Of 109 WHO Disease Outbreak News reports from 2010 through 2014 not titled as updates, we identified 73 (67%) that fit our inclusion criteria, of which 66 (61%) had reportable start dates. The geographic distribution of outbreaks was similar to that of the 281 outbreaks from 1996 through 2009. For all 347 outbreaks (Table 2) combined that fit the selection criteria and had known start dates, 54% were from Africa, 11% from the Western Pacific, 10% from the Eastern Mediterranean, 10% from the Americas, 7% from Europe, and 7% from South-East Asia. Change in HDI rank was stratified as −58 to −34; −33 to −14; −13 to −3; and −2 to +22. Polity score quartiles were stratified as −10 to −3; −2 to 0; 1 to 7; and 8 to 10. HDI change values were missing for 21% of outbreaks, and polity scores were missing for 11% of outbreaks.

Our primary outcomes of discovery delay and communication delay for all outbreaks from 1996 through 2014 had median times of 20 days (95% CI 16–25 days) and 32 days (95% CI 29–38 days), respectively. Median time to laboratory confirmation was 36 days (95% CI 32–46 days) and to WHO verification was 49 days (95% CI 44–55 days). These outcomes varied by WHO region, change in HDI quartile, and country-specific polity quartile in the year of the outbreak, and across time. For both primary milestones, median times were longest for Africa and the Eastern Mediterranean and shortest for the Western Pacific and South-East Asia (Table 3). Median discovery delay decreased monotonically with improving HDI rank quartile. Communication delay was uniform across the HDI variable aside from a large increase for the quartile with the greatest decline in HDI rank (Table 4). Polity showed a less clear trend; time lags increased with decreasing polity from the first through third polity quartiles, and then decreased for the quartile of lowest polity (Table 5).

Figure 1

Figure 1. Scatterplots with Loess curves of time to A) outbreak discovery and B) public communication in a study assessing global capacity for emerging infectious disease detection, 1996–2014. Gray shading around curve indicates...

Figure 2

Figure 2. Loess curves of time to A) outbreak discovery and B) public communication, by World Health Organization region, in a study assessing global capacity for emerging infectious disease detection, 1996–2014. Dashed line...

Figure 3

Figure 3. Loess curves of time to A) outbreak discovery and B) public communication, by quartile of change in Human Development Index rank, in a study assessing global capacity for emerging infectious disease...

Discovery and communication delays generally decreased over time but exhibited large fluctuations and substantial uncertainty because of the small number of outbreaks (Figure 1). Descriptive regional Loess curves show an overall trend of shortened discovery delay for each region but no clear trend for communication delay (Figure 2). However, the occurrence of extreme lags, defined as outliers relative to all recorded lags (>102-day and >117-day discovery and communication delays, respectively), has declined over time; the average annual number of extreme discovery delays decreased from 2.5 to 0 after 2007, and the average annual number of extreme communication delays decreased from 2.4 to 0.75. Our HDI-quartile Loess curves illustrate that progress in surveillance systems might vary according to changes in country-level human development metrics (Figure 3).

Cox regression estimates show that discovery and communication delays significantly decreased over time, with an average annual reduction of 5.5% in discovery delay (HR 1.055, 95% CI 1.034–1.076) and 2.5% in communication delay (HR 1.025, 95% CI 1.006–1.045). Days to laboratory confirmation, available for 82% of outbreaks, decreased by an average of 3.6% per year (HR 1.036, 95% CI 1.014−1.060), and days to WHO verification decreased by an average of 2.1% per year (HR 1.021, 95% CI 1.001–1.041).

The Eastern Mediterranean and Western Pacific regions showed the greatest improvements in discovery and communication delays, whereas Europe showed the least improvement in discovery delay, and Africa, Europe, South-East Asia, and the Americas all showed little improvement in communication delay (Table 6). When evaluated according to quartile of change in HDI rank, the 2 middle HDI quartiles exhibited the greatest improvement for both indicators (Table 7).

A Cox regression model with an indicator for whether the outbreak occurred before or after June 2007, when the WHO’s revised IHR were implemented, shows an 84% reduction in discovery delay after IHR implementation (HR 1.84, 95% CI 1.44–2.35) but no significant change in communication delay (HR 1.18, 95% CI 0.93–1.49). However, our sensitivity analysis comparing the reduction in discovery delay before and after June of other years shows that improvement peaked in 2005, 2 years before the IHR were implemented (HR 2.01, 95% CI 1.60–2.53).

Our results confirm that improved timeliness of outbreak discovery found by Chen et al. (13) has generally been sustained, but with smaller improvements since 2010. Median discovery delay and communication delay fluctuated over time; however, after we stratified by WHO region and HDI rank change, clear downward trends appear for most strata. We have also seen fewer extreme delays in recent years. Furthermore, all regional Cox regression models showed increases in timeliness. Although some were not statistically significant, partly because of the small numbers of outbreaks, we visualize the downward trend with Loess curves.

Because surveillance capabilities vary by region, yearly regional distribution of outbreaks has a strong influence on aggregate global outcomes. Although the majority of outbreaks in most years originated in Africa, where outbreak discovery tends to be slower (Table 3), there were years in which another region also had a large number of outbreaks, albeit fewer than Africa (e.g., Eastern Mediterranean in 2006 and Western Pacific in 2008), and years in which another region experienced more outbreaks than Africa (e.g., Western Pacific in 2007).

Despite regional disparities in timeliness of disease discovery, all regions showed some improvement. Variability between regions probably resulted from a combination of differences in culture, Internet availability, and previous outbreak experiences. Public communication must follow disease discovery, and additionally depends on recognition of outbreak severity, willingness to expose the outbreak, and belief that outbreak control will improve after the announcement. The 1996 start date of our study was during the early days of Internet-based surveillance, but some countries such as the United States, France, England, Norway, Scotland, and Sweden were already using digital technologies for centralizing and aggregating data (29–32), which set them ahead of the surveillance curve but left them with less room for improvement over the 18-year study period. We expect that local phenomena, such as the proliferation of Internet usage or the occurrence of particularly alarming regional outbreaks, were associated with local surveillance improvements, whereas global phenomena, such as the SARS epidemic and the adoption of the IHR (2005), encouraged more widespread surveillance system development.

Our analyses stratified by HDI and polity shed light on the role of government characteristics in surveillance system strength. Change in HDI rank from 1990 to 2013 captures a sustained movement toward improved (or stymied) development. Polity in the year of the outbreak, however, illustrates a time-constrained snapshot of government status. Although our analysis across levels of polity did not reveal noteworthy trends (Table 8), we found that the direction of change in HDI over time had a strong relationship with surveillance system success. Median discovery delay increased with worsening HDI quartile. Outbreaks in countries with the most positive change in HDI rank had the shortest median time to detection but the least reduction in delay (consistent with having the least room for improvement). The middle quartiles improved significantly, whereas the countries with the greatest decrease in HDI, or the most political and social instability, showed less improvement. Meanwhile, median communication delay was uniform across the HDI variable, aside from the lowest quartile, which had a significantly longer delay. That quartile was also the only quartile to show no improvement over time, although this was not significant (Figure 3, panel B). Although these results suggest that long-term national systemic improvements might foster surveillance system progress, it might be practical to begin with increased availability of and education regarding moderated digital disease reporting platforms. Additionally, international agencies can encourage surveillance efforts by bolstering consensus regarding transparency and developing improved mechanisms to enforce adherence to existing regulations.

Our finding of sustained improvement in timeliness of outbreak discovery for all regions in the current IHR period is encouraging. We can see that improvements in the lead up to adoption of the revised IHR have been maintained with some year-specific fluctuation, and extreme delays have been eliminated. There are likely several drivers behind these improvements, including enhancement of local surveillance systems to meet the requirements of the revised IHR, and continued development of informal surveillance sources such as GPHIN, ProMED, and HealthMap (33), and almost universal increases in human development scores.

We found that global and stratified regional trends were attenuated relative to our findings from 5 years ago, suggesting that progress has slowed. This finding is expected, given that there is now less room for improvement; however, different strategies will now be required to achieve further success. Initial drops in discovery and communication delays might have been partly attributable to the availability of digital surveillance technology, but further progress will require developments in local surveillance infrastructure. These include bolstering leadership and coordination, increasing access to medical care, and building trust in health systems, all costly and complex endeavors (34). Increased international collaboration can help diffuse the effort required for these undertakings (35).

We should note that the relatively small improvements in time to laboratory confirmation and WHO verification are not unexpected. Speed of laboratory testing varies considerably by disease and geography because of availability of laboratory capacity, particularly in the case of an unusual or nonendemic disease. Furthermore, because WHO verification generally requires laboratory confirmation and consent of national authorities, this announcement might also be delayed. Given that 33% of first public communications were WHO reports, it is not surprising that the timeliness of WHO report closely reflects the trend in timeliness of public communication.

Despite the general progress in timeliness of disease discovery, we find that there is room for improvement in timely public communication about outbreaks. Although the aggregate Cox regression shows an average improvement of 2.4% per year, we see that Africa and Europe barely improved at all, whereas South-East Asia and the Eastern Mediterranean improved initially but seem to have slowed in later years. Although detection of an outbreak triggers the initial alert for the surveillance system, it is public communication that will elicit a local response among the susceptible population and an international response among those governments and agencies that are equipped to assist in outbreak containment. In 2003, the first reports of SARS occurred >2 months after the first cases were discovered, after >300 people in China were already infected (36). In Saudi Arabia, Middle East respiratory syndrome was not announced until ≈3 months after the first symptomatic patient sought medical attention because of a lack of a definitive laboratory diagnosis (37). Similarly, the first cases of Ebola in Guinea in 2014 were not initially diagnosed, causing a delay of several months before the government recognized the outbreak of hemorrhagic fever and published an announcement (38). This delay provided time for the disease to spread to the large capital city of Conakry before control measures could be taken, at which point it became an unprecedented challenge to combat the outbreak (39).

The results of our analysis are subject to a few limitations. The small annual numbers of eligible outbreaks prevented us from stratifying our analyses by factors that might be associated with surveillance capacity, such as disease type (e.g., respiratory and hemorrhagic) or country-specific statistics on Internet usage. Additionally, many confounders or modifiers of interest, such as outbreak severity and size, local concurrent disease burden, and health system strength, are not available as validated metrics for analysis. Regions with the least room for improvement would have benefited by further stratification as a means to identify focus areas. We hope that future research will shed light on some of the more intricate heterogeneities in surveillance success around the world. The analysis was also limited by a large number of unreported outbreak start and discovery dates. We followed estimation procedures used by Chan et al. (13) to maximize the number of outbreaks we could include for our analysis, recognizing that our estimated dates were not entirely accurate. This approximation likely led to conservative estimates (i.e., overestimated times to surveillance milestones), with a stable degree of misclassification over time, thus introducing minimal bias on our final hazard ratios.

Our findings illustrate the general improvement in timeliness of outbreak discovery and the need for further improvement in timeliness of public communication. However, it is important that our conclusions be understood in the context of disease- and region-level heterogeneity. Although our data describe a single measure of progress in disease surveillance, we hope that our findings are considered in light of other research on contextually appropriate measures and indicators. We highlight the importance of international efforts to enforce regulations, identify regional strengths and weaknesses, and set appropriate goals for surveillance system strengthening, so that when an outbreak does occur, control measures can be set in motion quickly enough to avoid local and potentially pandemic disease spread.