Tuberculosis is a leading cause of morbidity and mortality, with an estimated 10.0 million new TB cases and 1.3 million deaths annually worldwide. Despite the availability of effective treatment, TB transmission and incidence are declining only very slowly in many high-burden countries, including Bangladesh.
The End TB Strategy, launched by the World Health Organization as part of its post-2015 agenda, set goals to reduce TB incidence by 50% by 2025, and by 90% by 2035, a rate similar to current levels of TB in low incidence countries. Unfortunately, given the present slow decline in TB incidence of 1.5% per year, it is unlikely that these targets will be met unless concerted efforts are made to rapidly increase the rate of decline in TB incidence.
One key factor to consider in leveraging current and future TB interventions is the geographic heterogeneity of TB prevalence. It is known that TB, along with many of its common risk factors, such as low socioeconomic status, poor living conditions, migration status, and HIV infection, tends to cluster in hyperendemic “hotspots”. These high-incidence areas act as reservoirs of infection and drive secondary transmissions and TB epidemics within the larger community. Therefore, targeting hotspots may be more effective in reducing TB incidence at the local (e.g., city) level compared with interventions that are delivered to the general population without any attempt at targeting those at highest risk.
This project had two primary aims. The first was to characterize heterogeneity in TB incidence at the ward level in Dhaka South City Corporation (DSCC). The second was to estimate the relative benefit of targeting TB interventions, in the form of active case finding and preventive therapy, to TB “hotspots” within DSCC. Using TB notification rates and individual-level data collected within selected reporting centers, we estimated the geographic distribution of TB in DSCC and identified hotspots in which notification rates were highest. We then developed mathematical models to estimate the impact of hotspot-targeted interventions in reducing TB incidence over a 20-year horizon.
Showing unadjusted notification rates (in units per 100,000 per year) and corresponding notification rates after adjustment for observed correlations between ward of residence and ward of reporting center. The first (left-most) panel shows data from 2010, the second (right-most) represents 2017 data.
Showing the average annual change (% per year) in estimated TB incidence between the years 2010 and 2017 at the ward level, and the corresponding data after adjustment for correlations between ward of residence and ward of reporting. Blue shading indicates a decline in TB incidence during the 7-year period (with darker shades representing steeper declines), whereas red shading indicates an increase (with darker shades representing greater increases).
On the left, we have the estimated ward-level TB incidence in 2017, indicated by the color. In the middle we have the projected TB incidence in 2022, after 5 year period of untargeted, citywide TB intervention. Colors indicate TB incidence on 2022, and the bubbles indicate the size of reductions (the reductions in number of incidence TB cases in 2022). The right-most map is the projected TB incidence in 2022, after 5 year period of targeted TB intervention. Colors indicate TB incidence on 2022, and the bubbles indicate the size of reductions (the reductions in number of incidence TB cases in 2022).
On the left, we have the estimated ward-level TB incidence in 2017, indicated by the color. In the middle we have the projected TB incidence in 2037, after 20 year period of untargeted, citywide TB intervention. Colors indicate TB incidence on 2037, and the bubbles indicate the size of reductions (the reductions in number of incidence TB cases in 2037). The right-most map is the projected TB incidence in 2037, after 20 year period of targeted TB intervention. Colors indicate TB incidence on 2037, and the bubbles indicate the size of reductions (the reductions in number of incidence TB cases in 2037).
Shown are percent reductions based on the baseline model, and ranges within the parentheses are based on low and high transmission models. Wards marked with an asterix were wards included in targeted interventions.
| Ward | ACF-citywide | ACF-targeted | IPT-citywide | IPT-targeted |
|---|---|---|---|---|
| 44 | 8.4 (5.8 – 11.1) | 0.0 | 14.8 (11.8 – 17.9) | 0.0 |
| 54 | 8.0 (5-6 – 10.7 | 0.0 | 14.3 (11.5 – 17.4) | 0.0 |
| 53 | 9.4 (6.4 – 12.4) | 0.0 | 15.9 (12.5 – 19.4) | 0.0 |
| 52 | 8.1 (5.6 – 10.8) | 0.0 | 14.4 (11.5 – 17.5) | 0.0 |
| 47 | 8.0 (5.5 – 10.6) | 0.0 | 14.3 (11.5 – 17.3) | 0.0 |
| 46 | 8.0 (5.6 – 10.6) | 0.0 | 14.3 (11.5 – 17.3) | 0.0 |
| 12 | 11.6 (7.8 – 14.8) | 0.0 | 18.5 (14.1 – 22.1) | 0.0 |
| 43 | 7.9 (5.5 – 10.5) | 0.0 | 14.1 (11.4 – 17.2) | 0.0 |
| 45 | 9.4 (6.4 – 12.4) | 0.0 | 16.0 (12.5 – 19.4) | 0.0 |
| 51 | 8.7 (6.0 – 11.5) | 0.0 | 15.1 (12.0 – 18.4) | 0.0 |
| 37 | 8.0 (5.5 – 10.5) | 0.0 | 14.3 (11.5 – 17.2) | 0.0 |
| 50* | 18.3 (13.6 – 21.4) | 54.2 (43.0 – 61.2) | 25.9 (20.7 – 29.3) | 73.6 (64.9 – 78.7) |
| 42 | 9.5 (6.4 – 12.5) | 0.0 | 16.1 (12.6 – 19.6) | 0.0 |
| 7 | 11.0 (7.5 – 14.2) | 0.0 | 17.8 (13.7 – 21.5) | 0.0 |
| 36 | 10.5 (7.1 – 13.6) | 0.0 | 17.2 (13.3 – 20.8) | 0.0 |
| 30 | 13.4 (9.3 – 16.8) | 0.0 | 20.5 (15.8 – 24.4) | 0.0 |
| 32 | 12.0 (8.2 – 15.4) | 0.0 | 19.0 (14.6 – 22.7) | 0.0 |
| 48 | 11.5 (7.9 – 10.8) | 0.0 | 18.4 (14.2 – 22.1) | 0.0 |
| 40 | 14.1 (9.9 – 17.5) | 0.0 | 21.3 (16.5 – 25.1) | 0.0 |
| 29 | 13.6 (9.4 – 17.0) | 0.0 | 20.7 (16.0 – 24.5) | 0.0 |
| 35* | 22.3 (18.2 – 24.8) | 63.2 (54.0 – 68.9) | 30.3 (25.8 – 33.1) | 79.8 (73.1 – 83.8) |
| 28* | 15.9 (11.4 – 19.2) | 48.9 (37.8 – 56.9) | 23.2 (18.2 – 27.0) | 69.8 (60.7 – 75.8) |
| 24 | 12.8 (8.8 – 16.2) | 0.0 | 19.9 (15.3 – 23.7) | 0.0 |
| 41 | 8.5 (5.8 – 11.3) | 0.0 | 14.9 (11.8 – 18.1) | 0.0 |
| 31* | 15.9 (11.4 – 19.3) | 49.1 (37.9 – 57.1) | 23.3 (18.3 – 27.1) | 69.9 (60.9 – 75.9) |
| 39 | 9.5 (6.5 – 12.5) | 0.0 | 16.1 (12.6 – 19.6) | 0.0 |
| 38 | 11.1 (7.5 – 14.3) | 0.0 | 17.9 (13.8 – 21.6) | 0.0 |
| 34 | 14.6 (10.2 – 18.0) | 0.0 | 21.8 (16.9 – 25.6) | 0.0 |
| 49 | 9.8 (6.7 – 12.9) | 0.0 | 16.5 (12.8 – 20.0) | 0.0 |
| 25* | 17.0 (12.5 – 20.4) | 51.5 (40.4 – 59.2) | 24.5 (19.4 – 28.2) | 71.6 (62.9 – 77.4) |
| 33* | 16.2 (11.6 – 19.5) | 49.6 (38.4 – 57.5) | 23.6 (18.5 – 27.3) | 70.2 (61.3 – 76.2) |
| 9 | 10.0 (6.8 – 13.1) | 0.0 | 16.7 (12.9 – 20.2) | 0.0 |
| 27 | 13.7 (9.5 – 17.1) | 0.0 | 20.9 (16.1 – 24.7) | 0.0 |
| 26 | 9.8 (6.6 – 12.8) | 0.0 | 16.4 (12.7 – 19.9) | 0.0 |
| 8* | 15.5 (11.0 – 18.9) | 48.0 (36.9 – 56.5) | 22.8 (17.8 – 26.6) | 69.1 (60.0 – 75.3) |
| 10* | 17.7 (13.2– 21.0) | 53.0 (42.0 – 60.5) | 25.3 (20.2 – 28.9) | 72.7 (64.2 – 78.2) |
| 23 | 14.1 (9.8 – 17.5) | 0.0 | 21.3 (16.5 – 25.1) | 0.0 |
| 22 | 10.6 (7.1 – 13.8) | 0.0 | 17.3 (13.4 – 21.0) | 0.0 |
| 6 | 11.7 (7.9 – 15.0) | 0.0 | 18.6 (14.3 – 22.3) | 0.0 |
| 20 | 10.1 (6.8 – 13.2) | 0.0 | 16.8 (13.0 – 20.3) | 0.0 |
| 18 | 14.8 (10.4 – 18.2) | 0.0 | 22.0 (17.1 – 25.8) | 0.0 |
| 13 | 110 (7.5 – 14.3) | 0.0 | 17.9 (13.7 – 21.5) | 0.0 |
| 5 | 13.7 (9.5 – 17.2) | 0.0 | 20.9 (16.1 – 24.7) | 0.0 |
| 14 | 13.8 (9.6 – 17.2) | 0.0 | 20.9 (16.2 – 24.8) | 0.0 |
| 4 | 13.5 (9.4 – 17.0) | 0.0 | 20.7 (15.9 – 24.5) | 0.0 |
| 11 | 13.8 (9.6 – 17.2) | 0.0 | 21.0 (16.2 – 24.8) | 0.0 |
| 21* | 16.0 (11.5 – 19.4) | 49.1 (38.0 – 57.1) | 23.4 (18.3 – 27.1) | 69.9 (60.9 – 76.0) |
| 19 | 11.1 (7.5 – 14.4) | 0.0 | 18.0 (13.8 – 21.7) | 0.0 |
| 16 | 12.0 (8.1 – 15.3) | 0.0 | 18.9 (14.5 – 22.7) | 0.0 |
| 2* | 15.3 (10.9 – 18.7) | 47.7 (36.5 – 55.8) | 22.6 (17.6 – 26.4) | 68.8 (59.7 – 75.0) |
| 1* | 17.7 (13.1 – 21.0) | 52.8 (41.9 – 60.4) | 25.2 (20.1 – 28.8) | 72.6 (64.0 – 78.2) |
| 15* | 20.4 (16.0 – 23.3) | 58.7 (48.6 – 65.4) | 28.2 (23.3 – 31.4) | 76.7 (69.1 – 81.4) |
| 17 | 14.2 (9.9 – 17.6) | 0.0 | 21.4 (16.6 – 25.2) | 0.0 |
| 3 | 10.9 (7.4 – 14.1) | 0.0 | 17.7 (13.6 – 21.4) | 0.0 |
| DSCC | 14.4 (10.3 – 17.7) | 20.1 (16.0 – 22.9) | 21.6 (17.0 – 25.2) | 27.6 (24.3 – 29.6) |
Our ward-specific compartmental model divided the population into three compartments based on their TB status: Uninfected (individuals that have not been exposed to TB), LTBI (individuals with latent TB infection), and Active TB (individuals with infectious TB disease). We modeled two interventions: Active Case Finding (ACF) was modeled to reduce time to diagnosis (thus, increasing the rate marked in blue); and preventive therapy (PT) was modeled to prevent reactivation and progression of disease (thus reducing the rates marked in red).
