Predictors of diagnosed cardiovascular diseases and their spatial heterogeneity in Lagos State, Nigeria

1 Introduction

Cardiovascular diseases (CVDs), cancers, chronic pulmonary diseases, and diabetes account for more than 80% of deaths from noncommunicable diseases (NCDs) globally [1]. However, over 75% of CVD deaths occur in low- and middle-income countries (LMICs) [1]; about one-quarter of such deaths occur before age 60 [2,3]. Broadly, the risk factors of CVD can be grouped into three categories: environmental, medical, and behavioral. The environment affects health [4]; thus, rapid unplanned urbanization, high levels of exposure to smoke, fine or ultrafine particulate air pollutants, and exposure to chemicals, such as polyaromatic hydrocarbons, aldehydes, and metals, are among the environmental risk factors of CVD [5,6,7,8]. While demographic ageing, high blood pressure, raised total cholesterol, elevated blood glucose, overweight, and obesity constitute the medical risk factors [6,9,10], excessive alcohol consumption, tobacco use, poor diet, and sedentary lifestyle are among the behavioral risk factors [1,11,12]. Even though these factors are global [6], nevertheless, they vary from place to place [13], from person to person [10,14], and from time to time [12,15]. Knowing the prominent predisposing factors of CVDs in an area is, therefore, of utmost importance to healthcare systems and managers [16], because it aids the control and prevention of the disease through focused intervention and appropriate policy development [7,8,12,17].

Gebreab and Roux [17] investigated variations in coronary heart disease (CHD) mortality rates between black and white populations in the continental United States from 1996 to 2006. They initially identified significant geographic differences in CHD mortality when considering age-adjusted rates at the county level. However, after factoring in county-level disparities related to poverty, racial segregation, and racial composition, the previously observed differences in CHD mortality across regions disappeared. The findings underscore how crucial it is to consider spatial variation in addressing and reducing racial inequalities in CHD mortality. Ford and Highfield [18] measured the association between social deprivation and CVD mortality in Harris County, Texas, at the census tract level. Accounting for age, income by race, education, and other factors, there was a significant and spatially varying association between the two variables. While underpinning the importance of geography and place of residence in the relationship between social deprivation and CVD mortality, neighborhoods at greatest risk for poor health outcomes and in need of prompt interventions could be identified. Using hospital records and population data, Fabiyi and Garuba [19] examined the spatial pattern of CVD and its biomedical risk factors among neighborhoods in Ibadan City, Nigeria, and found that the relationship between the spatial pattern of CVD and the biomedical risk factors was statistically significant, with a clustered pattern observable at the neighborhood level, suggesting a very strong positive spatial autocorrelation in the city. It was deduced that CVD is linked to both affluent and impoverished communities within the city.

Kauhl et al. [20] analyzed location-specific associations between hypertension, socio-demographic population characteristics, and area deprivation based on health insurance claims. While identifying age ≥45, area deprivation, and percentage of commuters to work outside residential municipality as the main risk factors, the findings revealed important regional variations, suggesting that the risk of hypertension will be greater in persons ageing in deprived areas. Thus, the future planning and allocation of primary healthcare in northeastern Germany should take into account the impact of area deprivation. To understand the role of urban green spaces (UGSs) in CVD prevention, Rahnama and Shaddel [4] explored the factors influencing the use of UGSs among women affected by CVD in Mashhad, Iran, at the block level; the result of the analysis showed that income, level of education, and access to UGSs were important predictors of the vulnerable blocks. Identifying blocks at risk could prove beneficial for crafting prevention strategies and location-specific interventions. Similarly, using data from the Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study, Tabb et al. [21] explored the spatial patterning in the racial differences in cardiovascular health (CVH) between blacks and whites across the United States and found a significant spatial patterning in racial differences, even beyond the well-known regions labelled as the Stroke Belt and Stroke Buckle. The blacks consistently had diminishing CVH compared with whites. Hence, efforts aimed at enhancing CVH and, in the long run, diminishing inequalities between blacks and whites necessitate the use of culturally sensitive approaches, with a particular emphasis on location-based interventions and policies.

Thus, understanding the significant underlying factors that contribute to CVDs in a particular region is crucial. This knowledge plays a pivotal role in enabling targeted interventions and the development of relevant policies to effectively control and prevent the disease. Even though traditional regression methods such as ordinary least squares and generalized linear model are veritable tools for determining the predictors of a phenomenon, the outcome of a regular regression analysis produces “global” or “average” parameter estimates that are assumed to be stationary everywhere in space [20,22]. However, given the variable nature of most phenomena, the assumption that such estimated regression parameters hold equal value everywhere in geographic space is doubtful. One of the statistical techniques that produce local estimates of the spatial variability of regression results across geographic space is the geographically weighted regression (GWR) [21,22]. Instead of a single global parameter estimate, GWR generates useful results on how parameters vary over space [23], evaluating the existence or otherwise of a linear regression relationship between the observed and expected prevalence of a phenomenon, thereby exposing inherent variations [24]. GWR has been widely applied in studies of the geographic variation of CVD and its predictors [16,18,20,25]. It reveals the locations of inherent variations, thus aiding in specific interventions to control and prevent the disease. However, the identification of the local variations in the incidence of CVDs and associated risk factors is limited in LMICs, especially in Lagos State, Nigeria, despite its status as a megacity.

Consequently, this study determined the predictors of CVDs in a megacity in an LMIC and investigated their spatial heterogeneity using hospital records of patients clinically diagnosed with CVDs between 2008 and 2018 in a tertiary healthcare facility.

2 Methods

3 Results

There were 6,780 cases of medically diagnosed CVDs for the study period 2008–2018, with 51% females. The mean age of the patients was 53.1 ± 15.6 years, although the males were slightly older than the females (53.4 ± 15.6 years). Men (75.0 ± 13.9 kg) weighed more than women, even though the overall average weight was 73.5 ± 14.7 kg. The mean height of men was 1.68 ± 0.1 m, while the average height of the population was 1.66 ± 0.1 m (Table 2). The average BMI of the population was 27.6 ± 5.8, while the average for women (28.2 ± 6.1) was higher. Their educational attainment showed that 59.7% had tertiary education (50.6% males), 15.4% had secondary school education (54.4% females), 3.0% had primary school education (57.1% females), 2.3% had other forms of education (51.6% male), and 5.6% had no formal education (55.2% females). There was no information on the educational attainment of 14% of the cases (51.6% females) (Table 2).

While 72.9% were married (51.3% females), 10.1% were single (60.6% males), 5.7% were widowed (70.5% females), 0.9% were separated (60.3% females), and 0.3% were divorced (63.6% females). The data on the marital status of 10% (51.9 males) were unavailable. Among the patients, those overweight were 39.1% (51.5% females), those obese were 21.85 (57.1% females), those underweight were 1.9% (55.6% females), and those with normal weight were 36% (50.3% males). However, the weight information of 1.3% (53.4% females) was not available.

Based on age categorization, CVD patients who were between 50 and 59 years had the highest number of cases (21.3%), followed by those between 60 and 69 years (21.1%). Those that were between 90 and 100 years inclusive (0.6%) and the adolescents (0.8%) group had the lowest number of cases. The data on the age category of 0.2% of the population were not available. Overall, the women (51.0%) were more than the men and also outnumbered men in eight of the ten age groups (Table 2).

The projected population of Lagos State in 2018 was 25,615,703 [31]. Alimosho LGA was the most populous, with a population figure of 2,987,306 (11.7%). Ajeromi-Ifelodun and Mushin LGA followed with 2,094,583 (8.2%) and 1,928,542 (7.5%), respectively. At the other end of the population spectrum were Badagry, Epe, and Ibeju-Lekki LGAs, with 555,162 (2.2%), 472,292 (1.8%), and 145,263 (0.6%) people, respectively.

Alimosho LGA had the highest number of diagnosed CVD cases (19.4%) over the study period, followed by Mushin (10.6%) and Agege (9.7%) LGAs. The LGAs with the lowest number of diagnosed CVD cases during the period were Badagry (0.9%), Ajeromi-Ifelodun (0.3%), Epe (0.3%), and Ibeju-Lekki (0.2%) LGAs. Maps of both the distributions of population and cases are presented in Figure 1a and b. The LGAs with highly diagnosed CVD cases were concentrated in the north-central part of the state, while those with lower figures were in the western, eastern, and southern parts of the state.

Figure 1

(a) The year 2018 population distribution map of the 20 LGAs of Lagos State, and (b) the map of the distribution of diagnosed CVD cases (2008–2018) in Lagos State.

Out of the 104 variables included in the stepwise regression, the female gender, the male gender, the forties age group, the seventies age group, the twenties age group, the thirties age group, transport profession, accounting profession, fashion profession, the built-up extent, drug abuse, past heart diseases, maritime profession, Yoruba ethnicity, the adolescent age group, and manufacturing profession were identified as the predictors of diagnosed CVD cases in Lagos State. Thus, these predictors can be summarized as gender, age, profession, environment, lifestyle, biomedical, and ethnicity, which conform to some known risk factors. From the model summary, the R-squared value of .998 indicated that the model explains 99.8% of the variance in the dependent variable. This means that the independent variables in the model are highly correlated with the dependent variable. The adjusted R-squared value of .998 is the same as the R-squared value, which suggests that the model does not have any unnecessary variables. The standard error of the estimate of 14.847 indicated the average distance that the observed values fall from the regression line, suggesting that the model’s predictions have relatively low variability around the actual values.

Checking the catch-all plot associated with the model showed no discernible pattern in the points, implying that the model was correctly specified. Plotting the residuals of the regression model to ascertain any spatial pattern revealed an apparent pattern.

The summary of the GWR coefficients (Table 3) revealed how each coefficient varied across the 20 LGAs of Lagos State, although the local coefficient variations observable in the fit were small. The coefficient of the female gender was almost stationary at 170.41, and that of the male gender at 157.88. However, there was a negative relationship between diagnosed CVD cases and seven of the variables: the forties age group, the seventies age group, the thirties age group, the built-up extent, the maritime occupational group, the adolescent age group, and the manufacturing occupational group. Nevertheless, the small values and the narrow range of the coefficients showed that their impact on diagnosed CVD might be negligible.

Conversely, a positive relationship existed between diagnosed CVD and the remaining nine variables: Female gender, male gender, twenties age group, transport occupational group, accounting profession, fashion occupational group, drug abuse, past heart disease, and Yoruba ethnic group.

The visualization of the spatial heterogeneity of the coefficients of the 16 model variables is presented in Figure 2. The coefficient of the female gender varied across the state, with the figures rising from the west to the east. The extreme figures were in the rural LGAs located at the two ends of the state. The 16 LGAs had female gender coefficients lower than the median value, including the LGAs with high populations. Similarly, the coefficient of the male gender had a west to the east gradient of variation, with a spatial pattern comparable to that of the female gender, and nine of the LGAs had male gender coefficient values lower than the median value.

Figure 2

Maps of the variation of the coefficient of the predictors of diagnosed CVD cases (2008–2018) in Lagos State. Variables: (1) female gender, (2) male gender, (3) age 40s, (4) age 70s, (5) age 20s, (6) age 30s, (7) transport sector, (8) accounting profession, (9) fashion profession, (10) built-up area, (11) drug abuse, (12) past heart disease, (13) maritime profession, (14) Yoruba ethnicity, (15) age adolescent, and (16) manufacturing profession.

The coefficients of the age groups (adolescent, twenties, thirties, forties, and seventies) varied oppositely to those of the genders, increasing from the east to the west. However, the most extreme values were observed in the rural LGAs located in the extremes of the state, with the four LGAs (Ikorodu, Epe, Eti-Osa, and Ibeju-Lekki) on the eastern half of the state recording coefficient values lower than the median value. Following a trend akin to that of the age groups, the coefficients of the maritime and manufacturing professions demonstrated an east–west gradient. While the values of the coefficient of the maritime profession were lower than the median value occurred in four LGAs, the values of the coefficient of the manufacturing profession that were lower than the median value were observable in five LGAs.

The variation of the coefficient of the transport profession revealed a pattern different from the previously observed east–west and west–east gradients. The coefficient values lower than the median value were in the western part of the state, rising eastward. However, there was a low–high–low pattern for the coefficient values higher than the median value. Unlike the patterns of the earlier coefficients, the extremely high values were not at the eastern extreme of the state. The spatial pattern of the coefficient of the built-up area was another deviation from the dominant east–west and west–east gradients. Eti-Osa was the only LGA with a built-up area coefficient value above the median; all the other LGAs in the state had values lower than the median. The coefficients of the remaining five predictors of diagnosed CVD cases (accounting profession, fashion profession, drug abuse, past heart disease, and Yoruba ethnicity) for the period 2008–2018 exhibited a west–east variation, with values above the median occurring in the eastern part of the state. Most of the LGAs in the state, with respect to these predictor variables, had coefficient values below the median values of each predictor.

At a distance of 0.96 km, as expected, the centroid of Ikeja LGA was the closest to the CVD clinic at the LASUTH, followed by that of Agege LGA (4.18 km). Whereas the centroid of Oshodi-Isolo LGA was 6.32 km from the healthcare facility, that of Mushin, Kosofe, Ifako-Ijaiye, Shomolu, and Alimosho were 6.70, 8.22, 8.66, 8.75, and 9.55 km, respectively, from the LASUTH. While all these LGAs had their centroids at less than 10 km from LASUTH, the remaining 12 LGAs of the state were more than 10 km from the CVD clinic. The farthest centroid from the clinic was Epe LGA (66.47 km), followed by Ibeju-Lekki (63.70 km), Badagry (49.36 km), Eti-Osa (27.01 km), Ojo (24.68 km), and Ikorodu (21.09 km). The correlation of diagnosed CVD cases at the LGA level with distances from the LGAs to the healthcare facility at LASUTH was −0.61, while the correlation of diagnosed CVD cases in LGAs with the LGA populations was 0.63.

4 Discussion

A total of 104 variables were considered as possible predictors of diagnosed CVD in Lagos State during the study period, and 16 variables were identified as contributing to the observed diagnosed CVD pattern. However, of these variables, the female gender explained 99.8% of the observed cases. This might not be due to biology, but it could be because women use healthcare services more than men [45]. The plausibility of this assumption is further strengthened by the number of women in the study population being more than the men in the age categories of adolescents to sixties. Wang et al. [46] observed that the most significant gap in healthcare facility utilization between men and women occurs between the ages of 16 and 60 years.

Also, the high contribution of females to the model might be due to their susceptibility to various risk factors for CVD such as hypertension, obesity, and diabetes, which are prevalent among women in Lagos State [47,48]. Women in the study group were more obese than men, and women aged 50–60 years were most diagnosed with CVD during the study period, which corroborates the fact that women’s risk for CVD increases with age, particularly after menopause due to hormonal profile changes [48]. In addition, the high contribution of the female gender could also be as a result of the residual effect of pregnancy-related complications, such as preeclampsia and gestational diabetes, which have been known to increase the risk of CVD in women later in life [48], and the urban nature of Lagos State makes the incidence of these complications rife [47].

The map in Figure 1b shows a concentration of LGAs with elevated numbers of diagnosed cases of CVD situated in the central region of the state, near LASUTH in Ikeja, which is also in the central part of the state. This clustering may be attributable to the inverse relationship between diagnosed CVD cases and the distances from the LGAs to the healthcare facility. With a correlation coefficient of −0.61, it indicates that as the distance to the healthcare facility increases, the number of cases decreases. While this does not necessarily imply a reduction in the occurrence of the disease, it does suggest that greater distances act as a barrier to seeking diagnosis. This finding aligns with that of Oviasu [49], who found that, for NCDs, proximity to a healthcare facility facilitates early diagnosis. Hence, geography plays a crucial role in explaining the observed pattern of diagnosed CVD cases in Lagos State, lending support to the use of the GWR model [18].

While there was a moderate correlation (0.63) between the population and the number of diagnosed cases at the LGA level, there were evident anomalies. A comparison of the maps presented in Figure 1 reveals that there was no straightforward one-to-one relationship between the population of LGAs and the incidence of diagnosed CVD within those LGAs. Some LGAs with larger populations had fewer diagnosed CVD cases compared to LGAs with smaller populations. For instance, Ajeromi-Ifelodun LGA, which ranked second in terms of population within the state, had the fewest diagnosed CVD cases during the study period. In addition, a considerable number of LGAs with high numbers of diagnosed CVD cases also exhibited high population densities. An earlier study [50] had already highlighted issues of residential overcrowding in these areas, confirming that housing could play a negative role in the health of the heart [18,51].

The alignment of the highest coefficient values for gender and age groups with rural LGAs may be linked to the state’s cosmopolitan character and its significant industrialization. In LMICs such as Nigeria, there is an observable difference in the life expectancy of adults, with rural inhabitants tending to live longer than their urban counterparts due to disparities in NCDs, traffic accidents, and other external factors [52,53].

The five occupational groups among the 16 predictors can be classified into two, exhibiting two distinct patterns of heterogeneity. The manufacturing and maritime occupation groups displayed a common east–west spatial pattern, whereas the accounting, transport, and fashion groups demonstrated a west-to-east variation. Besides the spatial similarity, these occupational groups also exhibited similar associations with diagnosed CVD. The latter group, consisting of sedentary occupations, showed a positive correlation with diagnosed CVD, while the former group, which is non-sedentary, displayed a negative relationship with diagnosed CVD. This aligns with prior research findings that suggest that a sedentary lifestyle elevates the risk of CVD [11,12,33].

Regarding the spatial pattern of the built-up area coefficient, it is noteworthy that only Eti-Osa LGA exhibited a coefficient value higher than the median. Eti-Osa LGA is characterized by upscale neighborhoods like Ikoyi, Victoria Island, and Lekki Lagoon, which are known for their meticulous planning and development. Previous research findings have indicated that living conditions at the neighborhood level can worsen diseases such as CVD, leading to poorer health outcomes [17,18,25]. Given that the built-up area coefficient showed a negative relationship with diagnosed CVD cases in the state, the affluence of these neighborhoods may have played a role in shaping the observed spatial pattern.

This study identified the predictors of diagnosed CVD in Lagos State for the study period 2008–2018, using GWR; it showed the association between the cases of the disease and its predictors, revealing their spatial variation across the study area. The ability of the technique to highlight how and where the strength of associations exists between a phenomenon and its explanatory variables makes the spatial regression model readily interpretable [20], while the method is also robust to multicollinearity among explanatory variables [21,42]. However, this study has some limitations. The data were restrictive in its spread, being hospital records of diagnosed cases attending the clinic at the LASUTH; a more extensive database might have revealed a more detailed picture of the variation of diagnosed CVD and its predictors in Lagos State. Also, the use of LGAs as the unit of analysis, as opposed to smaller and more homogeneous area units such as Wards, might reduce the applicability of the findings since interventions at curbing NCDs are best at community and neighborhood levels [18].

5 Conclusion

This study investigated the predictors of diagnosed CVD and their spatial heterogeneity. The findings underscore the significance of considering spatial variations when examining diagnosed CVD, highlighting gender as a prominent predictor. Notably, spatial disparities were observed across all 16 predictor variables. Consequently, it is recommended that gender-specific research and healthcare initiatives be implemented to better address the unique risk factors and challenges faced by women in relation to CVD. While this research provided plausible explanations for the observed spatial heterogeneity, the limitations stemming from the lack of comprehensive data at smaller geographical units highlight the necessity for further investigations to comprehensively unravel the dynamics of the predictor variables and their influence on diagnosed CVD in the study area.