Elbert Yevgeniy

Recent years' informatics advances have increased availability of various sources of health-monitoring information to agencies responsible for disease surveillance. These sources differ in clinical relevance and reliability, and range from streaming statistical indicator evidence to outbreak reports. Information-gathering advances have outpaced the capability to combine the disparate evidence for routine decision support. In view of the need for analytical tools to manage an increasingly complex data environment, a fusion module based on Bayesian networks (BN) was developed in 2011 for the Dept. of Defense (DoD) Electronic Surveillance System for the Early Notification of Community-Based Epidemics (ESSENCE). In 2012 this module was expanded with syndromic queries, data-sensitive algorithm selection, and hierarchical fusion network training [1]. Subsequent efforts have produced a full fusion-enabled version of ESSENCE for beta testing, further upgrades, and a software specification for live DoD integration. Beta test reviewers cited the reduced alert burden and the detailed evidence underlying each alert. However, only 39 reported historical events were available for training and calibration of 3 networks designed for fusion of influenza-like-illness, gastrointestinal, and fever syndrome categories. The current presentation describes advances to formalize the network training, calibrate the component alerting algorithms and decision nodes together for each BN, and implement a validation strategy aimed at both the ESSENCE public health user and machine learning communities.

Objective

This presentation aims to reduce the gap between multivariate analytic surveillance tools and public health acceptance and utility. We developed procedures to verify, calibrate, and validate an evidence fusion capability based on a combination of clinical and syndromic indicators and limited knowledge of historical outbreak events.

The Armed Forces Health Surveillance Center (AFHSC) supports the development of new analytical tools to improve alerting in the Electronic Surveillance System for the Early Notification of Community-based Epidemics (ESSENCE) disease-monitoring application used by the Department of Defense (DoD). Developers at the Johns Hopkins University Applied Physics Laboratory (JHU/APL) have added an analytic capability to alert the user when corroborating evidence exists across syndromic and clinical data streams including laboratory tests and filled prescription records. In addition, AFHSC epidemiologists have guided the addition of data streams related to case severity for monitoring of events expected to require expanded medical resources. Evaluation of the multi-level fusion capability for both accuracy and utility is a challenge that requires feedback from the user community before implementation and deployment so that changes to the design can be made to save both time and money. The current effort describes the design and results of a large evaluation exercise.

Objective

To evaluate, prior to launch, a surveillance system upgrade allowing analytical combination of weighted clinical and syndromic evidence with multiple severity indicators.

Data streams related to case severity have been added to the Electronic Surveillance System for the Early Notification of Community-based Epidemics (ESSENCE), a disease-monitoring application used by the Department of Defense (DoD), as an additional analytic capability to alert the user when indications for events requiring expanded medical resources exist in clinical data streams. Commonly used indicators are admission and death, but fatalities are rare and many DoD clinics lack admitting capability, so we sought to derive additional severity indicators from outpatient records. This abstract describes the technical details and the thought process behind two novel derived indicators: Sick-in-Quarters (SIQ) and Escalating Care.

Objective

To evaluate new severity indicators that mimic a public health professional or clinician's judgment in determining the severity of a public health event when detected by a surveillance system.

Influenza epidemics occur seasonally but with spatiotemporal variations in peak incidence. Many modeling studies examine transmission dynamics [1], but relatively few have examined spatiotemporal prediction of future outbreaks [2]. Bootsma et al [3] examined past influenza epidemics and found that the timing of public health interventions strongly affected the morbidity and mortality. Being able to predict when and where high influenza incidence levels will occur before they happen would provide additional lead time for public health professionals to plan mitigation strategies. These predictions are especially valuable to them when the positive predictive value is high and subsequently false positives are infrequent.

Objective

Advanced techniques in data mining and integrating evidence from multiple sources are used to predict levels of influenza incidence several weeks in advance and display results on a map in order to help public health professionals prepare mitigation measures.

An expanded ambulatory health record, the Comprehensive Ambulatory Patient Encounter Record (CAPER) will provide multiple types of data for use in DoD ESSENCE. A new type of data not previously available is the Reason for Visit (ROV), a free-text field analogous to the Chief Complaint (CC). Intake personnel ask patients why they have come to the clinic and record their responses. Traditionally, the text should reflect the patient's actual statement. In reality the staff often "translates" the statement and adds jargon. Text parsing maps key words or phrases to specific syndromes. Challenges exist given the vagaries of the English language and local idiomatic usage. Still, CC analysis by text parsing has been successful in civilian settings [1]. However, it was necessary to modify the parsing to reflect the characteristics of CAPER data and of the covered population. For example, consider the Shock/Coma syndrome. Loss of consciousness is relatively common in military settings due to prolonged standing, exertion in hot weather with dehydration, etc., whereas the main concern is shock/coma due to infectious causes. To reduce false positive mappings the parser now excludes terms such as syncope, fainting, electric shock, road march, parade formation, immunization, blood draw, diabetes, hypoglycemic, etc.

Objective

Rather than rely on diagnostic codes as the core data source for alert detection, this project sought to develop a Chief Complaint (CC) text parser to use in the U.S. Department of Defense (DoD) version of the Electronic Surveillance System for Early Notification of Community-Based Epidemics (ESSENCE), thereby providing an alternate evidence source. A secondary objective was to compare the diagnostic and CC data sources for complementarity.

Block 3 of the US Military Electronic Surveillance System for Early Notification of Community-Based Epidemics (ESSENCE) system affords routine access to multiple sources of data. These include administrative clinical encounter records in the Comprehensive Ambulatory Patient Encounter Record (CAPER), records of filled prescription orders in the Pharmacy Data Transaction Service, developed at the Department of Defense (DoD) Pharmacoeconomic Center, Laboratory test orders and results in HL7 format, and others. CAPER records include a free-text Reason for Visit field, analogous to chief complaint text in civilian records, and entered by screening personnel rather than the treating healthcare provider. Other CAPER data fields are related to case severity. DoD ESSENCE treats the multiple, recently available data sources separately, requiring users to integrate algorithm results from the various evidence types themselves. This project used a Bayes Network approach to create an ESSENCE module for analytic integration, combining medical expertise with analysis of 4 years of data using documented outbreaks.

Objective

The project objective was to develop and test a decision support module using the multiple data sources available in the U.S. DoD version of ESSENCE.

The VA has employed ESSENCE for health monitoring since 2006 [1]. Epidemiologists at the Office of Public Health (OPH) monitor the VA population at the national level. The system is also intended for facility-level monitoring to cover 152 medical centers, nearly 800 community-based outpatient clinics (CBOC), and other facilities serving all fifty states, the District of Columbia, and U.S. territories. For the entire set of facilities and current syndrome groupings, investigation of the full set of algorithmic alerts is impractical for the group of monitors using ESSENCE. Signals of interest may be masked by the nationwide alert burden. Customized querying features have been added to ESSENCE, but standardization and IP training are required to assure appropriate use.

Objective

The objective was to adapt and tailor the alerting methodology employed in the Electronic Surveillance System for Early Notification of Community-Based Epidemics (ESSENCE) used by Veterans Affairs (VA) for routine, efficient health surveillance by a small, VA headquarter medical epidemiology staff in addition to a nationwide group of infection preventionists (IPs) monitoring single facilities or facility groups.

The Johns Hopkins Applied Physics Laboratory and the Armed Forces Health Surveillance Center have developed a hybrid processing engine that alerts monitors when a severe health condition exists based on corroboration among several sources of data. The system was designed to ingest a day's worth of recent data and provide results to monitors daily. In some theaters, the health of the US Forces must be determined at near-real time rates requiring a reassessment of current surveillance practices. Challenges exist in both acquiring data in real-time and in modifying automated alerting processes to re-evaluate as a new piece of evidence is received.

Objective

To develop a real-time surveillance capability that processes, fuses and assesses when data is received using a new fusion processing methodology and multiple sources health indicator data.

Biosurveillance in resource-limited settings is essential because of both enhanced risk of diseases rarely seen elsewhere (e.g. cholera) and pandemic threats (e.g. avian influenza). However, access to care and laboratory test capability are typically inadequate in such areas, amplifying the importance of syndromic surveillance. Such surveillance in turn may be a challenge because of insufficient data history and systematic or seasonal behavior. The Suite for Automated Global Electronic bioSurveillance (SAGES) is a collection of modular, freely-available software tools to enable electronic surveillance in these settings. These tools require statistical alerting methods appropriate for SAGES data, and development of such methods is the subject of this effort. We evaluated alerting methods for two main uses: weekly surveillance for seasonal outbreaks such as dengue fever and influenza, and daily syndromic data for settings where monitoring and response on a daily basis are practical. The latter situation has the added complication that day-of-week clinical visit patterns differ widely, (e.g. clinic closure on Sundays and Thursdays) and may evolve over time.

Objective

The authors develop open-source temporal alerting algorithms for data environments characteristic of resource-limited geographic settings and recommend appropriate usage of each.

On 12/14/06, a windstorm in western Washington caused 4 million residents to lose power; within 24 hours, a surge in patients presented to emergency departments (EDs) with carbon monoxide (CO) poisoning. As previously described, records of all patients presenting to King County EDs with CO poisoning between 12/15/06 to 12/24/06 (n=279) were abstracted, of which 249 met the case definition and eligibility requirements. We attempted to identify each of the 249 confirmed cases of CO poisoning in our syndromic ED data set by comparing the hospital name, date, time, age, sex, zip code, chief complaint, and diagnoses across the two data sets. We designated each record as an exact match, likely match, possible match, or unmatched on the basis of the available fields.

Objective

We evaluated ED and emergency medical services data for describing an outbreak of CO poisoning following a windstorm, and determined whether loss of power was followed by an increase in other health conditions.