Article Text

Evaluation of hospital-onset bacteraemia and fungaemia in the USA as a potential healthcare quality measure: a cross-sectional study
  1. Surbhi Leekha1,
  2. Gwen L Robinson1,
  3. Jesse T Jacob2,
  4. Scott Fridkin2,
  5. Andi Shane3,
  6. Anna Sick-Samuels4,
  7. Aaron M Milstone4,
  8. Rajeshwari Nair5,
  9. Eli Perencevich5,
  10. Mireia Puig-Asensio5,
  11. Takaaki Kobayashi5,
  12. Jeanmarie Mayer6,
  13. Julia Lewis6,
  14. Susan Bleasdale7,
  15. Eric Wenzler7,
  16. Alfredo J Mena Lora7,
  17. Jonathan Baghdadi1,
  18. Gregory M Schrank1,
  19. Eli Wilber2,
  20. Amalia A Aldredge2,
  21. Joseph Sharp2,
  22. Kelly E Dyer2,
  23. Lea Kendrick8,
  24. Viraj Ambalam4,
  25. Scott Borgetti7,
  26. Anna Carmack1,
  27. Alexis Gushiken1,
  28. Ashka Patel1,
  29. Sujan Reddy9,
  30. Clayton H Brown1,
  31. Raymund B Dantes2,9,
  32. Anthony D Harris1
  33. On behalf of the CDC Prevention Epicenters Program
  1. 1Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland, USA
  2. 2Department of Medicine, Emory University School of Medicine, Atlanta, Georgia, USA
  3. 3Department of Pediatrics, Emory University, Atlanta, Georgia, USA
  4. 4Department of Pediatrics, Johns Hopkins School of Medicine, Baltimore, Maryland, USA
  5. 5Department of Internal Medicine, University of Iowa Hospitals and Clinics, Iowa City, Iowa, USA
  6. 6Department of Internal Medicine, University of Utah Health, Salt Lake City, Utah, USA
  7. 7Department of Medicine, University of Illinois College of Medicine, Chicago, Illinois, USA
  8. 8Children's Healthcare of Atlanta, Atlanta, Georgia, USA
  9. 9Divison of Healthcare Quality Promotion, Nationation Center for Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, USA
  1. Correspondence to Dr Surbhi Leekha, Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, MD 21201, USA; sleekha{at}som.umaryland.edu

Abstract

Background Hospital-onset bacteraemia and fungaemia (HOB) is being explored as a surveillance and quality metric. The objectives of the current study were to determine sources and preventability of HOB in hospitalised patients in the USA and to identify factors associated with perceived preventability.

Methods We conducted a cross-sectional study of HOB events at 10 academic and three community hospitals using structured chart review. HOB was defined as a blood culture on or after hospital day 4 with growth of one or more bacterial or fungal organisms. HOB events were stratified by commensal and non-commensal organisms. Medical resident physicians, infectious disease fellows or infection preventionists reviewed charts to determine HOB source, and infectious disease physicians with training in infection prevention/hospital epidemiology rated preventability from 1 to 6 (1=definitely preventable to 6=definitely not preventable) using a structured guide. Ratings of 1–3 were collectively considered ‘potentially preventable’ and 4–6 ‘potentially not preventable’.

Results Among 1789 HOB events with non-commensal organisms, gastrointestinal (including neutropenic translocation) (35%) and endovascular (32%) were the most common sources. Overall, 636/1789 (36%) non-commensal and 238/320 (74%) commensal HOB events were rated potentially preventable. In logistic regression analysis among non-commensal HOB events, events attributed to intravascular catheter-related infection, indwelling urinary catheter-related infection and surgical site infection had higher odds of being rated preventable while events with neutropenia, immunosuppression, gastrointestinal sources, polymicrobial cultures and previous positive blood culture in the same admission had lower odds of being rated preventable, compared with events without those attributes. Of 636 potentially preventable non-commensal HOB events, 47% were endovascular in origin, followed by gastrointestinal, respiratory and urinary sources; approximately 40% of those events would not be captured through existing healthcare-associated infection surveillance.

Discussion Factors identified as associated with higher or lower preventability should be used to guide inclusion, exclusion and risk adjustment for an HOB-related quality metric.

  • Quality measurement
  • Performance measures
  • Outcome Assessment, Health Care
  • Nosocomial infections

Data availability statement

Data are available upon reasonable request. Deidentified data are available upon reasonable request from the corresponding author.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • Hospital-onset bacteraemia and fungaemia (HOB) represents clinically significant events that can be measured using objective, electronically captured data, but our knowledge on the sources and preventability of HOB to inform its development as a surveillance and quality metric is limited.

WHAT THIS STUDY ADDS

  • The perceived preventability of HOB events caused by non-commensal (pathogenic) organisms is similar to or higher than other accepted measures of healthcare quality. A notable proportion of potentially preventable HOB events were associated with sources that would not be currently included in routine healthcare-associated infection surveillance at most US hospitals.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • HOB has potential to be used as an actionable healthcare quality metric with the application of inclusion and exclusion criteria and patient-level risk adjustment to account for events that are considered potentially not preventable.

Introduction

Hospital-onset bacteraemia and fungaemia (HOB) is being explored as an outcome measure for healthcare-associated infection (HAI) surveillance and as a potential healthcare quality metric by the US Centers for Medicare and Medicaid Services.1 HOB is a clinically significant measure given its association with high morbidity and mortality.2 3 Compared with central line-associated bloodstream infection (CLABSI), HOB is simpler to measure and more objective to classify, with a higher incidence of events.4 Gaps in our knowledge include sources or causes of HOB, and the proportion and types of HOB events considered preventable. Information on sources and potential preventability of HOB can define expectations for facilities initiating HOB surveillance, and guide refinement of HOB for the purposes of measuring quality of care via identification of factors associated with higher or lower preventability. In a pilot study of HOB cases at three US academic hospitals, 49% of 43 HOB cases with a non-skin commensal organism were considered potentially preventable.5 The limited scope and small sample size of that study does not allow for generalisation but did call for a framework to assess HOB preventability in a standardised manner. Therefore, prior to embarking on a larger evaluation of HOB, we developed and applied an expert consensus approach to refine and validate a preventability assessment guide for HOB.6

The objectives of the current study were (1) to determine the sources of HOB in a large, diverse sample of hospitalised patients in the USA, (2) to estimate the proportion of cases perceived to be preventable using the validated assessment guide and (3) to identify factors associated with perceived preventability. The results of this study will advance an understanding of prevention opportunities related to HOB and inform development of an appropriate subset of HOB as a potential quality measure.

Methods

We conducted an observational, cross-sectional study using in-depth chart review to evaluate the sources (ie, sites of infection or sources of bacteraemia and fungaemia) and potential preventability of HOB events among hospitalised adults and children at 13 hospitals (10 academic (including three free-standing children’s hospitals), three community) affiliated with six academic members of the Centers for Disease Control and Prevention’s (CDC) Prevention Epicenters Program (online supplemental methods). HOB was defined as a blood culture collected on or after hospitalisation calendar day 4 (where admission date is day 1) with growth of one or more bacterial or fungal organisms, regardless of blood culture results prior to day 4. Thus, the initial definition of HOB for this study was intentionally broad and inclusive of all blood cultures to facilitate unbiased identification of factors that could help refine an HOB metric suitable for surveillance and quality measurement.

Supplemental material

HOB event selection

Stratified sampling of HOB events was conducted from all positive blood cultures meeting HOB definition between 1 January 2016 and 30 June 2019. First, for a given patient, if there were multiple positive blood cultures (on or after day 4) during the same hospital admission, only one positive blood culture was randomly selected. Next, two separate lists were created at each hospital; one with HOB events with at least one pathogenic organism not considered to be a common skin commensal (henceforth referred to as ‘non-commensal’), and a second list of HOB events with only commensal organisms.7 Blood cultures were then randomly selected separately from the non-commensal and commensal lists to achieve the desired sample size of ~2000 total HOB events (online supplemental methods). Site-specific targets varied based on HOB incidence during the study period and overall desired ratios of commensal to non-commensal HOB (1:5) and paediatric to adult HOB (1:5) (online supplemental methods). Additionally, to maximise inclusion of cases from community hospitals (and therefore the generalisability of our findings), we included all non-commensal blood cultures from community hospitals given lower HOB incidence compared with academic hospitals. A patient was eligible to contribute more than one HOB event if the events occurred during separate hospital admissions.

Data collection and chart review

Primary reviewers assigned at each site (role: medical resident physician, infectious diseases fellow or infection preventionist) conducted in-depth medical record review to obtain information on patient clinical characteristics, procedures and surgeries, devices, results of clinical cultures and a narrative summary of the hospital course leading up to the HOB event. Reviewers were trained using a data collection guide as a template standardising the sequence, definitions and sources of information for medical record extraction (online supplemental methods). Study data were collected and managed using REDCap electronic data capture tools hosted at University of Maryland, Baltimore.8 9 Additionally, microbiology data (positive and negative cultures from blood and other body sites) for 14 days before and after the index blood culture were obtained through electronic data abstraction at each site.

Attribution of HOB source

The primary reviewer recorded their impression of the source or aetiology of HOB based on clinician documentation and their own clinical judgement and selected clinical criteria from a source-specific menu to support their impression. The US CDC National Healthcare Safety Network surveillance definitions10 were intentionally not used to allow the use of clinical judgement in adjudicating likely sources.

Determination of HOB preventability

At least one adult or paediatric infectious disease physician with training in infection prevention and hospital epidemiology was assigned as secondary reviewer at each site. Secondary reviewers were provided with a summary of the index case (online supplemental figure 1) and had access to all collected data with the ability to review electronic medical records to make a final determination of the source of HOB and indicate their agreement with the primary reviewer’s determination. Disagreements between primary and secondary reviewers were reconciled through discussion as necessary. Using the previously validated guide which provides a framework for assessing HOB preventability accounting for both patient risk and standard hospital-based post-admission prevention measures,6 a single secondary reviewer rated both the patient’s intrinsic HOB risk and extrinsic HOB preventability as low, medium or high for each HOB event (online supplemental methods). These ratings translated to a single 6-point Likert scale rating for overall HOB preventability (1=definitely preventable to 6=definitely not preventable, online supplemental figures 1 and 2).

Analyses

We summarised demographic and clinical characteristics of the cohort overall, and stratified by adult and paediatric patients, and by academic and community hospitalisations. Due to differences in their pathophysiology, we analysed commensal and non-commensal HOB cases separately for frequencies of microorganisms, HOB sources and preventability.

For non-commensal HOB cases, we compared organisms in the index blood culture, any additional positive blood cultures and cultures from other body sites with microbial growth. A culture was considered matching if at least one organism within (before or after) 14 days was the same as an organism in the index blood culture. We analysed the distribution of sources of HOB overall, and stratified by adult versus paediatric, and academic versus community hospitals. We calculated agreement on sources of HOB between primary and secondary reviewers as the proportion of cases where the secondary reviewer fully agreed with the primary reviewer on the suspected source(s) of bacteraemia, and additionally recorded partial agreement defined as episodes with multiple possible sources where the secondary reviewer agreed with at least one of the sources listed by the primary reviewer.

HOB preventability ratings were first summarised across the six categories of the 6-point Likert scale, and then analysed as a dichotomous variable ‘potentially preventable’ (rated 1–3) and ‘potentially not preventable’ (rated 4–6). Additional analyses among non-commensal HOB events were conducted as follows: We estimated mean (and SD) proportions of potentially preventable HOB among primary Epicenter sites, that is, health systems, and among individual hospitals using weights that were inversely proportional to the variance of the site-level and hospital-level proportions. Sites and hospitals supplying fewer cases would thus tend to have less weight. Weighted estimates were calculated separately for all commensal and non-commensal cases, and among non-commensal cases, separately for adult, paediatric, academic and community hospital cases. Using secondary reviewers’ justification to support their rating of extrinsic preventability (where provided), we summarised common themes and representative examples of reasons for low, medium and high preventability.

To inform the development of HOB as a quality metric, we conducted logistic regression analyses to evaluate the association between selected sociodemographic and clinical characteristics (including HOB source) with HOB preventability as a dichotomous outcome. Variables associated with preventability at p<0.1 in bivariate analyses were included in multivariable logistic regression after removing collinear variables identified using pairwise correlation coefficient >0.5. To understand implications for infection prevention practice and quality improvement efforts, we calculated the frequency of HOB sources only among those HOB events that were considered potentially preventable. To better understand the relationship of HOB to specific infections for which most US hospitals conduct routine surveillance and direct their current infection prevention efforts, we calculated the proportion of total and potentially preventable HOB events that were attributed to central line-related infection, indwelling urinary catheter-related infection, ventilator-associated pneumonia (VAP), surgical site infections (SSI) and methicillin-resistant Staphylococcus aureus (MRSA) bacteraemia. This analysis was based on clinical adjudication as described above and did not use CDC surveillance definitions.

Based on findings from our prior study6 which suggested that ratings of 3–4 capture uncertainty in preventability, in a sensitivity analysis, we also analysed the distribution of HOB preventability in three categories of preventable (1–2), uncertain preventability (3–4) and not preventable (5–6).

Results

Sample characteristics

We collected data on a total of 2109 HOB events in 2085 unique patients, 1754 (83%) in adult and 355 (17%) in paediatric patients. Consistent with the make-up of our sample of hospitals, most events occurred in academic hospitals (1768, 84%) (table 1). Overall, in 686 events (32% total; 28% of adult and 56% of paediatric events), the patient was hospitalised in an intensive care unit (ICU) in the 48 hours prior to the index blood culture. The median duration of hospital stay prior to the index blood culture was 11 days (IQR 6–20 days), and was longer for children (16 days, IQR 8–35) compared with adults (10 days, IQR 6–18), and for academic (12 days, IQR 7–21) compared with community hospitals (6 days, IQR 4–13). 28% of patients had undergone surgery in the preceding 30 days, and 39% had undergone other invasive procedures. Device use was common in this cohort: 1107 (65%) of adults and 259 (73%) of children had a central line on the day of or removed within 2 days before the HOB event. Receipt of total parenteral nutrition was reported in 345 (16%) patients, and 356 (17%) were receiving a form of renal replacement therapy (dialysis) (table 1).

Table 1

Demographic and clinical characteristics of participants with hospital-onset bacteraemia and fungaemia events from 13 US hospitals, January 2016 to June 2019, N=2109

Immunosuppression was prevalent in this cohort with 781 (37%) patients receiving a form of immunosuppressive therapy in the preceding 30 days; 464 (22%) were receiving chemotherapy for active malignancy, 143 (7%) were recipients of a solid organ transplant and 214 (10%) had undergone haematopoietic stem cell or bone marrow transplant. Neutropenia, defined as absolute neutrophil count less than 1.0x109/L, on at least two separate days before or after 7 days of the index positive blood culture, was present in 409 (19%) cases (table 1).

Microbiology

Among 1922 organisms in 1789 unique HOB events with at least one non-commensal organism, S. aureus was the most frequent organism reported, present in 23% of events overall and in 36% of events among patients in community hospitals, followed by Enterococcus spp, Escherichia coli, Candida spp, Klebsiella pneumoniae and Pseudomonas aeruginosa. Of 320 HOB cases with only commensal organisms, 267 (83%) contained coagulase-negative Staphylococcus; 67 (21%) had viridans group streptococci (table 2). Overall, 191 (9%) of HOB events and 179 (10%) of non-commensal HOB events were polymicrobial.

Table 2

Microorganisms associated with hospital-onset bacteraemia and fungaemia (HOB) events, stratified by non-commensal and commensal organisms

Of 1789 non-commensal HOB events, 459 (26%) had at least one other positive blood culture within (before or after) 14 days, during the same admission, 184 (10%) occurred prior to index HOB; 264 (15%) had a matching blood culture (127, 7% were prior to index HOB) and 553 (31%) had matching cultures from another body site (online supplemental table 1).

Sources of non-commensal HOB

For non-commensal HOB events, there was agreement on source between the primary and secondary reviewers in 1431/1789 (80%) (88% for community and 79% for academic hospital events), and partial agreement in additional 15% events. The most common sources of infection were gastrointestinal (including neutropenic translocation) (35%) and endovascular (32%), followed by respiratory (12%), urinary tract (10%) and skin and soft tissue (9%). No source could be identified in 17% of HOB events (14% adult and 35% paediatric) (figure 1 and online supplemental table 2). Among 622 events of gastrointestinal origin, 194 (31%) were attributed to neutropenic bowel translocation. Of 563 non-commensal HOB events with endovascular source, 488 (87%) were related to an intravascular catheter-related infection; 435 (77%) attributed to central line-related infection and 76 (14%) to a peripheral catheter. Specific types of central lines to which HOB events were attributed included peripherally inserted central catheter (155, 28%), tunnelled line or implanted port (157, 28%) and temporary central line (132, 23%). Three per cent (60/1789) of non-commensal HOB events were considered contaminated blood cultures (online supplemental table 1).

Figure 1

Distribution of major sources of hospital-onset bacteraemia and fungaemia (HOB) events, stratified by non-commensal and commensal organisms. *Includes neutropenic translocation.

Non-commensal HOB sources also varied by organism (online supplemental table 3); endovascular and skin and soft tissue were the top sources for S. aureus, and gastrointestinal and endovascular for Candida spp and Enterococcus spp. Among gram-negative bacteria, sources varied among Enterobacterales, P. aeruginosa and Acinetobacter spp. Likewise, S. aureus was the most frequent organism among HOB events of endovascular (38%), skin and soft tissue (63%) and bone and joint (77%) origin while E. coli, K. pneumoniae and Enterococcus spp were most frequent among HOB events from gastrointestinal and urinary sources.

Preventability of HOB

Among 1789 HOB events with non-commensal organisms, 636 (36%) were rated 1–3 or potentially preventable. Among 320 commensal organism HOB events, 238 (74%) were considered potentially preventable (figure 2). For non-commensal events, the mean proportion of potentially preventable cases weighted by Epicenter site was 35% (SD 4.5%), and weighted by hospital was 34% (SD 8.3%). Perceived preventability was higher for HOB events in community hospitals (40%, weighted mean 39% (SD 7.6%)) compared with academic (35%, weighted mean 34% (SD 8.1%)), and for HOB in adults (37%, weighted mean 36% (SD 8.2%)) compared with paediatric patients (29%, weighted mean 29% (SD 4%)). Common themes and representative examples of reasons associated with extrinsic preventability ratings are provided in online supplemental table 4.

Figure 2

Distribution of perceived preventability ratings of hospital-onset bacteraemia and fungaemia events, stratified by non-commensal and commensal organisms.

In multivariable logistic regression analysis among non-commensal HOB events, events attributed to intravascular catheter-related infection, indwelling urinary catheter-related infection, SSI and contaminated blood culture had higher odds of being rated potentially preventable compared with HOB cases from other sources. HOB events with the following characteristics had significantly lower odds of being rated potentially preventable: neutropenia, long-term immunosuppression, non-neutropenic, non-hepatobiliary gastrointestinal source, bone/joint source, no identified source, polymicrobial cultures and previous positive blood culture, compared with events without those characteristics (table 3).

Table 3

Association of demographic, clinical and microbiological attributes with preventability among non-commensal hospital-onset bacteraemia and fungaemia (HOB) events (N=1789)

When considering potential areas of focus for HOB prevention, among 636 potentially preventable non-commensal HOB events, the most common sources (not mutually exclusive) were endovascular (299, 47%, nearly all—297/299 intravascular catheter-related infection), gastrointestinal (112, 18%), urinary (85, 13%) and respiratory (83, 13%), and SSI (60, 9%). When evaluating HOB in the context of HAIs under routine surveillance, of 1789 non-commensal HOB cases, 751 (42%) were attributed to one of the following: central line infection, indwelling urinary catheter infection, VAP, SSI or MRSA bacteraemia. The perceived preventability of this subset of HOB events was 52% (387/751), and these events constituted 61% (387/636) of potentially preventable non-commensal HOB events.

In the sensitivity analysis of non-commensal HOB preventability in three (vs two) categories, we found that among 1789 non-commensal HOB events, 342 (19%), 379 (21%) and 1068 (60%) were considered preventable (1–2), uncertain (3–4) and not preventable (5–6), respectively. We found similar distributions of preventable (1–2) and not preventable HOB (5–6) by clinical characteristic and source, as with the dichotomous preventability outcome. Of 379 HOB events in the ‘uncertain’ preventability range (3–4), respiratory sources had the highest proportion (34%), while the uncertain proportion constituted 28%, 26% and 29% of events attributed to central line infection, urinary sources and SSI, respectively (online supplemental table 5).

Finally, among 320 HOB events with commensal organisms only, 182 (57%) were attributed to contaminated blood cultures and 168/182 (92%) were rated potentially preventable; 94 (29%) were considered bloodstream infections (BSI) and 47/94 (50%) were rated potentially preventable; in 34 (11%) events, both contamination and BSI were considered possible sources, and 21/34 (62%) were considered potentially preventable. Of total 128 commensal HOB events considered BSIs, endovascular was the most common source (59%) and of those, 56 (75%) were rated potentially preventable. No source was identified in 10 (3%) (online supplemental table 1).

Discussion

We studied over 2000 HOB events in adults and children at 13 US hospitals, with the motivation to inform development of HOB as a healthcare quality measure. Among 1789 HOB events with non-commensal organisms, gastrointestinal (including neutropenic translocation) (35%) and endovascular (32%) were the most common sources, and overall, 36% of non-commensal HOB events were rated potentially preventable. Among potentially preventable non-commensal HOB events, intravascular catheter-related infection was the most common source, followed by gastrointestinal, respiratory and urinary sources; approximately 40% of those preventable events would not be captured by existing routine surveillance at most US hospitals. Among HOB events with only commensal organisms, the most common source was contaminated blood cultures, with perceived preventability of 92%. These findings can be used to refine surveillance definitions for HOB-related quality measurement.

Our findings on HOB preventability are similar to prior studies both within and outside the USA. In a single-centre prospective assessment of hospital-onset BSIs in France in 2005–2006,11 among 272 BSI episodes with an identifiable source, using a 6-point Likert scale, 19% were considered preventable (rated 1–2), 25% uncertain (3–4) and 56% not preventable (5–6), with most preventable cases attributed to intravascular catheter-related infection. In a recent single-centre study in the USA, where device-associated bacteraemias, pneumonias, surgical complications and contaminated blood cultures were included in preventable HOB, 44% of 392 events were considered preventable.12 Finally, the preventability rating tool used in our study was also adapted for use in the study of HOB in India,13 where 45% of non-commensal HOB events were considered preventable, and infections associated with central lines, urinary catheter and mechanical ventilator accounted for 32% of preventable HOB events.

Our findings on HOB preventability have implications for developing HOB as a potential quality metric. First, these findings can help demonstrate to policymakers and the public that although commonly considered indicators of quality of care, HAIs—even when related to medical devices or procedures—may not be uniformly preventable. While a ‘gold standard’ per cent preventability for healthcare quality metrics is not defined, the estimated preventability of HOB is similar to or higher than reported potentially preventable proportions of non-infectious quality measures such as hospital readmissions (~27%)14 and hospital mortality (5–23%).2 15 16 Similarly, in a recent evaluation of 978 adverse events (including HAIs) at 11 Massachusetts hospitals, 23% were judged to be preventable.17 Accounting for patients’ intrinsic risk would also have lowered the perceived preventability of HOB events in our study, similar to a prior study in which estimates of preventability of hospital mortality decreased substantially when accounting for patients’ likelihood of being alive in 3 months.15

Second, when considering the use of HOB as a quality metric, ‘enriching’ the HOB measure by excluding non-preventable events, and risk adjusting to patient case mix, could help optimise allocation of quality improvement resources at the hospital level, improve buy-in with frontline clinicians12 and reduce bias in interhospital comparisons. In our analyses, a positive blood culture prior to the index HOB event, regardless of whether the organism ‘matched’ the organism in the index HOB event, was associated with lower odds of preventability. This suggests that multiple HOB events from the same patient should not be included in an HOB quality metric, as likely indicative of intrinsic patient risk rather than a preventable source of HOB. We also identified medical conditions associated with lower preventability, particularly neutropenia, immunosuppression and gastrointestinal sources, that should be accounted for in patient-level risk adjustment. Notably, our study cohort from mostly academic hospitals consists of patients at higher risk of poor outcomes relative to the population of all hospitalised patients, as indicated by high device use, duration of stay, immunosuppression and in-hospital mortality. This high mortality in patients with HOB noted here and in prior studies3 18–20 further underscores the need to account for patient risk factors in measurement of healthcare quality.

While intravascular catheter-related infections (including central line infections) constituted nearly half of all potentially preventable non-commensal HOB events, we estimated that approximately 40% of potentially preventable non-commensal HOB events would not be identified through routine HAI surveillance currently in place in most US hospitals. Gastrointestinal sources of HOB merit particular mention here given their high frequency overall, and while associated with a lower relative odds of preventability, these also constituted a notable absolute proportion of preventable non-commensal HOB events. In addition, 21% of non-commensal HOB events were rated in the ‘uncertain’ range (rated 3–4) for preventability. Together, these findings suggest that instituting HOB surveillance could allow recognition of events not detected by current surveillance for HAIs, and that further study of methods to prevent HOB from ‘non-standard’ sources is warranted.

The use of HOB as a modifiable infectious outcome is also supported by well-designed multicentre studies,21–23 showing that daily chlorhexidine bathing significantly reduces the risk of all-cause hospital-onset BSI in adult and paediatric ICU patients, including ‘primary’ bacteraemia or fungaemia without identified source. Mechanistically, this may be related to reduction in entry of organisms into the bloodstream via disruption of normal skin barriers during healthcare delivery—including reductions in intravascular catheter-related infections not meeting surveillance definitions and in contaminated blood cultures. These studies support the use of HOB as an expanded outcome for hospital infection prevention programmes when compared with the smaller proportion attributed to CLABSI.

Lastly, and not surprisingly, HOB cases with commensal organisms secondary to contaminated blood cultures had high preventability ratings. The association of contaminated blood cultures with unnecessary antimicrobial use, higher length of stay, costs and additional investigations, for example, echocardiograms,24 25 together with evidence for measures to reduce blood culture contamination,25–27 suggests that it would be appropriate to consider rates of contaminated blood cultures as a healthcare quality measure.

Strengths of this study include the diversity of our study sites and HOB events, inclusion of both adult and paediatric HOB events, engagement of paediatric and adult infectious disease specialists for case reviews and the use of a structured preventability rating guide validated against an expert panel consensus as the reference standard. Limitations include a heavy skew towards academic hospitals, and subjectivity and inherent uncertainty in the determination of both source and preventability. We attempted to minimise this through standardisation of the review process, but these assessments are naturally subjective particularly when done retrospectively and within the limitations of documentation in the medical record. Despite this, we found high agreement between primary (non-infectious disease experts) and secondary reviewers (infectious disease and hospital epidemiology experts) in source determination. The assessment of potential preventability is also limited by current assumptions on ability to prevent infections which may change with evolving knowledge, and only accounted for hospital-based prevention, whereas prehospital opportunities to prevent HOB were not considered. While this study was US focused, the potential use of HOB for automated HAI surveillance has gained traction in Europe,28 and the concept of using HOB to guide quality improvement is broadly applicable29 regardless of country and regulatory requirements. Future work should assess the association of HOB rates with the robustness of infection prevention programmes and nurse staffing levels, and include deeper prospective assessments of HOB ‘root causes’ to understand how HOB surveillance can guide hospital quality improvement efforts.

Supplemental material

Data availability statement

Data are available upon reasonable request. Deidentified data are available upon reasonable request from the corresponding author.

Ethics statements

Patient consent for publication

Ethics approval

This study was approved by the institutional review boards at Emory University (IRB00107002), Johns Hopkins University (IRB00189052), Utah University (IRB00116522), University of Illinois Chicago (2018-1471) and University of Maryland (HP-00082065). The University of Maryland IRB served as the IRB of record for University of Iowa.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • X @jestjac, @asicksamuels, @amenalor, @gregschrank

  • Contributors Author contributors: SL, JTJ, SF, AMM, EP, JM, SB, GMS, SR, RBD, ADH devised the study; SL, GLR, JTJ, RN, JM, SB, GMS, GN, SR, RBD, ADH developed the data collection instrument; SL, GLR, JTJ, SF, AS, AS-S, AMM, RN, MPA, TK, JM, JL, SB, EW, AJM-L, JB, GMS, GN, EW, AAA, JS, KED, LK, VA, SB, AC, AG, AP, ADH collected the data; SL, GLR, CHB analyzed the data; all authors contributed to interpretation of the results; SL took the lead in writing the manuscript; all authors provided critical feedback throughout the study and helped shape the implementation, analysis, and manuscript.

    Guarantor: SL accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.

  • Funding This study was funded by CDC (grant number: 1U54CK000450-01) and Iowa Institute for Clinical and Translational Science (grant number: UL1TR002537).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Linked Articles