Decision-making tool may help doctors cut unnecessary antibiotic prescribing

The authors estimate that if antibiotic prescribing in this low risk group was halved, and even if it increased to 90% in high risk patients, the new tool could reduce antibiotic prescribing to children with RTI and coughs by 10% overall, similar to other interventions used to combat antibiotic resistance.

The proposed tool called STARWAVe uses seven predictors of future hospitalization that can be easily identified by doctors and nurses during a patient visit -- short illness (less than 3 days), high temperature (?37.8°C on examination or parent reported severe fever in the previous 24 hours), aged under 2 years, respiratory distress, wheeze, asthma, and moderate/severe vomiting in the previous 24 hours. Children presenting with no more than one of these items are deemed at very low risk of future complications. The authors say that the rule now needs externally validating in a randomised trial, but could be a useful tool to improve the targeting of antibiotics to reduce the growing threat of antibiotic resistance.

Respiratory infections with cough is the most common reason people go to the doctor and the most frequent reason given for primary care antibiotic prescribing in children. Yet it is challenging for GPs and primary care nurses to easily identify serious respiratory infections, and up to a third of antibiotics prescribed in primary care are considered unnecessary.

"Excessive antibiotic use has contributed to the development of resistance to these drugs," explains lead author Professor Alastair Hay from the University of Bristol, Bristol, UK. "The aim of our study was to develop a simple, usable prediction tool based on symptoms and signs to help GPs and nurses identify children presenting in primary care at the lowest and highest risk of future complications and hospitalization, so that antibiotics can be targeted accordingly."

To create the tool, Hay and colleagues analysed data collected between July 2011 and May 2013 from almost 8400 children aged between 3 months and 16 years with acute (less than 28 days) cough and respiratory tract infection symptoms (eg, fever) who were seen at 247 GP practices across England. They used modelling to determine which of the 50 demographic characteristics, parent-reported symptoms and physical examination signs measured might be most useful and accurate in distinguishing good from poor prognosis illnesses, defined as those resulting in hospitalization for respiratory infection in the month following a visit to primary care.

Modelling showed that seven characteristics were independently linked with hospitalization -- short (<3 days) illness; temperature; age (<2 years); recession (signs of respiratory distress); wheeze; asthma; and vomiting (mnemonic "STARWAVe").

Using these findings, the authors then developed a seven-item scoring system for a child's risk of future hospitalization. For example, a child showing 0-1 of these characteristics would be at very low risk of hospitalization (0.3% risk; 67% of children in the study); a child with 2-3 of these characteristics would be at normal risk, similar to the general population (1.5% risk; 30% of children in the study); whilst a child showing 4 or more would be a high risk candidate for future hospitalization (11.8% risk; 3% of children in the study).

According to the authors, a 'no antibiotic' prescribing strategy would be appropriate for low risk children; whilst a 'no antibiotic or delayed antibiotic' treatment strategy would be best for normal risk children -- as recommended by NICE; and children deemed at high risk of hospitalization should be closely monitored for signs of deterioration and followed-up within 24 hours.

The accuracy of the rule was measured by a figure called the 'area under the receiver operating characteristic curve', or AUROC. An AUROC of 0.5 would mean the rule is about as good a predictor as flipping a coin. An AUROC of 1.0 is perfect. The new STARWAVe rule gave an AUROC of 0.81, which indicates it should predict the risk of hospitalization with high accuracy.

The authors note that the results are likely to be applicable to primary care systems similar to those in the UK, but as only 78 children were hospitalised during the study, further research is needed to externally validate the tool.

According to Professor Hay, "This is the first study of its kind, based on a large representative sample of children who visit the doctor with respiratory illness. We hope that our proposed clinical tool might eventually enable doctors to quickly and easily identify their lowest and highest risk patients, although more research will be needed to determine just how effective it is in clinical practice. The rule should supplement not replace clinical judgement, and doctors and nurses should still advise parents about the symptoms and signs they should look out for, and when to seek medical help."

In a linked Comment, Professor David Price, Chair of Primary Care Respiratory Medicine at the University of Aberdeen, Aberdeen, UK and colleagues discuss the need to test the tool in whole study populations and not just those recruiting and consenting to enter a study. They write, "Notwithstanding the inclusion of patients prescribed an antibiotic and the absence of an independent validation cohort, STARWAVe promises to achieve better targeting of antibiotics in primary care. There are few efficacious interventions for respiratory tract infection available to primary care clinicians beyond offering reassurance and self-management advice, so the modest benefit offered by antibiotics can persuade general practitioners to prescribe them. STARWAVe offers primary care clinicians an evidence-based practical tool to help guide antibiotic prescribing decisions and, through shared decision-making, has the potential to reduce prescribing based on prognostic uncertainty or on nonmedical grounds."