Table 3.
Effect of ART, CD4 cell count, and HIV viral load on COVID-19 in-hospital mortality among people with HIV admitted to hospital with laboratory-confirmed SARS-CoV-2 infection, South Africa
| Case fatality ratio unimputed | Case fatality ratio (95% CI) imputed | Unadjusted OR (95% CI) imputed | p value | Adjusted OR (95% CI) imputed* | p value | Adjusted OR (95% CI) imputed* | p value | |
|---|---|---|---|---|---|---|---|---|
| ART status | ||||||||
| HIV negative | 30 697/137 986 (22·2%) | 23·1% (22·9–23·3) | 1 (ref) | .. | 1 (ref) | .. | 0·77 (0·72–0·82) | <0·0001 |
| HIV positive on ART | 2046/7484 (27·3%) | 24·6% (23·8–25·4) | 0·85 (0·80–0·90) | <0·0001 | 1·30 (1·22–1·39) | <0·0001 | 1 (ref) | .. |
| HIV positive not on ART | 192/594 (32·3%) | 28·1% (25·2–31·1) | 0·99 (0·86–1·15) | 0·98 | 1·89 (1·60–2·23) | <0·0001 | 1·45 (1·22–1·72) | <0·0001 |
| CD4 count | ||||||||
| HIV negative | 30 697/137 986 (22·2%) | 23·1% (22·9–23·3) | 1 (ref) | .. | 1 (ref) | .. | 1·06 (0·93–1·20) | 0·37 |
| HIV positive CD4 count, ≥200 cells per μL | 368/1690 (21·8%) | 19·7% (17·7–21·7) | 0·64 (0·57–0·73) | <0·0001 | 0·95 (0·83–1·08) | 0·37 | 1 (ref) | .. |
| HIV positive CD4 count, <200 cells per μL | 380/1080 (35·2%) | 32·2% (29·9–34·5) | 1·23 (1·09–1·38) | 0·0023 | 2·19 (1·92–2·49) | <0·0001 | 2·31 (1·82–2·93) | <0·0001 |
| Viral load | ||||||||
| HIV negative | 30 697/137 986 (22·2%) | 23·1% (22·9–23·3) | 1 (ref) | .. | 1 (ref) | .. | 0·83 (0·76–0·90) | 0·0002 |
| HIV positive viral load, <1000 HIV RNA copies per mL | 316/1273 (24·8%) | 24·7% (23·0–26·4) | 0·86 (0·78–0·94) | 0·0029 | 1·21 (1·11–1·32) | 0·0002 | 1 (ref) | .. |
| HIV positive viral load, ≥1000 HIV RNA copies per mL | 128/443 (28·9%) | 25·2% (21·0–29·4) | 0·85 (0·69–1·05) | 0·13 | 1·88 (1·53–2·31) | <0·0001 | 1·55 (1·20–2·01) | 0·0029 |
Model adjusted for age, sex, race or ethnicity, health sector, province, month of admission, non-communicable comorbidities, and past or current tuberculosis. ART=antiretroviral therapy. OR=odds ratio.
*
Output from the same model but with different reference categories to assess the effect of the predictors compared with HIV-uninfected individuals (model 1) or individuals on ART, with high CD4 count, or with low viral load (model 2).