Modeling U.S. Infant Bronchiolitis Rates in the Presence of Spatial Uncertainty

• Activity Number: 538
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Candace Berrett* and Matthew Heaton and Chantel Sloan
• Companies: Brigham Young University and Brigham Young University and Brigham Young University
• Keywords: point process models ; intensity function ; bronchiolitis ; spatio-temporal

Abstract

Bronchiolitis is the single most common cause of infant hospitalization in the US, with approximately 1,915 per 100,000 infants hospitalized per year during winter virus season (Leader and Kohlhase 2003). To improve prevention strategies, a better understanding of the spatio-temporal spread of bronchiolitis across the US is required. The US Military Health System (MHS) has compiled data on about 140,000 cases of infant bronchiolitis between the years 2003-2013 and nearly 900,000 controls occurring across the US. However, for privacy reasons, patient home addresses have been randomized to within 400 meters of the original location creating spatial uncertainty in the geographic location of bronchiolitis incidences. Not accounting for this uncertainty may lead to incorrect inferences on bronchiolitis abundances. We develop a method to account for such spatial uncertainty in point process models by creating an intensity function that integrates over the possible space of true locations. Using this model and the MHS data, we provide inferences on the spatio-temporal pattern of national bronchiolitis rates.