Fighting the flu
From game theory to flu chips, researchers experiment with new technologies to help thwart upcoming health catastrophes
- By John Moore
- Feb 13, 2006
The westward movement of the H5N1 strain of avian flu has put the United States on alert. The virus first surfaced in East Asia, where it has killed millions of birds and more than 70 people since 2003. The disease, however, expanded its global reach late last year, with the discovery of infected birds in Europe. The World Health Organization (WHO) reports that the virus “is not yet spreading efficiently…among humans” but contends the “world is now closer to another influenza pandemic than at any time since 1968.”
The 1968-69 influenza pandemic, dubbed the Hong Kong flu, killed about 1 million people worldwide. WHO, citing a conservative estimate, believes that between 2 million and 7.4 million people would die if avian flu becomes a pandemic.
To blunt the effects of such a scenario, researchers from government, industry and academia are creating information technology-based systems. The projects described here fall into five categories: planning and modeling, prediction, surveillance, diagnosis, and supply chain management.
Planning and modeling systems let health care providers and public health officials run different outbreak scenarios to help anticipate hospital capacity and staffing needs. Prediction initiatives use commodity-like trading networks as a forecasting method. Flu surveillance systems aim to provide early detection and ongoing monitoring. Emerging influenza microarrays incorporate semiconductor technology to speed the identification of virus strains. Supply chain systems, meanwhile, have begun to automate the process of vaccine administration.
The projects catching on now aren’t necessarily new but have recently gained attention because of the H5N1 strain.
Martin Meltzer, senior health economist at the Centers for Disease Control and Prevention’s National Center for Infectious Diseases, said a flu modeling tool CDC launched in 2000 initially generated little interest. “There wasn’t the urgency,” he said. Several years later, the pandemic threat has changed that situation, and people worldwide are downloading the modeling software.
Avian flu has also raised the profile of surveillance systems, some of which were originally conceived to address bioterrorism concerns. In this area, considerable development has already occurred. At least four systems offer early event detection.
“We believe there is a lot of technology there,” said Susan Penfield, a vice president at Booz Allen Hamilton. “The big question is not necessarily the technology but…the sharing of the data and the integration of those systems.”
Preliminary integration work is under way for CDC’s BioSense program and the Defense Department’s Electronic Surveillance System for the Early Notification of Community-Based Epidemics (ESSENCE). Barry Rhodes, co-leader on BioSense at CDC’s National Center for Public Health Informatics, said the program is discussing harmonization with ESSENCE representatives. The objective is to cultivate a common infrastructure for early event detection.
From availability to acceptance
While the refinements continue, health officials can access a number of technology tools today. Delivery methods range from fee-based subscription services to free downloads. But availability doesn’t always translate into adoption. Some observers say they believe flu-fighting tools must blend into a health care organization’s daily routine to find acceptance.
Robert Egge, a project director at the Center for Health Transformation, a group based in Washington, D.C., that seeks to improve the U.S. health care system, questioned the practicality of fielding systems built to deal specifically with a flu pandemic. A complex initiative that medical professionals can only use once every 20 years or so won’t gain many followers, Egge said. “Realistically, you have to build your approach based on systems that will just become part of normal operations,” he said.
That’s the challenge facing public health officials and the researchers working on technology solutions.
Modeling and planning: Exploring flu scenarios
Health officials preparing for the next flu pandemic don’t have a lot of firsthand experience to go on because the most recent outbreak occurred in 1968.
In addition, familiarity with the annual flu cycle might not be much help. Pandemic flu follows a different path than the more commonplace varieties, experts say. For example, the typical flu strain hits the segment of the population aged 65 and older. The three pandemics in the 20th century, however, affected large numbers of people in younger age brackets.
But public health workers and hospital administrators can simulate what they have not experienced. Flu modeling tools aim to offer insight into an infrequently occurring event, and most are free to download.
CDC has built two such software tools, FluAid and FluSurge. FluAid lets users model the impact of a flu pandemic in their area. The software provides estimates of an outbreak’s effect in terms of deaths, hospitalizations and outpatient visits.
The 1968 pandemic is the basis for the modeling tool’s default values, Meltzer said. Because that pandemic was less lethal than the 1918 flu outbreak, estimates based on default assumptions will be conservatively low, he said. But users can change the default numbers to reflect other scenarios.
Planning for a surge in patients
While FluAid paints a pandemic in broad strokes, CDC’s FluSurge provides more details about an outbreak’s effect on hospitals. The tool compares existing hospital capacity with an estimated number of people who would be hospitalized, require intensive care and need ventilator support in the event of a pandemic.
Users can change the model’s underlying assumptions. “If you want to be drastic, you can put in a 1918 scenario,” Meltzer said. “By keeping the model simple, you can go back and forth and change whatever aspect you think is important.”
Hospital administrators can explore different scenarios to determine the number of beds they would need to handle a flu outbreak. The scenarios can also provide insight into staffing. Meltzer said many hospitals already experience nursing shortages, and the modeling tool might indicate the need to consider other staffing sources, such as retired nurses. Such information could prompt health officials to discuss how to create and maintain a list of supplemental staff.
“What the software has done is prodded them to think about things,” Meltzer said. “The software is really meant to generate conversation and not generate an absolute answer.”
Factory floor approach
Flu modeling tools aren’t the only planning resources available, however. Other software efforts help public health officials design vaccination clinics. Researchers at the University of Maryland’s Clark School of Engineering tapped principles of computer integrated manufacturing to create the Clinic Planning Model Generator. The software, designed to help planners ease congestion in clinics, has been available for download since September 2005.
In manufacturing, queuing models streamline the flow of parts in a production system. General models can also apply to customer service situations to help minimize the length of time people wait in lines. Vaccination clinics, however, face issues that general-purpose queuing models don’t cover because clinics are likely to deal with family groups, the elderly and people with disabilities.
“Those types of details make the problem more difficult,” said Jeffrey Herrmann, an associate professor at the Clark School’s Department of Mechanical Engineering and Institute for Systems Research. “We had to adapt the model.”
The clinic modeling project originally focused on smallpox. In 2004 researchers conducted a time study of a mass vaccination clinic exercise in Montgomery County, Md. Based on that experience, “we realized we could create models that could be useful for a wide variety of similar events,” Herrmann said.
Prediction: Borrowing from financial markets
Health professionals fight against time more than anything when it comes to containing an outbreak. The need for an early warning system is obvious. The source of inspiration for a couple of flu prediction efforts is less obvious — financial markets. Prediction markets have emerged in recent years to forecast election results. Those markets have now been tapped to predict the arrival of flu.
In a prediction market, people buy and sell futures contracts. But instead of investing in pork bellies or orange juice, they invest in the outcome of events. Iowa Electronic Markets, run by the University of Iowa’s Tippie College of Business, have focused primarily on political events. But in 2004 the business school set up its first flu market. That pilot project set out to predict flu activity in Iowa. Last fall the business school and Iowa’s Carver College of Medicine received a $1.1 million grant to expand the prediction market nationally.
Forrest Nelson, professor of economics at the University of Iowa, said he hopes public health officials will use the market forecast as a warning signal. Hospital administrators who anticipate an outbreak, as opposed to learning about one after the fact, have more time to prepare for an increasing number of flu cases, he said.
“People on the front lines seeing patients, administrators and managers — those are the kinds of people that this kind of market should benefit,” Nelson said.
For now, only Iowa health care workers will be able to get a heads-up from the prediction market. Participation in the 2005-2006 influenza market is limited to invited members of the Iowa health care community. The selected traders can purchase contracts corresponding to the level of flu activity in the state during a given week. The contracts are color-coded. A trader buying a white and yellow contract believes there will be no flu reports or activity. A red contract, in contrast, represents a widespread outbreak. Traders can buy contracts on flu activity as far as six weeks in advance. The higher the price, the stronger the feeling that an outbreak is imminent.
Traders ‘in the know’
The Iowa flu prediction market seeks to forecast outbreaks of the more typical strains of flu, which run their annual courses each winter. But Nelson said Iowa is considering an avian flu market. The key hurdle: finding traders in the know. “We need traders with access to the right kinds of information,” he said.
The Iowa flu market cultivates contacts in emergency rooms and pharmacies within the state. But an avian flu market, Nelson said, would require participants in Asia and traders with specialized profiles, such as microbiologists who examine data worldwide.
“We’ve not yet tapped into the right circles to get access to those traders,” he said. “That has kept us from opening an avian flu market.”
Meanwhile, Trade Exchange Network recently closed one avian flu prediction market and is winding down another. In the first market, traders bought and sold futures contracts on whether avian flu would be confirmed in the United States by the end of 2005. The other market has traders testing hunches regarding a confirmed U.S. outbreak by March 31. The flu market operates on Trade Exchange Network’s Tradesports Web site.
Mike Knesevitch, communications director at Trade Exchange Network, said the flu market is the first health-related trading arena the company has posted. The company now lists more than 2,000 markets, spanning basketball games to New York City snowfall amounts.
The Tradesports bird flu market differs from Iowa’s in that the former is open to anyone, not just people in the health field.
“Their exchange is limited to a fixed number of participants with a limited amount of capital,” Knesevitch said of the Iowa market. “Our markets have no limits, similar [to] the world financial markets, so we believe our data points are more representative of the true market.”
Knesevitch said the company’s markets possess a very high predictive value.
Surveillance: Developing tools for detection and situational awareness
To contain an infectious disease, public health officials need to be on top of an outbreak from its onset. An Imperial College of London study, based on a model of an avian flu outbreak in Thailand, found that the disease remained localized for the first 30 days and spread nationwide in 60 to 90 days, according to the journal Nature. Containment depends on rapid disease identification and response.
Against this backdrop, government and university researchers have assembled a number of early event detection systems, intended to keep public health officials apprised of disease outbreaks. The systems may provide a line of defense against a flu pandemic.
The University of Pittsburgh’s Real-time Outbreak and Disease Surveillance (RODS) system provides one example. The system analyzes data from hospitals and public health departments and searches for patterns that could signify an outbreak.
“The algorithms we created will…see if there’s an anomaly in the trending of the data,” said Steve DeFrancesco, director of IT at the university’s RODS Laboratory. If the software detects a spike in a certain syndrome, RODS dispatches an alert to public health officials.
The lab’s Public Health Data Center hosts the system’s software. Health care facilities link to the center through an interface that automatically transmits data such as emergency room registrations. The software processes the data and generates alerts for distribution via e-mail or pager. Health officials can view data as a map or graph via a secure Web interface.
Customers subscribe to the RODS service for a minimal fee, which is based on population and the geographic area covered, DeFrancesco said. He added that a little more than 250 hospitals connect to the system and actively send data to RODS. Health departments in 11 states also use RODS’ early detection feature.
CDC’s BioSense program is another system with an early event detection component. BioSense taps sources such as Department of Veterans Affairs treatment facilities, clinical laboratories and hospitals to keep tabs on health events as they unfold. The system is part of CDC’s Public Health Information Network (PHIN).
Other systems with an early detection thrust include Sandia National Laboratories’ Rapid Syndrome Validation Project and DOD’s ESSENCE. In addition, CDC’s National Electronic Disease Surveillance System (NEDSS), another PHIN element, provides a standard reporting format for disease data. Several states use NEDSS-compliant systems for reporting data to CDC.
But human observers — not automated systems — may play the greatest role in identifying an avian flu outbreak in the United States.
“Most likely, these surveillance systems are not going to be the predictors” of avian flu, said Dan Desmond, president of SIMI Group, a systems integrator and Microsoft partner that specializes in health care solutions. “At this point, people are going to be doing the surveillance.”
But surveillance systems would play a role in an outbreak, even if they aren’t first to identify one. “Monitoring what happens after [detection] is the real issue,” Rhodes said.
To that end, systems such as BioSense can provide situational awareness, he added. Once medical professionals confirm an outbreak, BioSenseRT, the system’s real-time component, could monitor the event as it unfolds. The idea is to use BioSense to shed light on the magnitude of an event, its duration and issues such as efficacy of treatment.
Rhodes said other systems such as ESSENCE and RODS could also address the situational awareness questions, assuming close-to-real-time data collection.
Diagnosis: Speeding virus Ids
To identify specific influenza strains and subtypes, medical examiners need several days to complete their tests.
Semiconductor-based microarrays, or biochips as they are sometimes called, promise to shrink detection time to a matter of hours. CombiMatrix, a biotechnology company based near Seattle, launched a DNA microarray last year for typing flu strains. In addition, the University of Colorado at Boulder has developed a Flu Chip that can determine flu strains from patient samples.
“Traditional testing takes days, because you have to grow a virus to get enough sample,” said Bret Undem, vice president of CombiMatrix. In October 2005, CombiMatrix debuted its Influenza Microarray, which the company said can detect and accurately type flu strains in less than four hours. A test last year at CDC found that the University of Colorado’s Flu Chip can uncover the genetic signatures of flu types and subtypes in about 11 hours.
A microarray consists of a chip that fits on a microscope slide. In the case of Colorado’s Flu Chip, an array robotically spots genetic bits of information that correspond to known flu strains, according to a university project summary.
Lab workers then place the microarray in a wash of flu gene fragments obtained from an infected individual. RNA fragments from the sample bind to the specific DNA segments on the microarray. A fluorescent label is applied to the captured RNA, and the array is then placed into a laser scanner.
The glowing portions of the array create a pattern indicative of a particular virus.
Chips gain acceptance
University of Colorado officials expect laboratories to adopt the Flu Chip within a year. The technology has been under development for two years, with funding from the National Institute of Infectious Diseases.
In CDC’s October 2005 review, the chip was tested for avian flu strain H5NI and two common human flu types, H1N1 and H3N2. The chip was more than 90 percent accurate, the university reported. At press time, CDC was expected to conduct an additional test of the Flu Chip.
Kathy Rowlen, a professor in the university’s chemistry and biochemistry department, said a commercial vendor will eventually manufacture the Flu Chip. “Several companies have expressed interest,” she added.
The CDC and WHO, meanwhile, have samples of CombiMatrix’s Influenza Microarray. “We are talking to other government agencies,” Undem said.
The Influenza Microarray can identify H5N1 avian flu and other strains of Influenza A, the company said. It can work with samples from birds, pigs, horses, dogs and humans.
Undem said the retail price is $550 for one chip, adding that the company is working on ways to make the chip less expensive.
Remote diagnoses ahead
CombiMatrix and the University of Colorado are seeking to refine their flu chip technology. The portability of flu testing gear is one limitation. The current crop of scanners used to read microarrays is too cumbersome to deploy outside a lab environment.
The ability to move diagnoses to remote areas, rather than confining the job to a centralized lab, would further speed the identification process. Undem said DOD has funded the company’s beta test of a portable field unit. The device will read microarrays electrochemically rather than optically. He said the unit is toaster-sized.
Plans to reduce the size of the Flu Chip are also in the works. Rowlen said InDevR, a company that makes biomedical instruments, is developing the “ancillary technology to enable field application of microarrays.” Rowlen, who is chief scientist at InDevR, added that it will take a couple of years to produce the field-portable version.
Supply chain management: Vaccine manufacturing process makes deliveries tough to predict
To combat an influenza pandemic, vaccines need to be delivered in the right quantities to the right place at the right time. Vaccine supply chain management, however, has not been a high-profile activity, industry observers say. Systems exist that automate elements of vaccine logistics, but no single system covers the supply chain from manufacturer to local health authorities.
“It’s a mistake there’s not more focus on that,” Penfield said.
Several issues complicate the timely delivery of flu vaccine. “On the supply side, manufacturers have opted out of the unpredictable vaccine market, causing a chronic shortage of production capacity,” according to a bulletin from the Massachusetts Institute of Technology’s Center for Transportation and Logistics. In addition, the vaccine manufacturing process makes yields hard to predict, according to MIT’s bulletin.
Another factor is the inflexibility of the supply chain. The retail industry routinely shifts supplies from areas in a surplus situation to areas experiencing shortages, but that approach is not common in health care, the bulletin states. The result is price gouging, a practice that is difficult to prevent because “no entity has an overall picture of product flows and where vaccine excesses and shortages are developing,” the bulletin states.
Vaccine finder system
CDC is putting systems in place to deal with the unpredictability of the flu vaccine market. In 2004 CDC launched its Flu Vaccine Finder System when the contamination of flu vaccine from a European manufacturer led to a shortfall in the United States. The system tracks flu vaccine supplies among manufacturers and distributors.
During the 2004 flu vaccine shortage, the system “matched available vaccine supply and production with population densities for priority groups to determine a monthly appropriation,” according to CDC’s Web site. The system also provides a Web-based interface that lets states view supply information, place orders and monitor order status.
The flu vaccine finder falls under PHIN’s Countermeasure and Response Administration. That system also provides a link to the Strategic National Stockpile, which houses medicine and supplies in secure warehouses nationwide. In the event of a bioterrorism attack or flu outbreak, CDC will deliver stockpiled supplies to affected states within 12 hours.
Federal and state roadblocks
Desmond said CDC has a Web-based application that manages the distribution of supplies to state drop-off centers. But each state must act independently when it comes to distributing vaccines. “There’s no overarching process,” he said.
A 2005 report published by the Trust for America’s Health says a gap in the supply chain exists between the federal distribution system and state and local governments. The report states that “only seven states and two cities have achieved ‘green’ status for the Strategic National Stockpile, which means being recognized by [CDC] as adequately prepared to administer and distribute vaccines and antidotes in the event of an emergency.”
Attempts to improve management of vaccine supplies in the United States could get a boost from international projects. In September 2005, the U.S. Agency for International Development awarded a $77 million contract to a consortium that will develop a supply chain management system to dispense HIV/AIDS-related drugs.
“Some of the lessons learned from this opportunity can be applied in developing a supply chain strategy for delivering drugs not only in the [United States], but globally,” said Penfield, whose company is a member of the USAID supply chain consortium.