Tag: statistics

Pharm Stat. 2026 Mar-Apr;25(2):e70086. doi: 10.1002/pst.70086.

ABSTRACT

Well-controlled clinical trials employ careful processes to reduce bias, often blinding investigators and sponsors to prevent knowledge of study outcomes and potential operational bias. Quality assurance of outcomes is also ensured through designation of unblinded data managers and statisticians, so that complex adaptive designs with multiple interim analyses can be executed. Our approach addresses potential ad-hoc requests by the Data and Safety Monitoring Board (DSMB) for monitoring safety, efficacy, and ethical oversight. A novel approach utilizing current trial data is proposed to predict trial outcomes for blinded decision-makers without unblinding those that should stay blinded. Bayesian predictive power, a trial prediction method, is employed and illustrated on simulated data. This study presents an approach for presenting updated Bayesian predictive power in complex adaptive designs, exemplified by the Hyperbaric Oxygen Brain Injury Treatment (HOBIT) trial. Simulation examples motivated from the trial demonstrate the utility of Bayesian predictive power in predicting trial outcomes and sample size distribution, aiding in resource allocation and decision-making with different reports for blinded and unblinded teams. Bayesian predictive power calculations offer valuable insights into future trial behavior for both blinded and unblinded groups, aiding in guidance during trial conduction. The approach outlined in this short communication can be applied to various trial designs.

PMID:41849677 | DOI:10.1002/pst.70086

Eur J Public Health. 2026 Mar 14;36(2):ckag022. doi: 10.1093/eurpub/ckag022.

ABSTRACT

Multimorbidity places increasing pressure on healthcare systems, requiring effective tools to assess clinical complexity. Existing comorbidity indices are often setting-specific and lack generalizability. The Multisource Comorbidity Score (MCS), developed in Italy, has shown strong predictive value. This study aimed to externally validate MCS and to test recalibrated and context-adapted versions to enhance its performance in a different healthcare system. A longitudinal observational study included 198 753 residents aged ≥50 in the Barcelona-Esquerra health district, followed between 2016 and 2019. The original MCS was validated, and two adapted versions were tested: a recalibrated MCS with locally derived weights and an enhanced MCS incorporating primary care data. Predictive validity for 1-year mortality (primary outcome) and secondary outcomes (4-year mortality, hospitalizations, and healthcare use) was assessed using the Area Under the Receiver Operating Characteristic (AUROC) curve, survival analysis, and net reclassification improvement (NRI). All MCS versions showed good discrimination. AUROCs for 1-year mortality were 0.742 (original), 0.756 (recalibrated), and 0.771 (enhanced). Adapted versions achieved better risk reclassification and higher discrimination for long-term mortality. Higher MCS scores were associated with progressively lower survival probabilities and increased healthcare resource utilization. The MCS demonstrated satisfactory external validity in the validation context, with adapted versions offering modest improvements.

PMID:41849674 | DOI:10.1093/eurpub/ckag022

J Med Internet Res. 2026 Mar 18;28:e81879. doi: 10.2196/81879.

ABSTRACT

BACKGROUND: Telehealth use surged during the COVID-19 pandemic and has stabilized at levels substantially above prepandemic baselines. However, concerns persist that the digital divide may reproduce or widen disparities in access. Understanding the determinants of telehealth use-and particularly modality choice between video and audio-is essential for designing policies that promote equitable access in the post-public health emergency era.

OBJECTIVE: This study aims to identify determinants of telehealth use and modality among US adults in 2022 and quantify the relative contributions of digital, geographic, clinical, and socioeconomic domains.

METHODS: We conducted a cross-sectional secondary analysis of the sixth cycle of the Health Information National Trends Survey, administered in 2022 by the National Cancer Institute, a nationally representative, 2-stage stratified random probability survey of civilian, noninstitutionalized US adults aged 18 years or older. Sampled households were recruited via mailed invitations, and 1 adult per household was randomly selected using the next birthday method and invited to complete a self-administered questionnaire between February 2022 and November 2022 (N=6252). The primary analytic sample included respondents with nonmissing telehealth modality responses (n=6046, 59.4% female; mean age of 55.1 y). Individual-level data were linked to county-level American Community Survey socioeconomic indicators and broadband availability measures. The primary outcome was telehealth use, categorized as video (n=1641, 27.2%; 95% CI 25.5%-29.1%), audio-only (n=876, 12.1%; 95% CI 10.9%-13.4%), or none (n=3529, 60.7%; 95% CI 58.6%-62.7%). We estimated 4 binary contrasts using survey-weighted linear probability models with jackknife variance estimation, reporting absolute risk differences in percentage points (pp) with 95% CIs. We applied Shorrocks-Shapley decomposition to quantify each predictor domain’s contribution to explained variance.

RESULTS: Nationally, 39.3% (n=2517; 95% CI 37.3%-41.4%) reported any telehealth use in the past 12 months. In survey-weighted linear probability models (α=.05), significant predictors of any telehealth vs none included: male sex (-9.7 pp, 95% CI -14.0 to -5.4), disability status (+22.5 pp, 95% CI 16.1-28.8), and health app use (+18.4 pp, 95% CI 12.0-24.8). For video vs audio-only telehealth, insurance coverage increased video use (+21.2 pp, 95% CI 13.0-29.3), while basic cell phone only (vs smartphone) decreased video use (-20.1 pp, 95% CI -33.5 to -6.8). Shorrocks-Shapley decomposition revealed that digital access and eHealth behaviors explained 40.4% of variance in video vs audio choice and 33.4% of video vs none; geography explained 40.5% of audio vs none; digital factors (25.7%), geography (19.7%), and health status and needs (15.5%) all contributed substantially to any vs none.

CONCLUSIONS: Digital access and eHealth behaviors collectively explain more variance in modality choice than traditional sociodemographic factors. Telehealth uptake reflects a combination of digital factors, geography, and clinical need, whereas video modality specifically hinges on digital readiness. Interventions pairing sustained insurance coverage with targeted investments in device access, affordable high-speed connectivity, and digital literacy training are most likely to narrow persistent telehealth gaps.

PMID:41849671 | DOI:10.2196/81879

JMIR Form Res. 2026 Mar 18;10:e79694. doi: 10.2196/79694.

ABSTRACT

BACKGROUND: Immune checkpoint inhibitors are widely used in oncology but can cause immune-related adverse events (irAEs), which may be severe or life-threatening if not detected early. Electronic patient-reported outcome (ePRO) symptom monitoring systems may facilitate timely recognition and management of irAEs. Usability testing is a critical stage in ePRO system development, yet no published examples of formal usability and acceptance testing exist.

OBJECTIVE: This study aims to assess the usability and acceptance of a co-designed ePRO symptom monitoring prototype for irAEs embedded within the Epic electronic medical record.

METHODS: Testing was conducted at an Australian quaternary cancer center. Eligible participants were patients who had received or were receiving immune checkpoint inhibitors, their caregivers, or clinicians (oncologists and nurse specialists). Participants completed baseline digital literacy assessments (16-item Mobile Device Proficiency Questionnaire [MDPQ-16] and 12-item Computer Proficiency Questionnaire [CPQ-12]) before a structured testing session. Each session involved role-specific tasks using the patient-facing Health Hub or the clinician-facing Epic electronic medical record. Usability was assessed using the System Usability Scale (SUS). Acceptance was assessed using a customized Unified Theory of Acceptance and Use of Technology (UTAUT) questionnaire. Semistructured interviews were used to capture qualitative feedback.

RESULTS: A total of 30 participants (7 patients, 3 caregivers, 10 oncologists, and 10 nurse specialists) completed 10 testing sessions. Median MDPQ-16 and CPQ-12 scores were higher for clinicians compared to patients and caregivers. Median SUS scores indicated high usability-patients and caregivers: 77.5% (IQR 70.0%-86.3%), oncologists: 82.5% (IQR 80.0%-90.0%), and nurse specialists: 80.0% (IQR 75.6%-94.4%). Median UTAUT scores demonstrated strong user acceptance-patients or caregivers: 4.27 (IQR 4.09-4.58), oncologists: 4.33 (IQR 4-4.63), and nurse specialists: 4.23 (IQR 3.87-4.57). Health Hub usability themes highlighted overall ease of navigation and efficiency of reporting, but a need for clearer survey navigation, simplification of the actions page, and improved organization of trend graphs. For clinicians, themes included efficient side effect capture and intuitive system design, but a need to improve navigation to results, optimize data display, and facilitate team-based alert management. Health Hub acceptance themes highlighted patient empowerment to self-manage, enhanced patient-clinician communication, and reinforcement of existing care. However, concerns were raised about digital equity for vulnerable groups. Clinicians reported that the system streamlined side effect management between visits, aligned with existing Epic workflows, and could be tailored to personal preferences. Concerns remained regarding additional workload and medico-legal responsibilities associated with real-time alerts.

CONCLUSIONS: The ePRO prototype demonstrated high levels of usability and acceptance across patients, caregivers, and clinicians. Limitations around navigation and data visualization, alongside equity and workload concerns, will guide refinements prior to implementation. These findings emphasize the value of rigorous formative usability and acceptance testing to optimize ePRO systems prior to deployment in routine cancer care.

PMID:41849661 | DOI:10.2196/79694

JMIR Res Protoc. 2026 Mar 18;15:e76470. doi: 10.2196/76470.

ABSTRACT

BACKGROUND: Advances in antiretroviral therapy (ART) have significantly improved the life expectancy of people living with HIV. However, maintaining retention in care-defined as ongoing engagement with medical services from diagnosis through regular follow-up-is essential for optimal clinical outcomes. Loss to follow-up (LTFU), commonly defined as the absence of ART prescription refills or medical visits for more than 90 days, has been associated with increased mortality, treatment failure, and continued community transmission. Although multiple individual and structural factors have been linked to LTFU, evidence from the Mexican context remains limited.

OBJECTIVE: This protocol describes a prospective cohort study designed to identify factors associated with LTFU among recently diagnosed people living with HIV in Mexico.

METHODS: We conducted a prospective cohort study at a national tertiary care hospital in Guadalajara, Jalisco, Mexico. Eligible participants were adults (≥18 years) who had initiated ART within 6 months prior to enrollment. Data on sociodemographic, clinical, HIV-related, and psychosocial variables were obtained from electronic medical records, pharmacy dispensing logs, and validated questionnaires (Simplified Medication Adherence Questionnaire, Berger HIV Stigma Scale, and the Medical Outcomes Study HIV Health Survey). The primary outcome is LTFU, defined as ≥90 consecutive days without a medical visit or ART refill, ascertained through institutional records and national ART dispensation systems. Participants will be followed for 24 months. Planned analyses include descriptive statistics, Kaplan-Meier curves for time to LTFU, and multivariable Cox and logistic regression models to identify factors independently associated with disengagement from care.

RESULTS: Recruitment took place between December 2023 and March 2024, yielding 164 enrolled participants who completed all baseline assessments. The 24-month follow-up period for this cohort extends from April 2024 through March 2026, with primary analyses and dissemination of results planned for the second half of 2026. Baseline data indicate that the cohort is characterized by substantial socioeconomic vulnerability, a high prevalence of late presentation, and notable levels of perceived stigma and reduced health-related quality of life.

CONCLUSIONS: This protocol outlines a prospective cohort study to evaluate factors associated with LTFU among people living with HIV in a Mexican tertiary care setting. The baseline findings highlight substantial socioeconomic, clinical, and psychosocial vulnerabilities that may compromise long-term retention in care. The longitudinal follow-up of this cohort will provide context-specific evidence to inform targeted interventions aimed at improving engagement in care and reducing LTFU in similar populations.

PMID:41849639 | DOI:10.2196/76470

Cerebrovasc Dis Extra. 2026 Mar 18:1-18. doi: 10.1159/000551531. Online ahead of print.

ABSTRACT

BACKGROUND AND AIMS: The benefit of intravenous thrombolysis (IVT) before endovascular therapy (EVT) in patients with acute ischemic stroke (AIS) with medium-large infarct core (MLIC) remains uncertain.

METHODS: We conducted a retrospective analysis of a prospective multicenter registry in Vietnam (August 2023-September 2024). We included patients with AIS-LVO in the anterior circulation within 4.5 hours of onset, an Alberta Stroke Program Early CT Score (ASPECTS) < 6, and a National Institutes of Health Stroke Scale (NIHSS) ≥ 6 at admission. The primary outcome was functional ambulation (defined as mRS 0-3) at 90 days of follow-up. Secondary outcomes were functional independence (mRS 0-2), mRS shift analysis, and rates of successful reperfusion (modified thrombolysis in cerebral infarction 2b-3). Safety outcomes were defined by symptomatic intracranial hemorrhage (ICH) according to SITS-MOST criteria and 90-day mortality. Outcomes between the bridging therapy and EVT alone groups were compared using propensity score-matched (PSM) analysis.

RESULTS: Of 403 MLIC patients undergoing EVT, 148 presented within 4.5 hours, 59 (39.9%) received bridging IVT. After PSM (n=72), with 36 in each group. The median age, proportion of males, baseline ASPECTS, and NIHSS scores were similar between the two groups. The bridging group achieved higher rates of functional ambulation (75% vs 41.7%, OR 4.2, 95% CI 1.54-11.46). Regarding safety, there was no statistically significant difference in symptomatic intracerebral hemorrhage (8.3% vs 11.1%, p = 1.0) or mortality (8.3% vs 19.4%, p = 0.17), though the confidence intervals were wide.

CONCLUSIONS: Our study suggests that bridging therapy in patients with acute medium-large ischemic core within 4.5 hours of onset results in better functional outcomes than EVT alone without increasing the sICH rate. Further studies are required to assess the safety and efficacy of bridging therapy.

PMID:41849637 | DOI:10.1159/000551531

Kidney Blood Press Res. 2026 Mar 18:1-21. doi: 10.1159/000551589. Online ahead of print.

ABSTRACT

INTRODUCTION: The identification of non-diabetic kidney disease (NDKD) in diabetic patients is critically important. Unlike diabetic nephropathy, NDKD often requires additional therapeutic interventions beyond standard diabetes care. There is a need to develop computational methods using electronic medical record data to identify NDKD in diabetic patients for whom kidney biopsy is not an option.

METHODS: The study included 1136 diabetic patients who underwent kidney biopsy at a tertiary teaching hospital. We collected 103 parameters from electronic medical records, including demographic characteristics, physical examination results, laboratory tests, and the status of diabetic retinopathy. We developed seven models to detect NDKD, including k-nearest neighbors, random forest, extreme gradient boosting (XGB), lasso Logistic regression, support vector machine, naïve bayes, and multilayer perceptron (MLP), in the training set (n=908), and compared their performances in the testing set (n=228). The SHapley Additive exPlanations (SHAP) approach was used to analyze the importance of features.

RESULTS: Biopsy-confirmed NDKD was present in 53% of the 1136 participants. In the testing set, the area under the receiver operating characteristic curve (AUC) for NDKD detection using XGB, Lasso regression, and MLP reached 0.8, with performances that were stable regardless of whether variable normalization was performed. Among them, XGB revealed the highest AUC (0.833; 95% CI: 0.800 to 0.864) without feature normalization, which was statistically superior to the other models according to DeLong’s tests. After feature normalization, SVM achieved the highest AUC of 0.841 (95% CI: 0.817to 0.861) among all models. In addition to established predictive factors for NDKD (e.g., hematuria and absence of diabetic retinopathy), SHAP analysis identified several features, such as low IgG levels, that contributed significantly to the differentiation models.

CONCLUSION: Despite performance variations in different modeling techniques, machine learning models may have the potential to facilitate the detection of NDKD for patients with contraindications for kidney biopsy. Further efforts are warranted to improve accuracy and facilitate their translation into clinical practice.

PMID:41849636 | DOI:10.1159/000551589

JMIR Form Res. 2026 Mar 18;10:e83090. doi: 10.2196/83090.

ABSTRACT

BACKGROUND: Smartwatches have gained popularity for their potential to provide accurate measurements of various physiological parameters. However, the validity of energy expenditure (EE) across different smartwatch models remains a topic of ongoing investigation. Discrepancies between results obtained from different models and gold standard methods are particularly critical across varying exercise intensities and types, as validation studies have demonstrated overestimation when wearable activity monitors are compared with indirect calorimetry.

OBJECTIVE: This study investigated the accuracy of 2 versions of the Samsung smartwatch (Galaxy Watch [GW] 6 and 7) in measuring EE during intermittent moderate-intensity running exercises, using indirect calorimetry as the gold standard method.

METHODS: This study included 148 healthy adults, comprising 80 men and 68 women. Participants performed intermittent treadmill running, consisting of walking at 5 km·h⁻¹ for 1 minute and running between 8 and 16 km·h⁻¹ for 2 minutes, based on participant preference, for a total duration of 27 minutes. The GW6 and GW7 models were used and EE was measured by indirect calorimetry using a wearable portable metabolic gas analysis system (K5; Cosmed), which is considered a gold standard method.

RESULTS: No statistically significant differences were found between the GW models and the K5. The K5 showed a mean EE of 213.60 (SD 43.04) kilocalories, compared with 219.53 (SD 35.70) kilocalories for the GW6 and 202.67 (SD 47.42) kilocalories for the GW7 (all P>.05). Good Spearman correlations (0.63-0.70) and moderate intraclass correlation coefficients (0.65-0.74) were found. Mean absolute percentage error values ranged from 10.10% to 12.55%. Bland-Altman analysis revealed limits of agreement for all comparisons (K5 vs GW6 and GW7, -61.93 to 65.80 kcal).

CONCLUSIONS: The GW6 and GW7 devices showed moderate validity for estimating EE during intermittent running exercises, demonstrating the suitability of the GW as a low-cost and practical wearable option for daily physical activities.

PMID:41849635 | DOI:10.2196/83090

Cardiology. 2026 Mar 18:1-24. doi: 10.1159/000551373. Online ahead of print.

ABSTRACT

BACKGROUND: Heart failure (HF) and various arrhythmias frequently co-occur in clinical practice, suggesting shared pathophysiological mechanisms. However, the extent and nature of their common genetic architecture remains incompletely understood. This study aimed to systematically investigate the genetic correlations and shared causal loci between HF-related traits and multiple arrhythmia phenotypes.

METHODS: We utilized GWAS summary statistics from European cohorts to analyze HF-related traits and ten common arrhythmias. Global genetic correlations were assessed using LDSC and HDL. Local genetic correlations were further investigated using LAVA, HESS, and SUPERGNOVA to identify regional overlaps. Pleiotropic loci were identified using PLACO, with Bayesian colocalization analysis (stringent threshold PP.H4 ≥ 0.75) to assess shared causality. Bidirectional Mendelian randomization (MR) was conducted to explore causal relationships, utilizing a discovery threshold (P < 5×10⁻⁶) and a validation threshold (P < 5×10⁻⁸) with independent FinnGen data.

RESULTS: Significant genome-wide genetic correlations were identified between HF and seven arrhythmia traits, with the strongest association for atrial fibrillation (LDSC rg = 0.42, P = 5.1×10⁻¹⁸; HDL rg = 0.63, P = 5.9×10⁻³⁷). Local genetic correlation analyses identified multiple genomic regions of significant overlap, particularly converging on a major hotspot at the 4q25/PITX2/ENPEP locus across all three methods. Pleiotropic analysis identified several high-confidence shared loci, including regions harboring BAG3 (PP.H4 = 0.990) and ZFHX3 (PP.H4 = 0.938). Bidirectional MR revealed significant causal effects of AF on HF development (IVW OR = 1.22, P = 4.83×10⁻¹⁸) and HF on reduced heart rate variability (P = 1.86×10⁻⁴), both validated in independent cohorts.

CONCLUSIONS: Our findings demonstrate substantial and complex shared genetic architecture between HF and multiple arrhythmia phenotypes. These insights identify specific pleiotropic genes, regional correlation hotspots, and causal pathways, potentially informing future precision medicine approaches for cardiovascular disease prevention and treatment.

PMID:41849624 | DOI:10.1159/000551373

Health Rep. 2026 Mar 18;37(3):15-27. doi: 10.25318/82-003-x202600300002-eng.

ABSTRACT

BACKGROUND: The accurate monitoring of population mental health requires repeated assessments using valid and reliable measures. The Warwick-Edinburgh Mental Well-being Scale (WEMWBS) and its short form (SWEMWBS) are widely used positive mental health (PMH) measures ([S]WEMWBS is used hereafter to refer to both). This study tested their validity among Canadian adults using representative health survey data.

DATA AND METHODS: Cross-sectional data from the 2024 Canadian Community Health Survey – Rapid Response on Sleep Quality and Positive Mental Health of adults (18 years and older) living in the provinces were used. The distributions of (S)WEMWBS responses and scores were examined. Confirmatory factor analysis (CFA) and bifactor exploratory structural equation modelling (ESEM) were conducted to assess factorial validity. Measurement invariance was tested across gender and age. Differences in (S)WEMWBS scores by gender, age, and other mental health indicators were examined. Cronbach’s alphas were used to investigate internal consistency.

RESULTS: (S)WEMWBS scores had relatively normal distributions, with no floor and minimal ceiling effects. A bifactor ESEM and bifactor CFA model for the WEMWBS and SWEMWBS, respectively, fit the data best, with indices suggesting that they were essentially unidimensional. Evidence was found for measurement invariance across gender and age. Older adults had higher (S)WEMWBS scores on average, as did men on the WEMWBS. The (S)WEMWBS had acceptable internal consistency and were associated with other mental health indicators.

INTERPRETATION: The (S)WEMWBS appear to be valid and reliable PMH measures for Canadian adults. The (S)WEMWBS could be regularly included in health surveys to support the surveillance of population-level changes in PMH.

PMID:41849617 | DOI:10.25318/82-003-x202600300002-eng