Barr, Bigdeli & Meyers (2022) — Prevalence, comorbidity and sociodemographic correlates of psychiatric disorders in All of Us
Not an interoception study. This is a psychiatric-epidemiology Brief Report — a descriptive characterization of how common psychiatric diagnoses are, how heavily they co-occur, and how they distribute across social groups in the NIH All of Us biobank. It measures no interoceptive variable, cites no interoception literature, and makes no claim about mechanism. It is kept in the wiki for one reason: it is the population-scale evidence for the comorbidity structure that the transdiagnostic framing of interoception presupposes but never itself demonstrates. Read it the way the wiki keeps Folkman & Lazarus (1980) and Friston (2005) — a thematic/structural neighbour that supplies context the interoception sources lean on, not an interoception result.
What the paper did
A cross-sectional analysis of All of Us release 5 (N = 331,380 enrolled May 2018–April 2021; analytic N = 329,038; 60.7% female, mean age 50.9, SD 16.8). Diagnoses were derived from linked electronic health records using phecodes (curated groupings of ICD-10-CM billing codes), with ≥2 codes taken as a diagnosis, across six broad domains: mood, anxiety, substance-use, stress-related, schizophrenia and personality disorders. The analysis reports (1) prevalence per disorder against nationally representative benchmarks, (2) comorbidity via tetrachoric correlations and an overlap (upset-style) plot, and (3) adjusted odds ratios across sex, gender, sexual orientation, age, race/ethnicity, country of origin, education, income, insurance and marital status.
The three findings, and which one matters here
Prevalence, and the healthy-participant bias. Mood disorders were most common (11.0%), then anxiety (10.1%), substance use (7.2%) and stress-related (2.9%); schizophrenia (0.8%), ADHD (0.8%), personality (0.5%), eating (0.2%) and OCD (0.2%) each ≤1%. Every disorder was less prevalent than in nationally representative samples (e.g. any anxiety 10.1% vs 28.8%; any SUD 7.2% vs 32.3%) — the well-known biobank volunteer bias, sharpened by the fact that diagnoses require having sought treatment and having it billed.
Comorbidity — the part the wiki cares about. About 51% of diagnoses involved overlap; the majority of participants with any disorder carried ≥2 diagnoses; cross-disorder tetrachoric correlations ran r = 0.43–0.75; and the single commonest comorbid cluster (64% of the multi-diagnosis group) was mood + anxiety + substance use. The authors read this as support for the idea that psychiatric disorders “may in part share common causes” — the same intercorrelation that motivates the p-factor / general-psychopathology and RDoC literatures.
Sociodemographic disparities. Highest risk for non-Hispanic White, low-SES, women/assigned-female and sexual-minority participants; foreign-born and higher-education/income participants lower on every disorder; schizophrenia the lone reversal (higher for Black, multiracial and other non-White participants, attributed by the authors to racism and diagnostic bias). Insurance was positively associated with each disorder — an access-to-care artefact, not a risk factor. All of this “recapitulated previous epidemiological research.”
Why it is in the wiki
The interoceptive-psychopathology page opens on a claim borrowed from Khalsa et al. (2018) and Greenwood & Garfinkel (2025): that interoceptive dysfunction is a transdiagnostic dimension, better probed by RDoC than by DSM category. That argument has two premises — (a) interoception is disturbed across many disorders, and (b) those disorders are in fact cross-cutting rather than cleanly separable — and the interoception literature only ever argues (a). Barr et al. is population-scale evidence for (b): at N ≈ 329,000, psychiatric diagnoses are heavily comorbid, intercorrelated at r = 0.43–0.75, and dominated by a mood/anxiety/SUD cluster, which is exactly the co-occurrence structure a dimensional, mechanism-first nosology is built to explain. It supplies the epidemiological warrant that the transdiagnostic move quietly assumes.
Two honest limits on that use:
- It says nothing about interoception. A shared-common-causes reading of comorbidity is compatible with a transdiagnostic interoceptive mechanism, but it is equally compatible with any other shared cause (shared genetics — the authors’ own follow-up direction — early adversity, socioeconomic stress, or measurement/ascertainment overlap). This is the cause-or-consequence gap one level up: the comorbidity is real, its interoceptive interpretation is not tested here.
- The disparity findings touch early-adverse-life-events only obliquely. Low SES, uninsurance and relationship disruption track higher diagnosis rates, consistent with the social-adversity thread the wiki runs through Bonaz et al., but this is EHR association in a treatment-selected biobank, not a developmental or mechanistic result.
Provenance and status
Open access (CC-BY), JAMA Psychiatry Brief Report, published online 20 April 2022, doi printed on the paper. Authored by psychiatric epidemiologists at SUNY Downstate / VA New York Harbor (no interoception researchers; no researcher page created, consistent with the wiki’s convention for out-of-field context sources). No hard contradiction with existing content — it neither confirms nor denies any interoceptive claim; it grounds a background assumption. No wiki claim was overwritten.