Category: Statistics

PLoS One. 2025 Apr 22;20(4):e0319383. doi: 10.1371/journal.pone.0319383. eCollection 2025.

ABSTRACT

The US COVID-19 Forecast Hub, a repository of COVID-19 forecasts from over 50 independent research groups, is used by the Centers for Disease Control and Prevention (CDC) for their official COVID-19 communications. As such, the Forecast Hub is a critical centralized resource to promote transparent decision making. While the Forecast Hub has provided valuable predictions focused on accuracy, there is an opportunity to evaluate model performance across social determinants such as race and urbanization level that have been known to play a role in the COVID-19 pandemic. In this paper, we carry out a comprehensive fairness analysis of the Forecast Hub model predictions and we show statistically significant diverse predictive performance across social determinants, with minority racial and ethnic groups as well as less urbanized areas often associated with higher prediction errors. We hope this work will encourage COVID-19 modelers and the CDC to report fairness metrics together with accuracy, and to reflect on the potential harms of the models on specific social groups and contexts.

PMID:40262087 | DOI:10.1371/journal.pone.0319383

PLoS One. 2025 Apr 22;20(4):e0318167. doi: 10.1371/journal.pone.0318167. eCollection 2025.

ABSTRACT

Background This investigation delves into the potential application of data-driven survival modeling approaches for prognostic assessments of breast cancer survival. The primary objective is to evaluate and compare the ability of machine learning (ML) models and conventional survival analysis techniques, to identify consistent key predictors of breast cancer survival outcomes. Methods This study employs data-driven survival modeling approaches to predict breast cancer survival, including survival-specific methods such as the Cox Proportional Hazards (CPH) model, Random Survival Forests (RSF), and Cox Proportional Deep Neural Networks (DeepSurv), as well as machine learning models like Random Forests (RF), XGBoost, Support Vector Machines (SVM) with an RBF Kernel, and LightGBM. The dataset, sourced from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program, comprises 4,024 women diagnosed with infiltrating duct and lobular carcinoma breast cancer between 2006 and 2010. To ensure interpretability across all models, the Shapley Additive Explanation (SHAP) method was applied to RSF, DeepSurv, Random Forests (RF), and XGBoost. This enabled the identification of key predictors influencing breast cancer survival, highlighting consistent factors across models while uncovering unique insights specific to each approach. Results The performance of survival-specific and ML models were evaluated using the Concordance index (C-index), Integrated Brier Score (IBS), mean accuracy, and mean AUC. The CPH model achieved a C-index of 0 . 71 ± 0 . 015 and an IBS of 0 . 08 ± 0 . 006, while RSF demonstrated slightly better discriminatory power with a C-index of 0 . 72 ± 0 . 0117. DeepSurv performed comparably, with a C-index of 0 . 71 ± 0 . 0095 and an IBS of 0 . 09 ± 0 . 0008. Both Cox and RSF models achieved the lowest IBS (0 . 08), indicating accurate survival probability predictions over time. For ML models, RF achieved a mean AUC of 0 . 74 ± 0 . 0021, and XGBoost with a mean AUC 0 . 69 ± 0 . 0183, reflecting fair discriminatory ability but not accounting for censoring in survival data. SHAP analysis for the top-performing models highlighted the extent of lymph node involvement, Regional Node-Positive (number of affected lymph nodes), tumor grade (cell abnormality and growth rate), progesterone status, and age as key predictors of breast cancer survival outcomes. Conclusions While ML models like XGBoost and RF can effectively identify important predictors and patterns in breast cancer outcomes, survival-specific methods such as the Cox model, RSF, and DeepSurv provide essential capabilities for handling time-to-event data and censoring, making them more suitable for accurate survival predictions. The primary objective of including ML models in this analysis was to leverage their interpretability in identifying key variables alongside survival-specific models, rather than to directly compare their performance against survival models. By examining both ML and survival models, this research highlights the complementary strengths of each approach. This study contributes to the integration of artificial intelligence in healthcare, emphasizing the value of data-driven survival modeling techniques in supporting healthcare professionals with accurate, personalized, and actionable insights for high-risk patients. Together, these approaches enhance the precision of survival predictions, paving the way for more informed clinical decision-making and improved patient care.

PMID:40262081 | DOI:10.1371/journal.pone.0318167

PLoS One. 2025 Apr 22;20(4):e0319775. doi: 10.1371/journal.pone.0319775. eCollection 2025.

ABSTRACT

The complexities of stock price data, characterized by its nonlinearity, non-stationarity, and intricate spatiotemporal patterns, make accurate prediction a substantial challenge. To address this, we propose the DCA-BiLSTM model, which combines dual-path convolutional neural networks with an attention mechanism (DCA) and bidirectional long short-term memory networks (BiLSTM). This model captures deep information and complex dependencies within time-series data. First, wavelet packet decomposition extracts high- and low-frequency features, followed by DCA for robust deep feature extraction, and finally, BiLSTM models bidirectional dependencies. Validated on datasets from Yahoo Finance, including Apple, Google, Tesla stocks, and the Nasdaq index, the model consistently outperforms traditional approaches. The DCA-BiLSTM achieves an [Formula: see text] of 0.9507 for Apple, 0.9595 for Google, 0.9077 for Tesla, and 0.9594 for the Nasdaq index, with significant reductions in error metrics across all datasets. These results demonstrate the model’s robustness and improved predictive accuracy, offering reliable insights for stock price forecasting.

PMID:40262076 | DOI:10.1371/journal.pone.0319775

JCO Clin Cancer Inform. 2025 Apr;9:e2400253. doi: 10.1200/CCI-24-00253. Epub 2025 Apr 22.

ABSTRACT

PURPOSE: Incorporating patient-reported outcomes (PROs) into health care processes can improve engagement with patients; however, adopting PROs at scale is challenging. The aim of this study was to describe the design, development, and adoption at scale of a comprehensive PRO program for standard of care and research at a cancer center.

METHODS: Requirements for a PRO program were obtained from multiple stakeholders. Components of the program included a governance process to assure a consistent and satisfactory experience for patients completing PRO questionnaires, tools to create and manage questionnaires and related content, methods to send questionnaires to relevant patients at the appropriate time, interactive tools for patients to complete the questionnaires as part of their portal experience, and integration of PRO data into workflows for clinicians. We used descriptive statistics to assess the use of the program from 2016 to 2023.

RESULTS: From program launch (on February 1, 2016) until December 31, 2023, 189 unique questionnaires were developed (101 for standard-of-care, 70 for research, and 18 for quality improvement). Of the 432,497 unique patients who were assigned at least one questionnaire, 314,685 (73%) completed at least one. Of 5,948,464 questionnaires sent, 3,098,574 (52%) were completed. The median completion time was 2 minutes.

CONCLUSION: Large-scale adoption of PROs at a cancer center is feasible. Key considerations for success include governance processes, attention to patient experience and clinician workflow, and the ability to manage complex inclusion criteria and timing of delivery of questionnaires. These principles should be disseminated so the full potential of PROs in health care can be realized.

PMID:40262062 | DOI:10.1200/CCI-24-00253

Health Rep. 2025 Apr 16;36(4):3-13. doi: 10.25318/82-003-x202500400001-eng.

ABSTRACT

BACKGROUND: It is not clear whether the increased mortality pattern observed in a prior analysis of the Canadian Census Health and Environment Cohorts for HIV/AIDS, diabetes, prostate cancer, and uterine cancer among Black adults is reflected in incident hospitalization (a marker of severity) or the diagnosis of these diseases, nor is it clear whether disparities exist regarding early screening and survivability.

METHODS: To understand the paths that contribute to differential mortality patterns, standard Cox proportional hazard models were used to assess the incidence risk of diagnosis (uterine and prostate cancer) and incident hospitalization (HIV and diabetes) among 161,520 Black adults, compared with 6,866,070 White adults. Competing risk regression was used to evaluate the cumulative risk of death for the four disease outcomes since diagnosis or hospitalization. For the observed differential cancer mortality, mediation analysis was conducted to investigate the role of cancer diagnosis at follow-up (a proxy for delayed diagnosis that is not entirely indicative of late-stage cancer).

RESULTS: Across all examined outcomes, except for uterine cancer, Black adults had elevated incident diagnoses or hospitalizations compared with White adults. Notably, Black males demonstrated a risk of incident prostate cancer and hospitalizations from HIV and diabetes twice as high relative to White males. For Black females, the risk of incident HIV hospitalization was 12 times as high. However, Black females were 15% less likely to be diagnosed with uterine cancer, compared with White females. Cumulative mortality risk analysis showed significantly lower survivability (two times lower) among Black females diagnosed with uterine cancer, relative to White females. Delayed diagnosis mediated a marginally higher proportion of the total differential uterine cancer mortality among Black females (14.9%; 95% confidence interval [CI]: 10.5% to 23.1%), compared with White females (8.9%; 95% CI: 6.3% to 13.9%).

INTERPRETATION: This study unveils substantial parallels between heightened incidence risk and relative mortality for most of the four explored outcomes between Black and White adults in Canada. Notably, the study highlights a lower incident diagnosis of uterine cancer among Black females, despite a relatively higher uterine cancer mortality. Three in every 20 uterine cancer deaths were mediated through the time of uterine cancer diagnosis (relatively delayed in Black females), underscoring the need for targeted interventions and early detection strategies to address health disparities in this population.

PMID:40262030 | DOI:10.25318/82-003-x202500400001-eng

Cardiol Rev. 2025 Apr 22. doi: 10.1097/CRD.0000000000000930. Online ahead of print.

ABSTRACT

Patients undergoing cardiac catheterization are advised not to take anything by mouth after midnight. However, limited scientific data exist on whether fasting before catheterization procedures improves clinical outcomes compared with nonfasting. A comprehensive literature search was performed by investigators using major bibliographic databases to identify studies that compared clinical outcomes for fasting versus nonfasting patient groups following cardiac catheterization procedures. The risk ratios (RR) and mean difference (MD) were pooled along with 95% confidence intervals (CIs) for dichotomous and continuous outcomes using R studios. A total of 9 trials were included in the review reporting data for 3432 patients (fasting: 1710 and nonfasting: 1702). There was no statistically significant difference between the 2 groups for incidence of procedural complications (RR: 1.05, 95% CI: 0.78-1.40; P = 0.757), 30-day mortality (RR: 0.83, 95% CI: 0.32-2.18; P = 0.71), 30-day readmissions (RR: 1.05, 95% CI: 0.74-1.49; P = 0.77), aspiration (RR: 0.45, 95% CI: 0.06-3.50; P = 0.45), contrast-associated acute kidney injury (RR: 0.90, 95% CI: 0.52-1.58; p 0.72), hypoglycemia (RR: 1.27, 95% CI: 0.74-2.17; P = 0.39), and nausea/vomiting (RR: 0.83, 95% CI: 0.46-1.51; P = 0.55). The nonfasting group was associated with significantly better satisfaction scores compared to the fasting group (standardized MD: 0.70, 95% CI: 0.13-1.27; P = 0.02). Before cardiac catheterization, a nonfasting approach is associated with higher satisfaction and similar procedural outcomes and adverse events compared to a fasting approach. The practice of routine fasting before cardiac catheterization should be reconsidered.

PMID:40262020 | DOI:10.1097/CRD.0000000000000930

WMJ. 2025;124(1):74-77.

NO ABSTRACT

PMID:40262014

WMJ. 2025;124(1):42-46.

ABSTRACT

INTRODUCTION: Implicit bias in patient care and outcomes is well documented. However, the presence of bias in hospital security interactions is a relatively new area of research. Flags placed on the electronic medical record identify patients considered high risk for negative outcomes, including those with security interactions.

OBJECTIVE: We sought to explore the types of flags and their frequency, differences among patients with flags, and their pattern over time.

METHODS: We conducted a retrospective chart review of flags placed on electronic medical records over 13 years of adults 18 years or older who were patients at a Midwest, tertiary, academic medical center. Descriptive statistics were used to explore patient demographic data. Chi-square tests were executed to compare patients with different flag types.

RESULTS: Three flag types were investigated: “communication alert,” “vulnerable/unsafe, behavior” and “risk management.” The communication alert flag was most common, although Black male patients were more likely to receive a vulnerable/unsafe behavior flag than a communication alert flag (P = 0.001). Patients who were prescribed anti-anxiety medications, antidepressants, antipsychotics, and psychotherapeutics also were more likely to receive a vulnerable/unsafe behavior flag than a communication alert flag (P = 0.001). The highest number of flags was placed during quarter 3 – the months of July, August, and September.

CONCLUSIONS: Records of patients with certain demographics and on certain medications were more likely to be labeled with vulnerable/unsafe behavior flags. There is no clear protocol to determine what behaviors elicit which flag. Standardized procedures could help provide transparency to this issue.

PMID:40262006

WMJ. 2025;124(1):17-21.

ABSTRACT

INTRODUCTION: People experiencing homelessness are more likely than the general population to have chronic health conditions and often encounter significant barriers to health care access. Many of these barriers can be affected by community-based factors, such as availability of reliable transportation, past experiences with health care systems, and community attitudes toward the unhoused population. This project aims to assess the needs and barriers to health care identified by people experiencing homelessness in a rural Midwestern city.

METHODS: The survey used was adapted from a survey previously conducted to assess the needs of the homeless population in Milwaukee, Wisconsin. Surveys were distributed during outreach around the city of Wausau, Wisconsin. Data were transcribed and reviewed, and descriptive statistics were calculated.

RESULTS: A total of 45 surveys were completed. Most participants identified as White, non-Hispanic males (n = 24, 53%) and were 46 to 55 years old (n = 14, 31%). Barriers to health care included lack of housing, cost, transportation, lack of a mailing address, inadequate hours, and disrespectful care. Eighty-six percent of participants (n = 38) reported having a mental health diagnosis, yet only 26% (n = 12) stated that they see a mental health professional.

CONCLUSIONS: Individuals experiencing homelessness in a rural community have broad and complex barriers to accessing health care. Given limited resources in smaller communities, innovative and holistic solutions should be considered when aiming to make care more equitable.

PMID:40262002

Sci Signal. 2025 Apr 22;18(883):eado3473. doi: 10.1126/scisignal.ado3473. Epub 2025 Apr 22.

ABSTRACT

Cellular plasticity mediates tissue development as well as cancer growth and progression. In breast cancer, a shift to a more epithelial phenotype (epithelialization) underlies a state of reversible cell growth arrest called tumor dormancy, which enables drug resistance, tumor recurrence, and metastasis. Here, we explored the mechanisms driving epithelialization and dormancy in aggressive mesenchymal-like breast cancer cells in three-dimensional cultures. Overexpressing either of the epithelial lineage-associated transcription factors OVOL1 or OVOL2 suppressed cell proliferation and migration and promoted transition to an epithelial morphology. The expression of OVOL1 (and of OVOL2 to a lesser extent) was regulated by steroid hormones and growth factors and was more abundant in tumors than in normal mammary cells. An uncharacterized and indirect target of OVOL1/2, C1ORF116, exhibited genetic and epigenetic aberrations in breast tumors, and its expression correlated with poor prognosis in patients. We further found that C1ORF116 was an autophagy receptor that directed the degradation of antioxidant proteins, including thioredoxin. Through C1ORF116 and unidentified mediators, OVOL1 expression dysregulated both redox homeostasis (in association with increased ROS, decreased glutathione, and redistribution of the transcription factor NRF2) and DNA damage and repair (in association with increased DNA oxidation and double-strand breaks and an altered interplay among the kinases p38-MAPK, ATM, and others). Because these effects, as they accumulate in cells, can promote metastasis and dormancy escape, the findings suggest that OVOLs not only promote dormancy entry and maintenance in breast cancer but also may ultimately drive dormancy exit and tumor recurrence.

PMID:40261955 | DOI:10.1126/scisignal.ado3473