Re: Clinicians’ gut feeling about serious infections in children: observational study
In the current study Van den Bruel, et al. reanalyse the dataset collected in 2004 which formed the basis of a previous publication.1 In that study they used the classification and regression tree (CART) analysis technique to construct a decision tree and aid the primary care clinician in identifying children with serious infections. They determined that a clinician feeling that ‘something is wrong’ was the most prominent predictor of serious infection. This item was therefore at the root node in the decision tree. In the current paper this item is relabelled “gut feeling”. The current logistic regression analysis seeks to identify consultation related variables that are associated with a gut feeling. It uses variables that have been included in the previous analysis and either found to be independent predictors of serious infection or discarded from the decision tree. The post hoc nature of the current paper mandates a cautious interpretation of its findings.
Furthermore we find data reported from the two studies to be inconsistent. The current study excludes 91 of the 3981 patients reported previously. The text suggests that these may be patients recruited from the emergency department, but excluding ED recruits should result in 136 exclusions. It is not clear why any patients have been excluded.
This reduction in study population to 3890 results in the inclusion of only 21 of the 31 children with serious infection reported in the 2007 study. There are 8 fewer children with a confirmed diagnosis of sepsis/meningitis and 6 fewer with pneumonia, but paradoxically one more child with pyelonephritis. Additionally, a case of bacterial lymphangitis is considered a serious infection. This is not included as a serious infection in the published study description.
These data do not sufficiently support the clinical adoption of “gut feeling” as a determinant of the decision to make a referral for hospital admission. The logistic regression analysis reveals seizure to be the clinical variable most highly associated with gut feeling. In a population of children who did not have “neurological illness” or an “exacerbation of a chronic condition”, patients presenting with seizure in the course of “an acute illness for a maximum of 5 days” duration should be suspected of having serious illness through deduction rather than gut feeling. The clinician who has a ‘gut feeling’ of serious illness should also be aware of all the symptoms and signs of potential serious illness elicited in that child. Thus ‘gut feeling’ is not independent of this awareness. The conflation in the study design of “gut feeling” suspicion with clinical assessment suspicion of serious infection limits the ability of the study to detect the determinants of gut feeling.
In assessing severity of children’s illness, we prefer to use objective severity scoring systems2 that can be used by all (medical, paramedical and nursing) members of the clinical team. There are a number of such scoring systems, primarily developed to alert the team to a deteriorating child, under observation in a clinical setting. However further developments using combinations of vital signs can give ‘one off’ assessments with 95% sensitivity for ‘serious or intermediate’ infection.3 We would advocate the modification for primary care of such hospital based scoring systems as a useful adjunct to standard clinical assessment. They would have the advantage over gut feeling that they are auditable and could therefore be part of a continuous process of quality improvement.
Sieber CR, Upton CJ, Armon K, Briars GL
1. Van den Bruel A, Aertgeerts B, Bruyninckx R, Aerts M, Buntinx F. Signs and symptoms for diagnosis of serious infections in children: a prospective study in primary care. Br. J. Gen. Pract. Jul 2007;57(540):538-546.
2. Edwards ED, Powell CV, Mason BW, Oliver A. Prospective cohort study to test the predictability of the Cardiff and Vale paediatric early warning system. Arch. Dis. Child. Aug 2009;94(8):602-606.
3. Thompson M, Coad N, Harnden A, Mayon-White R, Perera R, Mant D. How well do vital signs identify children with serious infections in paediatric emergency care? Arch. Dis. Child. Nov 2009;94(11):888-893.
Competing interests: No competing interests