Derivation of a tuberculosis screening rule for sub-Saharan African prisons

Int J Tuberc Lung Dis. 2014 Jul;18(7):774-80. doi: 10.5588/ijtld.13.0732.

Abstract

Setting: Lusaka Central Prison, Zambia.

Objective: To derive screening rules for tuberculosis (TB) using data collected during a prison-wide TB and human immunodeficiency virus (HIV) screening program.

Design: We derived rules with two methodologies: logistic regression and classification and regression trees (C&RT). We evaluated the performance of the derived rules as well as existing World Health Organization (WHO) screening recommendations in our cohort of inmates, as measured by sensitivity, specificity, and positive and negative predictive values.

Results: The C&RT-derived rule recommended diagnostic testing of all inmates who were underweight (defined as body mass index [BMI] < 18.5 kg/m(2)] or HIV-infected; the C&RT-derived rule had 60% sensitivity and 71% specificity. The logistic regression-derived rule recommended diagnostic testing of inmates who were underweight, HIV-infected or had chest pain; the logistic regression-derived rule had 74% sensitivity and 57% specificity. Two of the WHO recommendations had sensitivities that were similar to our logistic regression rule but had poorer specificities, resulting in a greater testing burden.

Conclusion: Low BMI and HIV infection were the most robust predictors of TB in our inmates; chest pain was additionally retained in one model. BMI and HIV should be further evaluated as the basis for TB screening rules for inmates, with modification as needed to improve the performance of the rules.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Body Mass Index
  • Chest Pain / epidemiology
  • Chest Pain / etiology
  • Female
  • HIV Infections / diagnosis
  • HIV Infections / epidemiology
  • Humans
  • Logistic Models
  • Male
  • Mass Screening / methods*
  • Predictive Value of Tests
  • Prisoners / statistics & numerical data
  • Prisons*
  • Sensitivity and Specificity
  • Tuberculosis / diagnosis*
  • Tuberculosis / epidemiology
  • World Health Organization
  • Zambia / epidemiology