key: cord-299465-c7ki3061 authors: Mallow, P. J.; Jones, M. title: When Second Best Might be the Best: Using Hospitalization Data to Monitor the Novel Coronavirus Pandemic date: 2020-05-18 journal: nan DOI: 10.1101/2020.05.11.20098475 sha: doc_id: 299465 cord_uid: c7ki3061 The novel coronavirus' high rate of asymptomatic transmission combined with a lack of testing kits call for a different approach to monitor its spread and severity. We proposed the use of hospitalizations and hospital utilization data to monitor the spread and severity. A proposed threshold of a declining 7-day moving average over a 14-day period, "7&14" was set to communicate when a wave of the novel coronavirus may have passed. The state of Ohio was chosen to illustrate this threshold. While not the ideal solution for monitoring the spread of the epidemic, the proposed approach is an easy to implement framework accounting for limitations of the data inherent in the current epidemic. Hospital administrators and policy makers may benefit from incorporating this approach into their decision making. Before government officials relax stay-at-home orders and hospitals resume elective procedures, decision-makers must accurately estimate the trend, severity, and prevalence of the novel coronavirus in a geographic region. Ideally, public health agencies would conduct active surveillance of infections in the general population (Sun, 2020; Nsubuga, 2006; Hashimoto, 2000) . The results from this first-best solution represent a coincident indicator of COVID-19's prevalence in a population. However, the fact that the novel coronavirus has a high rate of asymptomatic transmission hinders the usefulness of this approach (Gandhi, 2020) . Further hindering the disease surveillance is the limited number of novel coronavirus test kits as of April, 2020. Many states like Ohio prioritize the individuals who are eligible for testing (ODH, 2020a) . Ohio and many other states recommend that all individuals who exhibit symptoms should be tested. However, hospitalized individuals and healthcare workers are given first priority. Individuals in long-term care and first responders are given a lower priority, and individuals in the general population have the lowest priority. While this prioritization redirects resources to their most effective use, the number of positive cases represents a biased sample of the general population. This tradeoff suggests that the number of positive test cases in a population does not necessarily reflect the actual prevalence of COVID-19 nor the infection rate trend. As a lagging indicator, COVID-19 hospitalizations would normally be considered a second-best solution to measuring a trend in the infection rate. However, given the sample bias reflected in prioritized testing and asymptomatic transmission, we propose that COVID-19 hospitalizations combined with a capacity measure offer the best approach to measuring trends in COVID-19 infections. COVID-19 deaths . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 18, 2020. . present an even longer lag time than hospitalizations, and so they are not viewed as suitable of a measure. We chose the state of Ohio to illustrate our approach. The state of Ohio is one of several states that releases daily hospitalization data (ODH, 2020b). However they do not A 7-day moving average was calculated by adding the number of hospitalized COVID-19 patients over each seven-day window and dividing by the time period. The threshold for assessing the passing of a novel coronavirus wave was set at a declining 7day moving average over a 14-day period. The moving average period of 7-days was chosen to mitigate daily and weekend reporting effects and to be consistent with prior epidemiologic models Buckingham-Jeffery 2017; Rothman, 2008) . The length of time was chosen based on the current knowledge of the high end of the novel coronavirus incubation period (Lauer, 2020) . A further check included in the framework is stipulation that the 7-day moving average plus the historical occupancy level did not exceed the number . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 18, 2020. . https://doi.org/10. 1101 of staffed beds during this window. The research was conducted with de-identified publicly available data and is exempt from institutional review board review. All analysis was conducted in Microsoft Excel (Microsoft, Inc. Redmond, WA). The application of this approach to the state of Ohio found the first wave of the novel coronavirus passed on April 21, 2020. (Figure 1 ) During the period of January 7 to May 11, 2020, there were 4,413 COVID-19 hospitalizations. (Figure 2) Based on the median LOS of 4.9 days, these hospitalizations accounted for 21,624 hospital bed days ( Table 1 ). The peak bed utilization based on the 7-day moving average occurred on April 7 with 4,340 COVID-19. At the peak, COVID-19 patients occupied 13.5% of the total staffed beds in Ohio. Combined with the occupancy rate, approximately 73% (23,979) of staffed beds would have been in use on the peak day, remaining under capacity. The results were based on an imputed LOS and occupancy level for Ohio and were intended to illustrate this approach rather than informing decision making. A critical component of monitoring the novel coronavirus pandemic is availability of reliable and valid data. Preferably, we would have widespread testing data to inform our epidemiological models and provide a leading indicator of future demands of our healthcare system. However, the widespread asymptomatic community transmission and lack of testing kits prevents us from having a clear understanding of the novel coronavirus spread. In the absence of wide spread testing prior to or at the initial onset of the epidemic, hospitalizations and hospital utilization become the secondbest indicator to monitor the severity and progression of the novel coronavirus. Hospital utilization must be monitored to ensure that the hospitalization raw numbers do not become truncated. Once hospitals approach maximum capacity, the . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 18, 2020. Hospitalization Tracking Project, and our work expands upon the efforts of UM by incorporating hospital capacity and providing a means to assess the ongoing epidemic. Baker et al. 2017 proposed an approach for tracking influenza intensive care unit bed utilization to monitor severity of the influenza season (Baker, 2017) . However, many states are not reporting hospitalizations reliably or at all, let alone intensive care beds to provide usual information that can be aggregated. The proposed "7&14" framework has two key advantages. First, it can be implemented at the individual hospital level and aggregated by geographic regions. It requires three data inputs, hospitalizations, LOS, and occupancy. Second, one of the inherent benefits of using a moving average is to smooth . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted May 18, 2020 . . https://doi.org/10.1101 out random short-term fluctuations in daily hospitalizations. These two attributes combined creates an easy to understand dashboard at the chosen level of analysis to assess the severity and spread of the novel coronavirus epidemic. If or when additional healthcare system supply data becomes available (i.e. intensive care bed utilization), this approach can easily be expanded. The number COVID-19 occupied hospital beds is shown from January 7 to May 11, 2020 with the 7-day moving average. The number daily COVID-19 hospitalizations are shown from January 7 to May 11, 2020 with the 7-day moving average. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2020 . . https://doi.org/10.1101 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 18, 2020 . . https://doi.org/10.1101 Real-time surveillance of inluenza morbidty: tracking intensive care unit resource utilization. Annals of Correcting for day of the week and public holiday effects: improving a national daily syndromic surveillance service for detecting public health threats The Centers for Disesae Control and Prevention (CDC). National Center for Health Statistics. Occupancy Rates in Community Hospitals COVID-19 Capacity Predictor Aysmptomatic Transmission, the achilles' heel of current strategies to control COVID-19 Detection of epidemics in their early stage through infectious disease surveillance Calling All States to Report Standardized Information on COVID-19 Hospitalizations Severity of coronavirus respiratory tract infections in adults admitted to acute care in Toronto, Ontario The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application COVID-19 Presser Final Coronavirus (COVID-19). www.coronavirus.ohio.gov Evaluation of the Effectiveness of Surveillance and Containment Measures for the First 100 Patients with COVID-19 in Singapore Public Health Surveillance: A Tool for Targeting and Monitoring Interventions Presenting Characteristics, comorbidities, and outcomes among 5700 patients hospitalized with COVID-19 in the New York City Area Early epidemiological analysis of the coronavirus disease 2019 outbreak based on crowdsourced data: a population-level observational study Risk Factors Associated With Acute Respiratory Distress Syndrome and Death in Patients With Coronavirus Disease Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. The Lancet The authors would like to thank Michael Topmiller, PhD, Edmond A. Hooker, MD, DrPH, Dee Ellingwood, MS, and Jennifer Mallow, MBA for their thoughtful comments and suggestions. The authors did not receive any funding for this research nor have any conflicts of interest regarding this research to report. All authors contributed equally to all aspects of this reseach and manuscript.