Delivery of Automated Notifiable Condition Reports: A Success Story of the Regenstrief Institute's Notifiable Condition Detector Shaun Grannis, MD MS FAAFP Medical Informatics Research Scientist, Regenstrief Institute Assistant Professor of Family Medicine, Indiana University School of Medicine © 2008 Regenstrief Institute, Inc.
Overview: What We’ll Cover • The Problem: – Underreporting – Incomplete knowledge of disease burden • The Strategy: – Leverage existing flows – Automation – Sustainability • The System: – HIE – NCD • The Results • The Future © 2008 Regenstrief Institute, Inc.
The Problem (Premises) • To successfully manage the public health disease burden in a community, the true public health disease burden of a community must be ascertained • Disease burden is determined in large part using information generated in clinical care processes • Clinical care processes under-report to public health (Thacker) – Reporters overburdened/under-resourced – Reporters lack knowledge, reporting requirements change – Clinical data is scattered across disparate settings in different (non-standard) formats © 2008 Regenstrief Institute, Inc.
The Strategy • Leverage (re-use) existing clinical data flows to augment public health reporting • Minimize the need for human intervention in the reporting process by … • Standardizing data in a sustainable fashion so computers can automatically process © 2008 Regenstrief Institute, Inc.
A Note on Strategy to Consider • Who identifies whether a condition is reportable? © 2008 Regenstrief Institute, Inc.
The System: Health Information Exchange © 2008 Regenstrief Institute, Inc.
Regenstrief/INPC/IHIE Timeline • Regenstrief tasked in 1972 to “stitch” silos of medical information together -> the Regenstrief Medical Record System (RMRS/Wishard Memorial Hospital) • In 1994, began stitching together information across institutions (INPC/Indianapolis) • IN 2002 IHIE formed (services organization) • In 2003, working towards stitching the state together (INPC/Indiana) © 2008 Regenstrief Institute, Inc.
The Indiana Network for Patient Care An operational community wide electronic medical record © 2008 Regenstrief Institute, Inc.
Indianapolis MSA, Indiana 1. 5 million population base 12 th largest city in U. S. A. Indiana Health Information Exchange Home to Indiana’s only medical school Referral center for entire state (7 million) • Regenstrief Institute • State Department of Health • • • © 2008 Regenstrief Institute, Inc.
INPC Project Goal Demonstrate the feasibility and benefit of a community wide electronic medical record system in patient care. © 2008 Regenstrief Institute, Inc.
INPC Stakeholders • • • Indianapolis Med/Surg hospitals Physician offices Community health centers Homeless care network Public school clinics County and state public health departments National laboratories Pharmacies PBMs © 2008 Regenstrief Institute, Inc.
INPC Data sources • 17 hospitals including the 5 major hospital systems (99% of non-office care) and community hospitals • National and regional laboratories • Local imaging centers • All four homeless care systems • Public health departments (county and state) • Approximately 1/3 of ambulatory physicians © 2008 Regenstrief Institute, Inc.
Classes of Data shared by all Institutions • • • All ED and outpatient visits All hospital discharges (dx, procedures) All inpatient laboratory results All outpatient laboratory results Immunizations All discharge summaries/admissions summaries All operative notes All radiology reports All surgical pathology reports Inpatient medications Tumor registry data © 2008 Regenstrief Institute, Inc.
INPC Data Volumes • • 7 million registration “events” 60 million orders 900 million coded results 20 million dictated reports 8. 8 million radiology reports Hundreds of millions prescriptions 750, 000 EKG tracings 45 million radiology images © 2008 Regenstrief Institute, Inc.
Standardizing Clinical Data Hospital Lab Outpatient Clinic Lab Jane Doe’s 123 LMNOP Patient ID: hospital lab: Name: Jane Doe 30936 -9 3/1/04 GC DOB: 01/01/04 30938 -5 3/1/04 Blood. Cx SSN: N/A 33555 -4 3/1/04 555 Johnson Road Address: Urine. Cx 30943 -5 3/1/04 Hep. B City: Indianapolis State: Indiana ZIP: 46202 Patient ID: Doe’s clinic labs: Jane Ellen 6789 XYZ Name: Jane Ellen Doe 30936 -9 DOB: 01/01/04 5/1/04 Gonn SSN: 123 -45 -6789 30938 -5 5/1/04 BCx 33555 -4 Address: UCx Johnson Road 5/1/04 555 City: Indianapolis 30943 -5 7/9/04 Hepatits State: Indiana ZIP: 46202 Gonorrhea: Blood Culture: Urine Culture: Hepatitis B: 30936 -9 30938 -5 33555 -4 30943 -5 Global Patient Index Global ID: 45678 Name: Jane Ellen Doe Lots of Demographics. . MRF 1 ID: OU 81247 MRF 2 ID: 4564356 PH MRF ID: 123 LMNOP Concept MRF 3 ID: 6789 XYZ Dictionary
The Indiana Network for Patient Care Consolidating the Silos St. Vincent MRF Clarian MRF Global Patient Index Hospital Lab Global Patient Index Outpatient Clinic Lab Outpatient Clinic Concept Dictionary Wishard MRF Concept Dictionary Community MRF
Data Reuse Data Management Data Access & Use Hospitals Hospital Payers Health Information Exchange Physicians Labs Negotiated Access. Labs Data Repository • • • Results delivery Secure document transfer Shared EMR Credentialing Eligibility checking • • • Results delivery Secure document transfer Shared EMR CPOE Credentialing Eligibility checking • Results delivery Network Applications Public Health Physician Office • Surveillance • Reportable conditions • Results delivery Payer • Secure document transfer Researchers Outpatient RX • De-identified, longitudinal clinical data Public Health Ambulatory Centers © 2008 Regenstrief Institute, Inc.
A Note on Sources of Data for NCD • Four percent of physicians have an extensive, fully functional electronic-records system † • 13% report having a basic system† • Identify “low-hanging fruit” for data sources † “Electronic Health Records in Ambulatory Care: A National Survey of Physicians”, NEJM 359: 50 -60. July 3, 2008 © 2008 Regenstrief Institute, Inc.
Clinical Results ideally should Optimally Standardized, Coded observations Blood culture result in HL 7 format: MSH||CHARTLINC|SF_SOFTMED|||$Year. Month. Day|| PID|||$patient id$||$patient name$|| PV 1|||||||$attending doctor$||$consulting doctor$|| LOINC OBR||008^001|M 46123|BLC^BLOOD CULTURE^L|| OBX|1|ST|CULT^CULTURE^L^6463 -4^Bacteria identified^LN|BLC 1|METH. RESISTANT S. AUREUS (MRSA)^115329001^MRSA^SN|||A| SNOMED © 2008 Regenstrief Institute, Inc.
Real World Culture Report No LOINC or SNOMED codes MSH|^~&|PMLW|HNA|NCD||20081016114316||ORU^R 01|8|P|2. 3|25754 ZVX|K|PF|126134534| PID|1|384854^^^HNA^MRN|519074^^^HNA^MRN||Eccleston^Christopher||20080220|M||W|4784 Main Street^^Neverland^IN^75027||367 -433 -4206|306 -565 -8064|EN|S|CAT|710863^^^HNA^FIN NBR|274 -387 -3363|DL 4750617|902998^^^HNA^FIN NBR|||||N OBR|1|989923284^HNALAB|17 -494 -2482_459053|692571^Resp CX + Stn^CERLB|||20081119100049|||||||20080328201310|&CF&^^^&Sputum&|904379^Cortez^Ana-Lucia^^^^^^HNA^Personnel^^^HNA DOCTOR||||||20080711035105||MB|P|CD: 215871426&Cystic Fibrosis Birthday Visit Order Set. RH|1^^^20080123094900^^R~^^^^^R||154768586^HNA_ORDERID|||||||20080305023625 OBX|1|FT|666971^GS^CERLB||Less than 10 epithelial cells per lpf . br2+ (moderate) Gram Positive Cocci . br1+ (few) Gram Negative Rods . brRare White Blood Cells||||||F|||20080225164740|^CPL Micro^|536146^Tishler^Adair^^^^^^HNA^Personnel^^^HNA DOCTOR|^^220^CPL Micro OBX|2|CE|ORGANISM^^CERLB|1|310784^Staphylococcus aureus^CERLB|||||20081116062408|^CPL Micro^||^^220^CPL Micro OBX|3|CE|ORGANISM^^CERLB|3|310307^Haemophilus influenzae^CERLB|||||20080106025756|^CPL Micro^||^^220^CPL Micro OBX|4|FT|666972^Final^CERLB||4+ (many) Staphylococcus aureus (MRSA) . br2+ (moderate) Escherichia coli Extended spectrum Beta-lactamase producer confirmed . br1+ (few) Gram Negative Rods #2 Not identified . br4+ (many) Normal flora isolated||||||F|||20080302045155|^CPL Micro^|578549 OBR|2|806483421^HNALAB|16 -9622246_589620|312370^MIC^CERLB|||20080926121606||||&CF&|805953^Byrd^Eugene^^^^^^HNA^Personnel^^^HNA DOCTOR||||||20080615131702||MB||692571&Resp CX + Stn^1^310784&Staaur|1^^^^^R~^^^^^R|||||| OBX|1|CE|ORG^^CERLB|1|310784^Staphylococcus aureus^CERLB|||||20080615213954|^CPL Micro^||^^^CPL Micro OBX|2|NM|309643^Penicillin^CERLB|1|>=0. 5|||R|||F|||20080612075548|^CPL Micro^|209285^Watros^Libby^^^^^^HNA^Personnel^^^HNA DOCTOR|^^^CPL Micro OBX|3|NM|309669^Clindamycin^CERLB|1|<=0. 25|||S|||F|||20080222214017|^CPL Micro^|935308^Roundtree^Richard^^^^^^HNA^Personnel^^^HNA DOCTOR|^^^CPL Micro OBX|4|NM|309665^Erythromycin^CERLB|1|>=8|||R|||F|||20080603024932|^CPL Micro^|946525^Fernandez^Nikki^^^^^^HNA^Personnel^^^HNA DOCTOR|^^^CPL Micro OBX|5|NM|309663^Gentamicin^CERLB|1|<=0. 5|||S|||F|||20080218160702|^CPL Micro^|354065^French^Colby^^^^^^HNA^Personnel^^^HNA DOCTOR|^^^CPL Micro OBX|6|NM|309644^Oxacillin^CERLB|1|>=4|||R|||F|||20080815093826|^CPL Micro^|716621^Rousseau^Danielle^^^^^^HNA^Personnel^^^HNA DOCTOR|^^^CPL Micro OBX|7|NM|309640^Rifampin^CERLB|1|<=0. 5|||S|||F|||20081011194018|^CPL Micro^|296776^Clark^Josh^^^^^^HNA^Personnel^^^HNA DOCTOR|^^^CPL Micro OBX|4|FT|666972^Final^CERLB||4+ (many) Staphylococcus aureus (MRSA) . br2+ (moderate) Escherichia coli Extended spectrum Beta-lactamase producer confirmed . br1+ (few) Gram Negative Rods #2 Not identified . br4+ (many) Normal flora isolated||||||F|||20080302045155|^CPL Micro^|578549 © 2008 Regenstrief Institute, Inc.
High-Level Overview of Automated Notifiable Condition Detection © 2008 Regenstrief Institute, Inc.
Notifiable Condition Detector Realtime Inbound Message Compare to Dwyer I Potentially Reportable Record Count as denominator Abnormal flag, Organism name in Dwyer II, Value above threshold Reportable Condition E-mail Summar y Daily Batch To Public Health Reportable Conditions Databases Print Reports To Infection Control
Step-By-Step Process • Clinical result received • Test code (OBX-3) mapped to LOINC • LOINC code examined for potential reportability – PHIN NCMT – “Dwyer I” (e. g. , Blood culture is potentially reportable; Serum Prolactin Level is not) • If LOINC code is potentially reportable, then look up clinical conditions associated with that code – PHIN-NCMT – “Dwyer II”. • Look for those clinical conditions in the clinical concept field (OBX-5). “Fuzzy” string comparators used • If answer present, look for negation context • If no negation, flag as reportable © 2008 Regenstrief Institute, Inc.
“Fuzzy” String Comparator • Longest Common Subsequence (LCS) • If LCS (Condition, Result Text) > 0. 8, consider it a match • LCS (‘Hepatitis’, ‘Hepaitis’) : Hepatitis contains 9 characters Share subsequences ‘Hepa’ (4) and ‘itis’ (4) ( 4 + 4 ) / ( 9 ) = 88% → Considered a match © 2008 Regenstrief Institute, Inc.
Additional Detail • • • “Report All” sources Abnormal Flag ICD-9 diagnostic codes Key words Challenges © 2008 Regenstrief Institute, Inc.
standardize patient (MPI) Push data to recipients via clinical messaging standardize vocabulary standardize provider Monitor data flows P 1 P 2 P 3 P 4 P 5 P 6 Load data into repository Evaluate for Public Health Notifiable status P 7 Access Layer Inbound data Access Layer Data Receiver Raw Message Queue Normalized Message Queue Receive data by a Raw, non-normalized Normalized data variety of potential data available for multiple means: TCP port, uses directory harvester, /stakeholders/applicatio s. FTP, HTTP, email, etc ns WORKFLOW (from left to right in image): • Dr. Orders Chlamydia Culture • Lab performs test, generates electronic result • Electronic result sent via secure real-time connection to data receiver • Rates of flow from lab are monitored, alerts sent if rate falls outside expected range (P 1) • Data receiver performs minimal validation of inbound message, stores message in raw queue • Raw-queue processors augment/normalize data where necessary, eg: • Augment result with additional provider data present in the HIE (P 2) • convert local Chlamydia code to common vocabulary eg, LOINC, SNOMED, etc. (P 3) • Augment patient data with additional information in the HIE - MPI (P 4) • Message is place in normalized queue where 1 to N consumers can access each message for a specified use case: • results are pushed back to Dr. through clinical messaging (P 5) • results are stored in patient-centric community repository for access at point of care (P 6) • results are analyzed and if Chlamydia test is positive, is automatically sent to public health (P 7) Processor/ Consumer Layer Access Layer Data Layer
© 2008 Regenstrief Institute, Inc.
Results © 2008 Regenstrief Institute, Inc.
Public Health Reportable Conditions © 2008 Regenstrief Institute, Inc.
ELR Completeness† 4, 785 total reportable cases INPC– 4, 625 (97%) Health Dept – 905 (19%) Hospitals – 1, 142 (24%) †Overhage, Grannis, Mc. Donald. A Comparison of the Completeness and Timeliness of Automated ELR and Spontaneous Reporting of Notifiable Conditions. Am J Pub Health 2008 98: 344 -350.
Next Steps • • Evolving NCD / Open Source Critic Architecture Delivering reporting forms to providers Administrative/operational improvements – Monitoring – Denominator Counts – Cross jurisdictional reporting © 2008 Regenstrief Institute, Inc.
Open Source © 2008 Regenstrief Institute, Inc.
Public Health ? © 2008 Regenstrief Institute, Inc.
New “REX Critic” Architecture • Ability to identify more complex result structures • More robust “conditional” negation (using an ontology of negation) • Addresses at least 3 types of results categories: – Numeric • “Varicella Zoster Ig. G titre: 1. 51”, “Lead level: 22”, “Absolute CD 4 Helper: 284” – Discrete / Categorical • “Hep B surface Ag: Present”, “N gonorrhoeae, DNA Probe: Positive” – Freeform Text • e. g. , Blood culture reports © 2008 Regenstrief Institute, Inc.
MRSA Evaluation† Manual Review Positives REX Negatives Positives 39, 491 74 Negatives 17 24, 972 64, 554 total reports Sensitivity, Specificity, PPV > 99% †Friedlin J, Grannis S, Overhage JM. Using Natural Language Processing to Improve Accuracy of Automated Notifiable Disease Reporting. AMIA Proc Sym 2008. In Press. © 2008 Regenstrief Institute, Inc.
Example 1: All results found in OBX-5 OBX|1|TX|CLDWD^CULTURE WOUND, Deep^SERH|| Source: ABSCESS Collected: 09/09/08 00: 55 OBX|2|TX|CLDWD^CULTURE WOUND, Deep^SERH|| Site: ABDOMINAL abcess Received : 09/09/08 00: 55||||||F|||200809090055 OBX|3|TX|CLDWD^CULTURE WOUND, Deep^SERH||CULT Anaerobe ** FINAL * 09/14/08 06: 53||||||F|||200809090055 OBX|4|TX|CLDWD^CULTURE WOUND, Deep^SERH||09/14/08 No anaerobes isolated in 5 days. ||||||F|||200809090055 OBX|5|TX|CLDWD^CULTURE WOUND, Deep^SERH||CULTURE WOUND, Deep ** FINAL * 09/12/08 06: 29||||||F|||200809090055 OBX|6|TX|CLDWD^CULTURE WOUND, Deep^SERH|| # 01 Staphylococcus aureusmrsa||||||F|||200809090055 OBX|7|TX|CLDWD^CULTURE WOUND, Deep^SERH|| Moderate. ||||||F|||200809090055 OBX|8|TX|CLDWD^CULTURE WOUND, Deep^SERH|| Negative for inducible||||||F|||200809090055 OBX|9|TX|CLDWD^CULTURE WOUND, Deep^SERH|| Clindamycin resistance. ||||||F|||200809090055 © 2008 Regenstrief Institute, Inc.
Example 2: Results found in both OBX and NTE segments OBX|1|CE|SDES^SPECIMEN DESCRIPTION^MIDAM|1|BLUDC^BLOOD(C) CENTRL LINE DRAW^SQS||||||F|||200809090810|^| OBX|2|CE|SREQ^SPECIAL REQUESTS^MIDAM|1|OONLY^Aerobic bottle only received unable to determine presence or absence of^SQD||||||F|||200809090810|^| NTE|1|| strict anaerobic organisms. OBX|3|CE|CULT^CULTURE^MIDAM|1|ENT^ ENTEROCOCCUS SPECIES ^SQMO||||||F|||200809090810|^| NTE|1||METHICILLIN RESISTANT STAPH AUREUS NTE|2||PRESUMPTIVE NTE|3||VANCOMYCIN RESISTANT ENTEROCOCCUS NTE|4|| NTE|5||GPC CALLED 9/10/08 1009 AM AT 555 -3244 TO MARY SMITH RN, NTE|6|| NTE|7||MRSA CALLED TO 555 -3244 TO MARY SMITH, RN. 09/14/08 1044 © 2008 Regenstrief Institute, Inc.
Example 3: Incomplete sentences; Punctuation lacking Source: SWAB FROM SYNOVIAL/JOINT FLD Collected: 02/16/07 Site: Left Knee Received : 02/16/07 BACT. CULT- BODY FLUID 02/17/07 Notified MMS, Kathy 02/17/2007 07: 09. 02/18/07 **COPY OF REPORT SENT TO INFECTIONS CONTROL** Called to Carol 02/18/2007 16: 42. Called corrected report to Carol 02/18/2007 16: 48. ORGANISM 01 Staph aureus MANY SUSCEPTIBILITY TESTING TO FOLLOW. ***Please note, previously reported MRSA in error. After further testing, isolate is not MRSA*** © 2008 Regenstrief Institute, Inc.
Example 4: Ambiguous Phrases MRSA^MRSA by PCR Source: NOSE/NPH Collected: 12/31/07 14: 25 Site: NOSE Received : : 12/31/07 14: 32 NOSE Received 12/31/07 14: 32 CALLED 01/01/2008, 00: 27, positive mrsa called toto kathy at 0028, By 01/01/2008, 00: 27, positive mrsa called mary at 0028, By CCBL 7 CCAT 1 MRSA by PCR ** FINAL * 01/02/08 08: 18 12/31/07 POSITIVE 01/02/08 No viable methicillin resistant S. aureus (MRSA) for isolation and/or susceptibility studies. © 2008 Regenstrief Institute, Inc.
Example 5: No Punctuation EYE CULTURE No Collection Times Given No anaerobes seen ORG #5 MRSA Staphylococcus coagulase negative Enterococcus sp GNR ID to follow Bacillus species © 2008 Regenstrief Institute, Inc.
Augmenting MRSA reporting with NLP Average Monthly MRSA reporting rate = 1540 MRSA Reports NLP detecting average of 1, 157 more reports of MRSA/month Average Monthly MRSA reporting rate = 383 Months Is MRSA more prevalent than current estimates predict? © 2008 Regenstrief Institute, Inc.
Results Delivery – An Illustration © 2008 Regenstrief Institute, Inc.
© 2008 Regenstrief Institute, Inc.
© 2008 Regenstrief Institute, Inc.
Administrative / Operational improvements • • Monitoring reporting rates Denominator Counts Cross jurisdictional reporting Public Health Feedback © 2008 Regenstrief Institute, Inc.
Take Home Points • If we seek a better assessment of community health, a more complete picture is needed. Then either: – Address the reasons humans don’t report or – Un-encumber the human and begin to automate the process • If we seek to automate the reporting process, then standardized codes (test, not answer) are needed • If we want to sustainable management of standardized codes, we must find the multiple use cases (value propositions) that will support the ongoing mapping process © 2008 Regenstrief Institute, Inc.
Take Home Points • Notifiable disease surveillance is one such function that can be highly leveraged: Public health Notifiable Condition Detection shares many similarities with clinical decision support processes • In the near- to mid-term, public health information technology will become largely indistinguishable from clinical information technology © 2008 Regenstrief Institute, Inc.
Questions / Discussion Thank You! Shaun Grannis, MD MS FAAFP Medical Informatics Research Scientist, Regenstrief Institute Assistant Professor of Family Medicine, Indiana University School of Medicine © 2008 Regenstrief Institute, Inc.
Скачать презентацию Delivery of Automated Notifiable Condition Reports A Success
33196bf8e49cdc48e0e481b90d79e215.ppt