What underlying factors contributed to the unclear data picture that characterized much of the Ebola outbreak response, particularly in the early days? Several interviewees referred to a “fog of information,” describing the lack of timely, accurate, and accessible data, which clouded situational awareness, impeded effective decision-making, and stymied the response. This section summarizes common themes emerging across interviews that illustrate this fog. These themes demonstrate the impact of the physical, socio-cultural, organizational, and historical context of the Ebola outbreak on the response, including for data and information flows, and the use of data and digital systems.
The WHO declaration in August 2014 of a Public Health Emergency of International Concern (PHEIC), and the humanitarian emergency sparked by the Ebola outbreak, demanded significant public health and humanitarian responses. The hybrid nature of the crisis as both a health and a humanitarian emergency increased coordination challenges in the response and contributed to confusion among response actors. Although many humanitarian agencies had experience managing infectious disease epidemics, such as cholera or measles, or other health crises, such as malaria, few had experience in managing an epidemic of this type and scale, or of managing the required health facilities. The characterization of the Ebola outbreak as a public health emergency meant that the secondary impacts of the crisis on local populations received less attention, particularly in the early months of the epidemic.
As one interviewee put it, “We have blueprints for crisis and conflict, but nobody had a blueprint for a major health crisis.” The complexity of responding to an epidemic in a densely populated urban environment, with increased potential to spread the disease, compounded this issue. The fluidity and rapid pace of the outbreak forced responders to quickly adapt to changing circumstances. One humanitarian responder exclaimed, “Everything happened so fast. We were chasing the disease, [and] had to change quickly to adapt [to] what was happening. It was not like [a] conflict [response].”
"We have blueprints for crisis and conflict, but nobody had a blueprint for a major health crisis."
Although there was no set playbook for how to proceed in an epidemic of this type, scale, and speed, or of how to manage related secondary effects, some interviewees noted there were existing mechanisms and protocols that could have been better adapted and reused in the Ebola response. On the health side, one veteran of health responses pointed out that the protocols for responding to infectious disease, and polio in particular, were not consistently deployed in this response, despite similarities in how to contain the two diseases. He noted, “In middle of the outbreak, we were reinventing the wheel without taking into account and leveraging what exists. This is nonsense. We have case tracking for polio, with labs and contacts. This exists already. We can plan vaccination campaigns, which involve social mobilization. This knowledge was not used.” On the humanitarian side, many interviewees highlighted the absence of the Office for the Coordination of Humanitarian Affairs (OCHA) and the humanitarian cluster coordination system as contributing to confusion surrounding coordination and information sharing, and forcing a reinvention of the wheel when established practices already existed.
Within the UN system disease epidemics or pandemics are within the purview of the WHO, while coordination and information management in humanitarian emergencies fall under the remit of OCHA. To address the hybrid nature of the crisis and the perceived shortcomings of the WHO response in the early phases of the outbreak, the UN General Assembly passed a resolution creating the UN Mission for Ebola Emergency Response (UNMEER) as a temporary mission to coordinate the UN’s activities in support of a nationally-led Ebola response. This decision had ripple effects across the entire response.
Many interviewees mentioned the negative impact of the absence of the OCHA-led humanitarian coordination system on the collection and sharing of data, particularly at the beginning of the response. In particular, the normal OCHA humanitarian coordination system identifies a lead agency for each of the major “clusters” of humanitarian response (e.g., health, emergency telecommunications, refugee camp management). With many of the normal coordination channels absent, particularly in early phases, the collection and management of response-wide information suffered.
The governments of Guinea, Liberia, and Sierra Leone maintained a strong role in leading the response through their respective national coordinating bodies. Each country organized its national coordinating body differently in its operations, information management, and coordination structures. In Liberia, for example, the national coordinating body known as the Incident Management System (IMS) was limited in terms of who and how many individuals could participate in coordination meetings. This was necessary both to maintain effective command and control in a chaotic environment and to physically fit everyone into the room. At the peak of the crisis, the government convened daily IMS meetings. IMS participation was limited to two individuals per agency, and included the heads of various working groups and partners that played key roles in the response. Friday meetings were open and designed as a “briefing” meeting. In addition, the WHO hosted open information-sharing meetings on Saturdays, which were open to all agencies.
Without the typical humanitarian coordination systems or a clear pathway to engage with the national coordination bodies, many implementing organizations, both national and international, were excluded from discussions. NGOs and other response actors that were absent from those conversations had a limited view of the changing landscape of the outbreak and response. As a result, gaps and duplication in the response were less apparent and less easily resolved, especially during the chaotic early months of the crisis. For example, several interviewees mentioned that, especially at the beginning of the response, their organizations were the only ones operating in a particular part of the country. One interviewee observed that controlling the dissemination of information in this way allowed governments to maintain control of the response.
The absence of the typical humanitarian coordination system and the unique deployment of UNMEER as a coordinating body resulted in confusion about roles and responsibilities. A report investigating the international aid architecture in the Ebola response found that “UNMEER’s status as a new entity/invention forced it to define itself in the midst of a full-blown emergency..., where major response actors, from donors and governments to [international NGOs], had little knowledge of its purpose, objectives, or how to work with it. As late as March and April 2015, key implementing NGOs had little understanding of UNMEER’s role and functions.” Multiple interviewees shared their frustration with the absence of traditional coordinating mechanisms. One stated: “The [usual humanitarian coordination] systems did exist, but we weren’t allowed to use them.” Another said, “OCHA has structure, it’s predictable, people know what [OCHA is] going to do. … You know what to expect and that predictability is critical. To reinvent everything was really difficult.”
Finally, the decision to stand up UNMEER using staff from the UN Department of Peacekeeping Operations (DPKO) influenced how data and information were shared in the response. Multiple interviewees highlighted the diverging data cultures between a humanitarian operation, in which information-sharing is an essential part of the mandate, and a peacekeeping mission, in which information is largely regarded in the context of privacy and security. As one humanitarian official said, “DPKO does record keeping for the mission. They are not focused on public information ... or on making products for public consumption.” In contrast, OCHA regularly and openly distributes information, such as maps of agencies operating in a crisis (known as the “3Ws”: who is where, when, doing what), to support response efforts. Another interviewee expressed a similar sentiment, indicating DPKO’s approach to information management typically focused on “information security and what should be classified. It is internal information management, not external.”
Tensions Between Data Types and Uses
In discussing data use in the response, interviewees pointed out differences between operational data collected at a cluster level that were used to inform the activities of international and national actors, versus the more aggregated data used to influence advocacy and high-level planning efforts. This bifurcation between detailed, local data and aggregated, usually national-level data is characteristic of other emergency responses. In the Ebola outbreak response, however, the characteristics and uses of case data were new in the context of a humanitarian emergency. The data required to understand the nature and transmission of the disease, particularly in the early days of the response, were detailed, cumbersome to health responders given the scale of this outbreak, and oriented toward a post-epidemic analysis as opposed to real-time operational usage. As Jeremy Konyndyk, director of the Office of U.S. Foreign Disaster Assistance, stated: “One of the things we’ve struggled with is getting good operational epidemiological data. When you’re fighting a disease like this it is important to have a real-time picture of where it is growing, falling, and at the national level, where it is spreading and receding. The nature of the disease is that the spread is so particularly localized. Where is key. If you have that, then you have a greater picture of which interventions are working or not paying off. With the lag, it is harder to know what is effective or not.”
The hybrid nature of a combined health and humanitarian emergency surfaced tensions between the types of data collected and used. Those providing care for patients in treatment units needed data about symptoms and treatment regimens for patients. Those coordinating the response needed data about active hotspots and quarantine locations, information about essential supplies, and basic case data to implement the operational response. Those working in and with communities needed accurate information about specific behaviors and other measures needed to stop the spread of Ebola. Yet few data systems were designed to house these various types of data, and the data systems in use often were weak and disorganized. With a few exceptions there were no widely accepted places where response actors could share their data.
According to one health expert, “Multiple disconnected health data systems posed a major challenge to containing the disease. Standalone digital systems were often found to be in place for critical data sets that needed to be interpreted together--such as those for contact tracing, case investigation, and treatment of patients. For example, a reporting system tracking payments for health workers didn't link up with a separate system tallying their work hours. This blocked payments for nurses and burial experts risking their lives and led many of them to refuse to continue working.”
In addition, within the health domain, interviewees distinguished between the needs and skill sets of field epidemiologists and research epidemiologists. The former focused on investigating individual cases, and the latter on analyzing individual and aggregated data to understand the incidence and transmission of the disease. Despite these distinctions, all data types were interrelated and needed to support an effective response.
Finally, the hybrid nature of the Ebola emergency illustrated the differential impacts based on the type and granularity of data required. Whereas the operational side of most humanitarian responses can be managed using aggregated, often anonymized, data (e.g., estimated numbers of displaced persons in order to provide shelter, food, or non-food items), the Ebola outbreak required specific and detailed data. This included data about the health of individuals believed or confirmed to have contracted the disease, and the locations of those individuals with whom sick persons had come into contact. Aggregated information about groups of patients informed everything from district or national-level caseload data, vaccination programs, and epidemiological research about Ebola, to social mobilization efforts, the location of treatment centers, and food security programming. Yet detailed individual-level data also were critical to the response. In the case of contact tracing, a critical component of containing the outbreak, this comprised detailed data about individuals who had come into contact with an Ebola patient and who needed to be monitored for signs of Ebola for the 21-day incubation period. All this data needed to be managed and effectively used. Information about any one individual patient was linked to treatment, laboratory results, contact tracing, and isolation. Each of these sectors’ responder groups required tailored data with different levels of granularity.
In one example concerning the shipment and delivery of commodities, humanitarian logisticians and health facilities in Liberia tracked their requests and shipments using different metrics and standards. Estimates regarding supplies needed to equip medical facilities and health clinics were calculated using patient numbers. By contrast, logisticians tracked shipment of gloves into the country only by carton, with each carton containing various numbers of boxes of gloves, and boxes of varying quantities (e.g., boxes of 100/150/200 gloves). In this case, the aggregation of data that made sense for humanitarian logistics did not correspond to information required to appropriately equip health facilities based on patient levels.
Mapping the Outbreak
Another distinguishing factor of this hybrid emergency was the need to identify and track specific locations--of patients, health facilities, and supplies. In humanitarian responses to natural disasters or armed conflicts, aid agencies usually dispatch large amounts of food and services to a centralized location, such as a village or refugee or internally displaced persons camp. By contrast, in the Ebola outbreak response the food aid to quarantined households, for example, required the precise targeting of food and supplies to a single household, with hundreds of geographically dispersed distributions across the three most-affected countries. Response organizations in the three countries began collecting or using geographic information at various points in time; as a result geographic information may or may not have been associated with datasets.
A lack of readily available and detailed maps, particularly for the remote rural and the densely populated urban areas, compounded this need for timely, up-to-date, and specific geolocation data. In some cases, maps with relevant information, such as health facilities, were not readily available to all responders. This resulted in duplication of effort as multiple entities reproduced the same information. Initiatives such as the Ebola GeoNode and the Humanitarian Data Exchange attempted to alleviate this problem by providing open and widely accessible maps and other datasets.
For geographic information, GIS specialists needed shape files and information about the locations of health facilities, village names, and P-codes for three separate countries. Respondents in all three countries mentioned that village names are often the same or have similar spellings. The use of P-codes made it possible to eliminate data entry errors in cases where villages lacked names, shared names, or had multiple spellings due to local conventions or language. For example, French names for Liberian cities differed from the Liberian version, where Ganta became Gompa and Sanniquellie became Sannicoly. Employing a unique P-code for a village eliminated this problem. One interviewee indicated that the EpiInfo database Viral Hemorrhagic Fever (VHF) module used to track caseload data was customized for Guinea place names and spellings. It was then imported for Liberia, without the same customization. This led to errors in the location information associated with some cases in the VHF module.
Lack of consistent location-specific identifiers also posed a problem in urban areas. In Monrovia and Freetown, which both saw major outbreaks, houses often lacked distinct street names or addresses. Moreover, street names commonly were spelled different ways, and a single street name could be used in multiple locations. As a result, responders could not rely on street or family names to locate suspected or probable cases and their contacts. The lack of specific and accurate geographic information complicated a variety of aspects of the response, such as dispatching contact tracers, burial teams, and even for notifying loved ones of the location of a deceased family member. One responder explained, “One address might be shared with many families. There was some formal system but on the whole, you could have one number with five to six families sharing the same address.” Even after the emergency phase of the response ended, this continued to hamper efforts to notify family members, particularly in cases where place or family names were incorrect or missing.
Inconsistent Use of Unique Identifiers
Inconsistent use of unique identifiers for people also hampered the response within and across the three most-affected countries. Unique identifiers, such as a national identification number or unique code associated with government-provided services, match a specific person with a single identifying code. These identifiers, as well as correct names and addresses, are important in both emergencies and longer-term development--not only for data linkage and research purposes, but also for evaluation, public health programs, health service delivery, policy development and decision-making, and improving and saving lives.
In the countries most affected by the Ebola outbreak, multiple conventions existed to create unique identifiers, which varied depending on the organization and circumstances. For example, in the VHF module, laboratory samples each had a unique “LabID”’ and each patient a unique “EpiID,” making it possible to associate any given patient with multiple laboratory results. In some cases, however, individuals were assigned LabIDs instead of EpiIDs. In all three countries this created difficulties in efficiently or correctly associating a variety of datasets (e.g., laboratory data, patient care data, contact tracing data, burial data) with individual patients. Moreover, at the height of the response, it was not possible to adequately train all individuals collecting patient data about the importance of correctly using unique identifiers. According to one interviewee, “There were many challenges with implementing and not understanding the purpose of unique IDs. For example, we discovered 100 people who had the same unique ID and realized [one health center] had a form with a unique ID and then they photocopied the form. So you had 100 people with the same unique identifier. We only realized later the need to provide clear guidance on how this should be implemented.”
This challenge was compounded by common naming conventions. Many people in the three most-affected countries share first or surnames. Multiple interviewees cited this as further complicating efforts to correctly associate information with the appropriate individuals. In one instance, an interviewee recounted how a common name and a lack of unique identifiers affected contact tracing efforts. Without unique patient identifiers or place names, one responder asked, “How do you differentiate between 10 Mohammed Diallos?” Another pointed to the inverse problem: in many cases patients presented with symptoms at different clinics, resulting in their inclusion in a database multiple times, each time with a different location. In these cases, an individual patient could appear in a database multiple times, possibly with varying spellings or ages. Verifying and cross-checking master lists for this kind of duplication has presented an enormous challenge, yet are essential for using these data for follow-up interventions as well as research. Though compounded in the context of the weak health systems that characterized the Ebola-affected countries, such processes would be challenging even in countries with stronger health information infrastructure.
Tension Between Urgency of Action and Patient Privacy
The requirement to collect detailed individual data ran headlong into the imperative to act quickly during the height of the response, leading to data collection and use being prioritized over the identification and mitigation of individual privacy and security concerns. One health official noted that the “response was managed by getting lab results and CIFs [case investigation forms] to everyone all at once.” In some instances, this resulted in the sharing of information, including personally identifiable information (PII), without safeguards for patient privacy or confidentiality. Many interviewees indicated that responders shared patient information, often including PII, over email accounts or Google documents that were accessible to anyone with the link.
In the Ebola treatment units, personal details were sometimes openly shared out of necessity. For many clinical staff, who are trained to protect patient privacy and dignity, this caused distress. One doctor noted, “We were shouting information over the fence, saying this person has died, this one has diarrhea… This is a violation of patient confidentiality. As a doctor, you really feel this.” In other instances, concerns for patient privacy trumped relatives’ access to information about the location of family members who were taken to treatment centers. The inability to discover where relatives were taken or what happened to them also fed rumors and increased people’s fear and distrust of the response.
The field of CDR analytics for social good is growing rapidly. With CDRs, it has become possible to “follow” and map the movements and interactions of individuals or groups of individuals--or, rather, of their mobile phone or SIM card--to look for patterns and trends, especially when combined with other datasets. At the height of the Ebola epidemic some organizations called for and started exploring the use of mobile phone CDRs to support outbreak response efforts. Their intent was to access real-time, often aggregated, data about population movements in order to better target response efforts, and to enable predictive modeling of the disease’s spread based upon mobility patterns. The analysis of these aggregated data can provide relevant insights for future outbreaks and humanitarian emergencies.
Despite the valuable insights this particular category of “big data” can reveal, concerns about CDR data use remain. These include adequately ensuring individual privacy, even when using anonymized CDRs, establishing the value of CDR data to model and predict the spread of infectious diseases, including Ebola, and developing the appropriate legal and regulatory frameworks for sharing CDR while protecting proprietary data and user privacy.
To date, no coherent and comprehensive set of regulations or guidelines for use of CDR analytics exists. In emergencies, mobile network operators (MNOs) receive multiple and uncoordinated requests for services (e.g., free SMS messages or access to CDR data) and typically respond on a case-by-case basis. Current practice, policies, and legal agreements often are not suited to manage the risks or fully realize the opportunities of using CDR data. Although most agree that CDR analytics must be responsibly used for social good, it is not clear what ethical frameworks should apply, particularly because ethical concerns extend beyond privacy to those of equity and justice.
For example, concerns over the legal ownership--property rights--of CDR data, the importance of user consent, and risk of liability from civil or human rights suits, even under emergency situations, led the Liberian government to decide against releasing CDR data during the emergency. Conscious of the fragile balance between the public value and the privacy risks and concerns of CDR analysis, during the outbreak in Sierra Leone, however, UNICEF, together with UN Global Pulse, worked to negotiate protocols to allow the analysis of aggregated mobility patterns based on CDR data to support the response. They approached the National Ebola Response Committee (NERC), UNMEER, and the Government of Sierra Leone to allow the analysis of mobility aggregations based upon CDR data under certain conditions (e.g., the data remained proprietary and in the control of the mobile network operators, and only mobility aggregations were released for analysis in order to protect user privacy). The next issue, however, was setting up both the infrastructure to create mobility aggregations on the MNO’s premises and the proper channels to share these data with the data science teams of UNICEF and Global Pulse. While some MNOs had internal capacity and experience with aggregating CDRs, others relied upon third parties (such as Real Impact Analytics) to analyze their data for commercial purposes.
Deciding which data aggregations provided value to the response and setting up the infrastructure to create, share, and use them took time. According to Manuel Garcia Herranz, Data Scientist for UNICEF who worked on the CDR analysis, “Over four months we worked with partners to identify what aggregations can bring them value, and defined the privacy, sensitivity and governance protocols that make data-sharing possible while protecting individuals’ privacy. It took another 3 months to set up the capacity, technical infrastructure, secure connections, and initial analysis to inform the response. This is a long time, but it’s a big breakthrough. It was the first time we have been able to get real-time CDR aggregations during an epidemic, and we’ve been able to make a good case for why we need to set up capacity in advance of an epidemic.” As this case demonstrates, using CDR data and aggregations present significant capacity issues as well as appropriate policy and legal hurdles that require negotiation and preparation in advance of emergency operations.
Impact of Regional History and Language
A prominent feature of the Ebola response was its country-specific nature juxtaposed against a broader, regional outbreak. One official involved in the response remarked, “Data came just from a country-specific lens. But Ebola is moving across the region--it is not confined to geo-political boundaries. People and the virus didn’t respect this.” The challenges this presented were particularly apparent along border areas, where the monitoring of outbreaks and population movements required coordination both within and across affected countries. The complexity of border surveillance at numerous sites on land, at sea, and by air, with thousands of individual crossings, required significant coordination across multiple actors as well as capacity building, including training and monitoring of those conducting the surveillance activities. Across the three countries, the International Organization for Migration (IOM) adopted a mostly paper-based system of monitoring the location and direction of people’s movements together with basic demographics. This enabled them to identify hotspots, times, and places of particularly high movement across borders. In most instances, the use of paper--even in implementing one of the largest and longest border screening activities--proved sustainable, efficient, and practical, given the challenges of the rainy season or lack of electricity to charge electronic devices. In Sierra Leone, IOM did use solar-power tablets to collect data, which were uploaded into a central database.
In addition to a history of tension in the region, historical and language differences both across and within the three most-affected countries affected data use and collection. One UN official noted that although the outbreak required a regional response, “Operationally you’re talking about different cultures, coordination mechanisms, and situations.” She continued, “...these are three countries with a history of not getting along very well. This history undermined the collaboration that was needed.” In addition, whereas English is spoken in both Sierra Leone and Liberia, French is the dominant western language in Guinea. Within each country, dozens of local languages and dialects are spoken. This, and the mapping challenges cited earlier meant that it was not possible to simply replicate aspects of the response within or between countries without appropriate contextualization. Social mobilization messages had to be translated into multiple languages and interpreted within a country-specific and cultural context. Taken together, these factors increased the complexity of the response.
“Data came just from a country-specific lens. But Ebola is moving across the region--it is not confined to geo-political boundaries. People and the virus didn’t respect this.”
A Climate of Fear and the Importance of Trust
To stop the spread of a disease like Ebola requires information about the disease and its transmission. Social mobilization and health promotion, therefore, comprised a central pillar of the response. In all three countries, the presence or absence of trust influenced the ability to collect data about individual patients and to affect community behaviors. One government official observed, “There were many [community] leaders who were not convinced. This wasn’t good for the population that didn’t trust us. Up to now there are still people who do not believe in Ebola, who think it’s not true.”
Although trust and trusted networks always play a critical role in data and information flows, the role of trust was accentuated in this context in West Africa, with all three of the most-affected countries recently transitioning from a period of war or civil conflict. This legacy fueled rampant misinformation, rumors, and what became known in Guinea as la réticence--a reluctance, or in some cases opposition, to cooperation with Ebola responders due to the legacy of fear and mistrust of government and foreign actors. One study concluded, “Communities are not uncooperative because they are backward or uneducated. On the contrary, they harbor a distrust of Ebola response efforts that is completely rational, given their experience during recent decades of misrule and political tumult. … Such complex historical circumstances fuel distrust of formal power structures--and Ebola response efforts. Rumors abound that Ebola has been deliberately propagated as a ploy for entrenched interests to pocket money donated for the response.” One international responder in Guinea indicated that in some cases patients refused to provide their names when receiving treatment because of la réticence. Another individual working in Sierra Leone noted that the response was “complicated by Ebola being such an instigator of fear--people don’t want to give information or provide the wrong information, such as the location of house because they don’t want their possessions to be burned.” In the reporting of cases it was unclear from the case investigation forms (CIFs) whether an empty data field was accurate or simply missing due to widespread fear and mistrust. According to a UN official, “You had not just reticence but life-threatening reticence.” He immediately followed this comment with, “You don’t know if it is an empty field [in a form] because there is no risk factor or because it wasn’t filled out.”
In addition, to be effective, the response often had to be tailored to specific areas of a country. One interviewee involved in community engagement efforts in Guinea noted, “If you are meeting people in Haute-Guinée [the north part of the country] they listen to the elder. Elders are the source of wisdom and it is the elder who speaks. In Bas-Guinée [the southern part of the country], the person who is important is the person sending money to buy the rice. We need to make sure we adapt strategy in terms of the source of information.”
‘Building the Plane While Flying It’
Many interviewees discussed the difficulty of having to develop solutions while the emergency was underway. The duration of the response created challenges as well as opportunities for course correction and adaptation. Interviewees pointed to several effects of the outbreak duration on the response. For instance, although the response initially was slow to engage at-risk communities and to consider cultural practices and beliefs, as the response progressed organizations grew in their understanding of the importance of this engagement. This understanding eventually led to the expansion and escalation of health promotion messaging efforts.
Other organizations, as detailed in the case studies of Real-Time Information Flows, used digital technologies to respond to changing circumstances. This allowed them to increase their understanding of the disease’s progression, and to use real-time or near real-time information to make critical programming adaptations. For example, the Red Cross used updated maps of positive and negative Ebola cases visualized on a dashboard to determine where to focus their burial and social mobilization teams, and digital systems to gather daily reporting about everything from personal protective equipment (PPE) to lunch money.
Interviewees repeatedly highlighted the negative effects of frequent international staff turnover, in terms of the demands this placed on national staff leading the response, the negative implications for trusted relationships, and for continuity of the response overall. One NGO reported seven country directors in the space of several months. For CDC staff, the maximum length of stay without additional training was 29 days for one deployment. One national Ministry of Health (MOH) official quickly tired of the constant staff transitions and insisted that the CDC cycle the same people through the Ministry to provide support. The constant staff turnover resulted in a continual disruption to learning and uptake of data collection and management systems. One responder observed that staff turnover meant that international responders “did not see what resulted [from their actions or decisions] or the chronic or recurring issues that weren’t changing.” He continued, observing that the short-term rotations compromised the ability “to see problems through” or follow up on issues.
Examining the physical, socio-cultural, organizational, and historical context of the Ebola outbreak highlights a series of lessons from the Ebola outbreak response, and for health and humanitarian preparedness and response more generally.
- The Ebola outbreak entailed a prolonged yet fast-paced crisis response that enabled innovation and iteration on the one hand, but challenged the response with high turnover of international staff on the other. One interviewee likened the response to “a marathon and not a sprint. It was a long haul and people weren’t ready for it.” With regard to data and information flows, this both enabled adaptation over time, as responders modified their interventions, particularly related to behavior change messaging, and management approach. The prolonged crisis phase also negatively affected continuity of the response due to high frequency of international staff turnover.
- The publication of critical outbreak data in non-machine readable format slowed and narrowed the use of these data by actors who could have helped to contextualize the data and provide insight that could inform the response. This included epidemiologists and researchers who needed access to case data for predictive modeling. More timely data leads to better use of data, which leads to higher quality data, as evidenced by the Ebola Geo-Node and HDX examples.
- The outbreak magnified the shortcomings of existing health systems, including health information systems. Weaknesses in existing health information systems, a foundational component of public health delivery, made it difficult for governments of affected countries to understand health needs, target health interventions, allocate resources, and otherwise efficiently respond to the Ebola outbreak.
- Gaps in access to reliable electricity and/or digital connectivity contributed to significant delays in transmitting time-sensitive data. Temporary, emergency satellite-based communications systems met urgent responder needs but did not address either connectivity needs of the local population or longer term connectivity needs.
- Other ecosystem constraints that affected the collection and use of data included a lack of basic infrastructure, such as roads. Seasonal changes affected connectivity and transmission of timely data as well as the delivery of resources, including non-digital data (such as CIFs and laboratory specimens) to and from rural areas, because the roads were impassible.
- Where digital or physical infrastructure barriers or other constraints made the use of digital technologies impossible, paper-based systems were a reliable alternative--although at the cost of timeliness--particularly when data collection was reduced to the minimum necessary to support operational interventions.
- Information constraints affecting data and information flows included a lack of comprehensive and widely accessible digital maps, and commonly used unique citizen identifiers--elements that are critical to support digitization of data and information as well as health and other service delivery. Non-aligned naming conventions for places and people magnified the effect of the inconsistencies of unique identifiers, while a lack of widely accessible maps of affected areas and non-aligned conventions for place names, for example, stymied efforts to keep electronic records of commodities, health clinics, and traced contacts.
- Addressing privacy concerns for patients and contacts and negotiating agreement about critical datasets, particularly those that include PII in CDR data, take time. These are best addressed as part of preparedness protocols in advance of emergencies.
- The sociocultural context mattered, with delivery of data and information via digital technologies achieving full value only when tailored to take account of variations in local language, customs, cultures, and user context, including literacy and user behavior patterns.
- Effective behavior change and other messaging, whether delivered using digital technologies, word-of-mouth, or other channels proved most effective when delivered through existing affinity networks, by trusted messengers, and when structured to convey empathy. Messages that failed to do so amplified fear and mistrust, leading to unintended consequences, including the hiding of cases, secret burials, and in some extreme cases violence against health and other response workers.
Ebola response behavior change messaging, regardless of the channel through which it was delivered, fared best when it bore empathy in mind. A doctor working with an international NGO in Sierra Leone said, “When we told them ‘don’t touch’ [your loved ones], we failed to humanize something that is quite inhuman. We can’t make the mistake again--of neglecting empathy with those experiencing death in their families, and in their communities.” Where concerns about contagion prevented physical contact, in some cases digital technologies were used to help bridge the divide. In some treatment centers, responders provided mobile phones or tablets to patients in order to be able to connect them with loved ones outside of the treatment units, providing emotional support for patients and demystifying for families what happened inside of treatment units. Trust also played a critical role in the response, with messages being best received when they were delivered by trusted messengers and among existing affinity groups.
 Marc DuBois and others, The Ebola Response in West Africa: Exposing the Politics and Culture of International Aid, HPG Working Paper (London, 2015).
https://www.odi.org/sites/odi.org.uk/files/odi-assets/publications-opinion-files/9903.pdf. See also WHO report by the Ebola Assessment Panel, World Health Organization, Final Report of the Ebola Interim Assessment Panel, prepared by panel of independent experts , July 2015,
http://www.who.int/entity/csr/resources/publications/ebola/ebola-panel-report/en/index.html and World Health Organization, Sixty-eighth World Health Assembly, Ebola Interim Assessment Panel: Report by the Secretariat, in pursuance of resolution EBSS3.R1, A68/25, May 2015,
 Interview with humanitarian NGO official, March 2015.
 Interviews with international responders, November and December 2015, January and February 2016.
 Interview with international responder, December 2015.
 For more about previous Ebola outbreaks, see the following: David L. Heyman et al., “Ebola Hemorrhagic Fever: Lessons from Kikwit, Democratic Republic of the Congo,” Journal of Infectious Diseases 179, no.1 (February 1999): S283-S286, doi: 10.1086/514287,
http://jid.oxfordjournals.org/content/179/Supplement_1/S283.full; Barbara Kerstiëns and Francine Matthys, “Interventions to Control Virus Transmission During an Outbreak of Ebola Hemorrhagic Fever: Experience from Kikwit, Democratic Republic of the Congo, 1995,” Journal of Infectious Diseases 179, no.1 (February 1999): S263-267, doi: 10.1086/514320,
http://jid.oxfordjournals.org/content/179/Supplement_1/S263.full; William Bazeyo et al., “Ebola a Reality of Modern Public Health; Need for Surveillance, Preparedness and Response Training for Health Workers and Other Multidisciplinary Teams: a Case for Uganda,” Pan African Medical Journal., no.20 (April 2015): 404, doi: 10.11604/pamj.2015.20.404.6159, http://www.ncbi.nlm.nih.gov/pubmed/26301008 and Alicia Rosello et al., “Ebola Virus Disease in the Democratic Republic of the Congo, 1976-2014,” eLife, no. 4 (November 2015), doi: 10.7554/eLife.09015, https://elifesciences.org/content/4/e09015/article-info.
 Interview with international health expert, September 2015. Note: Nigeria did repurpose its polio outbreak protocol to quickly contain a single Ebola outbreak that affected the country in 2015.
 Interviews with international responders, January 2015. Although OCHA did provide support to the Ebola-affected countries early in 2014 and many of its staff members were seconded to UNMEER, this was less than would have been the case in a Level 3 emergency. (A “Level 3” emergency is the UN classification for the largest and most severe crises, a designation that triggers the deployment of surge staff and dedicated leadership). The UNMEER Lessons Learned exercise on coordination concluded: “Even if another entity is deployed in the lead of a crisis response, the coordination toolkit of the Office for the Coordination of Humanitarian Affairs still adds value and should be leveraged.” UN General Assembly, Seventieth Session, Agenda item 133, Lessons Learned Exercise on the Coordination Activities of the United Nations Mission for Ebola Emergency Response, prepared by the Secretary-General in pursuance of General Assembly Resolution 69/274B (13), A/70/737, 2016, accessed April 15, 2016, http://reliefweb.int/sites/reliefweb.int/files/resources/N1606127.pdf.
 “UN Mission for Ebola Emergency Response,” Global Ebola Response, http://ebolaresponse.un.org/un-mission-ebola-emergency-response-unmeer (accessed May 18, 2016).
 “Cluster Coordination,” United Nations Office for the Coordination of Humanitarian Affairs, http://www.unocha.org/what-we-do/coordination-tools/cluster-coordination (accessed May 18, 2016) and “What is the Cluster Approach?” Humanitarian Response,
https://www.humanitarianresponse.info/en/coordination/clusters/what-cluster-approach (accessed May 18, 2016).
 In Liberia, the national coordinating body for the response was known as the IMS, Incident Management System; in Sierra Leone as the NERC, National Ebola Response Center; and in Guinea, as the Coordination nationale de lutte contre Ebola (National Coordination for the Fight against Ebola). For more on the national response, see Marc DuBois and others, The Ebola Response in West Africa, 21-23. On Liberia’s response and IMS, see also Tolbert G. Nyenswah et al., “Ebola and Its Control in Liberia, 2014-2015,” Emerging Infectious Disease 22, no. 2 (2016),
 Interview with international responders in Liberia, May 2015 and January 2016.
 Interview with international responders in Liberia, May 2015 and January 2016.
 Interview with international responders, May and December 2015.
 Interviews with NGO responders, February 2016.
 Interview with international responder, January 2016.
 Marc DuBois and others, The Ebola response in West Africa, 27-28. The UNMEER Lessons Learned exercise echoes this conclusion. See UN General Assembly, Seventieth Session, Agenda item 133, “Lessons Learned Exercise on the Coordination Activities of the United Nations Mission for Ebola Emergency Response,” 2016.
 Interview with international responder, December 2015.
 Interview with UN official, May 2015.
 Interview with humanitarian official, January 2015. One USAID official pointed out that UN peacekeeping missions are funded by assessed dues, whereas humanitarian operations are funded with voluntary contributions. This could have influenced the operational culture and how the Mission initially engaged with the public.
 Interview with UN official, May 2015.
 Interviews with international responders, January, March, April, May, and June 2015.
 Interview with Jeremy Konyndyk, January 2015.
 Correspondence with an international health official, August 2016.
 Conversation with international health officials, January 2016.
 Interview with international official, February 2016.
 Email correspondence with health officials, August 2016.
 Interviews with international responders, September 2015 and January 2016.
 “Welcome to Ebola GeoNode,” Ebola GeoNode, http://ebolageonode.org/ (accessed May 18, 2016) and also Stefaan Verhulst and Andrew Young, “Battling Ebola in Sierra Leone: Data Sharing to Improve Crisis Response,” Open Data’s Impact (January 2016), accessed May 18, 2016, http://odimpact.org/case-battling-ebola-in-sierra-leone.html.
 “The Humanitarian Data Exchange,” HDX, https://data.humdata.org. For further discussion, see Stefaan Verhulst and Andrew Young, Battling Ebola in Sierra Leone, 8.
 A P-code is a unique geo-tag, the most precise identifier for a village or other location.
 Interviews with responders, January and February 2015, February 2016.
 Interview with CDC official, February 2016.
 Interview with NGO official, January 2016.
 Interview with CDC official, February 2016.
 Interview with health official, February 2016.
 Interview with international responder, February 2016.
 Interview with international responders, January 2015 and February 2016. As of February 2016, national government, WHO, and CDC officials were in the process of cleaning and verifying the Ebola datasets.
 Interview with health official, February 2016.
 Interviewees with international officials in Sierra Leone and Liberia, October 2015, January and February 2016.
 Interview with medical professional, October 2015. A similar sentiment was expressed in another interview with a medical professional, February 2016.
 Interview with international responder, December 2015.
 Emmanuel Letouzé and Patrick Vinck. “The Law, Politics and Ethics of Cell Phone Data Analytics.” Data-Pop Alliance White Paper Series. Data-Pop Alliance, World Bank Group, Harvard Humanitarian Initiative, MIT Media Lab and Overseas Development Institute. April 2015. http://datapopalliance.org/wp-content/uploads/2015/04/WPS_LawPoliticsEthicsCellPhoneDataAnalytics.pdf. See also Christopher Fabian, Mobile Technology in Emergencies: Principles and Practice. Developing TeleComs: Connected Citizens – Managing Crisis. August 2015, 21-26. http://www.unicefstories.org/wp-content/uploads/2013/08/devtelecoms-connected-citizens-aug-2015.pdf.
 Sean Martin McDonald, Ebola: A Big Data Disaster, CIS papers (Delhi: The Center for Internet and Society, 2016), 3, http://cis-india.org/papers/ebola-a-big-data-disaster.
 Interview with USG official, February 2015.
 Interview with IOM officials, February 2016. The IOM program in Sierra Leone continues as a regional effort in partnership with the US CDC, under the Global Health Security Agenda.
 Interview with UN official, May 2015.
 The CIA World Factbook lists some 27 ethnic group languages in Liberia, Guinea, and Sierra Leone. The World Factbook, Central Intelligence Agency, accessed April 26, 2016, https://www.cia.gov/library/publications/resources/the-world-factbook/index.html.
 Each country identified different pillars. For the health components, these included case management and isolation, safe burial, contact tracing, social mobilization, and infection prevention and control. Interview with USG official, December 2015.
 Interview with national government official, May 2015.
 One health data report observed that “data exchange moves at the speed of trust.” Robert Wood Johnson Foundation, “Data for Health: Learning What Works,” A report from the Data for Health Advisory Committee, April 2015, 5, accessed August 8, 2016, http://www.rwjf.org/content/dam/farm/reports/reports/2015/rwjf418628.
 While the term applies only to Guinea, community distrust of Ebola responders and messages appeared in all three countries.
 Ranu S. Dhillon, and J. Daniel Kelly, “Community Trust and the Ebola Endgame,” New England Journal of Medicine 373, no. 9 (2015):788, http://www.nejm.org/doi/pdf/10.1056/NEJMp1508413. See also Ashoka Mukpo, Surviving Ebola: Public Perception of Governance and the Outbreak Response in Liberia (London: International Alert, 2015), accessed April 26, 2016, http://www.international-alert.org/sites/default/files/Liberia_SurvivingEbola_EN_2015.pdf.
 Interview with NGO official, February 2016.
 Interview with international official, January 2015.
 Interview with UN official, February 2016.
 Interview with NGO official, February 2016
 The ODI report observed: “Ebola exposed the dangers of not getting it right when it came to engagement with the local community: into an atmosphere of intense distrust and fear, early efforts inserted news of an incurable killer disease, using foreigners (international and national) to tell remote villages about Ebola without telling them what they could do about it, and PPE-clad teams removing villagers who were never seen again. The resulting amplification of distrust, fear and resistance all boosted rather than reduced transmission.” Marc DuBois and others, The Ebola Response in West Africa, 34. See also ACAPS, Ebola Outbreak, Sierra Leone: Communication: Challenges and Good Practices, ALNAP (2015), accessed April 20, 2016, http://reliefweb.int/sites/reliefweb.int/files/resources/s-sierra-leone-ebola-outbreak-communications-challenges-and-good-practices.pdf.
 Interview with Amanda McClelland, December 2015.
 Interviews with various responders, January and September 2015, January and February 2016. Multiple interviewees also mentioned the difficulty in recruiting experienced professionals to respond, especially in the very early days of the crisis.
 Interview with NGO official, May 2015.
 Interview with CDC official, February 2016.
 Interview with international health official, January 2016.
 Interview with international responder, September 2015.