HearShare consortium
HearShareBuilding the first-in-field audiology knowledgebase.
The Harmonized Auditory Research Knowledgebase (HARK) will transform fragmented audiology and otology records into a secure, analysis-ready resource for discovery, clinical translation, and community use.
Vision
Hearing health research has the data. It does not yet have the shared resource.
Clinical audiology has collected vast amounts of real-world data, but those records remain fragmented across institutions, local schemas, vendor exports, and research cohorts. HearShare is building HARK to convert that distributed evidence into a governed, ontology-mapped knowledgebase for clinical discovery, regulatory science, and reusable community analytics.
Create data-driven hearing profiles that account for age, demographics, disease state, and longitudinal change.
Give uncommon conditions such as vestibular schwannoma enough statistical power for clinically relevant questions.
Make standards, tools, and derived products reusable by clinicians, researchers, regulators, and patient-facing teams.
HARK is designed to combine consortium clinical data with public cohorts including NHANES, BLSA, Framingham, HCHS, and NHATS.
Workstreams
Three workstreams to turn fragmented records into reusable knowledge.
Translational reach
HARK is designed for the people who need hearing evidence to move faster.
from pyhark import HARK
hark = HARK() # connects to BDSP-hosted HARK release
df = hark.query("""
SELECT subject_id, age, pta_right_db
FROM audiograms
WHERE noise_exposure_yrs > 10
""")
df.plot_audiogram(by="age_decade")
Programmatic access
Notebook-first analysis on shared infrastructure.
HARK is hosted alongside the analytic environment on BDSP, so approved investigators work in a notebook against the same release the consortium uses for its own studies. Code, derived cohorts, and methods can be shared as research artifacts.
- Open-source Python and R clients with a stable query API.
- Notebook-first workflows hosted alongside the data on BDSP.
- Optional LLM-assisted analytics; POD-Vis is one of several interfaces under development.
Harmonization
One reference record from many local conventions.
Audiometric data lives in heterogeneous local schemas, vendor-specific exports, and free-text fields. The consortium's harmonization work maps these sources to common audiologic and clinical terminologies so that a single query reaches across institutions.
Per-site fields
- thr_500_R
- PTA-L (db)
- noise_yrs
- audio_date
- sex_M_F
HARK canonical record
- threshold.right.500hz_db
- pta.left_db
- exposure.noise_years
- audiogram.measured_at
- demographics.sex
Consortium
Five academic medical centers, contributing data on BDSP.
Mass Eye and Ear
Boston, MA
Duke University
Durham, NC
Medical University of South Carolina
Charleston, SC
University of Maryland
Baltimore, MD
Oregon Health & Science University
Portland, OR
BDSP provides shared Python and analytics infrastructure, FAIR-aligned BIDS organization, tiered governance for sensitive data, and AWS Open Data sponsorship for long-term hosting.
M. Brandon Westover, BDSP / Beth Israel Deaconess Medical Center
Multi-PI leadership
Audiology and biomedical informatics co-PIs.
Kristal Riska
Contact PI
Duke University
Matthew Crowson
MPI
Mass Eye and Ear
Anup Mahurkar
MPI
University of Maryland School of Medicine
Kelly Reavis
MPI
OHSU / VA Portland
Scientific contributors: Judy Dubno (MUSC), M. Brandon Westover (BDSP, BIDMC), and the consortium.
Continue