Informatics Resources for Supporting Biomedical and

Environmental Research

BMIC TeamUniversity of Utah

Air Quality Competition

Overview

• Team Members• Background• OpenFurther• Objective• Methods• Next Steps• References

Team Members

Nicole Burnett

Peter Mo

Naresh Sundar Rajan

Randy Madsen

Ram Gouripeddi

Julio Facelli

Background

• Air Quality has been associated with various adverse health effects• Asthma• Cardiovascular disease• Respiratory infections• Cancers• Impaired glucose tolerance during pregnancies1–4.

Biomedical Research

• Researchers at the University of Utah are embarking on clinical studies to understand associations between the peculiar Air Quality patterns in Salt Lake City and clinical conditions:• Cerebral venous thrombosis• Exacerbations of idiopathic pulmonary fibrosis• Suicide• Reproductive outcomes• Cancers.

Salt Lake City Air Quality

• Prone to winter inversions where colder surface temperatures trap fine particulate matter (PM2.5) which poses serious health concerns5.• Summer months in the valley have

increased ozone (O3) levels.

Clinical Research Air Quality Requirements

• Elicited use-cases from clinical researchers.• Primary need: understand the risks associated with being exposed with

various air pollutants.• Manifestations following exposure could occur

• Immediately• After a lag phase• Could persist over long durations.

• Pathophysiology and mechanisms of many of these manifestations are not well understood at this time.• Current research mainly associates single pollutant and clinical conditions,

future areas of research could include exposures to multiple pollutants.

Utilizing Air Quality Data in Clinical Research: Actionable Causes

• Integrating air quality and biomedical data needs to support• Spatio-temporal variations of air pollutant concentrations• Location of individuals• Timing of the occurrence of conditions.

• Requirements for the granularity of AQ data vary from hourly readings to monthly averages depending on the study.• Difficulties in understanding and integrating air quality data with

clinical data is a limitation in performing research.

Technical Infrastructure

• In order to enable ease of federation/integration of air quality and clinical data, we are extending the OpenFurther platform to support environmental data.• OpenFurther6,7 is an informatics platform that supports federation and

integration of data from heterogeneous and disparate data sources.• Supports clinical and translational research by bringing data directly

to researchers without requiring the technical expertise to query large databases or knowledge about the data source.

OpenFurther Components

• Query Tool• Federated Query

Engine• Data Source Adapters• Admin & Security

Components• Virtual Identity

Resolution on the GO (VIRGO)

• Quality & Analytics Framework

• Metadata Repository• Terminology/Ontology

Server• Air Quality Modelling

Unit

Air Quality - Clinical Data Federation

• Demonstrated the feasibility of federating air quality data from the Environmental Protection Agency (EPA) with clinical data from the University of Utah using OpenFurther8,9.• We were able to select different cohorts of patients living in SLC

county and having clinical conditions (e.g. asthma) occurrences that were related to the temporal variations of air pollutant concentration.

Asthma in January 2014

• 615 patient with a diagnosis of asthma in Salt Lake County and average PM2.5 28 micrograms

Asthma in January 20th 2014

• 25 patient with a diagnosis of asthma in Salt Lake County and average PM2.5 50 micrograms• Worst Inversion Day

Air Quality Monitoring in SLC County

• Three monitoring stations in Salt Lake County• Need for cross-linking patient

locations and condition occurrences: High Resolution Spatio-temporal Air Quality Grid

High Resolution Spatio-temporal Air Quality Grid

• Current air quality modelling used in OpenFurther – EPA’s modelling10

• 12*12 kilometer resolution• Validated on the east coast• Altitude

• Current work with OpenFurther will include• Ability to select data models that provides best estimates for areas and times air

monitor measurements are unavailable along with uncertainties.• Extend the framework to federate other AQ data sources

• Balloon11

• Satellite-derived aerosol optical depth measurements4

Objective

• Validate EPA’s modeled data with local measurements for Salt Lake County.• Local Measurements: Data collected for the work in “Relationship

between particulate air pollution and meteorological variables in Utah’s Salt Lake Valley”9

Methods

• Analyze differences in Air Quality Index values using EPA modeled data and local measurements.• T-test using a random sample of EPA modeled data and local

measurements.• Clustering• Differences in intra-cluster and inter-cluster mean squared errors using EPA

modeled data and local measurements.

Air Quality IndexBreakpoints Equal these AQI's

O3 (ppm) 8-hour O3 (ppm) 1-hour1 PM2.5 (µg/m3) 24-hour PM10 (µg/m3) 24-hour CO (ppm) 8-hour SO2 (ppb) 1-hour NO2 (ppb) 1-hour AQI Category

0.000-0.059 0.0-12.0 0-54 0.0-4.4 0-35 0-53 0-50 Good

0.060-0.075 12.1-35.4 55-154 4.5-9.4 36-75 54-100 51-100 Moderate

0.076-0.095 0.125-0.164 35.5-55.4 155-254 9.5-12.4 76-185 101-360 101-150Unhealthy for

Sensitive Groups

0.096-0.115 0.165-0.204 355.5-150.4 255-354 12.5-15.4 4186-304 361-649 151-200 Unhealthy

0.116-0.374 0.205-0.404 3150.5-250.4 355-424 15.5-30.4 4305-604 650-1249 201-300 Very Unhealthy

(2)0.405-0.504 3250.5-350.4 425-504 30.5-40.4 4605-804 1250-1649 301-400 Hazardous

(2)0.505-0.604 3350.5-500.4 505-604 40.5-50.4 4805-1004 1650-2049 401-500

http://www.ecfr.gov/cgi-bin/text-idx?SID=0420a09b9ec269dff4cc7a976fc7341d&node=ap40.6.58_161.g&rgn=div9

Clustering• Seasonal clusters

generated using the simple K-means algorithm run on the Whiteman’s dataset.• X-Axis=Season• Y-Axis=PM2.5 Value• Winter has the highest

PM2.5 Values• Clusters (colors) naturally

fall into Seasons

Clustering• Monthly clusters

generated using the simple K-means algorithm run on the Whiteman’s dataset.• X-Axis=Month• Y-Axis=PM2.5 Value• Similar to Seasonal

clusters• Clusters (colors) naturally

fall into Seasons

Clustering• Day of Week clusters

generated using the simple K-means algorithm run on the Whiteman’s dataset.• X-Axis=Day of Week• Y-Axis=PM2.5 Value• Weekends vs. Weekdays

have similar values!

Clustering• Trend from year 1999

to 2013• X-Axis=Year• Y-Axis=PM2.5 Value• There are no noticeable

trend over the years!

Clustering• Morning vs. Afternoon• X-Axis=5am & 5pm• Y-Axis=PM2.5 Value• There are no noticeable

difference over morning or after work hours!

Next Steps

• Complete Validation• Validate other air quality models• Create a public repository for storing local measurements.• Harmonization and integration of local measurements• Incorporate into OpenFurther

References1. Kesten S, Szalai J, Dzyngel B. Air quality and the frequency of emergency room visits for asthma. Ann Allergy Asthma Immunol Off Publ Am Coll Allergy

Asthma Immunol. 1995 Mar;74(3):269–73. 2. Weisel CP, Zhang J, Turpin BJ, et al. Relationships of Indoor, Outdoor, and Personal Air (RIOPA). Part I. Collection methods and descriptive analyses. Res Rep

Health Eff Inst. 2005 Nov;(130 Pt 1):1–107; discussion 109–127.3. Fleisch AF, Gold DR, Rifas-Shiman SL, et al. Air Pollution Exposure and Abnormal Glucose Tolerance during Pregnancy: The Project Viva Cohort. Environ

Health Perspect. 2014 Feb 7 [cited 2014 Mar 7]; http://ehp.niehs.nih.gov/1307065/ 4. Kloog I, Koutrakis P, Coull BA, Lee HJ, Schwartz J. Assessing temporally and spatially resolved PM2.5 exposures for epidemiological studies using satellite

aerosol optical depth measurements. Atmos Environ. 2011 Nov;45(35):6267–75. 5. Utah Concludes Winter Inversion Season, Residents Proactively Engaged. [cited 2014 Mar 11]. http://

www.deq.utah.gov/News/docs/2014/03Mar/DAQ_NewRelease_AirQualityStats_draftv2.pdf 6. Openfurther.org7. Gouripeddi R, Schultz ND, Bradshaw RL, et al. FURTHeR: An Infrastructure for Clinical, Translational and Comparative Effectiveness Research. American

Medical Informatics Association, 2013 Annual Symposium; 2013 Nov 16; Washington, D.C. http://knowledge.amia.org/amia-55142-a2013e-1.580047/t-10-1.581994/f-010-1.581995/a-184-1.582011/ap-247-1.582014

8. Gouripeddi, R., and Julio C Facelli. “Programmatically Linking Air Quality Indicators with Clinical Data.” presented at the Air Quality, People and Health, 2nd Annual Retreat, University of Utah Guest House, University of Utah, Salt Lake City, April 14, 2014. http://www.airquality.utah.edu/files/2014/04/Ram_Programmatically-Linking-Air-Quality-Indicator-with-Clinical-Data.pdf.

9. Gouripeddi, R., Rajan, N.S., Madsen, R. Warner, P.B., Facelli, J.C., Federating Air Quality Data with Clinical Data, Annual Symposium of the American Medical Informatics Association, 2014

10. McMillan, Nancy J., David M. Holland, Michele Morara, and Jingyu Feng. “Combining Numerical Model Output and Particulate Data Using Bayesian Space–time Modeling.” Environmetrics 21, no. 1 (February 1, 2010): 48–65. doi:10.1002/env.984.

11. Whiteman CD, Hoch SW, Horel JD, Charland A. Relationship between particulate air pollution and meteorological variables in Utah’s Salt Lake Valley. Atmos Environ. 2014 Sep;94:742–53.

Thank You

OpenFurther.org