Early identification of individuals at risk for loss to follow-up of tuberculosis treatment: A generalised hierarchical analysis
Shirley Verônica Melo Almeida Lima, Karina Conceição Gomes Machado de Araújo, Marco Antônio Prado Nunes, Carla Nunes
Heliyon · 2021-04
Abstract
BACKGROUND: We characterise the loss to follow-up (locally termed abandoned) of tuberculosis treatment with individual and ecological health determinants and to identify the predictive capacity of these risk factors. METHODS: A cohort study with individual and ecological characterisation of patients diagnosed with tuberculosis in Sergipe/Brazil from 2015 to 2018 with either loss to follow-up or completion of treatment as a therapeutic outcome was performed. The examined variables were based on the social determinants of health with descriptive analysis, binary logistic regression, a generalised hierarchical model and graphical presentation using a nomogram. RESULTS: The loss to follow-up accounted for 18.21% of the 2,449 studied cases. The characteristics revealed that the highest abandonment percentages were people who: were male (20.0%), had black skin colour (20.3%), were aged 20-39 years (21.8%), had 4-7 years of schooling (23.6%), re-entered treatment after abandonment (36.5%), used alcohol (31.0%), used drugs (39.3%), were smokers (26.5%) and were homeless (55.4%). The ecological characteristics showed that individuals living in municipalities with a high human development index (HDI; odds ratio [OR]: 1.91) and high-income inequality (OR: 1.81) had a greater chance of not finishing the treatment. Most of these variables were identified as predictors in the generalised hierarchical model; the receiver operating characteristic curve (ROC) curve had 0.771 precision and 84.0% accuracy. CONCLUSION: The group of identified characteristics influenced the loss to follow-up of tuberculosis treatment. This data provides evidence for the early identification of individuals who are at greater risk of abandoning tuberculosis treatment.
MeSH terms
- Tuberculosis
- Medicine
- Logistic regression
- Receiver operating characteristic
- Nomogram
- Demography
- Abandonment (legal)
- Odds ratio
- Multilevel model
- Gerontology