Superspreaders and TB

For each study, we fitted single-predictor and multi-variable models to cough duration outcome.
Some of the individual predictors were statistically significant, but association with outcome was weak (low R²).
Multi-variate models fitted in a predictive framework (cross-validation) also did not perform well, no predictive power.
Single exception was U-Steps study, with variable “fraction visits to non TB healthcare providers” showing some correlation.

Discussion

There seems clear evidence that for TB - as for other diseases - there is heterogeneity in transmission potential.
Heterogeneity is likely in all 3 components: contact rate/infectiousness level/infectiousness duration.
Unclear how important each component is and if/how they co-vary.
Unclear if one can practically identify high-transmitters/superspreaders.

Melsew 2018, 2019, our study and many similar studies often find correlations between some variable and (proxies of) transmission.
Those findings are not tested on independent data, thus hard to know how useful such variables are.
To be useful, one would need to find robust and actionable variables.

Co-investigators: L Martinez, JN Sekandi, SE Bellan, L Zhu, C Chen, Q Liu, S Donkor, J Sutherland, PC Hill, RH Gilman, L Grandjean, CC Whalen
Funding: NIH A1093856, NIH AI093856-05W1, The Wellcome Trust 201470/Z/16/Z, NIH K01AI125830, NIH AI-45244-95383.
Contact: www.andreashandel.com, Email: ahandel@uga.edu, Twitter: @andreashandel.