Purpose MRSA Burden Over 126,000 persons are infected by MRSA in hospitals annually

1. Brad Doebbeling, Abel Kho, Heather Hagg, Jamie Workman, Mindy Flanagan, Kim McCoy, Shawn Hoke, Paul Dexter IU Center for Health Services & Outcomes Research, Regenstrief Institute VA Center of Excellence in Implementing Evidence-based Practice IU School of Medicine, Indiana University-Purdue University-Indianapolis (IUPUI) & Northwestern University Supported by a contract from AHRQ through ACTION.

2. Purpose • MRSA Burden • Over 126,000 persons are infected by MRSA in hospitals annually • ~ 4 MRSA infections per 1,000 hospital discharges • Over 5,000 die as a result of these infections • Over $2.5 billion excess healthcare costs • On average, for each MRSA patient this means: • 9.1 days excess LOS • Over $30,000 in excess cost per case (range $30,000-60,000) • 4% in excess in-hospital mortality • 1/3 patients acquiring MRSA will become infected.

3. Consistent Use of Known Practices Work • Target Modes of MRSA Transmission • Person-person via hands of health care providers • Personal equipment (e.g., stethoscopes, PDAs) and clothing • Environmental contamination • Healthcare environment • Home/Community environment

4. RN awareness of MRB status increased from 24% at baseline to 59% at 1 year. -93% at 1 year after notifying nurses. • Implementation of isolation precautions increased from 15% at baseline to 51% after 1st intervention and then to 90%. • RI electronic tool notifies staff of MRSA positive history at Wishard, based on micro data from all Indy hospitals (except VA). • 286 unique patients generated 587 admissions (4,335 inpatient days) where receiving hospital unaware of the prior history of MRSA. • An additional 10% of MRSA admissions received by project hospitalsover one year and over 3,600 inpatient days without contactisolation. Cac et al Arch Intern Med. 2007;167(19):2086-0 Kho et al J Am Med Inform Assoc 2008; 15:212-216

5. Implementation • “Testing Techniques to Radically Reduce Antibiotic Resistant Bacteria (MRSA)” • AHRQ funded Indiana ACTION Team effort over 18 months through the ACTION collaborative funding mechanism • Our interventions are based on the Pittsburgh model as specified by AHRQ: • conduct active surveillance of all incoming pts. in ICUs • improve rates of contact isolation • Improve hand hygiene rates

6. Interdisciplinary Research & Ops Teams • Clinicians, Health Services Researchers, Engineering/Technology Faculty, Purdue Communication faculty/students, Organizational Psychologists, Informaticists • Partnership with selected Hospital Clinical Staff • Integrated Lean/Positive Deviance Approach: • Identification of solutions from within, bottom up • Leadership support and buy-in • Standardization where evidence exists or to simplify • Customization to meet local redesign needs

7. What is Positive Deviance? • Technique to engage front line staff in owning & improving processes and sustaining change • Based on identification of practices of used by ‘positively deviant’ staff/departments • Critical for staff involvement/buy-in

8. Integrated Lean/PD approach Define the Problem Baseline Current Processes Discovery Identify Operational Barriers Develop Future State Process Action Process Control Strategy

9. Two ICU units in 3 original hospital systems • St. Francis (two ICUs in South Hospital) • Clarian (Methodist and University Hospital) • Community (Community East and Heart Hospital) • Early success encouraged 3 remaining systems to join the project • Wishard (two ICUs) • VA Medical Center (housewide) • St. Vincent's (two ICUs in north facility)

10. Our health care engineers partner with and train front-line workers to use lean-six sigma and positive deviance approaches • Focus on coaching front-line staff teams to lead instituting systems changes to systematize processes and sustain practices. • Emphasize regular measurement and feedback of adherence to enhance adoption. • Weekly Meeting of all hospital teams to identify barriers & facilitators, review and reinforce progress, share best practices, strategize about spread and solutions.

11. Preliminary pre and post intervention results for first three hospitals suggest average of 60% reduction on study units • ~ 20% reduction hospital wide • Currently investigating optimal biostatistical approach such as time series analysis to confirm

12. An Operational Citywide Electronic Infection Control Network: Results from 1st Year • Infection control is a regional problem, requiring a coordinated effort • Created a citywide electronic notification system to prospectively track all known patients with MRSA • Currently track 17,000 patients with a history of MRSA infection or colonization across Indianapolis. • Since May 2007, delivered 2698 admission alerts on patients with a history of MRSA, 19 percent based on data from another institution. • 20 infection control providers (ICPs) spanning 16 hospital Kho, Lemmon, Dexter, Doebbeling AMIA 2008

13. System redesign approach of training, consultation and coaching front-line staff seems to be strong, sustained approach • Importance of buy-in from highest institutional levels crucial • Enthusiasm builds from within because redesign teams own it! • Informatics tool helpful in identifying great cross-over of MRSA patients in hospitals

14. Our proposed data collection too intensive for most community hospitals • Need a better electronic data collection infrastructure relating to outcome data • Hospitals desire regular feedback on impact of interventions • Little time for paper writing (Hazard of short time lines for funding)

15. Engagement Process: The engagement process involves multiple social change processes centered on PD. • Outcome Data Feedback: Transmission and infection rates are shared with staff. • Process Data Feedback: Nursing unit staff track performance on infection control practices.

16. Learning Collaborative: Teams from hospitals connect with teams from other hospitals employing the MRSA Intervention Bundle to foster learning and innovation. • More extensive activities to train interdisciplinary teams within each of the participating health systems (compared with first study).

17. Acting as assessment, technical, and resource center for AHRQ and 5 other AHRQ ACTION Partners: • Denver Health and Parkland Health, Texas • HRET (Michigan Hospital Association Keystone Center ) • Iowa • American Institutes for Research (AIR) • Yale • Indy Investigators—IU Center for Health Services & Outcomes Research, Regenstrief Institute: • Jamie Workman-German; Mindy Flanagan; Amber Welsh; Shawn Hoke; Brad Doebbeling

18. Goal of overall initiative to identify factors associated with implementation of training that can assist hospitals in successfully reducing and sustaining the reduction of infections associated with the process of care. • Training tools aimed at VAP, BSI, and SSI

19. Facilitate and coordinate consistent collection of information across the HAI Initiative, • Provide technical assistance to the 34 hospitals, • Analyze and synthesize the information from the common information collection instruments, • Share lessons learned about the barriers, solutions and PD practices in implementing and sustaining infection safety.

20. Interdisciplinary group of collaborating investigators affiliated with VA sites distributed across the US. Improve the health of veterans through foundational and applied informatics research. Advance the effective use of unstructured text and other types of clinical data in the electronic health record. Indy Team--Merchant, French, Saleem, Burton, Friedlin, Flanagan, Allen, Doebbeling Salt Lake City VA, Indy VAMC, West Haven VA, Boston VA, Portland VA, Philadelphia VA, Nashville VA, Tampa VA, Regenstrief Inst., NLM, NIH, Mayo Clinic, Carnegie Mellon U., Vanderbilt U., U. of Pittsburgh, Oregon HS U., U. of Utah.

21. Concept extraction from clinical notes • Retrieve clinical concepts and their modifiers • Diagnoses, negation, temporality, ambiguity, etc. • Develop knowledge from concepts • Develop new or more knowledge about a patient • Timelines, symptomology, patient models • Apply knowledge • Use new knowledge to enhance patient care • PTSD • Predict suicidailty • Find undiagnosed patients • MRSA • Automate warnings to ICP & wards • Determine treatment efficacy

22. De-identification of sensitive data • Concept Extraction • Clinically-relevant inference and modeling • Document quality assessment • Evaluation and annotation methods

23. Systems Architecture Raw VistA text Annotated corpora Methods Terminologies NLP Systems Machine Learning Lexical analysis Lexicon Tokenization Structure analy. Classification Grammars POS tagging Pattern recog. Clustering Structured Data Syntactic analysis Vocabularies Association Analysis Parsing Chunking Ontologies Outlier Detection Semantic analysis Semantic Networks Semantic tagging WSD Code mapping Semantic analyzer De-identification systems Integrated databases New structured data

24. Twomajor applied projects to demonstrate the value of its research activities. --high priority for veteran’s health. • Methicillin-resistant Staphylococcus aureus (MRSA) infection (Indianapolis VAMC); • Post-traumatic stress disorder (PTSD). New Pilot projects

26. Develop, review and refine an ontology for clinically and epidemiologically relevant concepts to enable detection of MRSA • Index MRSA-related Concepts in Clinical Narrative.

27. Clinical Inference and Analysis of MRSA-Related Information Contained in the Medical Record • Develop and evaluate a prototype surveillance application that uses automatically processed VA electronic health record data

28. Evaluate algorithms for making multi-type predictions based on heterogeneous data, using MRSA as a clinical domain.

29. Funding AHRQ ACTION Network, Hospital Acquired Infections Collaborative • VA HSR&D and Consortium for Healthcare Informatics Research • Indianapolis Collaborators • AHRQ and VA Collaborators