Truace
Subscribe
The Index recomputed live from the record

What the evidence says.What the public feels.

The record holds 934 sourced gains and 771 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.

934 gains
771 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.

1,705 results
Show filters and sorting

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

AI gains · 934

285
HealthNewly addedModerate evidence · 1 source

NLP model classified incident MRONJ and identified antiresorptive drug holidays from free-text EHR notes with high precision and recall in validation sets.

Investigators developed and validated a natural language processing algorithm to identify incident medication-related osteonecrosis of the jaw and antiresorptive drug holidays from free-text clinical notes, using Veterans Health Administration records for adults 50 and older prescribed bisphosphonates or denosumab for fracture prevention between 1999 and 2022.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
98

Updated Oct 4, 2026 · TRV-2026-1272

68Index score
286
HealthNewly addedModerate evidence · 1 source

Conversational agents powered by generative AI may help transform healthcare delivery by collecting information, answering questions, documenting encounters, and supporting clinical decision-making through fluid contextual dialogue.

Published September 30, 2025, the article describes unprecedented challenges in healthcare communication linked to administrative burden and reduced clinician time with patients, and discusses conversational agents powered by generative AI as a potential way to collect information, answer questions, document encounters, and support clinical decision-making through fluid dialogue.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
98

Updated Oct 3, 2026 · TRV-2026-1267

68Index score
287
HealthNewly addedModerate evidence · 1 source

In 1,838 stroke survivors, machine learning models incorporating the albumin-bilirubin score predicted all-cause mortality and enabled risk stratification for individualized management.

Researchers evaluated the albumin-bilirubin score as a predictor of all-cause mortality in 1,838 stroke survivors from NHANES 1999-2018, with 78 months median follow-up. They selected variables with Boruta, tested associations with Kaplan-Meier and Cox models, and built five machine learning models with 10-fold cross-validation assessed by C-index, Brier score, time-dependent AUC, and calibration, interpreting the best model with SHAP.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
98

Updated Oct 3, 2026 · TRV-2026-1266

68Index score
288
HealthNewly addedModerate evidence · 1 source

Patient-level 500-view convolutional neural network improved acute myeloid leukemia subtype classification, achieving 0.8783 accuracy and 0.8622 F1 score compared to single-view analysis.

By the publication date of 2026-10-01, researchers had developed and evaluated a person-based multi-view CNN approach for AML subtype classification using 81,214 single-cell images from 189 individuals. The 500-view MV-CNN achieved accuracy 0.8783, F1 0.8622, sensitivity 0.8637, specificity 0.8774, precision 0.8616 and MCC 0.8441, outperforming a single-view CNN baseline in tenfold cross-validation.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
98

Updated Oct 3, 2026 · TRV-2026-1264

68Index score

AI problems · 771

285
HealthStableModerate evidence · 1 source

The linear LDA model showed unstable performance and poor discrimination in lymph-node-negative and pancreatic head tumor subgroups despite overall validation AUC.

In a retrospective study of 340 pancreatic ductal adenocarcinoma patients, investigators developed machine learning models combining CT radiomics and clinical predictors to predict synchronous liver metastasis preoperatively. In an independent validation cohort of 102 patients, the best linear model (LDA) reached AUC 0.828 and the nonlinear model (MLP) reached AUC 0.822, both showing good calibration with Hosmer-Lemeshow P values of 0.551 and 0.682.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
91

Updated Aug 27, 2026 · TRV-2026-0907

68Index score
286
HealthStableModerate evidence · 1 source

Clinical translation is limited by inconsistent evaluation, substantially lower performance on pathology-specific regions, and reliance on single-institution data with few reader studies or external validation.

Researchers systematically reviewed 41 studies through January 2025 that used deep learning to generate synthetic postcontrast T1-weighted MRI from precontrast images alone, aiming to reduce gadolinium use. Most work was in neuroimaging, using GANs and CNNs, and a targeted meta-analysis of 15 brain tumor studies reported high whole-image similarity metrics.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
91

Updated Aug 27, 2026 · TRV-2026-0904

68Index score
287
HealthStableModerate evidence · 1 source

In extended testing, the three-marker panel misclassified one febrile control, and the overall pilot size limits generalizability for rapid diagnosis.

By the publication date of 2026-08-25, a pilot study had collected breath samples from 17 melioidosis patients and 8 febrile controls and used two-dimensional gas chromatography mass spectrometry combined with machine learning feature selection to identify volatile signatures of infection and treatment course.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
91

Updated Aug 26, 2026 · TRV-2026-0896

68Index score
288
HealthStableModerate evidence · 1 source

AI-driven medicinal chemistry is limited by data scarcity and inconsistency, model generalizability and interpretability issues, and a persistent gap between in silico predictions and validated candidates, with systematic evidence of faster or better drug delivery still accruing.

This peer-reviewed review from August 2026 examines how machine learning, deep learning, NLP, and generative modeling are being used across medicinal chemistry, including target discovery, virtual screening, property prediction, de novo design, fragment optimization, ADMET assessment, and clinical trial design, with emphasis on multimodal data fusion and human-AI collaboration.

Impact 30%
49
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
91

Updated Aug 26, 2026 · TRV-2026-0893

68Index score

Recomputed live from the record · Oct 12, 2026, 12:20 PM