Designing Health Interventions for Hypertension Control Using Geospatial Analysis

Overview

1. Who is Boni ?
2. Defintion of key concepts
3. Problem statement in Sub-Saharan Africa context
4. Risk factors for hypertension and Why geospatial matter
5. Challenges in geospatial analysis of hypertension in Africa
6. Benin’s STEPS Survey and Geospatial data
7. Data Manipulation
8. Comparison of hypertension prevalence in 2008 and 2015
9. Discussion and Conclusion

Who is Boni ?

What is prevalence ?

What is hypertension (HTN) ?

• Blood pressure (BP) = force of blood against inner walls of arteries (major blood vessels in the body)
• HTN = high blood pressure
• You can’t see it, You can’t even feel it sometimes: “silent killer”
• Persistent high BP impact negatively your heart (heart attack), brain (stroke), eyes (blindness), kidney (kidney failure) and others.

How big is the thing in Africa ?

• Sub-Saharan Africa (SSA) is experiencing a surge in the burden of cardiovascular disease (CVD) (GBD, 2019)
• In 2019, more than one million deaths attributable to CVD in SSA (GBD Compare, 2019)
• CVD is set to overtake infectious diseases as the leading cause of mortality in the region in the next decade (WHO, 2014)
• Epidemiological changes due to profound cultural and socioeconomic transformation characterized by westernization of lifestyles with sedentary habits and unhealthy diets (WHO, 2014)

How big is the thing in Africa ?

• HTN is one of the strongest risk factors for CVD
• HTN is a major cause of premature death worldwide (WHO, 2021)
• Highest prevalence of HTN is in WHO African Region (27%) vs WHO Region of Americas (18%) (WHO, 2021)
• Overall, only 18% of HTN patients are treated and 7% are controlled (WHO, 2021)
• Highlight the need for implementation of timely and appropriate strategies for diagnosis, control, and prevention

Risk factors for HTN

Modifiable risk factors:

• Unhealthy diets
• Physical inactivity
• Consumption of tobacco and alcohol
• Overweight or obesity

Non-modifiable risk factors:

• Family history of hypertension
• Age (90% of chance by age 55 to 65)

Why geospatial analysis in HTN ?

• Beyond individual level factors, emerging research has demonstrated that environment, noise, may be important risk factors for CVD
• Studies has shown seasonal pattern with higher BP in winter and lower BP in summer in western countries (Argilés, 1998)
• The seasonal change in BP is correlated with several climatic parameters namely external temperature, humidity, rainfall, and daylight span suggesting that BP level and seasonality would vary across locations with different climates (Argilés, 1998, Spósito, 2000, Argilés, 2004)

Why geospatial analysis in HTN ?

• Understanding the temporal trend and geographic, seasonal patterns of hypertension may inform the delivery of targeted community health interventions
• It will help to link their location with factors like climate change (rain falls, sun), existing of green spaces, forest, air pollution, socio-anthropological factors (health behavior of a community)

World Heart Day

• Yesterday: 29th Sept was World Heart Day, WHF
• Opportunity for everyone to stop and consider how best to use ❤️ for humanity, for nature, and for you
• Beating cardiovascular disease is something that matters to every beating heart

Challenges in HTN geospatial analysis in Africa

• Not aware of existence of longitudinal geospatial health data focus on HTN in Africa
• Demographic Health Survey (DHS) collect geospatial data but no data on HTN apart from few countries in the recent DHS
• When BP collected in DHS, just a small-sample of the population
• No geospatial data collected in STEPS surveys
• Difficult to do a proper geospatial modelling in HTN

STEPS Survey and Geospatial data

• WHO STEPwise approach to non-communicable disease (NCD) risk factor surveillance (STEPS) is a simple, standardized method for collecting, analysing and disseminating data on key NCD risk factors in countries
• Analysis based on STEPS surveys reports conducted in Benin in 2008 and 2015
• Why Benin? At least data collected at two different years was needed
• GPS data of adminsitrative regions of Benin: obtained from The Humanitarian Data Exchange https://data.humdata.org/

Data Manipulation

At this stage we will import both databases and do some data manipulation.

Let’s install and load all required packages here:

# Load packages ----

pacman::p_load(
  dplyr,
  ggimage,
  ggplot2,
  glue,
  here,
  readr,
  sf,
  tidyr,
  tmap
)

Data manipulation

After loading the required packages, let’s load each datasets (prevalence data and geospatial data)

## loading prevalence data using Benin's steps survey report
prev_data_raw <- read_csv2(file = here("data/htn.csv"))

## loading geospatial data from The Humanitarian Data Exchange website 
benin_dpts_raw <- st_read(dsn = here("data/ben_adm_1m_salb_2019_shapes/ben_admbnda_adm1_1m_salb_20190816.shp"))

Data manipulation

Let’s clean our geodata and merge both databases

# Clean geospatial data
clean_benin_dpts_data <- function(x){
  x %>%
    mutate(
      adm1_name = ifelse(adm1_name == "Atakora", "Atacora", adm1_name)
    )
}
benin_dpts <- clean_benin_dpts_data(x = benin_dpts_raw)

# Merge prevalence data and geospatial data
merge_data <- function(x, y){
  
  left_join(
    x = x %>% select(adm1_name, shape_Leng, Shape_Area),
    y = y %>% select(-ic),
    by = c("adm1_name" = "department")
  ) %>%
    select(year, dpt = adm1_name, prev, shape_length = shape_Leng, shape_area = Shape_Area)
  
}
merged <- merge_data(x = benin_dpts, y = prev_data_raw)

Geographic differences in HTN prev from 2008 to 2015

Let’s compare the changes in prevalence of HTN in 2008 versus 2015 according to administrative regions in Benin

make_prev_map <- function(x, yr){
  tm_shape(shp = x %>% filter(yr == year)) +
    tm_polygons(col = "prev", breaks = seq(from = 15, to = 40, by = 5)) +
    tm_text(text = "dpt", size = 0.5) +
    tm_layout(legend.outside = TRUE)
}

prev_map_08 <- make_prev_map(x = merged, yr = 2008)

prev_map_15 <- make_prev_map(x = merged, yr = 2015)

facetted_prev_map <- tmap_arrange(prev_map_08, prev_map_15)

Geographic differences in HTN prev from 2008 to 2015

Inset a plot within the map

Here, we will inset a graph within the map. Let’s see how to do it!

create_inset_map <- function(x, y){
  
  # Get the locations of centroids ----
  
  dpts <- unique(x$dpt)
  
  coords <- x %>%
    select(dpt, geometry) %>%
    distinct() %>%
    st_centroid() %>%
    st_coordinates()
  
  dpt_centroids <- bind_cols(dpt = dpts, coords)
  
  # Get maximum prevalence (to set maximum of y-axis)
  
  max_prev <- max(x$prev) |> ceiling()
  
  # Create nested tibble of inset plots and centroid coordinates
  
  nested <- x %>%
    st_drop_geometry() %>%
    select(year, dpt, prev) %>%
    nest(data = -dpt) %>% 
    mutate(
      plots = purrr::map(data, function(d){
        ggplot(data = d, mapping = aes(x = year, y = prev)) +
          geom_col(mapping = aes(fill = as.factor(year))) +
          scale_x_continuous(breaks = c(2008, 2015)) +
          scale_y_continuous(limits = c(0, max_prev)) +
          scale_fill_manual(breaks = c(2008, 2015), values = c("blue", "red")) +
          labs(x = NULL, y = NULL) +
          theme_bw() +
          theme_transparent() +
          theme(
            legend.position = "none",
            axis.text.x = element_blank(),
            axis.ticks.x = element_blank()
          )
      }),
      width = 0.75,
      height = 0.75
    ) %>%
    select(-data) %>%
    left_join(dpt_centroids, by = "dpt")
  
  
  # Create map with inset plots ----
  
  ggplot() +
    geom_sf(data = y %>% select(adm1_name, geometry)) +
    theme_void() +
    geom_subview(
      data = nested,
      mapping = aes(
        x = X,
        y = Y,
        subview = plots,
        width = width,
        height = height
      )
    )
}

inset_map <- create_inset_map(x = merged, y = benin_dpts)

Inset a plot within the map

Map with diverging shades

Let’s visualise the quantitative changes from 2008 to 2015 using a map with diverging shades

create_diverging_palette_map <- function(x, y){ 
  
  # calculate the difference in prev of htn between 2015 and 2008
  diff_df <- x %>%
    st_drop_geometry() %>%
    select(year, dpt, prev) %>%
    tidyr::pivot_wider(names_from = year, values_from = prev) %>%
    mutate(diff = `2015` - `2008`)
  
  # join this difference with the geodata
  diff_geo <- left_join(
    x = y %>% select(adm1_name, geometry),
    y = diff_df %>% select(dpt, diff),
    by = c("adm1_name" = "dpt")
  )
  
  # create the map
  tm_shape(shp = diff_geo) +
    tm_fill(
      col = "diff",
      title = "Change in prevalence",
      style = "cont",
      palette = "-RdYlGn"
    ) +
    tm_text(text = "adm1_name", size = 0.75) +
    tm_borders() +
    tm_layout(legend.outside = TRUE) 
  
  
}  

diverging_map <- create_diverging_palette_map(x = merged, y = benin_dpts)

Map with diverging shades

Discussion and Conclusion

• Hypertension is starkly prevalent in Benin
• The burden of hypertension has increased from 2008 to 2015 and the highest prevalence has moved from Southern to Northern regions of the country
• Benin capital city is in the southern part with many other main cities around. They have been probably an epidemilogical transition there which may be the reason for the high prevalence in 2008 survey

Discussion and Conclusion

• The detection of higher prevalence in that region may have caused more prevention strategies implementation which has caused the reduction of the prevalence of hypertension
• At the same time, the Northern region have started to more industrialised which may be the cause of the raise in the burden of hypertension in the region

Discussion and Conclusion

• This mapping will serve just as a first step to show the need to conduct geospatial modelling.
• This would help understand the factors associated with the temporo-spatial difference and change in prevalence of hypertension in geographical regions and the result will be useful for targeted interventions in the community to control hypertension

References

1. GBD 2019 Diseases and Injuries Collaborators. Global burden of 369 diseases and injuries in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2020;396(10258):1204-1222
2. GBD Compare (2019). Institute for Health Metrics and Evaluation website.   https://vizhub.healthdata.org/gbd-compare/. Accessed September 26, 2022
3. World Health Organization. Global status report on noncommunicable diseases 2014. Geneva: World Health Organization; 2014

References

4. World Health Organization. Fact sheets hypertension 2021. Geneva: World Health Organization; 2021
5. Argilés A, Mourad G, Mion C. Seasonal changes in blood pressure in patients with end-stage renal disease treated with hemodialysis. N Engl J Med. 1998;339:1364–1370. doi: 10.1056/NEJM199811053391904
6. Spósito M, Nieto FJ, Ventura JE. Seasonal variations of blood pressure and overhydration in patients on chronic hemodialysis. Am J Kidney Dis. 2000;35:812–818

References

7. Argilés A, Lorho R, Servel MF, Chong G, Kerr PG, Mourad G. Seasonal modifications in blood pressure are mainly related to interdialytic body weight gain in dialysis patients. Kidney Int. 2004;65:1795–1801. doi: 10.1111/j.1523-1755.2004.00569.x

Acknowledgements

• Thanks to the Afrimapr for the opportunity
• Thanks to my R mentor Prof. Nono Gueye
• Thanks to my cheerleader for her tremendous support

Contact

• Personal website: https://bonimaximeale.com/
• Twitter: https://twitter.com/DrAleBoni
• LinkedIn: https://www.linkedin.com/in/boni-maxime-ale-md-msc-mph-ba541b207
• Health Data Acumen: https://healthdatacumen.com/