Surges of novel coronavirus 2019 disease (COVID-19) can severely strain healthcare systems. Safe and efficient capacity management is crucial for hospitals, but hospital-acquired COVID-19 frustrates those efforts by increasing lengths of stay, morbidity, and mortality. Reference Elkrief, Desilets and Papneja1–Reference Lessells, Moosa and de Oliveira3 Several issues complicate prevention of nosocomial COVID-19. First, symptoms of congestive heart failure, chronic obstructive pulmonary disease, or other chronic cardiorespiratory conditions can be indistinguishable from those of severe acute respiratory coronavirus virus 2 (SARS-CoV-2). In a recent report, patients with a delayed diagnosis of COVID-19 were more likely to present with heart failure and to have none of the cardinal symptoms of COVID-19 than patients who were diagnosed immediately upon admission (adjusted odds ratio [OR], 2.36; 95% confidence interval [CI], 1.15–4.84). Reference Pfoh, Hariri, Misra-Hebert, Deshpande, Jehi and Rothberg4 Another issue is the time-dependent nature of the SARS-CoV-2 real-time reverse transcription polymerase chain reaction (RT-PCR) assay. For example, collection and testing 4 days before symptom onset resulted in a false-negative test in 100% of samples, which decreased to 67% the day before symptom onset and 38% on the day of symptom onset. Reference Kucirka, Lauer, Laeyendecker, Boon and Lessler5
After fatal nosocomial outbreaks at our facility, we sought (1) to prevent future occurrences and (2) to evaluate the benefit and cost of a testing strategy consisting of retesting all inpatients after the second day of the hospitalization compared to a single RT-PCR on admission.
Methods
A nonrandomized intervention was conducted in an accelerated stepped-wedge manner across a 5-hospital (1,029 beds) healthcare system in the 9-county Finger Lakes region of New York. In this system, 19% of rooms (31% of the beds) are semiprivate. Infection control measures at our facility were described previously Reference Lesho, Walsh and Gutowski6 and mirrored those at other hospitals. Reference Rhee, Baker and Vaidya7
Under the existing testing program (P1), all patients were tested upon admission for SARS-CoV-2 infection with nasopharyngeal (NP) swabs that undergo RT-PCR on either the cobas 6800 System (Roche Diagnostics, Indianapolis, IN), the BD SARS-CoV-2 reagents for BD MAX (Becton Dickinson, Franklin Lakes, NJ), or the Simplexa COVID-19 Direct kit (DiaSorin Molecular, Minneapolis, MN) as described previously. Reference Lesho, Reno and Newhart8
In this study, we evaluated a second testing strategy (P2) consisting of automatically retesting all inpatients after their second day of admission on the same platforms. An opt-out clinical decision support (CDS) algorithm was established that activated after 48 hours of admission when any provider opened the patient’s chart. This tool consisted of a prewritten order for an NP swab for the RT-PCR. The CDS continued to pose an alarm until it was acted upon by either clicking on a “sign order” button or by opting out and providing a reason (Supplementary Fig. 1 online). The P2 intervention began with a 3-month preimplementation or ‘control’ phase from October 2020 to February 2021 on several units at the main hospital. Over the next month, it was sequentially implemented across all hospitals in the system.
Hospital-onset (HO) COVID-19 was defined according to the CDCdefinition: a negative admission test followed by a positive test ≥7 days (probable HO) or ≥14 days later (definite HO). HO–COVID-19 rates were reported as the proportion of nosocomial cases per 1,000 SARS-CoV-2–negative patients. The cost per additional detection was calculated as cost per subject screened ($50) × the number needed to test. The number needed to test was the average number of individuals who must be tested under a given strategy to identify a single case of SARS-CoV-2, also calculated as the inverse of efficiency. 9 Efficiency was the proportion of positive cases identified among all individuals tested. 9 Using only the number needed to test would overestimate the impact of the new testing strategy because it does not account for the cases already detecting by the existing strategy. Reference Honorio, Benzon and Molloy10 Therefore, we calculated the number needed to benefit as previously described and summarized below. Reference Honorio, Benzon and Molloy10
First, because identification of newly positive cases would trigger immediate enhanced isolation precautions and theoretically thwart further transmission, we defined ‘successful outcome’ (or benefit) as the identification of a conversion from negative to positive in an asymptomatic inpatient that would otherwise not have been retested. Second, we denoted the proportion of successful outcomes using the established mode of diagnosis (a single admission PCR test) as P1, and the proportion of successful comes under the new strategy (test efficiency of repeated PCR test on day 2) as P2. The inverse of the difference in success rates or 1/(P2 − P1) equals the number needed to benefit from the change to the new testing strategy. Reference Honorio, Benzon and Molloy10 Linear regression was used to analyze associations between community metrics and the outcome variables (the number needed to benefit and cost per additional detection).
Results
Community-level indicators such as positivity, vaccination, and new hospitalization rates are presented in Table 1 and Figure 1. The 7-day rolling average transmission rate fluctuated between 0.6% to 8.6%, and the vaccination rates ranged from 0.2% to 63%. The SARS-CoV-2 α (alpha) and δ (delta) variant lineages predominated in the first and last 5 months of the study period, respectively.
Table 1. Healthcare System–Level COVID-19 Indicators
Note. NNB, the number needed to benefit; NNT, test the number needed to test; Cd, cost per additional detection. Sources for the table: https://www.flvaccinehub.com/regional-data and https://forward.ny.gov/early-warning-monitoring-dashboard.
a % positive tests, 7-day rolling average.
b New cases per 100,000 population, 7-day rolling average.
c New hospitalizations per 100,000 population 7 day rolling average.
d No. of times the best practice advisory fired for ordering a SARS-CoV-2 PCR on second day of hospital admission
e No. of signed orders for SARS-CoV-2 PCR tests on second day of hospital admission.
f Number of patients eligible/available for testing based on inpatient census.
g Expected testing efficiency = the proportion of SARS-CoV-2–positive cases identified out of all individuals tested (cases detected per test); also P2.
h NNT = 1/Efficiency.
i Yield = total number of cases under a given testing strategy (eligible population × efficiency).
j HO-COVID-19 = hospital-onset infections per 1,000 non–COVID-19 patient days; also P1.
k P1, case rate identified by existing methods = nosocomial infection rate used as surrogate for established strategy (a single admission PCR) as surrogate. P2, case rate using new testing strategy (a repeat PCR on day 2).
l NNB = 1/(P2 – P1).
m Cd, NNT × $50 in USD.
Fig. 1. Positivity and hospitalization rates versus cost and benefit. Note. % Positive, % positive tests 7 day rolling average; Hospital, new hospitalizations per 100k population 7 day rolling average; NNB, number needed to benefit; $ cost (×10) = cost in USD ×10.
Healthcare system–level COVID-19 indicators, including the number of electronic reminders, the number of signed orders, and detection rates, are presented in Table 1. The mean monthly number of times the electronic reminder was triggered in patients’ charts and resulted in an order was 20,202 (range, 2,887–21,844), or 9.4% of the time. The 3 most common reasons for a day-2 order not resulting were (1) not part of treatment team, (2) patient will be discharged within 24 hours, and (3) patient refused test. Also, >90% of the repeated PCR tests were performed between inpatient days 2 and 4. The mean turnaround time for PCR results was 16.5 hours.
The diagnostic and financial impacts of the new testing strategy are presented in Table 1 and Figure 1. Testing efficiency ranged from 0.002% to 0.064%. The number needed to test ranged from 16 to 588. Cost per additional detection ranged from $800 to $29,400, and the number needed to benefit ranged from 16 to 769. Of the 3 community-level indicators evaluated, only the number of new hospitalizations was associated (negatively) with the number needed to benefit and the cost per additional detection (P = .04; adjusted R 2, 0.35 and P = .03; adjusted R 2, 0.39, respectively).
Discussion
The benefit and cost of a repeated RT-PCR testing on or after the second day of admission compared to a single admission test fluctuated as community prevalences and vaccination rates changed. Our healthcare system that has a mean daily census of 1000, and a mean length of stay of 5 days. Thus, applying these testing strategy performance characteristics (efficiency, <1%–9%) to day 2 testing could detect <10–90 HO–COVID-19 cases per month that would have otherwise been missed. These direct costs may seem excessive, but they are offset by having prevented additional nosocomial COVID-19 hospital days, staff exposures with the corresponding workforce effects, and additional morbidity and mortality.
This study had several limitations. Linear regression was applied to the few community factors that were available at the time. Therefore, the insight provided is incomplete because a number of factors are related to the properties of the healthcare system, such as staffing levels, that could not be captured in that analysis. Second, the true number needed to benefit and cost per additional detection could be less than reported due to the large number of opt-out responses.
Despite these limitations, these results could help other system administrators decide when repeat testing of asymptomatic inpatients might be most cost-effective. A provisional threshold for such an approach could be to test all inpatients until the vaccine coverage or level of immunity in the community reaches 50%. After that, only retest the asymptomatic when the 7-day rolling average of the new hospitalization rate is ≥2 per 100,000 residents.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2022.157
Acknowledgments
Financial support
No financial support was provided relevant to this article.
Conflicts of interest
All authors report no conflicts of interest relevant to this article.
