Re: Bad medicine: NICE’s traffic light system for febrile children
We note the views expressed by Spence(1) (BMJ 2014;348:g2056 doi: 10.1136/bmj.g2056) on the “traffic light” system employed within the Nice Clinical Guideline on Feverish Illness in Children (CG160).(2) He criticises the system as one that both overcomplicates and oversimplifies, leading to poor clinical practice. We disagree with this conclusion.
Fever is a common phenomenon in children and while most have a self-limiting disorder a few have a serious and potentially life-threatening bacterial infection. This poses a challenge for experienced and inexperienced healthcare professionals alike. Carefully considered guidance is clearly needed and should be interpreted using clinical judgement.
The purpose of the traffic light system is to assist clinicians in identifying those who may have a serious illness. This system categorises various symptoms and signs as ‘red’ (high risk), ‘amber’ (intermediate risk), or ‘green’ (low risk). Thus in remote assessment settings (for example telephone consultation) children with any ‘red’ feature should be urgently assessed in a face-to-face setting within two hours (sooner if clinically appropriate), and those with any 'amber' feature should have a face-to-face assessment with an urgency determined by clinical judgement. Those with ‘green’ features only can be cared for at home with appropriate advice for parents and carers, including when to seek further attention from the healthcare services.
Spence believes that the traffic light system is based on the “opinion and interpretation of a small panel of experts”. This is not correct. He disagrees with features of the system based on his own clinical experience. The evidence for the traffic light system was considered at length by Guideline Development Groups (GDG) both at the time of the original 2007 guideline (CG047) (3) and for the recent 2013 update(CG160) (2). These groups had a diverse membership including lay experts (2), nurses from a variety of clinical settings including ambulatory care(4), GPs (2), an expert from NHS Direct, specialist paediatric registrars (2), consultant paediatricians (4), and consultants in paediatric infectious disease and in emergency medicine. Moreover for both the original 2007 guideline and its 2013 update proposed recommendations in draft underwent an extensive process of public consultation. This is an integral part of the NICE Clinical Guideline development process.
Spence expressed concern that the cited research is from selected hospital populations in which the prevalence of bacterial infection is higher than would be found, for example, in low risk community populations. He correctly points out that the predictive value is dependent on the population prevalence of the condition and could therefore be misleading if incorrectly interpreted – the value might be very different in a primary care setting. In fact the clinical settings of the studies included in the evidence reviews varied, including GP surgeries, emergency departments and paediatric wards of general hospitals, emergency departments of paediatric hospitals, tertiary care paediatric units and tertiary care medical centres. The GDG took account of these settings when considering their interpretation of the evidence. Although it looked at a range of accuracy indices including positive and negative predictive value, sensitivity and specificity, it relied primarily on the positive and negative likelihood ratios in drawing their conclusions, and these ratios are not dependent on the population prevalence of the condition.
In the original 2007 guideline evidence based reviews were undertaken to identify the relationship of individual symptoms and signs with serious illness. Reviews were also undertaken to identify evidence on existing scoring systems to identify serious illness, and one (the Yale Observation System) contributed to the development of the traffic light system. For the 2013 update the symptoms and signs in the 2007 guideline were reassessed to ensure that the evidence still supported their inclusion and to identify any new inclusions if appropriate. The update focused on the diagnostic accuracy of these symptoms and signs, emphasising the importance of likelihood ratios in decision making. The update with some minor alterations provided further support for the original traffic light system. An updated review of the Yale Observation Scale was also undertaken to ensure its continued validity.
Spence cites a recent report examining the accuracy of the 2007 guideline’s traffic light system and states that it suggested a risk of false positives and false negatives (4). This study was undertaken in a paediatric emergency department setting which specifically examined the accuracy of the traffic light system when employed in isolation (without taking account of other aspects of the guideline) in identifying urinary tract infection, pneumonia and bacteraemia. In this limited context the overall sensitivity was 85.8% (95% CI 83.6 to 87.7%) and the specificity was 28.5% (95% CI 27.8% to 29.3%). We have commented previously on this useful report and concluded that when the traffic light system is used in conjunction with the guideline as a whole it supports its value in identifying serious illness. (5) The need for thorough and on-going evaluation of the traffic light system is clear and indeed the guideline includes a research recommendation that a large prospective UK-wide study is needed addressing all of the symptoms and signs covered in the guideline and including a variety of settings and locations including primary care.
The traffic light system was created to encourage healthcare professionals to consider symptoms and signs in their totality and individual symptoms clearly need to be considered in the context of the overall clinical picture. Combinations of manifestations may provide reasons for particular concern. We agree with Spence’s view that clinical assessment is a complex process based on experience, observation, continuity, and clinical intuition. It should be noted that both the evidence and the opinion of the guideline development groups coincided in the recommendation that when a child “appears ill to a healthcare professional” this was to be considered a “red” feature. We believe that the Clinical Guideline with its traffic light system will assist clinicians in the complex challenge of making safe and appropriate decisions.
1. Spence D. Bad medicine: NICE’s traffic light system for febrile children. BMJ 2014;348:g2056 doi: 10.1136/bmj.g2056
2. National Institute for Health and Clinical Excellence. Feverish illness: assessment and initial management in children younger than 5 years. London: NICE; 2013. www.nice.org.uk/CG160
3. National Institute for Health and Clinical Excellence. Feverish illness: assessment and initial management in children younger than 5 years. London: NICE; 2007. www.nice.org.uk/CG047
4. De S, Williams GJ, Hayen A, Macaskill P, McCaskill M, Isaacs D, Craig JC. Accuracy of the “traffic light” clinical decision rule for serious bacterial infections in young children with fever: a retrospective cohort study. BMJ 2013;346:f866.
5. Chard J, Richardson M, Murphy MS. Rapid response: Accuracy of the “traffic light” clinical decision rule for serious bacterial infections in young children with fever: a retrospective cohort study. BMJ 2013;346:f866. http://www.bmj.com/content/346/bmj.f866/rr/637155.
Competing interests: Stephen Murphy is Co-director in Child Health at the National Collaborating Centre for Women’s and Children’s Health, which produced the NICE guideline on feverish illness in children (2013). Mark Baker is director, NICE Centre for Clinical Practice.