Truace
Subscribe

Living evidence

Meta-Analysis

What the trials add up to, with every number traced to the paper it came from.

Truvace meta-analysesProtocol first, every number quoted, updated as trials appear

Protocol for a living systematic review and meta-analysis

AI chatbots and depressive symptoms: a living meta-analysis of randomized trials

In randomized trials, do people who are given a mental-health chatbot end up with fewer depressive symptoms than people who are not?

Status
Protocol published; results in checking
Protocol written
10 November 2026
Protocol
Version 1, f72a6565f8b3

Read the protocol

Search of 11 November 2026One square is one record

  1. 583Records found
  2. 358Screened on title and abstract
  3. 57Read in full
  4. Included trialsnumbers being checked against the papers

The pooled result and its forest plot appear here once an editor has checked every included trial’s numbers against the papers. Searched: PubMed, Europe PMC, OpenAlex.

Pooled results from published reviews

The reviews’ own figures, checked word for word by code. Truvace did not recalculate them.

59 pooled results from 19 reviews

  1. In a 110-study review, AI beat clinicians on urological cancer scans but lagged on prostate

    TRV-2026-06406 pooled results

    • SpecificitySpecificity of AI models in radiologic diagnosis of urological cancers0.83 95% CI 0.80 to 0.86110 studies
      AI models achieved pooled sensitivity and specificity of 0.85 (95% CI: 0.83-0.87) and 0.83 (95% CI: 0.80-0.86), with an AUC of 0.91 (95% CI: 0.88-0.93)

      The review’s own sentence. Open the review’s record (TRV-2026-0640)

    • SensitivitySensitivity of clinicians in radiologic diagnosis of urological cancers0.82 95% CI 0.79 to 0.85110 studies
      Clinicians demonstrated a pooled sensitivity of 0.82 (95% CI: 0.79-0.85) and specificity of 0.68 (95% CI: 0.62-0.73), with an AUC of 0.83 (95% CI: 0.80-0.86).

      The review’s own sentence. Open the review’s record (TRV-2026-0640)

    • AUCOverall diagnostic accuracy (area under the curve) of AI models for urological cancers0.91 95% CI 0.88 to 0.93110 studies
      AI models achieved pooled sensitivity and specificity of 0.85 (95% CI: 0.83-0.87) and 0.83 (95% CI: 0.80-0.86), with an AUC of 0.91 (95% CI: 0.88-0.93)

      The review’s own sentence. Open the review’s record (TRV-2026-0640)

    • SpecificitySpecificity of clinicians in radiologic diagnosis of urological cancers0.68 95% CI 0.62 to 0.73110 studies
      Clinicians demonstrated a pooled sensitivity of 0.82 (95% CI: 0.79-0.85) and specificity of 0.68 (95% CI: 0.62-0.73), with an AUC of 0.83 (95% CI: 0.80-0.86).

      The review’s own sentence. Open the review’s record (TRV-2026-0640)

    • SensitivitySensitivity of AI models in radiologic diagnosis of urological cancers0.85 95% CI 0.83 to 0.87110 studies
      AI models achieved pooled sensitivity and specificity of 0.85 (95% CI: 0.83-0.87) and 0.83 (95% CI: 0.80-0.86), with an AUC of 0.91 (95% CI: 0.88-0.93)

      The review’s own sentence. Open the review’s record (TRV-2026-0640)

    • AUCOverall diagnostic accuracy (area under the curve) of clinicians for urological cancers0.83 95% CI 0.80 to 0.86110 studies
      Clinicians demonstrated a pooled sensitivity of 0.82 (95% CI: 0.79-0.85) and specificity of 0.68 (95% CI: 0.62-0.73), with an AUC of 0.83 (95% CI: 0.80-0.86).

      The review’s own sentence. Open the review’s record (TRV-2026-0640)

  2. AI predicts gallbladder surgery difficulty well, but 16 of 18 studies were at high risk of bias

    TRV-2026-07872 pooled results

    • AUCPooled AUC of AI validation models for predicting difficulty of laparoscopic cholecystectomy0.818 95% CI 0.797 to 0.84018 studies
      For 32 validation models, the pooled AUC was 0.818 (95% CI, 0.797-0.840).

      The review’s own sentence. Open the review’s record (TRV-2026-0787)

    • AUCPooled AUC of AI training models for predicting difficulty of laparoscopic cholecystectomy0.848 95% CI 0.829 to 0.86818 studies
      The pooled AUC for 27 training models was 0.848 (95% CI, 0.829-0.868).

      The review’s own sentence. Open the review’s record (TRV-2026-0787)

  3. AI help raised fracture detection to 87% from 73%; junior doctors gained most, a review shows

    TRV-2026-11454 pooled results

    • SensitivitySensitivity of unassisted physicians in detecting bone fractures73% 95% CI 69% to 78%17 studies
      Compared to unassisted diagnosis, AI significantly improved pooled sensitivity (87%, 95% confidence interval [CI]: 84-89%, versus 73%, 95% CI: 69-78%) and maintained high specificity (95%, 95% CI: 92-97%, versus 94%, 95% CI: 89-96%).

      The review’s own sentence. Open the review’s record (TRV-2026-1145)

    • SpecificitySpecificity of AI-assisted physicians in detecting bone fractures95% 95% CI 92% to 97%17 studies
      Compared to unassisted diagnosis, AI significantly improved pooled sensitivity (87%, 95% confidence interval [CI]: 84-89%, versus 73%, 95% CI: 69-78%) and maintained high specificity (95%, 95% CI: 92-97%, versus 94%, 95% CI: 89-96%).

      The review’s own sentence. Open the review’s record (TRV-2026-1145)

    • SpecificitySpecificity of unassisted physicians in detecting bone fractures94% 95% CI 89% to 96%17 studies
      Compared to unassisted diagnosis, AI significantly improved pooled sensitivity (87%, 95% confidence interval [CI]: 84-89%, versus 73%, 95% CI: 69-78%) and maintained high specificity (95%, 95% CI: 92-97%, versus 94%, 95% CI: 89-96%).

      The review’s own sentence. Open the review’s record (TRV-2026-1145)

    • SensitivitySensitivity of AI-assisted physicians in detecting bone fractures87% 95% CI 84% to 89%17 studies
      Compared to unassisted diagnosis, AI significantly improved pooled sensitivity (87%, 95% confidence interval [CI]: 84-89%, versus 73%, 95% CI: 69-78%) and maintained high specificity (95%, 95% CI: 92-97%, versus 94%, 95% CI: 89-96%).

      The review’s own sentence. Open the review’s record (TRV-2026-1145)

  4. Machine learning's edge over standard statistics in predicting trauma deaths is small and may be inflated

    TRV-2026-10682 pooled results

    • AUC differenceDifference in AUC between best machine learning model and logistic regression for trauma mortality prediction0.026 95% CI 0.009 to 0.04317 studies, 243,324 people
      The pooled AUC difference favoring ML was 0.026 (95% CI 0.009-0.043)

      The review’s own sentence. Open the review’s record (TRV-2026-1068)

    • AUC differenceAUC difference between machine learning and logistic regression in studies reporting confidence intervals for both models0.017 95% CI 0.005 to 0.029
      the co-primary estimate was 0.017 (95% CI 0.005-0.029)

      The review’s own sentence. Open the review’s record (TRV-2026-1068)

  5. AI-made MRI scans match real ones overall but lag on diseased areas, a 41-study review found

    TRV-2026-09042 pooled results

    • SSIMImage similarity (SSIM) of synthesized postcontrast T1-weighted MRI in brain tumor studies0.92 95% CI 0.90 to 0.9315 studies
      Meta-analysis of 15 brain tumor studies (30 models) yielded pooled SSIM of 0.92 (95% CI, 0.90-0.93) and pooled PSNR of 30.6 dB (95% CI, 28.6-32.6).

      The review’s own sentence. Open the review’s record (TRV-2026-0904)

    • PSNRPeak signal-to-noise ratio (PSNR) of synthesized postcontrast T1-weighted MRI in brain tumor studies30.6 95% CI 28.6 to 32.615 studies
      Meta-analysis of 15 brain tumor studies (30 models) yielded pooled SSIM of 0.92 (95% CI, 0.90-0.93) and pooled PSNR of 30.6 dB (95% CI, 28.6-32.6).

      The review’s own sentence. Open the review’s record (TRV-2026-0904)

  6. Only 1 of 20 studies of AI fertility predictions had low risk of bias, a review shows

    TRV-2026-06743 pooled results

    • SensitivityPooled sensitivity of machine learning models for predicting pregnancy outcomes after assisted reproduction0.737 95% CI 0.662 to 0.79914 studies
      The pooled sensitivity was 0.737 (95% CI, 0.662-0.799) and the pooled specificity was 0.789 (95% CI, 0.709-0.851), with substantial heterogeneity (I2 = 97.7% and 99.0%, respectively).

      The review’s own sentence. Open the review’s record (TRV-2026-0674)

    • SpecificityPooled specificity of machine learning models for predicting pregnancy outcomes after assisted reproduction0.789 95% CI 0.709 to 0.85114 studies
      The pooled sensitivity was 0.737 (95% CI, 0.662-0.799) and the pooled specificity was 0.789 (95% CI, 0.709-0.851), with substantial heterogeneity (I2 = 97.7% and 99.0%, respectively).

      The review’s own sentence. Open the review’s record (TRV-2026-0674)

    • diagnostic odds ratioDiagnostic odds ratio of machine learning models for predicting pregnancy outcomes after assisted reproduction10.49 95% CI 6.28 to 17.5314 studies
      The pooled DOR was 10.49 (95% CI, 6.28-17.53), and the SROC curve indicated acceptable overall discrimination.

      The review’s own sentence. Open the review’s record (TRV-2026-0674)

  7. AI cut screening time over 95% in a brittle bone review, but showed bias in rating trials

    TRV-2026-03212 pooled results

    • percentage change in aBMD12-month change in lumbar spine bone mineral density with setrusumab in adults with osteogenesis imperfecta9.38% 95% CI 6.5% to 12.26%10 studies, 333 people
      In adults, setrusumab at 12 months yielded the highest improvement (9.38%, 95% CI 6.5%-12.26%).

      The review’s own sentence. Open the review’s record (TRV-2026-0321)

    • percentage change in aBMD12-month change in lumbar spine bone mineral density with denosumab in children with osteogenesis imperfecta25.49% 95% CI 17.14% to 33.84%10 studies, 333 people
      In children, denosumab produced the greatest 12-month increase in lumbar spine aBMD (25.49%, 95% CI 17.14%-33.84%).

      The review’s own sentence. Open the review’s record (TRV-2026-0321)

  8. For narrowed bile ducts, a review of five studies found AI hit 95% sensitivity

    TRV-2026-06796 pooled results

    • Accuracyaccuracy (SROC)Diagnostic accuracy (SROC) of AI-assisted cholangioscopy for biliary strictures97% 95% CI 95% to 98%
      The pooled sensitivity and specificity were 95% (95% CI: 85-98) and 88% (95% CI: 76-94), with a diagnostic accuracy (SROC) of 97% (95% CI: 95-98).

      The review’s own sentence. Open the review’s record (TRV-2026-0679)

    • Accuracyaccuracy (SROC)Diagnostic accuracy (SROC) of neural network cholangioscopy AI, neural network studies only97% 95% CI 95% to 98%4 studies, 538 people
      Sensitivity analysis of CNN studies (4 studies, 538 patients) demonstrated a pooled sensitivity, specificity, and accuracy (SROC) of 95% (95% CI: 82-99), 88% (95% CI: 72-95), and 97% (95% CI: 95-98), respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-0679)

    • SpecificitySpecificity of AI-assisted cholangioscopy for indeterminate or malignant biliary strictures88% 95% CI 76% to 94%
      The pooled sensitivity and specificity were 95% (95% CI: 85-98) and 88% (95% CI: 76-94), with a diagnostic accuracy (SROC) of 97% (95% CI: 95-98).

      The review’s own sentence. Open the review’s record (TRV-2026-0679)

    • SensitivitySensitivity of neural network cholangioscopy AI for biliary strictures, neural network studies only95% 95% CI 82% to 99%4 studies, 538 people
      Sensitivity analysis of CNN studies (4 studies, 538 patients) demonstrated a pooled sensitivity, specificity, and accuracy (SROC) of 95% (95% CI: 82-99), 88% (95% CI: 72-95), and 97% (95% CI: 95-98), respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-0679)

    • SensitivitySensitivity of AI-assisted cholangioscopy for indeterminate or malignant biliary strictures95% 95% CI 85% to 98%
      The pooled sensitivity and specificity were 95% (95% CI: 85-98) and 88% (95% CI: 76-94), with a diagnostic accuracy (SROC) of 97% (95% CI: 95-98).

      The review’s own sentence. Open the review’s record (TRV-2026-0679)

    • SpecificitySpecificity of neural network cholangioscopy AI for biliary strictures, neural network studies only88% 95% CI 72% to 95%4 studies, 538 people
      Sensitivity analysis of CNN studies (4 studies, 538 patients) demonstrated a pooled sensitivity, specificity, and accuracy (SROC) of 95% (95% CI: 82-99), 88% (95% CI: 72-95), and 97% (95% CI: 95-98), respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-0679)

  9. In 12 studies, AI-read ECGs spot weak heart pumping in people without symptoms, a review finds

    TRV-2026-18377 pooled results

    • AUCAI-ECG discrimination of left ventricular ejection fraction 35% or lower in symptomatic people0.89 95% CI 0.86 to 0.92
      Symptomatic cohorts showed AUROCs of 0.89 (95% CI 0.86-0.92), 0.90 (95% CI 0.88-0.92), and 0.85 (95% CI 0.83-0.86) respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-1837)

    • AUCAI-ECG discrimination of left ventricular ejection fraction 50% or lower in asymptomatic people0.89 95% CI 0.85 to 0.97
      In asymptomatic cohorts evaluating left ventricular systolic dysfunction (LVSD), AI-ECG demonstrated promising but heterogeneous discriminatory performance (I²>85%) with pooled AUROCs of 0.87 (95% CI 0.75-0.99) for LVEF ≤35%, 0.93 (95% CI 0.88-0.97) for LVEF ≤40%, and 0.89 (95% CI 0.85-0.97) for LVEF ≤50%.

      The review’s own sentence. Open the review’s record (TRV-2026-1837)

    • AUCAI-ECG discrimination of left ventricular ejection fraction 40% or lower in asymptomatic people0.93 95% CI 0.88 to 0.97
      In asymptomatic cohorts evaluating left ventricular systolic dysfunction (LVSD), AI-ECG demonstrated promising but heterogeneous discriminatory performance (I²>85%) with pooled AUROCs of 0.87 (95% CI 0.75-0.99) for LVEF ≤35%, 0.93 (95% CI 0.88-0.97) for LVEF ≤40%, and 0.89 (95% CI 0.85-0.97) for LVEF ≤50%.

      The review’s own sentence. Open the review’s record (TRV-2026-1837)

    • AUCAI-ECG discrimination of left ventricular ejection fraction 40% or lower in symptomatic people0.90 95% CI 0.88 to 0.92
      Symptomatic cohorts showed AUROCs of 0.89 (95% CI 0.86-0.92), 0.90 (95% CI 0.88-0.92), and 0.85 (95% CI 0.83-0.86) respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-1837)

    • AUCAI-ECG discrimination of left ventricular diastolic dysfunction in asymptomatic people0.76 95% CI 0.66 to 0.86
      Asymptomatic left ventricular diastolic dysfunction (LVDD) demonstrated moderate performance (AUROC 0.76, 95% CI 0.66-0.86) with no externally validated symptomatic LVDD studies identified.

      The review’s own sentence. Open the review’s record (TRV-2026-1837)

    • AUCAI-ECG discrimination of left ventricular ejection fraction 50% or lower in symptomatic people0.85 95% CI 0.83 to 0.86
      Symptomatic cohorts showed AUROCs of 0.89 (95% CI 0.86-0.92), 0.90 (95% CI 0.88-0.92), and 0.85 (95% CI 0.83-0.86) respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-1837)

    • AUCAI-ECG discrimination of left ventricular ejection fraction 35% or lower in asymptomatic people0.87 95% CI 0.75 to 0.99
      In asymptomatic cohorts evaluating left ventricular systolic dysfunction (LVSD), AI-ECG demonstrated promising but heterogeneous discriminatory performance (I²>85%) with pooled AUROCs of 0.87 (95% CI 0.75-0.99) for LVEF ≤35%, 0.93 (95% CI 0.88-0.97) for LVEF ≤40%, and 0.89 (95% CI 0.85-0.97) for LVEF ≤50%.

      The review’s own sentence. Open the review’s record (TRV-2026-1837)

  10. AI scan tools for glioblastoma show useful but imperfect accuracy in outside tests, with AUC 0.81

    TRV-2026-20754 pooled results

    • SensitivitySensitivity of MRI-based machine learning for distinguishing true progression from pseudoprogression in glioblastoma (external test cohorts)0.78 95% CI 0.63 to 0.88219 people
      Pooled sensitivity was 0.78 (95% confidence interval [CI], 0.63-0.88), specificity was 0.75 (95% CI, 0.63-0.84), and summary area under the receiver operating characteristic curve (AUC) was 0.81 (bootstrap 95% CI, 0.69-0.91).

      The review’s own sentence. Open the review’s record (TRV-2026-2075)

    • AUCSummary AUC of MRI-based machine learning for distinguishing true progression from pseudoprogression in glioblastoma (external test cohorts)0.81 95% CI 0.69 to 0.91219 people
      Pooled sensitivity was 0.78 (95% confidence interval [CI], 0.63-0.88), specificity was 0.75 (95% CI, 0.63-0.84), and summary area under the receiver operating characteristic curve (AUC) was 0.81 (bootstrap 95% CI, 0.69-0.91).

      The review’s own sentence. Open the review’s record (TRV-2026-2075)

    • SpecificitySpecificity of MRI-based machine learning for distinguishing true progression from pseudoprogression in glioblastoma (external test cohorts)0.75 95% CI 0.63 to 0.84219 people
      Pooled sensitivity was 0.78 (95% confidence interval [CI], 0.63-0.88), specificity was 0.75 (95% CI, 0.63-0.84), and summary area under the receiver operating characteristic curve (AUC) was 0.81 (bootstrap 95% CI, 0.69-0.91).

      The review’s own sentence. Open the review’s record (TRV-2026-2075)

    • AUC differenceDifference in AUC between internal and external test cohorts of MRI-based machine learning-0.09 95% CI -0.22 to 0.02
      Internal test cohorts showed higher apparent performance than external cohorts (AUC, 0.91 vs 0.81), but the difference was not statistically significant (difference, -0.09; 95% CI, -0.22 to 0.02; P = 0.12).

      The review’s own sentence. Open the review’s record (TRV-2026-2075)

  11. AI on heart ultrasounds spots a hard-to-diagnose heart failure, but evidence is thin

    TRV-2026-23114 pooled results

    • AUCarea under the summary ROC curveArea under summary ROC curve for AI echocardiography detecting HFpEF or elevated filling pressure0.875 95% CI 0.791 to 0.9302,770 people
      Pooled sensitivity was 84.0% (95% CI: 73.0-91.0%) and specificity 83.0% (95% CI: 73.0-89.0%), with an area under the summary ROC curve of 0.875 (95% CI: 0.791-0.930).

      The review’s own sentence. Open the review’s record (TRV-2026-2311)

    • diagnostic odds ratioDiagnostic odds ratio of AI echocardiography for detecting HFpEF or elevated filling pressure24.73 95% CI 18.46 to 33.122,770 people
      The diagnostic odds ratio was 24.73 (95% CI: 18.46-33.12)

      The review’s own sentence. Open the review’s record (TRV-2026-2311)

    • SensitivitySensitivity of AI-enabled echocardiography for detecting HFpEF, elevated filling pressure or diastolic dysfunction84.0% 95% CI 73.0% to 91.0%2,770 people
      Pooled sensitivity was 84.0% (95% CI: 73.0-91.0%) and specificity 83.0% (95% CI: 73.0-89.0%)

      The review’s own sentence. Open the review’s record (TRV-2026-2311)

    • SpecificitySpecificity of AI-enabled echocardiography for detecting HFpEF, elevated filling pressure or diastolic dysfunction83.0% 95% CI 73.0% to 89.0%2,770 people
      Pooled sensitivity was 84.0% (95% CI: 73.0-91.0%) and specificity 83.0% (95% CI: 73.0-89.0%)

      The review’s own sentence. Open the review’s record (TRV-2026-2311)

  12. A review pooling nine studies shows AI detects eye toxoplasmosis in photos with 0.965 sensitivity

    TRV-2026-17703 pooled results

    • SpecificitySpecificity of deep learning in detecting ocular toxoplasmosis from fundus photographs0.955 95% CI 0.929 to 0.972
      Pooled sensitivity was 0.965 (95% CI: 0.944-0.978) and pooled specificity was 0.955 (95% CI: 0.929-0.972).

      The review’s own sentence. Open the review’s record (TRV-2026-1770)

    • SensitivitySensitivity of deep learning in detecting ocular toxoplasmosis from fundus photographs0.965 95% CI 0.944 to 0.978
      Pooled sensitivity was 0.965 (95% CI: 0.944-0.978) and pooled specificity was 0.955 (95% CI: 0.929-0.972).

      The review’s own sentence. Open the review’s record (TRV-2026-1770)

    • diagnostic odds ratioDiagnostic odds ratio of deep learning for ocular toxoplasmosis on fundus photographs594.1 95% CI 263.3 to 1340.4
      The diagnostic odds ratio reached 594.1 (95% CI: 263.3-1340.4), with positive and negative likelihood ratios of 21.6 and 0.036.

      The review’s own sentence. Open the review’s record (TRV-2026-1770)

  13. AI predicts brain bleed outcomes well in 83 studies, but mostly in internal tests

    TRV-2026-06113 pooled results

    • AUCC-indexDiscrimination of clinical-radiomics models for predicting poor functional outcome in intracerebral hemorrhage0.850 95% CI 0.830 to 0.869
      models integrating both clinical and radiomics features achieved the highest discriminative performance across key prognostic prediction tasks: predicting HE (pooled C-index 0.822, 95% confidence interval [CI] 0.789-0.855), poor functional outcome (C-index 0.850, 95% CI 0.830-0.869), and mortality (C-index 0.860, 95% CI 0.809-0.911)

      The review’s own sentence. Open the review’s record (TRV-2026-0611)

    • AUCC-indexDiscrimination of clinical-radiomics models for predicting mortality in intracerebral hemorrhage0.860 95% CI 0.809 to 0.911
      models integrating both clinical and radiomics features achieved the highest discriminative performance across key prognostic prediction tasks: predicting HE (pooled C-index 0.822, 95% confidence interval [CI] 0.789-0.855), poor functional outcome (C-index 0.850, 95% CI 0.830-0.869), and mortality (C-index 0.860, 95% CI 0.809-0.911)

      The review’s own sentence. Open the review’s record (TRV-2026-0611)

    • AUCC-indexDiscrimination of clinical-radiomics models for predicting hematoma expansion in intracerebral hemorrhage0.822 95% CI 0.789 to 0.855
      models integrating both clinical and radiomics features achieved the highest discriminative performance across key prognostic prediction tasks: predicting HE (pooled C-index 0.822, 95% confidence interval [CI] 0.789-0.855), poor functional outcome (C-index 0.850, 95% CI 0.830-0.869), and mortality (C-index 0.860, 95% CI 0.809-0.911)

      The review’s own sentence. Open the review’s record (TRV-2026-0611)

  14. Generative AI was 13.0 points less accurate than physicians at radiology diagnosis, a review of 48 studies found

    TRV-2026-12843 pooled results

    • AccuracyPooled diagnostic accuracy of generative AI on free-text radiology diagnostic tasks42.9% 95% CI 35.8% to 50.1%
      Pooled diagnostic accuracy of generative AI was 42.9% (95% CI, 35.8-50.1%) for free-text tasks and 58.1% (95% CI, 50.0-66.2%) for choice tasks.

      The review’s own sentence. Open the review’s record (TRV-2026-1284)

    • difference in accuracy (percentage points)Difference in accuracy of generative AI between image-only and text-only input-25.9 95% CI -39.1 to -12.8
      Text-only input was associated with higher accuracy than image-only input (difference in accuracy [image-only minus text-only], -25.9 percentage points [95% CI, -39.1 to -12.8]; P = .001)

      The review’s own sentence. Open the review’s record (TRV-2026-1284)

    • AccuracyPooled diagnostic accuracy of generative AI on choice-based radiology diagnostic tasks58.1% 95% CI 50.0% to 66.2%
      Pooled diagnostic accuracy of generative AI was 42.9% (95% CI, 35.8-50.1%) for free-text tasks and 58.1% (95% CI, 50.0-66.2%) for choice tasks.

      The review’s own sentence. Open the review’s record (TRV-2026-1284)

  15. Review shows AI spots hip bone death at 88.8% sensitivity; only five tests used outside data

    TRV-2026-12802 pooled results

    • SensitivitySensitivity of imaging AI models in detecting avascular necrosis of the femoral head0.888 95% CI 0.789 to 0.944
      Pooled sensitivity was 0.888 (95% CI, 0.789-0.944), specificity was 0.937 (95% CI, 0.906-0.958), and SROC AUC was 0.967.

      The review’s own sentence. Open the review’s record (TRV-2026-1280)

    • SpecificitySpecificity of imaging AI models in detecting avascular necrosis of the femoral head0.937 95% CI 0.906 to 0.958
      Pooled sensitivity was 0.888 (95% CI, 0.789-0.944), specificity was 0.937 (95% CI, 0.906-0.958), and SROC AUC was 0.967.

      The review’s own sentence. Open the review’s record (TRV-2026-1280)

  16. AI hit 91% sensitivity on children's ear infections; a nine-study review shows it matched GPs

    TRV-2026-21822 pooled results

    • SensitivitySensitivity of AI systems in detecting acute otitis media91% 95% CI 77% to 97%
      The pooled sensitivity and specificity were 91% (95% confidence interval [CI]: 77%–97%) and 99% (95% CI: 94%–100%), respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-2182)

    • SpecificitySpecificity of AI systems in detecting acute otitis media99% 95% CI 94% to 100%
      The pooled sensitivity and specificity were 91% (95% confidence interval [CI]: 77%–97%) and 99% (95% CI: 94%–100%), respectively.

      The review’s own sentence. Open the review’s record (TRV-2026-2182)

  17. AI predicts death from heart valve infection well, but half the studies had bias risk

    TRV-2026-10492 pooled results

    • AUCDiscrimination of machine learning models for in-hospital or early mortality in infective endocarditis0.85 95% CI 0.81 to 0.89
      Pooled AUC was 0.85 (95% confidence interval [CI], 0.81-0.89) for in-hospital/early mortality (I2 = 35.3%) and 0.85 (95% CI, 0.82-0.88) for 6-month mortality (I2 = 0%)

      The review’s own sentence. Open the review’s record (TRV-2026-1049)

    • AUCDiscrimination of machine learning models for 6-month mortality in infective endocarditis0.85 95% CI 0.82 to 0.88
      Pooled AUC was 0.85 (95% confidence interval [CI], 0.81-0.89) for in-hospital/early mortality (I2 = 35.3%) and 0.85 (95% CI, 0.82-0.88) for 6-month mortality (I2 = 0%)

      The review’s own sentence. Open the review’s record (TRV-2026-1049)

  18. Surgical AI built on a US registry scored 0.71 vs 0.80 for other data in a 295-model review

    TRV-2026-14891 pooled result

    • adjusted regression coefficient (β) for AUROCAdjusted difference in AUROC of surgical prediction models, NSQIP vs non-NSQIP data sources-0.056 95% CI -0.105 to -0.007
      NSQIP remained associated with lower AUROC (β=-0.056, 95% CI -0.105 to -0.007; p=0.026)

      The review’s own sentence. Open the review’s record (TRV-2026-1489)

  19. AI models flag TB treatment failure, but only 23.5% of 34 studies tested them externally

    TRV-2026-08171 pooled result

    • AUCDiscrimination of AI/ML models predicting tuberculosis treatment failure0.836 95% CI 0.799 to 0.868100,790 people
      The pooled AUC was 0.836 (95% CI 0.799-0.868), with substantial heterogeneity (I² = 97.9%).

      The review’s own sentence. Open the review’s record (TRV-2026-0817)

  • sensitivity
  • specificity
  • AUC
  • accuracy
  • interval as printed

The short labels are an AI model’s reading of each review; open a row to read the review’s own sentence. Marks place sensitivity, specificity, AUC and accuracy on one 0 to 1 scale. Other measures have no common scale and show the figure only. A pooled figure is only as strong as the studies behind it.