Students

Nothing to sniffle at: Saving lives with software

Chaos in clinics" was what one TV news broadcast called 2009's shortage of H1N1 vaccine. Once vaccine did start trickling into the supply chain, it was up to county officials to decide which healthcare providers would get the few doses available, and those decisions had to be made on the fly. In Arizona's Maricopa County, public-health officials turned to Information Systems researchers from the W. P. Carey School of Business for a decision-making tool -- a software program -- that would help them distribute vaccines in an equitable and rational way.

2010-07-02

"Chaos in clinics" was what one TV news broadcast called 2009's shortage of H1N1 vaccine. For months, healthcare experts had warned the public that taking a shot in the arm -- or a snort of nasal-spray vaccine -- was the best defense against a coming and deadly swine flu. By late October, some 3,000 people already had died of the virus, and the vaccine to prevent it was still in short supply due to manufacturing delays. Once vaccine did start trickling into the supply chain, it was up to county officials to decide which healthcare providers would get the few doses available, and those decisions had to be made on the fly. In Arizona's Maricopa County, public-health officials turned to Information Systems researchers from the W. P. Carey School of Business for a decision-making tool -- a software program -- that would help them distribute vaccines in an equitable and rational way. "I don't think it's an exaggeration to say that there were lives saved because of that software," recalls Bob England, M.D., director of the Maricopa County Public Health Department. "Because of it, vaccines got to the people who needed them most." Decision making in the white-board room The Maricopa County Public Health Department is the third largest local public health jurisdiction in North America. Only New York City and Los Angeles County are bigger. The county itself, which encompasses the Phoenix metropolitan area, is home to 61 percent of the state's population and more than 1 percent of the U.S. population as a whole. For this expected pandemic, the federal government's Centers for Disease Control allocated vaccines to states, then the states contacted the counties for lists of providers who should receive the vaccine doses, he continues. "While policy is made at the federal and state level, public health service delivery happens at a county level," explains Ajay Vinze, a W. P. Carey School professor of Information Systems. "When we found out we would be given the role of deciding where the H1N1 vaccine would go, we were glad, because it meant we could prioritize distribution," England says. "It sounded much simpler in theory than it turned out to be," he adds with a laugh. Many factors made the decision-making a challenge. The first problem was inadequate supply. According to England, healthcare agency staffs expected plenty of vaccine to be available within the first few weeks of the vaccination program. They also expected the doses to be shipped weekly, so officials would only need to sit down once a week to decide where to send the supplies. "Instead, the vaccine dribbled out at a rate of about 1 percent of the population each week, and it was allocated on a daily basis." he explains. "You'd be told you had 3,100 doses of vaccine to give out that day, and it was of various types." "The vaccine comes in many different dosages, and there are different ways of administering it," notes Vinze. Certain doses are appropriate for children, as are certain administration methods, such as nasal spray versus injection. It was up to England and his team to correlate types of vaccine with physician requests. Deciding which providers received the vaccine had to happen within a couple of hours. Plus, the notice from state officials telling England's team how much vaccine would be available that day came via email. "Someone had to be getting the email on time, or they had even less time to make a decision," notes W. P. Carey School professor Raghu Santanam, another IS expert called in on the job. This distribution process was designed for the majority of U.S. counties, notes Trent Spaulding, a doctoral candidate at the W. P. Carey School and one of two students who handled the code work. "For most counties, that's not a problem. They have maybe 50 healthcare providers who could give the vaccine, so they can compare them and decide how to divide the day's supply. But Maricopa County had more than 1,000 providers signed up to receive the vaccines." As with any flu, H1N1 was more deadly for certain populations. Small children and pregnant women topped the list of people who should receive vaccines first. School-age children were high on the list, too. England and his team tried to target vaccines to the healthcare providers most likely to give the doses to vulnerable populations. "We had a white board and list of people on a spreadsheet," England recalls. It was a very long white board, and the spreadsheet contained some 30 pages showing how many patients a provider served in each target category. As soon as the vaccine was available, England says he stayed up nights trying to review all the information, match it against vaccination requests and get the vaccine to the places with the largest numbers of needy patients. "That white board was obsolete within a week," he says. "That's when we went begging to ASU." Flu shot in the dark This wasn't the first time England had called on ASU. Recognizing the value that business school researchers can bring to decisional and workflow processes, the county public health has formalized the relationship with Vinze and his team via an inter-governmental agreement (IGA). Vinze notes that "this helps us to be in a position to provide quick response when called on to share our expertise -- we have had an IGA set up with the county since 2002." Santanam points out that this project drew on work he and Vinze had done in the past. The professors had previously created a decision-making system to respond to pandemic flu. "It was a multi-agent environment where you could input how fast the flu was spreading, populations of different regions, how often those populations interacted, transportation between regions," and other factors. He says that vaccination is one response to pandemics, but there also are non-pharmaceutical approaches, such as closing transportation between regions or closing schools and offices. He and Vinze used their decision-making tool to chart the best way to respond to flu epidemics in different circumstances. Likewise, the team had looked at what Vinze calls "syndromic surveillance," in which you examine data to "see if there is something quirky going on that you should look at more closely." According to him, such surveillance might potentially be used for issues related to public health preparedness and response and even for concerns related to bioterrorism. On the H1N1 project, England had several specific goals he hoped to achieve with his flu-shot distribution. One was timely use. "We wanted to move the vaccines every day so none sat unused in refrigerators," he says. Spaulding notes that the Maricopa team also wanted to spread vaccines around the city. "They didn't want everyone having to come to one location. They wanted to distribute the vaccine as equitably as possible." And, there were tracking problems to address. "The county didn't always know when vaccines were shipped to providers," explains Aaron Baird, another W. P. Carey School IS Department doctoral candidate who did code work on the software. Plus, England had a distribution strategy he wanted to pursue involving what he calls "the herd-immunity effect." It holds that children are the primary transmitters of the flu, so if you can vaccinate 80 percent of school-age children, you can eliminate 90 percent of flu cases. "What has made the many childhood diseases that everyone used to get so rare isn't perfect vaccines," England explains. It's the herd effect that happens when enough people get vaccinated. At that point, if one person catches a germ, the germ will have trouble finding a second person to infect. "Take pertussis -- whopping cough," England continues. "That vaccine is only 80 percent effective. So, if the vaccine doesn't work in one out of five people, why doesn't one out of five people get sick? It's because of that herd immunity. When enough people are immune, you don't get outbreaks." Considering their goals and the limited supply of vaccine, the health agency needed a decision-making tool to tackle "constraint-based resource allocation," notes Vinze. The tool had to accommodate both the strategy behind decisions -- how to prioritize allocations -- as well as tactics, or how to deliver vaccines day by day. Vinze's colleague, Santanam, adds that "computers can't make such decisions. But, they can help." He explains that the software factored in orders and priorities for the week, then recommended how the available vaccine should be distributed. Under normal circumstances, developing such a decision support environment can take from three to six months when approached using a more traditional software development approach. The W. P. Carey team accomplished and set in place a working system in one week! Days after starting the code work, England and his team were using the tool's output as a starting point for their deliberations. Planning a day's distribution still took a couple of hours, but England says speed wasn't the point of the software. Smart decision-making was. "We couldn't look through that 30-page spreadsheet and evaluate how many patients each doctor had in each target category," England says. Without the software, he thinks his team might have resorted to sending vaccine out alphabetically rather than on a prioritized basis. He suspects some healthcare jurisdictions had to resort to that random approach. "With the software, we were able to ship vaccine in a rational way and keep track of what we'd shipped," England adds. "We were able to make decisions based on where the vaccine was needed, rather than send it someplace without having a clue if that was the right step." The software developed is highly customized, a one-of-a-kind solution, not intended for any other county or even any other flu. Still, England hopes to get it customized to help him deal with vaccine distribution next year. "Having this system and pulling this off convinces me that it's possible to get a herd-immunity effect in the future," he says. "If we get enough vaccine, that's what we'll do next year. We may have to rework the software, but I'm not ready to throw it away." Bottom Line:

• Shortages of vaccine to combat the deadly H1N1 flu led to cumbersome distribution methods. Counties were sent daily allocations of vaccine, and they had to decide where to send the doses.
• In Maricopa County, Ariz., more than 1,000 healthcare providers wanted the vaccines. County health officials had to prioritize distribution to reach the most vulnerable patients. Decisions had to be made quickly every day.
• Maricopa County Public Health Department turned to the W. P. Carey School's Department of Information Services for a decision-making tool to help with distribution of vaccine.
• In one week, the W. P. Carey School's team had workable software that has been credited by county officials as helping save lives.

KnowIT