Refining clinical algorithms for a neonatal digital platform for low-income countries: a modified Delphi technique

Abstract

Objectives To determine whether a panel of neonatal experts could address evidence gaps in neonatal guidelines by reaching a consensus on four clinical decision algorithms for a neonatal digital platform (NeoTree). Design Two-round, modified Delphi technique. Setting and participants Participants were neonatal experts from high-income and low-income countries (LICs). Methods This was a consensus-generating study. In round one, experts rated items for four clinical algorithms (neonatal sepsis, hypoxic ischaemic encephalopathy, respiratory distress of the newborn, hypothermia) and justified their responses. Items meeting consensus (>80% agreement) were included. Items not meeting consensus were either excluded, included following revisions or included if they contained core elements of evidence-based guidelines. In round two, experts rated items from round one that did not reach consensus. Results Fourteen experts participated in round one, ten in round two. Nine were from high-income countries, five from LICs. Experts included physicians and nurse practitioners with an average neonatal experience of 20 years, 12 in LICs. After two rounds, a consensus was reached on 43 of 84 items (52%). Experts consistently stated that items must be in line with local and WHO guidelines (irrespective of the level of supporting evidence or expert opinion). As a result, the final algorithms included 53 items (62%). Conclusion Four algorithms in a neonatal digital platform were reviewed and refined by consensus expert opinion. Revisions to the NeoTree application were made in response to these findings and will be clinically validated in an imminent study.

The Promise of Clinical Decision Support Systems Targetting Low-Resource Settings

“…This can be identified via a prospective trial, something that would be relatively feasible in this scenario given the minimal harm imposed on the patient by the decision of the algorithm. Most recently, Evans et al [116] convened 22 neonatal experts in order to gauge their agreement on algorithms used for the diagnosis of various neonatal conditions in low-resource settings. These conditions included neonatal sepsis, hypoxic ischaemic encephalopathy, respiratory distress, and hypothermia.…”