Summary
A prospective observational study at two tertiary-care hospitals found that broad whole-genome sequencing detected many Gram-negative genomic clusters involving patients without multidrug-resistant isolates or intensive-care exposure. The medRxiv preprint also found that detection increased with sequencing effort and that blood-culture-only surveillance would have identified very few clustered patients.
A prospective observational study at two tertiary-care hospitals found that genomic surveillance detects a substantially different set of Gram-negative bacterial clusters when sampling is broad rather than focused on drug resistance, intensive-care units or suspected outbreaks. The findings come from a medRxiv preprint analysing 25,888 bacterial genomes from 13,712 patients in hospitals in Copenhagen and Hannover.
The researchers sequenced isolates of Escherichia coli, Klebsiella pneumoniae and Pseudomonas aeruginosa regardless of resistance phenotype, specimen type or ward. They identified clonal clusters using single-linkage analysis, grouping isolates separated by no more than 30 single-nucleotide polymorphisms (SNPs). SNPs are individual differences in the DNA sequence; a distance threshold is used to identify isolates that are genetically close enough to warrant epidemiological investigation. The analysis used within-patient distances and was checked against records of patients’ movement through the hospitals.
A broader sample changed which patients were detected
The study classified 14.1% of patients in Copenhagen and 16.5% in Hannover as part of a genomic cluster. These percentages refer to the study’s sampled populations and its clustering definition.
Many of the clustered patients would not have been selected by surveillance strategies centred on the usual high-risk signals. In Copenhagen, 71.3% of clustered patients had no multidrug-resistant isolate; the corresponding figure in Hannover was 50.0%. Most also had no ICU-associated isolate: 86.1% in Copenhagen and 68.7% in Hannover.
The comparison between the hospitals also showed why sampling composition matters. After the researchers matched 18-month periods and adjusted for ICU association, resistance and specimen mix, clustering was still more common in Hannover, with an adjusted odds ratio of 1.49 and a 95% confidence interval of 1.28–1.73. The result indicates that differences in what was sampled explained much of the initial contrast, but not all of it.
Blood cultures alone captured very few clusters
The clearest difference appeared when the researchers modelled narrower sampling strategies. In Hannover, sequencing blood-culture isolates alone would have identified only 14 of 900 clustered patients, or 1.6%. Excluding colonisation-screening samples would have missed 240 clustered patients, equivalent to 26.7% of the total.
The analysis also found that no targeted strategy detected more clustered patients than random sampling of the same number of isolates. Detection increased as more isolates were sequenced, without reaching a plateau in the study analysis. In practical terms, both the composition of the sample and the amount of sequencing determined how much genomic clustering became visible.
A genomic cluster is a signal that isolates are closely related under a defined genetic threshold. For infection-control work, that signal is interpreted alongside patient location and movement data to investigate possible links and transmission routes. The study therefore addresses the sensitivity of surveillance—who is brought into view—rather than assigning a single cause to every cluster.
The authors conclude that the observed 14–16% range is a lower bound for the two hospitals’ sampled populations. They recommend broad, representative sampling and reporting both which isolates were sequenced and how much sequencing was performed. The study is a preprint based on three bacterial species and two tertiary-care hospitals in Copenhagen and Hannover, so the percentages should be treated as site- and design-specific rather than as a universal hospital rate.