Development of a Standardized Data Collection Tool for Evaluation and Management of Coronavirus Disease 2019.

BACKGROUND: Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2 ) is responsible for coronavirus disease 2019 (COVID-19), a disease that had not been previously described and for which clinicians need to rapidly adapt their daily practice. The novelty of SARS-CoV-2 produced significant gaps in harmonization of definitions, data collection, and outcome reporting to identify patients who would benefit from potential interventions.
METHODS: We describe a multicenter collaboration to develop a comprehensive data collection tool for the evaluation and management of COVID-19 in hospitalized patients. The proposed tool was developed by a multidisciplinary working group of infectious disease physicians, intensivists, and infectious diseases/antimicrobial stewardship pharmacists. The working group regularly reviewed literature to select important patient characteristics, diagnostics, and outcomes for inclusion. The data collection tool consisted of spreadsheets developed to collect data from the electronic medical record and track the clinical course after treatments.
RESULTS: Data collection focused on demographics and exposure epidemiology, prior medical history and medications, signs and symptoms, diagnostic test results, interventions, clinical outcomes, and complications. During the pilot validation phase, there was <10% missing data for most domains and components. Team members noted improved efficiency and decision making by using the tool during interdisciplinary rounds.
CONCLUSIONS: We present the development of a COVID-19 data collection tool and propose its use to effectively assemble harmonized data of hospitalized individuals with COVID-19. This tool can be used by clinicians, researchers, and quality improvement healthcare teams. It has the potential to facilitate interdisciplinary rounds, provide comparisons across different hospitalized populations, and adapt to emerging challenges posed by the pandemic.

In December 2019, severe pneumonia cases of unknown cause emerged in Wuhan, Hubei, China, with clinical presentations greatly resembling viral pneumonia [1]. A novel enveloped ribonucleic acid beta coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was deemed responsible for coronavirus disease 2019 (COVID-19) and rapidly became a pandemic virus. As of June 10, 2020, there have been over 7 million cases confirmed of COVID-19 worldwide and approximately 2 million cases in the United States [2]. The World Health Organization (WHO)/China Joint Mission estimated approximately 6.1% of all cases develop critical disease, with an overall case fatality rate of 3.8% [3].

Several patient characteristics have emerged as risk factors for severe illness and mortality. These include older age, male gender, smoking, black race, underlying comorbidities (ie, cardiovascular disease, diabetes, hypertension, chronic lung disease, cancer, chronic kidney disease, and obesity), symptomatology, and imaging findings at presentation to the hospital [4-14]. In addition, certain laboratory findings have also been associated with poor outcomes (ie, lymphopenia, high neutrophil-to-lymphocyte ratio, elevated transaminases, elevated lactate dehydrogenase [LDH], inflammatory biomarkers [C-reactive protein {CRP}, ferritin, and D-dimer]), red blood cell distribution width (RDW), troponin, as well as indicators of acute kidney injury [5, 15–23]. However, these biomarkers require further validation to guide clinical practice in a novel pandemic [3, 15].

Several agents have been used to treat patients during the current pandemic, based on in vitro antiviral/anti-inflammatory activity or experience in other illnesses [24-26]. Some are Food and Drug Administration (FDA)-approved products being used off-label, and others are investigational drugs available from the manufacturer via compassionate use programs or as part of ongoing clinical trials. As of June 10, 2020, no drugs have been licensed by the FDA for COVID-19, although remdesivir, an antiviral with potentially broad activity, has received FDA emergency use authorization for use in COVID-19 [25, 27]. However, results from the supporting clinical trials have only indicated efficacy in reducing illness duration and have not shown a statistically significant survival benefit [28, 29]. Limited literature [30] and anecdotal experience [31] suggest that timing of interventions and appropriate patient selection are important considerations. On April 21, 2020, the National Institutes of Health published national guidelines recommending that the use of therapeutic agents for the management of COVID-19 should be done in the context of a clinical trial [25]. Because not all clinicians have access to clinical trials and knowledge evolves very rapidly, it is critical to collect systematic data on presentation, management, and clinical outcomes to provide feedback to frontline clinicians and to improve patient safety and quality of care [26].

The large disease burden attributed to an emerging pathogen has led to a strong focus by the media, public health organizations, and the research community leading to an explosion of rapidly published scientific manuscripts of limited long-term analysis. Despite this, knowledge gaps remain, and there is a need for harmonized definitions, data collection, and outcome reporting across many centers to identify those most likely to benefit from certain therapies and those who are most likely to have poor outcomes [26].

In the setting of the considerable unknowns faced daily by our clinical teams during the pandemic, we developed a system to study the presentation and outcomes of patients seen across our diverse practice sites in Miami, Florida. In this study, we describe the development and pilot validation of a comprehensive data collection tool that was rapidly adopted in our hospital settings. The use of such a tool can harmonize systematic data collection within and between clinical sites to help guide clinical practice and research during this pandemic.

METHODS

Study Setting

This study was conducted in 3 large academic centers in Miami, Florida. As of June 10, 2020, the counties served by the participating centers (Broward and Miami-Dade) reported 27 625 cases of COVID-19 [32]. Multiple factors could potentially contribute to poor outcomes in our setting: older average age, significant volume of visiting travelers, and high proportion of those who are economically disadvantaged or lack health insurance [33, 34]. The University of Miami Hospital, Miami Veterans Affairs Medical Center (VAMC), and Jackson Health System hospitals serve a large and diverse patient population. This includes a high volume of specific patient populations for which outcomes are not well described in COVID-19 literature: veterans, persons with human immunodeficiency virus (HIV), malignancy, or other immunocompromising conditions.

Development of the Data Collection Tool

The proposed tool was a set of spreadsheets developed by a multidisciplinary team of infectious disease physicians, intensivists, and antimicrobial stewardship pharmacists (ASP). The development phase was conducted at the Miami VAMC and Jackson Memorial Hospital, and the pilot validation phase was conducted at the University of Miami Hospital and all Jackson Memorial Hospital, under an IRB-approved protocol (protocol number 20200424). The team members formed a working group that met via telephone or video conferences at least biweekly during the early phase of the COVID-19 pandemic arriving in Miami, Florida in early March 2020. Based on available literature and Centers for Disease Control and Prevention (CDC) guidance [35], we determined a set of sociodemographic and comorbidity risk factors, clinical signs and symptoms, and diagnostic tests that could be obtained from the electronic medical records (EMRs) (Table 1) or patient interview. Laboratory tests were chosen based on literature correlating elevation of infection or inflammatory biomarkers (eg, CRP, ferritin, LDH, D-Dimer, troponin, neutrophil-lymphocyte ratio, RDW, procalcitonin) with greater likelihood of severe/critical COVID-19 [5, 16–24], or to calculate intensive care unit (ICU) physiologic severity scores [36]. Antiviral, anti-inflammatory, and supportive therapies included in the data collection tool were selected based on available literature and upon review of hospital protocols at our institution and other settings [20, 24, 26].

Table 1.

Baseline Characteristics of the Pilot Validation Cohort

Characteristics	Total Cohort (n = 200)
Hospital
Jackson Memorial Hospital	113 (56.5%)
University of Miami Hospital	87 (43.5%)
Age (median, IQR)	63 (49–73)
Gender
Male	118 (59.0%)
Female	82 (41.0%)
Ethnicity
Hispanic	99 (49.5%)
Black	57 (28.5%)
White/Caucasian	25 (12.5%)
Asian/Pacific Islander	9 (4.5%)
Middle East/North Africa/Central Asia	2 (1.0%)
Other/not specified	6 (3.0%
Missing	2 (1.0%)
Comorbidities
Chronic ventilator dependence	4 (2.0%)
Asthma or COPD	33 (16.5%)
Obstructive Sleep Apnea	5 (2.5%)
Congestive Heart Failurei	18 (9.0%)
Hypertension	115 (57.5%)
Diabetes Mellitus	64 (32.0%)
Coronary artery disease (occlusive)	19 (9.5%)
End Stage Renal Disease	13 (6.5%)
Chronic Kidney Disease	24 (12.0%)
Cirrhosis	1 (0.5%)
Solid Organ Transplant	4 (2.0%)
HIV	10 (5.0%)
Malignancyii	19 (9.5%)
Smoking (recent)	14 (7.0%)
Alcohol use (recent)	19 (9.5%)
Obesity (BMI > 30)	88 (44.0%)
Pregnancy	1 (0.5%)
Exposure Epidemiology
Communityc	145 (72.5%)
Facilityd	29 (14.5)
Ship	17 (8.5%)
Foreign Travel	14 (7.0%)
Health Care Workere	7 (3.5%)
Symptom Duration (median, IQR)	4 (2–7)
Symptoms
Fever	145 (72.5%)
Cough	148 (74.0%)
Dyspnea	139 (69.5%)
Signsf
Fever	134 (67.0%)
Tachypnea	73 (36.5%)
Hypoxia	101 (50.5%)
Tachycardia	103 (51.5%)
Hypotension	11 (5.5%)
Disease Severityg
Mild	10 (5.0%)
Moderate	88 (44.0%)
Severe	83 (41.5%)
Critical	11 (5.5%)
Critical with MODS	8 (4.0%)
WHO Ordinal Severity Scale
3: Hospitalized, no oxygen	85 (42.5%)
4: Oxygen by nasal prongs or face mask	86 (43.0%)
5: Noninvasive ventilation/high-flow oxygen	11 (5.5%)
6: Intubation and Mechanical Ventilation	11 (5.5%)
7: Ventilation and additional support (RRT/ECMO/vasopressors)	8 (4.0%)

Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; ECMO, extracorporeal membrane oxygenation; HIV, human immunodeficiency virus; IQR, interquartile range; MODS, multiple organ dysfunction syndrome; RRT, renal replacement therapy; WHO, World Health Organization.

a Five patients had congestive heart failure with reduced ejection fraction (EF ≤ 40%).

bMalignancy: 11 solid, 9 hematologic, 1 patient with both; 4 on active chemotherapy.

cIf none of the other exposure locations was noted, the patient was labeled community exposure, by default.

dFacility: residence in nursing home, assisted living, long term care facility, prison, homeless shelter. Note: visiting one of the previously mentioned places did not qualify and there was a separate category for ship.

eHealthcare worker was someone performing their job duties in patient care areas of the hospital. patients who had visited a hospital did not qualify.

fDefinition of abnormal vital signs: fever - temperature ≥37.5°C; tachypnea - respiration rate ≥30/minute; hypoxia - SaO2 on room air ≤93% or PaO2; FiO2 ratio - <300 (mechanical or noninvasive ventilation); tachycardia - heart rate >90 beats per minute; hypotension - shock with vasopressor use.

gDefinition of disease severity: mild - no signs, symptoms, or imaging consistent with pneumonia; moderate - signs and/or symptoms consistent with pneumonia and compatible imaging; severe - tachypnea or hypoxia (as defined above); critical - respiratory failure requiring mechanical ventilation but no other organ dysfunction requiring support; critical with MODS - respiratory failure requiring mechanical ventilation and other organ failure requiring support (renal replacement therapy, vasopressors, ECMO). Liver injury and need for transfusion not included.

i,iiThese comorbidities have been associated with poor outcomes from COVID-19.

Based on the characteristics of the study (retrospective review of data obtained for purposes of clinical care and quality improvement), a waiver of informed consent was obtained from the Institutional Review Board (IRB). Health Insurance Portability and Accountability Act (HIPAA)-compliant spreadsheets were developed to track data trends and facilitate biweekly discussions. All elements included in the final data collection tool were discussed during the working group meetings, during international forums with clinicians from countries who encountered the pandemic early (Spain, Italy, China), and during interdisciplinary rounds with clinical providers—because this strategy has been shown to improve communication and foster agreement on the plan of care [37]. Factors were updated based on frequent forums and literature review until all members of the working group came to a mutual agreement. In addition, the working group decided to track important dates: illness onset, hospital admission, as well as beginning and end dates of ICU care, mechanical ventilation and other life-support treatments, and antiviral/anti-inflammatory therapies. Working group members created a note template to evaluate those with confirmed or suspected COVID-19 using the above elements, and this was distributed to the infectious disease teams for use at their discretion.

The working group decided on standardized definitions of terms and format of data reporting to facilitate data analysis. Binary (ie, 1 = yes, 0 = no) format was used unless additional complexity existed. In these cases, either a limited set of result categories was established, or free-text format was used. Care was taken to distinguish data that could not be obtained (left blank) versus normal/negative (recorded as “0”). If date of illness onset could not be ascertained, it was considered missing data (left blank). The data tool noted patients with chronic respiratory failure who were dependent on mechanical ventilation (yes/no) or supplemental oxygen at rest (yes/no).

Each vital sign was noted as normal (“0”) or abnormal (“1”) using the most severe value on day of admission: fever - temperature ≥37.5°C; tachypnea - respiration rate ≥30/minute; hypoxia - SaO2 on room air ≤93% or PaO2/FiO2 ratio <300 (mechanical or noninvasive ventilation); tachycardia - heart rate >90 beats per minute; hypotension - shock with vasopressor use.

The definitions for abnormal vital signs and COVID severity stage were based on early published natural history studies [3, 6, 16, 20], which were recommended for disease staging in later guidelines [26]. However, we distinguished 2 different stages within the previously defined “critical” stage: those with only respiratory failure were defined as “critical”, and those with respiratory failure and failure of other organ systems requiring support were defined as “critical with MODS” (multiple organ dysfunction syndrome)—mild - no signs, symptoms, or imaging consistent with pneumonia; moderate - signs and/or symptoms consistent with pneumonia and compatible imaging; severe - tachypnea or hypoxia (as defined above); critical - respiratory failure requiring mechanical ventilation, but no other organ dysfunction requiring support; critical with MODS - respiratory failure requiring mechanical ventilation and other organ failure requiring support (renal replacement therapy, vasopressors, extracorporeal membrane oxygenation [ECMO]). Liver injury and need for transfusion were not included.

This modification was made based on our treatment protocol and our experience that MODS confers different prognosis and responsiveness to interventions. The WHO ordinal scale for clinical improvement [38] was used to track clinical course and need for respiratory support.

In addition, in categories in which normal cutoffs were different than typical clinical practice (fever - ≥37.5°C vs ≥38.2°C typically used for nonneutropenic patients), we included a free text field to record the actual value, to facilitate later categorization or quantitative analysis. Time-updated vital signs were reported as daily ranges. Time-updated laboratory tests used first morning arterial blood gas (ABG) results and the results from other laboratory tests drawn closest in time to the ABG results. For radiology results, abnormal features were new findings compared with baseline imaging, and if no prior comparison was available, the findings were presumed to be new.

The order of specific variables and groups were modified based on feedback from regular meetings of the working group. Final groupings and order of variables represented the most user-friendly format balancing the ability to extract data within our EMR systems during real-time care, to allow easy quality control and to further analyze data. Specifically, organization of the data collection tool aimed to group data that would be extracted at specific time points during hospitalization.

Pilot Validation of the Data Collection Tool

During the pilot validation phase, the data collection tool was used to track data obtained from patients being evaluated for confirmed or suspected COVID-19 during 31 consecutive days (March 23–April 23, 2020) at the participating sites. During the development phase, CDC criteria for persons under investigation included either (1) compatible clinical syndrome (fever and either cough or dyspnea) and epidemiologic risk factor (travel to a high-transmission country or close contact with confirmed case within 14 days of symptom onset) or (2) severe febrile lower respiratory illness requiring hospitalization without an alternative diagnosis [39]. However, guidance was updated on March 4, 2020 to suggest testing be considered at the discretion of the treating clinician. Thus, patients studied during the pilot validation phase presented with one of the following: viral syndrome, fever, acute dyspnea, cough, hypoxia, or abnormal chest imaging. The tool was applied regardless of epidemiologic risk factors.

The pilot validation phase was performed in the context of normal duties by ASP and the infectious disease team (fellows and/or staff physicians), using a convenience sample size. Data were entered and results were shared with stakeholders (infectious disease team, intensivists, and ASP team) during interdisciplinary rounds. The clinical data collected during the pilot validation phase was input into secure clinical data management system at each site: SharePoint or RedCap. The percentage of missing data was calculated by the number of patients in whom the data was missing by the total participants included in the pilot validation process. Missing data for diagnostic testing was calculated based on any data obtained before beginning any antiviral or anti-inflammatory therapy. Missing data for interventions and clinical outcomes were calculated based on data available at date of discharge or death. Missing data for other outcomes were obtained after evaluation by the infectious disease team.

RESULTS

The data collection tool was pilot tested with 200 patients admitted to 1 of the 2 participating centers over a 31-day period, and patient characteristics are described in Table 1. The final data collection tool contained 2 main components with worksheets for infectious diseases and ASP teams (Supplemental Appendix; summarized in Table 2) and additional worksheets for definitions of abbreviations and terms, laboratory reference ranges, and data dictionary. The infectious diseases fellows and ASP found that the required data could be easily and efficiently collected, and all working group members were satisfied with its functionality. Stakeholders indicated that use of the tool improved decision making during interdisciplinary rounds by allowing rapid review of patient’s clinical presentation, treatment, and clinical course. The major data domains, specific data, and results from data collection during the pilot validation phase is specified in Table 2.

Table 2.

Data Collection Tool Domains, Timing, Variables, and Percentage of Missing Data From the Electronic Medical Record (EMR)

Domain	Timing and Source	Variables	Missing (%) N = 200
Demographics and Exposure Epidemiology	Timing: Baseline	Age, Gender, Race/Ethnicity Exposure	1.0% Race/ethnicity
	Primary Source: EMR	Epidemiology	7.0% Foreign Travel
	• Provider notes of patient-reported data	• Community	No missing data for age, gender, or other epidemiologic exposure groups
	Other source:	• Foreign Travel
	• Interdisciplinary rounds report	• Health Care Worker
		• Facility: nursing home/long- term care, assisted living, jail/prison
		• Cruise Ship
Prior Medical History and Medications	Timing: Baseline	Selected Home Medications	0%–2.0% in all comorbidities
	Primary Source: EMR	• Angiotensin Converting
	• Provider notes and medication reconciliation	Enzyme inhibitors
		• Angiotensin receptor blockers
	Other source:	• Nonsteroidal anti-inflammatory drugs
	• Interdisciplinary rounds report	• Corticosteroids
		Comorbidities/Substance Use
		• Smoking (current)
		• Alcohol (recent)
		Special Populations (eg)
		• Pregnancya
		• HIV infectionb
		• Malignancyc
Signs and Symptoms of Illness	Timing:	Vital signs at admission	6.0% Date of illness onset
	• Symptoms: Baseline	Date of onset of first symptom	No missing data for
	• Vital Signs: Baseline and time- updated (daily range, hospital day 1, 2, 3, etc)	All symptoms up to admission	• Most common symptoms (fever, cough, dyspnea)
	Primary Source: EMR		• Vital signs at admission <20% missing data on other nonspecific viral symptoms at admission (eg, sore throat, headache, nausea, diarrhea)
	• Symptoms: Provider notes documenting illness onset date per patient report		>20% missing data for uncommon symptomsd
	• Vital Signs: EMR flowsheets
	Other source:
	• Interdisciplinary rounds report
Diagnostic Test Results	Timing:	Arterial Blood Gas	97.5% Triglycerides
	• Baseline	Blood Count	56.6% Troponin
		Chemistry Indices	54.0% Procalcitonin
	• Time-updated: Hospital day 1,2,3 etc.e	C-Reactive Protein	46.0% Respiratory viral PCR panel
		Lactate Dehydrogenase	87.5% CT scan
	• End of follow upf	Ferritin
	Source: EMR	D-Dimer	All other tests: <10% missing datah
	• Laboratory	Interleukin-6
	• Microbiology	Triglycerides
	• Radiology Reports	Albumin
		Procalcitonin
		Troponin
		Respiratory Virus PCR Panelg
		SARS-CoV-2 PCR
		Chest x-ray
		Computerized Tomography
		Cultures (Blood, respiratory, other sterile sites)
Interventions	Timing	Antiviral Treatments	No missing data
Specific Treatments Supportive Care Modalities	•Baseline	• Hydroxychloroquine
	• Time-updated: Hospital day 1, 2, 3 etc	• Lopinavir-Ritonavir
	• End of follow up	• Azithromycin
	Source: EMR	• Oseltamivir
	• Provider Notes	• Remdesivir
	• Medication Administration Record
	•Flowsheets	• Tacrolimus (continuation of home medicine)
		Anti-inflammatory/other adjunctive treatments
		• Tocilizumab
		• Corticosteroids
		• Intravenous Immunoglobulin
		• Anticoagulation
		• Other Anti-inflammatory agents
		Other therapies: eg,
		• Pulmonary Vasodilators
		• Inhaled Nitric Oxide
		• Stem Cell Therapy
		• Convalescent Plasma
		Supportive care i
		• Supplemental Oxygen Delivery
		• Noninvasive Ventilation
		• Mechanical Ventilation
		• Extra-Corporal Membrane Oxygenation
		• Renal Replacement Therapy
		• Vasopressor Use
		WHO ordinal scale for clinical improvement j
Clinical outcomes and complications	Timing	Days requiring various life support modalities	No missing data
	•Baseline	• Intensive Care Unit admission
	• Time-updated: Hospital day 1, 2, 3, etc	• Intubation and Mechanical ventilation
	• End of follow-up	• Intensive Care Unit length of stay
	Source: EMR	• Hospital length of stay
	• Flowsheets: dates of intubation/ extubation, pressor/ECMO/RRT	Mortality (in-hospital)
	• Provider Notes: Complications, dates of ICU admission/ discharge, dates of hospital admission/ discharge	• 14 days
		• 30 days
		Adverse events

Abbreviations: CT, computed tomography; EMR, electronic medical records; ICU, intensive care unit; HIV, human immunodeficiency virus; PCR, polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2 virus; WHO, World Health Organization.

aPregnancy status and gestational age. All females of child-bearing age had a point-of-care pregnancy test unless they reported menopause.

bDuration of infection, viral load nadir CD4, recent CD4 (before COVID-19 illness) and adherence to antiretroviral therapy.

cSpecific type (organ/primary cell type), treatment (yes/no), specific treatment (free text therapy name).

dMany of which were described during or after the development phase of our study (eg, dysguesia, anosmia).

eUsed first morning arterial blood gas (ABG) results and laboratory tests drawn closest in time to the ABG results.

fEnd of follow-up for diagnostic data is defined as last data available before death or discharge from the hospital.

gAssay: BioFire for Jackson Memorial Hospital, GenMark for University of Miami Hospital.

hSignificant missing data for interleukin-6 and hepatitis B serologies since these were only drawn in patients considered for Tocilizumab.

iIncluding respiratory support parameters: oxygen liters per minute, PEEP, FiO2, respiratory rate, tidal volume, prone positioning, or paralytic use.

jIncorporates functional capacity and level of oxygenation and ventilation support needed to stabilize patient in the emergency department or initial level of care.

In both the development and pilot validation phase, scenarios arose that challenged our current framework of data reporting. When this occurred, a consensus decision was made within the working group to minimize reporting bias and maximize harmonization. After this, relevant prior data underwent review and reclassification, if necessary, to ensure consistency.

After data extraction, we assessed the frequency of missing data (Table 2). Among demographic data, we had no missing data for gender or age and only 1.0% without race/ethnicity recorded. We had no missing data for epidemiologic exposures except 7.0% for history of foreign travel. All females of child-bearing age had a point-of-care pregnancy test. Based on documentation available in the EMR of the participating institutions, there was <10% missing data on comorbidities of interest. Date of illness onset had 6.0% missing data, but presence of symptoms at admission was variably reported. There was no missing data for the most common symptoms (fever, cough, dyspnea), <20% missing data for other symptoms commonly associated with viral infections (eg, sore throat, headache, nausea, diarrhea), and a greater percentage of missing data for uncommon symptoms (eg, dysgeusia, anosmia)—which were often described during or after the development phase of our study. Data availability on most objective measures was excellent, and there was <10% missing data except for a few specific tests: respiratory viral polymerase chain reaction panel (46.0%: limited supply), computed tomography scan (87.5%: limited use based on infection control concerns), and certain laboratory tests such as triglycerides (97.5%), procalcitonin (54.0%), and troponin (56.6%). These laboratory tests were often not sent unless severe disease and potential association with severe COVID-19 was described during or after our development phase [19, 22, 40].There was also significant missing data for interleukin-6 level and hepatitis B serologies, but these were obtained only in patients who were candidates for tocilizumab. There was no missing data regarding use of specific interventions, complications, or clinical outcomes.

DISCUSSION

We present the development of a COVID-19 data collection tool, which efficiently tracked comprehensive clinical data and provided up-to-date information to guide decision making in an era of rapidly evolving data. We propose its use to effectively collect harmonized data of individuals admitted to hospitals with confirmed or suspected SARS-CoV-2 infection [39]. In the context of a pandemic, we recommend selecting patients for evaluation patients based on clinical presentation, as described above, without requiring specific epidemiologic risk factors. Given the gaps and rapidly evolving knowledge in understanding risk factors, clinical presentation, treatment options, and clinical outcomes, coordinated data collection tools are needed to maximize utility of clinical data and to improve patient safety and quality of care.

Timely risk stratification in patients infected with SARS-CoV-2 and patient selection for interventions has proved challenging. Some groups have reported favorable performance of risk prediction tools using early clinical laboratory tests and patient characteristics [17, 18]. However, to date, there is no widely validated diagnostic that can reliably identify those at high risk for respiratory failure requiring mechanical ventilation or death. Additional challenges of expanding the COVID-19 knowledge base include differences in biomarker assays, protocols specifying the roles of certain interventions that vary by institution and over time, and variable definitions for abnormal vital signs and disease staging. Infectious Diseases Society of America (IDSA) guidelines [26] specifically emphasized the need to report relevant objective clinical outcomes and use standardized disease staging definitions that use readily obtainable clinical criteria, like the WHO/China Joint Mission [3]. Data collection tools have been reported for emerging infectious diseases and other conditions, and they have had a positive impact on patient safety, quality improvement, research, and clinical care [41, 42].

The COVID-19 data collection tool can be used by clinicians, researchers, and quality control staff, and adapted to their own setting during this pandemic. It has the potential to enable comparisons across different hospitalized populations in the future and to be rapidly adapted to the emerging challenges posed by the pandemic. In addition, it can be used easily and safely in settings with limited technology by using spreadsheets within secure data collection systems (RedCap, SharePoint).

A few general observations in our study are worthy of discussion. The performance of the data tool during collection of data in all the major domains listed was enhanced by using the EMR, but use of the tool was not dependent on review of EMR documentation. Daily interdisciplinary rounds featured presentation of new patients by the primary team who personally evaluated and interviewed the patient. In addition, availability of data on diagnostic testing was dependent on the practice pattern of the primary medical team and infectious disease team. For example, the likelihood of missing data for diagnostic testing obtained before antiviral/anti-inflammatory therapy was influenced by the adherence to recommendations for baseline laboratory testing by the primary medical team. In other cases, detailed information regarding uncommon or rare symptoms was not gathered or documented. Often, there was low suspicion for COVID-19, and this information could not be obtained at the time of infectious disease evaluation due to patient condition. All of these examples highlight the need for systematic data collection, and we recommend creating a site-specific note template to prompt clinicians to obtain data relevant for care of those with confirmed or suspected COVID-19.

There are a few opportunities for improvement that should be noted. First, the 2 centers in the pilot validation phase had different EMR systems, but we did not study differences in missing data or time needed to extract data between the 2 sites. Such comparison would have been confounded by differences in data collectors and is inconsistent with our proposed recommendation for others to adapt and optimize this tool locally. Thus, we cannot determine which EMR system had organization and functionality best suited to our tool. However, some EMR have downloadable data functionality, and data tool organization should be optimized to receive output data in this manner, if available. Second, the scientific community should work toward consensus on how to report dynamic changes of certain biomarkers. This is important when data from multiple sites are combined for reporting because reference ranges and assays for laboratory tests may be different and may require normalizing procedures during analysis (eg, converting to percentage or fold change from baseline or pretreatment value). Third, we encountered few transplanted patients in the study. We believe a separate data collection tool should be developed and validated for recipients of solid organ and hematopoietic stem cell transplants. Likewise, there are other patient populations not well represented in our study that require further validation with this tool: chronic respiratory failure, obstructive sleep apnea, cirrhosis, HIV, and pregnancy. Fourth, there were no patients who had clinical documentation from recent care at other medical centers, but outside data would need to be accounted for in further improvements to the tool. In addition, insurance status and type are important variables not built into our data tool that should be included on further iterations. Uninsured patients who are hospitalized receive fewer services and are more likely to experience in-hospital mortality than insured patients [43]. Finally, this data tool only captured basic baseline data on patients with chronic respiratory failure; need for mechanical ventilation and need for supplemental oxygenation. More detailed data should be collected on such patients including support settings, level of exertion at assessment, and other factors—based on input from critical care specialists.

We believe these limitations can be overcome. Each center should include variables relevant to management of COVID-19 patients within local protocols and based on the monitoring and diagnostic capabilities of the medical center. This tool should then be adapted to the sophistication and organization of the EMR platform, validated locally, and improved based on feedback from members of interdisciplinary care teams.

At this time, no intervention has been shown to be effective in a randomized controlled trial for the most relevant clinical endpoints: mortality, rate of progression to acute respiratory distress syndrome, and need for mechanical ventilation [25, 26, 28, 29]. Thus, current guidelines do not recommend any specific therapies and recommend use of available products only within the context of a clinical trial [25, 26]. Until more trial results are published, clinicians and hospital systems are faced with treating patients who are currently ill, leading to difficult decisions on the role of multiple unproven therapies and on which clinical trials to pursue.

CONCLUSIONS

We developed this data collection tool to track patient data relevant to the initial evaluation and ongoing management of hospitalized patients with COVID-19 infection. It can be adapted and applied to facilitate research involving cohort studies investigating patient characteristics associated with poor outcomes, specific therapies associated with clinical improvement, and appropriate timing or patient selection for therapy [44]. Thus, data collected through this tool can inform future clinical practice, improve patient safety and quality of care, and provide feedback for the design and conduct of trials evaluating the efficacy of the existing and novel therapies for COVID-19.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Click here for additional data file.