This page explains how educational content across the Shrink Network is sourced, evaluated, written, and reviewed. It's the network wide standard. Every editorial site (Shrinkopedia, AnxietyResource, DepressionResource, PsychiatryRx, AnxietyResearch, Shrinktionary, shrinkDaily) follows it, and individual properties may publish their own methodology pages with property specific detail. Our goal is simple to state and hard to do well: say what the evidence actually says, tag how strong it is, name the sources, and have a board certified psychiatrist read every clinical claim before it goes live.
What evidence we use
We build clinical claims from primary and authoritative sources, in this order of preference:
- The DSM-5-TR for diagnostic criteria and definitions. When we describe what a condition is, how it's diagnosed, or what its formal criteria are, that's where we start.
- Peer reviewed primary research indexed in PubMed, with a preference for randomized controlled trials and prospective cohort studies over case series or cross sectional work.
- Systematic reviews and meta analyses, which pool many studies and carry more weight than any single trial. A Cochrane review or a large meta analysis is stronger evidence than one promising study.
- Specialty society guidelines from bodies such as the American Psychiatric Association, which synthesize the evidence into clinical recommendations.
- US government and agency sources, including FDA labeling, MedlinePlus, NIMH, CDC, SAMHSA, and HRSA.
- Standard psychiatric textbooks and reference works for established, settled knowledge.
What we don't use as primary sources
Some material is useful for context but can't carry a clinical claim on its own. We don't base medical statements on news articles, press releases, or marketing material from a manufacturer. We don't treat opinion pieces, blog posts, or social media as evidence. We don't build claims from individual patient experience or testimonial content; a single person's story can illustrate what something feels like, but it can't establish that a treatment works or how often a side effect occurs. When a popular claim is circulating that the evidence doesn't support, we'll sometimes address it directly, but we'll say plainly that the underlying source isn't reliable and explain what the better evidence shows.
How we tag evidence strength
Not all evidence is equal, and pretending it's misleads the reader. Every substantive clinical claim is tagged on a four level scale so you can see how much confidence to place in it.
- Strong. Backed by multiple high quality randomized controlled trials, large meta analyses, or consistent specialty society guidelines that agree. The finding has been replicated and the effect is well established. Example: that SSRIs are an effective first line treatment for major depressive disorder is a Strong claim, supported by decades of trials and every major guideline.
- Moderate. Supported by reasonable evidence (some randomized trials or consistent observational data) but with limits: fewer studies, smaller samples, mixed results, or effects that apply mainly to one population. The direction is fairly clear, but the certainty isn't complete. Example: that regular aerobic exercise reduces depressive symptoms is a Moderate claim; the evidence is real and consistent but the trials vary in quality and size.
- Limited. Based on a small number of studies, lower quality designs, or indirect evidence. The claim is plausible and worth knowing, but it could change as better studies come in. We say so explicitly. Example: many specific supplement claims for anxiety sit here, where a few small trials suggest a benefit that larger studies haven't confirmed.
- Emerging. Early research, preliminary findings, or active areas of investigation where the picture isn't settled. We include these when they're clinically relevant, but we flag clearly that the evidence is preliminary and shouldn't drive a treatment decision on its own. Example: much of the work on psychedelic assisted therapy for treatment resistant conditions is Emerging.
The tag isn't decoration. It changes how a claim is written. A Strong claim can be stated plainly. An Emerging claim is always hedged, always dated, and always framed as something to watch rather than something to act on.
How claims are written
Statements reflect what the evidence actually says. Where the literature is strong, the page says so. Where the evidence is mixed, limited, or applies only to a specific population, the page says that too. We don't present opinion as fact, and we don't extend findings beyond the population studied. If a drug was tested in adults, we don't quietly imply the same result in adolescents. Where it's implemented, sources are cited inline with links to the underlying study, agency page, or guideline, and we prefer links to primary sources over links to summaries.
How review happens
Every clinical entry goes through two layers. First, editorial review by a human editor for clarity, structure, and tone. Second, medical review by Shariq Refai, MD, MBA, a board certified psychiatrist in active clinical practice. He reads every clinical entry before it publishes. If a claim is wrong, unclear, or claims more certainty than the evidence supports, the entry goes back for revision. Where it's implemented, the reviewer's name, the date of review, and the date of the last source check appear on the page.
Review isn't a one time event. An entry is flagged for re-review when significant new evidence or guidance appears (a new meta analysis, a revised guideline, an FDA label change, a safety warning), on a recurring schedule even when nothing has obviously changed, and whenever a reader reports a possible error through the corrections process. Substantive corrections are logged publicly.
How this plays out on a real entry
Take a medication entry on PsychiatryRx for an SSRI. The claim that it's effective for major depression is tagged Strong and stated directly, with the guideline and meta analysis cited. The claim about how quickly it works is tagged Moderate, because the timeline varies and the studies don't all agree, so the entry gives a range rather than a single number. A claim about an off label use with only a few small trials behind it's tagged Limited, written cautiously, and explicitly framed as not yet established. If a new trial lands that changes the picture, the entry is flagged, the reviewer reads the new evidence, the claim and its tag get updated, and the change isn'ted. The reader sees not just what we think is true, but how confident anyone should be about it.
AI assistance
AI tools may be used for drafting, organization, summarization, and grammar review. AI doesn't independently publish content, doesn't make clinical judgments, and doesn't replace human editorial or medical review. The AI use policy describes this in detail.
What this methodology doesn't do
Educational content describes general patterns in research and clinical practice. It doesn't diagnose any reader, doesn't recommend a specific treatment for an individual, and doesn't substitute for care from a licensed clinician. For clinical care, see shrinkMD or another licensed provider.
This page covers how we handle evidence in educational content. Standards for the care itself are a separate question. For those, see the Digital Psychiatry Index, shrinkMD's standards framework for responsible telepsychiatry. shrinkMD is part of The Shrink Network, founded by Shariq Refai, MD.
Contact
Questions about sourcing or methodology can be sent to [email protected].