SERIOUS INFECTION IN ACUTELY ILL CHILDREN IN PRIMARY CARE: Validating clinical prediction rules and added value of vital signs and point-of-care tests

Background: Acute infection is the most common presentation of children to ambulatory care. In contrast, serious infections are rare and often present at an early stage. To avoid complications or death, early recognition and adequate referral are essential. In a recent large study children were included prospectively to construct a symptom-based decision tree with a sensitivity and negative predictive value of nearly 100%. To reduce the number of false positives, point-of-care (POC) tests might be useful, providing an immediate result at the bedside. The most probable candidate is C-reactive protein. Every clinician should reassure anxious parents of children with self-limiting illnesses. The improvement of diagnostic algorithms, the addition of technological devices and the sensible use of safety netting procedures could improve prognosis of seriously ill children.

Methods: First, we externally validated clinical prediction rules, identified by a systematic review, in 7 urgent-access datasets, as well as comparing these results to recent findings in other studies. After zooming in on the diagnostic value of the clinical prediction rules based on vital signs with potential to differentiate serious infections from the majority of self-limiting illnesses in an inpatient paediatric setting in the UK, we focused our analyses on the temporal & geographic validation of the decision tree based on signs and symptoms in a new but similar population in Flanders. We examined the analytical accuracy and user-friendliness of a POC test after careful selection of a device that meets all our requirements. Finally we explored the added value of the selected POC CRP test in a prospective diagnostic accuracy study in three different ambulatory care settings: general practice, outpatient paediatric clinic, and the emergency department.

Results: In low to intermediate prevalence settings the 4-step decision tree and evidence-based guidelines had high sensitivity, providing promising rule-out value for serious infections in 7 datasets. The paediatrician’s overall illness assessment was the most useful feature to rule in sepsis or meningitis in a study of hospitalized children. Temporal validation of the 4-step decision tree indicated that this practical tool for diagnostic triage of acutely ill children in primary care is valid and ready to be implemented in routine care, if appropriate safety netting or additional testing is applied. The selected POC CRP test was accurate in children and should be considered reliable and user-friendly. Adding point-of-care CRP test results to the 4-step decision tree helped identifying serious infections in the GP setting and can potentially reduce the number of investigations and admissions in children with non-serious infections. I propose a new decision tree to be used in specialist settings as a triage instrument to safely rule out serious infections.

Discussion: The incidence of serious infections has declined over the past few years, amongst other reasons, due to vaccination strategies and improvements in neonatal care. Before a clinical prediction rule can be implemented in routine care, it has to go through several stages of development and testing. A single test will never reach perfect sensitivity and specificity in real life. To tackle the ever-present clinical uncertainty, physicians often put a safety net in place, informing parents when to re-contact and which alarm signs are relevant to monitor. I put an emphasis on these issues and offer a perspective for future developments in the field of point-of-care testing in serious infections in paediatric primary care.