Discovering Multi-Scale Co-Occurrence Patterns of Asthma and Influenza with Oak Ridge Bio-Surveillance Toolkit

Abstract

We describe a data-driven unsupervised machine learning approach to extract geo-temporal co-occurrence patterns of asthma and the flu from large-scale electronic healthcare reimbursement claims (eHRC) datasets. Specifically, we examine the eHRC data from 2009 to 2010 pandemic H1N1 influenza season and analyze whether different geographic regions within the United States (US) showed an increase in co-occurrence patterns of the flu and asthma. Our analyses reveal that the temporal patterns extracted from the eHRC data show a distinct lag time between the peak incidence of the asthma and the flu. While the increased occurrence of asthma contributed to increased flu incidence during the pandemic, this co-occurrence is predominant for female patients. The geo-temporal patterns reveal that the co-occurrence of the flu and asthma are typically concentrated within the south-east US. Further, in agreement with previous studies, large urban areas (such as New York, Miami, and Los Angeles) exhibit co-occurrence patterns that suggest a peak incidence of asthma and flu significantly early in the spring and winter seasons. Together, our data-analytic approach, integrated within the Oak Ridge Bio-surveillance Toolkit platform, demonstrates how eHRC data can provide novel insights into co-occurring disease patterns.

Assessing Methods for Evaluating the Number of Components in Non-Negative Matrix Factorization

“…The results indicate that although normalization may speed up processing [21], it has an unpredictable effect on the estimation of the number of components k. We therefore recommend that, at least for the purpose of estimating k, data not be normalized. Indeed, this is the approach that has been taken in much of the public health and environmental science literature using NMF [5,6,8,9].…”

“…We therefore recommend that, at least for the purpose of estimating k , data not be normalized. Indeed, this is the approach that has been taken in much of the public health and environmental science literature using NMF [ 5 ], [ 6 ], [ 8 ], [ 9 ].…”

“…Sutherland-Stacey and Dexter [ 5 ], used projected gradient descent to identify two chemical factors corresponding to pollutants in the spectra measured from dairy processing wastewater. Ramanathan et al [ 6 ] used the alternate least squares algorithm to identify and characterize five geo-temporal patterns explaining the co-occurrence of asthma and flu based on ZIP code. In the context of public health, Stein-O’Brien et al [ 7 ] demonstrates the application of NMF to gene-expression data and reviews its utility in addressing questions ranging from systems-level to cell-level analysis in genetics.…”

“…However, although there are instances of normalization by column [ 4 ], there is frequently no mention of normalization [ 5 ], [ 6 ], [ 8 ], [ 9 ]. An examination of the effect of normalization on the estimation of is warranted.…”

Computing disease incidence, prevalence and comorbidity from electronic medical records

“…For younger patients, their age-dependent risk could be modeled, but this requires accurate determination of age of onset, which is problematic, and the older patients may be older than the EMR, so their diagnoses of early-onset diseases will be recorded at misleadingly later ages, if recorded at all. Unfortunately, co-occurrence of disease codes, without correction for age (or, if appropriate, gender), is often used to establish disease comorbidity [45, 46, 47, 48, 49, 50]. Wong et al [51] did use age and sex stratification to compute observed/expected ratios to examine multimorbidity.…”