Predicting the cost and impact of diagnostic tests for TB

Inside eLife
  • Views 266
  • Annotations

How can we improve the diagnosis of tuberculosis?

From the paper -

A user-friendly, open-source tool to project impact and cost of diagnostic tests for tuberculosis by Dowdy et al.

Tuberculosis is an infectious bacterial disease caused predominantly by the microorganism Mycobacterium tuberculosis. Although the number of deaths from tuberculosis has been falling in recent years, the disease still kills more than 1 million people every year, mainly in developing countries. Tuberculosis can be treated with antibiotics, but the emergence of bacteria that are resistant to existing drugs is threatening efforts to eradicate the disease.

Scanning electron microscope image of mycobacterium tuberculosis

Scanning electron microscope image of mycobacterium tuberculosis.

CCBY-NC 2.0 Sanofi Pasteur

Preventing the spread of tuberculosis is heavily dependent on accurate diagnosis of individuals with the disease. This is challenging because the initial symptoms are often mild, usually just a cough, which means that someone can spread the disease to many others over a period of several months before the symptoms become worse—fever, night sweats, and weight loss—and they realize that they are sick.

Multiple diagnostic strategies are available, from the relatively low-tech—examining sputum samples under a microscope to detect tuberculosis bacteria—to more sophisticated tests that can detect bacterial DNA and determine whether the bacteria are drug-resistant in less than 2 hr.

Choosing which diagnostic strategy to adopt can be challenging because the optimal solution in a region will depend on the specific local conditions. To overcome this problem, Dowdy et al. have developed a computer program that enables decision-makers to input four key parameters that describe the tuberculosis situation in their region, and to obtain 5-year projections of the rate of new infections, mortality, and total costs likely to result from adopting any of nine different diagnostic strategies.

The four parameters are the number of new cases of tuberculosis each year (incidence), the proportion of new cases that are multi-drug resistant, the proportion of the adult population that has HIV, and the local costs of various diagnostic techniques and treatments. Since the entire computer program is written in a freely available open-source programming language (Python), any user can tweak these parameters to provide a more precise fit to their own region. Alternatively, the standard version of the program can be run directly from a website without any need to interact with computer code.

This model is the first to enable local decision-makers to evaluate the impact of different diagnostic strategies for tuberculosis under the conditions specific to their region. The model predicts, for example, that in areas where there is little money available for up-front investment, same-day microscopy analysis of sputum samples and starting patients on treatment is the most cost-effective strategy for reducing the rate of new infections. Given the wide variation in conditions within even small geographical areas, this more flexible approach should lead to the more efficient use of resources and may, ultimately, help to reduce the spread of tuberculosis.