital and/or physical documents sent and received in public institutions, ensuring that a government employee addresses citizen’s requests and provides timely responses. It functions similarly to an official email server, where every document is legally binding, ensuring that communications and transactions within the platform adhere to strict legal standards and are enforceable within Ecuador’s legal framework. Moreover, documents within QUIPUX are authenticated using digital official signatures, which carry equal legal responsibility as physical signatures, further reinforcing their validity and legal standing. Government employees in Ecuador are required to respond to requests made through QUIPUX, and before replying, they must consult relevant departments and supervisors to verify the accuracy of the information. We requested this information under the Organic Law of Transparency and Access to Public Information (LOTAIP) [Ley Orgánica de Transparencia y Acceso a la Información Pública] (Ecuador’s version of the US Freedom of Information Act).
Health districts were responsible for overseeing the implementation of the EMB program within their respective jurisdictions (usually one or several cantons). Between July and September 2022, we reached out to all 107 existing health districts in Ecuador to gather the specified data. Responses were obtained from 93 health districts, representing 92% of the cantons documented in both ENSANUT surveys.
The implementation timing of the EMB program varies by canton, with distinct rollout dates for each. The first start date at the canton level is February 7th, 2017, and the last is November 27th, 2022. To illustrate the timing of EMB, Figure 1 shows the months of implementation in the X-axis; it first highlights, in the darker bar on top, the dates when EMB is rolled out; and in the lower, lighter bar, the dates when ENSANUT data is collected. Since both (EMB rollout and ENSANUT data collection) are staggered, what we obtain is a random window of observation in terms of exposure to the EMB program. Additionally, within Quito and Guayaquil, the country’s two largest cantons, EMB begins also on multiple or staggered dates. As a result, we further split these two (larger) cantons based on their respective (sub-cantonal, or parish-level) rollout dates for the analysis.
Figure 1.
Timing of the Estrategia Médico del Barrio (EMB) staggered rollout dates and ENSANUT 2018 data collection visit dates
Notes: Figure shows calendar months in the X-axis and type of activity in the Y-axis. The dark vertical line for ENSANUT 2018–19 represents the median survey visit date (December 24th, 2018). The maximum number of observed EMB treatment exposure was 97 weeks (about 24 months) starting from the first EMB date (February 7th, 2017). For example, two cantons with same start date (say, January 2018) can have either 12 months or 18 months of exposure, depending on whether survey data was collected in December 2018 or June 2019. Dates of baseline ENSANUT 2012 survey not shown.
In the Appendix, we display an illustrative example of a QUIPUX response to our citizen’s request (from one of the coauthors) from Health District 06D02, which oversees healthcare services in the cantons of Alausí, Chunchi, Pallatanga, and Cumandá. In this document, the health district provides the sought-after information in a table, detailing the EMB implementation details (specific EMB rollout dates and staff composition) for each canton under its jurisdiction and for every associated health center. We obtain similarly detailed documents for all cantons and health districts, and subsequently total the number of EMB primary care physicians and staff members to determine the overall count for each canton.
We extend our analysis by controlling for the climate and weather characteristics at the time and location of the surveys. We acquire daily weather data from the Climate Prediction Center (CPC) of the National Oceanic and Atmospheric Administration (NOAA). We calculate the weighted mean temperature and precipitation for each parish by extracting raster values within the boundaries of each parish. These results are then matched with the individual level survey dates to capture the exact maximum and minimum temperature and precipitation for each parish for the dates of, and seven days prior to, when the ENSANUT surveys are being conducted. The temperature data use a global gridded GTS format with a resolution of 0.5° × 0.5°, and the precipitation data are gauge-based with the same resolution (Chen et al., 2008).
Because all of the data are publicly available (including the household surveys: https://www.ecuadorencifras.gob.ec/salud-salud-reproductiva-y-nutricion/), the present study is not considered “human subjects research.” All replication materials are available at: https://doi.org/10.7910/DVN/NXWINX
3.2. Dependent and treatment variables
We focus on three primary dependent variables in our analysis. First, we examine the reported health problem in the 30 days prior to the ENSANUT survey dates. Specifically, the question asked is:
Did [person j] in the last 30 days [from … to…] have any illness, accident, burn, toothache, earache, or any other discomfort, even if it was temporary?
Second, we use healthcare utilization as a metric to address the reported health problem. The specific question posed is:
What did [person j] do as the first action to solve the (health problem)? a. Visited a hospital, dispensary, health center, or sub-center; consulted a doctor, healer, etc.? b. Got care at home from a doctor, nurse, healer, etc.? c. Self-medicated? d. Had to be admitted to a hospital, clinic, etc.? e. Did nothing.
We recode this variable into a dichotomous format where =0 signifies ‘did nothing’ or ‘self-medicated,’ and =1 represents all other responses. Note that by survey design, this question is not asked for those who do not have a health problem, which consequently reduces the total observations included in our models by 70%.
The third outcome pertains to preventive care. The specific question in the survey is as follows:
Now I am going to ask you questions about preventive care: In the last 30 days (from…to…), were you checked by a psychologist, dentist, healer, apothecary, or massage therapist? Or did the neighborhood doctor visit you at home? Or did you receive any preventive services such as: vaccinations, well-child check-ups, blood pressure checks, dental check-ups, etc.?
It is worth noting that this question is presented only to respondents who report no illnesses in the 30 days prior to the survey, which results in a smaller sample size.
Next, we operationalize the treatment variable in four ways. First, a dummy variable (EMB=1) indicates if an individual lives in a canton that at any point has been exposed to EMB when we can observe outcomes (i.e., prior to the ENSANUT survey date). Second, we measure treatment at the intensive margin via a dummy variable (=1) if the canton has been exposed to EMB for a longer time than the median number of potential weeks covered (i.e., more than 18.4 weeks out of the potential 97.2 weeks of program treatment that we could observe). A third and a fourth dummy variable (=1) are employed if the number of EMB primary care physicians, and staff members (nurses and community health workers), per 1000 population, respectively, exceeded the median value. Specifically, we obtain the data on the number of EMB team staff members in each canton (as mentioned before, via QUIPUX) and the population projections at the canton level from the National Statistics and Census Institute. We use the median for longerEMB, moreEMBdocs, moreEMBstaff to avoid the known issue of outliers having excessive weight on the mean. Other papers also use median values for similar dichotomizations (Altman & Royston 2006).
3.3. Control variables
Demographic controls, including age, gender, ethnicity, years of education, and log household per capita income, are sourced from the 2012 and 2018 ENSANUT surveys. Income is self-reported and reflects income from the month prior to the survey, lowering the risks of recall bias. It encompasses earnings from business ownership and salaries, as well as non-employment income such as investments, retirement pensions, remittances, and government cash transfers. Additionally, we include the average minimum and maximum temperature and precipitation for the seven days prior to the survey date, including the survey date (8 days in total). The weather variables, obtained from NOAA, are used to account for regional and seasonal weather characteristics that could affect individuals’ health status and utilization of health services. The inclusion of weather controls is based on evidence suggesting transient weather conditions can influence self-reported health outcomes (Frijters et al., 2020). For instance, studies indicate that weather can affect the likelihood of reporting poor mental health (Li et al., 2020). It is also conceivable that weather conditions might impact individuals’ willingness to travel to healthcare facilities; for example, rainy weather could deter visits, thereby affecting reported access to both curative and preventive care.
4. Identification Strategy and Econometric Methods
We observe outcomes at the individual level collected through nationally-representative ENSANUT surveys in 2012–13 and 2018–19. The treatment is at the cantón (or canton) level; the second highest administrative division level in Ecuador, below province. (As mentioned, additionally, in the largest cities of Quito and Guayaquil, the treatment can be differential at the sub-cantonal or parish level). The EMB program is rolled out in stages across different cantons, which allow us to evaluate the effects of EMB using a differences-in-differences (DD) approach.
For identification, we exploit the natural experiment created by the random combination of the different rollout dates of EMB start and the different data collection dates for the endline survey (ENSANUT 2018–19). This natural experiment generates geographic and temporal random variation in terms of exposure to treatment. Thus, the first DD specification is:
| [1] |
where Yit indicates the outcome of individual i at time t, POST is an indicator (=1) if the time is 2018–19, and POST=0 if time is 2012–13; EMBit is a dummy variable (=1) indicating any presence of Estrategia Médico Barrio at the canton level at time t; and where the coefficient of interest is the interaction α3. Second, as a robustness test, we use an alternative DD specification as follows:
| [2] |
where the treatment is measured by a dummy variable (=1) if the temporal coverage of EMB was above the national median (i.e., over 18.4 weeks out of the 97 observable weeks in the analytical sample). The coefficient β3 is the interaction of interest. The maximum potential number of weeks is defined as the difference between the last date of the endline survey and the first date of the EMB rollout (i.e., 97.2 weeks).
A third DD specification is of the form:
| [3] |
where we measure treatment by a dummy variable (=1) if the number of EMB primary care physicians per 1000 population at the canton level (moreEMBdocsit) at the time of the endline survey is above the national median; and where the coefficient γ3 is the interaction of interest. The rationale for this specification is that additional human resources for health in proportion to the total population may result in increased potential impact. We repeat this specification using EMB staff as well (i.e., nurses and community health workers per 1000 population at the canton level).
The triple differences (DDD) specifications use an additional dummy variable (=1) for cantons where indigenous self-identification is above the national mean, as done in related research (Galárraga & Harris, 2021). All DDD models are fully interacted (i.e., they include main effects and all second-order interactions). Moreover, all specifications (DD and DDD) include standard errors that account for the complex, multi-stage nature of the nationally representative data, using primary sample unit weights via the survey regression (svy reg) commands in Stata.
We test for the parallel trends assumption using data from 2012 and 2006. The empirical test proceeds using the same DD framework as presented above, but instead of having a “post” variable indicator for 2018, we have a “pre” indicator variable for 2006. The key is to observe the pre-treatment parallel trends. If the lines for the two groups run parallel before the intervention, it suggests that, in the absence of the treatment, the groups would have continued on similar paths.
5. Results
5.1. Summary statistics
Table 1 displays the summary statistics for the primary variables we use in our analysis, including outcomes, treatment variables, control variables, and intermediate variables. Health outcomes from 2012 to 2018 exhibit significant improvement. In 2012, 40 percent of individuals report experiencing health problems in the 30 days prior to the ENSANUT survey, a figure which decreases to 23 percent in 2018. Moreover, among these individuals, 52 percent in 2012 receive curative care to address the reported health problem, and this proportion increases to 68 percent in 2018. The use of preventive healthcare remains relatively stable, with only a slight decline from 9 percent to 8 percent.
Table 1:
Summary statistics
| Variable | Baseline 2012–2013 | Endline 2018–2019 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| N | Mean | SD | Min | Max | N | Mean | SD | Min | Max | |
| Dependent variables | ||||||||||
| Health problem in past 30 days | 73,691 | 0.40 | 0.49 | 0 | 1 | 138,490 | 0.23 | 0.42 | 0 | 1 |
| Received care to address health problem in past 30 days | 29,749 | 0.52 | 0.50 | 0 | 1 | 31,216 | 0.68 | 0.47 | 0 | 1 |
| Received preventive care in past 30 days | 50,915 | 0.09 | 0.29 | 0 | 1 | 107,287 | 0.08 | 0.27 | 0 | 1 |
| Treatment variables | ||||||||||
| Estrategia Médico del Barrio (EMB=1) | 138,490 | 0.57 | 0.49 | 0 | 1 | |||||
| High EMB temporal coverage | 138,490 | 0.49 | 0.50 | 0 | 1 | |||||
| High number of EMB primary care physicians per 1000 | 138,490 | 0.28 | 0.45 | 0 | 1 | |||||
| High number of EMB mobile health team staff per 1000 | 138,490 | 0.28 | 0.45 | 0 | 1 | |||||
| Control variables | ||||||||||
| Age | 64,673 | 28.33 | 17.87 | 5 | 99 | 120,762 | 31.70 | 19.74 | 5 | 99 |
| Female | 64,673 | 0.52 | 0.50 | 0 | 1 | 120,762 | 0.52 | 0.50 | 0 | 1 |
| Indigenous | 64,673 | 0.09 | 0.29 | 0 | 1 | 120,762 | 0.11 | 0.31 | 0 | 1 |
| Black | 64,673 | 0.04 | 0.19 | 0 | 1 | 120,762 | 0.04 | 0.21 | 0 | 1 |
| Years of education | 64,673 | 7.08 | 4.62 | 0 | 20 | 120,762 | 9.10 | 5.01 | 0 | 23 |
| Log household per capita income | 64,673 | 4.55 | 1.71 | 0 | 13.8155 | 120,762 | 4.95 | 1.40 | 0 | 14.51 |
| Average max temperature (°C) in past 7 days | 62,347 | 23.74 | 4.52 | 14.12 | 33.9337 | 115,942 | 24.43 | 4.34 | 14.11 | 32.69 |
| Average min temperature (°C) in past 7 days | 62,347 | 14.46 | 6.24 | 2.96 | 23.2282 | 115,942 | 16.78 | 6.07 | 4.12 | 25.11 |
| Average precipitation (mm) in past 7 days | 62,347 | 1.57 | 2.33 | 0 | 14.0013 | 115,942 | 3.22 | 2.97 | 0.00 | 19.73 |
| Intermediate variables | ||||||||||
| ENSANUT survey date (m/d/y) | 138,490 | 1/4/19 | 72.49 | 11/10/18 | 7/17/19 | |||||
| EMB implementation date (m/d/y) | 138,490 | 12/30/18 | 378.87 | 2/7/17 | 11/27/22 | |||||
| Actual number of weeks covered by EMB | 138,490 | −1.54 | 55.32 | −204.72 | 97.16 | |||||
| EMB primary care physicians per 1000 (if EMB=1) | 79,185 | 0.11 | 0.10 | 0.00 | 0.93 | |||||
| EMB mobile health team staff per 1000 (if EMB=1) | 79,185 | 0.88 | 0.84 | 0.02 | 6.17 | |||||
Notes: The dependent variables are observed for individuals surveyed in the 2012 and 2018 National Health and Nutrition Survey (ENSANUT). The table presents data for the 167 cantons for which we obtained official information about Estrategia Médico del Barrio (EMB) rollout dates as well as EMB mobile health teams’ number of personnel and composition; these 167 cantons correspond to 92% of the 182 cantons with data available in both surveys. EMB=1 flags individuals living in cantons were the EMB start date was before the ENSANUT survey date.
Regarding treatment variables, 57 percent of the sample in the 2018 ENSANUT are surveyed after the EMB program has been implemented in their respective cantons. About half (49 percent) of the respondents reside in cantons where the duration of EMB coverage exceeds the median. Additionally, 28 percent live in cantons where the number of EMB primary care physicians per 1000 population is above the median, and a similar 28 percent are from cantons where the number of EMB staff members (nurses and community health workers) per 1000 population exceeds the median.
Focusing on the intermediate variables, the earliest 2018 ENSANUT survey date is November 10th, 2018, and the last one is conducted on July 17th, 2019. The earliest EMB start of implementation date is February 7th, 2017, and the most recent is November 27th, 2022. The mean duration of the EMB program’s coverage spans 55.32 weeks, with the maximum duration being 97.2 weeks. The mean number of EMB primary care physicians per 1000 population is 0.11, and the mean number of EMB mobile health team staff per 1000 population is 0.88. These figures reflect only the healthcare personnel specifically assigned to the EMB program and do not represent the overall national physician density. According to WHO data, Ecuador’s physician density is relatively low, with only 2.2 physicians per 1,000 population, which is slightly below the standard of 2.3 physicians per 1,000 population needed to fulfill primary care needs (World Health Organization 2024 data.who.int, n.d.). This density is disproportionately lower in rural and remote areas, where some cantons do not have a single rural primary care physician (Romero-Alvarez et al., 2023).
5.2. Main results and effect mechanisms
Table 2 presents the results of implementing equations [1] and [2] on the three dependent variables (reported health problem, received curative care, and received preventive care). Panel A details the DD effects of EMB (measured by the interaction POST = 1 × EMB = 1), and Panel B examines the impact of EMB on individuals living in cantons where the number of weeks covered is greater than the median (represented by the interaction POST = 1 × longerEMB = 1). Results from column (3), which uses the full set of demographic and weather/climate controls, show a 4.69 percentage point reduction in the probability of reporting a health problem due to EMB (from a baseline of 35.84% in the untreated group), suggesting a positive impact of EMB on health outcomes. When focusing on cantons with EMB exposure longer than the median number of weeks, the effect of longerEMB is to decrease health problem reporting by 4.13 percentage points. We find no significant effects, however, of EMB on either curative or preventive care. This indicates that while the program reduces the number of health problems reported, it does not increase the use of curative or preventive care services.
Table 2:
DD effects of Estrategia Médico del Barrio (EMB)
| Outcome | (1) | (2) | (3) |
|---|---|---|---|
| Panel A: POST = 1 × EMB = 1 | |||
| Health problem in past 30 days | −0.0448*** (0.0136) |
−0.0455*** (0.0133) |
−0.0469*** (0.0124) |
| N | 212,181 | 185,435 | 178,289 |
| Received care to address health problem in past 30 days | −0.00794 (0.0177) |
−0.00682 (0.0187) |
−0.00453 (0.0185) |
| N | 60,965 | 49,733 | 47,880 |
| Received preventive services in past 30 days | 0.00445 (0.00687) |
0.000245 (0.00670) |
0.000714 (0.00660) |
| N | 158,202 | 141,954 | 136,448 |
| Panel B: POST = 1 × longerEMB = 1 | |||
| Health problem in past 30 days | −0.0403*** (0.0134) |
−0.0407*** (0.0130) |
−0.0413*** (0.0122) |
| N | 212,181 | 185,435 | 178,289 |
| Received care to address health problem in past 30 days | −0.0214 (0.0176) |
−0.0215 (0.0188) |
−0.0182 (0.0186) |
| N | 60,965 | 49,733 | 47,880 |
| Received preventive services in past 30 days | −0.000399 (0.00676) |
−0.00503 (0.00659) |
−0.00511 (0.00650) |
| N | 158,202 | 141,954 | 136,448 |
| Demographic controls | No | Yes | Yes |
| Weather and climate controls | No | No | Yes |
Notes: Demographic controls (in column 2) include age, gender, ethnicity, years of education, and log household per capita income. Weather and climate controls (in column 3) include average maximum and minimum temperature (°C) and precipitation (mm) in past 7 days. POST=1 indicates individuals surveyed in the 2018 ENSANUT, while POST=0 refers to those surveyed in the 2012 ENSANUT. EMB=1 flags individuals from cantons where the EMB program began before the ENSANUT survey was conducted. longerEMB=1 represents individuals living in cantons with EMB exposure longer than the median number of weeks. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
We explore causal mechanisms by looking at the impact of having more EMB primary care physicians as well as mobile health team staff (nurses and community health workers). Table 3 presents these results from running equation [3] on the selected health outcomes. Column (3) of panel A shows that individuals in cantons with more EMB primary care physicians see a 4.86 percentage point increase in reported health problems (from a baseline of 33.54% in the untreated group) and a 5.06 percentage point decrease in curative care received over the past 30 days (from a baseline of 41.99%). Additionally, individuals in cantons with more EMB team staff report a 4.68 percentage point increase in health issues (from a baseline of 33.34%) and a 1.52 percentage point increase in preventive care (from a baseline of 8.58%). The higher reporting of health problems can be attributed to increased diagnoses by the visiting EMB teams. Despite this, there is not a corresponding increase in curative treatments within the 30-day timeframe, potentially due to the time needed to access treatment. The increase in preventive care is likely because non-specialists, especially community health workers, can provide these services.
Table 3:
DD Effects of EMB personnel and composition
| Outcome | (1) | (2) | (3) |
|---|---|---|---|
| Panel A : POST = 1 × moreEMBdocs = 1 | |||
| Health problem in past 30 days | 0.0523*** (0.0143) |
0.0437*** (0.0142) |
0.0486*** (0.0135) |
| N | 212,181 | 185,435 | 178,289 |
| Received care to address health problem in past 30 days | −0.0376* (0.0198) |
−0.0507** (0.0225) |
−0.0506** (0.0227) |
| N | 60,965 | 49,733 | 47,880 |
| Received preventive services in past 30 days | 0.0132* (0.00755) |
0.0128* (0.00737) |
0.0116 (0.00727) |
| N | 158,202 | 141,954 | 136,448 |
| Panel A : POST = 1 × moreEMBstaff = 1 | |||
| Health problem in past 30 days | 0.0539*** (0.0152) |
0.0435*** (0.0148) |
0.0468*** (0.0141) |
| N | 212,181 | 185,435 | 178,289 |
| Received care to address health problem in past 30 days | −0.0145 (0.0206) |
−0.0319 (0.0232) |
−0.0297 (0.0236) |
| N | 60,965 | 49,733 | 47,880 |
| Received preventive services in past 30 days | 0.0158** (0.00775) |
0.0158** (0.00764) |
0.0152** (0.00757) |
| N | 158,202 | 141,954 | 136,448 |
| Demographic controls | No | Yes | Yes |
| Weather and climate controls | No | No | Yes |
Notes: Demographic controls (in column 2) include age, gender, ethnicity, years of education, and log household per capita income. Weather and climate controls (in column 3) include average maximum and minimum temperature (°C) and precipitation (mm) in past 7 days. POST=1 indicates individuals surveyed in the 2018 ENSANUT, while POST=0 refers to those surveyed in the 2012 ENSANUT. Standard errors in parentheses. moreEMBdocs=1 represents individuals living in cantons where the number of EMB primary care physicians per 1000 population exceeded the median. moreEMBstaff=1 represents those individuals living in cantons where the number of EMB mobile health staff members per 1000 population (nurses and community health workers) exceeded the median. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
5.3. Triple-differences results
In Table 4, we examine the effects on individuals residing in cantons with a high concentration of indigenous people using a DDD approach. In column (3) of Panel A, the DDD coefficient indicates that EMB increases the probability of reporting a health problem by 11.9 percentage points (from a baseline of 34.95% in the untreated group). Using the high EMB temporal coverage dummy in Panel B, the DDD coefficient shows a 9.81 percentage point increase in the probability of reporting a health problem (from a baseline of 35.43) and a 2.62 percentage point increase in the probability of receiving preventive care (from a baseline of 9.52). It is noteworthy that the effects of EMB on reported health problems in cantons with a larger indigenous population are positive and substantial. This can be attributed to the nature of the program itself, which targets vulnerable populations lacking access to healthcare services—a common issue in many rural and indigenous communities in Ecuador. Additionally, while there is no increase in curative care, likely due to the insufficient observation time for EMB teams to arrange medical appointments with primary care physicians, there is a significant increase in preventive care engagement. One potential explanation for this is because promoting health prevention through community health activities is one of the primary roles of EMB teams.
Table 4:
DDD effects of Estrategia Médico del Barrio (EMB)
| Outcome | (1) | (2) | (3) |
|---|---|---|---|
| Panel A: POST = 1 × EMB = 1 × ind = 1 | |||
| Health problem in past 30 days | 0.108*** (0.0268) |
0.106*** (0.0272) |
0.119*** (0.0269) |
| N | 212,181 | 185,435 | 178,289 |
| Received care to address health problem in past 30 days | 0.00959 (0.0395) |
−0.0000527 (0.0401) |
−0.0106 (0.0404) |
| N | 60,965 | 49,733 | 47,880 |
| Received preventive services in past 30 days | 0.0160 (0.0137) |
0.0175 (0.0129) |
0.0138 (0.0130) |
| N | 158,202 | 141,954 | 136,448 |
| Panel B: POST = 1 × longerEMB = 1 × ind = 1 | |||
| Health problem in past 30 days | 0.0860*** (0.0269) |
0.0847*** (0.0274) |
0.0981*** (0.0273) |
| N | 212,181 | 185,435 | 178,289 |
| Received care to address health problem in past 30 days | 0.0233 (0.0396) |
0.0154 (0.0403) |
0.00232 (0.0408) |
| N | 60,965 | 49,733 | 47,880 |
| Received preventive services in past 30 days | 0.0289** (0.0134) |
0.0307** (0.0126) |
0.0262** (0.0128) |
| N | 158,202 | 141,954 | 136,448 |
| Demographic controls | No | Yes | Yes |
| Weather and climate controls | No | No | Yes |
Notes: Demographic controls (in column 2) include age, gender, ethnicity, years of education, and log household per capita income. Weather and climate controls (in column 3) include average maximum and minimum temperature (°C) and precipitation (mm) in past 7 days. POST=1 indicates individuals surveyed in the 2018 ENSANUT, while POST=0 refers to those surveyed in the 2012 ENSANUT. EMB=1 flags individuals from cantons where the EMB program began before the ENSANUT survey was conducted. longerEMB=1 represents individuals living in cantons with EMB exposure longer than the median number of weeks. ind=1 represents cantons where the concentration of indigenous populations is larger than the mean at the national level. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
5.4. Parallel trends test
We conduct a parallel-trends test to check the validity of the assumptions underlying our DD estimation. The parallel-trends test investigates whether, prior to the EMB program implementation, the outcome variables exhibit similar trends across treatment and comparison groups. We specifically test the interaction term (PRE = 1 × EMB = 1) for our three dependent variables. The results of this analysis, presented in Table 5, indicate that there is no significant effect of the interaction term on any of the outcomes. This lack of significance suggests that the outcome trends are indeed parallel before the EMB program’s initiation. This finding provides additional support for the interpretation that the changes we observe after the program’s start are likely attributable to the EMB intervention itself, rather than to pre-existing trends.
Table 5:
Parallel trends test of EMB effects
| Health problem in past 30 days | Received care to address health problem | Received preventive services | |
|---|---|---|---|
| DD/ DDD interaction | (1) | (2) | (3) |
| Panel A: Difference-in-difference | |||
| PRE = 1 × EMB = 1 | −0.00857 (0.0158) |
0.00955 (0.0167) |
−0.00231 (0.0102) |
| N | 122,308 | 55,954 | 76,502 |
| PRE = 1 × longEMB = 1 | −0.00182 (0.0155) |
0.00587 (0.0166) |
0.00327 (0.0102) |
| N | 122,308 | 55,954 | 76,502 |
| Panel B: Triple-differences | |||
| PRE = 1 × EMB = 1 × ind = 1 | 0.0113 (0.0307) |
−0.0301 (0.0371) |
−0.000122 (0.0246) |
| N | 122,308 | 55,954 | 76,502 |
| PRE = 1 × longEMB = 1 × ind = 1 | −0.0268 (0.0306) |
−0.0244 (0.0374) |
0.000118 (0.0246) |
| N | 122,308 | 55,954 | 76,502 |
| Demographic controls | No | No | No |
| Weather and climate controls | No | No | No |
Notes: PRE=1 indicates individuals surveyed in the 2006 ECV, while PRE=0 refers to those surveyed in the 2012 ENSANUT. EMB=1 flags individuals from cantons where the EMB program began before the ENSANUT survey was conducted. longEMB=1 represents individuals living in cantons with EMB exposure longer than the median number of weeks. Ind=1 represents cantons where the concentration of indigenous populations is larger than the mean at the national level. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
5.5. Sensitivity analyses and robustness checks
This section presents sensitivity analyses and additional robustness checks. First, we present results from using a linear intensity variable (instead of the dummy indicator) to measure temporal exposure to the program as well as a linear intensity variable to measure the number of EMB mobile healthcare traveling team members per 1000 population. The variable to measure the time-related intensive margin (pweekscov) has as the denominator the maximum potential number of weeks that a canton can be exposed to EMB (counting from the first EMB rollout date to the latest ENSANUT survey data collection date), and the numerator is the actual number of weeks that the canton is exposed to EMB. [Note that these specifications of exposure intensity are similar in nature to that of the proportion of 18 to 20-year-olds that can legally drink in state s in time t in a panel fixed-effects evaluation of the minimum drinking age on morbidity and mortality in the US (Carpenter & Dobkin, 2011)]. Second, we present results from using only cantons that eventually adopt EMB as the comparison group (instead of all cantons in the analytical sample) in regressions evaluating the intensive margins with the same treatment dummy variables as before (i.e., longerEMB, moreEMBdocs, and moreEMBstaff as dichotomous treatments). Third, we discuss results from DDD regressions where we change the highly indigenous dummy (ind=1) to a linear variable of indigenous self-identification at the canton level (indrace, continuous variable); and also include an additional combined outcome variable for the intensity of treatment (time duration and personnel). Finally, we estimate the effects of longer exposure to EMB as well as more EMB primary care physicians and staff (nurses and community health workers) using the mean as the cutoff value for defining the treatment variables instead of the median.
5.5.1. Using linear variable for program coverage and intensity variable combining program coverage with number of EMB staff members per 1000 population
Table 6 illustrates the impact of EMB on reported individual-level health issues, employing a linear variable for the proportion of weeks covered by EMB (pweekscov) and a exposure variable of combined intensity (intensity) that accounts for the proportion of program coverage and the number of EMB staff per 1000 population (i.e., intensity = pweekscov × staffper1000pop). Column (1) of Panel A reveals a decrease of 4.69 percentage points in reported health problems (from a baseline of 34.17% in the group not exposed to EMB), replicating prior results. We then see a decrease of 5.72 percentage points in reported health problem when using the pweekscov linear exposure variable. This effect is interpretable as the aggregate impact of exposure to the maximum possible 97.2 weeks of coverage. We also observe an increase of 10 percentage points when applying the intensity variable. Columns (2) and (3) of Panel A indicate no significant DD effects of EMB on curative and preventive care with the linear EMB coverage variables. Panel B displays the DDD effects of EMB on individuals residing in cantons with a high indigenous population. Column (1) shows an 11.9 percentage point increase in reported health problem using the EMB=1 exposure, and a 21.5 percentage point positive impact on reported health problems using the linear pweekscov exposure variable. The intensity variable, however, does not yield any significant effects. Likewise, columns (2) and (3) reveal no significant impacts on curative and preventive care.
Table 6:
Robustness checks using linear variable for program coverage and in of total potential weeks covered by EMB
| Health problem in past 30 days | Received care to address health problem | Received preventive services | |
|---|---|---|---|
| DD/ DDD interaction | (1) | (2) | (3) |
| Panel A: Difference-in-difference | |||
| POST = 1 × EMB = 1 | −0.0469*** (0.0124) |
−0.00453 (0.0185) |
0.000714 (0.00660) |
| N | 178,289 | 47,880 | 136,448 |
| POST = 1 × pweekscov | −0.0572** (0.0286) |
−0.0367 (0.0418) |
0.00637 (0.0139) |
| N | 177,890 | 47,720 | 136,167 |
| POST = 1 × intensity | 0.100*** (0.0199) |
−0.0150 (0.0257) |
0.0118 (0.0111) |
| N | 96,759 | 25,287 | 74,578 |
| Panel B: Triple-differences | |||
| POST = 1 × EMB = 1 × ind = 1 | 0.119*** (0.0269) |
−0.0106 (0.0404) |
0.0138 (0.0130) |
| N | 178,289 | 47,880 | 136,448 |
| POST = 1 × pweekscov × ind = 1 | 0.215*** (0.0534) |
−0.0430 (0.0866) |
0.0308 (0.0260) |
| N | 177,890 | 47,720 | 136,167 |
| POST = 1 × intensity × ind = 1 | −0.0527 (0.0506) |
0.00852 (0.0906) |
−0.0129 (0.0248) |
| N | 96,759 | 25,287 | 74,578 |
| Demographic controls | Yes | Yes | Yes |
| Weather and climate controls | Yes | Yes | Yes |
Notes: Demographic controls include age, gender, ethnicity, years of education, and log household per capita income. Weather and climate controls include average maximum and minimum temperature (°C) and precipitation (mm) in past 7 days. POST=1 indicates individuals surveyed in the 2018 ENSANUT, while POST=0 refers to those surveyed in the 2012 ENSANUT. EMB=1 flags individuals from cantons where the EMB program began before the ENSANUT survey was conducted. pweekscov is a linear variable that represents the proportion of weeks covered by EMB. intensity is an exposure variable of combined program and exposure intensity that accounts for the proportion of program coverage and the number of EMB staff per 1000 population. ind=1 represents cantons where the concentration of indigenous populations is larger than the mean at the national level. Standard errors in parentheses. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
5.5.2. Early vs. late EMB adopters
In Table 7, we present a second set of robustness checks using a modified analytical sample: using only those cantons that eventually implement EMB as the comparison group (that is, rather than including all cantons as the comparison group, we now include only those not exposed to EMB prior to the ENSANUT survey). This approach is similar to studies featuring staggered-entry or stepped-wedge designs, which compare early adopters of a program with those adopting it later. In Panel A, columns (1) and (2) show no effect of early EMB exposure on reported health problems and access to curative care. However, column (3) reveals that individuals in cantons with earlier EMB exposure are 2.64 percentage points less likely to receive preventive care in the preceding 30 days compared to those in cantons exposed later. This suggests that EMB’s effectiveness wanes over time or that earlier beneficiaries of EMB do not report receiving preventive care, potentially because the care is received prior to the 30-day timeframe referenced in the survey. In contrast, Panel B’s column (3) indicates a significant and positive impact of 6.76 percentage points for early EMB exposure in cantons with a large indigenous population, meaning that the program’s benefits persist in these areas, effectively extending preventive care coverage over time.
Table 7:
Sensitivity analysis comparing early vs. late EMB adopters
| Health problem in past 30 days | Received care to address health problem | Received preventive services | |
|---|---|---|---|
| DD/ DDD interaction | (1) | (2) | (3) |
| Panel A: Difference-in-difference | |||
| POST = 1 × longerEMB = 1 | −0.00913 (0.0241) |
−0.0484 (0.0339) |
−0.0264** (0.0118) |
| N | 97,158 | 25,447 | 74,859 |
| POST = 1 × moreEMBdocs = 1 | 0.0930*** (0.0155) |
−0.0502* (0.0259) |
0.0149* (0.00820) |
| N | 97,158 | 25,447 | 74,859 |
| POST = 1 × moreEMBstaff = 1 | 0.0876*** (0.0159) |
−0.0292 (0.0267) |
0.0187** (0.00853) |
| N | 97,158 | 25,447 | 74,859 |
| Panel B: Triple-differences | |||
| POST = 1 × longerEMB = 1 × ind = 1 | −0.0489 (0.0510) |
0.0376 (0.103) |
0.0676** (0.0287) |
| N | 97,158 | 25,447 | 74,859 |
| POST = 1 × moreEMBdocs = 1 × ind = 1 | −0.119*** (0.0399) |
0.0744 (0.0663) |
−0.0299* (0.0178) |
| N | 97,158 | 25,447 | 74,859 |
| POST = 1 × moreEMBstaff = 1 × ind = 1 | −0.0797** (0.0379) |
0.0901 (0.0591) |
−0.0366** (0.0154) |
| N | 97,158 | 25,447 | 74,859 |
| Demographic controls | Yes | Yes | Yes |
| Weather and climate controls | Yes | Yes | Yes |
Notes: The sample includes only individuals in cantons where EMB was eventually implemented as the comparison group (rather than including all cantons, including those not exposed to EMB prior to the ENSANUT survey). Demographic controls include age, gender, ethnicity, years of education, and log household per capita income. Weather and climate controls include average maximum and minimum temperature (°C) and precipitation (mm) in past 7 days. longerEMB=1 represents individuals living in cantons with EMB exposure longer than the median number of weeks. moreEMBdocs=1 represents individuals living in cantons where the number of EMB primary care physicians per 1000 population exceeded the median. moreEMBstaff=1 represents those individuals living in cantons where the number of EMB mobile health staff members per 1000 population (nurses and community health workers) exceeded the median. ind=1 represents cantons where the concentration of indigenous populations is larger than the mean at the national level. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
In Panel A, columns (1) and (3) reveal that in cantons with a higher presence of EMB primary care physicians and staff, such as nurses and community health workers, there is an increased likelihood of health issues being reported and of individuals receiving preventive care. This increase could be attributed to enhanced diagnostic activity by healthcare providers or a higher propensity among individuals to report health issues if they believe that effective interventions are available. Conversely, Panel B presents a different scenario in cantons with high indigenous populations, where individuals are less likely to report health problems and access preventive care. This suggests that in these regions, service accessibility remains a challenge, and simply having a greater number of physicians and staff does not necessarily translate to improved health service delivery to these vulnerable communities. The findings indicate that the successful implementation of EMB in areas with high indigenous populations may require more than just an increase in healthcare personnel.
5.5.3. Linear definition of indigenous concentration and combined intensity program exposure
Table 8 presents the DDD results when employing continuous indigenous self-identification, our findings are consistent with the main results. The DDD coefficient for reported health issues (POST=1×pweekscov×indrace) is 0.446 (significant at p<0.05), suggesting that residing in a canton with the more extensive EMB coverage and more indigenous population has a highly noticeable effect on health problem reporting.
Table 8:
Robustness check using linear indigenous concentration and combined intensity program exposure
| Health problem in past 30 days | Received care to address health problem | Received preventive services | |
|---|---|---|---|
| DDD interaction | (1) | (2) | (3) |
| POST = 1 × pweekscov × indrace | 0.446** (0.188) |
−0.121 (0.288) |
0.0937 (0.0829) |
| N | 177,890 | 47,720 | 136,167 |
| POST = 1 × docsper1000pop × indrace | −0.244*** (0.0700) |
0.163 (0.105) |
−0.0532** (0.0239) |
| N | 97,158 | 25,447 | 74,859 |
| POST = 1 × staffper1000pop × indrace | −0.844 (0.640) |
2.141** (0.941) |
−0.360 (0.298) |
| N | 97,158 | 25,447 | 74,859 |
| Demographic controls | Yes | Yes | Yes |
| Weather and climate controls | Yes | Yes | Yes |
Notes: Demographic controls include age, gender, ethnicity, years of education, and log household per capita income. Weather and climate controls include average maximum and minimum temperature (°C) and precipitation (mm) in past 7 days. POST=1 indicates individuals surveyed in the 2018 ENSANUT, while POST=0 refers to those surveyed in the 2012 ENSANUT. pweekscov, docsper1000pop, staffper1000pop, and indrace are linear variables. pweekscov represents the proportion of weeks covered by EMB; docsper1000pop represents the number of primary care physicians per 1000 population; staffper1000pop represents the number of EMB staff members (i.e., nurses and community health workers) per 1000 population; and indrace represents indigenous self-identification proportion at the canton level. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
Additionally, we integrate a more nuanced exposure variable reflecting combined intensity (i.e., docsper1000pop × pweekscov and staffper1000pop × pweekscov) as an alternative DDD strategy. This approach employs the actual proportion of self-identified indigenous population, as opposed to a binary indicator for high indigenous concentration. For the continuous variable concerning primary care physicians, the DDD coefficient (POST=1×docsper1000pop×indrace) indicates a decrease in reported health problems by −0.244 (significant at p<0.01), meaning that the response slope decreases when both docsper1000pop and indrace increase; suggesting, again, that the actual number of primary care physicians per 1000 population, alone, does not seem to make a difference in reducing health problems. Similarly, we find a smaller, yet still significant, decrease in reported receipt of preventive care by −0.0532 (significant at p<0.05). Finally, using the continuous variable for staff members, the DDD coefficient (POST=1×staffper1000pop×indrace) is 2.141 (p<0.05) for curative care.
5.5.4. EMB effects using the mean as the cutoff value for defining the treatment variables
We use the mean instead of the median as the cutoff for defining the treatment variables for longer exposure to EMB, as well as more EMB primary care physicians and staff (i.e., longerEMB, moreEMBdocs, and moreEMBstaff as dichotomous treatments). Results in Table 9 demonstrate that when compared to the initial estimations using the median as the cutoff, as shown in Tables 2 and 3, the direction of the coefficients remains unchanged while the magnitude of the coefficients varies minimally. Furthermore, the statistical significance of the coefficients is consistent across all outcomes, with the exception of the impact of a higher number of EMB staff on preventive care utilization, which has now become statistically insignificant.
Table 9:
Robustness check using the mean as the cutoff value to define the treatment variables
| Health problem in past 30 days | Received care to address health problem | Received preventive services | |
|---|---|---|---|
| DD interaction | (1) | (2) | (3) |
| POST = 1 × longerEMB = 1 | −0.0434*** (0.0124) |
−0.00345 (0.0186) |
0.00117 (0.00665) |
| N | 178,289 | 47,880 | 136,448 |
| POST = 1 × moreEMBdocs = 1 | 0.0836*** (0.0151) |
−0.0755*** (0.0268) |
0.0165** (0.00818) |
| N | 178,289 | 47,880 | 136,448 |
| POST = 1 × moreEMBstaff = 1 | 0.0735*** (0.0156) |
−0.0364 (0.0255) |
0.00103 (0.0101) |
| N | 178,289 | 47,880 | 136,448 |
| Demographic controls | Yes | Yes | Yes |
| Weather and climate controls | Yes | Yes | Yes |
Notes: Demographic controls include age, gender, ethnicity, years of education, and log household per capita income. Weather and climate controls include average maximum and minimum temperature (°C) and precipitation (mm) in past 7 days. longerEMB=1 represents individuals living in cantons with EMB exposure longer than the mean number of weeks. moreEMBdocs=1 represents individuals living in cantons where the number of EMB primary care physicians per 1000 population exceeded the mean. moreEMBstaff=1 represents those individuals living in cantons where the number of EMB mobile health staff members per 1000 population (nurses and community health workers) exceeded the mean. ind=1 represents cantons where the concentration of indigenous populations is larger than the mean at the national level. Standard errors in parentheses:
p<0.10,
p<0.05,
p<0.01.
5.6. Limitations
This paper is not without limitations. There is a temporal gap between the pre and post periods in our analysis. The pre-period data predates the program rollout by a significant amount of time (six years), whereas the post-period data is collected shortly after the program’s implementation (in some cantons). This extended time difference between the two periods increases the potential for various external factors to change over time, and these changes may differ between control and treatment cantons. This aspect could introduce complexities in assessing the parallel trends assumption. Additionally, our ability to assess the plausibility of this assumption is limited due to the availability of only two pre-period data points, with the first one being considerably distant from the program’s implementation.
Furthermore, the dependent variables rely on self-reported data, making them susceptible to potential reporting errors that could introduce measurement inaccuracies. For example, when individuals report instances of illness, there is a possibility that they may inadvertently include healthcare utilization events rather than strictly reporting actual health problems. Nevertheless, we do use large, nationally-representative surveys in which there are no clear, a priori, incentives to misreport utilization or other health outcomes. Additionally, existing evidence supports the idea that self-reported illness and healthcare utilization over a 4-week reference period can serve as reliable proxies for objective health measures (Bourne, 2009; Short et al., 2009).
Finally, we are making a linearity assumption at the intensive margin (in terms of the time of program exposure and healthcare personnel); there is a possibility that some of the effects could be non-linear. Also, we do not explore heterogeneity by geographical/regional areas, population size (rural vs. peri-urban), or level of public health infrastructure.
In terms of future research, this line of inquiry could be extended to analyze potential effects on a number of additional outcomes: immunization rates (Cristia et al., 2015), mortality by age groups (i.e., infant vs. adult mortality) (Rocha & Soares, 2010), hospitalizations for ambulatory care sensitive conditions (da Silva & Powell-Jackson, 2017; Fontes et al., 2018), etc.
6. Discussion and Conclusion
This paper uses nationally representative data to evaluate the impact of a primary care program, Estrategia Médico del Barrio (EMB), which rolled out primary healthcare teams in rural and peri-urban zones in Ecuador during 2017–2019. Using double- and triple-difference models, we find DD evidence that is consistent with a positive effect of EMB on diagnostic and preventive healthcare; yet mixed evidence regarding an effect on curative healthcare.
The primary DD analysis indicates that the EMB program led to a 4.69 percentage point decrease in the likelihood of reporting health issues (from a baseline of 35.84%). This decrease of about 13% (i.e., 4.69 / 35.84) suggests that the program successfully impacted health outcomes and achieved its objective of enhancing healthcare access in underserved communities. This observation resonates with the preventive care perspective highlighted in our theoretical framework (Starfield et al., 2005). The diverse team composition within the EMB program, particularly the inclusion of community health workers, underscores the model’s emphasis on preventive care’s role in influencing healthcare outcomes diminishing reported health problems.
Delving deeper into this dynamic, the augmented presence of EMB staff, including primary care physicians, nurses, and community health workers, seemingly catalyzed a dual effect: not only did it increase the reporting of health issues but it also amplified preventive care engagement. This paradoxical rise in reported health issues in the presence of more EMB staff may be due to increased detection and diagnosis. In other words, as more mobile, traveling healthcare teams are able to identify more patients, more health issues can be detected. This finding can be partly explained through the lens of the Andersen Healthcare Utilization Model, where the EMB program acted as an enabling resource, influencing individuals’ propensity to utilize healthcare services (Andersen, 1995). This aligns with the model’s assertion that access to and utilization of healthcare services are influenced by available resources. A greater willingness to report health concerns may occur if individuals perceive more available and effective treatment options, or if they have become more aware of their health status due to the program’s intervention.
Our study reveals that the rise in preventive care uptake is linked to the increased number of EMB staff. This trend appears to be driven by the introduction of community health workers, whose role was to identify vulnerable patients and to engage in health promotion within their communities. Community health workers, although not professionally degreed, receive training from the Ministry of Public Health. The ability of these non-specialists to deliver preventive care could explain the increased uptake observed. The concept of task shifting, as discussed in our theoretical background, is exemplified here (McPake & Mensah, 2008). Redistributing healthcare tasks to non-physician workers, like community health workers within the EMB teams, optimizes healthcare delivery, particularly in preventive care (Ingenhoff et al., 2023).
Results show that, although preventive care improved, the EMB program’s impact on enhancing access to curative care was not changed. A potential factor in this limited effectiveness in increasing curative care may be the program’s reach, with only a small minority (1.65%) of ENSANUT respondents reporting that they received healthcare services directly at home (Ministry of Public Health of Ecuador, 2018). This may be due to a shortage of primary care physicians within the EMB program. Unlike community health workers, physicians offer specialized care, are more costly to employ, and are in shorter supply. Administrative data on the program implementation reveal that some cantons completely lacked primary care physicians, suggesting that the EMB program was rolled out in some regions without adequate medical staff. Table 1 shows that the canton with the minimum number of EMB primary care physicians was zero, and the maximum was 0.93. Such a shortfall may hinder the program’s overall success and its capacity to provide timely care to those in need.
Another plausible explanation for the observed lack of impact on curative care could be due to the time lag between patient identification and the scheduling of medical appointments. The ENSANUT surveys frame their questions to capture instances of illness and receipt of curative care within the past 30 days. It is possible that patients who were visited by EMB might have had to wait longer than this 30-day period to secure a medical appointment, a delay that is not unusual within the Ecuadorian healthcare system (Gómez & Rivera, 2019; Ibarra et al., 2019). Such a restricted timeframe in the surveys could potentially diminish the observed effect since individuals who were eventually covered by EMB, but received care after the 30-day window, would not have their treatment reflected in the survey responses. This scenario may lead to an underestimation of the program’s true impact on curative care delivery.
A distinguishing feature of this study is the use of a triple-differences approach to examine the heterogeneous effects of the EMB program on populations in cantons with substantial indigenous presence. Our DDD analysis reveals a significant 2.62 percentage point increase in the likelihood of receiving preventive care (from a baseline of 9.52%). These results are substantial, highlighting the EMB program’s success in delivering healthcare to communities historically lacking basic services, often the case in rural cantons with high indigenous populations.
The study’s findings that preventive care uptake increased in indigenous communities offer promising insights into developing health interventions that overcome cultural barriers to healthcare access. Qualitative research in Ecuador demonstrates that Indigenous populations can be wary of medical professionals. For example, indigenous populations tend to delay seeking medical help as there is a lack of belief that professional care will improve their health outcomes. In some cases, they prefer more traditional medicine and healthcare. Therefore, the medical professional influence might be counter-normative for those populations (Carpio-Arias et al., 2022; Goicolea et al., 2010). To address cultural nuances, the EMB program employs an intercultural strategy to tailor its outreach to these communities, incorporating cultural, ethnic, and linguistic considerations. For instance, in regions with a dense indigenous population, the title “neighborhood doctors” was adapted to “community doctors” to resonate with the indigenous concept of community, reflecting respect for their autonomy and cultural identity (Ministry of Public Health of Ecuador, 2017). The significant prolonged impact of the EMB program in indigenous communities can be attributed to its cultural competence approach. Tailoring healthcare services to align with the unique cultural narratives of these communities facilitates increased engagement with preventive care, demonstrating the importance of cultural competence in healthcare interventions (Govere & Govere, 2016).
When comparing early adopters to late adopters, our sensitivity analysis indicates that EMB had a more pronounced impact in its initial stages. We observe that late adopters—those with a shorter duration of EMB coverage—were less likely to report receiving preventive care, implying a reduction in the program’s influence over time. However, when employing a DDD approach to assess early- versus late-adoption specifically in cantons with high indigenous populations, we discover that late adopters in these areas report higher rates of preventive care. This suggests that, in contrast to the general trend, the program’s effectiveness is sustained over time in areas with large indigenous communities.
In terms of health policy implications, several key insights emerge from this study. It first underscores the critical need for a robustly funded and adequately staffed healthcare infrastructure to ensure prompt and effective healthcare delivery. The health budget, at $3.1 billion, accounting for only 2.78% of the nation’s GDP, falls notably short of the recommended 4 to 5% benchmark (El Comercio, 2019a). Additionally, the study notes that numerous cantons embarked on the EMB program with an insufficient number of primary care physicians. While Ecuador has made continuous efforts to expand its medical workforce, the deficit of specialists, especially in primary care, is an enduring challenge. The country faces a shortfall in postgraduate trained professionals in specialized fields of medicine and surgery (El Comercio, 2019b). Although a contingent of Cuban doctors was brought in after the 2009 health reform to mitigate this shortage, their departure in 2019 left Ecuador grappling once again with a deficit in medical practitioners (Ministry of Public Health of Cuba, 2019). This shortage has impeded the ability to meet the healthcare needs, especially in terms of doctors who can provide comprehensive care in the more rural and highly indigenous communities.
There is increasing evidence that despite limited healthcare workforce, limited financial resources, high burden of disease, rapid population growth, and challenges in extending healthcare to hard-to-reach populations, mobile traveling healthcare teams in low- and middle-income countries (LMICs) are emerging as useful and promising tools to address these healthcare system constraints (Yu et al., 2017). Our results also contribute to the incipient applied econometrics literature analyzing the effects of primary healthcare expansions in LMICs; and they are consistent with the overall mixed or modestly positive results (Fontes et al., 2018; Özçelik et al., 2020; Rocha & Soares, 2010).
In conclusion, this paper finds evidence to support the hypothesis that Estrategia Médico del Barrio improves health problem diagnoses and preventive healthcare utilization, including in highly indigenous cantons, yet it seems to have mixed results in terms of curative care use in Ecuador.
Highlights.
Estrategia Médico del Barrio (EMB) rolled out primary health teams in rural and peri-urban zones in Ecuador during 2017–19
Using difference-in-differences models, we find positive effects of EMB on diagnostic and preventive care; and mixed evidence on curative care.
EMB led to a 4.69-percentage-point decrease in the likelihood of reporting health issues (from a baseline of 35.84%).
Triple differences analysis shows a 2.62-percentage-point increase in the likelihood of receiving preventive care (from a baseline of 9.52%).
EMB had some success delivering healthcare to communities lacking basic services in rural cantons with high indigenous populations.
Acknowledgements
We thank useful comments by Melissa Knox, John Cawley and other participants at the American Economic Association / Allied Social Sciences Associations (AEA/ASSA) meetings in New Orleans, LA.; AcademyHealth (Global Health Interest Group) in Seattle, WA; International Health Association (iHEA) meetings in Cape Town, South Africa. Marta Wilson-Barthes and Annie Liang provided excellent research assistance; and our research interns at the Ecuadorian Development Research Lab, Cristhian López and Luis Reyes, provided invaluable assistance during the data gathering stage. The Population Studies and Training Center (PSTC) at Brown University, which receives funding from the US National Institutes of Health (through NIH Grant P2CHD041020), provided general support.
Appendix.
QUIPUX response from Health District 06D02 providing the requested information regarding EMB’s rollout dates and EMB mobile health teams’ number of personnel and composition.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- Ahmed S, Chase LE, Wagnild J, Akhter N, Sturridge S, Clarke A, Chowdhary P, Mukami D, Kasim A, & Hampshire K (2022). Community health workers and health equity in low- and middle-income countries: Systematic review and recommendations for policy and practice. International Journal for Equity in Health, 21(1), 49–49. 10.1186/s12939-021-01615-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- Altman DG, & Royston P (2006). The cost of dichotomising continuous variables. British Medical Journal, 332(7549), 1080. 10.1136/bmj.332.7549.1080 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Alexy BB, & Elnitsky C (1998). Rural Mobile Health Unit: Outcomes. Public Health Nursing, 15(1), 3–11. 10.1111/j.1525-1446.1998.tb00314.x [DOI] [PubMed] [Google Scholar]
- Andersen RM (1995). Revisiting the Behavioral Model and Access to Medical Care: Does it Matter? Journal of Health and Social Behavior, 36(1), 1–10. 10.2307/2137284 [DOI] [PubMed] [Google Scholar]
- Asante A, Price J, Hayen A, Jan S, & Wiseman V (2016). Equity in Health Care Financing in Low- and Middle-Income Countries: A Systematic Review of Evidence from Studies Using Benefit and Financing Incidence Analyses. PLOS ONE, 11(4), e0152866. 10.1371/journal.pone.0152866 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Aung K-K, Hill C, Bennet J, Song Z, & Oriol NE (2015). The Emerging Business Models and Value Proposition of Mobile Health Clinics. American Journal of Accountable Care, 3(4), 36–40. [PMC free article] [PubMed] [Google Scholar]
- Baylan S, Haig C, MacDonald M, Stiles C, Easto J, Thomson M, Cullen B, Quinn TJ, Stott D, & Mercer SW (2020). Measuring the effects of listening for leisure on outcome after stroke (MELLO): A pilot randomized controlled trial of mindful music listening. International Journal of Stroke, 15(2), 149–158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bitton A, Fifield J, Ratcliffe H, Karlage A, Wang H, Veillard JH, Schwarz D, & Hirschhorn LR (2019). Primary healthcare system performance in low-income and middle-income countries: A scoping review of the evidence from 2010 to 2017. BMJ Global Health, 4(Suppl 8), e001551. 10.1136/bmjgh-2019-001551 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bourne PA (2009). The validity of using self-reported illness to measure objective health. North American Journal of Medical Sciences, 1(5), 232–238. 10.4297/najms.2009.5232 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Braithwaite J, Zurynski Y, Ludlow K, Holt J, Augustsson H, & Campbell M (2019). Towards sustainable healthcare system performance in the 21st century in high-income countries: A protocol for a systematic review of the grey literature. BMJ Open, 9(1), e025892–e025892. 10.1136/bmjopen-2018-025892 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Callaghan M, Ford N, & Schneider H (2010). A systematic review of task- shifting for HIV treatment and care in Africa. Human Resources for Health, 8(1), 8. 10.1186/1478-4491-8-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Camacho-Leon G, Faytong-Haro M, Carrera K, Molero M, Melean F, Reyes Y, Mautong H, De La Hoz I, & Cherrez-Ojeda I (2022). A Narrative Review of Telemedicine in Latin America during the COVID-19 Pandemic. Healthcare, 10(8), Article 8. 10.3390/healthcare10081361 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carpenter C, & Dobkin C (2011). The Minimum Legal Drinking Age and Public Health. Journal of Economic Perspectives, 25(2), 133–156. 10.1257/jep.25.2.133 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carpio-Arias TV, Verdezoto N, Guijarro-Garvi M, Abril-Ulloa V, Mackintosh N, Eslambolchilar P, & Ruíz-Cantero MT (2022). Healthcare professionals’ experiences and perceptions regarding health care of indigenous pregnant women in Ecuador. BMC Pregnancy and Childbirth, 22(1), 1–11. 10.1186/s12884-022-04432-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chapman R, Keall M, Howden-Chapman P, Grams M, Witten K, Randal E, & Woodward A (2018). A Cost Benefit Analysis of an Active Travel Intervention with Health and Carbon Emission Reduction Benefits. International Journal of Environmental Research and Public Health, 15(5), Article 5. 10.3390/ijerph15050962 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chen M, Shi W, Xie P, Silva VBS, Kousky VE, Wayne Higgins R, & Janowiak JE (2008). Assessing objective techniques for gauge-based analyses of global daily precipitation. Journal of Geophysical Research: Atmospheres, 113(D4). [Google Scholar]
- Constituent Assembly of Ecuador (2008). Constitution of the Republic of Ecuador [Constitución de La República del Ecuador]. https://www.oas.org/juridico/PDFs/mesicic5_ecu_ane_cons.pdf.
- Cristia J, Evans WN, & Kim B (2015). Improving the Health Coverage of the Rural Poor: Does Contracting-Out Mobile Medical Teams Work? The Journal of Development Studies, 51(3), 247–261. 10.1080/00220388.2014.976617 [DOI] [Google Scholar]
- da Silva EN, & Powell-Jackson T (2017). Does expanding primary healthcare improve hospital efficiency? Evidence from a panel analysis of avoidable hospitalisations in 5506 municipalities in Brazil, 2000–2014. BMJ Global Health, 2(2), e000242–e000242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- El Comercio. (2019a). El presupuesto de salud 2019 prioriza prevención y servicios [The 2019 Health Budget Prioritizes Prevention and Services]. El Comercio. https://www.elcomercio.com/tendencias/sociedad/presupuesto-salud-prevencion-ecuador-servicios.html [Google Scholar]
- El Comercio. (2019b). En Ecuador, la falta de especialistas de salud aún preocupa [In Ecuador, the Shortage of Health Specialists Remains a Concern]. El Comercio. https://www.elcomercio.com/tendencias/sociedad/ecuador-falta-especialistas-salud-preocupacion.html [Google Scholar]
- Espinosa V, Acuña C, De la Torre D, & Tambini G (2017). La reforma en salud del Ecuador. Revista Panamericana de Salud Pública, 41. 10.26633/RPSP.2017.96 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Espinosa V, De la Torre D, Acuña C, & Cadena C (2017). Los recursos humanos en salud según el nuevo modelo de atención en Ecuador. Revista Panamericana de Salud Pública, 41, e52–e52. 10.26633/RPSP.2017.52 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fanelli S, Salvatore FP, De Pascale G, & Faccilongo N (2020). Insights for the future of health system partnerships in low- and middle-income countries: A systematic literature review. BMC Health Services Research, 20(1), 571–571. 10.1186/s12913-020-05435-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Flores Jimenez SE, & San Sebastián M (2021). Assessing the impact of the 2008 health reform in Ecuador on the performance of primary health care services: An interrupted time series analysis. International Journal for Equity in Health, 20(1), 169. 10.1186/s12939-021-01495-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fontes LFC, Conceição OC, & Jacinto P. de A. (2018). Evaluating the impact of physicians’ provision on primary healthcare: Evidence from Brazil’s More Doctors Program. Health Economics, 27(8), 1284–1299. 10.1002/hec.3775 [DOI] [PubMed] [Google Scholar]
- Freire WB, Belmont P, López-Cevallos DF, & Waters WF (2015). Ecuador’s National Health and Nutrition Survey: Objectives, design, and methods. Annals of Epidemiology, 25(11), 877–878. 10.1016/j.annepidem.2015.08.009 [DOI] [PubMed] [Google Scholar]
- Frijters P, Lalji C, & Pakrashi D (2020). Daily weather only has small effects on wellbeing in the US. Journal of Economic Behavior & Organization, 176, 747–762. 10.1016/j.jebo.2020.03.009 [DOI] [Google Scholar]
- Galárraga O, & Harris JE (2021). Effect of an abrupt change in sexual and reproductive health policy on teen birth rates in Ecuador, 2008–2017. Economics & Human Biology, 41, 100967. 10.1016/j.ehb.2020.100967 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Goicolea I, Wulff M, Sebastian MS, & Öhman A (2010). Adolescent pregnancies and girls’ sexual and reproductive rights in the amazon basin of Ecuador: An analysis of providers’ and policy makers’ discourses. BMC International Health and Human Rights, 10, 1–13. 10.1186/1472-698X-10-12 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gómez P, & Rivera J (2019). Un problema social: Tiempos de espera en la consulta externa del Hospital Carlos Andrade Marín. Estudios de La Gestión: Revista Internacional de Administración, 5, 121–146. [Google Scholar]
- Govere L, & Govere EM (2016). How Effective is Cultural Competence Training of Healthcare Providers on Improving Patient Satisfaction of Minority Groups? A Systematic Review of Literature. Worldviews on Evidence-Based Nursing, 13(6), 402–410. 10.1111/wvn.12176 [DOI] [PubMed] [Google Scholar]
- Granda ML, & Jimenez WG (2019). The evolution of socioeconomic health inequalities in Ecuador during a public health system reform (2006–2014). International Journal for Equity in Health, 18(1), 31–31. 10.1186/s12939-018-0905-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hampshire K, Mwase-Vuma T, Alemu K, Abane A, Munthali A, Awoke T, Mariwah S, Chamdimba E, Owusu SA, Robson E, Castelli M, Shkedy Z, Shawa N, Abel J, & Kasim A (2021). Informal mhealth at scale in Africa: Opportunities and challenges. World Development, 140, 105257. 10.1016/j.worlddev.2020.105257 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ibarra CPT, García WSV, & Alay TBR (2019). La calidad de la atención en los servicios de salud en la provincia de Manabí, Ecuador. Polo Del Conocimiento: Revista Científico-Profesional, 4(12), 135–148. [Google Scholar]
- Ingenhoff R, Munana R, Weswa I, Gaal J, Sekitoleko I, Mutabazi H, Bodnar BE, Rabin TL, Siddharthan T, Kalyesubula R, Knauf F, & Nalwadda CK (2023). Principles for task shifting hypertension and diabetes screening and referral: A qualitative study exploring patient, community health worker and healthcare professional perceptions in rural Uganda. BMC Public Health, 23(1), 881. 10.1186/s12889-023-15704-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lee J, & Smith JP (2012). The effect of health promotion on diagnosis and management of diabetes. J Epidemiol Community Health, 66(4), 366–371. 10.1136/jech.2009.087304 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Li M, Ferreira S, & Smith TA (2020). Temperature and self-reported mental health in the United States. PLOS ONE, 15(3), e0230316. 10.1371/journal.pone.0230316 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mahmoud K, Jaramillo C, & Barteit S (2022). Telemedicine in Low- and Middle-Income Countries During the COVID-19 Pandemic: A Scoping Review. Frontiers in Public Health, 10. 10.3389/fpubh.2022.914423 [DOI] [PMC free article] [PubMed] [Google Scholar]
- McPake B, & Mensah K (2008). Task shifting in health care in resource-poor countries. The Lancet, 372(9642), 870–871. 10.1016/S0140-6736(08)61375-6 [DOI] [PubMed] [Google Scholar]
- Ministerio de Salud Pública del Ecuador. (2019). Encuesta Nacional de Salud y Nutrición—Ensanut 2018 [dataset]. https://www.ecuadorencifras.gob.ec/salud-salud-reproductiva-y-nutricion/ [DOI] [PubMed]
- Ministry of Public Health of Cuba. (2019). Regresa a Cuba primer grupo de colaboradores de la Brigada Médica en Ecuador [First Group of Collaborators from the Medical Brigade Returns to Cuba from Ecuador]. Ministry of Public Health of Cuba. https://salud.msp.gob.cu/regresa-a-cuba-primer-grupo-de-colaboradores-de-la-brigada-medica-en-ecuador/ [Google Scholar]
- Ministry of Public Health of Ecuador. (2011). Ministerio de Salud presentará Modelo de Atención Integral en Salud en Ambato. http://instituciones.msp.gob.ec/dps/cotopaxi/index.php?view=article&catid=5%3Anoticias&id=271%3Aministerio-de-salud-presentara-modelo-de-atencion-integral-en-salud-en-ambato&format=pdf&option=com_content&Itemid=83#:~:text=Los%20requisitos%20para%20ser%20parte,de%20salud%20y%20la%20comunidad.
- Ministry of Public Health of Ecuador. (2012a). La Encuesta Nacional de Salud y Nutrición – ENSANUT 2012. Instituto Nacional de Estadística y Censos. https://ghdx.healthdata.org/record/ecuador-national-health-and-nutrition-survey-2012 [Google Scholar]
- Ministry of Public Health of Ecuador. (2012b). Manual del Modelo de Atención Integral de Salud-MAIS. Ministerio de Salud Pública Del Ecuador. http://librodigital.sangregorio.edu.ec/librosusgp/B0033.pdf [Google Scholar]
- Ministry of Public Health of Ecuador. (2017). Estrategia de Abordaje Médico del Barrio y su Equipo de Atención Integral en Salud (First). Dirección Nacional de Primer Nivel de Atención en Salud. https://www.salud.gob.ec/wp-content/uploads/2018/02/Manual_De_Estrategia_de_Abordaje_Medico_del_Barrio_2018.pdf [Google Scholar]
- Ministry of Public Health of Ecuador. (2018). La Encuesta Nacional de Salud y Nutrición – ENSANUT 2018. Instituto Nacional de Estadística y Censos. https://www.ecuadorencifras.gob.ec/salud-salud-reproductiva-y-nutricion/ [Google Scholar]
- Morphew T, Scott L, Li M, Galant SP, Wong W, Lloret MIG, Jones F, Bollinger ME, & Jones CA (2013). Mobile Health Care Operations and Return on Investment in Predominantly Underserved Children with Asthma: The Breathmobile Program. Population Health Management, 16(4), 261–269. 10.1089/pop.2012.0060 [DOI] [PubMed] [Google Scholar]
- National Institute of Statistics and Census of Ecuador. (2006). Encuesta de Condiciones de Vida – ECV 2006. https://anda.inec.gob.ec/anda/index.php/catalog/358
- Oriol NE, Cote PJ, Vavasis AP, Bennet J, DeLorenzo D, Blanc P, & Kohane I (2009). Calculating the return on investment of mobile healthcare. BMC Medicine, 7(1), 27. 10.1186/1741-7015-7-27 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ozawa S, Shankar R, Leopold C, & Orubu S (2019). Access to medicines through health systems in low- and middle-income countries. Health Policy and Planning, 34(Supplement_3), iii1–iii3. 10.1093/heapol/czz119 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Özçelik EA, Massuda A, McConnell M, & Castro MC (2020). Impact of Brazil’s More Doctors Program on hospitalizations for primary care sensitive cardiovascular conditions. SSM-Population Health, 12, 100695–100695. 10.1016/j.ssmph.2020.100695 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Papanicolas I, Woskie LR, & Jha AK (2018). Health Care Spending in the United States and Other High-Income Countries. JAMA, 319(10), 1024–1039. 10.1001/jama.2018.1150 [DOI] [PubMed] [Google Scholar]
- Quizhpe E, Teran E, Pulkki-Brännström A-M, & San Sebastián M (2022). Social inequalities in healthcare utilization during Ecuadorian healthcare reform (2007–2017): A before-and-after cross-sectional study. BMC Public Health, 22(1), 499–499. 10.1186/s12889-022-12884-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rocha R, & Soares RR (2010). Evaluating the impact of community-based health interventions: Evidence from Brazil’s Family Health Program. Health Economics, 19(S1), 126–158. 10.1002/hec.1607 [DOI] [PubMed] [Google Scholar]
- Roeper B, Mocko J, O’Connor LM, Zhou J, Castillo D, & Beck EH (2018). Mobile Integrated Healthcare Intervention and Impact Analysis with a Medicare Advantage Population. Population Health Management, 21(5), 349–356. 10.1089/pop.2017.0130 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Romero-Alvarez D, López-Cevallos DF, & Torres I (2023). Doctors for the people? The problematic distribution of rural service doctors in Ecuador. Health Policy and Planning, 38(7), 851–861. 10.1093/heapol/czad040 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rule J, Ngo DA, Oanh TTM, Asante A, Doyle J, Roberts G, & Taylor R (2014). Strengthening Primary Health Care in Low- and Middle-Income Countries: Generating Evidence Through Evaluation. Asia Pacific Journal of Public Health, 26(4), 339–348. 10.1177/1010539513503869 [DOI] [PubMed] [Google Scholar]
- Savaşer SK, & Kara BY (2022). Mobile healthcare services in rural areas: An application with periodic location routing problem. OR Spectrum, 44(3), 875–910. 10.1007/s00291-022-00670-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Short ME, Goetzel RZ, Pei X, Tabrizi MJ, Ozminkowski RJ, Gibson TB, DeJoy DM, & Wilson MG (2009). How Accurate are Self-Reports? Analysis of Self-Reported Health Care Utilization and Absence When Compared With Administrative Data. Journal of Occupational and Environmental Medicine, 51(7), 786. 10.1097/JOM.0b013e3181a86671 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sims-Gould J, & Martin-Matthews A (2010). We share the care: Family caregivers’ experiences of their older relative receiving home support services. Health & Social Care in the Community, 18(4), 415–423. 10.1111/j.1365-2524.2010.00913.x [DOI] [PubMed] [Google Scholar]
- Sims-Gould J, Tong CE, Wallis-Mayer L, & Ashe MC (2017). Reablement, Reactivation, Rehabilitation and Restorative Interventions With Older Adults in Receipt of Home Care: A Systematic Review. Journal of the American Medical Directors Association, 18(8), 653–663. 10.1016/j.jamda.2016.12.070 [DOI] [PubMed] [Google Scholar]
- Solomon EM, Wing H, Steiner JF, & Gottlieb LM (2020). Impact of Transportation Interventions on Health Care Outcomes: A Systematic Review. Medical Care, 58(4), 384. 10.1097/MLR.0000000000001292 [DOI] [PubMed] [Google Scholar]
- Sritart H, Tuntiwong K, Miyazaki H, & Taertulakarn S (2021). Disparities in healthcare services and spatial assessments of mobile health clinics in the border regions of Thailand. International Journal of Environmental Research and Public Health, 18(20), 10782–10782. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Starfield B, Shi L, & Macinko J (2005). Contribution of Primary Care to Health Systems and Health. The Milbank Quarterly, 83(3), 457–502. 10.1111/j.1468-0009.2005.00409.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- Walker GC, Martinez-Gómez V, LCDC C, & Gonzalez RO (2018). Mobile Traveling Healthcare Teams: An Innovative Delivery System for Underserved Populations. Online Journal of Issues in Nursing, 23(3), 1–10. 10.3912/OJIN.Vol23No03Man03 [DOI] [Google Scholar]
- Witmer A, Seifer SD, Finocchio L, Leslie J, & O’Neil EH (1995). Community health workers: Integral members of the health care work force. American Journal of Public Health, 85(8_Pt_1), 1055–1058. 10.2105/AJPH.85.8_Pt_1.1055 [DOI] [PMC free article] [PubMed] [Google Scholar]
- World Health Organization 2024 data.who.int. (n.d.). Density of physicians (per 10 000 population). Datadot. Retrieved March 25, 2024, from https://data.who.int/indicators/i/217795A [Google Scholar]
- Yu SWY, Hill C, Ricks ML, Bennet J, & Oriol NE (2017). The scope and impact of mobile health clinics in the United States: A literature review. International Journal for Equity in Health, 16(1), 178. 10.1186/s12939-017-0671-2 [DOI] [PMC free article] [PubMed] [Google Scholar]