Syndromic surveillance for neonatal sepsis via infant fever in EDs
Neonatal sepsis remains one of the most time-critical diagnoses in paediatrics. Every hour of delay between the first sign of illness and the first antibiotic dose can shift survival odds, yet the early clinical picture is often indistinguishable from a mild viral illness. A lethargic newborn who feels warm, feeds poorly, or breathes faster than usual may have a urinary tract infection, early-onset group B streptococcal disease, or a post-immunisation reaction.
Syndromic surveillance offers a faster net. Rather than waiting for laboratory-confirmed bloodstream infections to be notified days later, analysts monitor pre-diagnostic patterns such as the volume of infants arriving at emergency departments with recorded fever, the proportion triaged into high-acuity streams, or the rate of full septic work-ups. When those patterns drift from baseline, public health teams can investigate before an outbreak is declared. The same logic that powers school absenteeism monitoring can be redirected toward the smallest patients.
In Australia, where vast distances separate families from tertiary paediatric centres, the value of an early signal is amplified. A febrile newborn in regional Queensland or a remote Top End community may reach a hospital only after retrieval by the Royal Flying Doctor Service. Real-time triage data, once aggregated across states, can flag clusters that no single clinician would recognise. This approach complements existing notification pathways and slots into the country's syndromic surveillance platform.
The hour-glass nature of early-onset infection
Newborns present with sepsis in a way that compresses decisions. Early-onset disease, appearing within the first seven days, can progress from a subtle temperature elevation to septic shock within hours. Classic signs such as temperature instability, grunting, poor perfusion, and apnoea overlap with normal newborn physiology, and inflammatory markers rise too slowly to guide the first antibiotic dose. The Australasian College for Emergency Medicine febrile infant guideline recommends a full septic work-up for any newborn with a recorded fever of 38°C or higher.
That clinical conservatism generates a data stream that public health analysts can repurpose. The decision to perform a septic work-up is itself a syndromic event. If the daily count of "febrile infant under 28 days, full septic screen" rises above the seasonal baseline, the signal may reflect an evolving cluster rather than a change in triage habits. The same logic underpins many typhoid surveillance approaches in endemic regions, where pre-diagnostic patterns are tracked before culture confirmation.
Waiting for culture-confirmed bacteraemia reports is too slow. By the time a Group B streptococcus or Escherichia coli isolate reaches the National Notifiable Diseases Surveillance System, the clinical course has usually resolved. Syndromic surveillance repositions the data flow so that bedside indicators also feed the population-level early warning.
Mapping "fever in infants" to a usable indicator
The strength of any syndromic system depends on how cleanly the case definition translates from clinical notes to electronic records. In most Australian emergency departments, the triage nurse records a structured presenting complaint, and the clinician adds a working diagnosis. Both fields are imperfect proxies: a child may arrive afebrile after antipyretics given at home, or have a recorded temperature the clinician rounds up to "fever" in the notes.
For neonatal sepsis surveillance, analysts focus on infants under three months, where the prior probability of serious bacterial infection is highest. The case definition can include a triage temperature of 38°C or higher, a chief complaint of fever, a discharge diagnosis of suspected sepsis, or the performance of a blood culture. Combining these flags into a composite indicator smooths out coding differences. The Royal Children's Hospital Melbourne and Sydney Children's Hospitals Network use different triage vocabularies, yet their combined dataset yields a stable trend once harmonised.
Seasonality shapes the signal. In southern Australia, infant fever presentations peak with the winter respiratory virus season. In tropical north Queensland and the Northern Territory, the pattern is flatter and overlaid with year-round enteric and skin infections. Subtracting the expected seasonal component, or watching for a relative rise in the youngest age band, detects a true cluster earlier. Thresholds are set low enough that even small clusters in Tasmania or the ACT are not missed.
From triage note to real-time alert
Australian emergency departments record triage data within minutes of arrival. That timestamp is the entry point for syndromic surveillance. Each record, including age, presenting complaint, triage category, vital signs, and disposition, is pushed nightly or in near real-time to a state health department, and from there to a national aggregator.
The architecture is straightforward. Most jurisdictions already operate daily batch feeds for bed management and winter demand planning. Adding a parallel feed filtered to infants with a fever flag is a small extension. The harder work is governance: ensuring the data can be used under the Privacy Act 1988, that identifiers are hashed before leaving the hospital, and that alerts trigger action rather than another inbox message.
Once the data arrives, a baseline model, usually a negative binomial regression fitted to the previous three to five years, produces an expected count for each day and age stratum. An observed count exceeding the upper prediction interval, after adjusting for day of week and school holidays, generates an alert. In Brisbane's paediatric hospitals or Adelaide's Women's and Children's Hospital, an infectious diseases physician reviews such alerts within hours. The signal does not diagnose sepsis in any child; it tells the on-call team that the population pattern warrants a closer look.
Finding a signal in a sea of well babies
Neonatal sepsis is rare. Even in New South Wales, fewer than ten culture-confirmed early-onset cases may be recorded in a typical week, scattered across more than a hundred emergency departments. Syndromic surveillance works with the much larger pool of suspected cases that clinical work-ups generate.
Analysts use a layered approach. The first layer tracks broad, high-volume indicators such as all infant fever visits. The second narrows to infants under 28 days, where the prior probability of serious bacterial infection is highest. The third adds procedure flags: was a blood culture taken? Was a lumbar puncture performed? Was the child admitted to a special care nursery? Each layer adds specificity at the cost of sensitivity. A sensitive broad alert might prompt a reminder about sepsis screening, while a specific narrow alert might trigger an outbreak investigation.
Spatial dimension adds power in geographically uneven Australia. A cluster invisible in state-level data may stand out at the statistical area level 3, the smallest unit used by the Australian Bureau of Statistics. An increase in febrile infant presentations in a remote Western Australian community can be flagged even when the state total is unchanged. This matters where retrieval to a tertiary centre takes hours, and early recognition by a remote area nurse can shorten time to antibiotics.
Linking infant fever signals to pharmacy, ambulance, and lab data
No single data stream tells the whole story. The febrile infant who presents to a Sydney emergency department may have visited a general practitioner in the preceding 48 hours, been given paracetamol through the Pharmaceutical Benefits Scheme, and been transported by ambulance. Each touchpoint leaves a digital trace, and tying them together produces a richer picture than any one source.
Pharmacy dispensing offers a useful upstream signal. A spike in paediatric antipyretic sales in a postcode can foreshadow a wave of fever presentations several days later. Ambulance dispatch adds severity: a call coded as "neonate, febrile" is higher priority than a routine inter-hospital transfer. Microbiology labs provide the confirmation that the alert was pointing toward, with the next morning's blood culture log either validating the signal or prompting recalibration.
State health departments already operate several streams in parallel. New South Wales Health's emergency department surveillance system, Victoria's VEMD, and the Western Australian data collection each feed a national picture. Adding a neonatal-specific indicator is configuration rather than new infrastructure. The same approach has been trialled in Queensland during the wet season, when melioidosis can mimic early sepsis in remote Aboriginal communities.
From alert to action in diverse Australian settings
A surveillance signal is only as valuable as the response it triggers. In metropolitan Sydney or Melbourne, an alert might prompt a paediatric team to review recent infant admissions and reinforce sepsis screening education. In a regional centre such as Townsville or Cairns, it might trigger a telehealth consultation with a tertiary infectious diseases service and a check on retrieval logistics. In a remote Kimberley community or along the Torres Strait, it might prompt the local clinic to reinforce health promotion about when to seek care for a feverish newborn.
Equity shapes how alerts are interpreted. Aboriginal and Torres Strait Islander infants face higher rates of invasive bacterial infection and longer transit times to tertiary care, so a given threshold carries more weight in those communities. The Australian Institute of Health and Welfare has documented these disparities, and syndromic systems increasingly disaggregate signals by Indigenous status, remoteness area, and socioeconomic quintile.
Parental reassurance matters. A febrile infant visit does not mean a child has sepsis, and a syndromic alert does not mean parents should rush to emergency for every warm forehead. Plain-language communication about what the signal means belongs in the operational playbook. The Healthdirect Australia helpline and the Pregnancy, Birth and Baby service already support parents on individual concerns; materials about population-level surveillance help maintain trust in the data systems that protect the youngest patients.
What the signal cannot tell you
Syndromic surveillance for neonatal sepsis has clear limits. It cannot identify the pathogen, distinguish early-onset from late-onset disease, or replace clinical judgement at the bedside. The signal is statistical, not diagnostic, and a quiet week does not mean infections have stopped; it means only that the observed pattern has not crossed the alerting threshold.
Coding variability is the largest source of noise. Smaller emergency departments in rural South Australia or the ACT may use free-text triage notes that do not map cleanly to structured fields. Seasonal variation in respiratory virus activity, school terms, and media coverage of sepsis stories can shift parental health-seeking behaviour. Baselines must be re-fitted regularly, and alerts should always be reviewed by a clinician before any public-facing communication.
Privacy and governance impose their own constraints. Even hashed identifiers, combined with date of birth and postcode, can be re-identifiable for very young infants in small communities. Strict retention policies, role-based access controls, and transparent reporting of data use are essential. The Australian Health Protection Principal Committee has issued guidance on the proportional use of emergency department data, and any neonatal extension should sit within that framework.
If you work in a paediatric emergency department, a public health unit, or a research group studying early-onset infection, map your local triage fields to the case definition outlined here and pilot a baseline model on the past three years of infant presentations. Share your data dictionary, alerting thresholds, and false-positive experiences with colleagues across states, because the strength of the signal grows with the number of contributing sites. The smallest newborns in Australia deserve an early-warning system that is as attentive as the clinicians who care for them.