What to Expect when On-Boarding to ILHIE Master Patient Index (MPI) Presentation By : Alexander Danel
What Constitutes Demographics? MPI contains patient demographics. Patient information related to identification, contact, and environment. Examples: name, address, date-of-birth. Not clinical, yet can be subject to confidentiality concerns. Page 2
How is Demographics Managed? MPI is repository of patient demographics for entire HIE. The brand name HSPI is sometimes applied to the Illinois MPI. MPI = Repository + methods for maintenance and access Page 3
The MPI Responds to Queries Requestor and HIE must agree which patient is subject of query. Requestor sends (usually partial) demographics. MPI finds matching patients. Send back a list of candidates. Page 4
Search is Interactive Search might return a list of candidates. Requestor chooses which (if any) is correct. Even if just one candidate, requestor still needs to confirm. Fully populated demographic records help requestor make the correct decision. Page 5
ILHIE Assembles the Patient's Clinical Data Once the patient is identified, ILHIE assembles the consolidated record. ILHIE doesn t store clinical data; fetches it on-demand. Patient might have records at several facilities that are attached to ILHIE. ILHIE assembles a single, consolidated record from the several sources. Page 6
RLS Directs the Fetch Multiple partners might have information about the patient. Record locator service (RLS) knows which partners are familiar with the patient. RLS identifies which partners get queried by ILHIE. Page 7
Establishing and Maintaining the MPI Partner will establish the initial state of MPI using a bulk-load. After that, partner will maintain using message based transactions. Page 8
MPI Bulk-Load is Part of On-Boarding The bulk-load (a.k.a. back-load ) into production environment is done exactly once. Partner creates a text file that follows ILHIE specification. All participating patients from partner EHR must be present. Page 9
Bulk Load is a "Cut-Over" Critical timing a sequence of related actions must occur. The bulk-load brings MPI current. Messages must immediately start flowing to keep it current. Page 10
Parallel Development Efforts Bulk-load development. PIX messaging development. Both must be working reliably, then schedule cut-over. Page 11
Development Cycle for the Load Iterative cycle: develop-and-test, develop-and-test, etc. Goal is a reliable process. Actual load file won t get created until last minute, in cut-over. Need well defined, repeatable, scripted process. Page 12
Tasks in Developing the Load Process ETL extract, transform, load. Partner does business analysis what to extract and how to transform. Partner does extract, need SQL programmer. Partner does transform, need text programmer. ILHIE does the load. Page 13
Mapping EHR Fields into the Load-File Spec Mapping is a business analysis that matches source fields to target fields. Find the EHR table that has demographics. Simplest case: One-to-one match between source field and target field. Simplest case: Identical meanings, just re-order and re-name. Example: Field labeled family_name in EHR is labeled last_name in ILHIE. Page 14
More Complicated Field Mappings More complicated mappings require transformation and inference. Example: Parsed data might go to two fields. Example: Implied data might have no source except context. Business analyst must identify. Programmers must implement. SQL programmer and text programmer to agree about nulls, etc. Page 15
Known Complications to Field Mappings The tasks that need to get done might include: De-normalize Combine Convert Format Parse Infer Skip Re-code Compute Page 16
Filtering Records Sometimes the process should filter out records: Bad, very incomplete records. Non-participating patients. Other. Page 17
Incremental Test Files Optional: First attempt at bulk-load files can have: Just a few records Fictional patients Subsequent test files must be the real thing: All applicable records in the EMR OK to use e-mail for fictional patient records, but must use secure method for real patients. Page 18
The Result is a Process Testing proves a process. Process must be simple and reliable. Don t want trouble at cut-over time. Push-button, fool-proof scripts. Page 19
Continuous Updates after Cut-Over Admit, discharge, transfer (ADT,) or update at the partner EHR must trigger action. The EHR communicates with ILHIE about the event. Communication is via messages. Page 20
Message Standard for Synchronization Message protocol is IHE. Use PIX add-revise. When connection is via VPN, use PIX version 2. When connection is via web-services (SSL/TLS,) use PIX version 3. Page 21
Action Sequence for MPI Synchronization Partner actions upon ADT/update event. EHR trigger begins process Populate PIX revise Open communications channel Send PIX Receive acknowledgement Page 22
OK to Synchronize MPI with Daily Batch Basic ILHIE model is immediate update. Some partners can only do daily batch. ILHIE has no special technique for batch updates. Partner must convert batch into individual PIX revise transactions. Page 23
Parts of MPI Development Effort have a Dependency Connectivity is a pre-requisite for messages. Need secure network tunnel; either VPN or web-services (SSL/TLS.) Typically don t start PIX effort until connected, but could do some stuff, like populate. Page 24
An OID is Required Partner must identify self with an OID. ILHIE asks that partner acquire and use an HL7 OID. Have documents that show how to do it; quick, easy, and no cost. Page 25
OID and MRN Combination is Unique Partner medical record numbers (MRN) are unique only within partner s assigning domain. The OID is used to identify partner assigning authority domain. Combining the OID with the MRN restores the uniqueness. Assumes partner is maintaining internal uniqueness. Page 26
One Patient, Many Sources, Linked Within MPI, for each data source that treated a patient, there is a demographics record. Example: three hospitals treated a patient, then three demographics records in MPI. Want to link together records that are about the same patient. Have an automated link process. Will need manual intervention for some records. Decision making process is called link resolution. Page 27
Fields Used for Automated Linking Automated linking uses six fields Names Gender Addresses SSN Birth-date Telecoms Repetition fields telecoms and addresses compare all repetitions in one record with all repetitions in the other record. SSN is important do you have it available? Page 28
Manual Link Resolution Automated link-resolution assigns a score and compares score against a threshold. Record pairs slightly under threshold might none-the-less be for same patient. Human intervention might help; humans can: Apply common sense rules. Look at other demographics fields that were not used for automated scoring. Use external information, like partner DB, or even make phone calls. Page 29
Preventing False Positives ILHIE independently confirmed a large group of links. Found five rules for preventing false positives. Gender If genders are different, then put in review worklist DOB If three out of three (day/mon/year) are different, then put in review worklist NewBorn If DOB within 2 weeks from current date, then put in review worklist Twins If DOB and LN match and SSNs differ, then put in review worklist Jr/Sr If one has a suffix and DOB is different, then put in review worklist Page 30
MPI Calibration ILHIE's vendor offers a service called "MPI Calibration" which might be applied. Page 31
ILHIE s System Environments ILHIE maintains three environments: Test, UAT, and Prod. Partner does not need to imitate this. Partner will point interface engine at appropriate end-point to match work being done. Test UAT Prod Prelim testing, no patient info stored here. User Acceptance Testing. MPI goes here during link resolution. Final promotion. Page 32
Summary MPI stores, maintains, and makes use of patient demographics. Bulk-load initially, then continuous updates. Bulk-load uses file, continuous updates use messages. Cut-over is simultaneous bulk-load and message start-up. Need business analyst, SQL programmer, and text programmer. Messages follow the IHE standard, using the PIX profile. Partner EHR triggers message action upon local events. Message have prerequisite of connection. An OID is required, please get one from HL7 per ILHIE directions. Partner and ILHIE should agree on link resolution goal before starting. ILHIE has three environments; Test, UAT, and Prod end-points. Page 33
Appendix A -- More Information about the Bulk Load File Appendix A Contents: Fields in MPI Bulk Load File Format Field Delimiters Sample Load File Page 34
Fields in the MPI Bulk-Load File Unique Patient ID at partner (usually Medical Record Number.) For matching of pairs during link resolution, six field groups are used. Remaining fields are useful for human decision makers. Unique Patient ID Assigning Authority Is unique at partner domain; used in all messages. OID that identifies partner within ILHIE Fields used for matching pairs Name SSN Gender Birth-date Addresses Telecoms last, first, middle, prefix, suffix social security number M/F/U yyyy-mm-dd Street,city,state,postal-code,country, county, precinct country_code,area_code,phone_number, email_adrs, type, use Other fields other identifiers marital status religious affil VIP status other names mother s mdn nm ethnic group place of birth race language Page 35
File Format Plain text, one record per line. First line is a header that identifies the file and number of records. Need to figure out a way to get record count into header. Page 36
Field Delimiters Delimiters are just like those used for HL7 version 2.x Type Record Field Sub-field Repetition Escape Delimiter <cr> (pipe) ^ (carrot) ~ (tilde) \ (backslash) Page 37
Sample Load File Sample rows appear below. First line is the header. Next two lines are sample data. 2013-08-26_12:59:59-HFS HFS 2 2013-08-26 Sample file for facility HFS 10001 HFS S3098476234^USA^PPN 012-222-3333 Jones Jason James Mr. Jr. Jones^J.J.^^Mr^^Alias M 1980-09-23 1 Billings Road, Apt 1^Middlesex^Framingham^MA^02487^^US^HP~1 Mailing St^^Boston^MA^02487^^US^WP 1^617^4891111^^M^WP~1^782^2589338^^L^HP~^^^jjj@yahoo.com^^ si ngle Gonzalez caucasian baptist hispanic Spanish 0 10002 HFS 123456788^HFSID^PI 012-111-5555 Johnston Howard Henry M 1999-05-12 123 Main St.^^Malden^MA^02732^^US^WP ^^^test1@gmail.com^^ married caucasian English 0 Page 38
Appendix B: Connectivity A separate document will explain more about connectivity. Page 39