Census data more than just numbers
April 12, 2010
Filling out census data can now accomplish more than informing the government of how many people live in a residence and total household income. Since the 2000 census, scientists have been able to use the data to create models of how diseases can spread.
Using 2000 information such as household and business locations, scientists from RTI International, a scientifc research development institute, participated in the National Institute of General Medical Sciences’ Models of Infectious Disease Agent Study, or MIDAS, and created a synthetic population of the United States.
The information about the approximately 300 million people and 180 million households in the United States was plugged into a program that created the model of the country’s population.
The model can mimic how a disease could move within a population, based on how populated an area is and the age of its inhabitants. For example, a highly-congested area offers more potential hosts to a disease and very young or old populations tend to be higher risk because of weakened immune systems.
As young adults stay at home longer, or the elderly move back in with their families, the proximity of the varying age groups can create a cross-exposure that would not usually take place.
With data from this year’s census, scientists can create an even more realistic view of how an outnreak might occur.
Irene Eckstrand, director of the MIDAS program, said scientists need to think of how infectious diseases spread.“[We want to know] where people tend to be and how they bump into one another,” Eckstrand said. The MIDAS program aims to do just that.
This kind of information is important to determine who might get sick during an outbreak, she explained.
Different areas have different age groups and ethnicity concentrations; such factors influence how one reacts to a disease. Household size also becomes important when considering how many people are in proximity to each other and can pass on disease.
“This enables us to do things in computers you’d never do in human populations,” Eckstrand said. “[We can] run projects that would be unethical … it’s like a test tube.”
No identifiable human information is used, Eckstrand said, so no indviduals would ever be recognized from the sample.
“It gives us the same data without invading people’s privacy,” she said. “We don’t get that data, we don’t want that data.”
Such models are valuable because they provide a baseline of how epidemics spread and promote study on the impact of prevention and intervention, she explained.
“We know diseases don’t spread randomly,” Eckstrand said.
Eckstrand used the H1N1 epidemic as an example of how the synthetic population models are used. When the H1N1 vaccine was first created, there was a question of who to distribute it to first; the elderly, children, or those in a high-risk group. The answer was determined by who epidemiologists felt could be most affected by the flu as well as who was most likely to pass it on to others. Children were given the vaccine first because of their high exposure to others through school and home.
“Always in retrospect, [how to distribute] it seems obvious,” she said. “Now, we can know in advance. We’re gearing up to think about what might happen in the upcoming flu season.”
Bill Wheaton, a research geographer at RTI International, who is overseeing the study, said the project aims to advance the technology of computer simulations of infectious diseases. It’s important to simulate the right kind of interactions between individuals in a community, he explained.
“We want it to be realistic to provide information that’s helpful,” Wheaton said.
Different policies can be simulated as well, he added. Using H1N1 as an example, he said researchers were able to test if it would be more effective to close down a school or vaccinate the schoolchildren. If closing the school was more effective, the amount of days for an advisable closure could be experimented with.
In the event of a large-scale outbreak, a community may need data immediately to figure out how the epidemic may progress. In this case, a model can provide data right away, Wheaton said. The simulation can also be quickly altered to accommodate changing data.
Alfred Rademaker, professor in the department of Preventive Medicine at the Feinberg School of Medicine at Northwestern University, said disease models are very effective because they allow you to vary the parameters of a disease without affecting the population.
“Assumption is minimized because [you’re using] real data,” Rademaker said. “Other models assume a spectrum of scenarios.”
The wrong assumption is also less critical because it’s easier to predict what’s going to happen, he said.
Eckstrand said because of the economic downturn since the last census, bigger household sizes are expected when the new data are released. Young adults are moving out at later ages and the elderly move back in with their families as well. This can impact the disease models because larger households may lead to easier disease transmission, especially since two of the most vulnerable types of people— children and the elderly— are likely to be under the same roof again.
ern University, said disease models are
very effective because they allow you to vary the parameters of a disease without affecting the population.
“Assumption is minimized because [you’re using] real data,” Rademaker said. “Other models assume a spectrum
of scenarios.”
The wrong assumption is also less critical because it’s easier to predict
what’s going to happen, he said.
Eckstrand said because of the economic downturn since the last census, bigger
household sizes are expected when the new data are released. Young adults are moving out at later ages and the elderly move back in with their families as well.
This can impact the disease models because larger households
may lead to easier disease transmission, especially since two of the most vulnerable types of people— children and the elderly— are likely to be under the same roof again.