Table 2.
Heterogeneity assessments and sensitivity analysis
| Heterogeneity assessments | Sensitivity analysis | ||||
|---|---|---|---|---|---|
| Characteristic assessed | Findings | p valueƚ | Characteristic assessed | Findings | |
| CRP (all cut-offs) | Predetermined cut-off | Studies using a predetermined cut-off had slightly lower accuracy | 0.003 | Gestational age range | Excluding studies that included GA < 24 weeks resulted in a slightly lower accuracy |
| Interval between sampling and delivery | Studies > 72 h had lower accuracy | < 0.001 | Applicability concerns in patient selection domain | Excluding studies with high concerns did not change the SROC curve | |
| Risk of bias in patient selection domain | Studies with low risk score had lower accuracy | < 0.001 | Publication year after 2000 | Excluding studies published before year 2000 yielded a slightly lower accuracy | |
| Assay type | Studies with CRP assays after standardisation (year 1993) had lower accuracy | < 0.001 | |||
| PCT | Predetermined cut-off | Studies using a predetermined cut-off had slightly lower accuracy | 0.026 | Gestational age range | Excluding studies that included GA < 24 had no effect |
| Interval between sampling and delivery | No difference | 0.178 | Applicability concerns in patient selection domain | Excluding studies with high concerns resulted in a much lower accuracy and a change in shape of the curve | |
| Risk of bias in patient selection | Studies with low risk score had lower accuracy | < 0.001 | Publication year after 2000 | Not assessed as all studies were published after year 2000 | |
| IL6 | Predetermined cut-off | Not assessed as 1 subgroup had < 2 studies | Gestational age range | Excluding studies that included GA < 24 weeks resulted in a slightly higher accuracy and change in shape of SROC curve | |
| Interval between sampling and delivery | Studies ≤ 72 h had lower accuracy | < 0.001 | Applicability concerns in patient selection domain | Not assessed as all studies had low applicability concerns | |
| Risk of bias in patient selection | Not assessed as 1 subgroup had < 2 studies | Publication year after 2000 | Not assessed as all studies were published after year 2000 | ||