Development, Use, and Impact of a Global Laboratory Database During the 2014 Ebola Outbreak in West Africa

Summary

The usefulness and value of a multifunctional global laboratory database is far reaching, with uses including but not limited to informing local outbreak interventions, developing global outbreak response strategies, virtual biobanking, and disease forecasting.

Abstract

Background.

The international impact, rapid widespread transmission, and reporting delays during the 2014 Ebola outbreak in West Africa highlighted the need for a global, centralized database to inform outbreak response. The World Health Organization and Emerging and Dangerous Pathogens Laboratory Network addressed this need by supporting the development of a global laboratory database.

Methods.

Specimens were collected in the affected countries from patients and dead bodies meeting the case definitions for Ebola virus disease. Test results were entered in nationally standardized spreadsheets and consolidated onto a central server.

Results.

From March 2014 through August 2016, 256343 specimens tested for Ebola virus disease were captured in the database. Thirty-one specimen types were collected, and a variety of diagnostic tests were performed. Regular analysis of data described the functionality of laboratory and response systems, positivity rates, and the geographic distribution of specimens.

Conclusion.

With data standardization and end user buy-in, the collection and analysis of large amounts of data with multiple stakeholders and collaborators across various user-access levels was made possible and contributed to outbreak response needs. The usefulness and value of a multifunctional global laboratory database is far reaching, with uses including virtual biobanking, disease forecasting, and adaption to other disease outbreaks.

The 2014 Ebola outbreak in West Africa was declared an international public health emergency in August 2014 [1]. The high degree of population mobility across porous borders in West Africa introduced operational challenges and contributed to the rapid spread of Ebola virus disease (EVD) [2]. The regional and international impact of EVD demonstrated the importance of centrally compiling data to understand disease trends and guide public health interventions.

The effectiveness and efficiency of a public health response during an outbreak is dependent on having reliable and up-to-date information as the outbreak evolves [3, 4]. Lessons learned from past communicable disease outbreaks have emphasized the importance of having an accurate and real-time database that is accessible to relevant agencies involved in the outbreak response [5, 6]. Information from laboratories is particularly invaluable for outbreak control activities [7–9]. During the early stages of the 2014 Ebola outbreak, daily laboratory results provided the most reliable and timely information in comparison to delayed epidemiological reporting due to overwhelmed systems. Guidelines for laboratory data management, as part of the laboratory capacity requirements for the International Health Regulations, recommend computerized laboratory information systems, standardized data collection, and periodically disseminated activity reports at each level of health care service delivery [10, 11].

The World Health Organization (WHO) Emerging and Dangerous Pathogens Laboratory Network (EDPLN) is a network of high-security diagnostic laboratories that contribute to long-term outbreak preparedness and response efforts [12]. Established in 2008, the WHO EDPLN, with the support of Global Outbreak Alert and Response Network (GOARN), has overseen the global deployment of mobile field laboratories during outbreaks [13]. For the 2014 EVD outbreak, laboratories were deployed in the 3 most affected countries (Guinea, Liberia, and Sierra Leone).

To address the gap of a consolidated, centralized laboratory database that is necessary to inform response in a multicountry outbreak, the EDPLN supported the development, maintenance, and implementation of a global laboratory database. The database collated and reported data in near real time from 47 laboratories in Guinea, Liberia, and Sierra Leone. In contrast to other Web-based data systems, the spreadsheet-based system, which allows for maximum adaptability, information technology literacy, and coordination between data collection and collation. This article outlines the methods, practical applications, and limitations of the global laboratory database used to respond to the recent EVD outbreak.

METHODS

Laboratory Diagnostic Capabilities

Through the EDPLN and bilateral agreements with the EVD-affected countries, fixed and mobile laboratories were deployed to support diagnosis and confirmation of suspect and probable cases of Ebola, as well as the discharge decisions of convalescent patients with Ebola. The laboratories were owned and deployed by Belgium, Canada, China, England, France, Germany, Italy, Netherlands, Nigeria, Senegal, South Africa, and the United States of America. Some countries deployed multiples laboratories throughout the region (Figure 1).

Figure 1.

Mobile and fixed laboratories deployed in Guinea, Sierra Leone, and Liberia since March 2014.

Specimen Collection

Specimens were collected from patients presenting at hospitals, treatment centers, and clinics who had symptoms concordant with case definitions for EVD adopted by respective ministries of health (MoHs) and the WHO [14, 15]. Patients who tested negative for EVD with <48 hours between symptom onset and specimen collection were retested. Successive specimens were collected from EVD-positive patients until the patient received 2 negative test results, at which point they could be discharged. Oral swab specimens were also collected from dead bodies in the communities and health facilities according to protocols [16]. All specimens were transported to the laboratories for subsequent real-time qualitative reverse transcription–polymerase chain reaction testing.

Data Collection

At the start of the outbreak, laboratory staff collected ad hoc demographic data from the case investigation forms and entered them into an Excel spreadsheet (Microsoft, Redmond, WA), alongside laboratory results. In early 2015, all mobile and fixed laboratories that were members of EDPLN and/or were established de novo on the basis of bilateral agreements between donor countries and the 3 West African countries implemented nationally standardized spreadsheets for data collection.

Nationally standardized spreadsheets were developed using Excel 2010 and compatible versions by key stakeholders within Guinea, Liberia, and Sierra Leone, in collaboration with the WHO. Excel was chosen over tablet computers, smart phone–based applications, and Web-based platforms because it required relatively low computer literacy and inexpensive software investment and could be used in areas with a limited Internet connection. Each country’s MoH devised its own processes and procedures related to reporting requirements from the laboratories, including identifying which data fields were required for reporting. At a minimum, results were emailed once daily from the laboratory to the MoH, in addition to the relevant clinical teams.

Laboratory Results Consolidation

Results consolidation occurred at the national and global levels, with the processes being strengthened and standardized as the outbreak progressed. National database managers or focal points were responsible for consolidating data for in-country use and electronically sending data to the WHO. Each spreadsheet from the country was imported into a database by using extract, transform, and load tooling. Data-loading packages migrated data from spreadsheets with structural and contextual variations into a consolidated database. Data standardization was incorporated into the loading process by automatically formatting a series of expected inputs to standard outputs. Fields such as sex, age, test sequence, and dates were standardized in this way. Entries that did not match the expected list were manually reviewed prior to loading as part of a data verification process.

Record Verification and Cleaning

Record verification and cleaning was conducted during 3 stages of the data management process, to minimize and rectify manual and electronic errors that may have occurred during data collection and data loading. Prior to loading the data, fields were formatted to ensure successful importation into the database. Once uploaded, records were manually reviewed, and reference tables were updated according to standard operating procedures, to provide further data standardization and quality assurance. All staff involved in data management and cleaning were trained on implementing standard operating procedures, to ensure consistency in all data cleaning and verification processes. The last stage of verification occurred while conducting analyses to confirm accurate representation of the situation in-country. Gaps in information were resolved through regular communication and validation with MoHs, WHO country offices, and laboratories.

Data Analysis and Reporting

The database was linked directly to a secure Web-based dashboard where aggregate numbers, graphs, and charts were accessed in real time. Descriptive and statistical analyses were conducted with Stata 13 (StataCorp, College Station, TX) and Excel. Analyses were compiled into weekly reports and disseminated to MoHs, laboratories participating in the response, and other partners, as guided by the respective MoHs. Data were analyzed to identify local and regional trends, as well as to assess and adapt operational response strategies.

Data Ownership, Access, and Security

Ownership of the laboratory data remained with the respective MoHs. The system used a 3-tier data access system that enabled national, regional, and global stakeholders to directly export consolidated and cleaned data from a Web-based platform according to their predesignated access rights. Access to the data and reporting dashboard was granted via unique, secure usernames and passwords under the clearance and guidance of the respective MoH.

System Design

The laboratory database was designed to easily adapt to any infectious disease outbreak and with in-country end users in mind. To account for the unpredictability of outbreaks, data-loading packages were built to accommodate varying levels of data quality from a variety of data fields. Data loading, management, and analyses relied on Excel, to improve flexibility in data processes and reduce the need for computer programmers. Functions to maintain data integrity while improving quality were included to that ensure records could be updated as necessary and that changes could be traced back to the persons working within the data set. The database was also archived daily to allow for recovery of information if and when needed. Figure 2 depicts the flow of data from specimen collection to reporting and analysis.

Figure 2.

Data flow to and from laboratories and between levels of response.

RESULTS

The first set of laboratory results were received in March 2014. As of 31 August 2016, the database consisted of 256343 specimens from 4830 spreadsheets reported from 47 laboratories located in Sierra Leone, Guinea, and Liberia. On average, the data were analyzable within 24 hours after receiving the test result. The database captured 39 different types of specimens tested in these laboratories for EVD, the majority of which were blood specimens (38%) and oral swab specimens from dead bodies (41%). The remaining specimens tested include environmental swab specimens and samples of maternal blood, capillary blood, breast milk, semen, and other bodily fluids. Non-EVD tests, such as malaria diagnostic tests, blood biochemistry tests, and antibody tests, performed concurrently to EVD testing were included in the database, although these accounted for <5% of records.

With regard to data quality, analysis of data completeness showed variations by data field (range, 1.3%–100%) across the 3 countries (Table 1). Approximately half of the data fields within the database had >75% data completeness, although each country prioritized certain data fields to account for country-specific needs.

Table 1.

Database Field Completeness for Ebola Virus Disease (EVD)–Affected Countries

Field	Completeness, %
Laboratory name	100.0a
Unique patient identifier	92.3a
Laboratory sample identifier	96.9a
Name	99.1a
Surname	95.7a
Age	94.2a
Sex	86.9a
Country	100.0a
Location of specimen taken	72.7
Province, region, or county	91.3a
District or prefecture	90.7a
Chiefdom, subprefecture, or village	63.7
Initial or repeat test	81.1a
Specimen test sequenceb	10.4
Date of patient symptom onset	54.8
Date of patient hospitalization	5.6
Date of specimen collection	92.7a
Date of specimen received in laboratory	54.7
Date of test performed	97.7a
Date of patient deathc	8.6
Type of specimen collected	95.7a
Patient clinical status at time of testc	97.0a
EVD test result	99.6a
EVD test Ct value	13.6
Referral center	68.6
Results of non-EVD testsd	4.2
Patient contact information, address, profession	8.9
Symptoms	1.3
Comments	41.8

Abbreviations: Ct, cycle threshold

^aThe most commonly used data fields (completeness, >75%).

^bFor example, the first test, second test, third test, etc.

^cData were documented as of the time the sample was taken.

^dMalaria diagnostic testing; Lassa virus immunoglobulin M, immunoglobulin G, and antigen testing; and the Piccolo system.

Data completeness also varied over time. From August 2014 (testing week 31 in 2014) through November 2014 (testing week 48 in 2014), laboratory testing across the 3 countries increased at a rate of 11.7% per week. During this period, data completeness of key fields remained between 75% and 85%, after adjustment for country variations. By January 2015, data completeness rose to 89% and fluctuated between 89% and 96% through 2016 despite relatively sustained high levels of laboratory testing (Figure 3). This increase may be attributable to a decrease in the percentage of positive samples, as well as to efforts by the WHO teams to implement standardized spreadsheets and provide regular feedback, training, monitoring, and support to improve data quality and data collection.

Figure 3.

Data completeness in comparison to laboratory specimen test results in Guinea, Liberia, and Sierra Leone since March 2014

The broad range of data fields collected allowed for a multitude of analyses, which could be refined over the course of and after the outbreak. The laboratory data provided information necessary for monitoring, forecasting, supporting research questions, and providing epidemiological and strategic direction. Analyses from the database focused on but were not limited to understanding laboratory capacities, monitoring the functionality of the laboratory and response systems, understanding key demographic data of the patient population, mapping the geographic distributions of positive specimens, and illustrating the burden on clinical services and community burial teams (Figure 4).

Figure 4.

Example analyses created for informing response activities.

DISCUSSION

The geographic spread of cases, volume of laboratory tests performed, and speed at which information needed to be compiled to inform outbreak activities demanded an effective laboratory information system from the start of the 2014 EVD outbreak. Ideal data systems include those that are electronic [17], automated [17], free from political bias [5], and respectful of country sovereignty [8]. The laboratory database described in this article provided valuable and near real-time data that were used for quick decision making at various levels of the outbreak response, particularly when the epidemiological and surveillance systems were overwhelmed. The key lessons learned in the creation of a global laboratory database primarily pertain to the need to strengthen integrated data systems, to ensure standardized data collection tools and processes are in place, to invest in dedicated data management and information technology teams, and to maintain regular communication, feedback, and collaboration.

The usefulness of a database is predominately dependent on the quality of the data received. In resource-limited settings, laboratory staff may collect and manage case information through log books, paper records, and other nonstandardized paper methods [9, 18]. While these approaches provide a quick and convenient method to compiling information during small-scale disease outbreaks, the large number of cases in the 2014 EVD outbreak required a consolidated, electronic database that could provide real-time results across multiple locations. However, there were numerous challenges associated with the timely collection and reporting of data when existing data systems were overwhelmed. In the beginning of the outbreak, reported data varied considerably by laboratory and country, as did data quality. For example, data were reported on an ad hoc basis, dates were entered in both American and European standards, colloquial location names were instead of officially recognized names, and open-text fields introduced variation that made interpretation difficult (eg, the comments field contained additional information on specimens and clinical presentations). The use of standardized spreadsheets with dropdown menus and locked formatting reduced data entry errors and improved data integrity. The authors recommend prioritizing the strengthening of data systems during preparedness and recovery efforts, to ensure data usability and the connectedness between systems. If a laboratory system does not currently exist, deploying and implementing a standardized data collection and management package at the start of an outbreak to each of the reporting sites and/or laboratories should be considered. Efforts should also be made to ensure that all laboratories are uniformly entering the data and that protocols and procedures for data flow are established.

Previously documented laboratory data systems have attributed data quality challenges to limited computer literacy [8] and Internet availability [18]. However, an additional challenge in the 2014 Ebola outbreak was ensuring consistent and error-free data during high staff turnover and use of multiple data collectors or data entry specialists. For example, district surveillance officers and epidemiologists filled out paper-based case investigation forms, which detailed demographic data and unique patient identifiers. Laboratory technicians and data clerks, many of whom were on 4–6-week rotations, reentered data from the case investigation form along with the laboratory results after the diagnostic EVD tests were completed. Often the format of the laboratory spreadsheet changed with laboratory staff turnover and in-country needs and decisions. Without adequate human resource support and open communication, even the most efficient database processes, user-friendly data collection tools, and the most robust analyses will fall short. The authors recommend long-term investment into data management teams at the national level to ensure accurate and consistent data collection, management, and reporting. Further, developing a human resource plan for a data management team to ensure sufficient knowledge transfer in the event of staff turnover, designation of focal points with clear terms of reference, roles, and responsibilities to address data-related issues should be considered. Ensuring daily communication and feedback between all laboratory and data clerks/data managers is essential.

For a laboratory data management system to be useful in emergency situations, it must have mechanisms to adapt quickly, for changes to be reflected immediately, and for data to be deleted and modified without having to alter the entire system [8]. The size of the global laboratory database, variability of information collected and entered, sensitivity of the data, and need for reliable and up-to-date information demanded significant dedicated on-site information technology resources to immediately implement system changes on the basis of in-country needs. The authors suggest planning and budgeting for long-term information technology support when developing databases, to ensure efficiency, sustainability, and functionality over time.

A limitation of the global laboratory database is that records are at the specimen level instead of the patient level. Retrospectively linking data to the patient level and subsequently to other data sources, such as burial, clinical, and surveillance data, was not done during the outbreak and will require additional time and human resources. Developing and implementing a unique patient identification system at the beginning of an outbreak, actively maintaining this system throughout the outbreak, and developing a single database with complete laboratory, clinical, geographic, and epidemiologic data of all suspect cases, contacts, and survivors is critical to ensuring useable data at both the patient and specimen level. An additional limitation is that, despite the best efforts to verify and clean data, there may be inaccurate records (eg, the date of specimen collection is after date of test) in the global laboratory database if data collectors and data clerks are no longer contactable or if paper-based records are no longer available.

The usefulness and value of the laboratory database has been far reaching. With substantial amounts of data being shared and integrated on a common platform, the system was able to inform global response strategies and prompt decision-making, as well as address research questions that were not considered during the height of the 2014 EVD outbreak. The multifunctional global laboratory database can be easily adapted to any type of infectious disease outbreak and is quickly deployable with minimal training required for end users. Its usefulness for disease forecasting should be explored further. Additionally, the database has the potential to act as a virtual biobank, functioning as a digital repository where users can add, track, and share specimens. As demonstrated through the global laboratory database, it is possible to design, implement, manage, and analyze large amounts of data with multiple stakeholders and collaborators, different languages, various levels of user access, multiple data flows into and out of countries, and different end-user needs to inform immediate needs during outbreak response.

Notes

Acknowledgments. We thank Jaclyn Marrinan, Poushali Ganguli, Polly Oates, Reem Ghalib, Eliana Duncan, Ruth Zwizwai, Imad Faghmous, Rutvij Merchant, Antoine Coursier, Katherine Callaway, Richard Watkins, Amrita Ghataure, Beade Numbere, Ayibadiyepriye Ruth Igali, Nwoza Eshun, Shannon Boldon, Bobson Illoh, Jamie Currell, Allison Ng Zhou Ling, Isobel Firth, Stephanie Coward, Rosanna Jeffries, Dorothy Leonor, and Ravi Santhana Gopala Krishnan, for their support in cleaning the data, geographic information systems mapping, and information technology support; and the following institutes and laboratories, for their support and collaboration: Belgium—Centre for Applied Molecular Technologies, Université Catholique de Louvain and Defense Laboratory Department: Belgium Mobile Lab–Nzérékoré (Guinea [GN]). Canada—National Microbiology Laboratory Public Health Agency (PHA) of Canada: PHA Canada–Conakry (GN), PHA Canada–Kamsar (GN), PHA Canada–Magburaka (Sierra Leone [SL]), PHA Canada–Kailahun (SL). Guinea—Institut National de Santé Publique de Guinée, Laboratoire des Fièvres Hémorragiques: Boke Mobile Lab–Boke (GN), Tanene Mobile Lab–Tanene (GN), INSP/PFHG/IPD Lab–Nzérékoré (GN). Guinea and Senegal—Guinea National Public Health Lab with IP Dakar: IP Dakar-Dixinn Lab (GN). France—French Service de Santé des Armées: CTS Lab–Conakry (GN); K-Plan: K-Plan Mobile Lab–Conakry (GN), K-Plan Mobile Lab–Forecariah (GN); Institute Pasteur Lyon: IP France–Macenta (GN). European Union and France—European Union–IP Lyon Expertise France: EUWAM Lab–Conakry (GN). European Union and Germany—European Consortium with Bernhard-Nocht-Institute for Tropical Medicine: EU Mobile Lab–Guéckedou (GN), EU Mobile Lab–Coyah (GN), EU Mobile Lab–Nongo (GN), EU Mobile Lab–Foya (Liberia [LR]), EU Mobile Lab-Hastings (SL) within European Union and Germany—European Consortium with Bernhard-Nocht-Institute for Tropical Medicine list of labs. Italy—Lazzaro Spallanzani National Institute for Infectious Diseases: Italian Mobile Lab–Goderich (SL). Liberia – Ministry of Health Liberia: MOH–Montserrado (LR). Liberia and United States of America—Liberia National Reference Lab with USAMRIID: LIBR NRL/USAMRIID Lab–Margibi (LR). Netherlands and Liberia—RIVM Laboratory for Infectious Diseases and Screening, Centre for Infectious Disease Control and Ministry of Health Liberia: Dutch Mobile Lab–Sinje (LR). Netherlands and Sierra Leone—RIVM Laboratory for Infectious Diseases and Screening, Centre for Infectious Disease Control and Ministry of Health Sierra Leone: Dutch Mobile Lab–Kono (SL). Nigeria, European Union, and Germany—Nigeria with European consortium with Bernhard-Nocht-Institute for Tropical Medicine: Nigeria Mobile Lab–Kambia (SL), Nigeria Mobile Lab–Kingtom (SL). People’s Republic of China—Government of the People’s Republic of China: China-CDC Lab–Jui (SL). Russia—CREMS Pastoria: CREMS Lab–Kindia (GN). SENEGAL – Institute Pasteur Dakar: IP Dakar–Matoto (GN). Sierra Leone – Ministry of Health Sierra Leone, Princess Christian Maternity Hospital Emergency: MOH/Emergency–PCMH Lab–Freetown (SL). United Kingdom—Public Health England Porton Down: PH England Mobile Lab–Bo (SL), PH England Mobile Lab–Kenema (SL), PH England Mobile Lab–Kerry Town (SL), PH England Mobile Lab–Port Loko (SL), PH England Mobile Lab–Makeni (SL). South Africa, Sierra Leone – National Institute for Communicable Diseases (NICD) with Ministry of Health Sierra Leone: EMDF/NICD–Lakka (SL). United States of America—US Defense Threat Reduction Agency (DTRA): REDC Lab–Conakry (GN), US DTRA–Moyamba (SL), CPHRL/DTRA–Lakka (SL); Academic Consortium to Combat Ebola in Liberia: Tappita Lab–Nimba (LR), Redemption Hospital Lab–Montserrado (LR); Centers for Disease Control and Prevention (CDC) and National Institutes of Health: US-CDC Lab–Montserrado (LR); CDC: US-CDC Lab–Bo (SL); US Army–NMRC: OIC-NMRC Mobile Lab Bong (LR); US Department of Defense: US 1st Area Medical Laboratory–Grand Gedeh (LR), US 1st Area Medical Laboratory–Nimba (LR), US 1st Area Medical Laboratory–Sinoe (LR); US Navy: OIC-NMRC Mobile Lab Montserrado (LR).

The WHO thanks the WHO’s Ebola Response Programme for support of the Global Ebola Laboratory Database.

Potential conflicts of interest. All 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.