University of Indianapolis

Purpose This study explores how a privacy-preserving local large language model (Local-LLM) assists researchers in evaluating unstructured text data quality, early in the analysis process. It aims to accurately and effectively determine if a Local-LLM can evaluate the quality of open-ended survey responses compared to human coders during the initial analysis of unstructured survey data. Design/methodology/approach A Python tool using Llama 3.2 and 3.3 classified 604 survey responses. Human-coded sentiment labels served as a baseline; model performance was assessed with confusion matrices, F1, Cohen’s κ and Gwet’s AC1. All processing was offline to protect data privacy. Findings Llama 3.3 achieved top performance (F1 ≈ 0.97, AC1 ≈ 0.97), while Llama 3.2 also excelled on consumer hardware. Automated sentiment analysis reduced processing time from 4 h to 8 min and identified short responses that manual reviewers might miss, improving speed and data quality. Research limitations/implications The approach’s reliability beyond English or on longer narratives remains to be examined in future work. Practical implications The Local-LLMs allow researchers to rapidly filter and assess large volumes of unstructured text data before conducting deeper analysis. Originality/value This study demonstrates that Local-LLM’s semantic features can be used for sentiment analysis as a scalable, cost-effective and rapid method for evaluating unstructured text data, aligning with the research questions before analysis. It also illustrates how to use zero-shot prompting to interact with Ollama via Local-LLMs, simplifying AI integration for non-technical researchers.

ABSTRACT Introduction Low dietary diversity has been identified as a predictor of depression outcomes in high-income countries, while evidence is scarce from low-income settings where poor nutrition and depression often co-occur. In this study, we estimate the relationship between dietary diversity and depression among adults in rural western Kenya. Method We conducted a cross-sectional analysis of 311 participants enrolled in the Bridging Income Generation through Group Integrated Care program. We assessed depressive symptoms using the 20-item Centre for Epidemiologic Studies for Depression Scale (CES-D), and measured dietary diversity as the number of five food groups consumed in previous 24-hours using a validated dietary diversity scale. We used linear regression to estimate the association between high dietary diversity (consumption of all the five food groups) and continuous depression scores, adjusting for key covariates. We tested for effect measure modification by wealth status. We conducted a secondary analysis using quantile regression to explore variation across depression scores distributions. Results Higher dietary diversity was associated with fewer depressive symptoms [adjusted β (95% CI): -3.49 (-6.62, -0.38)]. The association was stronger among individuals with low wealth backgrounds [adjusted β (95% CI): -6.00 (-10.46, -1.42)] relative to those with high wealth backgrounds [adjusted β (95% CI): -0.53 (-4.76, 3.68); Wald p-value for interaction term =0.0003]. The effect sizes for the association were larger at higher quantiles, notably at 75th [adjusted β (95% CI): -4.00 (-10.13, 2.13)] and 90th [adjusted β (95% CI): -1.59 (-7.43, 4.25)] compared to those at lower quantiles for 10 th [adjusted β (95% CI): -0.59 (-2.46, 1.28) and 25 th [adjusted β (95% CI): -0.82 (-4.14, 2.50), though wide confidence intervals limited the precision of effect estimates. Conclusion In rural western Kenya, higher dietary diversity was associated with lower depression symptoms, particularly among participants from lower wealth backgrounds, and particularly among those with scores consistent with more severe depression symptoms. These findings suggest that improving dietary diversity may offer mental health benefit to the most socioeconomically disadvantaged individuals and could be a promising strategy to reduce depression in resource poor settings. Future work could leverage longitudinal and experimental studies for improved inference and should investigate mechanisms through which dietary diversity may influence depression.

In the United States, adult human papillomavirus (HPV) vaccination coverage remains low at 20-50%, depending on age, and cervical cancer (CC) screening rates range from 68 to 76%. Few studies have evaluated characteristics of women who are both HPV vaccinated and up to date (UTD) with screening as an integrated outcome. The purpose of the present study was to classify women into four prevention categories and examine factors associated with being double protected compared to unprotected. Data were gathered via an online survey from a sample of low-income women (household income < USD 50,000) provided by a research survey company (n = 719). Women were classified into four categories: vaccinated only, screened only, both vaccinated and screened (double protected), or neither (unprotected). Sociodemographic characteristics, healthcare access, and Health Belief Model constructs were assessed. Multivariable logistic regression compared women who were double protected with those unprotected (n = 274). Most women were UTD with screening only (57.8%), while 15.5% were double protected and 22.6% were unprotected. Younger age (Odds Ratio [OR = 0.93; 95% Confidence Interval [CI]: 0.89, 0.98), having ≥1 medical visit in the past year (OR = 4.16; 95% CI: 1.74, 9.95), higher perceived CC risk (OR = 3.65; 95% CI: 1.41, 9.43), greater perceived benefits of CC screening (OR = 1.96; 95% CI: 1.45, 2.66), and higher HPV knowledge (OR = 1.09; 95% CI: 1.01, 1.17) were associated with higher odds of being double protected. A substantial proportion of low-income women lack comprehensive CC prevention. Integrated, bundled prevention strategies that simultaneously promote HPV vaccination and screening may be important to reduce CC disparities.

Delirium, an acute and fluctuating neurocognitive disorder prevalent among hospitalized and geriatric surgical patients, remains a pervasive yet underrecognized clinical challenge. Leveraging Electronic Health Records (EHRs), Machine Learning (ML) models have emerged as promising tools for early prediction and intervention. This scoping review synthesizes the existing literature, identifies current research gaps, and outlines future directions to advance delirium prediction modeling. Following the PRISMA Extension for Scoping Reviews (PRISMA-ScR) guidelines, literature from 2020 to 2025 was systematically searched across Google Scholar, EMBASE, PubMed, Scopus, and Web of Science using a comprehensive query strategy. The review highlights a significant reliance on structured preoperative and intraoperative EHR for delirium prediction, despite the existence of abundant and highly informative unstructured clinical narratives. Furthermore, a substantial heterogeneity exists in the utilized delirium identification methodologies (e.g. Nursing Delirium Screening Scale (Nu-DESC), Delirium Observation Screening Scale (DOSS), International Classification of Diseases (ICD) criteria, 4AT delirium detection, Confusion Assessment Method (CAM), Intensive Care Delirium Screening Checklist (ICDSC), Cornell Assessment of Pediatric Delirium (CAPD) Diagnostic and Statistical Manual of Mental Disorders 5th version (DSM-5), natural language processing (NLP) based analysis), alongside a focus on specific surgical subgroups. This limited data utilization and methodological variation pose challenges to ML model generalizability and robustness. The literature also showed a research emphasis on critically ill patients, potentially overlooking subtle delirium in low-severity cases. Future research should focus on early risk stratification and prioritize four key areas: (1) expanded utilization of both tabular EHR and unstructured clinical notes; (2) development of integrated multimodal fusion models adaptable to dynamic patient states; (3) investigation of the temporal dynamics of delirium development using time-series analysis; and (4) application of causal inference methods to elucidate the relationships between risk factors and delirium. Superior prediction performance can be achieved by leveraging cutting-edge architectures (e.g. transformers) and parallel computing efficiencies to move beyond traditional machine learning. To enhance real-world adoption, future work should integrate Explainable AI tools such as Shapley Additive Explanations (SHAP) within EHR-based decision support systems, improving interpretability and mitigating subgroup disparities in localized risk assessment.

Ground subsidence in urban areas often reflects hidden failures within water and sewer infrastructure, amplified by hydrologic and seasonal conditions. This study analyzes 303 documented subsidence incidents in Gyeonggi Province, South Korea, from 2018 to 2024, focusing on infrastructure-related causes and their spatial and seasonal patterns. Incident records were reviewed to identify root causes, geographic distribution, seasonal trends, and impacts, including human injury and vehicle damage. Descriptive analysis showed that sewer pipe damage (39.3%) was the leading cause, followed by poor compaction or backfilling (22.8%) and excavation-related damage (14.2%). Subsidence linked to sewer systems occurred disproportionately during the summer monsoon, highlighting interactions between rainfall, pipe deterioration, and soil erosion. Statistical analysis using the Extended Fisher’s Exact Test revealed significant associations between subsidence causes and seasonality, vehicle damage, and regional location, but no significant link with human injury. Defective pipe construction contributed to elevated regional vulnerability, particularly in eastern municipalities, while excavation-related incidents were more common in spring. These results underscore the need for seasonally adaptive inspections, targeted rehabilitation of aging water and sewer networks, and region-specific asset management. By connecting subsurface failures with hydrologic conditions and infrastructure performance, this study offers data-driven insights to enhance proactive water infrastructure management and urban resilience.

A proper understanding of climatic consumption dynamics is critical for demand-side management, grid stability, and climate-sensitive residential energy planning. The non-linear interaction between meteorological conditions has necessitated intelligent predictive models. However, existing studies focused on machine learning (ML) models in their black-box nature, with limited interpretability of the impact of climatic-drivers, climatic-demand interactions, and hidden consumption regimes. This research fills this gap through an integrated framework that combines seasonal hypothesis-testing, k-means clustering, Shapley Additive exPlanations (SHAP)-based interpretability, and advanced predictive modeling using XGBoost, Random Forest (RF), long-short term memory (LSTM), Support Vector Machine (SVM), and Autoregressive Integrated moving average (ARIMA). The seasonal hypothesis-testing using ANOVA and Tukey’s HSD revealed statistically significant differences in energy consumption across seasons, with peak-demand during winter and summer extremes. SHAP-based feature ranking identified temperature and humidity as the most influential drivers of electricity-demand. The k-means clustering revealed three distinct groups/clusters, which reflect the climatic-consumption scenarios. The ensemble learning (RF and XGBoost) exhibited the lowest training-error. RF had the best training performance with MSE, RMSE, and MAE values of 1.6484, 1.2839, and 0.8480. The data-driven insights in this study provide useful intelligence that supports critical decision-making through a proper understanding of the residential climatic-consumption dynamics. • ANOVA and Tukey’s HSD revealed a significant difference in consumption across seasons • k-means clustering reveals 3 distinct climatic-consumption patterns and regimes • SHAP revealed temperature as a dominant consumption driver • The random forest gave the best training performance with RMSE = 1.2839 • The framework facilitates climate-sensitive demand-side management and planning.

Self-regulation is the basic capacity to regulate one's thoughts, emotions, and behaviors. The aim of the present study is to examine the psychometric properties of the Self-Regulation Scale (SRS) among male and female Portuguese youth participants (N = 559 youth, M = 16.51 years, SD = 1.07, range = 14-20 years). The three-factor model composed of the Emotional, Cognitive, and Behavioral regulation subscales obtained adequate fit, although the fit of the second-order model was also acceptable. Internal consistency as measured by the alpha and omega reliability estimators was good. Significant associations were found with psychometric measures of relevant constructs (e.g., self-control, psychopathic traits, criminogenic cognitions), and external criterion-related variables (e.g., engaging in activities against the law, alcohol and drug abuse). Cross-gender measurement invariance was supported, with females scoring significantly higher on the Cognitive regulation subscale, and males scoring significantly higher on the Emotional regulation subscale. The findings support the use of the SRS to validly and reliably measure self-regulation in Portuguese youth.

Spinal Cord Injury (SCI) rehabilitation is undergoing a transformative shift with the emergence of new treatment strategies. Historically, treatment options were limited, and few offered meaningful recovery. Recent work in human models has shown that neuromodulation specifically with spinal cord epidural stimulation (SCES) paired with task-specific training (TsT) can partially restore motor function such as the ability to stand, step, and perform volitional movements. Despite these advances, the recovery has been shown to plateau even with the combination of therapies. The recovery process typically leads to partial rather than complete restoration of function. This limitation arises because current approaches primarily reactivate existing circuits rather than repair the disrupted pathways. Scar tissue and loss of descending and ascending connections remain major barriers to full recovery, restricting the transmission of neural signals. We argue that the next phase of research should be a synergistic strategy building upon the successes of neuromodulation and TsT while incorporating a regenerative therapy such as stem-cell-based interventions. Whereas neuromodulation and task-specific training increases excitability and reorganizes existing networks, stem cells have the potential to repair structural damage and re-establish communication across injured regions or facilitating the establishment of dormant pathways. The future of SCI recovery relies on multi-modal synergistic interventions that are likely to maximize long-term functional outcomes. In the current perspective, we summarized the basic findings on applications of SCES on restoration of sensory-motor functions. We then projected on current interventions on utilizing stem cell therapy intervention. We highlighted the outcomes of randomized clinical trials, and the major barriers for considering the synergistic approach between SCES and stem cell intervention. We are hopeful that this perspective may lead to roundtable scientific discussion to bridge the gap on how to conduct numerous clinical trials in the field.

Abstract Background: Our understanding of the mechanisms of tumor immunosuppression in the tumor microenvironment (TME) of aggressive prostatic adenocarcinoma (PCa) at the single-cell level is limited. Integrated Visium-HD Spatial RNAseq analysis and 10X Flex sequencing of formalin-fixed paraffin-embedded (FFPE) tissue can provide unique insights into the mechanisms of immunosuppression in the ‘immune cold’ TME of PCa. Design: Visium-HD Spatial RNAseq analysis was performed on FFPE blocks from radical prostatectomy specimens of 12 patients of which six had lymph node metastases (LNM) and six did not. All specimens were also analyzed by 10X Flex sequencing, and data from each scRNAseq dataset were integrated with spatial RNAseq data. VisiumHD bins were processed with H&amp;E images to generate cell-level expression data using the bin2cell package, and cell types were annotated. Ligand-receptor interactions were identified using the CellWhisper algorithm. Results: Mean patient age was 63 years (range: 50-77 years). Gleason scores ranged from 7-10 (Grade group 3-5). Patients with LNM had more significant enrichment for gene signatures associated with inflammation and immune responses and high numbers of naïve regulatory T cells (Naïve Treg) near the invasive front of the aggressive tumor foci. The inflamed stroma adjacent to PCa in patients with LNM was enriched in expression of antigen-presenting genes, but also in immunosuppressive gene signatures. In addition, there was significant enrichment for spatial co-expression of the immunosuppressive receptor TIGIT and the NECTIN2, NECTIN3, and PVR ligands. TIGIT was expressed in CD4 and CD8 expressing cells and CD14+ monocytes, while NECTIN2, NECTIN3, and PVR were expressed in tumor cells, suggesting that this is one of the mechanisms of immunosuppression in aggressive PCa. Conclusions: High-definition spatial RNAseq analysis with integrated 10X Flex scRNAseq enabled identification of mechanisms of immunosuppression in the TME of PCa. Targeting TIGIT-NECTIN2/3 and/or TIGIT/PVR interactions may be a potential immunotherapeutic strategy for the management of patients with aggressive PCa. Citation Format: Jerry A. Ombor, Anurendra Kumar, Jatin Gandhi, Rakesh Shiradkar, Tilak Pathak, Cynthia L. Winham, Ismaheel O. Lawal, Olayinka A. Abiodun-Ojo, Lara Harik, Saurabh Sinha, Martin G. Sanda, David M. Schuster, Adeboye O. Osunkoya, Carlos S. Moreno. Integrated scRNAseq and spatial RNAseq analysis of aggressive prostatic adenocarcinoma identifies TIGIT-NECTIN 2/3 and TIGIT/PVR immunosuppressive interactions [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Prostate Cancer Research and Treatment; 2026 Jan 20-22; Philadelphia PA. Philadelphia (PA): AACR; Cancer Res 2026;86(2_Suppl):Abstract nr A046.