View list of archived news items.

InfoFusion: Utilization of Multi-Source Data
CDC Statistical Advisory Group Symposium
Atlanta, GA, April 7-8, 2009

InForMID members Julia Wenger and Dr. Steve Cohen presented their work at the 12th Biennial Symposium on Statistical Methods organized by CDC.

Wenger JB, Bhattacharyya S, Naumova EN . The Application of δ-Method for Seasonality Assessment of Influenza in Multi-Source Surveillance Data

Abstract:
Biosurveillance systems for infectious diseases typically deal with nonlinear time series arising as a result of non-Gaussian and non-stationary outcome processes. Characterization of nonlinear time series of abrupt outbreaks with short durations is challenging. However, if outbreaks appear with strong periodicity, a reliable description of time series can be provided. Well-defined seasonal patterns observed in host susceptibility, pathogen abundance and transmissibility are documented in infectious diseases. We illustrate the use of an analytical tool based on the δ-method for evaluation, examination, and comparison of seasonal behaviors for multi-source surveillance data. We consider three characteristics of seasonality: timing, duration, and magnitude of a seasonal peak. The measures related to timing are: position of the maximum point on the seasonal curve of disease incidence; position of the minimum point on the seasonal curve of disease incidence; and lag, the difference between the time of exposure and time of disease incidence maximum. Magnitude related measures are: maximum value on the seasonal curve of disease incidence; minimum value on the seasonal curve of disease incidence; amplitude, the difference between maximum and minimum values on the seasonal curve for disease incidence; and relative intensity, the ratio of maximum and minimum value on the seasonal curve. The estimation of duration is based on distribution thresholds. We demonstrate the proposed method using a comparative study of laboratory surveillance systems in Milwaukee, Wisconsin compared with reporting from hospital and mortality records. We detected differences in influenza seasonal patterns by various sources, suggesting differences in pathogen virulence or host susceptibility as measured by the impact of individual influenza seasons/strains on sensitive population. We also observed strong relations between peak timing, duration, and intensity of various disease-related outcomes.

Cohen SA and Naumova EN. Acceleration of influenza associated disease rates with age: The SIMPLE Method

Abstract:
Selecting appropriate summary measures that accurately and precisely capture disease burden and also changing population structure and dynamics that contribute to disease patterns is challenging. Although useful, some of the common measures of disease burden in the elderly, such as age-adjusted rates and age-categorization of rates can mask important population processes that contribute to such disease patterns, especially in the elderly. We used Medicare hospitalization data (1991-2004) and US Census Bureau counts to estimate age-specific, influenza-related hospitalization incidence in the elderly by influenza season on the state and regional levels. We observed that incidence increases exponentially with age, and therefore proposed a technique, the Slope Intercept Modeling for Population Linear Estimation (SIMPLE) method, that represented that trend by two parameters. There parameters consist of an intercept, which represents expected influenza incidence at age 65, and a slope, which represents the degree of exponential increase of influenza-related incidence by age for the elderly. R-squared values for these models ranged from 0.756 to 0.989 for states and between 0.935, 0.996 for regions, and 0.987-0.998 for the US as a whole. These results suggest spatiotemporal trends in influenza-related cases in the elderly can be estimated accurately and precisely using the SIMPLE method. In addition to utilizing the technique to assess single-season, period-based age acceleration in disease rates, the SIMPLE method can also be used to assess multiple-season age acceleration in rates by cohort. The two meaningful and relevant measures obtained from the SIMPLE technique can be used as summary measures of disease in a population and reflect the underlying demographics and epidemiology of disease, and can be applied to other diseases that exhibit similar age patterns in the elderly population on multiple geographic levels.