Table 3.
Characteristics of patient populations in the 33 included studies with prediction models for TB treatment outcomes
| Characteristics | Studies reporting characteristic, n (%) | Categories | N (%) or median (IQR) |
| Sample size | 33 (11) | – | 803 (291–4167) |
| Study duration, years | 32 (97) | – | 4 (2–7) |
| Study design | 33 (100) | Prospective cohort | 3 (9) |
| Retrospective cohort | 25 (76) | ||
| Nested case–control | 3 (9) | ||
| Non-nested case–control | 2 (6) | ||
| Data source | 33 (100) | Medical record | 6 (18) |
| National registry or surveillance system | 13 (39) | ||
| Local registry or surveillance system | 1 (3) | ||
| Regional registry or surveillance system | 2 (6) | ||
| Data collect form for study purposes | 11 (33) | ||
| Study region | 32 (97) | Africa | 8 (25) |
| Asia | 13 (41) | ||
| Europe | 6 (19) | ||
| North America | 4 (12) | ||
| South America | 0 (0) | ||
| Global | 1 (3) | ||
| High burden TB setting* | 31 (94) | All | 143 (42) |
| Some | 1 (3) | ||
| None | 17 (55) | ||
| Missing data | 18 (54) | Complete case analysis | 9 (50) |
| Missing indicator method | 4 (22) | ||
| Heckman’s method | 1 (6) | ||
| Simple imputation | 2 (12) | ||
| Sensitivity analysis with imputation | 1 (6) | ||
| Other | 1 (5) | ||
| Number of models developed | 33 (100) | 1 | 25 (76) |
| 2 | 4 (12) | ||
| 3 | 1 (3) | ||
| 4 | 2 (6) | ||
| 7 | 1 (3) | ||
| Reasons for multiple models developed | 8 (24) | Different outcomes | 1 (12) |
| Different predictors considered | 4 (50) | ||
| Different methods | 2 (25) | ||
| Different outcomes | 1 (12) | ||
| Different populations and outcomes | 1 (12) |
*Determined based on study location and WHO list of 30 countries with high-burden TB in the 2019 Global Tuberculosis Report (1).
TB, tuberculosis.