Online Program

Actionable public health information from a temporal milwaukee beach water quality database

Wednesday, November 6, 2013 : 1:10 p.m. - 1:30 p.m.

Aurash Mohaimani, BS, Doctoral Program in Biomedical and Health Informatics, University of Wisconsin-Milwaukee, Milwaukee, WI
John Hernandez, BS, Zilber School of Public Health, University of Wisconsin-Milwaukee, Milwaukee, WI
Chelsea Weirich, BS, Zilber School of Public Health, University of Wisconsin-Milwaukee, Milwaukee, WI
Todd Miller, Ph.D, Zilber School of Public Health, University of Wisconsin-Milwaukee, Milwaukee, WI
Peter Tonellato, PhD, Zilber School of Public Health, University of Wisconsin-Milwaukee, Milwaukee, WI
Public health authorities post advisories or close beaches based on the presence of indicator bacteria (e.g. E. coli) that may suggest the presence of additional pathogenic species. The evidence supporting issuance of a public notice may be confounded by sample quality, methodology and timeliness of analysis. Timely E. coli counts, coupled with environment-based predictive modeling of E. coli trends at beaches may mitigate these complications and thus improve public health decisions. With the University of Wisconsin-Milwaukee, Zilber School of Public Health joined the Milwaukee Health Department (MHD) to devise and build a data Extract, Integrate, Translate and Load (EITL) process and database to support E. coli predictive modeling at Milwaukee beaches. Lake Michigan water samples were collected from three popular beaches, June to August 2012. Levels of E. coli were assessed by Colilert-18 assays and qPCR. EPA guided survey data were recorded for each beach. Our EITL quality assurance steps produced a robust database supporting predictive modeling. Retrospective analysis was conducted by preparing collected data in conjunction with automated satellite data obtained from the Environmental Data Discovery and Transformation service offered by the United States Geological Survey. Transformed data was input into Virtual Beach (VB), a software package for developing multiple linear regression models to predict pathogen indicator levels at recreational beaches. Combinations of four-variable models were investigated, and overlapping informative variables were selected for generation of a five-variable model demonstrating 83% predictive sensitivity to incidences of E. coli threshold (EPA regulatory standard: 235 most probable number per 100 mL freshwater). Higher order [six- and seven-variable] models were also investigated but did not improve accuracy. This work produced an automated EITL web interface delivering accurate and timely actionable public health information.

Learning Areas:

Protection of the public in relation to communicable diseases including prevention or control
Public health administration or related administration

Learning Objectives:
Name and define the steps involved in the EITL process for robust database construction in the context of public health freshwater quality monitoring. Describe the process employed in creating a predictive, high-sensitivity model for pathogen indicator levels.

Keyword(s): Water Quality, Decision-Making

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: I am qualified as PI and Director of the Laboratory for Public Health Informatics and Genomics which conducts the studies for which this abstract is based. My lab develops methods, collaborates on studies and conducts validation analysis on a wide spectrum of research with population-wide health impacts.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.