The impact of the COVID-19 pandemic on blood culture practices and bloodstream infections

ABSTRACT The COVID-19 pandemic has likely influenced the epidemiology of bacterial infections through wide-ranging changes to clinical practices and infection control and prevention interventions. We sought to determine how the detection and incidence of bloodstream infections (BSIs) have been influenced by the pandemic. We performed a retrospective analysis of blood culture data in the province of Ontario, Canada, from 1 January 2017 to 31 December 2020. Outcomes included a weekly incidence of blood culture tests, BSIs, and contaminant results. Results were stratified by hospital, community, and long-term care (LTC) settings. An interrupted time series analysis using segmented regression models was used to determine changes in outcome incidence/prevalence during the pre- and peri-pandemic periods. Of the 14,083,853 individuals included, 129,329 (0.92%) developed a bloodstream infection. The blood culture ordering rate increased during the pandemic in the hospital setting only [Incidence rate ratio (IRR) 1.09, 95% confidence interval (CI) 1.01–1.19]. There was a decline in the incidence of community-acquired (IRR 0.95, 95% CI 0.91–0.99) and LTC-acquired (IRR 0.85, 95% CI 0.76–0.94) BSIs. Hospital-acquired BSIs were unchanged. The proportion of blood culture contaminants increased in the community (7% increase, P < 0.01) and LTC settings (14% increase, P < 0.05). There was decreased incidence of community-acquired Streptococcus pneumoniae (IRR 0.43, 95% CI 0.33–0.57) and Staphylococcus aureus (IRR 0.91, 95% CI 0.84–0.99) bacteremia. Pandemic-related changes in the performance of blood cultures and the epidemiology of BSIs have implications for current and future pandemic antimicrobial use, healthcare resource allocation, and hospital and laboratory policies. IMPORTANCE Bacterial infections are a significant cause of morbidity and mortality worldwide. In the wake of the COVID-19 pandemic, previous studies have demonstrated pandemic-related shifts in the epidemiology of bacterial bloodstream infections (BSIs) in the general population and in specific hospital systems. Our study uses a large, comprehensive data set stratified by setting [community, long-term care (LTC), and hospital] to uniquely demonstrate how the effect of the COVID-19 pandemic on BSIs and testing practices varies by healthcare setting. We showed that, while the number of false-positive blood culture results generally increased during the pandemic, this effect did not apply to hospitalized patients. We also found that many infections were likely under-recognized in patients in the community and in LTC, demonstrating the importance of maintaining healthcare for these groups during crises. Last, we found a decrease in infections caused by certain pathogens in the community, suggesting some secondary benefits of pandemic-related public health measures.

of CoNS blood culture contamination in community/LTC cultures (decreased), and in the rate of blood stream infection detection in community/LTC (decreased).In general, the observed effect sizes are small.The manuscript is clearly written.

Major Points 1. Choice of Impact Model
The authors do not provide an explanation for their choice of ITS impact model.While looking for a level-change makes intuitive sense, there are additional considerations when considering the impact of the pandemic on the studied outcomes.Some changes due to the COVID-19 pandemic such as staffing/supply challenges may not be expected necessarily to manifest immediately and might be best modeled with a slope-change model and/or a lag-change model (see Bernal 2017 PMID 27283160) or even a combination of level and slope change models.For example, there visually appears to be a difference in the blood culture ordering rate between the pre-and peri-pandemic periods in the community setting with a transient dip occurring at the time of the declaration of the pandemic.A level + slope change model could be used to determine if this increase is the result of a pre-existing trend or if the onset of the pandemic may have contributed.

Inclusion of summary statistics/Interpretation of ITS
The authors use of a rigorous statistical method (ITS) is commendable.However, there are some instances where I think the results are misapplied and a "zoomed out" perspective might be instructive.For example, in Figure 3B it appears that the peripandemic hospital BSI rate is higher than the pre-pandemic rate.However, this could be due to continuation of a pre-pandemic trend based on simple visual review of the data.The authors claim, based on the ITS analysis, that there was no change in hospital BSI but this visually seems to not be the case.It would be informative to see the median pre-and peri-pandemic hospital BSI rates compared.ITS (perhaps with a level + slope change impact model, see above) could then be used to determine how likely any observed difference is to be attributable to the pandemic.In short, a non-significant ITS co-efficient does not mean that there is no difference, just that the data does not support an immediate effect (in a level change model) of the onset of the pandemic.Figure 1A appears to be another example of this.
A related example is Figure 3A (community BSI).Here visual inspection shows higher BSI rates in the peri-pandemic period compared to pre-pandemic.Inclusion of summary statistics would confirm this.However, the ITS model returned a significant coefficient for a level change and so a significant decrease in community BSI is reported by the authors.This is likely driven by the small cluster of datapoints in March/April 2020 that are substantially below the overall trend.

Determination of blood culture contamination
The authors deal with suspected blood culture contamination in two separate ways without clear explanation as to why.All blood cultures with Corynebacterium spp., Cutibacterium acnes, Micrococcus spp., or Bacillus spp.were excluded.Conversely, CoNS identified only on one day in a 14-day period was used as proxy for contamination whereas if the same CoNS species was identified more than once in a 14-day period it was treated as a true BSI.If the data is available, including the other skin flora in the calculation of contamination rate would give a more complete picture.

Limitations
The authors discuss several important limitations of their work including concerns about missing data particularly in regards to LTC patients.Limitations also include the inability to identify which aspects of the pandemic and the response to it might have contributed to the observed changes.This affects the impact of these observations in terms of their ability to direct future research.
Minor Points 1. Supplemental Figure 2-Panel A is listed as being by week but data points appear to be graphed by month.Conversely, Panel B appears to be by week but is labeled as by month.

Reviewer #2 (Comments for the Author):
This is a large study involving at the population-level of a province in Canada.The method of analysis using segmented regression models and stratified by hospital, community, and long-term care settings is adapted.
Only the first 9 months of the pandemic were taken into account.It would have been interesting to include the year 2021, especially as we are already more than 3 years from the start of the pandemic.Moreover, there are already many articles on the impact of the COVID-19 pandemic on BSIs, including reviews, and the article presented does not really add anything new.
There are a several methodological considerations: -The time taken to differentiate between nosocomial and community-acquired is 48 hours in international definitions, not 72 hours.
-The definition of CoNS BSI does not match international definitions either.Without clinical data (initiation of antibiotherapy, etc.), it is difficult to conclude to conclude that it is a CoNS infection.
-The number of patients in the community, in LTCs, and in hospitals are specified.These data evolve day by day.In addition, the date considered to be the beginning of the pandemic period is not specified.
-The cohort is large enough to present only significant results.The terms "did not reach statistical significance" or "trend" or "statistical power limited" should not be used.
In addition, the results section is short compared to the methods and discussion sections, while many results are presented in tables and figures.
The results show significantly fewer contaminants in the hospital and more in the community and LTC.The proposed explanation ("attributable to new PPE protocols, increased workload, reassignment of nursing staff to new units, or other factors that primarily affected emergency departments and ambulatory laboratories") is also valid for the hospital, but does not explain why a decrease is observed there.
Line 53: "the effects of the pandemic on bacterial infections remains to be established".This is not correct, the existing literature should be cited and the contribution of the article to this literature should be specified.Line 59: "non-transmissible infections were likely under-recognized".The definition of "non-transmissible infections" is not clear, and the choice of the term "non-transmissible" does not seem appropriate, since infections are caused by transmissible germs.Line 57: "blood tests for infection were more likely represent false-positive results, which likely cause unnecessarily antimicrobials and hospital stays for patients".Contaminant in an isolated blood culture does not lead to a prescription.
Bacterial names should appear in italics in the figures.

Preparing Revision Guidelines
To submit your modified manuscript, log onto the eJP submission site at https://spectrum.msubmit.net/cgi-bin/main.plex.Go to Author Tasks and click the appropriate manuscript title to begin the revision process.The information that you entered when you first submitted the paper will be displayed.Please update the information as necessary.Here are a few examples of required updates that authors must address: • Point-by-point responses to the issues raised by the reviewers in a file named "Response to Reviewers," NOT IN YOUR COVER LETTER.
• Upload a compare copy of the manuscript (without figures) as a "Marked-Up Manuscript" file.
• Each figure must be uploaded as a separate file, and any multipanel figures must be assembled into one file.
• Manuscript: A .DOC version of the revised manuscript • Figures: Editable, high-resolution, individual figure files are required at revision, TIFF or EPS files are preferred For complete guidelines on revision requirements, please see the journal Submission and Review Process requirements at https://journals.asm.org/journal/Spectrum/submission-review-process.Submissions of a paper that does not conform to Microbiology Spectrum guidelines will delay acceptance of your manuscript." Please return the manuscript within 60 days; if you cannot complete the modification within this time period, please contact me.If you do not wish to modify the manuscript and prefer to submit it to another journal, please notify me of your decision immediately so that the manuscript may be formally withdrawn from consideration by Microbiology Spectrum.
If your manuscript is accepted for publication, you will be contacted separately about payment when the proofs are issued; please follow the instructions in that e-mail.Arrangements for payment must be made before your article is published.For a complete list of Publication Fees, including supplemental material costs, please visit our website.
The authors conducted a retrospec ve review of blood culture and administra ve data to analyze the impact of the early COVID-19 pandemic (March -December 2020) on the incidence of blood culture collec ons, blood stream infec ons, and various secondary outcomes in Ontario, Canada.They use a level-change interrupted me series (ITS) analysis to make these comparisons.They report observed differences in the rate of blood culture collec on in hospitalized pa ents (increased), the rate of CoNS blood culture contamina on in community/LTC cultures (decreased), and in the rate of blood stream infec on detec on in community/LTC (decreased).In general, the observed effect sizes are small.The manuscript is clearly wri en.

Major Points 1. Choice of Impact Model
The authors do not provide an explana on for their choice of ITS impact model.While looking for a level-change makes intui ve sense, there are addi onal considera ons when considering the impact of the pandemic on the studied outcomes.Some changes due to the COVID-19 pandemic such as staffing/supply challenges may not be expected necessarily to manifest immediately and might be best modeled with a slope-change model and/or a lag-change model (see Bernal 2017 PMID 27283160) or even a combina on of level and slope change models.For example, there visually appears to be a difference in the blood culture ordering rate between the pre-and peri-pandemic periods in the community se ng with a transient dip occurring at the me of the declara on of the pandemic.A level + slope change model could be used to determine if this increase is the result of a pre-exis ng trend or if the onset of the pandemic may have contributed.

Inclusion of summary sta s cs/Interpreta on of ITS
The authors use of a rigorous sta s cal method (ITS) is commendable.However, there are some instances where I think the results are misapplied and a "zoomed out" perspec ve might be instruc ve.For example, in Figure 3B it appears that the peri-pandemic hospital BSI rate is higher than the pre-pandemic rate.However, this could be due to con nua on of a prepandemic trend based on simple visual review of the data.The authors claim, based on the ITS analysis, that there was no change in hospital BSI but this visually seems to not be the case.It would be informa ve to see the median pre-and peri-pandemic hospital BSI rates compared.ITS (perhaps with a level + slope change impact model, see above) could then be used to determine how likely any observed difference is to be a ributable to the pandemic.In short, a nonsignificant ITS co-efficient does not mean that there is no difference, just that the data does not support an immediate effect (in a level change model) of the onset of the pandemic.Figure 1A appears to be another example of this.
A related example is Figure 3A (community BSI).Here visual inspec on shows higher BSI rates in the peri-pandemic period compared to pre-pandemic.Inclusion of summary sta s cs would confirm this.However, the ITS model returned a significant coefficient for a level change and so a significant decrease in community BSI is reported by the authors.This is likely driven by the small cluster of datapoints in March/April 2020 that are substan ally below the overall trend.

Determina on of blood culture contamina on
The authors deal with suspected blood culture contamina on in two separate ways without clear explana on as to why.All blood cultures with Corynebacterium spp., Cu bacterium acnes, Micrococcus spp., or Bacillus spp.were excluded.Conversely, CoNS iden fied only on one day in a 14-day period was used as proxy for contamina on whereas if the same CoNS species was iden fied more than once in a 14-day period it was treated as a true BSI.If the data is available, including the other skin flora in the calcula on of contamina on rate would give a more complete picture.

Limita ons
The authors discuss several important limita ons of their work including concerns about missing data par cularly in regards to LTC pa ents.Limita ons also include the inability to iden fy which aspects of the pandemic and the response to it might have contributed to the observed changes.This affects the impact of these observa ons in terms of their ability to direct future research.
Minor Points 1. Supplemental Figure 2-Panel A is listed as being by week but data points appear to be graphed by month.Conversely, Panel B appears to be by week but is labeled as by month.
This is a large study involving at the population-level of a province in Canada.The method of analysis using segmented regression models and stratified by hospital, community, and long-term care settings is adapted.
Only the first 9 months of the pandemic were taken into account.It would have been interesting to include the year 2021, especially as we are already more than 3 years from the start of the pandemic.Moreover, there are already many articles on the impact of the COVID-19 pandemic on BSIs, including reviews, and the article presented does not really add anything new.
There are a several methodological considerations: -The time taken to differentiate between nosocomial and community-acquired is 48 hours in international definitions, not 72 hours.
-The definition of CoNS BSI does not match international definitions either.Without clinical data (initiation of antibiotherapy, etc.), it is difficult to conclude to conclude that it is a CoNS infection.
-The number of patients in the community, in LTCs, and in hospitals are specified.These data evolve day by day.In addition, the date considered to be the beginning of the pandemic period is not specified.
-The cohort is large enough to present only significant results.The terms "did not reach statistical significance" or "trend" or "statistical power limited" should not be used.
In addition, the results section is short compared to the methods and discussion sections, while many results are presented in tables and figures.
The results show significantly fewer contaminants in the hospital and more in the community and LTC.The proposed explanation ("attributable to new PPE protocols, increased workload, reassignment of nursing staff to new units, or other factors that primarily affected emergency departments and ambulatory laboratories") is also valid for the hospital, but does not explain why a decrease is observed there.
Line 53: "the effects of the pandemic on bacterial infections remains to be established".This is not correct, the existing literature should be cited and the contribution of the article to this literature should be specified.
Line 59: "non-transmissible infections were likely under-recognized".The definition of "nontransmissible infections" is not clear, and the choice of the term "non-transmissible" does not seem appropriate, since infections are caused by transmissible germs.
Line 57: "blood tests for infection were more likely represent false-positive results, which likely cause unnecessarily antimicrobials and hospital stays for patients".Contaminant in an isolated blood culture does not lead to a prescription.
Bacterial names should appear in italics in the figures.