Methodology: how we source, report, review, and correct

Every statistical claim on Black Health cites a primary source in the same paragraph as the claim. Primary sources are: peer-reviewed papers (with a PubMed ID or DOI), federal datasets (CDC, NIH, HRSA, FDA, CMS, USPSTF), named-expert on-record interviews, court filings, public-record transcripts, and major medical-society guidelines (ACOG, AHA, ACS, NCCN, APA, and the relevant specialty society for the topic).

What we do not cite as primary: press releases when the underlying paper is available; the abstract of a paper we have not read in full; a secondary news write-up when the primary report is public; or another aggregator's coverage of a study. "Studies show" without a study is a kill signal for our editor.

When a study's sample was mostly white, the article says so before citing the effect size. We do not silently extrapolate non-Black research to Black patients. When a number is older than 24 months we label it with the year ("In 2022, the CDC reported...") rather than implying freshness that does not exist.

A piece moves through five hands before publish. The news scout surfaces the candidate and the angle. The research lead produces an evidence memo: study list with PMIDs, effect-size synthesis, Black-specific findings, quotable stats with exact phrasing, open questions. The editor briefs the writer with the angle and the evidence the piece must rest on. The writer drafts against the evidence memo and verifies that every URL in the citations resolves before handoff. The editor reads end to end, enforces voice, and either approves or returns with specific revision notes.

A piece does not publish without an original-value element. Acceptable originals: a named-source interview with an on-the-record quote; an original analysis we ran against a public dataset; a field observation from a reporting visit; a synthesis across three or more primary sources that no upstream article combines; a reader-mail synthesis with aggregated patterns; an original chart or table built from public data with the source cited. A summary of a single press release or a paraphrase of a single paper is not an original-value element; the editor kills it.

Every clinical article is reviewed by a licensed clinician separate from the writer. The reviewer reads the full draft, checks every clinical claim against current guidelines, flags out-of-date guidance for rewrite, and signs off only when the piece reflects the current state of evidence.

Disagreements between the writer and the reviewer go to the editor. The editor resolves them by re-reading the primary source, not by averaging the two positions. If the source is ambiguous, the article says so in the body.

The reviewer's full name and credentials appear on the byline of every article they reviewed, and link to a public profile page that lists every other piece they have reviewed for us. Anonymous medical review does not exist on this site. The full active board is published at /medical-reviewers/.

AI tools assist our team with research compilation, structure suggestion, citation cross-checking, brand-voice grep, and dataset summarization for editor briefs. AI does not author the prose on our published pages. Every published article has a named human author or editorial team byline that represents a person standing behind the reporting; we do not present AI as the author and we do not give AI a byline.

When AI materially shapes a piece, for example summarizing a 200-page dataset for the editor brief, the methodology behind the analysis is documented in the article body or here. We do not publish AI-generated body content with no human writing pass. AI-drafted text that has not been reviewed, edited, and approved by a named human on our team does not ship.

Errors happen. When a fact in a published piece is challenged and we cannot re-verify it, we issue a dated correction note at the bottom of the article, strike through the incorrect claim, and link to the correction from any other Black Health URL that cited the original. Corrections are never silent edits.

To flag an error, email hello@blackhealth.org with the page URL, the specific claim, and any supporting documentation. We acknowledge every correction request within five business days. The acknowledgement either confirms the correction, explains why we are not making the change with the primary source that supports our wording, or asks for additional information so we can adjudicate fairly.

Our running record of published corrections is at /policies/corrections/.

Evergreen clinical articles run a 12-to-18-month re-review cadence by default. Faster-moving treatment topics, including any article touching FDA approvals, USPSTF guideline changes, or major society guideline updates, are reviewed every 6 months. News dispatches are dated at publish and not re-reviewed; if the underlying news changes substantively we publish a new dispatch and link the two.

The "Last medically reviewed" date on an article byline is the date the current reviewer signed off, not the date we touched a layout pixel or fixed a typo. We do not bump dates to fake freshness.

Black Health is editorially independent. No sponsor, advertiser, or partner influences which stories we cover, how we cover them, or the clinical conclusions we reach.

Affiliate relationships, when present, are disclosed at the top of the relevant content, with the disclosure visible at the same scroll position as the affiliate link, per FTC guidance. We disclose in plain English ("We earn a commission if you buy through this link") rather than relying on a generic footer link.

Sponsored content is labeled "Sponsored" with the sponsor's name above the fold. Sponsored articles do not appear in our medically-reviewed editorial workflow and are not eligible for our evidence-based trust badge. Display ads are programmatic Google AdSense; we do not control which ads serve, and ad placement does not influence editorial selection.

For sponsorship, ad, or partnership inquiries: hello@blackhealth.org. For an editorial pitch or correction: same address.