Penn LDI Research Initiative on Improving Primary Care for Older Adults

Proposals must be led by an LDI Senior Fellow. Penn faculty without an LDI affiliation and LDI Associate Fellows may participate as co-investigators. External collaborators are permitted with justification. Funding priority will be given to observational studies using existing data and rigorous empirical methods, especially those examining the effect of value-based payment (VBP) and alternative delivery models on chronic disease management and outcomes for older adults, and those identifying opportunities to improve chronic disease management for older adults in primary care settings.

The Grant Review Committee will select up to two projects, each lasting 15 months. Project budgets may not exceed $175,000. Faculty salaries (up to the FY26 NIH cap) and benefits may be included only with strong justification and may not exceed 20% of the award.

Applications will be due Monday, November 10, 2025, at 11:59 p.m. ET. Funding decisions are expected to be made in December 2025. IRB approval is required before the release of funds. PIs must submit their projects to the IRB for approval on notice of funding. Selected projects are expected to start by March 2026.

Can an investigator mention relevant data they have access to in their proposal?

Yes, we recommend receiving a letter of support from the data owner to use data for your specific project.

Can investigator teams have more than one Principal Investigator?

Yes, teams can have up to two Principal Investigators.

Does the ban on indirect costs also apply to sub-awards for faculty co-investigators at other universities?

Yes.

Are graduate students or nurses working on projects subject to the 30% salary and benefits limit?

No, the 30% salary and benefits limit applies to only faculty (including research track/non-standing), not staff.

Are hyperlinks to online data (ie. video material) allowed to be included in the application?

Yes, applicants are allowed to link to online data in their application.

Are investigators allowed to include publication costs and/or software in the budget?

Yes, but if the grant pays 100% for the software, then it should be justified within the proposal.

Do I have to use the CenterWell Humana Dataset in my research?

While use of the dataset is encouraged, it is not required.

How is provider type represented in the data?

• The claims tables include standard taxonomy codes indicating provider specialty and provider type (e.g., physicians, nurse practitioners, physician assistants).
• There is an additional table for primary care provider attribution, which attributes patients to primary care organizations (at tax ID level). This table does not contain provider specialty or provider type information, because attribution occurs at the organizational level, not the individual provider level.

Does the “paid amount” variable include both plan and patient payments?

Only the total amount paid will be available, which includes both the plan share and the patient share (out-of-pocket).

Can urgent care visits be identified in the data?

• Yes. There is a “place of service” code that can be used to identify urgent care centers.
• There are no additional proprietary identifiers for urgent care in this dataset.

Can the data differentiate between annual preventive visits and other primary care visits?

• For Medicare patients, you can use G-codes to distinguish between:
  • Initial annual visit
  • Subsequent annual visit
• Otherwise, determining “unscheduled” visits is possible only through custom logic, which this claims data doesn’t directly indicate.

What information is available on value-based payment models?

• The “financial arrangement” variables categorically classifies primary care providers based on risk-sharing arrangements with Humana.
• Main Categories include:
  • No risk (fee-for-service)
  • Upside-only risk
  • Two-sided risk
• There are 9 total subcategories, offering more granularity within those three main types.
  • For example, within “full risk,” it distinguishes between:
    • Global budgets (100% responsibility for Parts A & B)
    • Partial risk (e.g., 100% Part A, 50% Part B)

Are unique identifiers available across data files to link members?

Yes. Each file includes unique pseudo-IDs, allowing enrollees to be matched across data sets.

How is PCP attribution handled for members who haven’t seen their PCP recently?

• HMO members are required to select a PCP (TIN) each year, so attribution is nearly complete.
• PPO member attribution and adjustments for HMO members are handled through an internal claims-based algorithm that assigns then subsequently adjusts attribution using factors such as:
  • Frequency (plurality) of visits
  • Distance between provider and patient

Is there overlap between CenterWell enrollees and Penn Medicine patients? Can the data link to Penn Medicine’s EHR?

• The data is de-identified under HIPAA expert determination and cannot be combined with other datasets by default.
• Any linkage would require a new expert determination and explicit approval from Humana.
• Practically, overlap is likely minimal, as only about 10% of members are in the Northeast, where Penn Medicine operates.

What data are available on health-related social needs (HRSN)?

• There is a separate, linkable dataset with HRSN information, collected via survey.
• It is limited:
  • Covers a subset of members over 2 years
  • Cross-sectional, repeated survey
  • Data is about 5 years old
• It includes member-reported data on multiple HRSN domains.
• The dataset is not included in the standard release, but researchers can request access if relevant to their proposal.

Can CenterWell data be linked with external datasets like the Social Vulnerability Index or Area Health Resource File?

• Possibly, but it requires review and approval to maintain HIPAA compliance.
• Publicly available data linkages can be proposed as part of the research design.
• CenterWell/Humana supports such proposals in principle but requires them to go through the expert determination process before combining data.
• The CenterWell/Humana team has prior experience linking data to indices such as ADI or AHRQ SDOH datasets.
• The additional data file contains patient assignment/attribution to a primary care TIN.

Have a question?

Please email Caleb Hearn.