Figure 1 (click for full size)
Edward Swing, VSTI, a SAS Company, Ed.Swing@vsticorp.com [PRIMARY Contact]
Kevin Boone, VSTI, a SAS Company, Kevin.Boone@vsticorp.com
Wesley Mann, VSTI, a SAS Company, Wesley.Mann@vsticorp.com
SAS Institute's Base SAS 9.2 was used to prepare the microblogs dataset for analysis in the GIS software. Approximately six hours were needed to create and execute the required programs in Base SAS 9.2, specifically to achieve the following:
Esri's ArcGIS 10.0 was used for visual analysis of the outbreak over location and time. Approximately 80 hours were needed to complete the analysis and arrive at a conclusion (presented in the answers section). ArcGIS 10.0 was utilized to perform these tasks:
MC 1.1 Origin and Epidemic Spread: Identify approximately where the outbreak started on the map (ground zero location). If possible, outline the affected area. Explain how you arrived at your conclusion.
Microblog records were analyzed to determine if an epidemic occurred in Vastopolis. The data included over 28,000 records that indicated someone had at least one of several symptoms, which were categorized into stomach problems (stomachache, nausea, diarrhea, or vomiting) and flu-like (flu, fever, pneumonia, or shortness of breath). Analysis on a daily basis to see how symptoms spread over time revealed two outbreaks: stomach problems in the southwest starting on May 19 (figure 1) and flu-like symptoms concentrated in the city-center starting on May 18 (figure 2). These resulted from a truck accident on the I-610 Bridge on May 17. It closed the road for ten hours and included two public health threats: a fire that spread flu symptoms east with the prevailing wind, and a hazardous materials spill into the Vast River that caused stomach problems downstream.
Figure 1 (click for full size)
Figure 2 (click for full size)
MC 1.2 Epidemic Spread: Present a hypothesis on how the infection is being transmitted. For example, is the method of transmission person-to-person, airborne, waterborne, or something else? Identify the trends that support your hypothesis. Is the outbreak contained? Is it necessary for emergency management personnel to deploy treatment resources outside the affected area? Explain your reasoning.
Process
The analysis process first attempted to establish the pattern of the outbreak over location and time, and then focused on determination of the source and/or “ground zero” location of the outbreak. With the established knowledge that two different categories of symptoms, flu-like and stomach problems, existed, both categories of symptoms were viewed separately on a day-by-day basis in ArcGIS 10.0. The software’s time slider tool made it easy to a) visualize the “sick” records for any given day, and b) step forward or backward one day at a time to see patterns of change.
Since flu-like symptoms are usually transmitted from person-to-person, a visual illustration of how people move through Vastopolis on a daily basis was also created to aid in the analysis. Figure 3 shows the daily change between daytime and nighttime population counts in each of Vastopolis’ thirteen zones. It is evident that the central zones (Uptown, Downtown, and Westside) have larger populations during the day, and that most people reside in the surrounding areas, especially south and east, at night.
Figure 3 (click for full size)
Flu symptoms
The day-by-day visualization showed few flu-like reports from April 30 to May 17. May 18 was the first day with significant numbers of reports. May 19 through 20 also reported many cases, but not enough to establish a pattern of dispersion. To further establish a pattern, the analyst visualized the flu-like reports in six hour intervals starting at midnight on May 18. The first reports happened overnight in the downtown and uptown areas, and continued in those locations as well as started to spread east from there until 6:00pm (see figure 4).
Figure 4 (click for full size)
The areas surrounding the city center started to include many more reports after 6:00pm on the 18th, while uptown and downtown continued to show many cases, as shown in figure 5. By May 20 there was a widespread epidemic, with most reports centered in the downtown/uptown zones and a few miles east of there, as shown in figure 6. Many people went to the hospital with these symptoms on May 20, illustrated by the small clusters of dots in the green ovals on figure 6.
Figure 5 (click for full size)
Figure 6 (click for full size)
Winds were blowing to the east on May 17 and 18, so any airborne germs would have primarily gone in that direction. Winds shifted slightly by May 20, blowing to the southeast.
Results of this visual analysis show that the flu-like symptoms were spread as an airborne contaminant and via human-to-human transfer.
Stomach Symptoms
The reports of stomach-related problems were very few until May 19. On that date many instances of stomach issues occurred in the southwest of the Vastopolis metro area, most within 1.5 miles of the Vast River. The number of reports quadrupled by May 20, all in the same location. It was determined that over 98% of the reported stomach problems occurred in the same area over those two dates. Figure 7 illustrates the outbreak of these symptoms.
Figure 7 (click for full size)
The visual analysis on the map indicate that stomach problems are being spread in epidemic fashion as a waterborne contamination in the Vast River, one of the main water supplies for Vastopolis.
Conclusions
All of the stomachache records are on one side of the I-610 Bridge over the Vast River, which flows in a southerly direction. Therefore the analyst searched the dataset for any records that mentioned “610”, and quickly found about 140 messages indicating a truck accident on the bridge on May 17. The accident stopped traffic for over ten hours, according to the timestamps in the relevant messages. Other messages indicated that the accident included a fire. It is highly likely that this catastrophe also led to hazardous materials (cargo and/or fuel) leaking into the river, thus contaminating the water supply and starting the outbreak of stomach problems downstream two days later. Continued waterborne spread is likely, as the pattern of dissemination is not contracting.
Further examination indicated that the truck accident was responsible for the outbreak of flu-like symptoms as well. That outbreak started directly east of the accident site, soon after the wreck was first reported. From there, the epidemic spread as previously discussed, in an airborne manner with the winds and via human-to-human transfer. This epidemic also appears to still be spreading rampantly.
Since the outbreaks were continuing to spread at the end of the time period covered by the data, Vastopolis managers should warn other surrounding regions about the problems. The waterborne contamination must be cleaned immediately. Hospitals and other healthcare providers should quickly determine if it is truly a flu epidemic or another sickness with similar symptoms, and be prepared to treat thousands of patients accordingly.
Other Explanations
The analyst also examined the data for other events that possibly caused or contributed to the spread of the outbreaks. Terms such as alarm, fire, explosion, gas, convention center, court, stadium, and others were searched for in the messages, and visualized on the map when warranted. If a search term yielded several hits that were contextually valid, their position on the map was compared to the patterns of dispersion of each outbreak. Noteworthy events included a fire in the vicinity of the Capital on May 14, and a chemical plant explosion in the southwest of Vastopolis on May 17. The timing and patterns of symptom origin and dispersion ruled these out as source or contributing events, however, and no alternative events were found to have affected the epidemics.
The truck accident on I-610 on May 17 led to airborne and waterborne hazardous materials contaminations. These caused epidemic outbreaks of flu-like symptoms and stomach problems that started on May 18 and 19, respectively. The outbreaks continued to spread throughout May 20, the last day for which data exists.