Prescriptions, Paradoxes, and Perversities

An original-research data-journalism post (warned up front as amateur and unconfirmed). Scott aggregates antidepressant patient-ratings from Drugs.com/WebMD/AskAPatient (which inter-correlate ~0.7, so there's a real patient consensus) and surfaces three escalating findings. (I) A unified patient-favorite ranking with the old MAOIs (Nardil, Parnate) on top and the newest drugs at the bottom. (II) Doctor ratings (RateRx) correlate -0.21 with patient ratings, and the more patients like a drug the LESS it is prescribed -- doctors love what patients hate. (III) Patient rating correlates -0.71 with year of release: older antidepressants are rated better, robustly. He rigorously works the confounds -- crucially noting that older drugs are reserved for treatment-resistant ('fighting Cthulhu') cases, which should bias AGAINST them yet they still win -- enumerates and rebuts several alternative hypotheses, and lands on the pharma-patent-cycle explanation (companies fund research to push new on-patent drugs; the lone MAOI-vs-SSRI head-to-head study was authored by Eli Lilly employees, which he establishes via LinkedIn detective work). He then validates by re-running the analysis on 55,498 ratings across antipsychotics/antihypertensives/antidiabetics/anticonvulsants, showing the older-is-better effect is PURELY psychiatric while the doctor-patient anticorrelation holds across all classes, and publishes the spreadsheet for replication. A model of careful amateur empirical investigation with surprising, policy-relevant conclusions; an early seed of his later 'legible expertise' (WebMD) themes. Caveat: rests on self-selected online-review data (selection bias, the dead-don't-review problem he himself flags), so the specific magnitudes are soft even though the cross-validation is impressive.

Quality 78 · Excellent. Excellent (78): genuinely original empirical work -- surprising, well-supported findings (the doctor/patient rating anticorrelation; psychiatry-specific older-is-better), exemplary confound analysis and robustness checks, and real-world COI sleuthing -- that changes how a careful reader thinks about choosing an antidepressant. A clear early-cohort standout scored on merit. Held below the 80+ landmark tier because the data source is self-selected online reviews (heavily caveated) and the conclusions are suggestive rather than definitive.

Claude’s paradigm shift 58 · Moderate. Moderate-to-Notable (58): the empirical discoveries (the anticorrelation; the psychiatry-specific decline) and the pharma-patent-cycle synthesis were genuinely novel and non-obvious, though they build on existing skepticism about the SSRI-era and on standard correlational methods.

Real-world impact 2 · Minor. Genuinely original empirical data-journalism (antidepressant patient-ratings; the doctor/patient rating anticorrelation; older-is-better in psychiatry) that changes how a careful reader weighs antidepressant choice. Real topical relevance, but amateur/exploratory and discourse-level with no material change — low RWI.