622 - Diagnostic Accuracy of Clinical Sign Algorithms to Predict Mortality among Young Infants aged 0-59 days: A Systematic Review

Poster Number: 622
Publication Number: 622.355

Background: Annually, 2.3 million newborns globally die from prematurity, birth asphyxia, and sepsis. Clinical sign algorithms can potentially identify infants at high-risk for mortality in both community and hospital settings. The diagnostic accuracy of these algorithms to predict mortality is critical to understand.

Objective: To systematically review the evidence on the diagnostic accuracy of infant clinical sign algorithms to predict mortality among young infants (0-59 days).

Design/Methods: We conducted a systematic review using Cochrane methods. Medline, Embase, CINAHL, Global Index Medicus and Cochrane CENTRAL Registry of Trials were searched in May 2023 using terms for infants (0-59 days), mortality, prediction and diagnostic accuracy. Two reviewers screened articles, extracted data and assessed risk of bias with QUAPAS, and used GRADE to assess certainty of evidence. We included studies that evaluated the diagnostic accuracy of clinical sign algorithms to predict mortality in young infants.

Results: Of 6,701 studies identified, 9 met inclusion criteria (total 107,199 infants) (Table 1). Algorithms included checklists (n=2), weighted scores (n=4), prediction models (n=2) or combined prediction model/weighted score (n=1). Most were intended for hospital/NICU settings (n=7), with 2 developed for the community settings (n=2). Algorithms were assessed in both upper/high income (n=6) as well as low/low-and middle income countries (n=13). A full description of the identified studies/algorithms are shown in (Table 1). The SEARCH algorithm required 2 out of 7 signs (2 studies, n=10491) and had a pooled sensitivity of 0.39 (95% CI 0.31,0.49) and specificity of 0.98 (95% CI 0.98,0.98) for predicting death in infants (0-28 days). Diagnostic accuracy parameters for other models are shown in Table 2 and were not pooled. The certainty of evidence was GRADED very low across studies for both sensitivity and specificity (Table 2)

Conclusion(s): Existing infant sign algorithms have varying accuracy in their ability to predict mortality among infants 0-59 days . While some algorithms demonstrate promising results, the overall certainty of evidence remains very low. In the high mortality contexts, more rigorous research is essential to develop, validate, and scale accurate methods to identify the highest risk young infants.
Table 1.jpeg

Table 2.jpeg

Figure 1.jpeg