DAY 6 READING MEDICAL ARTICLE

Abstract

Background

Despite the substantial attention to primary care (PC), few studies have addressed the relationship between patients’ experience with PC and their health status in low-and middle-income countries. This study aimed to (1) test the association between overall patient-centered PC experience (OPCE) and self-rated health (SRH) and (2) identify specific features of patient-centered PC associated with better SRH (i.e., excellent or very good SRH) in 6 Latin American and Caribbean countries.

Methods and findings

We conducted a secondary analysis of a 2013 public opinion cross-sectional survey on perceptions and experiences with healthcare systems in Brazil, Colombia, El Salvador, Jamaica, Mexico, and Panama; the data were nationally representative for urban populations. We analyzed 9 features of patient-centered PC. We calculated OPCE score as the arithmetic mean of the PC features. OPCE score ranged from 0 to 1, where 0 meant that the participant did not have any of the 9 patient-centered PC experiences, while 1 meant that he/she reported having all these experiences. After testing for interaction on the additive scale, we analyzed countries pooled for aim 1, with an interaction term for Mexico, and each country separately for aim 2. We used multiple Poisson regression models double-weighted by survey and inverse probability weights to deal with the survey design and missing data. The study included 6,100 participants. The percentage of participants with excellent or very good SRH ranged from 29.5% in Mexico to 52.4% in Jamaica. OPCE was associated with reporting excellent or very good SRH in all countries: adjusting for socio-demographic and health covariates, patients with an OPCE score of 1 in Brazil, Colombia, El Salvador, Jamaica, and Panama were more likely to report excellent or very good SRH than those with a score of 0 (adjusted prevalence ratio [aPR] 1.61, 95% CI 1.37–1.90, p < 0.001); in Mexico, this association was even stronger (aPR 4.27, 95% CI 2.34–7.81, p < 0.001). The specific features of patient-centered PC associated with better SRH differed by country. The perception that PC providers solve most health problems was associated with excellent or very good SRH in Colombia (aPR 1.38, 95% CI 1.01–1.91, p =0.046) and Jamaica (aPR 1.21, 95% CI 1.02–1.43, p = 0.030). Having a provider who knows relevant medical history was positively associated with better SRH in Mexico (aPR 1.47, 95% CI 1.03–2.12, p = 0.036) but was negatively associated with better SRH in Brazil (aPR 0.71, 95% CI 0.56–0.89, p = 0.003). Finally, easy contact with PC facility (Mexico: aPR 1.35, 95% CI 1.04–1.74, p = 0.023), coordination of care (Mexico: aPR 1.53, 95% CI 1.19–1.98, p = 0.001), and opportunity to ask questions (Brazil: aPR 1.42, 95% CI 1.11–1.83, p = 0.006) were each associated with better SRH. The main study limitation consists in the analysis being of cross-sectional data, which does not allow making causal inferences or identifying the direction of the association between the variables.

Conclusions

Overall, a higher OPCE score was associated with better SRH in these 6 Latin American and Caribbean countries; associations between specific characteristics of patient-centered PC and SRH differed by country. The findings underscore the importance of high-quality, patient-centered PC as a path to improved population health.

Introduction

Primary care (PC) has been described as being uniquely positioned to promote health and well-being at the population level [1–3]. Its central role in providing adequate, efficient, and equitable access to preventive and curative healthcare was strongly emphasized by the Declaration of Alma-Ata 40 years ago [2]. However, PC’s centrality has been questioned by findings of its poor quality, especially in low- and middle-income countries (LMICs), and its limitations in adapting to urbanization and the epidemiological transition, as well as increasing population demand for responsive, high-quality services [4–6].

More recently, patient-centered healthcare has emerged as a person-oriented model of care aimed at meeting population needs, expectations, and preferences. Studies from the United States and the United Kingdom have shown the positive effect of patient-centered healthcare in improving the quality of the processes of care, reducing hospitalizations and emergency visits (and consequently healthcare costs), and improving users’ satisfaction and self-management [7–9]. Within PC, a number of patient-centered healthcare attributes have been shown to be associated with perception of good healthcare quality, such as the availability of a PC provider who “knows relevant information about a patient’s medical history,” “solves most of the health problems,” “spends enough time with the patient,” “coordinates healthcare,” and “is easy to communicate with” [10].

Previous work using ecological data has shown positive effects of PC on population health outcomes, such as child mortality, and avoidable hospitalizations, both in high-income countries and in LMICs [11,12]. However, despite the substantial attention and policy emphasis on PC, few studies have addressed the relationship between patient experience with PC and health in LMICs; none to our knowledge have done so with a cross-country perspective [13,14].

Self-rated health (SRH) is a broadly used measure: individuals evaluate their health status through a Likert scale or compare their health status with individuals of the same age [15]. Though SRH is a subjective indicator of health status, it has been found to be a robust predictor of mortality [15,16]; also, low SRH is associated with increased hospitalization and outpatient care in elderly populations [17]. Research studies addressing the relationship between health service characteristics and SRH have reported that in the US, individuals living in states with a higher ratio of PC physicians to population were more likely to report good SRH than those with a lower ratio of PC physicians [18]. Enhanced accessibility and continuity of PC in the US [19,20] and high total PC quality scores in South Korea were associated with better SRH of health service users [21].

The objectives of the present study were (1) to test the association between overall patient-centered PC experience (OPCE) and SRH in 6 Latin American and Caribbean (LAC) countries and (2) to identify specific features of patient-centered PC associated with better SRH. This work can help inform financing and policies at a moment of renewed global attention to PC.

Methods

We performed a secondary analysis of a recent (2013) public opinion survey focusing on perceptions and experiences with healthcare systems in 6 LAC countries: Brazil, Colombia, El Salvador, Jamaica, Mexico, and Panama. The detailed methodology of this survey was previously reported elsewhere [10,22,23]. In each country, the survey included a nationally representative urban sample of the population that comprised between 1,500 and 1,506 adults per country. According to the 2017 revision of World Population Prospects [24], the urban population constitutes the majority in these countries, ranging from 54% in Jamaica to 85% in Brazil. In total, 330 million individuals reside in urban areas in these countries.

During 2012 and 2013, Harris Interactive collected the data through telephone interviews. The sample frame for the survey consisted of random digit dialing listings of landline and mobile phone numbers in each country. The survey used an adapted version of the methodology and questionnaire that the Commonwealth Fund has been applying in Europe, Australia, Canada, and the US over the past 15 years [25]. The selection criteria considered any household member aged 18+ years. Only 1 adult per household was interviewed.

Analysis plan

We did not have a formal prospective analysis plan. Prior to seeing the data, we identified the public opinion survey as a unique resource to test associations of healthcare quality and SRH in a representative, multi-country sample. We then reviewed the literature on patient-centered PC and identified 5 key domains (contact with clinic, time spent with provider, patient–provider communication, technical quality and solving problems, and healthcare coordination) relevant for patient-centered healthcare [10,23,25,26]; we mapped items from the survey to these domains and created single-item summaries as well as an overall score. We defined covariates based on relevance to health status and healthcare utilization. We planned to assess all countries in a pooled sample; on identifying substantial variation in the level of SRH between countries, we tested for interaction between patient-centered PC variables and country on the additive scale and report stratified models where evidence of interaction was found.

Study variables

The dependent variable was “excellent or very good SRH,” obtained from the general SRH report and categorized as 1 = “excellent” or “very good” and 0 = “good,” “fair,” “poor,” or “not sure”.

The survey specified that PC is care provided by the doctors or other health professionals (i.e., nurses, social workers) at the family doctor’s practice or clinic. We selected items related to PC that fall into the domains of patient-centered healthcare identified in the literature [10,23,25,26] and organized them by domain:

1. Contact with PC clinic
   • PC facility is easy to contact by telephone during regular office hours
2. Time spent with provider
   • PC provider spends enough time with patient
3. Patient–provider communication
   • PC provider gives the patient an opportunity to ask questions
   • PC provider explains things in a way that is easy to understand
4. Technical quality and solving problems
   • PC provider knows relevant information about the patient’s medical history
   • PC provider advises about healthy lifestyles (healthy food, regular physical activity, and possible stressors)
   • Preventive exams are up to date
   • PC provider solves most of the patient’s health problems
5. Healthcare coordination
   • PC provider helps to coordinate care with other physicians or sources of care

The variable “preventive exams up to date” was defined as “yes” when the respondent reported having blood pressure measurement in the last year, serum cholesterol in the last 5 years, and, for women over 40 years, cervical cytology (Pap test) and mammography in the last 3 years.

All other PC variables were measured on a 5-options Likert scale and categorized as 1 = yes (“always” or “often”) and 0 = no (“sometimes,” “rarely or never,” and “not sure”). The decision to categorize the variables this way was based on previous studies [10,23,25,26]. We calculated OPCE as the arithmetic mean of these items, following the recommendation of previous research on the use of patient experience surveys to assess service provision [27]. OPCE score ranged from 0 to 1, where 0 meant that participant did not have any of the 9 patient-centered PC experiences, while 1 meant that he/she reported having all these experiences. We assumed that each component of OPCE score contributed equally to patients’ experiences and that a difference in patient experiences had a constant effect on SRH. We maintained the 9 binary items as individual components of patient-centered PC.

Several socio-demographic and health-service-related factors are associated with poor SRH. Individual factors linked to lower health status include unhealthy lifestyle [28–30] and chronic diseases that affect mental and physical health [31–34]. Although some aspects of the relationship between socio-demographic factors and SRH are still inconclusive, it has been reported that older age, low schooling, low socio-economic status, low social capital, and low health insurance (HI) coverage are associated with poor SRH [35–37].

Based on survey data availability, we included the following covariates: sex, age, education, chronic disease, and the type of HI. The variable education defines the level of education for participants who answered the survey in all countries except El Salvador, where it describes the education of the head of the household. We identified the participant as having a chronic disease if he/she reported that a doctor had told him/her of having arthritis, asthma or chronic lung disease, cancer, diabetes, heart disease, hypertension, or depression. The type of HI was categorized as: government HI (publicly subsidized insurance not related to job affiliation), social security HI (contributory insurance related to job affiliation), and private HI (voluntary private insurance; also, in Brazil and Jamaica, those who reported having private HI provided by workplace). Furthermore, respondents reporting both government and social-security-based HI (4.1% of participants in Colombia and 21.5% in Mexico) were grouped under social security HI. Participants who reported not having HI were placed in the government HI group, because, in all these countries, government HI is freely available for those without social security or private HI.

Statistical analysis

We used descriptive statistics to analyze the characteristics and PC experiences of the study participants. We performed a bivariable analysis including chi-squared tests between the dependent variable (SRH) and each independent variable (PC experience) or categorical covariate. We used Student t tests for comparison of the continuous variable OPCE score between people who reported excellent or very good SRH and those who reported good, fair, or poor (or not sure) SRH.

The survey asked the complete set of questions about PC experiences only to respondents who affirmed “having a regular doctor or regular place for primary care.” This skip pattern results in a high percentage of missing data, given that lack of access to a regular source of PC ranged from 16.3% in Jamaica to 43.1% in El Salvador; in addition, several PC variables also had missing information (S1 Table). In sum, in the 6 countries there were 6,100 participants with complete information, which represented 67.7% of the initial sample of 9,012. Thus, we applied a double-weighted strategy with the use of survey weights to account for the survey sample design and stabilized inverse probability (IP) weights to correct for potential selection bias [38]. This technique consists in assigning a weight to individuals with complete information so that they account for themselves as well as for those with similar characteristics who had missing information. It assumes that those with missing information are similar to those with complete information who share the same measured covariates [38]. In particular, to apply this technique to adjust for the missingness induced by not having a regular PC clinic or doctor, we assumed that the PC experience of individuals without a regular PC clinic or doctor can be represented by those with a regular PC clinic or doctor conditional on the specified covariates, i.e., that there are no unmeasured confounders that are a common cause of both having access to a regular PC clinic or doctor and SRH. We first compared the number of PC visits between those with and without a regular PC clinic or doctor. We found that the mean number of visits in the last 12 months in the group with a regular doctor was 3.07 and in the group without a regular doctor was 2.09 (p < 0.001). We then generated the denominator and numerator of the IP weights. The denominator for stabilized IP weights was the probability of having missing data conditional on the following covariates: sex, age, education, type of HI, and presence of chronic disease. The numerator was the probability of having missing data regardless of the covariates.

We used Poisson regression models with robust variance as recommended for cross-sectional studies with high-prevalence binary outcomes [39]. We initially fit pooled models across all 6 countries, and then calculated the relative excess risk due to interaction (RERI) between each country and PC experience as a measure of interaction on the additive scale [40]; additive interaction is more indicative of underlying causal interaction than interaction on the relative (ratio) scale. Where evidence of interaction was identified (RERI significant at p ≤ 0.05), we included interaction terms for country in the pooled multiple regression model, or stratified the model by country in the case of multiple interactions identified. Each multiple Poisson regression model included the dependent variable, independent variables, and conceptually relevant covariates. The pooled model included fixed effects for countries to control for country-level heterogeneity and to focus on the effect of the individual-level predictors [41,42].

Finally, we performed a sensitivity analysis in which the IP weights were calculated after the individuals without a regular PC clinic or doctor were dropped. The results were similar to those of the main analysis, suggesting that our findings were not distorted by including everyone when calculating the IP weights. All analyses were performed using the software Stata 14 and considering estimates with p ≤ 0.05 to be statistically significant.

Ethical considerations

The study consists of a secondary data analysis of a public opinion survey focusing on perceptions and experiences with healthcare systems in 6 LAC countries. The survey was commissioned by the Inter-American Development Bank, and the contracted surveying firm was responsible for obtaining all necessary regulatory approvals and verifying compliance with the ethical standards of the ICC/ESOMAR Code on Market, Opinion and Social Research. The survey data for this secondary data analysis were made available by, and their use approved by, the Inter-American Development Bank.

Results

Tables 1 and 2 present the characteristics of study participants from 6 LAC countries (n =6,100) double-weighted by survey and stabilized IP weights. Slightly more women than men participated in all countries (52.2% versus 47.8% in the full sample). Participants reported lower education levels in Brazil, Panama, and El Salvador (62.7%, 37.5%, and 33.4% with elementary school or less, respectively), while approximately half the sample had completed secondary school in Mexico, Colombia, and Jamaica. Government HI predominated in Brazil and Jamaica (76.5% and 61.5%, respectively), while social security HI was more common in Colombia, Panama, Mexico, and El Salvador (65.1%, 62.4%, 48.9%, and 47.4%, respectively). The proportion with private HI ranged from 9.8% in Colombia and El Salvador to 38.5% in Jamaica. Report prevalence of chronic conditions ranged from 31.5% in El Salvador to 52.2% in Jamaica. Finally, the percentage of participants with excellent or very good SRH was highest in Jamaica (52.4%), declining to a low of 29.5% in Mexico.