Category: Statistics

Neuroradiology. 2026 Jul 24. doi: 10.1007/s00234-026-04115-4. Online ahead of print.

ABSTRACT

BACKGROUND: Accurate prognostication after aneurysmal subarachnoid hemorrhage (aSAH) remains challenging. Conventional clinical and radiological grading systems, including the World Federation of Neurosurgical Societies (WFNS), Hunt-Hess, and Fisher scales, are widely used but have limited discriminative capacity. This study aimed to systematically compare machine learning (ML)-based prognostic models with conventional grading systems for predicting functional outcomes and mortality after aSAH, and to evaluate factors influencing ML performance.

METHODS: A systematic review and meta-analysis were conducted according to PRISMA 2020 guidelines. PubMed, Embase, Scopus, Web of Science, and the Cochrane Library were searched for studies published between 2010 and 2025. Eligible studies evaluated ML-based models for outcome prediction in adult aSAH patients and reported performance of conventional grading systems. Prognostic discrimination was pooled using random-effects meta-analysis of the area under the receiver operating characteristic curve (AUC), with predefined subgroup analyses.

RESULTS: Fourteen studies including 6,247 patients were analyzed. ML models demonstrated good to excellent discrimination, with AUCs ranging from 0.81 to 0.97. The pooled ML AUC for predicting unfavourable neurological outcome was 0.86 (95% CI 0.83-0.89; p < 0.0001), with substantial heterogeneity (I² = 96.1%). ML models outperformed conventional grading systems in most studies and showed comparable performance in the remainder. Subgroup analyses confirmed statistically significant prognostic accuracy across clinical-only, imaging-based, and multimodal ML models.

CONCLUSION: Machine learning-based prognostic models demonstrate statistically significant and clinically meaningful performance for outcome prediction after aSAH, exceeding conventional grading systems and supporting their role as complementary risk stratification tools.

PMID:42496900 | DOI:10.1007/s00234-026-04115-4

Trop Anim Health Prod. 2026 Jul 24;58(7):446. doi: 10.1007/s11250-026-05261-w.

ABSTRACT

For the first time, this study employs a century-long dataset (1925-2024) to reveal key factors that would be in relationship with sheep population (NSheep) in Türkiye using state-of-the-art machine learning algorithms. Due to the existence of missing values in the original dataset, missing observations were addressed through four imputation techniques-Next Observation Carried Backward (NOCB), Mean, MIDASpy, and Random Forest (RF)-generating four distinct datasets for comparative analysis. For revealing the key production factors related with NSheep, Extreme Gradient Boosting (XGB) and Multilayer Perceptron (MLP) algorithms were modeled via 5-fold cross-validation and multiple performance metrics (R², MSE, RMSE, MAE, and MdAPE). MLP produced lower prediction errors than XGB across all imputation techniques, though this difference was statistically confirmed only under NOCB and RF imputation (Diebold-Mariano test, P < 0.01 and P < 0.05, respectively); differences under MEAN and MIDASpy imputation were not significant. The highest overall accuracy was achieved by MLP with NOCB imputation (R² = 0.975), while XGB with RF imputation showed the weakest fit (R² = 0.917). Feature importance analyses consistently identified cattle population (NBovine) as the dominant variable associated with NSheep across all four imputation techniques and both algorithms, followed by meadow and pasture area (M&PH) for XGBoost and a more evenly distributed set of variables (M&PH, sheep meat production, goat population) for MLP. Given that NBovine and NSheep both increased steadily over the study period, this association is interpreted as reflecting shared structural growth among livestock subsectors rather than a causal effect of cattle population on sheep numbers. These dataset-specific, associative findings support the value of combining multiple imputation strategies with flexible machine learning algorithms to characterize structural interdependencies in long-term agricultural production data. Future studies incorporating chronologically ordered validation schemes and explicitly modeling structural breaks and policy shifts could further clarify the robustness of these associations, and extending this approach to other livestock species and regions would help establish their broader generalizability.

PMID:42496899 | DOI:10.1007/s11250-026-05261-w

JMIR Mhealth Uhealth. 2026 Jul 24;14:e88382. doi: 10.2196/88382.

ABSTRACT

BACKGROUND: Engagement is a necessary precondition for the effectiveness of mobile health (mHealth) apps for long-term physical health conditions (LTCs), particularly as health systems increasingly prioritize the deployment of scalable, self-guided digital interventions. Outside controlled research settings, where clinician involvement often drives engagement, little is known about whether, how, and why people engage with mHealth apps based on intrinsic motivation alone. Existing systematic reviews have cataloged behavioral engagement indicators but rarely assess the mechanisms underlying engagement.

OBJECTIVE: This overview of systematic reviews aimed to (1) synthesize evidence on how engagement with mHealth apps for LTCs is defined, measured, and associated with health outcomes, (2) explore intrinsic and extrinsic motivational processes underlying engagement, and (3) provide practical guidance for developing scalable, user-centered digital health interventions that sustain sufficient engagement with minimal reliance on external drivers. Uniquely, we interpreted modifiable barriers and facilitators through a motivational lens that distinguishes extrinsic from intrinsic motives, mapping intrinsic motives onto autonomy, competence, and relatedness, as proposed by Self-Determination Theory (SDT).

METHODS: Searches of MEDLINE, Web of Science, Epistemonikos, and gray literature (inception to June 9, 2025) identified systematic reviews reporting engagement indicators, engagement-outcome associations, or barriers and facilitators among adults with LTCs. Quantitative reviews were appraised using AMSTAR 2 (A Measurement Tool to Assess Systematic Reviews 2), and qualitative and mixed methods reviews were appraised using CASP (Critical Appraisal Skills Programme). A narrative synthesis was undertaken, and modifiable barriers and facilitators were independently mapped by 2 reviewers to extrinsic and intrinsic motivation and, for intrinsic factors, to the SDT constructs of autonomy, competence, and relatedness. Discrepancies were resolved through discussion with the wider research team.

RESULTS: Nineteen reviews (12 quantitative, 5 mixed methods, and 2 qualitative) were included from 4684 records. Fourteen (74%) reviews did not define engagement, and the remaining 5 equated it with “usage” or “adherence,” precluding meta-analysis. Fourteen reviews reported microlevel behavioral indicators, but none captured macrolevel or effective engagement. Eight assessed engagement-outcome links; 7 reported positive associations, and 1 reported no effect. Seven reviews included nonmodifiable factors that influence engagement (eg, ethnicity), while 13 included modifiable factors. SDT mapping revealed that modifiable factors influencing autonomy (eg, personal relevance, flexibility), competence (eg, usability, technical support), and relatedness (eg, clinician endorsement, peer connection) underpin intrinsic engagement, whereas extrinsic barriers include restrictions to access (including cost).

CONCLUSIONS: Current evidence on engagement with mHealth apps remains conceptually inconsistent and methodologically fragmented, but motivational patterns are clear: engagement depends largely on intrinsic motives once external conditions are satisfied. Applying SDT provides the first mechanism-oriented explanation of how engagement operates, enabling practical recommendations for evaluating existing apps and designing future mHealth interventions that support autonomy, competence, and relatedness.

PMID:42496884 | DOI:10.2196/88382

J Autism Dev Disord. 2026 Jul 24. doi: 10.1007/s10803-026-07459-1. Online ahead of print.

ABSTRACT

PURPOSE: Problem behaviours exacerbate the challenges associated with the core characteristics of autism during adolescence. Parenting interventions have demonstrated efficacy in addressing such behavioural challenges in younger autistic individuals, however, their evidence for autistic adolescents is sparse, particularly in low-income settings.

METHODS: This systematic review narratively synthesised studies outlining parenting interventions targeting problem behaviours in autistic adolescents.

RESULTS: Seven studies were included with three providing comparable quantitatively data for their efficacy in improving problem behaviours. Five out of seven studies reported positive evidence on reducing problem behaviours. Characteristics including fewer sessions, instructional mode of intervention delivery and use of behavioural management skills were most commonly shared among statistically significant studies. Parenting interventions had a positive effect on adaptive behaviours for children, and improved parent wellbeing and knowledge. The overall satisfaction with interventions was high, however, only one study was conducted in a lower- and middle-income country.

CONCLUSION: The findings underscore the encouraging evidence in an understudied area. The results emphasise the need to conduct further research for autistic adolescents and highlight potential parent-mediated interventions carry, given their acceptability and logistically scalable nature.

PMID:42496858 | DOI:10.1007/s10803-026-07459-1

Rheumatol Ther. 2026 Jul 24. doi: 10.1007/s40744-026-00861-2. Online ahead of print.

ABSTRACT

INTRODUCTION: Persistence to therapy to treat rheumatoid arthritis (RA) is an indirect measure of tolerability and effectiveness in the real-world setting. Previous clinical trials suggested that seropositive patients with RA may particularly benefit from abatacept. The risk of non-persistence after initiating abatacept, tumor necrosis factor inhibitors (TNFi), and Janus kinase inhibitors (JAKi) as first-line biologic or targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs) was examined.

METHODS: We utilized the 100% Medicare Fee-For-Service sample linked claims with Prognos laboratory data from 2012 to 2019. Newly-initiating, anti-cyclic citrullinated peptide (anti-CCP)-positive and rheumatoid factor-(RF) positive patients were identified. Index date was initiation of abatacept, TNFi, or JAKi as first-line b/tsDMARD treatment. Persistence (measured at 12 months post-index) was defined as the absence of a treatment gap ≥ 60 days, > 90 days off-treatment, or switch in therapy. Cox regression was used to investigate risk of non-persistence between the groups.

RESULTS: Of 3105 patients identified, 487 received abatacept (16%), 2330 TNFi (75%), and 288 JAKi (9%). Abatacept, TNFi, and JAKi twelve-month persistence was 48%, 33%, and 39%, respectively. Beneficiaries initiating with abatacept were more likely to be persistent than those initiating with TNFi or JAKi (hazard ratio [HR] for TNFi, 1.45, 95% CI 1.27-1.64; HR for JAKi, 1.43, 95% CI 1.19-1.72).

CONCLUSION: These real-world findings suggest that abatacept as a 1L treatment has the potential to improve persistence in patients with anti-CCP+ and RF+RA compared to the most commonly used 1L alternatives.

PMID:42496853 | DOI:10.1007/s40744-026-00861-2

Child Psychiatry Hum Dev. 2026 Jul 24. doi: 10.1007/s10578-026-02067-7. Online ahead of print.

ABSTRACT

In this article, we present a computational framework for predicting treatment response to neurofeedback (NF) among patients with Attention-Deficit/Hyperactivity Disorder (ADHD). The proposed framework uses functional brain connectivity analysis of electroencephalogram (EEG) signals acquired during an early-to-mid NF treatment window to classify participants as eventual responders or non-responders. The six-stage algorithm was evaluated using an open-access EEG dataset from the Mendeley Data repository comprising 60 children with ADHD aged 6-12 years. The framework includes a preprocessing pipeline designed to reduce EEG artifacts and noise. Next, spectral features, specifically of the alpha and beta frequency bands, were extracted from the noise-reduced signals. In the fourth stage, functional connectivity was estimated by calculating Phase Locking Value (PLV) between all electrode pairs, thereby quantifying inter-channel phase synchronization. The fifth stage, which is an essential stage, was dimensionality reduction to find the most discriminative features. Dimensionality reduction was achieved in a two-process manner; for the first process, statistical screening was performed using Welch’s t-test with FDR correction, followed by GA-based channel selection to identify the most discriminative electrode subset. The GA analysis identified a compact six-channel subset consisting of C3, C4, Cz, Fz, Fp1, and T6 from the original 32-channel montage. In the last classification stage, the reduced feature vectors were input as part of an ensemble of machine learning classifiers to achieve classification. Model performance was evaluated using subject-wise grouped cross-validation, with the best configuration achieving an accuracy of 84.72%. These results suggest that the proposed data-driven framework may support future research on individualized NF response prediction, pending validation on independent clinical datasets.

PMID:42496846 | DOI:10.1007/s10578-026-02067-7

J Med Internet Res. 2026 Jul 23;28:e99136. doi: 10.2196/99136.

ABSTRACT

BACKGROUND: Molecular Tumor Boards (MTBs) generate highly technical recommendations. The language used in their protocols is rarely accessible to patients. Lay-language patient protocols could support patient-clinician communication, yet manual production is difficult to sustain in high-volume oncology settings. Large language models (LLMs) may offer scalable drafting assistance, yet clinical usability remains largely uninvestigated under real-world deployment constraints. Existing evaluations rely predominantly on synthetic data or closed-source models that are incompatible with strict data protection requirements.

OBJECTIVE: This study evaluated whether open-weight LLMs can provide clinically usable drafting support for German MTB patient protocols under real-world deployment constraints and developed a transferable evaluation framework for patient-facing text generation.

METHODS: Eight open-weight LLMs were evaluated under zero-shot (A1) and one-shot (A2) prompting with constrained decoding, which ensures section-schema compliance. Automatic evaluation used ROUGE-1 (Recall-Oriented Understudy for Gisting Evaluation), BERTScore-F1 (Bidirectional Encoder Representations From Transformers Score), Wiener Sachtextformel version 4, and DistilBERT (Distilled Version of Bidirectional Encoder Representations From Transformers)-based complexity using a corpus of 316 MTB protocols and 47 expert-written patient protocols. For expert evaluation, 7 medical oncologists evaluated 50 protocols from the best-performing model across 3 International Organization for Standardization 9241-11 usability dimensions using fine-grained error annotation, perceived postediting effort (PPEE), and net promoter score. Critical errors were defined as bearing the risk of patient harm.

RESULTS: Llama-3.3-70B-Instruct achieved the strongest automatic performance. Across models, A2 significantly improved most automatic metrics compared to A1. However, expert usability evaluation of Llama-3.3-70B-Instruct showed the opposite picture: the proportion of protocols containing at least 1 critical error doubled under A2 (10/25, 40% vs 5/25, 20%) compared with A1, and the dominant error type shifted from language (40/108, 37%) errors to factual errors (69/145, 48%). Overall, 16% (230/1420) of the annotated paragraphs contained errors. Median PPEE was 2 (IQR 2.0-3.0; low), and median net promoter score was 7 (IQR 5.0-9.0). Detractors (46/100, 46%) outweighed promoters (29/100, 29%), which suggests hesitation toward routine adoption. These differences in expert evaluation between A2 and A1 were directionally consistent but did not reach individual statistical significance for the paired samples (n=25).

CONCLUSIONS: Prompting strategies that improve automatic metrics can simultaneously increase the number of critical errors. Surface-level metric gains were, therefore, insufficient proxies for clinical safety. This was observed as a consistent directional pattern for a single model, but generalization to other models remains to be investigated. Nonetheless, the low paragraph-level error rate and favorable PPEE suggest that structured open-weight LLM generation may be a useful drafting support in a clinician-supervised setting. The proposed evaluation framework provides a text-quality-focused basis for future assessment of patient-facing LLM applications in real-world clinical settings.

PMID:42495812 | DOI:10.2196/99136

Int Endod J. 2026 Jul 24. doi: 10.1111/iej.70236. Online ahead of print.

ABSTRACT

INTRODUCTION: This study evaluated whether the SCN9A rs6746030 (G>A; R1150W) polymorphism is associated with the clinical success or failure of pulpal anaesthesia following an inferior alveolar nerve block (IANB).

METHODS: In this prospective clinical genetic association study, 320 patients with symptomatic irreversible pulpitis in mandibular molars were recruited. Each patient received a standardized IANB with 2% lidocaine and 1:80000 epinephrine. Anaesthesia was considered successful if the Heft-Parker Visual Analogue Scale (HP-VAS) score was below 54 mm during access cavity preparation/instrumentation; scores of 54 mm or higher were classified as failures. Buccal swab samples were collected for genotyping of rs6746030 using a TaqMan SNP assay. Genotype distributions were compared between the success and failure groups. Statistical analysis included Hardy-Weinberg equilibrium testing, chi-square comparisons, and binary logistic regression with genotype, age and gender as predictors.

RESULTS: Out of 320 patients analysed, 199 experienced failed anaesthesia, while 121 achieved successful outcomes. A significant deviation from Hardy-Weinberg equilibrium was found in the failed anaesthesia group (p < 0.05), suggesting potential genetic influence, while the successful group showed no deviation (p = 0.99). Logistic regression revealed that patients with the AA genotype had significantly lower odds of successful anaesthesia (OR = 0.22; 95% CI: 0.049-0.993). No significant associations were observed for the GA genotype, age, or gender. The model’s predictive ability was limited, with an AUC of 0.53.

CONCLUSION: The rs6746030 polymorphism of SCN9A in its AA form is associated with a higher likelihood of IANB failure in mandibular molars with symptomatic irreversible pulpitis.

PMID:42495805 | DOI:10.1111/iej.70236

J Biopharm Stat. 2026 Jul 24:1-18. doi: 10.1080/10543406.2026.2699849. Online ahead of print.

ABSTRACT

Oncology dose optimization has moved beyond the maximum tolerated dose paradigm, yet many programs still implicitly target a threshold-style minimum effective dose (MED). In serious cancers, deliberately sub-therapeutic comparators are rarely ethical or approvable, so any sharp MED threshold is typically a fragile and only partially identifiable target. We propose a statistical and operational framework that reframes dose-finding as region-based optimization. First, we define a constraint-based Dose Optimization Region (DOR) where clinically meaningful benefit (potentially multi-endpoint and PD-informed) is credible, unacceptable toxicity is bounded, exposure targets are attainable within a prespecified window, and implementation is feasible. Second, within the DOR, we identify an Operational Optimal Dose (OOD) – a label-ready dosing strategy that integrates dose, schedule, exposure targets, and adjustmentrules – using comparative evidence across binary and time-to-eventendpoints, exposure – response (PK/PD) analyses, and time-to-event methods that account for delayed effects where follow-up allows. This approach turns regulatory evidence domains into region-definingconstraints and shifts the target from a single-point estimate to a strategy-level deliverable. It is intended to sit on top of conventional dose-finding designs and PK/PD modeling as a region-defining and reporting standard, rather than to replace existing methods. We illustrate its use with small-samplevisualizations based on pairwise superiority probabilities across in-region arms and with a real-world-inspired example to show how DOR→OOD can summarize the totality of evidence in practice. The framework provides statisticians and clinicians with practical design patterns and reporting checklists, aligning with contemporary regulatory expectations and supporting dose optimization in oncology, with potential adaptation to other therapeutic areas.

PMID:42495769 | DOI:10.1080/10543406.2026.2699849

Euro Surveill. 2026 Jul;31(29). doi: 10.2807/1560-7917.ES.2026.31.29.2600567.

NO ABSTRACT

PMID:42495763 | DOI:10.2807/1560-7917.ES.2026.31.29.2600567