How ScienceDex Rankings Are Built

ScienceDex is built from official NIH public award data. Current-year data comes from the NIH RePORTER API v2, which exposes project numbers, award amounts, fiscal years, principal investigators, institutions, department codes, and related award metadata.

Historical records are reconciled against NIH ExPORTER bulk files where flat-file downloads are better suited for reproducible, multi-year analysis.

1. 1

   Ingest public NIH records

   Fetch award records from NIH RePORTER API v2 and ExPORTER bulk files, then publish a versioned fiscal-year snapshot. The displayed release remains fixed until the ingestion and ranking refresh completes.

2. 2

   Deduplicate awards

   Group project records by project number and fiscal year, then sum award amounts across related subprojects when needed.

3. 3

   Normalize institutions

   Apply curated alias mappings so name variants, campus labels, and renamed entities roll up to canonical institution records.

4. 4

   Classify departments

   Map NIH IMPAC II department codes into 27 standard basic-science and clinical-science department categories.

5. 5

   Aggregate and rank

   Sum qualifying award dollars by ranking type, sort descending, and use grant count and investigator count as deterministic tie-breakers.

6. 6

   Publish indexed profiles

   Store ranking tables and profile data for institution, medical-school, department, investigator, geographic, and discipline pages.

NIH records often contain multiple names for the same organization. A single university can appear under campus names, hospital affiliates, old legal names, abbreviations, or medical-school labels. ScienceDex applies curated alias mappings so records roll up to a canonical institution before ranking.

Default grouping. Medical-school and department rankings count awards at affiliated teaching hospitals toward their medical school, the convention used in medical-school funding tables. For example, Massachusetts General Hospital, Brigham and Women's Hospital, Beth Israel Deaconess Medical Center, Boston Children's Hospital and Mass Eye and Ear count toward Harvard Medical School. The all-institution ranking lists every organization separately. Default rankings always use this grouping.

Institution pages list every department with NIH funding. The national rank is the institution's place in that department's ranking. For an affiliated hospital it is the rank of the medical school its awards count toward, shown with the school's name. Departments outside the 27 ranked categories read "Not ranked".

Department people pages. Each department on an institution or health-system page, and each row of a department ranking, opens a page listing the principal investigators on those awards. NIH names only principal investigators, and the department belongs to the award, so the page is not a list of the department's members. Contact PIs come first. Other PIs on those awards are listed as co-investigators, with the institution and department of the awards they lead themselves. Each award shows where its department came from: NIH's department code, the investigator's affiliation, or the award's topic. Investigator profiles show the institution and department of the awards the investigator leads.

Health systems. NIH funds a health system's hospitals as separate organizations and lists almost no awards under a system's name. A health-system page adds up the fiscal year's NIH awards of the system's members. Totals include every award, as on institution pages, and investigators are counted once across members. Department figures follow the department-ranking rules. These figures are computed when the fiscal year's rankings are rebuilt, not when a page is opened.

• Members. A system's own published member list, cited, overrides AHRQ for that system. Otherwise members come from AHRQ's Compendium of U.S. Health Systems, 2023 (hospital linkage file, data year 2023: 6,800 hospitals, 4,193 of them in 639 systems). A research institute or system office counts with its system when the system's own website says it belongs to it.
• Matching. Each NIH-funded hospital organization was matched to AHRQ's hospitals by name, city and state, ignoring spelling differences such as apostrophes, a leading "The", "Inc" and word order. A match on all three with exactly one AHRQ hospital passed automatically (43 organizations). Every other case was checked by hand against the AHRQ row and, where needed, the organization's own website (70). Organizations outside the United States, organizations that are not hospitals, and uncertain cases were left out (10, $31.8M).
• Coverage, FY2026. 123 hospital organizations with $4.2B in NIH awards were checked. 100 of them ($3.1B) belong to one of 83 health systems, and 13 ($1.1B) are hospitals in no system, such as St. Jude Children's Research Hospital and Memorial Sloan Kettering.
• Research arms. Medical schools and universities stay where the default rankings put them. Three that NIH classifies as a medical school or university but that are their systems' research arms also count with their systems on system pages and in the separate department view: Mayo Clinic (Mayo Clinic Health System), Vanderbilt University Medical Center (Vanderbilt Health) and the Cleveland Clinic Lerner College of Medicine (Cleveland Clinic), with $815.5M in FY2026 NIH awards. This was decided on September 27, 2026. Other schools and universities that run hospitals are not moved.
• Override. AHRQ's 2023 Compendium lists Dana-Farber Cancer Institute in Mass General Brigham. Mass General Brigham's own member list does not, so it is not counted.

Hospitals and health systems separately. Each department ranking has a second view. In it no hospital is grouped under a medical school: an award counts toward a listed health system when its organization is a member, and otherwise toward the organization NIH funded. Every award counts once, so the department total is the same in both views, and a medical school's row holds only awards made to the school itself. The default view does not change.

Department rankings use NIH department codes and map them into 27 standard medical department categories. The taxonomy separates basic-science departments from clinical departments so users can compare similar research areas across schools.

ScienceDex ranks by simple summation instead of hidden weighting. The goal is for a researcher, administrator, or vendor to understand exactly why one record appears above another.

ScienceDex extends the transparent ranking tradition used by independent NIH funding tables such as the Blue Ridge Institute for Medical Research. ScienceDex publishes its own cohort rules, source snapshots, metric versions, coverage, and release identifiers so each result can be inspected and reproduced.

Every published portfolio table and CSV identifies the NIH snapshot, fiscal year, membership rule, analysis release, SHARE release when applicable, and metric definition version. A new release never silently rewrites an older published result.

• Subcontracts: NIH subcontract records can make multi-site awards appear in both prime and subcontract contexts.
• Timing: NIH records can be corrected after publication, so rankings represent the data snapshot used for the release.
• Department codes: Awards without department codes remain in institution and investigator views but cannot be attributed to a department ranking. Hospital awards use the contact PI's own department from PubMed affiliations, or a topic-inferred department when no affiliation names one.
• Multi-PI awards: Awards are attributed from NIH award metadata. Each named PI is linked to the full award amount, so investigator totals and PI-award assignment counts exceed unique-award totals. They should not be read as the investigator's personal funding share.
• Nominal dollars: Historical funding is shown in reported dollars, not inflation-adjusted dollars.

ScienceDex methods are designed to be inspectable from this page: data sources, normalization rules, ranking computations, and limitations are documented together so users can understand how each ranking view is produced.

This Methods page is maintained alongside the ConductScience ScienceDex integration. Current displayed rankings use FY2027 data.