Galen Healthcare Solutions: Allscripts Consultants Enterprise EHR

Our clients environments are both sophisticated and complex, often times with different vendors in the fold for the different healthcare information systems that are utilized by the organizations. For those clients that are Managed Service Organizations (MSOs) or have different sub-entities, this is even more pronounced. Consider for a moment a scenario where an Integrated Delivery Network (IDN) consists of four physician groups under its umbrella. Some of these physician groups were added via acquisition – and as such were using existing systems such as EHRs or PMs from vendors different than those of the organization they were joining. The following mosaic illustrates such a case:

Given the graphic above, one can appreciate the complexity involved with the following core enterprise organizational functions:

• Interoperability – Most systems do not easily interoperate with one another and thus require interfaces to be developed to facilitate communication between the systems
• Patient Matching – uniquely identifying a patient across the enterprise in a system-agnostic fashion.
• Reporting and Analytics – Each of the systems may have different database technologies at their core, and additionally the structure of the data is sure to be different.  This creates a challenge in reporting metrics to exhibit adherence to meaningful use criterion for instance or to
• Trust – Which patient data should be shared across which systems?

A recent presentation at a NEHIMSS last month illustrated these points above and did a great job of communicating how Partners Healthcare is addressing the Healthcare Information System (HIS) mosaic via their COMPASS project. The COMPASS project is an aggressive initiative which implements a common administrative system and processes to streamline revenue cycle management and help manage costs through a “holistic, patient-centric, workflow-driven approach.”

The efficiency of the mosaic of systems (ala Claude Shannon for those EE nerds out there) is subpar at best. But this is the environment organizations find themselves. The alternative would be to consolidate to utilize one vendor across all systems ala the COMPASS project. However, some vendor systems are better at functions than others and the cost for conversion may be prohibitive or in some instances not feasible. For those organizations seeking out advice or recommendations for healthcare information systems, check out the folks at Software Advice as they offer great resources.

Contact us today if your organization seeks assistance with data conversion or integration of healthcare information systems.

Tags: Data Conversion, IDN, Integration, Interfaces, Interoperability, Meaningful Use, MSO, Multi-Org, Reporting

Tags: Allscripts Consultants, Allscripts Demo, Allscripts Enterprise EHR, Data Conversion, EHR Implementation, Enterprise EHR, Extensibility, Integration, Meaningful Use, ScanMD

Drawing on our past experience and expertise with data conversions, we recently assisted one of our clients with a conversion of administrative ICD9 diagnostic data extracted from their Practice Management system to clinical Medcin-based  problem data within the EHR. The project ultimately saved a tremendous amount of data entry time. Upon completion of the data-conversion, clinicians were then able to review the problem list in “Past Medical History” section of the patient chart within the EHR and categorize by either choosing to make the problem “active” or mark redundant or resolved problems as “Entered in Error”.

As with any data conversion, one must be cautious in terms of negative implications. For instance, administrative data has its limitations, and an example or where the process can go wrong is the highly-publicized case of e-Patient Dave.  Ultimately, problem conversions can be useful, but the data needs to be reviewed, and almost treated as suspect.  The value in the conversion is saving the entry of the problems that are accurate – say 80-90%.  Any that are incorrect, will be reviewed with the patient and can easily be marked EIE.

Statistics:

• 1,007,238 problems were loaded to the EHR for 205,831 patients via the interface engine, taking about 11 hours to process totally.
• PM Extract file statistics:
  • Total matchups of ICD9s to patients: 5,405,874
  • Total Unique ICD9s: 8346
  • ICD9s that only match up with 1 patient:1295
  • ICD9s that match up with 100 or more patients: 2027

Approach and Components:

• Master approved “ICD9” list provided by client
• Extract of ICD9 data from PM system provided by PM vendor
• Automated macro that attempts to match ICD9 to Medcin. Potential matches include the following:
  • 1 to 1
  • One to many (20 or less)
  • One to many (20 plus)
  • One to none
  • Each of the different flavors of matches were marked with an annotation (highlighted via an asterisk) to identify to clinicians the logic that was used in importing the problems:

• Once the translation was finalized, it was loaded into the interface engine and mapping logic loaded problems into the patient chart in the EHR via the API (existing stored procedure).

Known Issues Mitigated:

• Due to incorrect logic, some ICD9s were linked to patient profiles improperly. To mitigate this, a script was run to mark these problems as “entered in error”
• Problems were loaded to the “Past Medical History” section of the patient chart with a status of active. However, given this status, it didn’t facilitate providers to easily change the problem to be an active problem linked to a note.

Lessons Learned:

• Execute a proof-of-concept and as with any technical project, get clinician feedback. The client had a pilot group of 5 clinicians to vet out issues and bless the data before the live conversion was run.
• Do NOT use spreadsheets to track the cross-walk between administrative ICD9 diagnoses and clinical Medcin problems. Rather utilize a staging DB to serve as a single repository in developing ICD9 to Medcin translations. Also, the data from flat-file export from PM can be loaded into a staging environment via SSIS such that it can be analyzed and summarized while facilitating persistence.
• Make sure to tie the problem conversion load to a specific provider, that way if side effects or issues are identified after the fact, there is a clear way to identify which problems were loaded in the conversion via the provider they are tied to. The interface log should also have a record of this, but most organizations set the retention time to 90 days.
• Workflow validation – ensure that the workflow to move problems from PMH to Active will not be a barrier to use.

If your organization is looking for assistance in data conversion, please contact sales@galenhealthcare.com and visit our website for more information regarding our technical service offerings.

Tags: Allscripts, Allscripts Enterprise EHR, Data Conversion, ICD9, Interfaces, Medcin, Problem Conversion, TouchWorks Problem

One of the many challenges in interfacing to the Electronic Healthcare Record is patient identification and matching. Results and documents from outside systems need to link to the correct patient record. This is especially profound in data conversion initiatives. Given the scenario of an organization converting to utilize an EHR, aside from the plethora of documents being scanned in and associated with the chart as well as “bulk loads”  from the practice management system, there are could also be several data silos which need to feed data into the EHR.

We encountered one such scenario with one of our clients. Our client had been processing and loading flat-files from its legacy systems into the EHR. The client loaded approximately 15 years of legacy data (equating to millions of records). In the import process, the client had followed a strict patient matching criteria and received a patient matching error rate of approximately 5% which may be considered a reasonable matching rate.

However, the client’s help desk was getting a multitude of calls reporting missing legacy system records in the EHR (suspected to be in the 5% that did not make the conversion). The issue working against the client was a drop-dead date upon which these legacy systems were being deprecated and thus the clinicians would no-longer have a “fall-back” plan to access the records – the repercussions of which were potential patient care issues.

As such, Galen was engaged to analyze the records that did not meet the strict patient matching criteria , determine which records could be successfully loaded to the EHR under relaxed patient matching rules, and describe the impact of relaxing the patient matching. In the analysis that followed, it was recognized that in the data set that erred due to patient matching errors, identifier fields (namely first name, last name, DOB, MRN, Other Number1 and Other Number2) exhibited typos and inconsistencies. Enter Microsoft SQL Server Integration Studio’s (SSIS) Fuzzy Lookup Transformation. For those unfamiliar with fuzzy logic, it is “the process of reaching conclusions based on information and facts that are not 100 percent certain.”

The underlying algorithm to the Fuzzy Matching transformation is the SOUNDEX function:

• In the late nineteenth century, United States census officials faced a dilemma. During the process of counting the huddled masses, our public servants created a huge paperwork trail that the law required them to preserve for future historians. With amazing forethought, they realized that people searching for records might not know the exact spelling of their ancestor’s name. Was it Smith or Smythe? Chapple, Chapel or Chapelle?

• To ease these searches, census officials turned to the Soundex phonetic filing system. This system uses a simple phonetic algorithm to reduce each name to a four character alphanumeric code. The first letter of the code corresponds to the first letter of the last name. The remainder of the code consists of three digits derived from the syllables of the word.

• Largely unused outside of the halls of government and genealogy, the Soundex system is making a comeback in modern databases. Database developers have long struggled with the problem of matching words that might not look alike, but actually sound alike.

Thus to reclaim some of the records that erred in matching to a patient chart in the EHR, the Fuzzy Matching transformation was utilized. Flat-files output from legacy data silos were input, pre-processed and then fed to the transformation. Given previous studies, the matching criterion utilized was as follows:  Match on LastName and FirstName Similarity Threshold >.8 AND DOB matches exactly AND one of three (MRN, OtherNumber, OtherNumber2) cross-referenced match exactly. The end result was reclamation of close to 25% of those legacy system patient records that originally failed patient matching.

If your organization is looking for assistance in data conversion, please contact sales@galenhealthcare.com and visit our website for more information regarding our technical service offerings.

Tags: Allscripts Enterprise EHR, Data Conversion, Database, Integration, Patient Matching, Technical, TouchWorks