Clinical impact of rapid influenza PCR in the adult emergency department on patient management, ED length of stay, and nosocomial infection rate

Abstract Background Seasonal influenza causes significant morbidity and mortality and incurs large economic costs. Influenza like illness is a common presenting concern to Emergency Departments (ED), and optimizing the diagnosis of influenza in the ED has the potential to positively affect patient management and outcomes. Therapeutic guidelines have been established to identify which patients most likely will benefit from anti‐viral therapy. Objectives We assessed the impact of rapid influenza PCR testing of ED patients on laboratory result generation and patient management across two influenza seasons. Methods A pre‐post study was performed following a multifaceted clinical redesign including the implementation of rapid influenza PCR at three diverse EDs comparing the 2016‐2017 and 2017‐2018 influenza seasons. Testing parameters including turn‐around‐time and diagnostic efficiency were measured along with rates of bed transfers, hospital‐acquired (HA) influenza, and ED length of stay (LOS). Results More testing of discharged patients was performed in the post‐intervention period, but influenza rates were the same. Identification of influenza‐positive patients was significantly faster, and there was faster and more appropriate prescription of anti‐influenza medication. There were no differences in bed transfer rates or HA influenza, but ED LOS was reduced by 74 minutes following clinical redesign. Conclusions Multifaceted clinical redesign to optimize ED workflow incorporating rapid influenza PCR testing can be successfully deployed across different ED environments. Adoption of rapid influenza PCR can streamline testing and improve antiviral stewardship and ED workflow including reducing LOS. Further study is needed to determine if other outcomes including bed transfers and rates of HA influenza can be affected by improved testing practices.


| INTRODUC TI ON
Influenza remains a global health care burden. The average estimated cost of seasonal influenza in the United States is $11.2 billion annually. 1 Influenza infection can lead to severe morbidity and mortality among all patients, but the very young, the elderly, pregnant women, and persons with underlying disease are at higher risk. The efficacy of vaccination, our main defense against influenza, is variable and suboptimal.
During influenza season, outpatient and emergency department visits increase as patients seek care. In symptomatic patients, a rapid and accurate influenza diagnosis can improve management. 2,3 Delayed influenza diagnosis in the ED can adversely impact patient management. Hospital admission may be delayed while awaiting test information for bed assignment. Appropriate infection prevention measures are needed to reduce hospital-acquired (HA) influenza infections, a significant contributor to patient morbidity/mortality and healthcare costs. 4,5 Additionally, longer boarding times in the ED can lead to higher mortality and longer hospital length of stay (LOS). 6 Furthermore, delays in diagnosis can delay initiation of antiviral therapy beyond the recommended 48 hours after symptom onset. Finally, for non-admitted patients, lack of influenza test results before ED discharge can contribute to unnecessary antibiotic prescriptions. 7 Several laboratory methods are available to diagnose influenza, but historically there has been a trade-off between turn-aroundtime (TAT) and sensitivity. Less sensitive rapid influenza diagnostic tests (RIDTs) and direct fluorescent antigen testing (DFA) generate results within 15 minutes up to a few hours, while more sensitive nucleic acid amplified tests (NAAT) have substantially longer TAT, having typically been performed once a day in the local laboratory or sent to a reference laboratory. At our institution, methods for influenza diagnosis have historically varied depending on patient location and time of day.
While RIDTs can improve patient management in the ED, 3 poor sensitivity, especially in adults, has propelled a transition to NAAT especially for hospitalized patients. 8 Recently, several highly sensitive NAAT assays have become available that can be performed on demand with minimal hands-on time and short in-laboratory TAT.
Some of these are available as point-of-care tests (POCT), and some detect pathogens in addition to influenza.
We sought to optimize influenza testing practices among three EDs and to provide a PCR result within 1 hour of sample receipt in the laboratory, regardless of time of day, to streamline patient admissions, promote timely and appropriate use of anti-influenza medications, and hopefully reduce nosocomial transmission. We undertook a clinical redesign with stakeholders from emergency medicine, laboratory medicine, infection prevention, and bed management and implemented a series of changes during the 2017-2018 influenza season. We identified key metrics including test volume, test TAT, ED LOS, oseltamivir prescriptions, bed transfers, and rate of HA influenza infection to assess program effectiveness. Yale-New Haven Hospital (YNHH) is an urban 1541 bed academic tertiary care hospital located in New Haven, CT. YNHH has a mandatory influenza vaccination requirement for all directly employed staff and non-employed medical staff. YNHH operates EDs at three locations: YNHH main campus (Site 1), a smaller urban hospital approximately 0.5 miles from the main campus (Site 2), and a free standing community ED 15 miles from the main campus (Site 3). These three sites differ in size, population, and cachement area.

| Clinical redesign intervention
A multidisciplinary team involving stakeholders from the emergency department, laboratory medicine, infection prevention, bed management, and hospital Clinical Redesign Team was formed and met weekly to standardize influenza testing practices among EDs to optimize appropriate use of rapid influenza test results for bed management, infection prevention, and identification of patients for whom anti-influenza medications would be beneficial.
The following interventions were developed as a result of the clinical redesign process: implementation of a rapid influenza A/B PCR assay performed 24/7 at on-site laboratories for all ED locations, ED staff education, modification of the ED "Quick Pick List" to promote rapid PCR ordering, creation of a direct sample tubing pathway to the microbiology laboratory at Site 1, and modification of bed management processes (see below).

| Bed management
YNHH has a combination of single-and multi-occupancy rooms.
Patients with the same respiratory viruses can be cohorted, but once K E Y W O R D S bed management, ED length of stay, hospital-acquired influenza, influenza PCR, nosocomial influenza, rapid PCR a patient is identified as requiring respiratory virus isolation, either the patient or roommate(s) are transferred to an appropriate room.
During the 2016-2017 influenza season, bed assignments incorporated positive RIDT and DFA results, if available. After the intervention in 2017-2018, bed assignments for patients admitted through the ED were not made until results of rapid influenza PCR were available whenever possible. to Site 2 and Site 3 EDs 24/7, and at Site 1 from the hours of 10 pm and 7 am. Respiratory virus DFA testing was performed at Site 1 from 7 am to 10 pm and included influenza A, influenza B, respiratory syncytial virus (RSV), adenovirus, and parainfluenza viruses 1-3. 9 RVP was performed two to three times a day at Site 1 and included influenza A, influenza B, RSV, adenovirus, parainfluenza viruses 1-3, human metapneumovirus and human rhinovirus. All components of the RVP were LDTs. 10 The stand-alone influenza LDT PCR was run on the same schedule as the RVP. 11 Choice of testing was at the discretion of the ordering provider. When DFA, RVP, or influenza LDT PCR were ordered at Sites 2 and 3, samples were couriered to Site 1 for testing.

| Influenza testing
Following the clinical redesign process and intervention, a rapid, on demand influenza by PCR assay (Xpert Xpress Flu, Cepheid) was made available at all sites by 1/16/2018. The option to directly send specimens to microbiology/virology via pneumatic tube system was added at Site 1. Results at all sites were entered into the EMR and verified manually. Rapid PCR replaced RIDT, DFA, and stand-alone influenza PCR for ED ordering; RVP remained available primarily for admitted patients. All rapid influenza PCR testing was performed on-site in the local laboratories 24/7 as a STAT test, and the RVP was performed two to three times per day at the core virology laboratory at Site 1.

| Data collection
Data were extracted from the EMR (EPIC Systems) by the Joint Data Analytics Team at YNHH and Yale University. All ED encounters during which influenza testing was ordered in the ED were extracted along with associated clinical and demographic data: patient medical record number, date of birth, gender, unique visit identification number, ED location, ED visit start time, ED discharge time, admis- Separate data extractions were performed to identify all ED visits regardless of influenza testing or order location, with associated ED visit data and all influenza tests with associated laboratory data.
Cases of HA influenza are tracked as part of routine infection prevention.

| Exclusion criteria
The following exclusion criteria were applied to encounters: patient age <18 years, patient leaving the ED against medical advice or without being seen by a provider, patient dying while in the ED, patients triaged to psychiatry or obstetrics services, patients lacking complete demographic information, laboratory studies lacking complete order information, encounters with irresolvable or incomplete admission or triage information, and encounters with atypical laboratory order patterns. Atypical laboratory ordering patterns (29 total encounters) were considered visits with two or more tests ordered and/or specimens collected for which the order, collection, and/or resulting date/time stamps did not allow for clear calculation of time intervals under investigation. Due to differences in practice and workflow, the pediatric emergency department was not included in the clinical redesign process, and pediatric patients presenting to off-site EDs were excluded for consistency across all sites.

| Hospital-acquired infections
All encounters where patients with laboratory confirmed influenza diagnosed using specimens collected greater than 72 hours after admission underwent chart review. 12 Patients with onset of signs and symptoms of acute respiratory tract infection within the first 72 hours of hospitalization were considered community-acquired The rate of HA influenza was calculated as: for the time period in question.

| RE SULTS
A total of 5272 encounters met the case-finding criteria, and, after applying the exclusion criteria, 5118 encounters underwent further analysis (Table 1) (Table 1).
Looking at all specimens tested within 24 hours of ED presentation, we found that delayed identification of influenza was significantly reduced during the 2017-2018 influenza season (Table 2 We next examined rates of oseltamivir prescription among patients tested for influenza in the ED. There were significant treatment differences among the two influenza seasons for patients testing positive or negative for influenza ( Table 3)

| D ISCUSS I ON
We showed that implementation of rapid influenza PCR in conjunc- Note: Data include all oseltamivir prescriptions entered in the electronic medical record for both admitted and non-admitted patients. Rates of oseltamivir prescription were significantly different for both influenzanegative (P < .001) and influenza-positive (P < .001) patients between the two influenza seasons. more timely and targeted administration of anti-influenza medications and decreased ED LOS for patients undergoing influenza testing.
There was a significant decrease in the interval between ED triage and influenza test ordering in the 2017-2018 cohort. This is likely a result of the education campaign raising awareness of influenza testing and the clinical redesign project. When combined with a rapid and sensitive PCR assay performed 24/7, the more rapid generation of results likely led to the positive outcomes we observed.
As expected and previously shown, patients with laboratory identified influenza were diagnosed more quickly following the implementation of rapid PCR testing. 2,13 However, previous studies often look at only in-laboratory TAT, 2 while we reported the times from patient arrival in the ED to order placement and result reporting. Institutions seeking to optimize testing practices and reduce global test TAT should consider interventions that promote prompt test ordering and specimen collection, submission, and testing.
There was no significant increase in influenza positivity rate following transition to rapid PCR. This was possibly due to the Note: "Yes" includes only transfers of the same acuity. De-escalation from an ICU unit to a non-ICU unit was considered "No" transfer. Data does not include "Not Admitted" patients or patients placed in "Observation" (2016-2017 n = 595, 2017-2018 n = 1780). Data were not significant by Chi-squared analysis (P = .814).

F I G U R E 3
Hospital-acquired (HA) and total inpatient influenza infections. HA influenza cases were defined as described in the Methods. Flu-positive inpatients represent unique patients. The vertical dashed line (--) indicates the month in which rapid influenza PCR was available at all three EDs F I G U R E 4 ED length of stay. Median lengths of stay in minutes and interquartile ranges are shown. Time of arrival in ED was considered t = 0 min. Medians were compared by Mann-Whitney U Test. *** indicates P < .001, and ** indicates P < .01 shown a reduction in ED LOS. 15,16 In contrast, we found a significantly reduced ED LOS in influenza-tested patients, with median ED LOS decreasing 74 mintues (19.8%). We also looked at an important control population over the same period to capture the effects of changes in ED workflow independent of influenza testing. We found a significant, but small, difference in the median LOS of patients not undergoing influenza testing over the same period, but for those patients, the median LOS decreased by only 5 minutes (2.1%).
Rapid testing can facilitate bed management decisions for admitted patients and improve infection prevention even in the ED among patients who will be discharged. However, we were unable to identify a difference in the rate of bed transfers between the two influenza seasons. There were a smaller number of transfers than anticipated. It is possible that many of the bed transfers in the hospital are for patients directly admitted rather than those passing through the ED. Alternatively, the contribution of respiratory viruses to bed transfers could have been overestimated. Additionally, our case finding would not capture the movement of influenza-negative patients out of rooms with influenza-positive patients.
We hypothesized that implementation of early diagnostic testing for influenza would facilitate rapid identification of infected patients expediting the implementation of appropriate infection control measures and bed management decisions to collectively decrease the rate of HA influenza. However, we did not find a significant difference in the rate of HA influenza between the two study periods.
Youngs et al 17 reported that implementation of a comprehensive program including rapid influenza PCR testing reduced rates of HA flu. However, pre-intervention rates of HA influenza in their study were higher than our rate at baseline, suggesting that there was greater opportunity for improvement. Additionally, at our institution, prior to the availability of rapid influenza PCR, many patients underwent testing for influenza by LDT PCR shortly after admission.
Thus, infected patients may have been identified and placed on isolation relatively early in admission limiting patient-to-patient spread. In summary, clinical redesign featuring the implementation of a sensitive rapid influenza PCR led to a simplified testing algorithm, faster test TAT, more appropriate and earlier administration of antiviral therapy, and significantly shorter ED LOS. However, we found no impact on inpatient transfers and rates of HA influenza. No increase in influenza positivity was observed, likely due to prior availability of LDT PCR. Rapid, user-friendly FDA-cleared NAAT are significantly more expensive than RIDT, DFA, and LDT PCRs. To justify the added expense, improved patient outcomes should be documented. Our experience reinforced the critical importance multidisciplinary teams working together to reengineer workflows, streamline ordering, educate providers, and use rapid test results to promptly guide correct actions by caregivers. Even with the positive impacts observed in our study, opportunities for further improvements are evident and will be pursued in the future.

ACK N OWLED G EM ENTS
We thank YNHH Bed Management, Laboratory Medicine, and Emergency Medicine staff for assistance with implementation of the changes outlined. We also thank the Joint Data Analytics Team for providing data extracts for analysis. This project and study were funded entirely through clinical operations without external funding.

CO N FLI C T O F I NTE R E S T
Dr Martinello served on a scientific advisory board for Genetech for baloxivir. There are not other conflicts of interest.