Our pragmatic RCT is an important contribution to the literature on the clinical impact of rapid respiratory viral testing for four major reasons: 1) we are the first RCT to show reduction in ancillary studies in both adult and paediatric populations, 2) we assessed multiple clinical and health care utilization outcomes in both paediatric and adult populations, 3) assessed these outcomes in these populations in a randomized manner, and 4) we generated results within the context and confines of complex health and accredited laboratory systems, thus representing the outcomes that would be seen when formally implemented.
The primary outcome for PLC (improved oseltamivir utilization) was met, but not for ACH (reduction in LOS). At both study sites there was a significant reduction in oseltamivir prescriptions and number of doses administered to influenza negative patients in the ROST arm compared to the standard arm, concordant with other adult and paediatric studies (17, 18), including two RCTs (8, 9). We found a reduced mean time to first dose of oseltamivir in the ROST arm for patients with influenza A or B. This reached statistical significance at the ACH site with mean time to first dose being 18 hours shorter for the ROST arm, thereby increasing the chance of patients receiving oseltamivir within the ideal 48 hours from symptom onset window (19). A recent RCT in adults has similarly demonstrated a decreased time to first dose (8). Within the pediatric population, this has been previously shown in a retrospective cohort study in pediatric inpatients, but not a paediatric RCT (20).
In contrast to several studies showing increased antiviral prescription in influenza positive patients (8, 9, 18, 20), we did not find a significant difference in oseltamivir prescription in influenza positive patients. Although this may have been due to insufficient power in this subgroup, our findings support the need for education on appropriate prescribing.
We found no significant difference in LOS between the ROST and standard arm at either ACH or PLC sites. Although there was an indication that LOS in paediatric inpatients may be decreased, our study was underpowered to adequately assess this outcome. The impact of rapid respiratory tests on LOS varies in the literature with before/after designs in paediatric populations showing either shorter admission durations (21) or no significance difference (18). Similarly, results vary within the adult population. In a non-blinded RCT by Andrews et.al., found no difference in length of stay (22), however due to delays in specimen processing their rapid test TAT was 19 hours. Recently, a quasi-RCT (randomizing by week) by Bouzid et al., comparing POCT and central testing with Qiagen QIAstat-Dx Respiratory Panel V2® failed to show reduction in LOS with rapid testing with a TAT of 1.1 hour (15). Likewise, no change to admission duration in adult inpatients was seen in a recent RCT by Clark, et al., comparing molecular point-of-care test with TAT 1.2 hour with standard care (8). In contrast, Brendish et al., in a parallel-group, open-label RCT with research staff performing testing, found a significantly shorter mean length of admission in adult inpatients tested with the Biofire Filmarray® respiratory panel group compared to conventional laboratory polymerase chain reaction (PCR) (9). Other studies showing significant reduction in LOS are limited to specific populations or require multivariate analysis to adjust for clinical variables (21, 23, 24). We were unable to ascertain differences based on clinical variables given no chart reviews were performed.
Regarding our secondary outcomes, we found a significantly shorter TAT in the ROST arm with results reported in a mean 26 hours or more before the standard arm. Significantly reduced TAT with rapid testing is a consistent finding in rapid/ROST studies of various designs and quality (5, 8, 15, 20, 23, 25). However, the optimal time for a respiratory NAT result has not been defined and is important to identify as it would guide which tests and locations of testing can lead to clinical benefit.
Our RCT is the first pragmatic RCT to show a decrease in CXR orders associated with ROST. In the ACH study, we found a significant decrease in CXR orders in the ROST arm which has been shown previously in non-RCTs in both adult (26) and paediatric populations (20). Only one RCT showed a decrease in CXR orders, however they had intervened following rapid molecular testing to review the impact of the results with the ordering physician (13). In our subgroup analysis the proportion of patients with a CXR ordered was lower in the ROST arm for all subgroups, but only statistically significant in the influenza and RSV negative group, perhaps due to insufficient power in the positive subgroups. This supports the need for a quality improvement initiative to decrease unnecessary CXRs at ACH (27).
ROST was also associated with improved laboratory test utilization, with a significant reduction seen in the subgroup analysis for PLC inpatients testing positive for influenza. This is an important finding, given previous studies (non-RCTs) have shown no significant difference in laboratory test utilization (11, 18). Two non-RCTs which analysed subsets of diagnostic tests found significantly less orders post-rapid molecular implementation for ancillary microbiology testing (25) and C-reactive protein, complete blood count, and electrolyte analysis (26) in the rapid molecular cohorts. We did not differentiate between specific laboratory tests, however this would be important for future economic analysis.
Finally, we found no significant difference between antibiotic prescriptions between the ROST and standard arm at either site. This is consistent with other published randomized trials in adult (8, 9, 12, 15) and paediatric (12) populations, in addition to non-randomized trials in paediatric (18, 21) and adult (26) populations. This highlights that rapid test results in and of themselves are unlikely to change antimicrobial prescribing practice (13). Antibiotic stewardship interventions within the ED and hospital settings need to be developed and supported.
Overall, the impact of the ROST in our blinded study may have been reduced as physicians did not know the randomization until the result was reported. Although this methodology has its advantages, this could have impacted diagnostic and therapeutic decisions. In a randomized study by Bonner et al. where the rapid influenza testing results were known prior to clinician evaluation of the patient, rapid testing reduced antibiotic use, laboratory testing and radiographs (13).
The main limitation of our study was the lack of prospective recording of symptom onset date, diagnosis, and co-morbidities precluding the ability to control for these factors and perform associated subanalysis. Due to operational limitations and assay procurement regulations, we were unable to continue our study long enough to enrol the intended number of participants for the ACH primary outcome. In terms of randomization, although the pattern of rapid results could be ascertained by ordering clinicians, this was thought to be mitigated by the large size of these emergency departments and hospitals with the cycling of multiple care providers with various shift schedules. Additionally, our study did not address if rapid testing for pathogens other than influenza A/B or RSV on the standard of care panel would impact clinical management. Finally, we did not account for patients with tests positive for multiple targets. These patients only occurred in the ACH ED patients and were excluded only from subanalysis.
We did not assess isolation measurements for several reasons. Isolation days are not systemically recorded therefore recording this would require prospective monitoring which was beyond our funding. Additionally, per our Infection Prevention and Control Program, droplet isolation is based on symptoms and specific respiratory pathogen test result would only impact isolation if cohorting was required. This is compared to other regions where isolation to single rooms is based on positive respiratory pathogen testing (8, 15).
Although our enrolment period was partially eclipsed by the introduction by SARS-CoV-2 into Calgary, there was minimal impact given the low case number.