Prototype Statewide Interactive Mapping Service Preterm birth, 2000-2002 Environmental Health Investigations Branch California Department of Health Services.

Prototype Statewide Interactive Mapping Service

Preterm birth, 2000-2002

Environmental Health Investigations Branch

California Department of Health Services

2

High-resolution, continuous health outcome maps

• Tap into narratives that stakeholders want told

– People identify communities, including vulnerable communities, by their geography

– Policy makers are invested in understanding the geography of health problems and the resources to address them

– Concerns about environmental impacts on health are often expressed in spatial terms

• Can be an effective communication tool

3

Needs for a mapping function• Flexibility of views

• Interactive

• Accessible without specialized technology

• Complete protection of confidentiality

• Provides two kinds of information:

– Rate of health outcome

– Whether rate is significantly elevated relative to a standard

4

Representing disease risk• Each map has potential to mislead

• Which map facilitates the most productive conversations?

Asthma hospitalizations by ZIP code

Asthma ER visits by density estimation

5

Representing disease riskChoropleth

• Data from this map corresponds directly to what you would get from a table

Continuous

• Uses point (address level) data

• Risk varies continously (→more intuitive)

• A mathematical function is used to represent rate “at a point” (→less intuitive?)

• Example: Spatial loess function

ZIP CODE RATE

94612 8.3%

94705 9.1%

94610 7.2%

94701 9.9%

6

Interactive Mapping

7

Interactive Mapping

8

Interactive Mapping

9

Interactive Mapping

10

Interactive Mapping

11

Future plans• Presently only have prototype; ultimately want

publicly-viewable maps with supporting text

• Revise calculation algorithm to eliminate “phantom” hot spots, streamline computation

• Birth outcomes: – Preterm birth

– Low birthweight at term

• Cohorts:1996 – 1997 2002 – 20031998 – 1999 2004 – 20052000 – 2001 2005 – 2006

12

Considerations• This works for non-aggregated point

data

• Both cases and controls (ie, numerator and denominator) must be point data

• If the outcome rate is based on a birth cohort (preterm birth, SIDS, pediatric cancers, birth defects), controls can be birth records

• For outcomes among older children or adults, is there a source of control data?

13

Considerations• This approach may not calculate meaningful

rates if the outcome of interest is very rare

• Perhaps in this case a continuous rate surface doesn’t really interest us, but rather the detection of clusters?

• Cluster detection should be considered a different category of functions

– Statistical significance contours of loess

– SatScan

– SpatClus

– Etc.

14

Questions for TIAG• Would you or your constituents ever use a

service like this? If so, how?

• What other information or services would be necessary to make them useful?

– Supporting/explanatory material?

– Tabular data?

• What other health data would it make sense to display like this?

– What outcomes are most important?

– What outcomes do we have point data for?