ORBiT: Oak Ridge biosurveillance toolkit for public health dynamics

Abstract

BackgroundThe digitization of health-related information through electronic health records (EHR) and electronic healthcare reimbursement claims and the continued growth of self-reported health information through social media provides both tremendous opportunities and challenges in developing effective biosurveillance tools. With novel emerging infectious diseases being reported across different parts of the world, there is a need to build systems that can track, monitor and report such events in a timely manner. Further, it is also important to identify susceptible geographic regions and populations where emerging diseases may have a significant impact.MethodsIn this paper, we present an overview of Oak Ridge Biosurveillance Toolkit (ORBiT), which we have developed specifically to address data analytic challenges in the realm of public health surveillance. In particular, ORBiT provides an extensible environment to pull together diverse, large-scale datasets and analyze them to identify spatial and temporal patterns for various biosurveillance-related tasks.ResultsWe demonstrate the utility of ORBiT in automatically extracting a small number of spatial and temporal patterns during the 2009-2010 pandemic H1N1 flu season using claims data. These patterns provide quantitative insights into the dynamics of how the pandemic flu spread across different parts of the country. We discovered that the claims data exhibits multi-scale patterns from which we could identify a small number of states in the United States (US) that act as "bridge regions" contributing to one or more specific influenza spread patterns. Similar to previous studies, the patterns show that the south-eastern regions of the US were widely affected by the H1N1 flu pandemic. Several of these south-eastern states act as bridge regions, which connect the north-east and central US in terms of flu occurrences.ConclusionsThese quantitative insights show how the claims data combined with novel analytical techniques can provide important information to decision makers when an epidemic spreads throughout the country. Taken together ORBiT provides a scalable and extensible platform for public health surveillance.

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

“…Tracking the residual errors using the Frobenius norm for both training and testing data, we performed a total of 250 iterations. In our analysis, as shown in previous work ( 17 , 22 ), we identified s to be 5. Once we chose s , the most stable version of the basis matrices ( W , H ) by computing the Kullback–Leibler (KL) divergence between every pair of the 250 instances of W (or H ) from the training set and picking W (or H ) with the lowest KL divergence value.…”

“…In our previous work, we showed how diagnostic eHRC transactions are comparable to standard public health surveillance data, such as the CDC ILINet ( 17 ). Further, we also showed that the consolidated eHRCs at local (zip code level information), regional (county, metropolitan, city, state, etc.…”

“…Prior to our study, we obtained internal Institutional Review Board approval for analyzing the IMS Health datasets. In this study, we analyzed the IMS Health ambulatory care reimbursement claims data from the 2009–2010 pandemic (H1N1) flu season; the details of the datasets are provided in our previous paper ( 17 ). Note that the data from IMS Health are already processed by a third party to remove any form of personally identifiable information before it receives the claims data from its suppliers.…”

“…To answer these questions, we used non-negative matrix factorization (NMF) to extract a small set of spatial and temporal patterns from the flu and asthma eHRCs. As we have shown in our previous paper ( 17 ), we chose NMF as an unsupervised machine learning technique to analyze the data primarily based on several empirical observations about the data. First, the data matrix A consists of only non-negative entries – because the total number of patients at any given zip code will be ≥0.…”

“…First, the data matrix A consists of only non-negative entries – because the total number of patients at any given zip code will be ≥0. Further, we observed that the individual zip codes exhibit a small number of distinct patterns in the occurrence of the flu ( 17 ), suggesting that in spite of a high dimensional setting of the A asthma and A flu matrices (with more than 14,000 individual zip codes and 365 days), there might be only a small number of geo-temporal patterns that could best capture the co-occurrence of these two conditions. Second, while several other types of analyses are possible to examine the data, our choice of analysis was motivated by the need to discover the underlying geo-temporal patterns in an unsupervised manner.…”