Researchers Develop New Methods for Imputing Data for Geographic Analysis

Mar 08, 2010

(PhysOrg.com) -- Researchers at the Johns Hopkins Bloomberg School of Public Health have developed new methods for analyzing health data geographically.

Typically, data are plotted spatially through a process known as geocoding in which mailing address information is translated into map coordinates. However, not all addresses can be converted successfully (nongeocodable). Rural postal routes, post office boxes, and addresses with errors or missing information cannot be mapped using geocoding. linked with these types of addresses have traditionally been discarded from analysis leading to concerns of bias and underreporting. In a study published February 10 in the journal , the researchers develop and evaluate strategies for including nongeocoded data in spatial analysis.

“Our methods allow for the analyses of complete data, which is a notable improvement over the traditional practice of discarding information from addresses that did not geocode,” said Frank C. Curriero, PhD, lead author of the study and associate professor in the Bloomberg School’s Department of Environmental Health Sciences and Department of Biostatistics. “Although we evaluated our methods using , these strategies could be applied in other analysis settings.”

For the study, the research team evaluated prostate cancer data from the Maryland Cancer Registry and considered total case counts at the Census county, tract, and block group level as the analysis objective. Three strategies were developed and evaluated. The third strategy tested used a weighted probability method to assign nongeocoded addresses to these Census units using available information on the age and race distribution within the assumed known zip code. According to the study, the weighted assignment strategy performed best overall.

Additional authors of “Using Imputation to Provide Location Information for Nongeocoded Addresses” are Martin Kulldorff, PhD, of Harvard University, Francis P. Boscoe, PhD, of the New York State Department of Health, and Ann C. Klassen, PhD, associate professor in the Bloomberg School’s Department of Health, Behavior and Society.

Explore further: UN: Goals helped lift one billion people from extreme poverty

More information: www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0008998

Related Stories

Cost-effective measures could stop child pneumonia deaths

Jun 01, 2009

Implementing measures to improve nutrition, indoor air pollution, immunization coverage and the management of pneumonia cases could be cost-effective and significantly reduce child mortality from pneumonia, according to a ...

Sleep duration may play important role in childhood obesity

Feb 07, 2008

Less sleep can increase a child’s risk of being overweight or obese, according to a study by researchers at the Johns Hopkins Bloomberg School of Public Health. Their analysis of epidemiological studies found that with ...

Consumer electronics can help improve patient health

Oct 27, 2009

Electronic tools and technology applications for consumers can help improve health care processes, such as adherence to medication and clinical outcomes like smoking cessation, according to a report by researchers at the ...

Recommended for you

WHO says too few countries taxing tobacco products enough

2 hours ago

Taxing cigarettes up to 75 percent of their retail price is among the most effective ways to reduce tobacco use, but too few governments levy high enough taxes, according to a World Health Organization global report released ...

Can four fish oil pills a day keep the doctor away?

3 hours ago

Fish oil is one of the most popular dietary supplements in the U.S. because of the perceived cardiovascular benefits of the omega-3 it contains. However, scientific findings on its effectiveness have been conflicting. New ...

User comments : 0

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.