Published systematic reviews display a heterogeneous methodological quality, which can impact decision-making. Large language models (LLMs) can support and make the assessment of the methodological quality of systematic reviews more efficient, aiding in the incorporation of their evidence in guideline recommendations. We aimed to develop an LLM-based tool for supporting the assessment of the methodological quality of systematic reviews. We assessed the performance of 8 LLMs in evaluating the methodological quality of systematic reviews. In particular, we provided 100 systematic reviews for eight LLMs (five base models and three fine-tuned models) to evaluate their methodological quality based on a 27-item validated tool (Reported Methodological Quality (ReMarQ)). The fine-tuned models had been trained with a different sample of 300 manually assessed systematic reviews. We compared the answers provided by LLMs with those independently provided by human reviewers, computing the accuracy, kappa coefficient and F1-score for this comparison. The best performing LLM was a fine-tuned GPT-3.5 model (mean accuracy = 96.5% [95% CI = 89.9%-100%]; mean kappa coefficient = 0.90 [95% CI = 0.71-1.00]; mean F1-score = 0.91 [95% CI = 0.83-1.00]). This model displayed an accuracy >80% and a kappa coefficient >0.60 for all individual items. When we made this LLM assess 60 times the same set of systematic reviews, answers to 18 of 27 items were always consistent (ie, were always the same) and only 11% of assessed systematic reviews showed inconsistency. Overall, LLMs have the potential to accurately support the assessment of the methodological quality of systematic reviews based on a validated tool comprising dichotomous items.

Objectives: Analyze the diagnostic and prognostic value of the sFlt-1/PlGF ratio in pregnant women with at least one sign/symptom of suspected/diagnosed pre-eclampsia. Methods: This retrospective observational study included 170 pregnant women with at least one sign/symptom of pre-eclampsia, who had sFlt-1/PlGF ratio values. The following information was evaluated: pregnant women’s demographic data and clinical history; laboratory data (urine protein/creatinine ratio; sFlt-1/PlGF ratio); signs and symptoms presented; clinical outcome; fetal complications; data related to childbirth. Statistical analysis was performed by R Software Version 3.5.2. Results: Among the 170 patients, 78 presented pre-eclampsia. The median sFlt-1/PlGF ratio was significantly higher [143.1 (2.2–2,927.1)] for women who presented pre-eclampsia than for women without pre-eclampsia [33.5 (0.8–400.2)]. The negative predictive value of sFlt-1/PlGF ratio 85 or 110 (for late onset pre-eclampsia) was 76.4 % (95 % CI, 66.2–84.8 %). sFlt-1/PlGF >85 or 110 was associated with pre-eclampsia clinical development, fetal complications, shorter gestational age at birth, higher number of caesarean deliveries and lower birth weight. Conclusions: The sFlt-1/PlGF ratio, together with the standard diagnostic criteria, can be used to rule out pre-eclampsia, identify high-risk patients and predict the occurrence of adverse outcomes.

Background: Telemedicine has gained prominence in primary care, but evidence comparing its modalities is scarce. Methods: We conducted a cross-sectional study of 184,308 video and phone consultations from a telemedicine platform in Portugal, comparing quality indicators based on the Institute of Medicine (IOM) domains. Results: Video consultations were associated with lower recurrence rates, higher patient-reported helpfulness (95.4% vs. 92.3%), and greater satisfaction. Time to appointment was shorter for phone consultations, but video had a longer duration. Conclusions: In this cross-sectional analysis, video consultations were associated with higher patient-reported helpfulness and lower short-term recurrence than telephone consultations, while telephone offered faster access. Because telephone users were substantially older and the dataset lacked measures of comorbidity, illness severity, and socioeconomic status, these associations are likely confounded by indication and should not be read as causal evidence of superior quality. The findings nonetheless point to differences between modalities and to the importance of equitable access to information-richer consultations.

Background: Anxiety and depression are major contributors to mental-health burden and frequently co-occur in clinical practice. In tele–mental health, routinely captured operational variables such as consultation duration, visit frequency, and follow-up cadence may provide clinical digital phenotypes that complement conventional symptom scales. This study aimed to characterize anxiety–depression comorbidity in a large real-world tele–mental health cohort and to determine whether symptom severity was associated with distinct patterns of healthcare utilization. Methods: We conducted a retrospective real-world study of 3467 patients followed in psychiatry and psychology teleconsultations. Patients were classified as anxiety only, depression only, comorbid anxiety–depression, or neither. Symptom severity was categorized as mild, moderate, or severe using validated questionnaire-based measures; to improve comparability across instruments, scores were additionally harmonized using z-score normalization. Associations between anxiety and depression severity within the comorbid subgroup were examined using a chi-square framework. Telehealth utilization endpoints included consultation duration, number of consultations, and inter-visit interval, analysed overall and stratified by sex, age group, and symptom severity. Results: Anxiety and/or depression were present in 61.7% of the cohort (2140/3467), and anxiety–depression comorbidity accounted for 43.8% of all patients (1520/3467), indicating substantial real-world overlap. Within comorbid cases, anxiety and depression severity were strongly coupled, with depression severity varying systematically across anxiety severity strata (chi-square p = 9.88 × 10⁻¹⁰²). Compared with isolated anxiety or depression, comorbidity was associated with a more intensive healthcare-utilization profile, characterized by a higher mean number of consultations and shorter inter-visit intervals. Among comorbid patients, females showed greater longitudinal service use than males, with more visits and closer follow-up. Resource use also varied according to symptom burden, mainly in depression, supporting a graded relationship between clinical severity and operational care demand. Conclusions: In this large real-world tele–mental health cohort, anxiety–depression comorbidity was highly prevalent, clinically structured, and associated with distinct and measurable resource-use signatures. These findings highlight the novelty and practical value of integrating symptom severity with operational telehealth data to derive pragmatic digital phenotypes of care intensity. Such phenotypes may support risk stratification, triage, follow-up scheduling, and capacity planning in tele–mental health, with potential translational relevance for broader mental healthcare systems. However, these findings should be considered descriptive and hypothesis-generating and warrant further longitudinal validation in other clinical settings.