Spatial and temporal projections of the prevalence of active tuberculosis in Cambodia

Abstract

Introduction: Cambodia is among the 30 highest burden of tuberculosis (TB) countries. Active TB prevalence has been estimated using nationally representative multistage sampling that represents urban, rural and remote parts of the country, but the prevalence in non-sampled communes remains unknown. This study uses geospatial Bayesian statistics to estimate point prevalence across Cambodia, and demographic modelling that accounts for secular trends in fertility, mortality, urbanisation and prevalence rates to project the future burden of active TB.

Methods: A Bayesian hierarchical model was developed for the 2011 National Tuberculosis Prevalence survey to estimate the differential effect of age, sex and geographic stratum on active TB prevalence; these estimates were then married with high-resolution geographic information system layers to project prevalence across Cambodia. Future TB projections under alternative scenarios were then derived by interfacing these estimates with an individual-based demographic model.

Results: Strong differences in risk by age and sex, together with geographically varying population structures, yielded the first estimated prevalence map at a 1 km scale. The projected number of active TB cases within the catchment area of each existing government healthcare facility was derived, together with projections to the year 2030 under three scenarios: no future improvement, c ontinual r eduction and GDP projection.

Conclusion: Synthesis of health and geographic data allows likely disease rates to be mapped at a high resolution to facilitate resource planning, while demographic modelling allows scenarios to be projected, demonstrating the need for the acceleration of control efforts to achieve a substantive impact on the future burden of TB in Cambodia.

Keywords: geographic information systems; mathematical modelling; tuberculosis.

Figure 1

Methodology and data. Overview of the data sources and model framework in the manuscript is presented in this flow chart. National Tuberculosis Prevalence Surveys (NTP) for 2011 were made available by the National Center for Tuberculosis and Leprosy Control (CENAT) in Cambodia. A summary of the methodology is represented by the model framework: (A) TB-NTP model, and (B) demographic epidemiological model of Cambodia (DEMOKH). The projections to the year 2030 were modelled under three scenarios: no future improvement (should the prevalence rate be maintained at the 2011 rates), continual reduction (should the prevalence rate decline at the same rate from 2002 to 2011 and gross domestic product ( GDP) projections (should the fall in the prevalence rate mirror the projected rises in GDP) as represented in the bottom panel. TB, tuberculosis.

Figure 2

Differences in prevalence of bacteriologically positive tuberculosis (TB) across exposures. Modelled bacteriologically positive TB prevalence per 100 000 population estimates are represented in the log base 10 scale and are based on individual-level data from the 2011 NTP survey. The prevalence of active TB varies across age and sex of the individual, and stratum level of his/her living environment. The interpretation of violin plots is like box plots: they display the probability density of the prevalence estimates at different values. Points are posterior median prevalence estimates, curves are posterior distributions of the parameters truncated to within 95% CI.

Figure 3

Population age distribution by province and urbanisation category. The population pyramids by age and sex of the 24 provinces of Cambodia in 2008 is classified by urbanisation extent (urban, the darker area; or rural, the lighter area).

Figure 4

Projected point prevalence and cases of active tuberculosis (TB) across Cambodia, 2010–2030. Darker regions indicate higher prevalence of active TB. Three scenarios were modelled—(1) no future improvement: future TB prevalence is projected based on the estimates of age specific point prevalence in 2011 assuming no further improvement to the TB control programme; (2) continual reduction: the reduction in age-specific point prevalence between the 2002 and 2011 NTP surveys was extrapolated to 2030; and (3) gross domestic product ( GDP) projections: the reduction in age-specific point prevalence of TB brought about by rapid development projected for Cambodia until 2030. No future improvement and continual reduction may constitute lower and upper plausible bounds for future prevalence of active TB.

Figure 5

Projected change in cumulative active tuberculosis (TB) cases and point prevalence in Cambodia from 2010 to 2030. Of the three scenarios presented, no future improvement and continual reduction may constitute lower and upper plausible bounds for future prevalence of active TB in Cambodia; the intermediate third scenario, g ross domestic product ( GDP) projections, is based on the assumption that TB point prevalence falls mirror projected rises in GDP. The no future improvement scenario assumes no further improvement to the TB control programme after 2011; the future TB prevalence projected are based on the estimates of age-specific point prevalence in 2011. The continual reduction model accounts for the reduction in age-specific point prevalence between the 2002 and 2011 NTP surveys, which was then extrapolated to 2030. the GDP projections model accounts for the rapid development Cambodia could experience and the reduction it brings about to age-specific point prevalence of TB. The darker lines indicate the median values.

Figure 6

Projected catchment size of the government healthcare facility in Cambodia in 2030. The size of the points corresponds to the number of active tuberculosis (TB) cases within the catchment area of a government healthcare facility. Two scenarios for 2030 are presented, no future improvement (points in black) and continual reduction (points in hues of red). The burden of the healthcare facility is expected to change over time, with 2010 in blue vs 2030 in dark grey (for the no future improvement scenario) and red (for a continual reduction scenario).