We undertook a retrospective study of responses to all substantiated disease outbreaks reported to WHO by member states in the African Region during 2017–2019. Outbreaks were reported using the Integrated Disease Surveillance and Response (IDSR) strategy. A substantiated outbreak was defined as one whose hazard was confirmed or in which the occurrence of human cases was clearly in excess of normal expectancy; that is, the epidemic threshold was reached or surpassed. The substantiated disease outbreaks constituted a subset of all the public health events reported to WHO, which also encompasses humanitarian and other health emergencies.

The primary data source was the public health event database maintained by the WHE Programme at the Regional Office for Africa, which contains all formally reported public health events in the African Region including those verified through epidemic intelligence activities. Variables captured include country name, event name, etiology, date of onset, date of detection, date of notification to WHO, date of end of outbreak, and total number of cases and deaths, as well as a short description of events. Two additional variables were derived from the public health event database: each country reporting an outbreak was assigned to the Central, West, Eastern, or Southern African subregion on the basis of the WHO intercountry support team structure (10), and disease category was assigned on the basis of either similarity in mode of transmission, type of etiologic agent, or diseases requiring the same type of public health intervention (e.g., vaccination).

We used secondary sources to complete missing data, including the WHO event management system, IDSR weekly bulletins produced by member states in the African Region, outbreak investigation and situation reports submitted to WHO via email, published articles in peer-reviewed scientific journals, and the World Development Indicators database. The event management system is an online central repository for all globally reported events that may constitute a public health risk to countries through the international spread of disease or that may require a coordinated international response as required by IHR 2005 regulations (11). The World Development Indicators database is the World Bank’s compilation of cross-country data on development, providing internationally comparable statistics on global development and the fight against poverty (12). We extracted data from the database for 2017–2019 for selected variables that could influence our outcomes of interest (13): income level, population density, health expenditure as a percentage of gross domestic product, the percentage of internally displaced population, and the refugee population living in the countries in which the outbreaks occurred. We chose these predictor variables guided by available theoretical and empirical literature to determine how well our results agree with those of other published studies concerning these predictors and also to identify what is unique to the African Region.

Humanitarian emergencies and other public health events reported to WHO that did not constitute a disease outbreak were initially excluded because we could not estimate key milestone dates and their end dates did not necessarily depend on timely implementation of public health response given the context in which they occurred. Only substantiated disease outbreaks were selected for the study. Before analyzing the outcomes of interest, for each variable we discarded data from any outbreaks missing key milestone dates, while retaining data from that outbreak in other analyses that did not involve variables with missing data (14). We calculated the proportion of outbreaks sampled across each variable subcategory (e.g., by income level) and compared them with the proportion in all outbreaks to ensure representativeness and make inferences.

We adapted the definition of milestones used by Chan et al. (15). We chose 4 key dates from each outbreak to determine milestones for the analysis: dates of onset, detection, notification, and end of outbreak. For date of onset we used the reported date of symptom onset for the first case found by the investigators. For date of detection we used the date on which national authorities were alerted to the outbreak. For date of notification we used the date the event was first reported to WHO by a member state. For date of end we used either the date when a country declared an outbreak to have ended, or a defined length of time over which no new cases were reported. As a proxy, we used twice the maximum incubation period from the date of recovery or death for the last case. We chose this definition to ensure that disease transmission was no longer occurring and the outbreak was over.

We measured time to detection as the number of days between the dates of onset (first reported case) and detection, time to notification as the number of days between the dates of detection and notification, and time to end as the number of days between the dates of onset and end of the outbreak. A total of 29 events that began but had not ended during the study period were right censored from the study on December 31, 2019.

We performed univariable and multivariable Cox proportional hazards regression analyses in 3 separate models to assess the rates of change over time to outbreak detection, notification, and end. Predictor variables were outbreak start year, WHO African subregion, income, number of refugees, percentage of internally displaced population, current health expenditure as percentage of gross domestic product, population density, and disease category. Each outbreak milestone was treated independently.

We computed the median time and interquartile range (IQR) for the time to reach each of the milestones for all the outbreaks included in the study, then stratified results by year of outbreak start, WHO Africa subregion (Central, West, Eastern, or Southern), and disease category (food/waterborne, vaccine-preventable, vectorborne, viral hemorrhagic fever, or others). We used the World Bank’s classification of countries’ economies (16) to stratify results by income level (low vs. middle and high). We used median values (i.e., in-country value for each variable compared with median value for that variable for all African Region countries) to stratify results by number of refugees (low or high), percentage of internally displaced population (low or high), current health expenditure as percentage of gross domestic product (low or high), and population density (low or high). The results are indicated as hazard ratios (HR) with 95% CIs. An HR >1 indicates improvement in the time to each of the milestones, whereas an HR <1 signals a regression in the time to each milestone. R version 3.5.2 (https://www.r-project.org) was used for statistical analysis and ESRI ArcGIS Desktop 10.6.1 (https://desktop.arcgis.com) for mapping of outbreaks included in our study.

Figure 1

Figure 1. Exclusion criteria used to select subset of substantiated disease outbreaks reported to the WHO African Region, 2017–2019. WHO, World Health Organization.

Of the 348 substantiated public health events reported to WHO in the African Region from 2017 to 2019, we selected 296 disease outbreaks for the study. Key milestone dates were often missing, and therefore we only included 184 (62%) events for time to detection, 232 (78%) events for time to notification, and 201 (68%) events for time to end (Figure 1). The percentage of events included in the study for each outcome of interest slightly decreased from 2017 through 2019, after exclusion of outbreaks with missing data, across each year (Tables 1, 2).

Figure 2

Figure 2. Geographic distribution of substantiated disease outbreaks selected in study of timeliness of key outbreak milestones in the WHO African Region, 2017–2019. CCHF, Crimean-Congo hemorrhagic fever; WHO, World Health Organization. cVDPV2,...

Events meeting our selection criteria occurred in 41 out of 47 WHO member states in the African Region. Uganda registered the highest number of events (n = 21); The Gambia, Rwanda, and Seychelles registered the fewest (n = 1). São Tomé and Príncipe, Guinea-Bissau, Gabon, Equatorial Guinea, Eritrea, and Eswatini did not report any disease outbreak with a start date during the study period (Figure 2). Cholera (18.6%) was the most frequently reported outbreak disease, followed by measles (11.1%), dengue fever (7.8%), Crimean-Congo hemorrhagic fever (7.4%), poliomyelitis (5.7%), and meningococcal disease (5.4%) (Table 2).

Early detection and rapid response to outbreaks greatly contributed to reducing the illness and death rates during these events. Timeliness of response is an essential element of surveillance systems (17,18). Systematically capturing and analyzing the timeliness of key milestones in outbreak detection and response can also provide critical information to public health decision makers and stakeholders on progress made towards implementing IHR requirements. Our analyses explored trends in the timeliness of detection, notification, and end of outbreaks over 3 years (2017–2019) in the WHO African Region and predictor variables that could explain changes. Overall, the findings showed a decrease in the median time to detection and end of outbreaks over the years, signaling an improvement in capacities in these areas. In contrast, the median time to notification increased.

Our finding of overall improvement in time to detection of outbreaks in the African Region was consistent with 2 other studies, although the period studied, geography, and variables analyzed differed (15,19). In our study, the disease category was associated with differences in time to detection. Time to clinical diagnosis and laboratory confirmation of a disease affect time to detection. The median time to detection of food/waterborne disease outbreaks was found to be shorter than for other disease categories. Our final dataset contained a high number of food/waterborne diseases, with outbreaks of cholera the most frequent. Cholera has a relatively short incubation period compared with other diseases and the potential to spread rapidly, causing large-scale outbreaks (20). The main approach to clinical diagnosis and surveillance of cholera has been the use of the case definitions (21) contained in the IDSR guidelines. The scale-up of IDSR in recent years (22), coupled with the short incubation period, contribute to early detection of outbreaks of cholera. In addition, laboratory capacities for confirming Vibrio cholerae, particularly in cholera-prone settings, have improved (23), reducing the time needed to confirm outbreaks. However, challenges remain, as surveillance and diagnostic capacities across member states vary and serious gaps exist (24).

The lower hazard ratio for time to detection of vaccine-preventable diseases compared with food/waterborne diseases could be explained in the context of the outbreak threshold for these diseases and how they are monitored. The IDSR outbreak threshold for cholera is one confirmed case while most vaccine-preventable diseases require several cases over a period of time to reach the epidemic threshold (21). Cholera outbreaks, therefore, are more likely to generate a quick alarm. Furthermore, monitoring of diseases reaching an outbreak threshold in the region continues to be performed mainly through indicator-based surveillance that relies on structured weekly IDSR reports from health facilities. This structure could likely result in delays in early detection of vaccine-preventable disease outbreaks because this system captures only those cases from people visiting a health facility. Event-based surveillance systems are designed to address this limitation by capturing reports from a wide variety of sources. For example, introducing community event-based surveillance contributed to increased reporting of suspected cases of measles in Liberia, leading to early detection of measles outbreaks (25). However, event-based surveillance is not yet well developed in most countries in the region and there remain major gaps in its implementation (22).

No other predictor variable had a notable association with time to detection. Differences in national income levels and health expenditures by country GDP could provide plausible explanations for variations in the efficiency of health systems (26), with higher expenditures possibly leading to capacities gained for detecting outbreaks faster, even though we did not find any significant association (p>0.05) with the time to detection in this study. Also, delayed detection in the Central African subregion could be partly attributed to protracted humanitarian crises resulting from armed conflicts that have plagued many countries in this subregion (27,28). Notwithstanding, our study also did not find a significant association (p>0.05).

Other than outbreak start year, none of the predictor variables had a statistically significant association (p>0.05) with the time to notification. Rapid notification is aimed at preventing or limiting the international spread of diseases and avoiding disruption to international trade (29). Under this guideline, member states need to notify WHO within 24 hours of assessment of any event in their territory “that may constitute a public health emergency of international concern” (6). Reversing the increase over the years in time to notification will require continuous engagement with member states and capacity building for the national IHR focal points. Initiatives such as the open-access online “Global Health Security, Solidarity, and Sustainability through the International Health Regulations” course (30) may help enhance this capacity.

Many countries in the WHO African Region continue to experience recurrent disease outbreaks. Lessons gained from each outbreak response have helped to improve response to subsequent outbreaks, a possible reason for the reduction in time to end. The lower HR for controlling outbreaks of vaccine-preventable diseases compared with food/waterborne disease outbreaks may reflect challenges associated with vaccine acquisition, uptake, and access, particularly in hard-to-reach areas.

Our list of predictor variables was not exhaustive and several other factors may have influenced the improvement in time to detection and time to end, for example, the roles of WHE and other high-level initiatives and partnerships. WHE’s enhanced use of a digital disease detection approach through media monitoring platforms such as Hazard Detection and Risk Assessment and Epidemic Intelligence from Open Sources (31–33) has expanded the window of opportunity for capturing event-based surveillance information. In 2018, about one quarter of the events in the African Region were verified by member states and reported to WHO following media monitoring activities undertaken at the Regional Office for Africa (34). In addition, several other high-level initiatives have been undertaken in the aftermath of the 2014–2016 Ebola outbreak to bolster global capacities for preparedness and response to infectious disease outbreaks (35).

There are 3 main limitations of the study. First, key milestone dates were missing in some of the reports sent to WHO, which excluded some outbreaks from the analysis. This omission may have had some effect on the findings, although efforts were made to ensure that the sample was representative of the study population of events (Appendix). Second, the number of outbreaks considered in our analysis is not an exhaustive list of all the outbreaks that occurred in the region. Only those reported to WHO by member states on the basis of IHR (2005) requirements, using the Annex 2 decision instrument (36), were included in the study. Third, the application of right censoring at the end of the study may have affected the results, especially for time to end of outbreaks in 2019, although the difference is unlikely to be substantial because relatively few events were right censored.

In spite of these limitations, our study shows that outbreak metrics can be collected and measured to allow countries to monitor the timeliness of outbreak detection and response and provides an estimate of improvement of outbreak detection and end over time. The next step is to enable countries to set up systems in which these measurements are routinely collected, analyzed, and used to improve surveillance and response interventions.

In conclusion, our study has established that the use of simple, easy-to-collect, and verifiable metrics is key to monitoring the timeliness of outbreak detection and response in the WHO African Region. The findings of improvement in early outbreak detection and rapid control over the studied period should be interpreted within the context of the variations and multiple factors that influence these outcomes. Further studies could shed better light on these variations within the member states, and across subregions and disease categories. The momentum needs to be sustained and member states supported in building capacities for early detection, notification, and rapid control of outbreaks. It is equally necessary to support member states to enable them to track key milestones systematically for continuous measurement of outbreak response performance.