Introduction

The coronavirus disease 2019 (COVID-19) pandemic continues to escalate worldwide, with over 18 million people infected and over 696,000 deaths as of August, 2020. While some geographic regions experience peak surges, followed by consistent decreases in the number of patients testing positive for Severe Acute Respiratory Syndrome coronavirus 2 (SARS-CoV-2), other regions remain inundated with high volumes of infections, hospitalizations, and deaths [1]. Scientific reports from various regions are prolific, published with the intent of rapidly sharing knowledge regarding prevalence, treatment, and outcomes of infected individuals. Numerous reports of neurological symptoms and manifestations seemingly associated with SARS-CoV-2 infection continue to emerge, with spectrum of neurological manifestations ranging from mild (anosmia, ageusia, and headache) to severe (coma, seizures, hypoxic brain injury encephalopathy, stroke, cerebral hemorrhage, posterior reversible encephalopathy, and Guillain-Barré syndrome) [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]. Similarly, prevalence estimates across these reports range from 3.5 to 84% [5,6,7, 12, 16], with many studies not controlling for confounding factors or differentiating between levels of care or severity of systemic infection. Given the continued escalation of the global COVID-19 pandemic and the rapidity of emerging information from regional reports on neurological manifestations of COVID-19, it has become increasingly difficult to ascertain accurate estimates on prevalence and severity of neurological dysfunctions. As a result, critical data on causality versus recrudescence of preexisting neurological conditions versus coincident findings of COVID-19 are lacking. Furthermore, recent attempts at pooled analysis are proven unsuccessful due to substantial heterogeneity across populations, sites, and data components [6].

A global, systematic, and collaborative approach to knowledge development in this pandemic is critical [18]. Fragmented reports create ambiguity about case definitions, clinical findings, and empiric treatments, potentially leading to administration of ineffective or harmful regimens and public alarm due to inaccurate conclusions about causation in the absence of sound scientific methodology [14, 19]. Data gathering in a global pandemic must include diversity in age, sex, race, ethnicity, geographic region, and resource settings. Researchers must utilize a collaborative and pragmatic approach feasible for members of the healthcare team, as many research teams do not have access to hospitalized patients during a pandemic crisis. To this end, we rapidly established the Global Consortium Study of Neurological Dysfunction in COVID-19 (GCS-NeuroCOVID) and launched a Tier 1 basic study to gather essential data on acute neurological manifestations of COVID-19 captured by frontline providers in the midst of this pandemic [20]. The objective of this study is to pragmatically and systematically determine the major phenotypes of neurological symptoms and estimate their global prevalence among hospitalized COVID-19 patients. All acute care hospitals caring for COVID-19 patients are eligible to register as sites through a central Web portal [21]. The consortium employs an accelerated process to develop consensus common data element (CDE) and definitions as previously described. Whenever possible, we utilized any consensus CDEs already developed through the National Institute of Health (NIH) CDE process [22, 23].

As COVID-19 infection is associated with age- and race-dependent disease prevalence, severity, symptoms, and outcomes, the GCS-NeuroCOVID consortium study uses CDEs specifically designed for adult and pediatric populations that include race and ethnicity designations. Additionally, the GCS-NeuroCOVID consortium collaborated closely with partners in Latin America to harmonize data elements and definitions, resulting in the development of Spanish and Portuguese versions of case report forms (CRFs).

Methods

The aims, design, and methodology of the GCS-NeuroCOVID research program are previously reported [20]. In brief, the study employs a nimble, tiered design to determine the prevalence, severity, predictors, and outcomes of neurological manifestations among hospitalized patients across the age span with COVID-19 infection. A key element is CRFs that are readily available in multiple languages and applicable across sites and settings. International guidelines for good clinical practices in research define CRFs as “…specialized documents designed to record all protocol-required information to be reported on each subject” [24]. As such, CRFs are a critical component to ensure internal validity of any study, particularly for multisite investigations.

GCS-NeuroCOVID Consortium CRF: Adult

Guiding principles underlying development of the CRF for the GCS-NeuroCOVID investigation center on feasibility and harmonization. Feasibility refers to the ability of frontline clinicians to record observations of neurological manifestations among hospitalized COVID-19 patients with minimal burden, care delivery disruption, and no additional risk of exposure to providers. Given the sustained surge of cases worldwide, lack of adequate personal protective equipment (PPE), and limited resources for clinical research particularly during a pandemic, it is paramount to consider practical aspects to ensure feasibility of study initiation and completion. The acquisition of these data elements must not involve increased SARS-CoV-2 exposure risk for frontline clinical providers or require additional PPE use for research purposes alone. Data elements must have clear, simple, and specific definitions to optimize collection of high-quality data with minimal errors or missing values. Data definitions must be easy to use and not require specialized expertise in neurological examinations for accurate data acquisition. Data elements should be commonly available regardless of resource levels of the acute care hospital to encourage participation and broad inclusion of all regions and hospitals that care for acute COVID-19 patients. Data elements need to capture overall severity of illness, because severe single- or multi-organ failures can result in secondary neurological sequelae, regardless of the original etiology of organ failure.

Creation of these initial highly pragmatic data elements provides the foundation for the Tier 1 study of GCS-NeuroCOVID. The aim of the Tier 1 study is to systematically record data to identify the major phenotypes and prevalence of neurological symptoms among hospitalized patients with COVID-19 infection. As such, it is designed to be a minimal dataset that captures the most high-value data elements that are feasible to obtain in the challenging clinical environment during a pandemic where many routine diagnostic data, such as imaging studies, may not be available due to infection containment considerations or an overwhelmed health system. To address this, we further divided this basic and minimal set of Tier 1 CDEs into Core and Supplemental CDEs (Table 1). Core elements represent the minimal dataset that can be rapidly and easily captured during initial patient triage and care. Core CDEs should be recorded as completely as possible. Supplemental elements capture additional clinical characteristics, basic pre-morbid conditions, laboratory values, and outcome measures beyond acute hospitalization.

Table 1 GCS-NeuroCOVID Tier 1 adult study data elements and definitions
Full size table

An added benefit of pragmatic data elements is the possibility for an expedited institutional review board (IRB) process at single-center sites. During pandemic surges, many sites limit research operations to reduce risk of exposure or facilitate re-allocation of staff to needed areas. As such, there may be limited personnel to review and approve lengthy or complex research protocols, and mechanisms for contract negotiations for data sharing may be temporarily suspended. A pragmatic protocol deemed not greater than minimal risk may be feasibly approved at sites using a single-center approval process. Subsequent amendments when resources are not as limited may then be made to establish data-sharing agreements via a central data coordinating center to pool data from participating sites for analysis and ensure use of secure platforms for protected health information (PHI).

The second principle of global CRF development includes data harmonization. Alignment of data elements and definitions across sites and studies serves the vital role of reducing heterogeneity of findings across studies, thus allowing for data pooling, systematic reviews, meta-analyses, and development of evidence-based guidelines based on a body of high-quality evidence. These factors are of critical importance when developing CRFs for a global consortium during a pandemic. As data rapidly become available, it is crucial to be able to cumulatively synthesize findings across studies to draw accurate conclusions and drive care decisions in real time.

To this end, the GCS-NeuroCOVID adult CRF aligns data fields and definitions with CDEs from the NIH where applicable when documenting neurological manifestations observed in COVID-19 infection. Our investigator team also engaged with emerging and existing consortiums to align common definitions across COVID-19 populations and to have CRFs available in other languages to encourage participation across geographic regions and resource settings.

Development of Adult Spanish Version CRF

The GCS-NeuroCOVID consortium includes many sites located in primary Spanish-speaking regions. The Regional South American chapter of the Neurocritical Care Society was instrumental in aligning collaborations to facilitate participation among member sites and regional groups. To minimize data collection burden for the frontline clinicians at those sites and promote data fidelity, we developed Spanish language versions of study documents and CRF (Table 2). Spanish translation was performed by trained research team member (VA) and neurologist (CV) at the central coordinating center (University of Pittsburgh) and reviewed and edited by global partners in primary Spanish-speaking regions (JM, NM). Data elements were reviewed and adjusted for applicability to the local region. Specifically, we avoided using medical abbreviations as there are variabilities in regional customs.

Table 2 Spanish GCS-NeuroCOVID Tier 1 adult study data elements and definitions
Full size table

The Latin American Brain Injury Consortium (LABIC) is the association of neurointensivists in Latin America. LABIC’s missions are (1) to promote education in neurointensive care and (2) to foster neurocritical care research and thereby improve clinical care and patient outcome throughout Latin America. LABIC consists of over 400 critical care professionals from Patagonia to México and the Caribbean countries who are dedicated to improve care and outcome of their neurological patients. The LABIC consortium has a track record of successful multicenter collaboration and research such as the publication of consensus guidelines for traumatic brain injury management [25]. The LABIC consortium is actively developing research initiatives and collaborations with global partners such as the South American chapter of the Neurocritical Care Society (NCS) and the GCS-NeuroCOVID consortium. The GCS-NeuroCOVID Spanish data elements and CRF are developed in close collaboration with LABIC consortia. This harmonization of data elements and definitions is an important step toward harmonized data capture of COVID-19 neurological manifestations across the North, Central, and South America.

Harmonization with POSSIBLE Network and Development of Adult Portuguese Version CRF

The GCS-NeuroCOVID consortium is endorsed by and collaborates closely with the Prospective Observational Study of Subarachnoid hemorrhage and IntracereBral hemorrhage patients in Latin America (POSSIBLE) network. The POSSIBLE network was created in 2017 to investigate the epidemiology and the medical practices in hemorrhagic stroke and subarachnoid hemorrhage in 33 Latin America critical care units including 12 primary Portuguese-speaking sites in Brazil. Non-Portuguese-speaking sites include critical care units from Argentina, Bolivia, Chile, Colombia, Cuba, Equator, Paraguay, Peru, Puerto Rico, Uruguay, and Venezuela. There are three coordinating sites: Brain's Institute in Rio de Janeiro, led by CR and PK (for centers from Brazil) and Fundación Valle del Lili, and Hospital Universitário (Cali, Colombia) and Hospital de Especialidades Eugenio Espejo (Quito, Equator), led by JM and NM, respectively (for the non-Portuguese-speaking sites). All data collected are stored in a RedCap database linked to the Brain's Institute.

As soon as the COVID-19 pandemic emerged in South America, two POSSIBLE network investigators (CR and PK) developed a common database for neurological manifestations in COVID-19 patients by leveraging the existing infrastructure of POSSIBLE network in Brazil. In collaboration with the GCS-NeuroCOVID consortium, lead POSSIBLE investigators (CR and PK) harmonized the GCS-NeuroCOVID Tier 1 CDEs with existing POSSIBLE CDEs to create a single neurological COVID-19 CRF for the POSSIBLE network. They further translated all data elements and the data dictionary into Portuguese (Table 3). Some variables were adjusted due to regional heterogeneity. For example, the discharge disposition was changed to free text, because there are few long-term facilities like hospice units or long-term care units in Brazil.

Table 3 Adult CRF Portuguese version
Full size table

Development of Pediatric CRFs (English and Spanish)

The pediatric CRF (Table 4) was approached with similar principles as the adult CRF with some intentional differences. COVID-19 infection prevalence in hospitalized children is estimated to be 9.4 and 5.4 per 100,000 in children aged 0–4 and 5–17 years of age, respectively, compared with 316.9 per 100,000 adults > 65 years of age [26]. The lower patient volume in the pediatric population makes collecting more detailed data elements possible for the pediatric CRF. Conversely, the low volume expected per country demands multicenter collaboration to attain a large enough sample size to produce robust results to inform clinical care and research initiatives. Some data elements, such as those specific to intensive care admission (e.g., mechanical ventilation, risk of mortality scores), are already standardized and commonly collected in pediatric intensive care units (PICUs). However, many of the baseline and COVID-19-related data elements should harmonize with other pediatric COVID-19 studies, none of which (to our knowledge) includes detailed neurological manifestations. Finally, unique data elements specific to pediatrics are necessary in order to account for the following: (1) child age and developmental stage [27, 28]; (2) pediatric-specific risk adjustment [29, 30]; (3) baseline hospital characteristics; (4) differences in pediatric versus adult morbidities, symptoms, treatments, and outcomes; (5) inclusion of multisystem inflammatory syndrome related to COVID-19 (MISC) data elements [31], a pediatric-specific manifestation; and (6) child- and family-validated outcome measures.

Table 4 Pediatric case report form (CRF)
Full size table

Similar to the adult study, the pediatric collective is actively working to establish partnerships with centers and networks affected by COVID-19 in high-, middle-, and low-income countries. To facilitate this aim, we engaged with the LARed (“Red Colaborativa Pediatrica Latinoamericana, or Collaborative Pediatric Latin American Network), which in turn reached out to partner networks including CLaNi (Colegio Latinoamericano de Neurointensivismo), AMCI (Asociacion Colombiana de Medicina Critica y Cuidados Intensivos), and SLACIP (Sociedad Latinoamericana de Cuidados Intensivos Pediatricos SM) to recruit sites outside of LARed. At the time of this writing, these efforts have resulted in 22 registered sites spanning from Mexico to Chile and Argentina. We developed Spanish language versions of study documents and CRF (Table 5). Spanish translation was performed by team members (MS, JDR.) and reviewed and edited by global partners in primary Spanish-speaking regions (PV, SG). As in the adult CRF, data elements were reviewed and adjusted for applicability to the local region and medical abbreviations were avoided.

Table 5 Pediatric CRF Spanish version (formato para reporte de casos)
Full size table

Conclusion

We report the general guiding principles and framework for the development of CRFs for a global consortium during a global pandemic crisis. Unique features and considerations include: (1) timeline and speed—in order to address an explosive pandemic, we had to adopt an extremely accelerated and succinct process for CRF development, consensus CDE development, and data harmonization; (2) pragmatism and feasibility in a pandemic—the CRF and data elements present minimal burden to frontline clinicians who populate these data, including minimizing exposure risk and PPE use; (3) adaptation to rapid change—new data and information rapidly emerge in this new pandemic and studies must run on an accelerated timeline to provide timely and accurate information to the public; (4) inclusion of the life span to understand age-related effects; and (5) global partnerships and adaptation of CRF into multiple languages. As a result of this initiative, to date, there are 218 registered sites for the study representing 109 countries spanning six continents across the globe, and formal collaborations with other large research networks such as the European Academy of Neurology (EAN) NeuroCOVID Registry (ENERGY) [32] have been established. Participation on working groups within the World Health Organization is ongoing to align efforts with other initiatives related to neurological implications of COVID-19 [18, 33]. Throughout this work, the adaptable, layered design of the GCS-NeuroCOVID consortium affords a nimble yet systematic and scientific approach that may ultimately serve as a model for future studies that require rapid execution in the midst of pandemics or other overwhelming natural disasters.