Barry Rhodes

1
NHSN InnovationsCreation of Synthetic Data Sets
For Ventilator Associated Events Funded
Through the CHIIC Awards2014

• Barry Rhodes

2
What is the National Healthcare Safety Network
(NHSN)?

• A secure, internet-based system for monitoring
  healthcare-associated adverse events and
  practices
• Participating healthcare facilities use the NHSN
  web application to enter, analyze, and share data
• Data entry is predominantly manual with some
  electronic submission via Clinical Document
  Architecture (CDA) documents
• States get HAI data from NHSN based on state
  mandates
• CMS gets aggregated data for quality monitoring
  and payment purposes

3
NHSN Supports

• Over 36,000 users representing 14,000 facilities
• 3,400 facilities 25 reporting at least some
  data electronically via CDA
• National-level public health surveillance (CDC)
• State reporting mandates (30 states and DC)
• CMS reimbursement programs (Inpatient Prospective
  Payment System)

4
NHSN Electronic Data Flow Diagram
Quality Measurement Data
CDA Messages
CMS
Healthcare Facility
NHSN
De-identified Data Sets
Identified Data Sets (SAS, CSV, Excel)
State Public Health
CDC Analysts
5
Public Health Electronic Reporting Food Chain
We are here!

• Public Health or Facility Action
• Validation (State or Facility Based)
• Electronic Reporting
• CDA creation
• Event Determination (algorithmic or human)
• Data aggregation (EHR)
• Data normalization (vocab mapping)

6
So Whats the Problem?

• How do we know the definitions are being applied
  properly?
• Working with data providers to create conformant
  CDA messages is a good thing.
• The CDA represents the providers assessment that
  there is something to report based on NHSN
  definitions
• There is a validation step but that is manual,
  labor intensive and can be limited in scope

7
Whats Needed?

• Need a way to electronically validate that the
  data provider is applying the NHSN definitions
  properly
• So that when a CDA message is sent to NHSN, we
  can have some level of comfort that it accurately
  represents an HAI event.
• Two approaches to this problem
• Web Services
• Synthetic Data Sets

8
Funded CHIIC Project

• Synthetic Patient Dataset for the Determination
  of Ventilator-Associated Events in Electronic
  Health Record Systems
• Barry Rhodes, PhD
• Shelley Magill, MD, PhD
• Cindy Gross, MT (ASCP)

9
Procedure

• Working with NHSN SMEs and a healthcare facility,
  create synthetic EHR patient data seeded with
  Ventilator-Associated Events (VAEs)
• We used a de-identified data set from a partner
  facility to make the data as real as possible.
• We tried to think up all the normal
  possibilities for VAEs and as many edge cases
  as seems appropriate.
• Created the data set in XML
• Also created a result set with gives the VAE
  determination for each patient

10
Procedure (continued)

• Provide the data set and result set to vendors
  and other data providers to import into their
  systems
• Allow them to self-certify that their local
  implementations of the VAE definitions are valid.

11

• lt?xml version"1.0" encoding"ISO-8859-1" ?gt
• lt?xml-model href"vaeData.xsd"?gt
• ltvaeData xmlns"http//nhsn.phiresearchlab.org/ns"
  xmlnsxsi"http//www.w3.org/2001/XMLSchema-insta
  nce" xsinoNamespaceSchemaLocation"vaeData.xsd"gt
• ltpatientgt
• ltpatientIndexgt1lt/patientIndexgt
• ltventDaygt
• ltdaygt1lt/daygt
• ltpeepgt2.4lt/peepgt
• ltfio2gt33.0lt/fio2gt
• ltminTempgt38.4lt/minTempgt
• ltmaxTempgt38.7lt/maxTempgt
• ltminWbcgt4850lt/minWbcgt
• ltmaxWbcgt5200lt/maxWbcgt
• lt/ventDaygt
• ltventDaygt
• ltdaygt2lt/daygt
• ltpeepgt2.4lt/peepgt
• ltfio2gt33.0lt/fio2gt
• ltminTempgt38.4lt/minTempgt

12
Current Status

• Sent dataset to vendor community
• Received feedback from EPIC (they found an error
  actually!)
• Working with Partners Healthcare in Boston to
  validate their VAE efforts
• Waiting for more feedback to publish the dataset
  with instructions to the NHSN web site.

13
Where is This Pilot Going?

• Repeat the process for other HAIs
• Extend this scenario to a more formal EHR
  certification process.
• Provide synthetic dataset to vendor
• Vendor submits CDAs extracted from the data set
  to an NSHN certification site
• NHSN certification checks the CDAs against the
  known result set and certifies the vendor.

14
NHSN Innovations Philosophy and Why CHIIC is such
a Good Thing for CDC

• This project is one of several that are ongoing
  in NHSN under the heading of Informatics
  Innovations
• I believe there are many low cost, low
  technology, quick turn around projects that may
  have a broad impact on very difficult issues.
• Projects like this are good because by virtue of
  their low cost, they have the privilege of
  failure, i.e. If we do ten small projects and
  8 fail, we still have two successful ones and
  weve learned 8 new things!

15
Comments/ Questions?Please send accolades to
brhodes_at_cdc.gov