Protecting Patient Confidentiality in Merged HIV/STI Surveillance Data ‘Virtual Integration’ Protecting Patient Confidentiality in Merged HIV/STI Surveillance Data ‘Virtual Integration’ National STD Prevention Conference, Chicago 2008 Mark Stenger Epidemiologist III Washington State Department of Health Infectious Disease & Reproductive Health Assessment Unit
Guiding Principles o Patient confidentiality is paramount o The integrity of surveillance data must be preserved o As much as possible, unintended consequences of linking disease registries should be anticipated o Stakeholders should be consulted
The Legal Environment o Washington State Administrative Code ( ) authorizes using HIV surveillance data: o To link with other name-based public health disease registries when doing so will improve ability to provide needed care services and counseling and disease prevention o For contacting the HIV-infected individual to provide post-test counseling or to contact sex and injection equipment-sharing partners, including spouses; or o For carrying out an investigation of conduct endangering the public health or of behaviors presenting an imminent danger to the public health pursuant to RCW or
Confidentiality Issues in STI/HIV Registry Matching o Infection with bacterial & viral STIs may provide evidence of ongoing HIV transmission risk behaviors at the individual patient level o Purposeful or negligent HIV exposure may have legal ramifications in some jurisdictions o Match sensitivity/specificity issues should ethically preclude use of matched records for any patient-level action
Why Should We Care About HIV/STI Co-morbidity? o Infection with bacterial & viral STIs increases susceptibility to HIV infection and increases the risk of HIV transmission o Management of STIs among people with HIV is potentially more difficult, leading to more complications and greater cost of treatment o Synergistic relationship between HIV and STI epidemics Core risk groups Bridging populations
What Would We Want to Know About HIV/STI Co-morbidity? o Person: What are the characteristics of people living with HIV disease? Who gets STIs? Who is HIV+ and gets infected with STIs? Who is HIV+ among people getting STIs? What are the behavioral risks for all of the above? o Place: Are there specific geographic clusters of STI incidence among HIV+? What clinics/clinic types have the highest prevalence of HIV among STI patients? o Time: Has incidence of STIs increased, decreased or remained stable among HIV+? How do trends in co-infection compare to overall STI rates? Is the prevalence of HIV among STI cases changing over time? o Disease: What specific STIs are most likely to be diagnosed among HIV+? Are treatment outcomes or incidence of repeat infection different among HIV+ versus HIV- patients?
How Are HIV and STI Information Combined in Washington State? 1) Records with HIV+ patient’s information are matched against list with STI patients 2) Temporary file with matching records created: no names in this temporary file HIV Patient Records (HARS) STI Patient & Case Records (PHIMS-STD)
How Are HIV and STI Information Combined? 3) Temporary data warehouse constructed with STI incidence, HIV status, date of HIV dx, specific STI, date of STI dx, HIV risk, facility of HIV, facility of STD and patient demographics 2) Temporary file with unique ID numbers for matching records HIV Patient RecordsSTI Case Registry 4) Analysis of co-morbidity from temporary data warehouse
How Are HIV and STI Information Combined? 5) Temporary file with ID Numbers for matching records and temporary data warehouse are securely wiped – no permanent link preserved between HIV and STI surveillance systems! 6) Value-added analysis such as charts, graphs, rates and maps are preserved and distributed to stakeholders
What Data Elements Are Needed to Match Across Registries? HIV Patient Data from HARS or other data source – Patient First and Last Name – Date of Birth – Sex/Gender/Demographics – Unique record ID STI Patient Registry (Unique Patients) from STD-MIS or other source – Patient First and Last Name – Date of Birth – Sex/Gender/Demographics – Unique record ID Patient identifiers used only for the initial match!
What Data Elements Are Used For Analysis? HIV Data – Risk/Mode of Exposure – Dates of Diagnosis for HIV and AIDS – County at diagnosis or other geographic unit of interest – Facility or facility type at diagnosis STI Data – Disease – Date of diagnosis/exam – Treatment status – Provider characteristics – County of residence at exam or other geography
How Are Data Organized For Analysis? STI Perspective STI Incidence Records have a variable indicating HIV status at the date of exam which allows calculation of %HIV+ among incident STI cases HIV Prevalence among Gonorrhea Cases by Year of Diagnosis Washington State,
How Are Data Organized For Analysis? HIV Perspective Male Gonorrhea Rate per 10,000 Among HIV+, 18 – 44 Washington State, HIV/AIDS Case Records have a variable indicating whether this person had an STI subsequent to their HIV infection
What Other Kinds of Analysis Are Possible? Moderately simple statistical analysis can identify characteristics associated with HIV status among STI patients or risks among HIV positive for STI incidence: Factors Associated with Being HIV+ at Time of STI Diagnosis STI Cases Diagnosed in WA, 1996 – 2006
What Other Kinds of Analysis Are Possible? Trend Analyses, Comparison of GC Rates for HIV+ vs. HIV- Males Trend analysis and comparisons between HIV+ and presumed HIV- populations can be fairly done in a straight forward way
What Tools Are Needed For Virtual Integration of HIV and STI Data? Software – Excel or other charting programs – SAS for running matching/merging programs (free from CDC) Staff Capacity – Knowledge of underlying databases – Basic knowledge of Excel or graphics – Ability to export data from HIV and STI surveillance systems – Basic SAS knowledge Expert coding not required to run existing code and basic analyses More advanced training for statistical models
Match Quality Concerns o Manual review of matched records demonstrates that approximately 85% of ‘true’ matches can be captured in automated processes – manual review can enhance sensitivity o Specificity is good: manual review found few ‘false’ matches at higher point values
“Every great advance in science begins with the audacity of imagination.” J. Dewy
Acknowledgements Washington State Department of Health, IDRH Assessment Unit, STD/TB Services Section – Jason Carr – Maria Courogen U.S. Center For Disease Control and Prevention – – OASIS Workgroup – Lori Newman For additional information or help implementing matching contact: Mark Stenger Washington State Department of Health
Sample Co-infection Analyses from Washington State
Persons Presumed Living with HIV* Washington State, *Persons diagnosed in Washington State and not known to be deceased, reported through 11/2006 Source: Washington State Department of Health, HIV/AIDS Surveillance Program, Nov 2006
Chlamydia Cases & Incidence Rates* Washington State, *Crude rate, not age adjusted Source: Washington State Department of Health, STD/TB Services Section, Oct 2006
Washington State, Gonorrhea Cases & Incidence Rates* Washington State, *Crude rate, not age adjusted Source: Washington State Department of Health, STD/TB Services Section, Oct 2006
Washington State, Early Syphilis Cases & Incidence Rates* Washington State, *Early syphilis, crude rate, not age adjusted Source: Washington State Department of Health, STD/TB Services Section, Oct 2006
Methods o HIV (>18,000 records) and STD (256K records) case registries were matched using multi-element, weighted deterministic match algorithm o Point-value cutoff determined to balance greater specificity against slightly lower sensitivity o STD and HIV-specific variables merged for matched cases to facilitate analyses
Match Results o 3,812 unique person-matches between STD and HIV registries For HIV/AIDS diagnosed and reported through 11/2006 and for STDs diagnosed and reported between 1/1992 and 11/2006 o 2,753 episodes of STD infection subsequent to HIV infection among 1,597 unique persons 1992 – episodes of CT infection 1397 episodes of GC infection 457 episodes of infection syphilis 121 episodes of late syphilis 188 episodes of other STS (herpes, NGU, etc.)
Match Results (cont.) o 1,023 (64%) HIV+ persons with 1 episode of STD subsequent to HIV infection o 574 (36%) HIV+ persons with 2 or more episodes of STD infection (range 2 = 19) subsequent to HIV infection o 1,495 (93.6%) Male, 102 (6.4%) Female
Race & Ethnicity HIV+ people diagnosed with STDs Subsequent to HIV Infection Males Females
Mode of Exposure HIV+ people diagnosed with STDs Subsequent to HIV Infection Males Females
STDs Diagnosed Subsequent to HIV Infection Males Females
Trends in Co-infection
HIV Prevalence among Chlamydia Cases by Year of Diagnosis Washington State,
HIV Prevalence among Gonorrhea Cases by Year of Diagnosis Washington State,
HIV Prevalence among Early Syphilis Cases* by Year of Diagnosis Washington State,
Chlamydia Incidence Rate per 1,000 HIV+ vs. Presumed HIV- Persons Washington State,
Gonorrhea Incidence Rate per 1,000 HIV+ vs. Presumed HIV- Persons Washington State,
Syphilis* Incidence Rate per 1,000 HIV+ vs. Presumed HIV- Persons Washington State, *Primary, Secondary & Early Latent
Gonorrhea among 18 – 44 Year Old Males HIV+ verses Presumed HIV-
Male Gonorrhea Rate per 10,000 Presumed HIV-, 18 – 44 Washington State,
Male Gonorrhea Rate per 10,000 Known HIV+, 18 – 44 Washington State,
Trend is significant: at p < Trend Analyses, GC Rate for HIV- Males
Trend is significant: at p < Trend Analyses, GC Rate for HIV+ Males
Trend Analyses, Comparison of GC Rates for HIV+ vs. HIV- Males Z = 1.7x10 5, trends are significantly different
New HIV diagnoses by gender, Washington State, % of recently diagnosed cases have been female Cases reported as of 8/31/06
Caveats HIV and STD Surveillance data vary in completeness by year and by disease: Syphilis > HIV > GC > CT Registry match sensitivity/specificity issues introduce uncertainty