I came across a preprint entitled “Evaluating the economic and health impact of proactive genomic epidemiology in a hospital setting”. In hospitals, we try to limit the spread of bacteria, especially Multi-Drug Resistant Organisms (MDRO), for obvious reasons. This study quantified the costs and benefits of routinely performing whole-genome sequencing (WGS) of all clinical isolates of 19 bacterial species, regardless of their site of isolation or antibiotic susceptibility profile: 18,940 clinical isolates from 7,760 patients in a single hospital during 28 months. WGS was combined with patient movement data to determine clusters of spread (epi-links).

Then they modelled the number of infections that would be prevented (and the associated costs saved) if infection control was based on the WGS results. “The holistic, proactive genomic epidemiology approach … could potentially generate net savings of €1.25 million annually and avoid more than 750 disability adjusted life years (DALYs).”

A landmark study? If the conclusions are valid and generalisable, yes. But are they? Some things need clarification first.

Clusters occurred predominantly for MDRO, and among them almost exclusively for VRE (362 isolates) and MRSA (14 isolates). In the Netherlands, infection control measures are installed for each patient with VRE or MRSA, regardless of WGS profile or epi-link. To me, it remains unclear what the standard of care in this Danish hospital was and to what extent WGS information would change that standard.

In the modelling it was assumed that all infection control based on WGS was 100% effective, that each infection prevented also prevented 7-8 days in hospital (MRDO or non-MDRO, €350/day), as well as Disability Adjusted Life Years (8 for bloodstream infection), all based on published data. But these published data were rough estimates from rough estimates in – predominantly – retrospective studies, and, therefore, not very precise. Together with another Danish hospital we were unable to quantify the attributable effects of VRE infections. In other words, there might be none. Yet, in this new study, most clusters were VRE and 75% of the 1,818 DALYs gained came from premature deaths prevented, mostly from VRE infections.

Fun fact: when quantifying the number of isolates that need to be sequenced to detect 1 cluster, VRE was by far the lowest (Figure 9), suggesting that WGS is not needed to detect clusters of VRE.

I trust that this study will now be peer-reviewed and I look forward to seeing the official publication.