Table 1.
Characteristics of the included studies grouped by used causal inference method.
| IPTW (N = 42) No (%) | Parametric G formula (N = 5) No (%) | TMLE (N = 1) No (%) | RL (N = 31) No (%) | All (N = 79) No (%) | |
|---|---|---|---|---|---|
| Exposure of interest (categorized) | |||||
| Vasopressors & intra-venous fluids | 0 (0) | 0 (0) | 0 (0) | 10 (32) | 10 (13) |
| Anti-inflammatory drugs | 7 (17) | 0 (0) | 0 (0) | 2 (6) | 9 (11) |
| Hospital-acquired complications | 8 (19) | 0 (0) | 0 (0) | 0 (0) | 8 (10) |
| Sedatives & analgesics | 1 (2) | 0 (0) | 0 (0) | 6 (19) | 7 (9) |
| Antimicrobials | 4 (10) | 0 (0) | 0 (0) | 0 (0) | 4 (5) |
| Mechanical ventilation | 1 (2) | 2 (40) | 0 (0) | 3 (10) | 6 (8) |
| Anticoagulants | 1 (2) | 0 (0) | 0 (0) | 2 (6) | 3 (4) |
| Diuretics | 3 (7) | 0 (0) | 0 (0) | 0 (0) | 3 (4) |
| Renal replacement therapy | 3 (7) | 0 (0) | 0 (0) | 0 (0) | 3 (4) |
| Sodium bicarbonate | 3 (7) | 0 (0) | 0 (0) | 0 (0) | 3 (4) |
| Blood transfusion | 2 (5) | 0 (0) | 0 (0) | 1 (3) | 3 (4) |
| Other | 9 (21) | 3 (60) | 1 (100) | 7 (23) | 20 (25) |
| Primary outcome (categorized) | |||||
| Mortality | 32 (76) | 2 (40) | 1 (100) | 11 (35) | 46 (58) |
| Combined | 0 (0) | 0 (0) | 0 (0) | 13 (42) | 13 (16) |
| Maintenance of clinical target value | 0 (0) | 0 (0) | 0 (0) | 6 (19) | 6 (8) |
| Hospital-acquired complications | 1 (2) | 2 (40) | 0 (0) | 0 (0) | 3 (4) |
| Need for mechanical ventilation | 2 (5) | 0 (0) | 0 (0) | 0 (0) | 2 (3) |
| Other | 7 (17) | 1 (20) | 0 (0) | 1 (3) | 9 (11) |
| Number of included ICUsa | |||||
| 1 | 16 (38) | 2 (40) | 0 (0) | 20 (77) | 38 (51) |
| 2–4 | 5 (12) | 1 (20) | 1 (100) | 4 (15) | 11 (15) |
| 5–10 | 3 (7) | 1 (20) | 0 (0) | 0 (0) | 4 (5) |
| 11–20 | 4 (10) | 1 (20) | 0 (0) | 0 (0) | 5 (7) |
| 21–100 | 5 (12) | 0 (0) | 0 (0) | 0 (0) | 5 (7) |
| >100 | 9 (21) | 0 (0) | 0 (0) | 2 (8) | 11 (15) |
| Utilized open source databases | |||||
| MIMIC-II | 0 (0) | 0 (0) | 0 (0) | 1 (3) | 1 (1) |
| MIMIC-III | 6 (14) | 0 (0) | 0 (0) | 18 (58) | 24 (30) |
| MIMIC-IV | 3 (7) | 1 (20) | 0 (0) | 3 (10) | 7 (9) |
| eICU | 2 (5) | 0 (0) | 0 (0) | 2 (6) | 4 (5) |
| AmsterdamUMCdb | 0 (0) | 1 (20) | 0 (0) | 2 (6) | 3 (4) |
| Study size (n patients)a | |||||
| 0–100 | 2 (5) | 0 (0) | 0 (0) | 0 (0) | 2 (3) |
| 100–500 | 7 (17) | 1 (20) | 0 (0) | 0 (0) | 8 (11) |
| 501–1000 | 9 (21) | 0 (0) | 1 (100) | 1 (4) | 11 (15) |
| 1001–5000 | 16 (38) | 3 (60) | 0 (0) | 6 (23) | 25 (34) |
| >5000 | 8 (19) | 1 (20) | 0 (0) | 19 (73) | 28 (38) |
| Type of exposure regime | |||||
| Static | 36 (86) | 3 (60) | 1 (100) | 0 (0) | 40 (51) |
| Dynamic | 6 (14) | 2 (40) | 0 (0) | 31 (100) | 39 (49) |
IPTW inverse probability of treatment weighting, TMLE targeted minimum loss-based estimation, RL reinforcement learning.
aFor the number of included ICUs and study size, the studies that used simulated patient data (n = 5) are not taken into account and therefore, the number of RL and all studies add up to 26 and 74, respectively.