Loading BCONZ...
HARM · Healthcare data readiness assessment
Healthcare data readiness assessment for research and AI
Before a hospital or laboratory enters a research data partnership, it needs to know what its data can actually support. HARM is BCONZ's data readiness assessment: it reads your data exports inside your own environment and produces a readiness report on patients, longitudinal follow-up, coding and identifiers.
It is read-only by design. HARM blocks network use in the tool itself, never copies or moves patient data, and never writes patient values into its report — so your IT and governance teams can verify the assessment is safe before anything is shared.
What a data readiness assessment checks
A research-ready dataset is more than a large one. HARM's data readiness checklist covers the questions a data partnership actually turns on:
- Volume: how many patients, visits and clinical records exist
- Longitudinal depth: how many patients have two or more visits, and the median follow-up span
- Coding: whether diagnoses (e.g. ICD), medications and laboratory results are coded or free text
- Identifiers: which columns hold direct or quasi-identifiers that must be removed before research use
- Structure: whether tables can be linked by patient and visit and mapped to a research data model
- Cohort floor counts: patients matching given diagnosis terms, reported as a floor, not a clinical cohort
Data readiness for AI
Healthcare AI projects fail most often on the data, not the model: too little follow-up to define outcomes, labels trapped in free text, or identifiers that block access. A data readiness assessment for AI surfaces those issues before a model development or validation project is scoped.
A governance-first assessment
HARM is also a data governance readiness assessment. Because it counts identifier columns and never outputs their values, it gives governance and ethics committees an evidence base for de-identification planning without exposing patient information during the assessment itself.
Frequently asked questions
HARM: common questions
What is a healthcare data readiness assessment?
It is a structured review of whether a healthcare dataset can support a specific kind of research — measuring volume, longitudinal follow-up, coding quality, identifiers and structure — before a partnership or project is committed to.
Does HARM send our data to BCONZ?
No. HARM runs on your own systems, blocks network use, and writes only aggregate counts to a report that you review before choosing whether to share it.
What file formats does HARM read?
Delimited exports such as CSV and TSV, and Excel workbooks where the environment supports them.
Is the cohort count a guarantee of eligible patients?
No. Cohort counts are a floor based on diagnosis text and code matching only, with no clinical adjudication or eligibility criteria applied. The report says so explicitly.
Related capabilities
DIA · Demand Intelligence Agent
Responsible healthcare data monetization starts with demand: DIA shows who has published a need for data like yours, and which of their needs your dataset meets.
PIA · Provider Intelligence Agent
Evidence-led clinical trial site selection and data partner identification: find the providers most likely to hold the patients and data a study needs.
See HARM on your own question or dataset.