Enhancing STD Surveillance by Matching to Other Data

Enhancing STD Surveillance by Matching to Other Data Sources: A Hot Topic • • Michael C. Samuel, Dr. PH California Department of Health Services Lori Newman, MD Centers for Disease Control and Prevention

Defining Matching • • Case-based • Matching individually line-listed data to another individually line-listed source of data Ecologic • Correlate stratum-specific (e. g. county level) rates of one disease or condition with rates of another

Why Match? • • • Assess co-morbidity or the co-occurrence of diseases/conditions –> identify “hot spots” Answer specific research questions Complete missing data or correct data Case finding Analyze patterns of re-infection

Why Match? • • • Encourage collaboration and communication between programs “Mining” existing data Prioritize program activities / target limited resources

Data Sources • Diseases • • • Syphilis Gonorrhea Chlamydia NGU Herpes AIDS/HIV Cancer TB Enterics • • Vital Statistics • • Births Deaths Other related data • • Substance use Tx Incarceration Records Behavioral Data • e. g. , BRFS SES, etc. Data • e. g. , Census

Technical Issues • • • Confidentiality/Security Data formats Software • • SAS, Access, etc. Dataflux (and other matching software) STD*MIS and HARS NEDSS

Matching Criteria • • • Unique identifiers Algorithms • • Incorrect matches (false positive) Missed matches (false negative) Database size

Matching Examples: Assessing Co-Morbidity

STDs and HIV/AIDS Co-morbidity and STDs as markers of HIV risk Chlamydia Gonorrhea Syphilis HIV

California Matching Algorithms • Match 1 (Automated Exact Match) • Exact matches on: Last Name, First Name, DOB • Match 2 (“Best” Match) • Exact matches + manually reviewed matches with point values ≥ 35 • Match 3 (Loosest Match) • “Best” match + HARS records with no names that match STD records on SOUNDEX, DOB, SEX

Point System Variable Name Description Points First 3 letters of first name 15 All letters in first and last names match 10 Month of birth date 10 Day of birth date 5 YEAR Year of birth date within 5 years 15 IDENTICALYEAR Year matches identically 5 MDY Month, day, year of birth date all match 10 TRANSPOSITION Month and day are transposed 15 FIRST ALLNAME MONTH DAY *All matches with a total point value ≥ 35 were manually reviewed by two individuals to determine match validity

Co-morbidity from Three Matches Matching Algorithm Syphilis-AIDS Cases 1990 -2001 Exact Match 150 "Best" Match 184 Loosest Match 244

Percent with AIDS Diagnosis Percent of Male Syphilis Cases with AIDS Diagnosis California Department of Health Services, Office of AIDS. Epidemiological Studies Section

Washington State - HIV Prevalence Among Infectious* Syphilis Cases, 1994 - 2002 100 Number of Cases Percent HIV+ All Infectious Syphilis Cases t Percent HIV+ t 80 t t t 30 40 20 20 0 50 40 t 60 60 t t 1994 1995 t 1996 10 t 1997 1998 Year *Primary, secondary and early latent syphilis 1999 2000 2001 2002 0

Washington State - HIV Prevalence Among Reported Chlamydia Cases, 1994 - 2002

Trend in Rate of Change, Reported STDs*, PLWHA and STDs Reported Among PLWHA 1998 - 2002 35 Percentage Increase l t PLWHA 30 l STDs Among HIV+ 25 s All STD Cases l 20 l 15 10 t 5 s 0 98 -99 t s 99 -00 *Chlamydia, gonorrhea, P, S & EL syphilis only Interval s 00 -01 l t 01 -02

Detroit HIV/STD Match • • 1997 -2004 2. 8% to 4. 9% (per year) of syphilis cases coinfected with HIV • 67% of these were infected with syphilis after HIV diagnosis

Matching Example: Answering a Research Question

California Chlamydia/Birth Match • • • Assess adverse birth outcomes associated with chlamydia (CT) during pregnancy 1997 -1999; 675, 000 births, 101, 000 female CT cases 14, 000 matched cases with CT during pregnancy

CA Chlamydia/Birth Match Results Low birth weight (LBW): • 6. 6% LBW among women with CT • 4. 7% LBW among women without CT • Adjusted (for age, race, education, prenatal care) Odds Ratio = 1. 2 (95% CI 1. 1 -1. 3)

Matching Example: Completing Data

California “Family PACT” Administrative / Unilab Chlamydia Test Data, 2000 Data Elements Test Results Race/Ethnicity Gender Unilab Data Administrative Data Merged Data Complete Missing 100% Complete Missing 7% Complete

Unilab and FPACT Claims Data : Female CT Positivity By Age and Race/ Ethnicity Dec 00 -Jul 01

Family PACT Match Results/Conclusions • Precise estimates of age/race specific • • chlamydia prevalence rates Demonstrates racial disparities in CT rates from large state “safety net” provider, not otherwise available Required no additional data collection

Matching Example: Case Finding

Virginia HIV/AIDS Case Finding • TB match with HIV/AIDS found few new cases, but helped complete risk factor data (IDU) • ADAP (AIDS Drug Assistance Program) match with HIV/AIDS identified many new cases and improved timeliness of reporting

Matching Example: Re-infection

California – Repeat Gonorrhea Infection Assessment • • • Exact match on name and date of birth 1/1/2001 -12/31/2002 >26, 000 unique cases • >1, 650 (6%) re-infections or duplicates

Patients with Two or More Gonorrhea Infections* California Project Area, 2001– 2002 Duplicate? Treatment Failure? True Re-infections? * Repeat infections identifier based on patient last name and date of birth.

OASIS Matching Findings • • Substantial and increasing STD cases after HIV/AIDS; highlights potential for HIV transmission (CA, SF, WA, MA…) Lack of chlamydia / HIV co-morbidity screening of CT cases for HIV not resource efficient (WA) Little TB / STD co-morbidity (multiple sites) Successful for building data mart across diseases (NY)

Strengths of Matching • • Inexpensive, efficient way to augment knowledge Can be made easy/simple • • Can help build bridges Can provide actionable results • • • Automated matches Data warehouses NEDSS-like systems Interpret carefully Even negative match can provide info

Weakness/Limitations of Matching • • • Technically may be difficult or impossible • • No unique identifiers Database/registry may cover small and/or biased population Can be time consuming and difficult May be better ways to get data • e. g. , ask cases with one disease if they have another Confidentiality concerns May not provide information for action

General Recommendations • • • Know data sources Assure data protection Assess technical capacity and technical issues before beginning Assess likely “juice for squeeze” Collaborate with OASIS team Think ……………. …. . outside the box

Thanks to the California Matching Team STD Control Branch • Joan Chow • Denise Gilson • Mi-Suk Kang Office of AIDS • Maya Tholandi • Allison Ellman • Juan Ruiz And, • • • Kathryn Macomber, Michigan Department of Health Mark Stenger, Washington State Department of Health Jeff Stover, Virginia Department of Health

For more information contact: Michael C. Samuel msamuel@dhs. ca. gov 510 -540 -2311 or Lori Newman len 4@cdc. ogv 404 -639 -6183

Timing of Syphilis-AIDS Diagnoses (1999 -2001, “Best” Match) Timing of Infections “Best” Match (%) Syphilis >1 after AIDS diagnosis 29 (76) Syphilis within 1 year of AIDS diagnosis 9 (24) Syphilis >1 before AIDS diagnosis 0 (0) Total 38 California Department of Health Services, Office of AIDS. Epidemiological Studies Section

Scatter plot of Gonorrhea and Chlamydia Rates by Gender and State, United States 2002