1. Biosurveillance/BioSense Evaluation Project Overview of State Assessments August 28, 2008 Peter L. Elkin, MD, FACP, FACMI Anna Orlova, PhD Walter G. Suarez, MD, MPH Brett Trusko, PhD Katie Skeen-Morris, MPH Lorna Will, RN, MPH Prya Rajamani, MPH Jennifer Ellsworth, MPH David Froehling, MD Dietlind Wahner-Roedler, MD Zelalem Temesgen, MD Martin LaVenture, PhD, MPH Larry Hanrahan, PhD Roland Gramache, MD

3. Background Current public health surveillance and investigation often involves manual reporting of cases to public health agencies and phone calls to healthcare providers for more detailed information The timeliness, completeness, and breadth of coverage of these manual processes can be problematic, especially during a public health emergency With increasing amounts of healthcare and health-related data in electronic form, the progressive adoption and use of electronic health records and the move towards interoperable health information exchanges (HIEs) at the local, regional, state and national level, there are now significant opportunities to leverage health IT and HIE to better support public health functions including preparedness and disease surveillance.

4. The BioSense Mission  The purpose of the BioSense program is to improve the nation's capabilities for real-time biosurveillance and situational awareness Provide immediate, constant, and comparable information needed to inform local, state, and national public health, to support preparedness efforts; and Assess health data about current patients as needed from hospitals nationwide to identify disease clusters (i.e. outbreak) or unusual health indicators (i.e. bioterrorism attack). 4

5. The BioSense Vision  Provide local, state, and nationwide health situational awareness For suspect illness Cases of disease Before, during, and after a health event  Help confirm or refute the existence of an event  Monitor disease outbreak Size Location Rate of spread 5

6. The BioSense Approach  Real-time delivery of healthcare data to BioSense from: Hospitals Laboratories Ambulatory care settings Other health data sources  Electronic “views,” analytics, and reports for: National public health, CDC State and local public health Contributing healthcare organizations 6

7. Defining the Intent  Early Event Detection: Prediction of an outbreak or other public health emergency at as early a stage as possible through review of supporting clinical data. (Does NOT imply only initial event detection). May assist in detecting the very first upswing of the epidemic curve.  Health Situational Awareness Literally, the ability to know what’s going on; accomplished by monitoring extent of disease or disease indicators over time and geographically, especially in an emergency context. Emphasis is placed on monitoring after the initial upswing of the epidemic curve. 7

8. Clinical Care Data of Interest: ED, Outpatient, Inpatient Care Settings Foundational: demographics (minus obvious identifiers), chief complaint, discharge diagnoses, disposition, hospital utilization Clinical (ED, Ambulatory, Inpatient): vitals, triage notes, working diagnosis, clinical records, discharge summary Laboratory: orders, microbiology results Pharmacy: medication orders Radiology: orders, interpretation results 8

9. Implementation Targets and Status: National Healthcare Data Sources : Data SourceRationale Orders & results from 3 major commercial clinical laboratories Represent 20% of all US lab testing; 60% of independent testing; critical to many PH efforts Real-time data from VA150 hospitals and ~1000 ambulatory care clinics; share data with many state and local PH communities Real-time data from DoD45 US hospitals and ~800 ambulatory; share data Poison Control Centers call data All 62 poison control centers; display and compare with other community health data 9

10. BioSense Hospitals and Target Cities  Atlanta Austin  Baltimore  Boston Buffalo / Rochester Charlotte  Chicago Cincinnati Cleveland  Columbus Dallas – Ft. Worth  Denver Detroit  El Paso Houston Indianapolis Jacksonville Kansas City  Las Vegas Los Angeles Louisville Memphis Miami  Milwaukee Minneapolis New Orleans New York Newark Norfolk Orlando  Philadelphia Phoenix Pittsburgh Portland Salt Lake City San Antonio  San Diego San Francisco San Jose Seattle St. Louis Tampa  Washington D.C. Last updated 1Oct06 10

11. Biosense National Public Health System that Receives, Analyzes and Visualizes Health Record data –As of June 2008 => –558 acute-care hospitals –826 Veterans Affairs –354 Department of Defense facilities –Transmitting chief complaint or diagnosis –42 that send text reports for chest and skeletal radiographs

12. Long-term Performance Measure For CDC Biosurveillance Capacity 2006 OMB PART Measure By 2010, CDC’s biosurveillance activities (including BioSense) will reduce the time needed from a triggering surveillance event to initiate event-specific standard operating procedures for all infectious, occupational or environmental (whether man-made or naturally occurring) threats of national importance. 12

13. Primary Tensions Early warning system, primary terrorism detection VS System to enable PH access to real-time existing healthcare information, for any situation that arises Developing a single national system, one architecture, cross-jurisdictional info VS Local responsibility, under jurisdiction of States, necessary for local relationship-building 13

14. Suggested Improvements to Biosense  Suggestions for improved system functionality Expand and enhance geographic mapping functions Track visits over time (for same individual) Enable more ways to share, print, export, download data Enable views of aggregated national data Use BioSense data as overlay with existing data, such as air quality index, sewer maps Whole Record Surveillance Use of Standardized Ontologies to code and surveill clinical data 14

15. Biosense Evaluation  Evaluation is critical Systematic evaluation of use for emergency detection and response critical Develop systematic approach to evaluate inclusion of new data streams (value, data use, etc) Pilot new data types before broad use Need specific plan to ensure and enhance data quality 15

16. Cooperative Agreement for BioSense Program Evaluation CDC awarded cooperative agreement for key evaluation aspects including: Baseline data – to determine PH impact Data accuracy and validity System usability and utility Cost effectiveness 16

17. CDC’s Health Protection Goals Healthy People in Every Stage of Life: All people, and especially those at greater risk of health disparities, will achieve their optimal lifespan with the best possible quality of health in every stage of life. Healthy People in Healthy Places: The places where people live, work, learn, and play will protect and promote their health and safety, especially those at greater risk of health disparities. People Prepared for Emerging Health Threats: People in all communities will be protected from infectious, occupational environmental, and terrorist threats. Healthy People in a Healthy World: People around the world will live safer, healthier, and longer lives through health promotion, health protection, and health diplomacy. 17

18. BioSense Aligned with Four of CDC’s Nine Preparedness Goals Pre –EventEventPost-Event Prevent 1.Increase the use and development of interventions known to prevent human illness from chemical, biological, radiological agents, and naturally occurring health threats. Detect 2.Decrease the time needed to classify health events as terrorism or naturally occurring in partnership with other agencies. 3.Decrease the time needed to detect and report chemical, biological, radiological agents in tissue, food or environmental samples that cause threats to the public ’ s health. 4.Improve the timeliness and accuracy of communications regarding threats to the public ’ s health. Investigate 5.Decrease the time to identify causes, risk factors, and appropriate interventions for those affected by threats to the public ’ s health. Recover 7.Decrease the time needed to restore health services and environmental safety to pre-event levels. 8.Improve the long- term follow-up provided to those affected by threats to the public ’ s health. Control 6.Decrease the time needed to provide countermeasures and health guidance to those affected by threats to the public ’ s health. Improve 9.Decrease the time needed to implement recommendations from after-action reports following threats to the public ’ s health. 18

19. Awardees should evaluate BioSense with respect to one or more of the following areas:  Description of BioSense system & stakeholders  Appropriateness of project goals  Appropriateness of data sources & data elements  Rationale (proof-of- concept) for system utility in emergency situations. Assess needs of emergency responders, inventories of other data sources, and mathematical modeling Overall Approach Data Accuracy Data Use System Utility  At the data source, the accuracy of electronic records when compared with an independent source & the accuracy of standard codes  Similarity of received electronic data to data that is sent from the data source & the accuracy of data preprocessing  Aggregate data from source & CDC  Using sample and/or simulated data, confirm data accepted & stored by system is the same as that being transmitted by the source  Appropriateness of methods and protocols as well as whether the methods are accurately applied  Specific data uses to be evaluated: data analysis, data display, monitoring, reporting, & actions taken  For day-to-day use, the costs & benefits to identify communicable or reportable diseases  For early event detection, the sensitivity & predictive values for finding disease clusters (including ILIs), cost to users of evaluating data anomalies, & benefit from having identified data anomalies  For health situational awareness, cost & benefit to assist in investigation & mgmt of a disease outbreak  Certain characteristics (e.g. timeliness, flexibility, acceptability, stability) are relevant for all uses BioSense Evaluation Areas 19

20. Proposed Methodology System utility rationale (proof-of-concept) –Catalog all methods and protocols along with the use cases that led to their generation; perform analysis of workflow Clinical accuracy review and descriptive statistics –Two reviewers will review 1,000 randomly selected clinical records associated with BioSense submissions (from three sites, one being Johns Hopkins University) Cost-benefit analysis –Provide predictions of potential cost savings when shifting the response to an earlier time when using BioSense Mayo Clinic-Mount Sinai School of Medicine Proposal 20

21. What are we really attempting to do? –Opportunity to move forward the bi-directional inter- relationship between Electronic Health Records and Public Health Information Systems –Help give shape to the future of the integration between public health and clinical practice to improve the quality of health care Project Objectives

22. Example Surveys & Findings Public Health Authorities Healthcare Organizations

24. Known problems with case/event detection and management 52% Did not identify any problems Problems reported: –Need higher resolution than CC, triage notes, nurses notes, diagnoses data –Need better data resolution/ quality of data –Understaffed IT department –Siloed databases –Queries do not meet daily needs –Incorrect parsing –Poor connections, system goes down regularly –LOINC codes and local codes are used –Receive late reports w/ incorrect intervetions implemented –Difficult system that does not meet daily needs –No support for current program –Current database is over capacity, just need a new system

25. Data Gaps in System 57% did not identify any data gaps Data gaps reported: –Need data from Urgent Care Clinics and Primary Care Physicians –E-lab needs better patient demographics –Demographic data gaps –In long term care, rates for death from flu not reported –Data needs to be reported in a better format –Current forms may not capture all risk factors

26. Gaps in Current System 52% did not report any gaps in current system Gaps reported: –No archive-ability –Graph analysis for longitudinal data analysis –E-mail alerts –Integrating lab data –Integrating diagnosis data –School absenteeism integration –School clinic diagnosis data –Need better syndroming and trending for communicable disease surveillance –More linkages with epi system EHR-S, GIS, RHIO –Data that is not real-time is not useful, data needs to be real time and a flag should produce as soon as the info hits the EMR. Differential Diagnosis would also be helpful. –Need to integrate with other data sources so there is one system, one password –Need to be able to add new fields with drop-down lists of answers for data entry –Automated analysis of incoming data with alarm that is pushed to me –Need to integrate laboratory results file with clinical case files –Ability to link patients w/ same conditions across jurisdictions and investigations –System integration for surveillance and case mgmt. data, threshold alerts, geocoding and jurisdiction names

27. Next Steps Accuracy of the Biosense Data –Data Sharing Agreements with JHU in place Is the data submitted accurate at JHU and at the CDC? –Only 47 of 3000 ICD9-CM Codes matched the data from the chief complaints of patient data submitted on the same day by the same patient to the CDC. Could the data be better using SNOMED CT based whole record analysis? –Discussions are ongoing toward formulating a data sharing agreement with Aurora Healthcare System PHIRMS

28. Public Health Information Resource Management Solution (PHIRMS) HK00014 (U38) Collaborators And our CDC collaborators

29. PHIRMS Goal Solution to capture the data needed to: –Evaluate BioSense Public Health Survey –Evaluate any Public Health Intervention in terms of Cost-Benefit Ratios –Benchmark current and future Public Health practice

30. PHIRMS Conceptual Framework

31. PHIRMS Data PHIRMS data consists of: –Healthcare Data: History Physical Exam (PE) Lab (Results, Orders) Procedure Notes Radiology (Results, Orders) Nursing notes Impressions –Demographic Data Symptoms, signs, findings, diagnoses and procedures.

32. PHIRMS Data (continued) PHIRMS data consists of: Resource Data Costs Personnel Available Utilized Resources Available Utilized –By Activities

33. LBI Codified Health Record The Mayo Clinic’s Laboratory: Biomedical Informatics (LBI) stores patient Healthcare and Demographic data in a codified Health Record to which other PH data could be added.Healthcare and Demographic –SNOMED CT –LOINC –RxNorm –ICD9-CM –CPT –Others

38. Biomedical Informatics Research Collaborative: LBI Codified Resource Record The Laboratory of Biomedical Informatics (LBI) proposes to add resource related data to codified health record data

39. Public Health Workflow (Handling of Epidemics: Case/Event Detection, Monitoring and Response Management) Case/Event Detection –Anonymization of Records for Data Mining –Epidemics Clusters of Cases –Cases Suspected Confirmed –Mandatory Public Health Reporting –Ad Hoc Case Reports

40. Conceptual Organization of Knowledge Case Case Cluster Cluster Epidemic

41. Public Health Workflow (Continued) Case/Event Management –Case/Event Monitoring –Case/Event Response Resources Management –Resource Identification –Resource Allocation and Utilization

42. Local PH Data Flow Local PH communicates with HCO, State PH and CDC Local PH HCO & RHIO State PH CDC

43. Public Health Agencies Data Flow State Public Health communicates with Local Public Health and CDC State PH Local PHCDC

44. PHIRMS Vision Simultaneous information sharing to improve workflow. HCO State CDC Local

45. Work Flow – Data Flow HCO Data flowing from the Healthcare Organization (HCO) and RHIOs HCO & RHIO CDC State PH Local PH Biosense PHIN (NEDSS) Public Health Grid

46. Utilities – Data Aggregate Data could be Aggregated by: –HCO –Zip code –Local PH –State PH –National –Diagnosis Suspected Confirmed –Symptoms/Signs

47. Utilities – Data Analysis Analysis of the data would provide information regarding: –Trends –Time Series data –Survey Data Satisfaction Other –Costs (Resource Utilization by Case-Cluster- Epidemic) –Benefits’ Analysis

48. Data Communication Data could be made available in various formats: –By Aggregation / Stratification –Graphical Representations Use of Active Graphics –Push and Pull Subscribe and Publish Web / e-mail / Phone Calls

49. Laboratory Ambulatory Care Pharmacy School eQuality System run by HCO / RHIOs PHIRMS: eQuality Monitoring Solution DHHS 4 – Prescribe Medication and Treatment Plan 1 – Conduct Routine Check-ups 7 – Report Data to Schools 2 – Order cholesterol test Media 10 – Conduct Health Education 6 – Fill Prescription 9 - Monitor ER visits, hospitalizations data from EMRs & utilization data Hospital State Public Health Surveillance System Payor 11 – Send reports 3 – Report test results 8– Coordinate Care PUBLICPUBLIC 5 – Monitor Treatment EHRS 12– Conduct Surveys (BRFSS) Encoded Data Internal Quality Reporting Aggregate Anonymized Reports Encoded Biosurveillance & Quality Guidelines Decision Support PHIRMS Expert System Resource Management

50. Healthcare Organization EHR SNOMED CT Encoded EHR Data Health Information Exchange (HIE) EHR SNOMED CT Encoded EHR Data And / Or CDC State Public Health Authority Local Public Health Authority Expert System Shell Biosurveillance & Adhoc Rule Sets Healthcare Organization

51. EHR Database Trigger sends HL7 v2.5 message to MVCS coding service "DOC _ I D SEC_I D PROP _ I D CONCEPT_CO DE INEX_T Y P E PTS" "25212368500071" "252125530200971" "25217961900971" "25225305900171" "252223767900471" "25221109200171" "25234943600471" "251134336400171" "2511425870700071" "2511425526000171" "2511410252200971" "2511522663000971" "251155144000271" "2511518233400371" "251154336400171" "2564322979900171" "2564325868400471" Codified Data for the Secondary use of EHR data

52. BioSense Evaluation Cooperative Agreement (U38) We need to work together toward a common set of Public Health Goals We can leverage an intelligent electronic environment with a robust set of tools to mine the knowledge necessary to accomplish these goals Success is a community reward and a community responsibility PHIRMSPlease help us to design PHIRMS