Open-Source Methodology

We use two official NIH data systems, both maintained by the Office of Extramural Research:

Raw data goes through a multi-step pipeline before producing rankings. Each step is implemented as an auditable script in our open-source repository:

1. 1

   Ingest

   Fetch all NIH awards for the target fiscal year from the RePORTER API (paginated, 500 records per request). For historical years, parse ExPORTER CSV files. Store raw award records with project number, award amount, fiscal year, organization name, department code, and PI details.

2. 2

   Deduplicate

   NIH projects can have multiple sub-projects and supplements. We deduplicate by project number + fiscal year, summing award amounts across sub-projects. Each unique project is counted once toward grant counts.

3. 3

   Normalize Institutions

   The NIH database contains hundreds of name variations for the same institution (see Section 3 below). We apply ~200 alias mappings to consolidate these into canonical institution records, each with a unique slug, city, state, and organization type classification.

4. 4

   Classify Departments

   Map NIH department codes to our 27 standard department categories (see Section 4 below). Awards with unrecognized or missing department codes are included in institution-level rankings but excluded from department-specific rankings.

5. 5

   Classify Organization Types

   Each institution is classified as School of Medicine, University, Hospital, Research Institute, Government Lab, or Independent Organization. This classification determines which ranking tables an institution appears in — only Schools of Medicine appear in the medical school rankings.

6. 6

   Aggregate & Rank

   For each ranking type (medical school, all institutions, department, state, PI), sum total funding across all qualifying awards and sort descending. Rank ties are broken by grant count, then PI count. No weighting, normalization, or adjustment is applied — rankings reflect raw total award dollars.

7. 7

   Store & Index

   Final rankings are stored in PostgreSQL with full-text search indexes on institution and PI names. Historical rankings are preserved for trend analysis. The web application queries this database directly for all ranking pages and API endpoints.

The single most critical step in producing accurate rankings is institution name normalization. The NIH database contains hundreds of variations for the same institution, which must be consolidated to avoid double-counting and produce meaningful rankings.

Why this matters

Without normalization, a university with three separately named entities in the NIH system would appear as three separate institutions in the rankings, each with a fraction of their true funding. Standardization ensures that all funding flows to the correct canonical institution.

Alias mapping examples

We maintain approximately 200 alias mappings as open-source code. These are curated manually, cross-referenced against official NIH institution records (IPF numbers), and updated when institutions merge, rename, or restructure. The full alias list is available in our GitHub repository.

Harvard: a special case

Harvard Medical School presents a unique challenge. NIH awards flow to individual hospitals (MGH, BWH, BIDMC, etc.) rather than to "Harvard" directly. For medical school rankings, we aggregate funding from all major Harvard-affiliated teaching hospitals under "Harvard Medical School." For the all-institution rankings, each hospital appears separately, reflecting where the research is actually performed.

NIH awards include a department code from the IMPAC II system, which we map to 27 standard medical departments organized into two categories. These correspond to departments as they typically exist in US medical schools.

See all departments and their aggregate funding on the Departments page and Disciplines overview.

Department combining rules

Some NIH department codes map to combined categories. For example:

• INTERNAL MEDICINE/MEDICINE — Combines internal medicine and general medicine divisions
• MICROBIOLOGY/IMMUN/VIROLOGY — Combines microbiology, immunology, and virology departments
• PUBLIC HEALTH & PREV MEDICINE — Combines public health and preventive medicine
• RADIATION-DIAGNOSTIC/ONCOLOGY — Mapped to "Radiology" (includes diagnostic radiology and radiation oncology)
• PHYSICAL MEDICINE & REHAB — Combines PM&R and rehabilitation medicine

Awards with department codes outside these 27 categories (e.g., Dentistry, Psychology, Veterinary Sciences) are included in institution-level and PI-level rankings but do not appear in the department breakdown tables.

ScienceDex produces six distinct ranking tables, each with its own inclusion criteria:

Rankings are computed by simple summation and sorting. We deliberately avoid complex weighting or normalization schemes to maintain transparency and reproducibility.

Geographic rankings aggregate NIH funding by state (using the institution's state field from the NIH database). Per-capita funding is calculated using US Census Bureau population estimates (2024 vintage).

• Total funding by state: Sum of all NIH awards to institutions in that state
• Per-capita funding: Total state funding / state population (Census 2024 estimates)
• Year-over-year change: Percentage change in total state funding from the previous fiscal year
• City-level breakdown: Available within each state detail page, aggregated by the city field in NIH records

DC and Puerto Rico are included as separate entities. US territories other than Puerto Rico are not separately broken out due to small sample sizes.

Rankings are available from FY2001 through FY2024. Historical data enables:

• Rank trajectories: How an institution's position has changed over time (visible as sparklines in tables and in the interactive bump chart)
• Funding trends: Total dollar amounts by year for any institution, department, or state
• Year-over-year movers: Institutions with the largest rank changes between consecutive fiscal years (see Trends)

Note on inflation: All dollar amounts are nominal (not adjusted for inflation or purchasing power). This is consistent with how NIH reports its own funding data. Users should account for inflation when comparing dollar amounts across widely separated fiscal years.

We document the following limitations transparently:

• Subcontract inclusion: The RePORTER API includes subcontract awards. This means some funding may be double-counted — once for the prime awardee and once for the subcontract performer. This primarily affects large multi-site clinical trials.
• Data timing: NIH data is a point-in-time snapshot. Awards may be added, corrected, or reclassified after our ingestion date. We re-ingest when significant corrections are published.
• Department codes: Not all NIH awards have department codes. Awards without department codes are included in institution-level rankings but cannot be attributed to specific departments.
• Multi-PI awards: Awards with multiple PIs are attributed to the contact PI's institution. Funding is not split across co-PI institutions unless separate subcontracts exist.
• Institution classification: Organization type classification (School of Medicine vs. University vs. Hospital) is based on NIH's own institutional profile, supplemented by manual curation. Edge cases exist for institutions with both a medical school and a broader university identity.
• Nominal dollars: Rankings reflect nominal (current-year) dollars. No inflation adjustment is applied.

All ranking data is available programmatically via JSON API endpoints. These are the same endpoints that power the ScienceDex website.

Common query parameters: year (fiscal year), limit (pagination), offset (pagination), search (text search), sort (sort field).

Our entire pipeline is open-source under the MIT license:

Anyone with access to the NIH RePORTER API can reproduce our rankings by running these scripts. We welcome corrections, additions, and improvements via pull requests or email.