What the Coronavirus Disease 2019 (COVID-19) Pandemic Has Reinforced: The Need for Accurate Data

Abstract

The coronavirus disease 2019 (COVID-19) pandemic has challenged the United States’ existing national public health and laboratory infrastructure. Central to fighting the pandemic is accurate and timely reporting of COVID-19 tests and case patients. Achieving this requires multiple elements, including the following: unambiguous data requests; clearly defined variables; consistent labeling and processing of laboratory samples; uniform reports; accurate and complete data collection, aggregation, and transmission; and trained personnel to identify quality issues. In the United States, COVID-19 interventions and prevention strategies have been decentralized and not harmonized across jurisdictions. Data complexity has resulted in reporting inaccuracies and delays at all levels of government, curtailing efforts to interrupt the spread of COVID-19. Multiple governmental and nongovernmental entities’ requests for COVID-19–related data from healthcare institutions are duplicative, ambiguous, underspecified, and lacking in granularity. Despite efforts to limit information gathering to the extent sufficient for informing policy and the public, the various requests commonly overlap, revealing poor communication between data requestors and lack of understanding of data availability, structure, and organization. Entities with legitimate data needs commonly require reports in various formats (faxes, pdfs, csv or xls files, e-mails, HL7 feeds, online portals) and have varying deadline intervals (daily, weekly, monthly). Currently, our regional healthcare system reports to 7 organizations (2 city, 1 county, 1 regional, 1 state, 2 national). Data requests include information about testing (6/7), hospital census (5/7), ventilators (3/7), and staffing/capacity (1/7). Although electronic case reporting from established electronic health records (EHRs) to public health agencies can be automated, uniform adoption has not been achieved due to high costs and complexity to vendors and institutions. The uncoordinated data collection, lack of sharing among entities, and inability of the EHR to automatically generate reports required us to create a reporting team composed of 11.85 full-time equivalents from the clinical laboratory, data warehouse, hospital operations, and hospital quality team. Not every institution has the required resources to create these complex reports. Data requests from public health and other authorities should be precise, actionable, and unambiguous. Early in the COVID-19 pandemic, confusion existed as to which date should be attributed to a positive test result (eg, date of first symptoms, specimen collection, specimen resulting, or specimen reporting) [1]. For a single specimen, these dates varied by weeks. Reporting requests lacking necessary specificity have led to confusion. The White House Coronavirus Taskforce requested daily COVID-19 test results from hospital-based laboratories [2]. The instructions (since reversed due to duplicate reporting) were “If all of your COVID-19 testing is sent out to private labs and performed by one of the commercial laboratories on the list below, you do not need to report using this spreadsheet.” Hence, patient samples collected at institutions not performing COVID-19 testing and sent to one of the specified commercial laboratories for processing resulted in 1 report to the government. However, patient samples collected at institutions not performing COVID-19 testing and sent to laboratories for processing that were not on the list may have been reported twice to the government—once by the hospital laboratory sending the specimen and once by the laboratory processing the specimen. Ambiguity was also seen in local (city, county, and state) data requests. Researchers voiced concerns about COVID-19 results being reported multiple times [3] and some requesting entities revealed that they could not address duplicated results. For example, the Texas Department of State Health Services is unable to de-duplicate the test numbers from private laboratories [4]. Because the extent and detection of these errors across entities are unknown, data quality remains in doubt. Fulfilling data requests by different entities is challenging for laboratories when associated specimen data are missing. Accredited laboratories require 2 forms of patient identifiers (eg, name and date of birth), collection date, and collection time to accept specimens. While severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) testing sites often provide the minimum data to meet laboratory acceptance criteria, they do not uniformly collect all data requested by the various entities. Specimens are frequently shipped from third parties (eg, nursing homes, physician’s offices, etc) to affiliated hospitals to be shipped to reference laboratories. When receiving specimens with missing reportable information, the laboratory must reject the specimen, delay specimen processing until information is received, or process the specimen with missing reportable information. In the latter case, the laboratory must choose to delay reporting or report incomplete information. For our medical center, collecting missing reportable information is a manual, unfunded process that has taken between 1 and 20 days. Once missing data are collected and reported, downstream practices can affect data accuracy depending on entities’ reporting policies. Using the date that a result was reported to an entity versus specimen collection date will skew the shape and the trajectory of the epidemic curve. Local and regional data-aggregation strategies contribute to reporting delays depending upon the automation, integration, and interoperability of public health departments’ electronic systems. Public health information technology infrastructure and capabilities vary widely. Resources allocated to local health departments depend on the investment and infrastructure provided by the federal government, state health departments, and categorical funding through initiatives and grants. Because categorical funding addresses specific public health goals, this funding is rarely used to develop or maintain flexible infrastructures to address future issues. Prior to COVID-19, health departments across the country received reportable infectious disease information through a combination of electronic laboratory reporting systems, faxes, and e-mails [5]. Health department staffing was usually sufficient to handle the volume of disparate reports. The volume generated by SARS-CoV-2 testing increased demands and stressed resources, creating reporting failures [6]. The confluence of the aforementioned problems, in addition to reporting time frames at an institutional level (Figure 1), resulted in inaccurate monitoring of the pandemic. For example, in Dallas County from 27 April 2020 to 27 May 2020, the daily incidence of COVID-19–positive cases artifactually decreased according to data reported by authorities on 2 June (Figure 2a). However, on 9 June, after additional backlogged data were added for the same time frame, the revised data show the daily incidence had been stable or increasing. The 2 different daily incidences for the same time period affected the calculated Rt, the virus’s transmission rate (Figure 2b). Using the data from 2 June, the Rt for 27 May was 0.73, a rate that could quench the spread of disease. However, using the revised data from 9 June, the Rt for 27 May was 1.03—a rate that could fuel the spread of disease. The data discrepancy affected SIR (Susceptible, Infectious, or Recovered) modeling for the estimated and 14-day projection of COVID-19 cases (Figure 2c). For the 2 June data, the projected COVID-19 case numbers in Dallas County were 1005; however, when corrected with backlogged data reported on 9 June, the projected COVID-19 cases were 2054. Data-reporting delays directly affected policymakers’ ability to interpret the trajectory of the epidemic, evaluate trends in cases and SARS-CoV-2 testing penetration and capacity, assess the role of interventions, and respond to the emerging data in a timely manner. Median and range of time (with possible sources of delay listed) for the stages from COVID-19 symptoms onset to data publication by requesting entity between 27 April 2020 and 27 May 2020 at a large academic medical center, which also serves as a reference laboratory. In 689 of 12 182 (5.7%) SARS-CoV-2 tests performed, patients were asked about date of symptom onset at the point of specimen collection with a median of 3 days (0–31 days) of symptoms onset prior to specimen collection. The median turnaround time from specimen collection to specimen result for all 12 182 specimens was 32 hours (<1–151 hours). The median courier time to the laboratory for all 12 182 specimens was 10 hours (<1–92 hours). The median turnaround time for the laboratory to report to a requesting entity for all 12 182 specimens was 29 hours (2–157 hours). It is unknown how much time requesting entities needed to publish the data they received. *If performed. Abbreviations: COVID-19, coronavirus disease 2019; hrs, hours; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2. A, Daily incidence of COVID-19–positive cases reported by the Dallas County Health Department based on specimen collection date from 27 April to 27 May. Gray bars represent the daily incidence for the time frame as reported on 2 June. Black bars represent additional cases (backlog) for the same time frame revised on 9 June 2020. B, Estimated COVID-19 Rt values from 27 April to 27 May using the incidence of positive cases reported on 2 June (gray line) and the revised Rt based on additional cases revised on 9 June (black line). C, SIR modeling for the estimated and 14-day projected trajectory of COVID-19–positive cases using the incidence of positive cases reported on 2 June (gray solid line, estimated; gray dashed line, projected) and when backlogged cases were added on 9 June (black solid line, estimated; black dashed line, projected). Abbreviations: COVID-19, coronavirus disease 2019; Est., estimated; Proj., projected; Rt, virus transmission rate; SIR, Susceptible, Infectious, or Recovered. To successfully address an emerging pathogenic infectious disease with pandemic potential and limited public health and therapeutic interventions, like COVID-19 or H1N1, the availability of accurate, reliable, and timely laboratory and diagnostic data is critical for researchers, policymakers, and the public. Informatics pitfalls experienced during the H1N1 pandemic [7], now re-experienced during the COVID-19 pandemic, have shown how data robustness complements surveillance strategies and assures that public health and government entities can identify and trace cases and contacts, assess disease trends and patterns, and evaluate the epidemiology and transmission dynamics and assess mitigation strategies. It is imperative that our national informatics strategy be improved upon. Local, regional, and national partners who share common data needs need to identify a finite list of clearly defined variables for a minimum reportable data set to share across entities. This will reduce the reporting burden felt by healthcare providers [6, 8]. Existing public health information technology infrastructure must assure high-quality data and enable real-time data sharing. Healthcare institutions cannot opt out of electronic reporting to public health agencies. Electronic health record vendors must increase interoperability among institutions and facilitate the basic public health reporting requirements to local, regional, and national public health entities without additional cost to an institution. Unique identifiers, like a universal healthcare identifier, should be implemented to minimize errors in providing needed demographic information and increase the integration of healthcare data with non-healthcare data. The restrictions that limit data sharing through the Health Insurance Portability and Accountability Act (HIPAA) should be revisited and criteria should be developed for the use of novel digital data, such as cellular data, with transparent public communication [3, 9–11]. Increased data-management practices must be developed to avoid preventable methodologic and interpretation errors, such as by standardizing analytic strategies [12, 13] and publishing best practices for data presentation through dashboards [14, 15]. These investments will facilitate responses to the COVID-19 pandemic and to the inevitable next infectious disease outbreak. Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Topics & Concepts