Imagine this – a six-week-old infant with a fever of 38.3⁰C presents to your emergency department. They look well and, apart from fever, have no other symptoms. There are no sick contacts at home. Although they look well, you don’t want to miss a serious diagnosis, so what do you do?
You’ve so many questions. Which of the many febrile infant clinical decision tools should you
use? How far should you investigate? Should you send urine? Bloods (and if so, which ones)? Lumbar puncture? Should you give antibiotics or observe? Should you admit or discharge?
Managing febrile infants is challenging. Somewhere between one and four in every 100 febrile infants will have an invasive bacterial infection (IBI) with positive blood or cerebrospinal fluid cultures. But how do we identify which of the many febrile infants we see have an IBI and which don’t?
Internationally, guidelines vary both in their recommendations for investigations and in their recommended clinical management. Over the last decade, there has been a shift towards incorporating age, clinical findings, and blood test biomarkers into clinical decision-making tools.
The 2021 AAP guideline uses absolute neutrophil count, CRP and procalcitonin (PCT) to help clinicians identify which febrile infants are low-risk and do not need a lumbar puncture or intravenous antibiotics. The AAP guideline is not alone in using PCT – several other, well-validated tools use PCT, including StepByStep and PECARN. But what if your hospital lab doesn’t run PCT? Can you still use these rules? PCT is not widely available in the UK and Ireland, so this is not an uncommon dilemma.
In the UK, there are two national febrile infant guidelines: NICE NG143 (Fever Under Five Years) and the British Society for Antimicrobial Chemotherapy (BSAC) guideline. Add the AAP and Aronson guidelines, and several clinical decision aids (CDAs) are available.
Enter the FIDO study. FIDO was designed to assess the management of young febrile infants presenting to emergency departments in the UK and Ireland, report risk factors for IBI, and apply CDAs that do not exclusively rely on PCT testing.
Who were the patients?
This prospective multicenter cohort study was run in 35 sites in the PERUKI network over 13 months between July 2022 and August 2023. Febrile infants were eligible if they were:
- aged 90 days or younger
- with a fever ≥38°C in the ED or Assessment Unit
- or with a history of fever ≥38°C recorded by anyone via any thermometer in the 24 hours before presentation.
1821 infants were recruited, with a median age of 46 days (IQR, 30 to 64).
260 (14%) had prior medical conditions, and more than half (1058, 58%) looked unwell.
What was the intervention?
This was an observational, rather than interventional, study. Data was collected on investigations sent.
- 1483 (81%) had a urine test
- 1626 (89%) had blood tests
- 937 (52%) had a lumbar puncture
- 1395 (77%) had a respiratory viral test.
An additional 1 ml of blood was taken for PCT from some infants.
What were the outcomes measured?
The primary outcome was the performance accuracy of CDAs in identifying which febrile infants had an IBI.
CDA performance was assessed using sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV).
Secondary outcome measures included:
- The organism grown from blood or CSF
- Clinical predictors that could identify which infants are at risk of an IBI are age, comorbidities, time to presentation from fever onset, unwell contacts, clinical appearance, vital signs, symptoms on presentation such as cough, vomiting, diarrhoea, and clinical exam findings.
- The mean cost per patient for each CDA. A cost profile comprised the costs of ED attendance, hospital admission, and follow-up costs such as reattendance to the ED or ambulatory care.
What happened to the infants?
- 1242 (68%) infants received intravenous antibiotics.
- 1407 (77%) were admitted to hospital.
- The median length of hospital stay was two nights (IQR 2 to 3).
- 247 (14%) were discharged from the ED, 167 (9%) were ambulated, and 165 (9%) were admitted for observation without antibiotics.
- 16 (0.9%) were admitted to paediatric intensive care units (PICU).
- No infants died.
- 114 (6%) infants reattended emergency care within 7 days of discharge. Of these, seven were admitted, with two diagnosed with a UTI. None were subsequently diagnosed with IBI.
How did the Clinical Decision Aids perform (the primary outcome)?
- BSAC and AAP were the most sensitive tools, with the same sensitivity of 0.99 (95% CI 0.92 – 1.00). For those who need a reminder about sensitivity, this means that 99% of patients with an IBI would test positive with the BSAC and AAP tools (the true positives).
- NICE and Aronson had lower sensitivities of 0.93 (95% CI 0.83 – 0.98) and 0.90 (95% CI 0.80 – 0.96), respectively. However, these differences in sensitivities were not statistically significant.
- BSAC and AAP each misclassified one infant with an IBI as low risk. Both patients presented within six hours of the onset of fever. Aronson misclassified five infants with an IBI as low risk, and NICE NG143 misclassified seven.
- NICE and Aronson were the most specific, with the highest specificities of 0.27 (95% CI 0.25 – 30) and 0.30 (95% CI 0.28 – 0.32). This means that only one in three infants without an IBI would test negative (the true negative), with many false positive results. These specificities were statistically significantly higher than the specificities of the other CDAs—AAP’s specificity was 0.23 (95% CI 0.21 – 0.25), and BSACs’ specificity was 0.20 (95% CI 0.18 – 0.22).
- All four CDAs had a high negative predictive value, with BSAC and AAP adapted for use without PCT having the best NPV of 1.0 (95% CI 0.98 – 1.0). This means that approximately 100% (all) of the infants with a negative test would not have the disease.
- When the 260 infants with comorbidities were excluded, the specificities and sensitivities of the CDAs did not differ.
Secondary outcomes
67 (3.7%) infants had an IBI – 62 with bacteraemia and 9 with bacterial meningitis. Four had both bacteraemia and meningitis.
The most common organisms were E. coli and Streptococcus agalactiae (Group B Strep)—just over half of the infants with bacteraemia and 11% of those with meningitis grew E. coli, while Group B strep was grown in 16% of the blood cultures and one-third of the CSFs of infants with meningitis.
Four risk factors for IBI were identified:
1. Clinician gestalt – when the clinician suspected that an IBI was likely
2. No unwell contacts at home
3. No rash
4. No cough or coryza
The AAP rule performed just as well when used without PCT in a cohort of 466 infants, decreasing the cost per infant from £1,396 to £1,278.
Aronson and NICE were the cheapest tools, at £1,171 and £1,218 per infant, compared to a treat-all approach—sending blood, urine, and CSF and treating all febrile infants with intravenous antibiotics.
What were the limitations?
Variations in standard clinical practice between participating centres may have introduced potential selection bias.
It was assumed that infants who did not have blood tests had blood results within normal values. However, the authors tried to reduce potential bias by performing a sub-analysis with an incomplete biomarker dataset (CRP, WCC, and absolute neutrophil count) and excluding these patients without changing the results.
Only a few patients had PCT measured, and not all infants had blood tests sent, potentially biasing the interpretation of blood investigations. Although the study was adequately powered to detect the primary outcome, it was underpowered to detect small differences in the cohort that used PCT testing.
Reattendance and later diagnosis of an IBI were assessed via hospital chart review, which may have potentially led to an underestimation of the number of IBIs.
CASP checklist for cohort studies – How good was the paper?
- Does this address a clearly focused issue? Yes.
- Was the cohort recruited in an acceptable way? Yes. The study included prospectively enrolled consecutive febrile infants.
- Was the exposure accurately measured to minimise bias? Yes.
- Was the outcome accurately measured to minimise bias? Yes. Sensitivity, specificity, NPV, and PPV of each CDA were estimated. A sensitivity analysis excluding infants with prior comorbidities was performed.
- Have the authors identified all-important confounding factors? Yes. They also acknowledge the limitations of the study design, including the lack of PCT testing and variability in clinical practice between participating centres.
- Was the follow-up of subjects complete and accurate? No. Reattendance and eventual IBI diagnosis were assessed via hospital chart review. This may lead to a potential underestimation of IBIs. However, the IBI rate in the study is similar to prior literature findings.
- What are the results? Described above.
- Do you believe the results? Yes. The study was methodologically robust, and the authors applied validated clinical decision aids for febrile infants in a large multicentre prospective study.
- Can the results be applied to a local population? Yes. This was a multicentre prospective study in the UK and Ireland with thirty-five participating centres.
- Do the results fit with other evidence available? Yes. The rate of IBIs was similar to prior literature findings.
What did the authors conclude, and what does it mean for current practice?
In this large multicentre study evaluating the management of febrile infants across the UK and Ireland, 3.7% had an invasive bacterial infection.
The AAP and BSAC tools were the most sensitive, while the Aronson and NICE rules were the most specific, making them the most cost-effective. CDAs should be applied with caution in infants with early-onset fever.
The AAP rule performed just as well when CRP was used instead of PCT.
The results from FIDO suggest we can adopt a more tailored approach to managing febrile infants. This could potentially include observation for lower-risk infants while also reducing lumbar punctures and intravenous antibiotics in well-appearing febrile infants under 90 days.
What about that well-appearing nine-week-old infant with a fever of 38.3⁰C, no cough or coryza, and no sick contacts at home?
FIDO suggests you could use either the AAP or BSAC guidelines, sending just urine and blood tests without the need for PCT.
If the results were normal and it had been more than 6 hours since the fever onset, you could safely offer an observation period without the risk of missing an invasive bacterial infection.
And finally, a note from the author, Etimbuk Umana
Managing febrile infants poses challenges for clinicians, especially in conservative health systems that typically encourage a “treat all” approach. However, international studies have shown that a tailored approach using clinical decision aids (CDAs) can effectively identify a low-risk cohort who can be safely managed without invasive testing, antibiotics, or admission. The FIDO qualitative study revealed that both parents and clinicians support tailored care, with parents favouring fewer investigations.
The FIDO study provides updated reference data on the invasive bacterial infection (IBI) rate within UK and Irish populations. Recognising that populations and practices differ between healthcare systems and countries, we focused on the application rather than validation of CDAs. This distinction is crucial, as replicating the exact conditions of the original CDA derivations was not feasible. Encouragingly, when applied to our environment, these CDAs successfully identified a low-risk cohort for whom invasive testing and antibiotic administration could be avoided. However, caution is advised when using CDAs for infants presenting with early-onset fever.
The FIDO study offers clinicians insight into the post-test probability of IBI when these CDAs are applied, facilitating discussions with parents and caregivers. It also provides cost estimates for UK and Irish institutions and health systems considering CDA adoption. Additionally, we observed an evolving approach to febrile infant care, with the practice of ambulating both high and low-risk patients emerging as an alternative to admission. Overall, a tailored approach has proven to be both cost-effective and safe in managing febrile infants.
