Home › Research › Long-term cohort studies
Intelligence Breaks the System It Controls
Ewing MA.. Research Square, 2026-08-27
Topic: Health services, long-term care and technology
Using two large-scale longitudinal datasets covering more than 75,000 patients or stays and 7.8 million clinical observations—spanning emergency vasopressor titration in sepsis (MIMIC- IV; 6,243 ICU admissions drawn from 73,141 encounters) and serial cognitive assessment in Alzheimer’s disease (ADNI; 2,262 longitudinal participants)—we train a 13-family computa- tional model zoo and perform a novel experiment: auditing whether clinician and AI treatment decisions jointly or individually match the subsequent patient trajectory, and testing whether cooperation between the two actors resolves their disagreements. The experiment reveals that in acute care, where clinician–patient contact is brief, both clinician and AI share the same blind spot—neither matched the subsequent trajectory in 53% of treatment events, with neg- ligible cooperation surplus (< 0.003). In longitudinal care, the blind spot shrank to 11%, and the clinician’s accumulated private knowledge added up to 25 percentage points of directional information, but the model did not improve—clinical interaction enriches the clinician, not the model. Serial norepinephrine adjustments drove branch balance from 0.816 to 0.994, confirming that repeated treatment progressively erases directional identifiability. Seven dose-recommender families over-dosed at MAP 65 relative to clinicians by 12.5–104.2%. Direct measurement of serial trajectory deformation—how far the patient’s physiologic profile has drifted from its own initial shape—recovered 10–19 percentage points of directional resolution where action exchange added nothing. These findings show that the obstacle to safe clinical AI is not that clinicians and models need to cooperate better with the information they have. It is that neither agent currently measures the right thing: how treatment has changed the patient’s trajectory relative to its own history.
Key findings
- These findings show that the obstacle to safe clinical AI is not that clinicians and models need to cooperate better with the information they have.
Study details
- Study designlongitudinal study
- Participants75,000
- Effect size104.2%
- Evidence strengthModerate
How this study scored 16/100
Preprint — not yet peer-reviewed. It scored 50 on the same rubric, but non-peer-reviewed sources are capped at 16, so it shows as 16. Scored zero for randomised / blinded because the record does not state it. We have no data on times cited, citations per year, impact for its field and year, cited by studies here, still being cited, senior author's track record, institution's track record and multi-centre collaboration, so those were left out: it earned 31 of the 62 points we could judge it on, which is 50 out of 100.
| Component | Points | Why |
|---|---|---|
| Journal tier | 2 / 10 | Tier 4 — not a peer-reviewed journal (Research Square) |
| Journal's citation record | 0 / 4 | Not a journal article, so the journal's record does not count |
| Study design | 7 / 12 | A longitudinal study following the same people over time |
| Strength of evidence | 3 / 5 | Rated “moderate” in our summary |
| Peer-review validation | 4 / 10 | Preprint — not yet peer-reviewed. Hosted on a preprint server. |
| How recent | 5 / 5 | 2026 — within 1 year(s) of the newest study here |
| Sample size | 4 / 4 | 75,000 participants |
| Effect reported | 2 / 3 | Effect reported as a number, without an interval or test |
| Studied in people | 3 / 3 | Studied in people (participant count stated) |
| Randomised / blinded | 0 / 2 | No randomisation or blinding stated — scores zero (not stated in the record — scored 0) |
| Times cited | — / 7 | No citation count (not found in OpenAlex) — left out of this score |
| Citations per year | — / 5 | No citation count or no year — left out of this score |
| Impact for its field and year | — / 6 | No field-weighted citation impact (OpenAlex has none yet) — left out of this score |
| Cited by studies here | — / 6 | Not yet checked against this library's citation graph — left out of this score |
| Still being cited | — / 4 | No yearly citation counts — left out of this score |
| Senior author's track record | — / 4 | Senior author's record not available — left out of this score |
| Institution's track record | — / 3 | No institution listed in OpenAlex — left out of this score |
| Multi-centre collaboration | — / 3 | Institutions not listed — left out of this score |
| Indexed in PubMed | 1 / 2 | Not in PubMed, but has a DOI |
| Free to read | 0 / 2 | Behind a paywall — scores zero |
| Points | 31 / 62 | Scored on 12 of 20 parts. We have no data for the rest, so they are left out: 31 of the 62 points we could judge it on is 50 out of 100. |
| Total | 50 / 100 | Capped to 16 because this is not a peer-reviewed article. |
What each of these 20 components measures, and what this score is not →
Read the original
Free copy unknown: no free-copy check has an answer for this record yet. That does not mean it is paywalled.
Cite this study APA, Vancouver, BibTeX · download .bib or .ris
Built from the record alzhub holds for this paper, not looked up. Volume, issue and page numbers are not held in the record. Check the citation against the original before you use it.
APA 7th
Ewing, M. A. (2026). Intelligence Breaks the System It Controls. Research Square. https://doi.org/10.21203/rs.3.rs-10283184/v2
Vancouver
Ewing MA. Intelligence Breaks the System It Controls. Research Square. 2026. doi:10.21203/rs.3.rs-10283184/v2.
BibTeX
@article{ewing2026-38225c61,
author = {Ewing, M. A.},
title = {{Intelligence Breaks the System It Controls}},
journal = {Research Square},
year = {2026},
doi = {10.21203/rs.3.rs-10283184/v2},
url = {https://doi.org/10.21203/rs.3.rs-10283184/v2}
}