No Slide Title

Download Report

Transcript No Slide Title

Acknowledgements
Epidemiologic Query & Mapping System
Sherrilynn Fuller
Principal Investigator
Public Health System Linkages – Bench to Bedside and Beyond
Richard Hoskins
Cathy O’Connor
Clark Johnson
Patrick O’Carroll
EpiQMS URL
http://198.187.0.45/EpiQMS/
Also EPIQMS in www.google.com
Components
Components
• WWW based - high speed access by local health
• Three user levels (public, practitioner, need-to-know)
• Rates with statistical measures
• Charts & graphs, time trends
• Deals with small numbers (empirical Bayes spatial modeling)
• Static, dynamic and full GIS mapping
• Multiple geographies
• Queries a dynamic database (no or little on-line calculation)
• Allows central Q&A (software & data & statistical measures)
• Comprehensive security model
• Individual id protection (available dimensions)
• Only aggregated data
• Tutorials
ABORTIONS 1991-1999
Datasets for EpiQMS
BIRTHS CERTIFICATE 1980-1999
CANCER Registry
CENSUS DATA 1980 1990 – 1998 2000
(state, county, census tract, legislative district, school district, Zip Code,
SES cluster zones, climate zones, rural areas)
COMMUNICABLE DISEASE 1980-1997
DEATHS 1980-2000 ICD9 – ICD10
HOSPITALIZATIONS 1990 -1999
HOSPITALIZATIONS (EPI FILE) 1990-1999
INFANT DEATHS 1981-1999
STD 1993-1999
Available now
TUBERCULOSIS 1992-1999
In preparation
Requested
HIV 1992 – 1999
Health of Washington State
Youth Violence
Crime , housing
Why do this?
• Original and still primary objective:
Communicable disease tracking
•
•
•
•
Geographically oriented (maps)
Small numbers
Ease of use and access
Low cost for users
Objectives
• Ease of access to public health data by all citizens
while paying strict attention to individual privacy.
• Allow medical practitioners routine access to support
assessment and surveillance in local health departments,
communities, WA DOH, and public health research.
• Get people who use public health data to
think geographically. Many geographies, some non-standard.
• Uniformity of epidemiologic measures.
• Offer on-line instruction in how to use and intrepret
public health data.
• Software burden is on DOH not users.
• Allow down loading of information – tables, charts, maps.
How it works …
Diagram
DOH databases
Population data
Web server
Geocoded data
SAS
preprosessing
SAS
PreProcessing
Data formating
Dynamic mapping
engine
Full GIS
DOH:Secure Data Server
No identifiers !
Static Maps
Aggregated data
EpiQMS
database
EpiQMS
Internet
Engine
SQL server
EpiQMS: Data Server
WWW users
citizen users
practitioners need-to-know
Indexing - primary key
Aggregating events by:
Disease
Geography
Year
Age
Race
Yes
Sex
Yes
No
Yes
No
No
Generates Index:
15
Breast cancer
X
O
X
O
O
15XOXOO
Cancer Registry security model level I
User class
Public
Available selected selected
5 or 10 year
Dimensions Disease Geography Year Age group Race
3
11
County
1990-99
Zip Code
all
10 year
legislative district
Sex
A
M
B
F
I
T
W
all
Practitioner
4
all
County
1990-99
Zipcode
all
10 year
A
M
B
F
legislative district
I
T
census tract
W
all
Need-to-Know
all
all
County
1990-99
Zipcode
all
legislative district
census tract
block group
SES clusters
5 year
10 year
A
B
I
W
all
M
F
T
Concurrent dimensions security model level II
User class
Available
Dimensions
selected
Disease
selected
Geography
3
11
Public
4
all
Year
5 or 10 year
Age group
Race
Sex
3 out of 6
all
10 year
all
all
x
x
x
x
x
x
x
x
4 out of 6
10 year
all
x
x
x
x
x
x
x
all
all
5, 10 year
all
x
x
x
x
x
x
x
x
x
x
x
x
x
Practitioner
x
all
Need-to-Know
all
x
x
x
x
x
all
x
x
x
x
all
x
x
Who decides which users get various levels of
access?
User class
Available
Dimensions
selected
Disease
selected
Geography
Year
5 or 10 year
Age group
Race
Sex
3
11
3 out of 6
all
10 year
all
all
x
x
x
x
x
x
x
Public
x
4
all
4 out of 6
10 year
all
x
x
x
x
x
x
x
all
all
5, 10 year
all
x
x
x
x
x
x
x
x
x
x
x
x
x
Practitioner
x
all
Need-to-Know
all
x
x
x
x
x
all
x
x
x
x
all
x
x
Tools to help data owners decide:
• Probability studies
• Count suppression
• Data “owners”
• Not EpiQMS team
Map
Beginning
to think geographically ...
Thematic Maps
Trivial pursuit
word: choropleth
Issues in
Map
thematic
mapping
Natural break
Equal ranges
Equal counts
Different
conclusions
can be drawn from
maps of the
same data.
The Modifiable Areal
Unit Problem (MAUP)
MAUP
A form of ecological fallacy associated
with the aggregation of data into
areal units for geographical analysis.
There are two effects:
Scale effect:
The larger the unit of aggregation, the larger, on
average, is the correlation between two variables.
Aggregation effect: By aggregating data into different blocks, you
can get different correlations.
1960 election:
+0.44 correlation between rural non-farm voting for Nixon in
using Census nine-region division
-0. 22 correlation using the Census four-region division.
How to estimate disease
rates in “small” areas?
Empirical Bayes estimation
• Smoothing to reflect confidence
of local estimation of risk
• Prior knowledge of about rates and the
observed data are used to develop a prior
distribution
posterior
likelihood

distribution
of data

• previous data
• intuition
• good guess (or even a bad one)
• the data itself - Empirical Bayes
prior
distribution
Mean (smoothed rates)
std error (Bayesian confidence intervals)
Map
Deaths from breast
cancer in age 35-44
women
No Bayes
Blank areas indicate no deaths
Bayes
Breast Cancer
Bayesian Rate Ratio
0.00 to 0.75
0.75 to 1.25
1.25 to 2.25
2.25 to 12.00
Other
0
30
60
90
Miles
Zipcodes
What does it take to run EpiQMS?
Fast!
User
• Internet Explorer
• Internet connection 56k or >
• Two plug-ins which are easy to deal with.
(SVG for maps, ChartFX for charts)
DOH
• SQL server
• ChartFX – charting software
• SAS for the prep of data
• Visual Interdev – standard Internet site development tool.
• RoboHelp – help system development package
http://198.187.0.45/EpiQMS/