Impact of personalised risk predictions on preventive treatment initiation in tuberculosis

Background Preventive treatment for clinically latent Mycobacterium tuberculosis (Mtb) infection has low uptake, partly due to poor individual risk stratification. We evaluated whether personalised tuberculosis (TB) risk prediction influenced treatment initiation. Methods We undertook a retrospective service evaluation at a single TB clinic in London May 2022–September 2024, in which the PERISKOPE-TB risk prediction model was presented within consultations with patients eligible for preventative treatment for clinically latent Mtb infection. We then examined the association between predicted risk and treatment acceptance using logistic regression, with restricted cubic splines to account for a non-linear relationship. Results are presented as an odds ratio (OR) with 95% confidence intervals (CI’s). Results 283 patients were included in the evaluation, where 62% initiated treatment. Higher predicted TB risk was associated with increased treatment initiation; each 1% increase in risk more than doubled the odds of acceptance (OR 2.12, 95% CI 1.59–2.89). Conclusions Personalised risk prediction may improve targeting and uptake of preventive TB treatment in those that need it the most.