Data Analytics

The intrinsic variability that exists in the cases counting data for aggregated-area maps amounts to a corresponding uncertainty in the delineation of the most likely cluster found by methods based on the spatial scan statistics [3]. If this cluster turns out to be statistically significant it allows the characterization of a possible localized anomaly, dividing the areas in the map in two classes: those inside and outside the cluster. But, what about the areas that are outside the cluster but adjacent to it, sometimes sharing a physical border with an area inside the cluster? Should we simply discard them in a disease prevention program? Do all the areas inside the detected cluster have the same priority concerning public health actions? The intensity function [2], a recently introduced visualization method, answers those questions assigning a plausibility to each area of the study map to belong to the most likely cluster detected by the scan statistics. We use the intensity function to study cases of diabetes in Minas Gerais state, Brazil.

Objective

Cluster finder tools like SaTScan[1] usually do not assess the uncertainty in the location of spatial disease clusters. Using the nonparametric intensity function[2], a recently introduced visualization method of spatial clusters, we study the occurrence of several non-contageous diseases in Minas Gerais state, in Southeast Brazil.

The status of each Intensive Care Unit (ICU) patient is routinely monitored and a number of vital signs are recorded at sub-second frequencies which results in large amounts of data. We propose an approach to transform this stream of raw vital measurements into a sparse sequence of discrete events. Each such event represents significant departure of an observed vital sequence from the null distribution learned from reference data. Any substantial departure may be indicative of an upcoming adverse health episode. Our method searches the space of such events for correlations with near-future changes in health status. Automatically extracted events with significant correlations can be used to predict impending undesirable changes in the patient's health. The ultimate goal is to equip ICU physicians with a surveillance tool that will issue probabilistic alerts of upcoming patient status escalations in sufficient advance to take preventative actions before undesirable conditions actually set in.

Objective

To present a statistical data mining approach designed to: 1. Identify change points in vital signs which may be indicative of impending critical health events in ICU patients and 2. Identify utility of these change points in predicting the critical events.

Two significant barriers to greater use of syndromic surveillance techniques are computational time and software complexity. Computational time refers to the time for many methods (for example, scan statistics and AMOEBA statistics) to create reliable results. Software complexity refers to the difficulty of setting up and configuring suites of software to collect data, analyze it, and visualize the results. Both of these barriers can be partially surmounted by the use of cloud computing resources.

Objective

To describe how use of cloud computing resources can improve the timely provision of disease surveillance analyses.

Cardiovascular event prediction has long been of interest in the practice of intensive care. It has been approached using signal-processing of vital signs [1-4], including the use of graphical models [3,4]. Our approach is novel in making data segmentation as well as hidden state segmentation an unsupervised process, and in simultaneously tracking evolution of multiple vital signs. The proposed models are adaptable to the individual patient's vitals online and in real time, without requiring patient-specific training data if the patient-specific feedback signal is available. Additionally, they can incorporate expert interventions, produce explanations for alarm predictions, and consider effects of medication on state changes to reduce false alert probability.

Objective

To enable prediction of clinical alerts via joint monitoring of multiple vital signs, while enabling timely adaptation of the model to particulars of a given patient.

The Voronoi Based Scan (VBScan)[1] is a fast method for the detection and inference of point data set space-time disease clusters. A Voronoi diagram is built for points representing population individuals (cases and controls). The number of Voronoi cells boundaries intercepted by the line segment joining two cases points defines the Voronoi distance between those points. That distance is used to approximate the density of the heterogeneous population and build the Voronoi distance Minimum Spanning Tree (MST) linking the cases. The successive removal of its edges generates sub-trees which are the potential space-time clusters, which are evaluated through the scan statistic [2]. Monte Carlo replications of the original data are used to evaluate cluster significance. In the present work we modify VBScan to find the best partition dividing the map into multiple low and high risk regions.

Objective

We describe a method to determine the partition of a map consisting of point event data, identifying all the multiple significant anomalies, which may be of high or low risk, thus monitoring the existence of possible outbreaks.

Disease surveillance data often has an underlying network structure (e.g. for outbreaks which spread by person-to-person contact). If the underlying graph structure is known, detection methods such as GraphScan (1) can be used to identify an anomalous subgraph which might be indicative of an emerging event. Typically, however, the network structure is unknown, and must be learned from unlabeled data, given only the time series of observed counts (e.g. daily hospital visits for each zip code).

Objective

Our goal is to learn the underlying network structure along which a disease outbreak might spread, and use the learned network to improve the timeliness and accuracy of outbreak detection.

Detection of biological threat agents (BTAs) is critical to the rapid initiation of treatment, infection control measures, and public health emergency response plans. Due to the rarity of BTAs, standard methodology for developing syndrome definitions and measuring their validity is lacking.

Objective

The objective of this study is to outline and demonstrate the robust methodology used by Geographic Utilization of Artificial Intelligence in Real-Time for Disease Identification and Alert Notification surveillance system to generate and validate BTA profiles.

Optimal sequential management of disease outbreaks has been shown to dramatically improve the realized outbreak costs when the number of newly infected and recovered individuals is assumed to be known (1,2). This assumption has been relaxed so that infected and recovered individuals are sampled and therefore the rate of information gain about the infectiousness and morbidity of a particular outbreak is proportional to the sampling rate (3). We study the effect of no recovered sampling and signal delay, features common to surveillance systems, on the costs associated with an outbreak.

Objective

Development of general methodology for optimal decisions during disease outbreaks that incorporate uncertainty in both parameters governing the outbreak and the current outbreak state in terms of the number of current infected, immune, and susceptible individuals.

With the increase in GPS enabled devices, pin-point spatial data is an obvious future growth area for cluster detection research. The FBSSS handles binary labelled point data, but requires Monte Carlo testing to obtain inference [1]. In the Bayesian Poisson SSS [2], Monte Carlo is replaced by use of historic data, manifoldly speeding up processing. Following [2], [3] derived the BBSSS, replacing historic data with expert knowledge on cluster relative risk. This paper compares the spatial accuracy of BBSSS and FBSSS using new measure [4] which, being independent of inference level, permits direct comparison between Bayesian and frequentist methods. To compare the spatial accuracy of a Bayesian Bernoulli spatial scan statistic (BBSSS) and the frequentist Bernoulli spatial scan statistic (FBSSS), using benchmark trials.

Disease outbreak detection based on traditional surveillance datasets, such as disease cases reported from hospitals, is potentially limited in that the collection of clinic information is costly and time consuming. However, social media provides the vast amount of data available in real time on the internet at almost no cost. Our solution, NPHGS, provides a nonparametric statistical approach for outbreak detection that well addresses the key technical challenges in social media streams.

Objective

We present a new method for disease outbreak detection, the 'Non-Parametric Heterogeneous Graph Scan (NPHGS)'. NPHGS enables fast and accurate detection of emerging space-time clusters using Twitter and other social media streams where standard parametric model assumptions are incorrect.