{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,5]],"date-time":"2026-10-05T19:55:08Z","timestamp":1791230108010,"version":"4.1.0"},"reference-count":37,"publisher":"JMIR Publications Inc.","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["J Med Internet Res"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec sec-type=\"background\">\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Voice-based deep learning models for Parkinson disease (PD) and dementia screening report areas under the curve (AUCs) of 0.85\u20100.97, but rarely audit demographic confounding. Because speech changes substantially with age, case-control age imbalance alone can produce high classification performance independent of disease.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec sec-type=\"objective\">\n                    <jats:title>Objective<\/jats:title>\n                    <jats:p>We audited the Bridge to Artificial Intelligence (Bridge2AI) Voice Dataset v3.0.0 with three objectives: (1) quantify age and site confounding in voice screening for PD and dementia, (2) evaluate whether a disease-specific acoustic signal persists after demographic adjustment, and (3) propose minimum reporting standards.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec sec-type=\"methods\">\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We fine-tuned an audio spectrogram transformer (AST; 86.4 million parameters) using 5-fold participant-level cross-validation for PD (n=253) and dementia (n=221). Logistic regression on age and sex provided a demographic-only baseline under identical splits. Confounding was assessed by (1) restricting evaluation to ages 60\u201080 years, (2) restricting to US participants because all Canadian PD (n=62) and all Canadian dementia (n=70) participants were cases with no Canadian controls, (3) retraining AST from scratch on the age-restricted subgroup, (4) 1:1 nearest-neighbor propensity-score matching on age and sex in the US age-restricted PD subgroup, and (5) applying v3.0.0-trained models to v2.0.1 spectrograms of the same participants.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec sec-type=\"results\">\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      On the full cohort, age alone was statistically indistinguishable from the AST (PD: AUC 0.875 vs 0.843, DeLong\n                      <jats:italic>P<\/jats:italic>\n                      =.36; dementia: 0.905 vs 0.895,\n                      <jats:italic>P<\/jats:italic>\n                      =.75), indicating a demographic shortcut; because both cohorts share the same 148-person control group, this parallel pattern is one dataset-level artifact, not 2 independent confirmations. Within ages 60\u201080 years, the AST exceeded age-only for both conditions (PD: 0.787 vs 0.568, \u0394AUC +0.225, 95% CI +0.089 to +0.361;\n                      <jats:italic>P<\/jats:italic>\n                      =.001, n=129; dementia: 0.809 vs 0.598, \u0394AUC +0.215, 95% CI +0.056 to +0.375;\n                      <jats:italic>P<\/jats:italic>\n                      =.008, n=95), though the dementia result reflects residual site confounding. Three convergent estimates of PD-specific signal, post hoc age-restricted (0.787), de novo retrained (0.798), and propensity-matched (US, 28 pairs; 0.760;\n                      <jats:italic>P<\/jats:italic>\n                      =.02 vs age-only), all fell substantially below the 0.843 full-cohort AUC. Under the most conservative adjustment (US-only, ages 60\u201080 years), AST AUC was 0.726, which we consider the least confounded estimate. Cross-version preprocessing testing produced AUC 0.618, indicating brittle representations. Of 13 published voice-PD studies, only 7 reported case and control age distributions, and none reported a demographic-only baseline or an age-restricted evaluation.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec sec-type=\"conclusions\">\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>Demographic confounding dominates full-cohort voice-screening metrics on the Bridge2AI Voice Dataset for both PD and dementia. After age restriction, site control, and propensity matching, residual AST performance for PD (AUC 0.726\u20100.798) remained significantly above demographic baselines, supporting a disease-specific acoustic signal substantially weaker than full-cohort metrics suggest; for dementia, insufficient US cases precluded similar estimates. Voice biomarker studies should report case and control demographic distributions, demographic-only baselines under identical splits, and age-restricted performance alongside full-cohort metrics.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.2196\/95609","type":"journal-article","created":{"date-parts":[[2026,7,6]],"date-time":"2026-07-06T13:15:06Z","timestamp":1783343706000},"page":"e95609-e95609","source":"Crossref","is-referenced-by-count":1,"title":["Demographic Confounding in Voice-Based Parkinson Disease Screening: Methodological Analysis of the Bridge2AI Voice Dataset"],"prefix":"10.2196","volume":"28","author":[{"ORCID":"https:\/\/orcid.org\/0009-0003-8941-3396","authenticated-orcid":false,"given":"Shikhar","family":"Shukla","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-5917-4558","authenticated-orcid":false,"given":"Parvati","family":"Naliyatthaliyazchayil","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1097-316X","authenticated-orcid":false,"given":"Judy W","family":"Gichoya","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3625-534X","authenticated-orcid":false,"given":"Saptarshi","family":"Purkayastha","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1010","published-online":{"date-parts":[[2026,8,20]]},"reference":[{"issue":"11","key":"R1","doi-asserted-by":"publisher","DOI":"10.3390\/s25113424","article-title":"The human voice as a digital health solution leveraging artificial intelligence","volume":"25","author":"Muddaloor","journal-title":"Sensors (Basel)"},{"key":"R2","doi-asserted-by":"publisher","DOI":"10.1016\/j.artmed.2025.103109","article-title":"Voice analysis in Parkinson\u2019s disease - a systematic literature review","volume":"163","author":"Xavier","journal-title":"Artif Intell Med"},{"issue":"11","key":"R3","doi-asserted-by":"publisher","DOI":"10.3390\/bioengineering12111279","article-title":"Voice-based detection of Parkinson\u2019s disease using machine and deep learning approaches: a systematic review","volume":"12","author":"Sedigh Malekroodi","journal-title":"Bioengineering (Basel)"},{"key":"R4","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2015.01.040","article-title":"Sensory-motor networks involved in speech production and motor control: an fMRI study","volume":"109","author":"Behroozmand","journal-title":"Neuroimage"},{"issue":"5","key":"R5","doi-asserted-by":"publisher","first-page":"1167","DOI":"10.1044\/1092-4388(2010\/09-0154)","article-title":"Voice and fluency changes as a function of speech task and deep brain stimulation","volume":"53","author":"Van Lancker Sidtis","journal-title":"J Speech Lang Hear Res"},{"issue":"12","key":"R6","doi-asserted-by":"publisher","first-page":"5708","DOI":"10.1044\/2025_JSLHR-24-00268","article-title":"Natural speech analysis can reveal individual differences in executive function across the adult lifespan","volume":"68","author":"Wei","journal-title":"J Speech Lang Hear Res"},{"key":"R7","doi-asserted-by":"crossref","unstructured":"Gong Y Chung YA Glass J . AST: audio spectrogram transformer. Presented at: Interspeech 2021; Aug 30 to Sep 3, 2021:571-575; Brno, Czechia. [doi: 10.21437\/Interspeech.2021-698]","DOI":"10.21437\/Interspeech.2021-698"},{"key":"R8","doi-asserted-by":"publisher","DOI":"10.1016\/j.jvoice.2024.11.008","article-title":"Vocal biomarkers for Parkinson\u2019s disease classification using audio spectrogram transformers","author":"Madusanka","journal-title":"J Voice"},{"issue":"7","key":"R9","doi-asserted-by":"publisher","DOI":"10.3390\/bioengineering12070728","article-title":"Speech-based Parkinson\u2019s detection using pre-trained self-supervised automatic speech recognition (ASR) models and supervised contrastive learning","volume":"12","author":"Sedigh Malekroodi","journal-title":"Bioengineering (Basel)"},{"key":"R10","doi-asserted-by":"crossref","unstructured":"La Quatra M Turco MF Svendsen T Salvi G Orozco-Arroyave JR Siniscalchi SM . Exploiting foundation models and speech enhancement for parkinson\u2019s disease detection from speech in real-world operative conditions. Presented at: Interspeech 2024; Sep 1-5, 2024:1405-1409; Kos Island, Greece. [doi: 10.21437\/Interspeech.2024-522]","DOI":"10.21437\/Interspeech.2024-522"},{"key":"R11","doi-asserted-by":"publisher","DOI":"10.3389\/fneur.2025.1706317","article-title":"Parkinson\u2019s disease detection using spectrogram-based multi-model feature fusion networks","volume":"16","author":"Chen","journal-title":"Front Neurol"},{"key":"R12","doi-asserted-by":"publisher","DOI":"10.1016\/j.cmpbup.2022.100074","article-title":"Voice pathology detection using convolutional neural networks with electroglottographic (EGG) and speech signals","volume":"2","author":"Islam","journal-title":"Computer Methods and Programs in Biomedicine Update"},{"issue":"1","key":"R13","doi-asserted-by":"publisher","DOI":"10.1038\/s41531-025-00913-4","article-title":"Speech and language biomarkers for Parkinson\u2019s disease prediction, early diagnosis and progression","volume":"11","author":"Cao","journal-title":"NPJ Parkinsons Dis"},{"key":"R14","doi-asserted-by":"publisher","DOI":"10.4103\/abr.abr_254_21","article-title":"Evaluating the effect of Parkinson\u2019s disease on Jitter and Shimmer speech features","volume":"10","author":"Azadi","journal-title":"Adv Biomed Res"},{"key":"R15","doi-asserted-by":"publisher","DOI":"10.1016\/j.jvoice.2024.05.018","article-title":"A study on voice measures in patients with Parkinson\u2019s disease","author":"Xiu","journal-title":"J Voice"},{"key":"R16","doi-asserted-by":"publisher","DOI":"10.1136\/bmj-2023-078378","article-title":"TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods","volume":"385","author":"Collins","journal-title":"BMJ"},{"key":"R17","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijporl.2025.112455","article-title":"Developing age-specific protocols for pediatric voice databases in artificial intelligence research","volume":"196","author":"Russell","journal-title":"Int J Pediatr Otorhinolaryngol"},{"key":"R18","doi-asserted-by":"publisher","DOI":"10.3233\/SHTI230079","article-title":"Classification of Parkinson\u2019s disease from voice - analysis of data selection bias","volume":"302","author":"Brenner","journal-title":"Stud Health Technol Inform"},{"issue":"3","key":"R19","doi-asserted-by":"publisher","first-page":"991","DOI":"10.3390\/app12030991","article-title":"Things to consider when automatically detecting Parkinson\u2019s disease using the phonation of sustained vowels: analysis of methodological issues","volume":"12","author":"Ozbolt","journal-title":"Appl Sci"},{"issue":"10","key":"R20","doi-asserted-by":"publisher","DOI":"10.3390\/bioengineering12101052","article-title":"Voice-based early diagnosis of Parkinson\u2019s disease using spectrogram features and AI models","volume":"12","author":"Quamar","journal-title":"Bioengineering (Basel)"},{"issue":"10","key":"R21","doi-asserted-by":"publisher","DOI":"10.2196\/26305","article-title":"Detecting Parkinson disease using a web-based speech task: observational study","volume":"23","author":"Rahman","journal-title":"J Med Internet Res"},{"issue":"6","key":"R22","doi-asserted-by":"publisher","first-page":"4293","DOI":"10.1007\/s10278-024-01367-5","article-title":"Innovating challenges and experiences in Emory Health AI Bias Datathon: experience report","volume":"38","author":"Paddo","journal-title":"J Imaging Inform Med"},{"key":"R23","unstructured":"Datasets. Bridge2AI. URL: https:\/\/bridge2ai.org\/datasets\/ [Accessed 23-07-2026]"},{"key":"R24","unstructured":"Bensoussan Y Sigaras A Rameau A et al. Bridge2AI-voice: an ethically-sourced, diverse voice dataset linked to health information. PhysioNet. URL: https:\/\/physionet.org\/content\/b2ai-voice\/3.0.0\/ [Accessed 23-07-2026]"},{"key":"R25","doi-asserted-by":"crossref","unstructured":"Churchwell C Morrison M Pardo B . High-fidelity neural phonetic posteriorgrams. Presented at: 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW); Apr 14-19, 2024:823-827; Seoul, Korea, Republic of. [doi: 10.1109\/ICASSPW62465.2024.10669905]","DOI":"10.1109\/ICASSPW62465.2024.10669905"},{"key":"R26","unstructured":"Shukla S . Bridge2ai-voice-parkinsons-ast. GitHub. URL: https:\/\/github.com\/Amorfati123\/bridge2ai-voice-parkinsons-ast [Accessed 23-7-2026]"},{"key":"R27","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijmedinf.2023.105237","article-title":"On the inter-dataset generalization of machine learning approaches to Parkinson\u2019s disease detection from voice","volume":"179","author":"Hire\u0161","journal-title":"Int J Med Inform"},{"issue":"9","key":"R28","doi-asserted-by":"publisher","first-page":"842","DOI":"10.1016\/j.jacr.2023.06.025","article-title":"\u201cShortcuts\u201d causing bias in radiology artificial intelligence: causes, evaluation, and mitigation","volume":"20","author":"Banerjee","journal-title":"J Am Coll Radiol"},{"issue":"13","key":"R29","doi-asserted-by":"publisher","DOI":"10.1126\/sciadv.adq0305","article-title":"Demographic bias of expert-level vision-language foundation models in medical imaging","volume":"11","author":"Yang","journal-title":"Sci Adv"},{"key":"R30","doi-asserted-by":"crossref","unstructured":"Orozco-Arroyave JR Arias-Londo\u00f1o JD Vargas-Bonilla JF Gonz\u00e1lez-R\u00e1tiva MC N\u00f6th E . New spanish speech corpus database for the analysis of people suffering from parkinson\u2019s disease. In: Calzolari N Choukri K Declerck T , editors. Presented at: Ninth International Conference on Language Resources and Evaluation; May 26-31, 2014:342-347; Reykjavik, Iceland. 2014.URL: https:\/\/aclanthology.org\/L14-1549\/ [Accessed 17-03-2026]","DOI":"10.63317\/3c643i6hcfzo"},{"issue":"4","key":"R31","doi-asserted-by":"publisher","first-page":"828","DOI":"10.1109\/JBHI.2013.2245674","article-title":"Collection and analysis of a Parkinson speech dataset with multiple types of sound recordings","volume":"17","author":"Sakar","journal-title":"IEEE J Biomed Health Inform"},{"key":"R32","doi-asserted-by":"crossref","unstructured":"Wroge TJ Ozkanca Y Demiroglu C Si D Atkins DC Ghomi RH . Parkinson\u2019s disease diagnosis using machine learning and voice. Presented at: 2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB); Dec 1, 2018:1-7; Philadelphia, PA. 2018.[doi: 10.1109\/SPMB.2018.8615607]","DOI":"10.1109\/SPMB.2018.8615607"},{"key":"R33","doi-asserted-by":"publisher","DOI":"10.1016\/j.jbi.2019.103362","article-title":"Investigating voice as a biomarker: Deep phenotyping methods for early detection of Parkinson\u2019s disease","volume":"104","author":"Tracy","journal-title":"J Biomed Inform"},{"issue":"1","key":"R34","doi-asserted-by":"publisher","DOI":"10.1007\/s13755-021-00162-8","article-title":"An algorithm for Parkinson\u2019s disease speech classification based on isolated words analysis","volume":"9","author":"Amato","journal-title":"Health Inf Sci Syst"},{"issue":"14","key":"R35","doi-asserted-by":"publisher","DOI":"10.3390\/s24144625","article-title":"Exploring spectrogram-based audio classification for Parkinson\u2019s disease: a study on speech classification and qualitative reliability verification","volume":"24","author":"Jeong","journal-title":"Sensors (Basel)"},{"key":"R36","unstructured":"Piao R Lu Y Kemps H Xia T Saeed A . Unified acoustic representations for screening neurological and respiratory pathologies from voice. arXiv. Preprint posted online on Dec 19, 2025. [doi: 10.48550\/arXiv.2508.20717]"},{"key":"R37","doi-asserted-by":"crossref","unstructured":"Luz S Haider F Fuente S de la Fromm D MacWhinney B . Alzheimer\u2019s dementia recognition through spontaneous speech: the ADReSS challenge. Presented at: Interspeech 2020; Oct 25-29, 2020:2172-2176; Shanghai, China. [doi: 10.21437\/Interspeech.2020-2571]","DOI":"10.21437\/Interspeech.2020-2571"}],"container-title":["Journal of Medical Internet Research"],"original-title":[],"language":"en","deposited":{"date-parts":[[2026,10,5]],"date-time":"2026-10-05T19:24:25Z","timestamp":1791228265000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.jmir.org\/2026\/1\/e95609"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,20]]},"references-count":37,"URL":"https:\/\/doi.org\/10.2196\/95609","relation":{},"ISSN":["1438-8871"],"issn-type":[{"value":"1438-8871","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,20]]},"article-number":"v28i11e95609"}}