Tag: statistics

J Imaging Inform Med. 2026 Jul 14. doi: 10.1007/s10278-026-02113-9. Online ahead of print.

ABSTRACT

The use of artificial intelligence (AI) models to assist clinical decisions for diagnosis and prediction has been shown to improve patient outcomes and clinical decision-making. However, the performance of an AI model deployed in a clinical setting can vary over time or between initial evaluations and clinical use because of data drift. Monitoring the performance of a deployed AI model may help to (i) ensure the AI model performs as expected in the clinical environment and (ii) detect performance deviations and alert stakeholders to these deviations. In this study, we investigate how a change in the performance of an AI model caused by an abrupt data drift can be detected using a cumulative sum (CUSUM) control chart. We demonstrate the use of CUSUM for computer-aided breast cancer detection using data from the publicly available Emory Breast Imaging Dataset (EMBED). Our results indicate that, when the magnitude of the drift is 1.5 times the standard deviation of the performance metric, CUSUM is able to detect changes in the test negativity rate of an AI model within an average of 5 days following the onset of performance drift, with a long duration ( $\sim$ 293 days) between false alarms. Our results also indicate that when the average number of samples is 40/day, leading to a reasonable standard deviation for the performance metric, almost no false alarms are expected in a period of 60 days, with the choice of CUSUM parameters that lead to a small true alarm detection delay. We also demonstrate that CUSUM can be applied for monitoring the performance of AI models even when ground-truth labels are not available. We evaluate the robustness of the CUSUM chart to non-normal distributions and show that when tuned to detect relatively small parameter shifts, the CUSUM chart can be robust to non-normal distributions. We demonstrate that the sensitivity of CUSUM can be controlled to balance the mean time between false alarms and the delay in detecting a true change in performance. We conclude that CUSUM, which is a well-studied statistical process control method, can easily be adapted for monitoring the performance of AI models targeted for assisting in medical diagnosis.

PMID:42449079 | DOI:10.1007/s10278-026-02113-9

Nurse Educ Today. 2026 Jul 12;166:107277. doi: 10.1016/j.nedt.2026.107277. Online ahead of print.

ABSTRACT

BACKGROUND: Artificial intelligence (AI) has emerged as a transformative, fundamentally reshaping how we access and apply knowledge. Achievement motivation (AM) remains a critical component of students’ academic and personal development. Examining the association between nursing students’ dependence on AI and their AM is a pertinent area of research, particularly in light of the growing incorporation of AI within educational settings.

AIM OF THE STUDY: To explore the association between undergraduate nursing students’ dependence on artificial intelligence and achievement motivation.

METHODS: A multicenter descriptive correlational study was conducted across five nursing faculties affiliated with Port-Said, Damietta, Alexandria, Mansoura, and Sohag universities, involving 1412 nursing students. Data were collected using a Personal and Academic Characteristics Questionnaire, the Achievement Motivation Scale (AMS), and the Dependence on Artificial Intelligence Scale (DIA). Data were analyzed using descriptive statistics, Pearson correlation, and simple linear regression.

RESULTS: Nursing students demonstrated a moderate level of dependence on AI (M = 19.46, SD = 11.33) and a moderate level of AM (M = 6.32, SD = 2.13). A statistically significant moderate negative association was found between AI dependence and AM (r = -0.389, p < .001). This indicates that increased dependence on AI is associated with lower achievement motivation. Regression analysis indicated that AI dependence accounted for 19.6% of the variance in AM (R² = 0.196), while the remaining variance is likely influenced by other unmeasured factors.

CONCLUSION: Nursing students exhibited moderate AI dependence and achievement motivation. The findings indicate a significant negative association between AI dependence and achievement motivation, emphasizing the need to consider students’ technology usage patterns concerning academic motivation.

IMPLICATIONS FOR PRACTICE: Educational programs should promote informed and balanced use of AI by clarifying its benefits and limitations as a learning support tool. Learning strategies should also be formulated to foster independent thinking, self-regulation, and goal-directed behavior to support students’ achievement motivation.

PMID:42447606 | DOI:10.1016/j.nedt.2026.107277

Acta Psychol (Amst). 2026 Jul 14;269:107456. doi: 10.1016/j.actpsy.2026.107456. Online ahead of print.

ABSTRACT

Practical assignments are increasingly regarded as an important means of supporting students’ broader development, yet evidence on their associations with non-cognitive skills remains limited. In the present study, physics-related practical assignments are understood as outcome-oriented learning tasks that go beyond routine written exercises and require students to apply, analyze, and sometimes create using physics knowledge in authentic, hands-on, or problem-solving contexts. In this sense, they are defined as assignments linked to observable learning performance rather than to mere repetition of subject content. This study examined whether physics-related practical assignments were associated with lower secondary students’ non-cognitive skills and whether these associations varied across regional contexts. Using large-scale survey data from 7929 Grade 8 students in Jiangsu Province, China, we applied propensity score matching to compare students who were assigned physics-related practical assignments with matched peers who were not. The results indicated positive associations between practical assignments and five dimensions of non-cognitive skills-interpersonal skills, collaboration, openness, emotional regulation, and task competence-as well as with scientific critical-reflective disposition as a related developmental outcome. These patterns were robust across alternative matching methods. However, the associations varied by region: they were significantly positive for urban and township students, but not statistically significant for rural students. These findings extend research on homework design and adolescent development by suggesting that subject-specific practical assignments may support students’ broader competencies, while also highlighting the importance of contextual conditions in shaping their benefits.

PMID:42447577 | DOI:10.1016/j.actpsy.2026.107456

Acta Psychol (Amst). 2026 Jul 14;269:107402. doi: 10.1016/j.actpsy.2026.107402. Online ahead of print.

ABSTRACT

Despite the growing influence of user-generated content (UGC) on tourism decisions and well-being, little research has examined how UGC characteristics affect tourists’ subjective well-being (TSWB). This study integrates the Stimulus-Organism-Response (S-O-R) framework and the Elaboration Likelihood Model (ELM) to investigate how review text length, sentiment polarity, and reviewer identity influence TSWB, mediated by review usefulness and moderated by photo quantity. Analyzing 31,676 reviews from Ctrip combining a BERT-Attention-BiLSTM deep learning and regression model, we find that (1) text length and reviewer identity positively predict review usefulness, while sentiment polarity has a negative effect; (2) review usefulness mediates the relationship between UGC features and TSWB; (3) photo quantity amplifies the effects of text length and reviewer identity on usefulness, but not sentiment polarity, suggesting emotional cues depend more on text than visuals. This study clarifies the mechanism linking UGC to well-being and offers practical insights for platforms and tourists.

PMID:42447576 | DOI:10.1016/j.actpsy.2026.107402

Int Dent J. 2026 Jul 14;76(5):109748. doi: 10.1016/j.identj.2026.109748. Online ahead of print.

ABSTRACT

BACKGROUND: The rising incidence of multiple primary cancers among patients with oral squamous cell carcinoma (OSCC-MPCs) presents a profound clinical challenge. This study aimed to characterize the clinicopathological phenotypes and survival outcomes of OSCC-MPCs compared with those of single primary OSCC (SPOSCC), and to develop robust prognostic nomograms for the OSCC-MPCs population.

METHODS: In this single-centre retrospective cohort study (2015-2025), clinicopathological and systemic inflammatory indices were extracted. Survival outcomes were compared using Kaplan-Meier analysis. Least Absolute Shrinkage and Selection Operator (LASSO) coupled with multivariable Cox regression were employed to identify independent predictors and construct nomograms for overall survival (OS) and progression-free survival (PFS).

RESULTS: Among the 1,909 patients (OSCC-MPCs: n = 249; SPOSCC: n = 1,660), the OSCC-MPCs cohort exhibited an older age at diagnosis, lower rates of smoking and alcohol consumption, higher T but lower N categories, and significantly elevated preoperative neutrophil-to-lymphocyte ratios (NLR). Survival analysis based on complete survival data (OSCC-MPCs: n = 249; SPOSCC: n = 1,546) revealed significantly inferior outcomes in the OSCC-MPCs group compared with the SPOSCC group, demonstrating markedly reduced 5-year OS (51.28% vs. 79.85%, P < .001) and PFS (37.97% vs. 63.05%, P < .001) rates. LASSO-Cox regression identified age > 60 gt; 60 years, advanced T/N categories, NLR > 2.5, and specific treatment modalities as independent prognostic factors. The resulting dual-endpoint nomograms demonstrated reliable discrimination, optimal calibration, and substantial clinical net benefit.

CONCLUSION: OSCC-MPCs exhibit highly distinct clinical phenotypes and are associated with significantly poorer survival compared with SPOSCC. Our nomograms, integrating clinicopathological features with NLR, provide robust, non-invasive quantitative tools to tailor individualized therapeutic interventions and optimize long-term surveillance strategies.

PMID:42447544 | DOI:10.1016/j.identj.2026.109748

Musculoskelet Sci Pract. 2026 Jul 11;85:103618. doi: 10.1016/j.msksp.2026.103618. Online ahead of print.

ABSTRACT

OBJECTIVE: To develop an expert-derived, consensus based framework for defining and classifying unintended responses (UR) and adverse events (AE) following dry needling (DN) through an exploratory, descriptive modified Delphi process.

METHODS: One hundred DN experts were invited to participate in a 3-round electronic Delphi. Eligibility required ≥3 years of DN practice, ≥5 years of clinical experience, and DN-related teaching and/or scholarship. A-priori consensus criteria were ≥80% agreement, median ≥3, and interquartile range ≤1. Round-1 used open-ended questions exploring UR and AE definitions and classifications. Rounds 2 and 3 required participants to rate agreement with statements derived from thematic analysis on a 4-point Likert scale. Descriptive statistics and Kendall’s coefficient of concordance and Wilcoxon rank-sum tests were calculated.

RESULTS: Sixty-five experts met inclusion criteria and completed Round 1. Fifty-two participants (80%) from nine countries completed all three rounds. Key consensus findings included: (1) UR categories including bleeding/bruising, neurologic responses, and autonomic responses; (2) classification of AE along a severity spectrum; (3) prioritizing severity and functional impact over response type when distinguishing UR from AE; and (4) four operational AE categories ranging from Minor Expected to Severe. Notably, pain and discomfort did not achieve consensus as an UR category.

CONCLUSION: This exploratory Delphi produced a preliminary, expert-derived framework for defining and classifying DN AE along a severity spectrum, pending prospective validation. Emphasizing severity and functional impact over prescriptive event typologies, this framework may support more consistent clinical documentation, patient communication, and research reporting.

PMID:42447540 | DOI:10.1016/j.msksp.2026.103618

Clin Biomech (Bristol). 2026 Jul 11;139:106917. doi: 10.1016/j.clinbiomech.2026.106917. Online ahead of print.

ABSTRACT

BACKGROUND: Clinicians specializing in hip preservation surgery perform procedures to increase longevity of the diseased hip. To assess patient ambulation, clinicians use temporospatial parameters (TP), to identify underlying deficits, along with asymmetry. The purpose of this study was to investigate the differences in TP and gait symmetry in a group of adolescent and young adult patients undergoing HPS.

METHODS: A retrospective query was conducted for these diagnoses: acetabular hip dysplasia (AHD), femoroacetabular impingement (FAI) and Legg-Calvé-Perthes disease (LCP). Pre-surgery and post-surgery gait lab visits (minimum of 1 year) were required. 91 un-impaired individuals (Control) were utilized for comparison. Statistical tests were run comparing the pre-surgery visit to the Control cohort for all TP and a symmetry index measure. Before and after surgery analysis was also performed.

FINDINGS: 162 patients (16.0 ± 2.8 years) were included. Prior to surgery, FAI and LCP patients walked with a reduced cadence and all groups walked slower compared to Control (p < 0.05). FAI and LCP patients had a longer stride time and stride length was shorter in AHD and LCP patients (p < 0.05). LCP patients walked with greater step time asymmetry while AHD and LCP patients walked with greater single limb support asymmetry (p < 0.05). There were minimal changes between before and after surgery.

INTERPRETATION: Patients presenting for surgery have statistically significant deficits in temporospatial parameters when compared to their unimpaired peers though many of these differences could be considered to not be clinically significant. Patients with previous surgery or diagnosed with unilateral involvement presented with greater asymmetry.

PMID:42447536 | DOI:10.1016/j.clinbiomech.2026.106917

Cognition. 2026 Jul 14;276:106648. doi: 10.1016/j.cognition.2026.106648. Online ahead of print.

ABSTRACT

From a brief exposure to an artificial language, infants and adults can segment out the embedded words in the speech input. The fact that segmentation is successful when the speech contains no prosodic cues to word boundaries has been taken as evidence that humans compute the statistics of syllable transitions to discover the underlying words. However, this explanation cannot account for experiments that show that word segmentation is hindered when the words in the artificial language differ in length, despite the availability of informative statistical properties. In this paper, we offer an alternative theoretical account of these findings. We propose that word segmentation is driven by previously unrecognized rhythmic properties inherent in the artificial language input used in prior studies. In Study 1, we demonstrate that the speech signal used in past studies of word segmentation contains periodicities which are informative as to the length of the embedded words. In Study 2, we present a computational model that uses this rhythmic property to segment speech and to rate possible word candidates. The model not only accounts for cases where learners have been found to segment artificial language streams, but also accounts for cases where they have not, such as when the artificial language stream contains words of variable length. Next, we report on a meta-analysis of the infant statistical word segmentation literature, including segmentation studies with syllable sequences void of natural speech prosody. We found that only sequences made from words that were uniform in length, but not mixed in length, were successfully learned. Taken together, we argue that the perception of rhythm as a fundamental mechanism that is ubiquitous in language and other perceptual domains can likely explain many word-segmentation studies previously attributed to the computation of transitional probabilities.

PMID:42447525 | DOI:10.1016/j.cognition.2026.106648

Genetics. 2026 Jul 14:iyag180. doi: 10.1093/genetics/iyag180. Online ahead of print.

ABSTRACT

Across many complex traits, genetic variants with larger effect sizes tend to occur at lower frequencies, which is often interpreted as a signature of stabilizing selection. In statistical genetics, the so-called α-model captures this relationship by assuming that effect size variance is inversely proportional to heterozygosity raised to a power 0 ≤ α ≤ 1. Although empirically useful, the α-model is phenomenological rather than mechanistic and lacks a direct population genetic interpretation. In this paper, we derive an alternative to the α-model based on evolutionary theory. Our approach yields a linear mixed model in which frequency dependence emerges naturally as a function of interpretable evolutionary quantities describing mutational variance, selection intensity, and coupling between the focal and selected traits. These quantities enter through two identifiable variance components that can be estimated by restricted maximum likelihood (REML). The resulting framework links a fitness landscape model to standard mixed-model methodology, enabling both inference on evolutionary parameters and downstream prediction by best linear unbiased prediction (BLUP). In forward simulations, the model accurately recovers the focal-trait variance and generally improves genetic prediction relative to conventional α-model baselines.

PMID:42447493 | DOI:10.1093/genetics/iyag180

JMIR Med Inform. 2026 Jul 14;14:e90146. doi: 10.2196/90146.

ABSTRACT

BACKGROUND: The Swiss Personalized Health Network (SPHN) facilitates the interoperability and secure sharing of health-related data for research in Switzerland, in line with the findable, accessible, interoperable, and reusable (FAIR) principles. Since medical datasets can be highly sensitive, access is often governed by complex legal and regulatory requirements. Enabling researchers to discover, understand, and evaluate datasets through rich, well-structured metadata is therefore essential to support informed decisions about data suitability and reuse.

OBJECTIVE: This study describes the design and functionality of the SPHN Metadata Catalog and its role in supporting the discovery, exploration, and reuse assessment of health-related datasets.

METHODS: The SPHN Metadata Catalog is a FAIR Data Point-compliant infrastructure that provides rich, structured metadata in both human and machine-readable form. Dataset descriptions are based on the HealthDCAT (Health Data Catalog Vocabulary) Application Profile, ensuring a standardized representation of health data catalogs. Beyond the descriptive metadata typically offered by other catalogs, the SPHN Metadata Catalog includes extensive dataset-level statistics expressed using the Vocabulary of Interlinked Datasets. An interactive visualization component further enables users to explore graph-based schemas and datasets, including entities, attributes, relationships, and their relative abundances.

RESULTS: The SPHN Metadata Catalog enables users to explore the semantic structure of graph schemas and statistics of datasets prior to requesting access. Researchers can examine data structures, relationships, attributes, and the abundances of individual data elements. This functionality supports feasibility assessments and informed evaluations of dataset suitability and reuse conditions.

CONCLUSIONS: By combining HealthDCAT Application Profile-based descriptions with rich statistical metadata and interactive exploration capabilities, the SPHN Metadata Catalog enhances dataset discoverability and supports FAIR-compliant data reuse. As a key component of Switzerland’s health data research infrastructure, the SPHN Metadata Catalog provides a foundation for future interoperability initiatives, including potential alignment with emerging frameworks such as the European Health Data Space.

PMID:42447472 | DOI:10.2196/90146