Mathematical modelling and spatial data analysis to inform TB care and control strategies in high TB incidence settings

9.0 million people developed tuberculosis (TB) in 2013, with 1.5 million dying. Detecting TB disease at an early stage minimises the harm it causes to a person's health, and reduces the number of other people whom they infect. In richer countries with little TB, anyone who has had extended contact with someone with TB is likely to be tested. Poorer countries with much higher rates of TB mostly rely on people going to health centres themselves when they have symptoms of TB. Screening the general population for TB results in cases being detected earlier, but is too expensive to be widely used in high-TB countries. Developing a better understanding of the spatial distribution of TB will enable screening to be targeted at areas where it will have the greatest effect.

I will look at TB in four settings with moderate or high levels of TB: Blantyre city, Malawi; Karonga district, Malawi; Zambia; and Western Cape Province, South Africa. There are three main questions I will explore:

1) I will identify areas within the four settings where rates of TB are highest, and determine what characteristics of areas are associated with high levels of TB. For instance, TB is often concentrated in the poorest, most crowded areas of cities, and/or in areas with high rates of HIV.

2) Most TB screening programs use one of two approaches: offering quick and convenient screening at a set location (which may change on a regular basis), or screening people at their own homes (either because they live with someone who has been diagnosed with TB, or because screening is being offered to all households in an area). In the former case, uptake of screening is likely to be highest amongst people living close to a screening location, and drop off as the distance increases. Very little is known about how quickly screening rates drop off with distance however, or how quickly levels of undiagnosed TB increase again after a screening program has stopped. I will investigate these questions using data from a completed trial of two intervention strategies in Zambia and South Africa.

3) It is possible that data on the household locations of people diagnosed with TB can be used to develop more cost-effective interventions against TB. At the moment, the World Health Organization recommends testing all people who live in the same house as a TB patient. Testing people in their neighbourhood may also be beneficial however, for two reasons. Firstly, M. tb. (the bacteria that causes most TB disease) is infectious and the patient may therefore have transmitted M. tb. to or caught M. tb. from someone who lives nearby. Secondly, certain factors such as HIV or malnutrition increase the odds of someone having TB, and these factors are often locally clustered. I will use data from the settings listed above (which include the dates and household locations of all people diagnosed with TB) to estimate the number of TB cases that could have been detected earlier if neighbourhood screening had been carried out. I will see if cost-effectiveness can be improved by only screening the neighbourhoods of patients with certain characteristics, for instance HIV positive cases. Finally, I will use computer simulations to estimate the number of new TB cases that could be prevented by screening.

The planned research will give us a much better understanding of how the distribution of TB varies within high-TB cities and districts. It also will improve our knowledge of how current interventions against TB work at a small scale, and will suggest new interventions that can be used to reduce TB in resource poor settings.

The overall aim of my Fellowship is to determine how spatial data from TB cases presenting at clinics can best be used to improve the impact and cost-effectiveness of TB care and control interventions in high incidence settings. My objectives and methods are:

Theme 1: Using spatial data analysis to understand the distribution of TB, and the effects of TB care and control interventions.
1a: Investigate the distribution of TB disease, and of individual and population level TB risk factors.
1b: Use existing data to understand the spatial and temporal effects of completed community and household based TB case finding interventions.
1c: Identify and cost candidate spatially targeted intervention approaches.

Theme 2: Using mathematical modelling to inform the development of spatially targeted case finding strategies.
2a: Develop a spatially structured mathematical model of M. tb. transmission in high incidence settings
2b: Investigate the impact and cost-effectiveness of spatially targeted TB care and control interventions.

I will use datasets from four high incidence African settings, all of which have data on the locations of TB cases' households. They are the ZAMSTAR study populations in South Africa and Zambia, and Blantyre and Karonga in Malawi.

The work will identify strategies that could improve the impact and cost-effectiveness of TB care and control measures, using spatially targeted interventions. A funding proposal for a pilot study of one or more of the approaches suggested by my work will be developed near the end of the Fellowship.

Recent WHO guidelines recommend that community-wide screening programs be considered in the highest prevalence populations, and in subpopulations with poor access to care. They also caution that indiscriminate mass screening should be avoided however, due to high costs and potential for overtreatment and harm in low risk populations. Methods that would allow screening programs and other interventions to be spatially targeted at areas where they are likely to have the greatest effect will therefore be of huge interest to a wide range of organisations. These include the WHO, and governmental and non-governmental organisations involved in TB care and control.

More effective TB care and control measures will also have a wider societal impact. Informed spatial targeting has the potential to greatly increase both the impact and cost-effectiveness of control measures, ensuring that optimum use is made of the limited resources available for TB care and control. Reducing the mean time to diagnosis for TB cases will decrease both morbidity and mortality, and reduce the economic impacts of poor health. It will also reduce onward transmission of M. tb. (the bacterium that causes the majority of cases of TB disease), reducing the incidence and prevalence of TB disease, and contributing towards TB elimination.

In addition to this, my research will have an impact on the control of other infectious diseases, by demonstrating the potential applications to disease care and control of spatial data collected from cases presenting to health centres. Data of these kind are currently not widely used in routine disease control in low-income settings, but are likely to become much more widespread in the near future as technology improves and costs continue to fall.