By Anais Cottage-stone · 25 August 2026
On Thursday, 19th June 2025, NeoNET AFRICA, in collaboration with the African Neonatal Association and KlebNET-GSP, held the second seminar in the Neonatal Outbreaks in Africa series. Chaired by Professor Angela Dramowski (Stellenbosch University, South Africa), the session moved from describing outbreaks to detecting them: how you know an outbreak is happening, how you find where it is coming from, and what your laboratory can realistically tell you.
Watch the full seminar recording here.
Dr Chiara Minotti (UKBB, the university paediatric hospital in Basel, Switzerland) walked through a Klebsiella oxytoca outbreak in a 32-cot NICU, traced across nearly two years with whole-genome sequencing alongside conventional investigation.
The trigger was modest: two late-onset sepsis episodes in November and December 2021. Screening and core genome multi-locus sequence typing identified an initial wave of eight confirmed cases, all preterm or very preterm, caused by a fully susceptible K. oxytoca ST18 strain. Targeted environmental sampling followed, covering water and upstream water sources, sinks, syphons, a bathtub, milk refrigerators, incubators and shared equipment, and the source was removed in February 2022.
Then, in July 2022, twins presented with the same sequence type, one with sepsis and one colonised, and their strains carried different resistance profiles. Sequencing tied both back to the original wave. What emerged over the second wave was not one outbreak but three running concurrently: K. oxytoca ST18 (38 confirmed cases), K. oxytoca ST199 (29 confirmed cases, also recovered from pooled milk refrigerator samples), and K. michiganensis ST381 (3 confirmed cases, with one isolate from a washbasin). Across the whole period, 816 infants were screened, and 152 isolates of Klebsiella were cultured from patient samples. The outbreak was declared contained at the end of June 2023.
Dr Minotti was frank about the constraints. Cost meant not every sample could be sequenced, and no cost-benefit analysis was done comparing this approach with conventional methods. Where sequencing is not available, she pointed to rep-PCR and other comparative genotyping methods, and to collaboration with reference laboratories, with the practical advice to make the best use of whatever typing method your own laboratory already has. The outbreak report has been accepted for publication in Antimicrobial Resistance and Infection Control.
Dr Lauren Hookham (Brighton and Sussex Medical School and MRC Uganda) gave a practical guide to environmental sampling, drawing on work in Uganda, Kenya and the UK.
Her central argument was that sampling should start with the organism, not the swab. Pseudomonas aeruginosa favours moist, biofilm-prone surfaces, so sinks, drains, taps and respiratory equipment are the places to look. Acinetobacter survives desiccation and belongs to dry, high-touch surfaces such as bed rails, monitors and door handles. Klebsiella spans both, turning up in bedding, incubators, suction bottles, sinks and staff gloves. Where the organism is unknown, she suggested a default spread: sinks and drains, the bedspace and its shared equipment, floors and corners for spore or dust-borne organisms, and air vents if airborne transmission is suspected.
Before any swab is opened, she recommended simply watching the ward. Informal sketches, quick logs of who touches what and how often, and photographs where appropriate all help identify the surfaces that actually matter. Without that step, sampling risks becoming what she called token sampling of irrelevant surfaces. Her example observation checklist recorded touches per hour, who did the touching, whether hands were gloved, and whether anything was cleaned in between, which is often where the real finding sits.
She then compared the tools themselves, including contact plates, dipslides, PetriFilm, swabs and sponges, noting that contact plates give quantitative results in CFU per square centimetre and so allow threshold-based interpretation, while swabs remain the most commonly used method. Pre-analytical choices matter more than they might appear: whether a swab is wetted by dipping or dropping affects both sterility and saturation, and results are further shaped by surface material, contact time, transport medium and technician skill. She also flagged publication bias in the existing literature, and made a plea for photographing sampling sites so that results remain reproducible and traceable.
Professor Andrew Whitelaw (Stellenbosch University and NHLS Tygerberg Hospital, South Africa) closed the session by turning the question around, asking not what clinicians can do for the laboratory but what the laboratory can do for them.
Laboratory surveillance, he noted, is almost always passive. To show what that means in practice, he took a common arrangement in which the laboratory notifies clinicians about organisms of interest (ESBL-E, MRSA, CRE, VRE, CRAB) and about unusual clusters from the same ward, then tested it against three cases. A two week old neonate with a line infection and a Staphylococcus aureusblood culture resistant to penicillin only triggers nothing, because that organism is not on the list. A similar neonate with a negative blood culture triggers nothing at all. Widening the rules to include every positive culture from the NICU and every positive blood culture catches more, at the cost of information overload.
Cluster detection carries its own assumptions: that you know what a normal number looks like, that someone is keeping a record of what has been seen recently, and that the laboratory communicates internally. He was clear about the limits, too. A laboratory cannot give you healthcare-associated infection rates, because not every positive culture is an infection and not every infection yields a positive culture, and no outbreak detection system is infallible.
On automation, he cited a comparison of three systems, WHONet, CLAR from Berlin and P75 from Utrecht, run over a seven year dataset. WHONet generated far fewer alerts but none for sporadic organisms, illustrating the sensitivity and specificity trade-off, and all three required curated data and updated thresholds. On genomics, he pointed to work comparing reactive, IPC-driven sequencing with routine sequencing combined with machine learning on electronic patient data. Across 2,752 unique patient isolates, 297 isolates formed 99 clusters, and potential transmission routes were identified for 65% of them. Reactive investigation had identified two.
His closing list for clinicians was refreshingly concrete: send good quality specimens, identify them properly, complete the request forms, talk to the laboratory, respect its limitations, and advocate for its resources.
Seminar 3, on outbreak control and prevention, followed on 24th July 2025.
