Cardiology: Mutation that Disrupts Calcium Signaling May Be One Cause of Heart Failure Public Health: Software Rings Early Alarm on Bioterrorism Cell Biology: Molecular Movies Catch Mitochondria Dividing In Memoriam: Microbiology Department's Harold Amos Dies Second Year Show: Second-year's Put On Swell Show Letter to the Editor   Details Reported on Caspase-independent Cell Death Pathway Researchers Tie Leptin to Obesity Pathway Distinct from Hormone's Role in Reproduction Evidence of Safety and Efficacy Halt Trial of Low-dose Blood Thinner   Proceedings of the HMS Faculty Council FAQs on the HMS Faculty Survey Dean's Community Service Award Nominations Honors and Advances   Mentoring Program Pairs Students with Youths at Risk   Insuring Americans Both Efficiently and Fairly   Front Page

PUBLIC HEALTH Software Rings Early Alarm on Bioterrorism Surveillance Refined Through Modeling Outbreak Patterns, Using Wider Time Window When it comes to detecting bioterrorism, the basic premise has not changed much since John Graunt presented his Observations on the Bills of Mortality to the Royal Society in London in 1662. The biosurveillance manuscript was the first book on statistics, defining the normal fluctuations in deaths ranging from "affrighted" to "worms" and including the biggest threat, the plague. Graunt distilled patterns in the data from 60 years worth of "bills of mortality" posted every Thursday except for Christmas. Ever since, periodic retrospective reports have been a mainstay of public health monitoring. Digital detection. Hospital visit data normally fluctuate widely from day to day, creating a biosurveillance challenge to detect a signal amid the noise. Here, the closest row shows that the most popular one-day analysis method cannot detect a simulated attack resulting in 20 extra visits a day in two one-week periods. The attack shows up best in the back row, after the data has been "filtered" through an algorithm theoretically suited to detect exponential growth over seven days. Kenneth Mandl, Marcello Pagano, and Ben Reis (below, l to r) achieved earlier and better detection of simulated bioterrorism attacks in hospital emergency visit data when they expanded the analysis time window from one day to one week and looked for more specific patterns of different types of biological outbreaks. (Image adapted from original by Ben Reis. Photo by Steve Gilbert) With renewed concerns about bioterrorism in the last year, people are seeking earlier and more sensitive ways to spot emerging outbreaks. Syndromic surveillance systems are being set up at local, state, and federal levels to collect, analyze, and interpret health data in hopes of a quick, effective response to a biological threat. The first signs of a germ warfare attack, for example, might show up as clusters of people with flulike symptoms seeking care at emergency rooms. "In this paper, we showed you shouldn't just look at what happened today. You should put what happened today in the context of what happened the day before, and the day before that." --Marcello Pagano The problem is, influenza and other diseases also can show up in the data as clusters of people with flulike symptoms seeking health care at emergency rooms. "Efforts to detect covert bioterrorist attacks from increases in hospital visit rates are plagued by the unpredictable nature of these rates," said Ben Reis, first author of a new study that describes two ways to improve the emerging science of biosurveillance. "With only narrow time windows available for effective public health response," he said, "knowing early is often as important as knowing at all." The findings, published in the Feb. 18 (online Feb. 3) Proceedings of the National Academy of Sciences, have immediate practical application. Another researcher already has reported back by e-mail that the paper's fundamental recommendations appear to translate to other datasets in other settings, according to preliminary results. Discerning the Damage For the study, Reis and his colleagues at Children's Hospital and the Harvard School of Public Health started with 10 years of data adding up to more than 500,000 visits to the emergency department at Children's Hospital. They designed four different mathematical formulas, or "filters," loosely correlated to different biological paradigms--a one-day spike in visits, a fixed number of additional visits (moving average), a linearly increasing number of visits, and an exponentially increasing number of visits. More importantly, in each simulation, Reis looked at a week's worth of new data through a sliding seven-day window, instead of the standard one-day perspective, which had the effect of smoothing out random blips in the data. Then they simulated and tested three potential patterns of attack, lasting from three days to two weeks, and affecting different numbers of people. "This is the first published paper I'm aware of that uses simulation of different types of attacks to test the detection algorithm and, further, that looks at tuning detection algorithms to be best matched with detection of specific attacks," said senior author Kenneth Mandl, HMS assistant professor of pediatrics and research director at the Children's Hospital Emergency Department. Filter Performance Generally, their temporally enhanced filters detected more outbreaks within the first two days, the researchers found. Each of the filters was slightly better for different stages and sizes of outbreak. The standard one-day filter was best for the first day of an outbreak but most vulnerable to false alarms. The moving-average filter improved over time and was best on the seventh day. The linear filter was better in the earlier and middle stages of an outbreak. The exponential filter was strongest at detecting the earliest stages of an outbreak. The researchers calibrated all the filters to allow an average of one false alarm per month. "In this paper, we showed you shouldn't just look at what happened today," said co-author Marcello Pagano, HSPH professor of statistical computing. "You should put what happened today in the context of what happened the day before, and the day before that, etc. It's a simple idea that improved things tremendously." "People around the country are setting up these systems and building expensive digital pipes to bring the data together to analyze it," said Reis, who now works at the Markle Foundation in New York, which is trying to set up uniform standards for public health interactions so medical records and other data are compatible between hospitals. "The smaller part is writing the software code to analyze the data once you have it in one place. The most significant aspect of the paper is, with a small tweak to the analysis code, you can use the system you have in place, in some cases more than doubling its sensitivity." Mandl and Pagano are exploring detection algorithms based on where people may have been exposed. Using a small simulated attack and home addresses of patients, they found increased detection power, according to preliminary results presented in May at the Society for Pediatric Research annual meeting in Baltimore. They are also working on ways to analyze different data sources and applying the same algorithms to a real-time analysis of Beth Israel Deaconess Medical Center adult emergency visits. Meanwhile, Childrens' Emergency Department uses the surveillance software, known as EDScope, for daily operations. A computerized readout--updated approximately every 15 minutes--shows all the activity in the ED that day and for the previous week. It also compares the data from a given day to the more than 500,000 emergency visits in the last 11 years and raises a red flag if patterns develop. "Public health has never had real-time data to work with before," said Mandl, who is also on the faculty of the Children's Informatics Program and an early investigator in the new science. "The demands of bioterrorism are focusing attention and funding, but these systems will only be successful in the long run if they have dual uses and track many different kinds of outbreak. For example, we may be able to detect West Nile virus, an episode of food poisoning, or more asthma in certain parts of the city without doing lengthy investigations." --Carol Cruzan Morton