Zoish: A Novel Feature Selection Approach Leveraging Shapley Additive Values for Machine Learning Applications in Healthcare.

In the intricate landscape of healthcare analytics, effective feature selection is a prerequisite for generating robust predictive models, especially given the common challenges of sample sizes and potential biases. Zoish uniquely addresses these issues by employing Shapley additive values-an idea rooted in cooperative game theory-to enable both transparent and automated feature selection. Unlike existing tools, Zoish is versatile, designed to seamlessly integrate with an array of machine learning libraries including scikit-learn, XGBoost, CatBoost, and imbalanced-learn.The distinct advantage of Zoish lies in its dual algorithmic approach for calculating Shapley values, allowing it to efficiently manage both large and small datasets. This adaptability renders it exceptionally suitable for a wide spectrum of healthcare-related tasks. The tool also places a strong emphasis on interpretability, providing comprehensive visualizations for analyzed features. Its customizable settings offer users fine-grained control over feature selection, thus optimizing for specific predictive objectives.This manuscript elucidates the mathematical framework underpinning Zoish and how it uniquely combines local and global feature selection into a single, streamlined process. To validate Zoish's efficiency and adaptability, we present case studies in breast cancer prediction and Montreal Cognitive Assessment (MoCA) prediction in Parkinson's disease, along with evaluations on 300 synthetic datasets. These applications underscore Zoish's unparalleled performance in diverse healthcare contexts and against its counterparts.

Mental health outcomes of hospital staff during the COVID-19 pandemic in Iran.

BACKGROUND: The COVID-19 pandemic, which had recorded 769 million cases and resulted in 6.95 million deaths by August 2023, has put pressure on healthcare systems. Frontline medical professionals face stress, potentially leading to health challenges. This research aimed to examine the mental health of staff during the COVID-19 pandemic. METHODS: This cross-sectional descriptive-analytical study was conducted in several hospitals in Tehran, Kerman, and Golpayegan between 2021 and 2022. The study encompassed a population of 1,231 nurses and physicians. Data collection was done using the General Health Questionnaire-28 (GHQ-28). We applied the K-means clustering algorithm to unveil hidden patterns within the data and extract valuable insights from participants' responses to the GHQ-28. This method was chosen because our dataset lacked explicit labels, making grouping individuals with similar characteristics necessary. The primary aim was to delineate how the COVID-19 pandemic affected the mental health of hospital staff and identify which factors played a more significant role in this process. RESULTS: We have observed that Cluster two exhibits the highest scores in response to the GHQ-28 questions, indicating a more significant degree of mental distress. Within this cluster, 83.0% of individuals identify as female, 71.0% hold bachelor's degrees and 42.8% are nurses who have experienced the most substantial impact. Among these individuals, 90.4% did not have a history of smoking. Additionally, 59.7% are married, suggesting that these mental health issues may also affect their families. CONCLUSION: Given that the most critical subscale is related to anxiety/insomnia within the second cluster, it is necessary to implement management plans aimed at appropriately redistributing night shifts to improve employee health.

Comparison of habitual and meal-specific dietary patterns identified by latent class analysis and confirmatory factor analysis in adults: A cross-sectional study.

OBJECTIVE: The aim of this study was to compare characteristics of habitual and meal-specific dietary patterns identified by latent class analysis (LCA) and confirmatory factor analysis (CFA). METHODS: Participants included 778 adults selected for the present cross-sectional study from local health care centers in Tehran, Iran. Three 24-h dietary recalls evaluated dietary intake. LCA was used to group study participants into exclusive subgroups of individuals with similar patterns of dietary intake. CFA was applied to identify patterns of habitual and meal-specific dietary intake. Analysis of variance was used to compare the average scores of habitual and meal-specific CFA-derived dietary patterns across classes identified by LCA. RESULTS: Using habitual dietary intake, CFA grouped correlated food items into three major factors: fruits and vegetables, mixed, and Western dietary patterns. LCA grouped study participants with similar patterns of habitual intake into four subgroups of individuals: fruits and vegetables, mixed, Western, and low consumer classes. LCA-fruits and vegetables, LCA-Western, and LCA-mixed classes had, respectively, higher mean scores of CFA-fruits and vegetables, CFA-Western, and CFA-mixed dietary patterns compared with other classes (P < 0.001). Similar findings were observed for meal-specific dietary intake, where classes identified by LCA had the highest mean scores of their corresponding dietary pattern identified by CFA. CONCLUSION: Habitual and meal-specific classes identified by LCA were well characterized by the dietary patterns derived by CFA, suggesting that LCA may be an appropriate statistical approach to classify study participants with similar patterns of intake into exclusive subgroups of individuals.

Temporal patterns of energy intake identified by the latent class analysis in relation to prevalence of overweight and obesity in Iranian adults.

We aimed to identify temporal patterns of energy intake and investigate their association with adiposity. We performed a cross-sectional study of 775 adults in Iran. Information about eating occasions across the day was collected by three 24-h dietary recalls. Latent class analysis (LCA) was used to identify temporal eating patterns based on whether or not an eating occasion occurred within each hour of the day. We applied binary logistic regression to estimate the OR and 95 % CI of overweight and obesity (defined as BMI of 25-29·9 and ≥ 30 kg/m2, respectively) across temporal eating patterns while controlling for potential confounders. LCA grouped participants into three exclusive sub-groups named 'Conventional', 'Earlier breakfast' and 'Later lunch'. The 'Conventional' class was characterised by high probability of eating occasions at conventional meal times. 'Earlier breakfast' class was characterised by high probability of a breakfast eating occasion 1 h before the conventional pattern and a dinner eating occasion 1 h after the conventional pattern, and the 'Later lunch' class was characterised by a high probability of a lunch eating occasion 1 h after the conventional pattern. Participants in the 'Earlier breakfast' pattern had a lower likelihood of obesity (adjusted OR: 0·56, 95 % CI: 0·35, 0·95) as compared with the 'Conventional' pattern. There was no difference in the prevalence of obesity or overweight between participants in the 'Later lunch' and the 'Conventional' patterns. We found an inverse association between earlier eating pattern and the likelihood of obesity, but reverse causation may be a plausible explanation.