Health Tracking and Disease Registries Environmental Health Investigations

Health Tracking and Disease Registries Environmental Health Investigations Branch California Department of Health Services Eric M Roberts, MD Ph. D

Discussion for Today • Connecting health and air pollution data: the underlying logic • What kinds of data can we use for communities? • Disease registries – In California – Diseases not amenable to registries – Data sources that function like registries • Health Tracking: beyond registries 2

Motivations for Air Pollution Measurement and Modeling • Which populations are burdened by pollution, and how much? • Could patterns of illness in communities be related to patterns of air pollution? • Generally it is this second question that gets us interested in disease registries and health tracking 3

Examples of data sources • Convenience samples • • • No data are Snowball samples good or bad; The question Community surveys is what kinds General surveys of questions Disease registries they can be used to Administrative or planning address data 4

Descriptive data • A child with asthma lives near the factory • 20% of the children near the factory have asthma 5

Comparisons • 20% of the children near the factory have asthma • 10% of the children far from the factory have asthma • In epidemiology, we call the patterns observed through the comparison of two or more groups associations 6

Details • Covariates – Extra variables that may also play a role in the outcome (e. g. smoking) – Which variable is of interest and which ones are just covariates depends only on inclination • Sample size – Anecdotes can be powerful when they can describe a small number of people in-depth – Large samples can be powerful because they can (often) be used to generalize – Usually we only have an opportunity to do one of these things well 7

Building blocks for observing associations Hazard or exposure Data Health Data Analysis 8

Building blocks for observing associations Hazard or exposure Data Health Data categorizes Has disease No disease Analysis 9

Building blocks for observing associations Health Data Hazard or exposure Data categorizes* Has disease No disease Exposed Not exposed Analysis *Exposure classification must describe period of time relevant to disease development 10

Building blocks for observing associations Health Data categorizes Has disease linkage No disease Hazard or exposure Data categorizes* Exposed Not exposed Analysis *Exposure classification must describe period of time relevant to disease development 11

Building blocks for observing associations Health Data categorizes Has disease linkage No disease Hazard or exposure Data categorizes* Exposed Not exposed Analysis • Effect size: How striking is this pattern? • Could pattern have happened by chance? • Do alternative variables (covariates) explain pattern? *Exposure classification must describe period of time relevant to disease development 12

Health data sources (partial list) • General survey--For a sample drawn from many groups of people, ask about disease, symptoms, covariate data • Single community survey--For a community of concern, ask about disease, symptoms, covariate data • Disease registry--For all reported cases, follow up with chart review, clinical verification, covariate data 13

Studying associations: Requirements for health data (1) • Uniform definitions of disease • Everyone should have same chance for inclusion • Many people with disease included (“sample power”) • People without disease included: denominator or control data • Linkable 14

Studying associations: Requirements for health data (2) • People with different exposures must be included • Also helpful: – Covariate data describing individual characteristics – Covariate data describing exposure patterns (commuting patterns, residential histories, etc) 15

Assuming well-designed data collection: Requirement Uniform definitions, chances for inclusion, etc. Covariate data available Linkable Many people with disease included People without disease included Range of exposures included Disease registry Community survey General survey x ? x ? ? 16

California Cancer Registry • Statewide, population-based reporting mandated since 1985 • Adds ~140, 000 cases annually • Near 100% reporting of all types (except common skin and non-invasive cervical cancers) • Includes information on demographics, cancer type, extent of disease at diagnosis, treatment, and survival 17

California Cancer Registry • The CCR is a three-tiered system: – Medical treatment facilities collect and report cancer data from their medical records. Physicians report information of cancer patients who are not referred to a medical treatment facility. – A network of eight regional registries receives these data and checks for accuracy, performs analyses, and conducts studies specific to the local area. – The Cancer Surveillance Section in Sacramento collates these data, performs additional quality control and analyzes the data on a statewide basis. 18

California Cancer Registry • Linkage potential – Linkage to external data (e. g. birth records) can provide additional information such as address at birth and/or demographics – Linkage can also facilitate identification of control subjects 19

California Birth Defect Registry • Selected counties representing ~40% of state’s births Fresno Kern Kings Los Angeles Madera Merced Orange Riverside San Bernardino San Diego San Joaquin Stanislaus Tulare • ~400 parental interviews conducted per year • Emphasis on maintaining validity of diagnoses, including delayed manifestations 20

California Parkinson’s Disease Registry • AB 2248 signed in 2004 • CDPH deputized Parkinson’s Disease Institute and UCLA • 2 -year pilot project beginning 2007 examining use of – Pharmacy records – Physician office reports 21

Diseases not amenable to registries • When do you say someone has a disease? • When do you say they no longer have the disease? • Are people arguing about who has the disease? 22

Poorly defined or contested disease definitions • Asthma – How bad do symptoms need to be before there is a diagnosis? – Does anyone ever stop having asthma? How do you know? • Autism – Who is diagnosing? – Are the shifting degrees of stigma and social privilege playing a role in diagnosis? 23

Poorly defined or contested disease definitions • SIDS – Diagnosis dependent on county coroner • Infertility – Only reported when people are trying to conceive 24

Registry-like data sources • Vital (birth and death) records – Outcomes: • Preterm birth • Low birthweight • Infant mortality – Linkable – Built-in source for denominator/control data 25

Registry-like data sources • Hospital discharge records for acute events – Outcomes: • Asthma • Heart attacks • Stroke – Limited linkage (geographic resolution = ZIP code) – Limited denominator data (census) 26

Advances • Use of address level data for geocoding • “Smoother” functions – Can show small-scale variations in risk within city boundaries – Can “borrow” regional data to stabilize risk estimates in rural areas – May help account for spatial autocorrelation--often key to understanding associations with pollution • Facilitation of uptake of technological and statistical methods is main Tracking Program focus 27

Outcome risk Smoother function: One-dimensional example Spatial coordinate 28

Outcome risk Smoother function: One-dimensional example Spatial coordinate 29

Outcome risk Smoother function: One-dimensional example Spatial coordinate 30

Preterm birth in California 31

Preterm birth in California 32

Preterm birth in California 33

Preterm birth in California 34