Truace
Subscribe
The Index recomputed live from the record

What the evidence says.What the public feels.

The record holds 929 sourced gains and 770 sourced problems, averaging 68 and 66 on the index score. Readers have logged 22 public signals on the Pulse, which is kept apart and never counted as evidence.

929 gains
770 problems
Every sourced claim in the Index, one square each, shaded by the strength of its evidence. High Moderate Emerging

Ranks distinct AI gain and problem claims from the published record. Scores reward impact, independent source strength, scale, confidence, and recency.

1,699 results
Show filters and sorting

Download every matching row, not just this page:Export CSVExport JSON

AI gains · 929

25
HealthStableModerate evidence · 1 source

Interpretable random survival forest model predicted all-cause mortality in adults with unrepaired PAH-CHD and stratified survival in both Eisenmenger and non-Eisenmenger subgroups where the ESC model did not.

Researchers developed and internally validated an interpretable machine learning risk model for adults with unrepaired pulmonary arterial hypertension associated with congenital heart disease using data from 601 patients in a Chinese national prospective registry followed for a median 76 months. A random survival forest achieved a bootstrapping C-index of 0.773, and SHAP analysis highlighted predictors such as hemoglobin, BMI, systolic blood pressure, and diastolic pulmonary artery pressure to build new risk strata.

Impact 30%
49
Evidence 25%
95
Scale 20%
85
Confidence 15%
87
Recency 10%
95

Updated Sep 14, 2026 · TRV-2026-1083

78Index score
26
HealthStableModerate evidence · 1 source

AI systems are being developed to fuse facial expressions, voice, and physiological signals to objectively quantify pain and to automatically segment spine, nerves, and needle tips to improve identification accuracy.

A September 2026 review in Journal of Translational Medicine surveyed AI research in pain medicine across three areas: objective pain assessment by fusing facial expressions, voice and physiological signals, automated segmentation of spine, nerves and needle tips from medical images, and AI support for classification, treatment decisions and prognosis in conditions from osteoarthritis to cancer pain.

Impact 30%
49
Evidence 25%
95
Scale 20%
85
Confidence 15%
87
Recency 10%
95

Updated Sep 14, 2026 · TRV-2026-1080

78Index score
27
HealthStableModerate evidence · 1 source

Systematic review finds AI and deep learning models achieve high accuracy on cytology and histopathology images and offer gains in consistency, speed, cost-effectiveness, and reduced pathologist workload in breast cancer screening and diagnosis.

A systematic review published September 12, 2026 reviewed AI and machine learning techniques for breast cancer screening, diagnosis, classification and tumor marker scoring. It examined advanced deep learning models including ANNs, SVMs, CNNs and faster R-CNN applied to cytology, histopathology and combined imaging-pathology data.

Impact 30%
49
Evidence 25%
95
Scale 20%
85
Confidence 15%
87
Recency 10%
95

Updated Sep 14, 2026 · TRV-2026-1079

78Index score
28
EducationStableModerate evidence · 1 source

Among 503 Polish state university students, habit and performance expectancy increased behavioral intention to use ChatGPT, and behavioral intention increased actual use behavior in higher education.

By November 2023, researchers surveyed 503 Polish state university students to test an extended UTAUT2 model of ChatGPT acceptance. Using PLS-SEM, they found habit, performance expectancy, and hedonic motivation predicted behavioral intention, while behavioral intention, habit, and facilitating conditions predicted actual use behavior.

Impact 30%
49
Evidence 25%
95
Scale 20%
85
Confidence 15%
87
Recency 10%
94

Updated Sep 9, 2026 · TRV-2026-1038

78Index score

AI problems · 770

25
HealthStableModerate evidence · 1 source

Most models lacked external validation, only one was low risk of bias, publication bias was detected, and performance dropped in HIV-positive populations, leaving models not ready for routine clinical implementation.

By August 2026, a systematic review and meta-analysis of 34 studies evaluated AI and machine learning models to predict tuberculosis treatment failure. Nineteen studies with 100,790 participants were pooled, yielding an AUC of 0.836 with high heterogeneity, with tree-based and multimodal approaches common and most publications appearing after 2019.

Impact 30%
49
Evidence 25%
95
Scale 20%
85
Confidence 15%
87
Recency 10%
90

Updated Aug 18, 2026 · TRV-2026-0817

78Index score
26
HealthStableModerate evidence · 1 source

Only 59% of patients felt meaningfully involved in decisions about their own care, and clinicians reported implementation barriers including poor connectivity, time pressures and training burden.

Between January 2024 and January 2026, a health district in Australia implemented a digital wound model of care combining an AI-enabled app with a virtual command centre across four hospitals and five community health centres. A post-implementation evaluation surveyed and interviewed 94 patients, 75 frontline clinicians, 9 senior wound nurses and a product manager, reviewing governance minutes to assess acceptability and perceived benefit.

Impact 30%
63
Evidence 25%
95
Scale 20%
60
Confidence 15%
87
Recency 10%
92

Updated Sep 1, 2026 · TRV-2026-0949

77Index score
27
PolicyStableModerate evidence · 1 source

Stakeholders flag unresolved risks for AI in the medicine lifecycle around accuracy and reliability, data governance confidentiality and consent, and ethics fairness and bias prevention requiring further regulatory science research.

On 2026-08-13, a peer-reviewed article reported a European-wide survey to set regulatory science research priorities for AI use in the medicine lifecycle. Authors developed 28 research questions across seven domains and collected 273 responses from regulators, industry, patients and consumers, academics, and healthcare professionals, finding convergence in rankings across groups.

Impact 30%
49
Evidence 25%
95
Scale 20%
85
Confidence 15%
87
Recency 10%
89

Updated Aug 15, 2026 · TRV-2026-0772

77Index score
28
HealthStableModerate evidence · 1 source

Clinical translation of AI mortality prediction after road traffic crashes remains limited by insufficient external validation, inconsistent handling of class imbalance, and incomplete reporting of tuning and missing data strategies.

A systematic review published 7 August 2026 examined 18 retrospective studies from 2014-2025 that used AI or machine learning to predict death after road traffic crashes, drawing mostly on national or regional databases, hospital records, and police or insurance tabular data.

Impact 30%
49
Evidence 25%
95
Scale 20%
85
Confidence 15%
87
Recency 10%
88

Updated Aug 10, 2026 · TRV-2026-0729

77Index score

Recomputed live from the record · Oct 11, 2026, 2:32 AM