Conectar
Artificial intelligence (AI)-driven chatbots have been rapidly adopted across research, education, business, marketing and medicine. Most interactions, however, come from non-experts using chatbots like search engines, including for everyday health and medical queries.
We conducted an original study to audit chatbot responses in health and medical fields prone to misinformation.
Five popular chatbots were assessed: Gemini (Google), DeepSeek (High-Flyer), Meta AI (Meta), ChatGPT (OpenAI) and Grok (xAI). In February 2025, each chatbot was prompted with 10 questions from five categories: cancer, vaccines, stem cells, nutrition and athletic performance. We deployed an adversarial-like framework, using open- and closed-ended prompts designed to strain models toward misinformation or contraindicated advice. Two experts from each category rated responses as ‘non-problematic’, ‘somewhat problematic’ or ‘highly problematic’ using a coding matrix based on objective, predefined criteria. Citations were scored for accuracy and completeness, and each response was given a Flesch Reading Ease score.
Nearly half (49.6%) of responses were problematic: 30% somewhat problematic and 19.6% highly problematic. Response quality did not differ significantly among chatbots (p=0.566) but Grok generated significantly more highly problematic responses than would be expected under a random distribution (z-score +2.07, p=0.038). Performance was strongest in vaccines (mean z-score –2.57) and cancer (–2.12), and weakest in stem cells (+1.25), athletic performance (+3.74) and nutrition (+4.35). Chatbot outputs were consistently expressed with confidence and certainty; from 250 total questions, there were only two refusals to answer (0.8%), both from Meta AI. Reference quality was poor, with a median completeness score of 40% (Q1–Q3: 20–67%). Chatbot hallucinations and fabricated citations precluded any chatbot from producing a fully accurate reference list. All readability scores were graded as ‘Difficult’ (30–50), equivalent to college sophomore–senior level.
The audited chatbots performed poorly when answering questions in misinformation-prone health and medical fields. Continued deployment without public education and oversight risks amplifying misinformation.
Unplanned hospital readmissions within 30 days of discharge measure the quality of healthcare. This study aims to identify the characteristics of patients at higher risk of readmission.
Retrospective cohort study.
North-east Italy (Marca Trevigiana Local Health Authority).
The study examined a total of 39 467 index admissions from hospital discharges (SDO) in the 890 000-inhabitant area during 2022.
Readmission rates and 95% CIs were computed by risk factor, age and type of admission (surgical or medical). A logistic mixed-effects model was used to estimate readmission OR, adjusting for potential confounders.
A total of 2197 readmissions occurred within 30 days of the index admission, resulting in an overall rate of 30-day readmissions of 6.7% (CI 6.4% to 7.0%). The median time to readmission was 11 days (IQR 5 to 20). In the multivariate analysis, after adjusting for age and sex, the following clinical conditions were associated with a higher risk of readmission: alcohol-related disease (OR=2.06, CI 1.36 to 3.13), metastatic cancer (OR=1.98, CI 1.57 to 2.50), epilepsy (OR=1.93, CI 1.36 to 2.75), dialysis or end-stage kidney disease (OR=1.92, CI 1.39 to 2.66), chronic obstructive pulmonary disease (OR=1.88, CI 1.49 to 2.36), stoma (OR=1.72, CI 1.22 to 2.44), transplant (OR=1.62, CI 1.03 to 2.55), being bedridden (OR=1.57, CI 1.28 to 1.93), anaemia (OR=1.57, CI 1.35 to 1.83), urinary tract infection (OR=1.54, CI 1.30 to 1.83), pneumonia (OR=1.52, CI 1.31 to 1.75), dementia (OR=1.49, CI 1.24 to 1.79), diabetes (OR=1.37, CI 1.17 to 1.61) and transfusion (OR=1.34, CI 1.03 to 1.73).
Several chronic and acute conditions at index admission significantly increased the risk of readmission. Strengthening transitional care, outpatient services and palliative care could mitigate readmissions.
Identify and describe patient engagement interventions used to improve medication management in older adults during transitions of care.
A mixed-methods systematic review.
A comprehensive search of all study designs was conducted. Studies were categorised using the ladder of patient and family engagement, a framework that positions engagement from low (passive) to high (active partnership) patient engagement.
Six databases were searched from inception to April 2024.
The search yielded 29 reports, with 25 classified as studies. Most interventions (n = 19, 76%) were low-level interventions that comprised informing patients in a passive manner. Interventions that facilitated high-level engagement (n = 6, 24%) where patients were integrated in the decision-making process were associated with consistently improved patient and healthcare long-term outcomes.
While low and high-level engagement interventions were associated with significantly decreased hospital readmission rates, high-level interventions consistently demonstrated positive patient outcomes. Interventions supporting older adults beyond discharge achieved meaningful and lasting patient and healthcare outcomes for older adults.
Findings provide clinical reference for designing engagement interventions, highlighting long-term benefits of partnership-based approaches and continuity beyond discharge.
Engagement in medication management during transitions of care varied significantly. High-level engagement was consistently linked to improved patient and healthcare outcomes but was often resource intensive. This review identifies the need to design balanced interventions that align with the preferences of older adults and real-world contextual healthcare settings.
PRISMA guidelines.
No patient or public contribution.
PROSPERO (registration number CRD42024557385).
The global rise in the population aged over 65 has led to a corresponding increase in cognitive impairment diagnoses, with dementia as a predominant condition characterised by diverse aetiopathogenetic profiles. Behavioural and psychological symptoms of dementia (BPSD) encompass a range of psychiatric, behavioural and cognitive symptoms associated with cognitive impairment. BPSD significantly affects patients, caregivers and healthcare providers, often necessitating interventions like sedatives or physical restraints that may worsen patient outcomes. Emerging evidence underscores the need for innovative, non-pharmacological interventions to manage BPSD effectively.
The current study intends to investigate the feasibility and acceptability of customised, immersive virtual reality environments (iVRe) to reduce responsive behaviours among individuals with dementia. Building on prior findings demonstrating virtual reality (VR) potential in reducing anxiety and fostering positive emotional states, this pilot study assesses the feasibility, safety and user engagement of customised iVRe interventions.
A longitudinal, mixed-methods design will be employed, enrolling 20 elderly participants with varying levels of cognitive impairment from the APSP ‘Margherita Grazioli’ long-term care facility in Trento. Participants undergo three VR exposure sessions featuring tailored environments adjusted in real-time for visual and auditory preferences. Data collection integrates standardised self-report questionnaires, observational tools and clinical records. Measures include the Functional Assessment Staging Tool, Neuropsychiatric Inventory and Cohen-Mansfield Agitation Inventory, as well as tools assessing pain, anxiety and emotional states before, during and after VR sessions.
The study protocol has been approved by the Comitato Etico Territoriale della Provincia Autonoma di Trento per le Sperimentazioni Cliniche, Azienda Provinciale per i Servizi Sanitari—Trento, Italy (Rep. Int. 12090, 15 May 2025). All participants or their legal representatives will provide written informed consent prior to enrolment. Deidentified data will be securely stored on institutional servers at the Fondazione Bruno Kessler and the APSP ‘Margherita Grazioli’, curated in compliance with the General Data Protection Regulation, and retained for 3 years after study completion. Any data shared externally will be provided in fully anonymised form, and only for scientific purposes, subject to prior ethical and legal approval. Study findings will be disseminated through peer-reviewed publications, conference presentations and executive summaries shared with participating institutions and stakeholders.
FreshRSS 