New rapid algorithm could help trace hospital-derived SARS-CoV-2 infections
By Dr. Liji Thomas, MDNov 17 2020
A major concern for both patients and hospital staff in the current coronavirus disease 2019 (COVID-19) pandemic has been the high risk of acquiring SARS-CoV-2 infection during a visit or stay at a hospital facility.
A new study reports a rapid method that uses two different routes to swiftly trace the onset of infection to a putative hospital source. This could be immensely helpful for preventing and controlling infection, once validated.
Study: Rapid feedback on hospital onset SARS-CoV-2 infections combining epidemiological and sequencing data. Transmission electron micrograph of SARS-CoV-2 virus particles, isolated from a patient. Image captured and color-enhanced at the NIAID Integrated Research Facility (IRF) in Fort Detrick, Maryland. Image Credit: NIAID / Flickr
The importance of sequencing
Infection prevention and control (IPC) practices have long been the mainstay of identifying and preventing routes of nosocomial transmission. However, there is now much evidence that this can be improved substantially by incorporating genome sequencing data. Such data can help link patients with hospital-onset COVID-19 infections (HOCIs) by matching their viral sequences to those found in the hospital during that period.
This is now possible using rapid sequencing methods that can yield results within 24-48 hours. Moreover, the genomic differences between SARS-CoV-2 sequences belonging to different strains provide the basis for different potential chains of transmission.
Nonetheless, phylogeny (the study of genetic evolutions) alone cannot provide strong support to link infections, or to indicate in which direction the spread occurred. This is why clinical data relating to the patient is also required.
Novel reporting tool for nosocomial spread
The researchers in the current study, which was part of the COVID-19 Genomics (COG) UK initiative, developed a novel tool – the sequence reporting tool (SRT).
This is designed to use both genomic sequencing data and epidemiologic data to identify patients with healthcare-associated infections (HCAIs, or nosocomial infections) among those with HOCIs.
HOCIs, HCAIs and viral isolates
HOCIs are inpatients who were not suspected to have COVID-19 on admission, and first test positive or have their earliest symptom for COVID-19 less than 48 hours after admission. The role of the SRT is to calculate the estimated probability that a given HOCI is a healthcare-associated infection (HCAI).
Public Health England (PHE) has its own IPC criteria to identify definite and probable HCAIs. If the first positive test is at 15 or more days after admission, the patient is termed to have definite HCAI. At 8-14 days, it is termed probable, and between 3 and 7 days, indeterminate.
PHE defines outbreaks of COVID-19 as two or more cases in the same ward, one being definite or probable HCAI. The SRT algorithm was used to identify one or more clear cluster of similar sequences within each such outbreak, using a difference of two single nucleotide polymorphisms as the cut-off for each cluster.
The COG-UK program sequences multiple viral isolates from both hospitals and communities in the UK. In the current study, the sequences were collected from Glasgow and Sheffield from February to May 2020.
The Glasgow samples had around 1,200 sequences, of which about 430 were from the community and the rest from hospitals. The sample does not accurately confirm those that came from healthcare workers (HCWs) except in 15 cases. The SRT algorithm was used to analyze 125 HOCI cases in the sample.
The Sheffield data came from around 1,600 sequences, with 800 from HCWs and around 830 from hospitals. For the sake of comparison, the 447/714 samples from inpatients taken at admission were considered to be community-onset cases. The SRT was applied to 200 HOCIs.
SRT identifies HCAIs by integrating matching sequences
PHE-defined definite HCAIs accounted for about 62% and 35% of the HOCIs, in Glasgow and Sheffield datasets, respectively. The SRT algorithm found close sequence matches on the same ward in 66% and 62% of definite and probable HCAIs, indicating that the virus probably spread within the ward. The researchers note, “When one or more close sequence matches was identified on the ward of the focus sequence, the SRT probability of infection on the ward was >0.5 in 185/189 cases.”
- Discovery of novel hidden gene in SARS-CoV-2
- Study shows SARS-CoV-2 direct heart muscle cell infection, cell death, loss of contractility
- Novel mutations in SARS-CoV-2 may explain high pathogenicity of COVID-19
The incomplete sequencing from hospitalized SARS-CoV-2 cases, with the presence of asymptomatic and untested carriers, may account for the roughly 7% of probable or definite HCAIs for which no matching sequence was found in hospital sequences. Only 40% or less of patients in Glasgow, and very few staff, were sequenced, while around 75% of patients and staff in Sheffield had sequences.
Another explanation may be that the virus spread from visitors; the researchers noted 26 HOCIs in Sheffield who had visitors allowed at the time of sampling. Three of the 26 had an estimated probability of infection from a visitor of 0.4 to 0.5, all being at 18 or more days from admission.
For indeterminate cases, the SRT showed that around 40% (33 out of 82) of cases were probably HCAI, with a probability estimate of >0.5, and with over 80% of the 33 having a close sequence match on the ward.
HCW sequences showed a match with HOCIs in 87% and 11% of sequences from Sheffield and Glasgow, obviously because the former was much richer in HCW sequences.
Comparison with IPC – Sheffield
In Sheffield, IPC identified 18/201 to be the index HOCIs – the first to be detected in that setting. Of these, 14 were the first isolated sequences on their wards, while one was the second. SRT showed one or more close matches on the same ward for 11/18 index cases, with the median probability of HCAI being 0.70.
PHE considered 144/201 HOCIs to be part of local outbreaks. The SRT showed the median probability of HOCI to be 0.98. Of these, 104 had at least one other matching sequence on the same word.
IPC considered the remaining 39/201 to be separate outbreaks, with 10/39 not being identified as HOCIs. SRT produced the median probability that these were HCAI of 0.74, and 7/39 showed close sequence matches on the same ward.
Multiple outbreaks from the same ward (six wards with two clusters of two or more patients each, and three wards with three clusters each) were seen in the Sheffield dataset. This detail of resolution achieved by the SRT is new, since IPC practices put them all into the same category. Again, there were 10 and 44 unique or genetically unrelated HOCIs in the Glasgow and Sheffield samples. Two were closely matched, but the admission to sampling date was two or more days.
What are the implications?
The authors sum up, “We have developed a novel approach for identification and investigation of hospital-acquired SARS-CoV-2 infections combining epidemiological and sequencing data, designed to provide rapid and concise feedback to IPC teams working to prevent nosocomial transmission.”
Phylogenetic analysis would be more accurate in confirming direct viral spread but takes far more time, and cannot tell the direction of spread. In over 90% of cases, the SRT tool allows nosocomial infection to be diagnosed, even without an index case, by identifying closely matched sequence networks linked to a known HOCI.
The tool was retrospectively validated, showing its ability to confirm most PHE-identified HCAIs, both probable and definite, and to provide a probability estimate for PHE-indeterminate HOCIs. In about a fifth of around 280 cases, infections thought to be part of a ward outbreak were actually unrelated.
The study also shows the key location of HCWs in the transmission network within a hospital. The SRT will help explore how a HOCI is linked to an HCW with a closely matched sequence.
The biggest benefit of this tool is the rapid identification of matching sequences within the hospital, with the estimated probability of nosocomial transmission, as and when sequencing data becomes available. This can enable the timely prospective application of sequencing data to IPC practice, allowing optimal use of resources where nosocomial transmission appears likely.
The no-frills probability estimation model that supports this tool requires little computational power, nor does it need to be refined and adjusted to local models. Further development may be carried out, using patient movement and HCW shift location data in the SRT. A prospective trial is planned for later this year. Finally, this may well be used to evaluate outpatient nosocomial transmission as well.
medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.