Risk Stratification Running Meeting Notes

Follow Up - CHCN

• NACHC met with Alameda and health center leaders at P&I . They are interested but there might be some administrative barriers.

• CHCN has access to claims data, 97% of patients have SDOH/PRAPARE data.

• Concerns were regarding security of the data, cloud etc.

• DUAs - NACHC executes an agreement with the network. Network is responsible for having the appropriate agreements in place with health centers. Individual health centers can opt in or opt out of the project.

  • Could ask for dummy health centers names instead of real health center names.

Next Steps for CHCN

• CHCN will have internal discussion and estimate level of effort.

Budget

• NACHC is waiting on approval of the full budget/workplan.

• NACHC presented this project a month ago to the sponsor - Aetna/CVS Health.

AWS, Other Data Sources

• AWS has two open data stores that we could potentially look into.

  • Health data hub?

Considering Speaking to Asian Health Services (Member of CHCN)

This health center usually holds back on sending EMR data to the network, very protective. Only as needed.

Use carryover from Adult Immunizations?

Risk Stratification Phases

Phase 1: Collect Data and Validate

Phase 2: Testing and Outcomes

Phase 3: Implementing the Tool in EHRs

CHCN

Meeting with them during P&I

CHCN Meeting

• Ongoing project on patients experiencing homelessness and housing insecurity

• 95% of patients have SDOH data

• CHCN data team is escalating the Risk Stratification proposal to CHCN leadership (CIO and CMO)

  • Aiming to set up a meeting before P&I or during the P&I conference

• Tentative yes - Need to determine level of effort for data pull, slowest step would be NACHC contracting

• If we are moving forward, need to draft SOW

  • Task 1: Provide subject matter expertise on data set and experience with PRAPARE, VBC

  • Task 2: Provide data.

  • Due end of July, can execute no cost extension if needed.

• Ideally, CHCN would send NACHC data, AWS can help with data cleaning and normalization.

  • Need to have the right agreements with AWS - No selling the work. Free for health centers.

Epic

• “App orchard” - people can access applications via Epic

  • Can explore building the application via this method or in Epic.

• CHCN is on the OCHIN Epic instance as well

• NACHC has had conversations with Epic about ODH data. Still need to discuss SDOH.

Funding

• Focusing in CHCN and Siouxland

RFA-DK-22-038: Pilot Interventions to Integrate Social Care and Medical Care to Improve Health Equity (R01 Clinical Trial Required) (nih.gov)

Due end of September. Could focus on diabetes for Siouxland and housing insecurity for CHCN.

Integrating Risk Stratification into Epic

Challenges with PRAPARE licensing?

Some health centers have trouble getting data out of Epic.

Siouxland

• Originally developed the Risk Calculator in i2i but Epic says they can build the calculator in Epic.

  • Currently running reports to identify high risk patients (Score of 17 or higher) for follow up.

  • Integration with the EHR is ideal. OCHIN is the implementation partner - using OCHIN Epic.

• Siouxland has access to all the data needed included utilization data.

  • When they get an ER report, there is an obs term for ER visits/admission.

• Providing care in Iowa, Nebraska and South Dakota

• Estimate of level of effort? How are patients engaged?

Behavioral health, RN health coaches and CHWs are using the i2i system/reports. Total of 15 people.

Workflow

1. Teams know who is due for PRAPARE screening annually

2. CHW or provider administers and provides intervention when needed

3. 50% of patients seen in the last year have PRAPARE screening

Tool also built into NextGen by Heartland

Use in EHR vs Population Management System

Most EHR data warehouses do not integrate claim data, usually in a data warehouse that is external to the EHR.

Dave: Epic gets claim data and there are other ways to get that utilization data other than claims data.

Thoughts on piloting?

Depends on load for Epic implementation

Contacts at Epic

Nee to ensure that the tool will be no cost for health centers

Dave has a contact at OCHIN Epic. Julia has a contact at Epic.

NACHC Feedback on the Risk Stratification Model Validation Concept

• Goal

  • Test how well the algorithm works and refine as needed.

  • Is it accurate?

  • Is it helpful?

  • Health centers can plan targeted interventions, prioritize care team workload, etc.

• Might be better to start with who has access to the data that we need (Pedro)

• Different stages of evaluation using the same data (Jen)

  • How well does it fit when we remove certain variables?

  • Proxys for those that don’t have the data?

• How many patients worth of data is needed for validation?

  • Aim 1: Need organizations that have 51% of more patients with PRAPARE data

    • General validation

  • Aim 3: Need organizations that have 90% of more patients with diabetes and PRAPARE data

    • Clinical utility of algorithm with a specific population - Diabetes

Utilization data including emergency department (ED) visit and inpatient hospital (IP) stays by CPT codes

Number of days in hospital? Number of encounters?

Roles

NACHC: Data collection.

AAPCHO: Once dataset is received and organized, work with analytics partner.

Siouxland

Now on Epic and wants to build the tool on Epic.

Per Julia, data warehouse may not be in the EHR. Might more make sense to work with data aggregation partners.

NACHC Data Update

• AllianceChicago

  • Check which PCAs/HCCNs have the most complete data.

  • NACHC did not explicitly ask for many of the data elements.

  • Probably need to start over - select partners, new data request, etc.

• NACHC is mainly working on remapping which is a heavy lift - especially SDOH data.

Contracting

• Data Validation

  • Consider:

    • Siouxland

    • Heartland

    • Colorado

    • CHCN is preliminarily interested per Jen

• AWS

Next Steps

• Partner selection (Siouxland, Heartland, Colorado)

  • Check in with CyncHealth, CHCN is probably a better option

• N3C data

• Draft proposal

Original Proposal

• Palmira is involved in talking to CHCs and determining the pathway - how does the PRAPARE tool inform and make a difference for the patient? If centers are using the tool, what happens? Referral? Follow up?

  • Identification of social needs and steps for follow up.

  • Nothing to do with risk.

Risk Stratification

• Stratification tool developed with Learning Communities. Clinics say that it is useful but validation (check data for predictive and clinical utility) is needed.

  • Siouxland is interested. Possibly Heartland.

• How many patients are needed to validate? Number of patients with positive finding in a specific domain?

  • Rosy estimates at least 1000 per site.

  • Some health centers think diabetes could be a good outcome or domain.

• Consider N3C data - 16 million patients - real claims and admission data. Total cost of care data might be missing.

• Consider working with OCHIN and AllianceChicago to send additional SDOH data.

  • Could be a control/comparison.

  • Funding?

• What outcomes are predicted using this algorithm?

  • Cost (health centers pulled this information based on chart review - need to check if cost was based on claims data).

  • Could also explore other outcomes like diabetes.

• Need to refine domains - mental health, SDOH.

• Cost is a proxy for a lot of other things.

Heartland

• SDOH data is PRAPARE and directly in their EMR.

• Generating the utilization data by asking the patient - large burden. 100 patients with full populated data, goal is 400.

• Slightly modified tool - weighted everything equally.

• CyncHealth has admission data from the HIE - probably our best bet.

• Displayed interest in participating.

Amazon Proposal

Risk stratification and AWS work are related but not dependent on one another.

AWS Work:

• Look at external data that is not coded at PRAPARE.

• Machine learning on data that we have.

• Timeline: June 2023. We can also apply for another grant.

Ultimately, we do not need to stick to the AWS proposal and timeline.

Next Meeting

Cancel December 21. Meet on January 4.

Original Proposal

There are concerns about the direction of the project and timeline. Need to focus on validating the algorithm.

Space for Notes, Document Storage, Planning

Confluence.

Review of NACHC SDOH Data

2022-11-16 Risk Stratification SDOH - COVID data Summary.pptm

• Current dataset does not include all variables included in the algorithm. Additional data from partners needed.

  • Funding source?

• Explore removing certain variables from the algorithm (if performance of the algorithm remains optimal).