Development and validation of a risk score to predict mortality during TB treatment in patients with TB-diabetes comorbidity

Duc T Nguyen, Edward A Graviss

Research output: Contribution to journalArticlepeer-review

5 Scopus citations


Background: Making an accurate prognosis for mortality during tuberculosis (TB) treatment in TB-diabetes (TB-DM) comorbid patients remains a challenge for health professionals, especially in low TB prevalent populations, due to the lack of a standardized prognostic model. Methods: Using de-identified data from TB-DM patients from Texas, who received TB treatment had a treatment outcome of completed treatment or died before completion, reported to the National TB Surveillance System from January 2010-December 2016, we developed and internally validated a mortality scoring system, based on the regression coefficients. Results: Of 1227 included TB-DM patients, 112 (9.1%) died during treatment. The score used nine characteristics routinely collected by most TB programs. Patients were divided into three groups based on their score: low-risk (< 12 points), medium-risk (12-21 points) and high-risk (≥22 points). The model had good performance (with an area under the receiver operating characteristic (ROC) curve of 0.83 in development and 0.82 in validation), and good calibration. A practical mobile calculator app was also created ( Conclusion: Using demographic and clinical characteristics which are available from most TB programs at the patient's initial visits, our simple scoring system had good performance and may be a practical clinical tool for TB health professionals in identifying TB-DM comorbid patients with a high mortality risk.

Original languageEnglish (US)
Article number10
Pages (from-to)10
JournalBMC Infectious Diseases
Issue number1
StatePublished - Jan 5 2019


  • Diabetes
  • Mortality
  • Risk score
  • TB-DM
  • TB-diabetes
  • Tuberculosis

ASJC Scopus subject areas

  • Infectious Diseases


Dive into the research topics of 'Development and validation of a risk score to predict mortality during TB treatment in patients with TB-diabetes comorbidity'. Together they form a unique fingerprint.

Cite this