Using the NAACCR Geocoder to Perform In-House Geocoding of the Alaska Cancer Registry Database David O’Brien, PhD, GISP Data Analyst Alaska Cancer Registry Section of Chronic Disease Prevention & Health Promotion Alaska Department of Health and Social Services Presentation Topics Geocoding nature resource data vs people Alaska Cancer Registry & cancer data reporting laws Cancer basics Why do people get cancer? Why geocode cancer data? Cancer cluster studies Geocoding the Alaska Cancer Registry database Geocoding Natural Resource Data Most geocoding in Alaska related to natural resources Oil wells, gas wells, mines, gravel pits Commercial fishing, lumber, aquaculture, mariculture, fish hatcheries Harbor seal haulouts, sea lion haulouts, salmon streams, sea bird colonies, bald eagle nests How Were These Natural Resources Geocoded? Field notes & maps transcribed to final paper maps, paper maps published either alone, in reports, or in atlases, maps digitized GPS coordinates for some features (bald eagle nests, sea bird colonies) Sometimes problems with data layers digitized in the wrong datum or arbitrary coordinate system So How Do You Geocode People? People are geocoded differently from natural resources Data from cancer patients obtained from confidential medical records Types of location data: Where they live, where they work, what hospital or doctor they went to Cancer registries concentrate on location of residence at time of diagnosis About the Alaska Cancer Registry Consists of a database of all cancer cases diagnosed and treated in Alaska since 1996, plus Alaska residents diagnosed or treated in other states AK Department of Health & Social Services Funded by the National Program of Cancer Registries (NPCR) through the US Centers for Disease Control and Prevention (CDC) U.S. Cancer Reporting Laws Public Law 102-515 (Oct 1992): Cancer Registries Amendment Act: “each form of invasive cancer with the exception of basal cell and squamous cell carcinoma of the skin and each form of in situ cancer except for carcinoma in situ of the cervix uteri” Benign Brain Tumor Cancer Registries Amendment Act (Oct 2002): Cases diagnosed on or after 1/1/2004 Alaska Cancer Reporting Laws 7 AAC 27.011 (Jan 1996): Established reporting requirements for our statewide cancer registry Benign brain reporting added to 7 AAC 27.011 Jan 2004 Who is Required to Report to ACR? Hospitals Physicians Independent pathology laboratories Outpatient centers Home health agencies Hospices Nursing homes Intermediate care facilities Purpose of ACR To identify all reportable cases of cancer in Alaska in order to provide information on the burden of cancer, types of cancer, and changing patterns of cancer among residents of our State. Goals of ACR Obtain complete, accurate and timely data on all newly diagnosed cancer cases Monitor trends for unusual patterns Identify areas of the state with unusually high incidence or mortality rates Provide data to: AK Comprehensive Cancer Control Program AK Breast & Cervical Cancer Health Check Program Some Basic Information about Cancer Cancer is not one disease Cancer is more common than most people realize About 50% of men and 33% of women will develop cancer About 24% of all deaths nation-wide are from cancer Highest Ranked Cancers for AK Incidence: Breast, Prostate, Lung, Colorectal, Bladder (56%) Mortality: Lung, Colorectal, Breast, Pancreas, Prostate (55%) Lung alone is 29% of all cancer deaths Why Do People Get Cancer? Personal behavior Smoking & chewing tobacco Excessive alcohol consumption (esp. in smokers) Obesity / lack of exercise Type of diet (high in red meats, processed meats, low in fruits & vegetables Why Do People Get Cancer? Workplace exposure Example: asbestos, certain industrial chemicals Genetics / family history Unknown factors Why Do People Get Cancer? Other Risk Factors: Previous exposure to ionizing radiation for medical treatment of chest (breast & lung cancers) or head and neck (thyroid cancer) Infection from Human Papilloma Virus (cervical, anal, & mouth cancers) Stomach infection from H. Pylori bacteria (stomach & pancreatic cancers) Certain illnesses: Type 2 diabetes & inflammatory bowel disease (colorectal cancer); Hepatitis B & C (liver cancer) Why Do People Get Cancer? Other Risk Factors: Long-term use of oral contraceptives or hormone replacement therapy, esp. estrogen plus progesterone (breast cancer) Radon gas exposure in the home & secondhand smoke exposure (lung cancer in non-smokers) Excessive sun exposure and use of tanning beds, especially at a young age (skin cancer) Why Do People Get Cancer? Very rarely due to exposure to pollution found in the environment, but this tends to be the highest concern Why Geocode Cancer Patients? Incidence case counts and rates reported by borough/census area Mortality case counts and rates reported by borough/census area Rate per 100,000 102.8 to 218 92.6 to 102.8 82.4 to 92.6 72.2 to 82.4 28 to 72.2 (5) (1) (5) (5) (9) Lung Cancer Incidence Rates by Borough/Census Area, 1996-2004 Rate per 100,000 220 to 270 170 to 220 120 to 170 70 to 120 20 to 70 (2) (6) (8) (2) (4) Prostate Cancer Incidence Rates by Borough/Census Area, 1996-2004 # of Cases 248 to 1,839 49 to 247 33 to 48 16 to 32 2 to 15 (5) (5) (5) (6) (6) 32 32 25 25 33 33 40 40 473 473 44 44 14 14 16 16 378 378 66 66 1,839 1,839 39 39 343 343 24 24 44 66 22 15 15 248 26 248 26 67 67 93 93 49 49 37 37 116 116 10 10 22 22 Number of In Situ & Malignant Breast Cancer Cases, AK Residents, 1996-2005 Why Geocode Cancer Patients? Determine census tract for each cancer case Census tracts can be linked with socio-economic data collected by the US Census Bureau Researchers can conduct quality of care studies Geocode cancer data necessary to conduct cancer studies for the general public and communities with cancer concerns Why Are There Cancer Concerns? General Public Personal experience with cancer Neighbors getting cancer Something in the water? Communities Usually some sort of past event as a trigger Oil or chemical spill, pollution from industrial or military facility Responding to the General Public Phone call: Listen, ask questions, be understanding, the caller is usually upset Concern is usually due to neighborhood aging Extended phone interview if caller not satisfied Review information and determine level of investigation Responding to the General Public Provide copy of Cancer and the Environment, by National Cancer Institute Responding to Communities Usually referred to us by AK DHSS’s Section of Epidemiology or CDC’s Agency for Toxic Substances and Disease Registry (ATSDR) Proceed with a full cancer study if there is a triggering event of concern If no triggering event, start with a smaller study as a “screening” measure Past Community Cancer Studies Chemical spill at refinery in North Pole Pollution from old military facilities near Savoonga and Gambell on St. Lawrence Island Pollution from old military facilities near Yakutat Cancer Study Addresses These Questions: Are the number and types of cancers reported in the community unusual (cases & deaths)? Are the number of observed cancers greater than the number of expected cancers (cases & deaths)? First Study: Number and Types of Cancers Alaska Cancer Registry for reported cancer cases Bureau of Vital Statistics’ State Mortality Database for reported cancer deaths Cancers per year Compare most common cancers to top ranked cancers in the state Second Study: Observed vs. Expected Cancers Determine the number of reported cancers for the community or for an affected census tract in the community Calculate the number of expected cancers Is observed greater than expected? If so, is the difference statistically significant? Past ACR Geocoding Practices Traditional geocoding was not done Only about 10 towns in Alaska have more than one census tract Used a table of census tract numbers and town names to populate the database Census tracts only 34.3% complete Preparing for Geocoding Create special “Address at Geocoding” Fields Reduce number of records w/ unknown addresses Export data and put into format acceptable to the geocoding software “Address at Geocoding” Fields Stores address that was used for geocoding Initially the same as the patient address Over time, patient address might get updated For future geocoding, 2 addresses are compared Geocode record if: Address at Geocoding is blank The 2 addresses are different Reduce Unknown Addresses Identify records w/ unknown addresses 14,679 records (36.3%) PO Box, rural route, pouch, general delivery, unknown P.O. Box, PO Box, P O Box, P. O. Box, PO Bx, POB, Box, BX HC, PS, SR, RT, RR, Route Unk, Unknown, General Delivery, No Address, Bad Address, Not Deliverable Link unknown address records with source records 260 records resolved Reduce Unknown Addresses Link unknown records with the Alaska Permanent Fund Dividend (PFD) Application database 578 records resolved from mailing address 4023 records resolved from physical address Reduce Unknown Addresses Manually review remaining non-numeric addresses Sorted all addresses in reverse-alpha order and Manually review addresses starting with letters Apartment buildings Mobile home parks Long-term care facilities Correctional facilities 25 records resolved Reduce Unknown Addresses Total of 40,389 records Started with 14,679 unknown addresses (36.3%) Ended with 9,793 unknown addresses (24.2%) Format Data for Geocoding Exported data from database Prepared a geocoding data file with specific fields Year of Diagnosis Unique ID: combination of Patient ID and Tumor Number Address at diagnosis -- Street Address at diagnosis -- City Address at diagnosis -- State Address at diagnosis -- Postal Code Interpreting the Geocoder Output Don’t just assume all the output is correct! Verify all records have been geocoded Problem with geocoder timing out? Review “GIS Coordinate Quality” codes GIS Coordinate Quality Manually research records geocoded to the state or borough centroid Review records geocoded to city centroid OK for PO Boxes Not OK for street addresses Census tract assignment for cities with multiple census tracts need to be recoded as 999999 Re-geocode corrected addresses Geocoding Results Certain towns were not recognized by the geocoder and were geocoded to the state centroid Kongignack, Auke Bay, Nikiski, Ward Cove, Funter Bay, Cube Cove Ketchikan with zip code 99950 Manually assigned to the city centroid and the corresponding census tract No cases assigned to a borough centroid Geocoding Results 160 numerical street addresses geocoded to city centroid Manually research, corrected in database, re-geocoded All cases geocoded to city centroid in Anchorage, Fairbanks, Wasilla, Juneau, Kodiak were PO Boxes or UNKNOWN Zip codes specific to a PO Box and do not have a Zip Code Tabulation Area (ZCTA) Manually changed census tract to 999999 Geocoding Results Used unique ID field to link back to the cancer registry database Updated geocode-related data fields 1996-2010 cancer data geocoded with Census 2000 data (35,315 records) 2011+ cancer data geocoded with Census 2010 data (3,302 records) Successfully geocoded 38,617 of 40,436 records with known census tracts Census tract completeness increased from 34.3% to 95.5% Acknowledgements This presentation was supported by the cooperative agreement number U58/DP003856-01 (DP-1205-01) from The Centers for Disease Control and Prevention. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of CDC.
© Copyright 2018