Social determinants and cardiovascular disease mortality in Panama, 2012–2016

We conducted a registry-based analysis of mortality due to CVD, from the years 2012 to 2016. Several studies evaluated the data obtained from the Panamanian civil registration and vital statistics and classified it as high quality [17, 18]. CVD was defined and coded, following the tenth revision of the International Classification of Diseases (ICD-10). ICD-10 codes I20-I25 and I60-I69 were utilized for IHD and stroke, respectively.

The HS is a tool used to identify the employment status and socioeconomic conditions of the general population. It is widely used by governments to measure the characteristics of the working market and report unemployment. The World Bank, The International Labour Organization and the Oxford Poverty and Human Development Initative use the HS data to measure inequality calculate unemployment trends and construct the MPI, respectively. Other studies concerning reproductive, maternal, newborn, and child health have used HS as the primary data source to evaluate coverage indicators [19, 20].

In Panama, the HS has been applied biannually since 1963 (March and August) [21] . Households are randomly selected from the census performed in the country with the aim to be representative by provinces. Households are defined as the location used by a family or a group of people, with or without bonds, which share a unit or a person that lives alone.

The data is collected through person-in-person interviews, including all the members of the household at the moment of the interview and the questions are structured to correspond to the information of the week prior the interview. We analyzed all the data of surveys within each household, and thus, number of study participants may vary per year. Annually, 25% of households are replaced whereas 75% of households are revisited 4 times before they are replaced. In order to capture part of the economically active population, the reference period is extended to 6 months explaining why the survey is applied biannually. In Panama, March is a summer month where labor market widens, primarily because of seasonal employment and the beginning of the school year. Therefore, in the present study, we utilized the data obtained in August for the time frame 2012–2016, as March may not represent the market behavior compared to other months. An average of 15,919 households was surveyed per year. Since we aimed to address the economically active population, study participants below 15 years of age were excluded (n = 64,202). Study participants with missing data for monthly income (n = 73) and years of education (n = 122) were further excluded. Thus, a total of n = 154,019 study participants were enrolled in the present study.

Monthly income was analyzed as a continuous variable and defined as the total income an individual received per month in US dollar ($), including monetary and species earnings, subsidies and grants (the national currency is the Panamanian balboa, however it is pegged with the US dollar at par). It included primary income and the income earned with other economic activities. Social security coverage was categorized as the access to public social security health services, represented as contributor (active worker), beneficiary, retired, pensioner (permanent disability), retired or pensioner from another country and with no coverage. We dichotomized the variable as having or not social security coverage. Years of education were explored as a continuous variable, specifying the maximum number of years of education reached. The education in Panama is characterized in three levels: basic general education, middle education and higher education. Basic general education is compulsory and consists of a total of 11 years, including preschool, elementary and middle school. Middle education consists of an additional 3 years and is diversified in general, pedagogical and technical education. Higher education is subdivided in university and non-university education. Special education is imparted to children with limiting mental or physical conditions.

Dichotomic variables were expressed as percentages and quantitative variables as mean and standard deviations (SD). Estimates were weighted to represent the total population of Panama per province and regions. The expansion factor was calculated and provided by the INEC.

Specific mortality rates for IHD and stroke were calculated separately for each year, province and sex, adjusted by age, based on the 2010 national census estimates and projections of the population per province and region. These mortality rates were standardized by direct method, using the World Health Organization (WHO) standard population. The groups were categorized in 5-year age groups; however, we reported specific mortality rates from the population aged 35 to 79 years.

To determine if there were changes throughout the years (within variation) and provinces/regions (between variation), the mortality rates and socioeconomic variables were evaluated together in an econometric analysis based on panel data models. The analysis was performed separately for IHD and stroke and by sex. Due to a high underreporting in mortality data, indigenous territories were not included in the panel data analysis [22]. The data was described, summarized and sorted for data panel analysis. The Durbin–Wu–Hausman test was applied to choose between the fixed and random effect models. The Breusch Pagan Lagrange Multiplier test was implemented to choose between random effect and pooled models. The results of this test suggested that a pooled model regression was not suitable. Moreover, within and between variations in standard deviation were found for IHD and stroke.

Because of these findings, socioeconomic variables were merged in a polychoric correlation matrix in order to perform a factor analysis by principal components. Three factors were introduced in the analysis. The common factor was extracted by the principal-component method and then an orthogonal varimax rotation was done in order to distribute the factor loadings. A Kaiser-Meyer-Olkin test was performed to measure how suitable the data was for factor analysis. This factor was predicted for each value and then a SEI was obtained per province/region. Provinces and regions were categorized in tertiles according to the value of its SEI score median.

Analysis were performed in Microsoft Excel 2013 and in Stata IC version 14.0.

From 2012 to 2016, 15,661 deaths were attributed to CVD (IHD n = 8144 and stroke n = 7517). As shown in Fig. 1 (men), standardized and age-adjusted specific mortality rates (× 100,000 inhabitants) for IHD (Panel A) were higher in Colón for all the years, [2012 (151.9), 2014 (116.8), 2015 (112.1), 2016 (95.3)], except in 2013 (Darién 144.3). Likewise, in women (Fig. 2, Panel A), Colon presented the highest IHD mortality rates 66.2 and 58.8 (× 100,000 inhabitants) in 2012 and 2015, respectively. The region of San Miguelito had the highest IHD mortality rates in women (71.9 and 64.7) in 2013 and 2016.

In men, the provinces with the highest mortality rates (× 100,000) due to stroke (Fig. 1, Panel B) were Colón (89.4 in 2012), Coclé (81.3 in 2013), Herrera (105.2 in 2014), Darién (96.7 in 2015) and the region of San Miguelito (76.5 in 2016). In women, compared to other provinces (Fig. 2, Panel B), Herrera presented higher stroke mortality rates in 2013 (59.4) and 2015 (50.5), whereas Los Santos had the highest mortality rate in 2016 (70.6).

Additional file 1 presents the baseline characteristics of socioeconomic determinants in the HS per region and year. Compared to other regions, the indigenous territories and Darién had the most limited access to social security (4–15% and 17–22%, respectively), followed by East Panama region (32–42%). Panama region and San Miguelito region had the highest monthly incomes whereas the population of Darién (except for 2014) had the lowest mean of monthly income (below US$215.00).

Only the study population of the province of Panama (Panama, San Miguelito and West Panama) and the province of Colón achieved the compulsory years of education in the study period (mean). Overall, years of education were higher in women compared to men in all the regions for all the years (except for indigenous territories); however, average monthly income was higher in men doubling women’s monthly income.

The data panel analysis assessed temporal and geographical changes. As shown in Additional file 2, there were higher variations in all the socioeconomic variables and in IHD mortality in men throughout the provinces and regions (between). Similarly, higher variations were observed in IHD mortality in women and stroke mortality throughout the years (within variation).

Fig. 3 (Panel A) shows the SEI score ranking per province and region. The regions of San Miguelito, Panama and West Panama ranked as first, second and third, respectively, with the highest SEI scores. Darien ranked last, preceded by Bocas del Toro. Panel B shows the geographical distribution according to the SEI score. SEI scores were categorized into tertiles. The rationale behind this division was to observe if provinces with similar socioeconomic index had similar patterns with respect to the cardiovascular mortality rates. The lowest tertile was composed by Darién and Bocas del Toro. The middle tertile had Coclé, Veraguas, East Panama region, Chiriquí, Herrera, Los Santos and Colón whereas the highest tertile consisted of West Panama region, Panama region and San Miguelito region. When integrating the indigenous territories in the SEI, we observed that Emberá-Wounaan, Guna Yala and Ngöbe-Buglé were included in the lowest tertile. Darién and Bocas del Toro were recategorized to the middle tertile, while Colón was re-categorized to the highest tertile

Factorial analysis extracted the common factor (factor 1). This factor had the highest proportion (0.6044) compared to the other factors and therefore it was utilized. As shown in Table 2, the variable with the strongest correlation to factor 1 after rotation of factor loadings was years of education (0.9205), followed by social security access (0.8865). The overall Kaiser-Meyer-Olkin was 0.643, an acceptable value that indicates the correlation matrix is appropriate for factor analysis [23].

The constituent elements of our SEI index: income, education, and access to social security are key factors associated with health by being wholly or partially classed as social determinants [24‐27]. Our results indicated regional disparities in SDH within the country, yet disparities between years were stable. Education, one of the most important determinants of health, presented the highest burden in our SEI index. In Panama, the current government expenditure on education represents 3.2% of the GDP, a low percentage compared to neighboring countries like Costa Rica (7.06%), Cuba (12.83%) and Chile (4.90%) [28]. According to data from the World Bank, the government expenditure was 4.04% in the 1970s [28].

Although the average years of education was higher in women, the monthly income in men doubles the one of women. Similarly, a previous study using data from the HS concluded that men had an income 28% higher than women, despite men had higher rates of incomplete elementary and high school education [29]. In the context of these findings, gender-related policies need to be investigated-in depth. We found that social security coverage was 57–58% at a national level, a lower percentage than that reported by the INEC (an average of 75.5% from 2012 to 2016) [30]. This discrepancy might be explained because during the HS, participants were asked if they have social security coverage the week prior the survey. In addition, there has been an increase in informal employment in the country, with 40.2% of the employed population in informal employment in 2016 [31]. Remarkably, the socioeconomic variable with the strongest correlation with the SEI was social security coverage when including the indigenous population. In agreement with our results, the MPI reported that in almost all indigenous territories, more than 90% of the population was multidimensional poor, followed by Bocas del Toro (45%) and, Darién (40%) [13].

Our results suggest that there was a decrease in mortality rates in IHD and stroke during the study period. Similarly, a recent study reported a decline in the trend of stroke mortality from 2001 to 2014 and a decrease in IHD mortality that started in 2010 [10]. Whether the decrease in trends occurred similarly in provinces and regions is less clear. Overall, we observed that mortality rates varied through years and provinces, especially mortality related to IHD. This variation could be explained, in part, by certifications of the cause of death, higher coverage of ambulance services since 2007 and greater availability of health services [32] . Less variation in mortality rates was found in the regions of Panama, West Panama and San Miguelito, possibly due to the proximity to tertiary care hospitals and better death registration and certification.

For most of the years studied, Colon presented higher mortality rates for IHD in both women and men. Similarly, the highest mortality rate due to Diabetes Mellitus (DM) had been documented in this province [16]. This is despite the fact that Colon SEI score ranks above the middle tertile. Previous findings from a population-based study performed in Panama and Colon reported elevated prevalence of high blood pressure (HBP) among individuals with elementary education and illiteracy, compared to those with higher levels of education [33]. Colon, located 80 km from the capital city, is the third most populated province in the country with the largest African descent population [34]. Colon has a Free Trade Zone, the second largest in the world, several important maritime ports and part of the Panama Canal. GDP is the second in the country, superposed only by the province of Panama [35].

Colón has 74 health centers (1 secondary care hospital and 73 primary health care centers) [36] . Notably, more than half of the population of Colon had access to social security and therefore, it could be speculated that multiple factors may be playing an important role in the mortality rates observed in this province. Among the potential explanations, access and adherence to treatment deserves mention. In Colon, studies have reported that 65.6% of the population diagnosed with HBP were under antihypertensive treatment [33] and 72.2% of Colón’s population cannot afford out-of pocket expenditure in the private sector of medications [37]. In agreement with our findings, higher CVD mortality rates have been reported in municipalities that are closer to the US-Mexico border, despite having superior infrastructure and proximity to health care [25], probably explained by US acculturation, unequal access to healthcare system or specific factors in the Mexican population living in the border area [25]. Further, ethnicity, as a proxy for biological characteristics and cultural norms has been associated with increase in cardiovascular risk factors [38].

Los Santos and Herrera, with a known aging population, ranked in the middle tertile, and presented higher mortality rates for stroke, particularly in women. In line with our results, a study has reported higher self-reported prevalence of hypercholesterolemia, HBP, DM, and overweight in these provinces, but lower socioeconomic determinants [16].

Darién (with higher mortality rates for stroke in 2015 in men), was categorized in the lowest tertile of the SEI. Remarkably, Darién (along with indigenous territories) had lower amount of years of education (mean of 7 years), compared to the rest of the country. Mortality including premature mortality, increased with decreasing educational level and with increasing number of underlying diseases [24]. Similarly, there is evidence that socioeconomic risk factors may have a significant impact on stroke and mortality [26, 39]. Darien (located at the eastern end of the country-bordering Colombia) is an admixed population isolated by urban developments and limited of health public services. As of the year 2016, Darien had health centers from the MoH, yet no hospitals. Noteworthy, in the mortality context, deaths are underreported in Darien by 52%, therefore results must be interpreted with caution [28]. Interestingly, San Miguelito, a region with a high SEI index, albeit a high population density presented the highest stroke mortality rates in some years.

Several limitations must be addressed. First, we used socioeconomic variables that did not correspond to the individual’s mortality data, so causality must not be inferred and the present findings must be interpreted with caution, in order to avoid an ecological fallacy. Second, to avoid bias, we excluded indigenous territories from data panel analysis. Of note, the NMR underreporting is still an important limitation in indigenous territories. In the HS, socioeconomic variables were obtained from indigenous territories, however, because of social inequality; socioeconomic determinants in the indigenous territories differed importantly compared to other regions. When integrating this data in the factorial analysis and obtaining the SEI, the other regions had an overestimated SEI, limiting the comparisons in these regions. Finally, only five years were considered in the study because since 2012, the HS had a mayor variation in the sampling, and thus trend analysis was not applicable. However, to the best of our knowledge, this is the first study that attempts to explore CVD mortality in relation to socioeconomic determinants per province and region in Panama.