Transfer of Data from a Hospital System to a Local Health Department

1. Transfer of Data from a Hospital System to a Local Health Department Evelyn O. Talbott, Dr.P.H., M.P.H LuAnn L. Brink, Ph.D. February 24-26, 2009 TRACKS2009 Washington, DC

2. Thanks to… • Ron Voorhees, ACHD • Rao Guduru, ACHD • Linda Gmitter, UPMC • Adam Gronsky, UPMC • Tony Fabio, University of Pittsburgh • Tanya Kenkre, University of Pittsburgh

4. Cooperation • Data provided to Allegheny County Health Department as part of Public Health Surveillance • Data provided by the University of Pittsburgh Medical Center to improve Public Health • Only aggregate data will be provided to investigators

5. Via site-to-site VPN, • UPMC will send: • patient name, • SSN, • race • date of visit, • time of visit, • address, • date of birth, • gender, • type of insurance, • Chief complaint • ICD9 diagnosis code • disposition • For all hospital visits with a final diagnosis code of 460-520 from Allegheny County hospitals Matched respiratory data UPMC will purge all ED data that do not match a corresponding discharge code of 460-520 Irrelevant data

6. Data Transfer • Site-to-site VPN has been dedicated for data transfer • Firewalls to secure data are in place at ACHD • Firewall exceptions have been added to allow data from hospital to pass continuously to ACHD

7. Data format • Medical record data arrives at ACHD in a flat file in real time when the final diagnosis is applied to an asthma Emergency Department visit (usually less than 24 hours) • Data is then stored on a secure Oracle server that was provided by Pitt APEX • Data is accessible to ACHD epidemiologists through Oracle Discoverer

9. VPN Connection shared with UPMC If no acknowledgement, the connection is broken If Medipac data is present for a visit with acute reactive airway disease is there, it is transferred ACHD acknowledges the data receipt yes ACHD dumps data to an Oracle database The data can then be analyzed using the password-protected web-based interface of Oracle Discoverer at ACHD Data is de-identified and made available to APEX researchers

10. Interruption in service

11. De-identification of UPMC Data File • Creation of unique Identifier and removal of UPMC and patient identifiers • Create Cpi-T variable [integer; 14 digits] • Create VN-T variable [integer; 10 digits] • Let Cpi-T = formula(Cpi) • Let VN-T = formula(Visit number)

12. De-identification of UPMC Data File • Calculation of Age variable and removal of Date of Birth • Create “Admit Day” variable [CCYYMMDD] • Create “Age” variable [numeric, 3 digits] • Create Age Groups • Let “Admit Day” = truncated “Visit Admit Date”; keep first 8 digits only • Let “Admit Day” = truncated “Visit Admit Date”; keep first 8 digits only

13. De-identification of UPMC Data File • Assigning of Census Tract and removal of patient address information • Create “Census Tract” variable • Assign census tract using program (80% hit rate) • Manually assign remaining Census Tracts

14. De-identification of UPMC Data File • Delete Identifying Information • Delete Account Number • Delete Cpi • Delete Visit number • Delete Patient Name Last • Delete Patient Name First • Delete Patient Name Middle • Delete “Date of Birth” field • Delete “Age” field • Delete “Patient Address- Line 1” field • Delete “Patient Address- Line 2” field

15. Preliminary data 2009 – provided by ACHD • 7921 messages transmitted Feb 1-12, 2009 • 548 had diagnosis of 493.x (6.9%) • 5228 of these were final (abstracted) diagnoses • 389 had diagnosis of 493.x (7.4%) • These represent 2623 individuals • Age range is 0-99 • Mean age is 46.93

17. Expected achievements • Innovative, cost-effective surveillance strategy • Near real-time surveillance • Manageable amount of information (~50 asthma cases/day) • Based upon a diagnosis, not free-text • Mechanism to collect other information of public health importance

18. Future Directions • This secure connection can be utilized to speed up reporting of notifiable diseases to benefit the health department • It can be utilized to transmit information on lab tests for metals and other biological markers for biomonitoring to benefit environmental public health tracking

19. Im W. Schneider D. Effect of Weed Pollen on Children’s Hospital Admissions for Asthma During the Fall Season. Archives of Environmental and Occupational Health. 60(5), September-October 2005, 257-65.

20. Im W. Schneider D. Effect of Weed Pollen on Children’s Hospital Admissions for Asthma During the Fall Season. Archives of Environmental and Occupational Health. 60(5), September-October 2005, 257-65.

22. Case-Crossover Analysis of Air Pollution and Cardiorespiratory Hospitalizations: Using Routinely Collected Health and Environmental Data for Tracking • Significant associations were observed between the fourth quartile in PM10 and cardiorespiratory hospitalizations (odds ratio [OR]: 1.12; 95% CI: 1.02–1.23) and cardiovascular hospitalizations only (ICD-9: 390–459) (OR: 1.13; 95% CI: 1.01–1.26) before the plant closure. • After closure of the plant, PM10 was not significantly associated with cardiorespiratory or cardiovascular disease hospitalizations. Moreover, the referent sampling approaches did not greatly alter the estimations in the case-crossover analysis. • The findings suggest that closure of the steel coke plant was associated with a reduction risk of the cardiovascular hospitalizations.