TL;DR
Turn quiet afternoons into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
Researchers at Rensselaer Polytechnic Institute developed Quest2Dx, an AI framework that analyzes health questionnaire responses for patterns associated with Parkinson’s disease. In a study published in Nature Aging, it performed better than comparison models and nearly matched its full-questionnaire results using a 30-question survey; clinical use and effects on patient outcomes remain unestablished.
Researchers at Rensselaer Polytechnic Institute have developed an AI framework called Quest2Dx to screen health questionnaires for patterns associated with Parkinson’s disease, with the aim of helping clinicians identify people who may need further testing. In a study published in Nature Aging, the team reported that the model outperformed the comparison approaches it evaluated; it is intended to flag potential cases, not diagnose Parkinson’s.
The team trained Quest2Dx using data from the Parkinson’s Progression Markers Initiative and evaluated it against a separate dataset from the Fox Insight Study. The questionnaires included one with 187 questions and another with more than 300. Many respondents skipped items, and the researchers said that missing answers and differences in question wording or order can make conventional machine-learning analysis less reliable.
Quest2Dx was designed to handle missing responses, work across questionnaires with different wording despite being trained on one, and identify which combinations of questions and answers were most informative for an individual. The researchers said it used a large language model to interpret the meaning and context of questions and responses. The paper reports that its performance exceeded earlier models, including an approach based on Llama 2.
The researchers also tested a shortened questionnaire containing the 30 questions ranked as most relevant by Quest2Dx. They reported that the model’s predictions using the shorter survey were nearly as accurate as those using the full questionnaires. The report does not establish that the tool has been deployed in clinics or show that its use leads to earlier treatment or better health outcomes.
A Shorter Route to Specialist Referral
Parkinson’s disease may be present for years before recognizable motor symptoms emerge, while people may not reach specialists until later in the disease course. A questionnaire-based screen could give primary-care clinicians another way to notice patterns that might otherwise be missed in long forms, including when patients leave questions unanswered. That could support decisions about referral for further assessment, rather than replace clinical judgment or diagnostic testing.
The 30-question result points to a possible practical advantage: a shorter survey may be easier for patients to complete and for clinics to use. But the study’s reported model performance is not the same as proof of benefit in everyday care. The key question for patients and health systems is whether testing the model in real clinical settings confirms that it can identify people who warrant assessment without generating excessive or misleading referrals.
Products Worth Considering
Fitbit Sense 2 Health and Fitness Smartwatch, Shadow Grey/Graphite
Withings Sleep Under-Mattress Sleep Tracking Pad
Blueair Blue Pure 211i Max Air Purifier for Large Rooms
As an affiliate, we earn on qualifying purchases.
Why Questionnaire Design Matters
Parkinson’s assessment has long involved observation of symptoms, particularly movement, alongside other clinical evaluation. The disease can also involve features that appear before prominent motor symptoms, such as changes in smell, sleep, handwriting, or mood. These features are not unique to Parkinson’s, and a single questionnaire response cannot establish the condition.
The RPI-led researchers argue that lengthy forms can make subtle combinations of responses harder to detect. They also say standard machine-learning methods may fill in missing answers through a process called imputation, potentially introducing errors, or may rely too heavily on a question’s position rather than its meaning. Quest2Dx was designed to address these challenges by modeling response meaning and relationships across questions. The study draws on two questionnaire datasets, but it does not show that every health system’s forms or patient population will produce similar results.
“We built this AI model to screen for potential cases earlier.”
— Pingkun Yan, RPI biomedical engineering department head and study leader
Clinical Performance Still Needs Testing
The study reports comparisons using existing questionnaire datasets, but the available material does not provide enough detail to assess how Quest2Dx performs across different patient groups, clinics, or rates of Parkinson’s in a screened population. It also does not establish a clinically useful threshold for deciding who should be referred, or quantify the risk of false positives and missed cases in routine practice.
Several questionnaire responses highlighted by the model—including sleep problems, anxiety, stress, and difficulty with everyday tasks—can occur in many conditions. The authors’ findings therefore describe patterns associated with Parkinson’s, not proof that a respondent has the disease. It remains unclear whether the shorter survey would retain its reported performance when administered prospectively in primary care, and whether referrals based on its results would improve the timing of diagnosis.
Validation Before Primary Care Use
The researchers say their findings could help design simplified questionnaires for Parkinson’s screening and support future studies of tools for primary-care settings. The next steps would need to establish how Quest2Dx performs when used prospectively with patients, how its results compare with clinical assessments, and whether it can be integrated into referral workflows. The source material does not specify a timeline for those studies or announce a clinical launch.
For now, Quest2Dx is a research framework with a proposed screening role. Any future implementation would need to show that it works reliably across populations and questionnaire formats, while clinicians remain responsible for evaluating symptoms and deciding whether diagnostic testing is appropriate.
Key Questions
Does Quest2Dx diagnose Parkinson’s disease?
No. The researchers describe it as a screening tool that may flag people for further testing. A questionnaire result does not confirm Parkinson’s.
What information does the AI analyze?
It analyzes responses to health questionnaires, including patterns across questions and answers. The model was designed to account for skipped responses and differences in question wording.
How did the researchers test the system?
The team trained the model using data from the Parkinson’s Progression Markers Initiative and evaluated it with a separate Fox Insight Study dataset. The paper also reports tests using a 30-question survey.
Could a person’s sleep or mood answers indicate Parkinson’s?
Such responses may contribute to patterns the model identifies, but sleep or mood changes can have many causes. The authors caution that no single answer is diagnostic.
Is Quest2Dx available in doctors’ offices now?
The source material describes a research study and future potential for primary-care use. It does not report that Quest2Dx has been deployed in clinics.
Source: news-medical.net
