Performance Of Logistic Regression Model Research Articles

Radiological predictors of PCP in HIV-positive adults in South Africa: A matched case-control study.

Definition of chest X-ray (CXR) features associated with laboratory-confirmed pneumocystis pneumonia (PCP) among HIV-positive adults is needed to improve diagnosis in high-burden settings. Our primary objective was to identify CXR features associated with confirmed PCP diagnosis and severe PCP (defined by hypoxia, intensive care unit referral/admission, and/or in-hospital death). We also explored the performance of logistic regression models, incorporating selected clinical and CXR predictors, for PCP diagnosis and severe PCP. We conducted a case-control study involving HIV-positive adults with laboratory-confirmed PCP and a matched cohort with non-PCP respiratory presentations at regional hospitals in Cape Town, South Africa (2012-2020). Records from 104 adults (52 PCP cases and 52 non-PCP controls) were included. Diffuse versus patchy ground-glass opacification was associated with increased odds of PCP diagnosis (adjusted odds ratio [aOR]: 6.2, 95% confidence interval [CI]: 1.6-28.9, P = 0.01) and severe PCP (aOR: 4.5, 95% CI: 1.6-14.4, P = 0.008). Consolidation was associated with severe PCP (aOR: 3.3, 95% CI: 1.2-11.0, P = 0.03) as was increasing ground-glass zone involvement (aOR: 2.1 for each one-unit increase in involved zone; 95% CI: 1.4-3.2, P = 0.0004). Models incorporating hypoxia (hypoxia model) or tachypnoea (respiratory rate model) with diffuse ground-glass opacities, absence of pleural effusion or reticular/reticulonodular changes on CXR performed well in predicting PCP (area under the receiver operating characteristic curve 0.828 [hypoxia model] and 0.857 [respiratory rate model]). CXR evaluation alongside bedside clinical information offers good accuracy for discriminating definite PCP from other HIV-associated respiratory diseases.

Processing imbalanced medical data at the data level with assisted-reproduction data as an example

ObjectiveData imbalance is a pervasive issue in medical data mining, often leading to biased and unreliable predictive models. This study aims to address the urgent need for effective strategies to mitigate the impact of data imbalance on classification models. We focus on quantifying the effects of different imbalance degrees and sample sizes on model performance, identifying optimal cut-off values, and evaluating the efficacy of various methods to enhance model accuracy in highly imbalanced and small sample size scenarios.MethodsWe collected medical records of patients receiving assisted reproductive treatment in a reproductive medicine center. Random forest was used to screen the key variables for the prediction target. Various datasets with different imbalance degrees and sample sizes were constructed to compare the classification performance of logistic regression models. Metrics such as AUC, G-mean, F1-Score, Accuracy, Recall, and Precision were used for evaluation. Four imbalance treatment methods (SMOTE, ADASYN, OSS, and CNN) were applied to datasets with low positive rates and small sample sizes to assess their effectiveness.ResultsThe logistic model’s performance was low when the positive rate was below 10% but stabilized beyond this threshold. Similarly, sample sizes below 1200 yielded poor results, with improvement seen above this threshold. For robustness, the optimal cut-offs for positive rate and sample size were identified as 15% and 1500, respectively. SMOTE and ADASYN oversampling significantly improved classification performance in datasets with low positive rates and small sample sizes.ConclusionsThe study identifies a positive rate of 15% and a sample size of 1500 as optimal cut-offs for stable logistic model performance. For datasets with low positive rates and small sample sizes, SMOTE and ADASYN are recommended to improve balance and model accuracy.

Making a Better Logistic Regression Model Using Ensemble Learning

Abstract: Machine learning is a hot research topic in today's society and an important direction in the field of image classification research. Image classification is a supervised learning method used to classify images. There are many image classification algorithms that perform differently in different conditions. Paper focuses on improving the performance of image classification algorithms by using ensemble learning. Paper analyses the performance of logistic regression model and support vector machine model and then uses ensemble learning to increase the performance of logistic regression model. In the research work, both theoretical and empirical approaches were followed. For the theoretical approach a review of both secondary data as well as data based on results obtained by application on the tools is studied. Secondary data was acquired from the research articles, text books, journals, technical reports, published thesis, websites, e-journals, software tool manuals, conference proceedings and any other research articles published in the related domain. The empirical study was carried out on the set of experiments, using software tools. The results obtained from the experiments were analyzed for the finding of the research. The paper compares the results of three algorithms when tested on same dataset, in same environment and on same system. The final results of the research prove that ensemble techniques give best results and this type of learning is effective.

Using the Electronic Health Record to Develop a Gastric Cancer Risk Prediction Model

Background and aimsGastric cancer (GC) is a leading cause of cancer incidence and mortality globally. Population screening is limited by the low incidence and prevalence of GC in the United States. A risk prediction algorithm to identify high-risk patients allows for targeted GC screening. We aimed to determine the feasibility and performance of a logistic regression model based on electronic health records (EHR) to identify individuals at high risk for non-cardia gastric cancer (NCGC). MethodsWe included 614 patients who had a diagnosis of NCGC between ages 40-80 years and who were seen at our large tertiary medical center in multiple states between 2010 and 2021. Controls without a diagnosis of NCGC were randomly selected in a 1:10 ratio of cases to controls. Multiple imputation by chained equation for missing data followed by logistic regression on imputed datasets was used to estimate the probability of NCGC. Area under the curve (AUC) and the 0.632 estimator was used as the estimate for discrimination. ResultsThe 0.632 estimator value was 0.731, indicating robust model performance. Probability of NCGC was higher with increasing age (odds ratio [OR]=1.16, 95% confidence interval [CI]: 1.04 – 1.3), male sex (OR = 1.97; 95% CI: 1.64-2.36), Black (OR= 3.07; 95% CI: 2.46-3.83) or Asian race (OR=4.39; 95% CI: 2.60-7.42), tobacco use (OR=1.61; 95% CI: 1.34-1.94), anemia (OR=1.35; 95% CI: 1.09-1.68) and pernicious anemia (OR=6.12, 95% CI: 3.42-10.95). ConclusionWe demonstrate the feasibility and good performance of an EHR-based logistic regression model for estimating the probability of NCGC. Future studies will refine and validate this model, ultimately identifying a high-risk cohort who could be eligible for NCGC screening.

Statistical Approach to Enhancing Performance of Logistic Regression Model: Application to HIV/AIDS Data

This paper considered a robust method for modeling and predicting HIV/AIDS status of patients using logistic regression model enhanced with principal component analysis (PCA) and K-medians. In particular, the study developed a computational method for disease classification; and then identified key haematological predictors of HIV/AIDS status. Based on quantitative research design, the utility of the methods is exemplified using real HIV/AIDS data obtained from a polyclinic in the Greater Accra region of Ghana. The data consists of one hundred and fifty (150) patients, eighty (80) of whom are known to have tested positive for HIV/AIDS. The study findings revealed that enhancement in predictive accuracy for a logistic regression is possible by means of incorporating PCA and K-Medians with robust centers. Model 5 was found to be the best predictor of HIV/AIDS status of a patient. It is an integration of both robust principal component analysis and K-Medians clustering into a binary logistic regression model. Its predictive accuracy is over 93%, and with 98% probability per the ROC criterion. The study thus recommends the incorporation of both RPCA and K-Medians with robust centers into binary logistic regression model to enhance its predictive performance.

Comparison of the Performance of Logistic Regression Model in the Presence and Absence of Mediation

Over the last decade major global efforts mounted to address the HIV epidemic has realized notable successes in combating the pandemic. Sub Saharan Africa (SSA) still remains a global epicenter of the disease, accounting for more than 70% of the global burden of infections. Despite wide spread use of various intervention strategies that act as mediation factors in Human Immunodeficiency Virus (HIV) prevention, HIV prevalence still remains a challenge especially in some geographic areas and populations. Therefore, how mediation factors interact with the prevailing HIV risk factors to cause an impact on its prevalence remains a question not answered. This study considered Exposure to HIV related media as a mediator variable in the relationship between HIV risk factors and HIV prevalence. Two logistic regression models, one in presence of mediation and another in absence of mediation were formulated and compared to establish the best performing model. Models were fitted to real data from the Kenya Population-based HIV Impact Assessment survey-2018 and model parameters were estimated using Maximum Likelihood Estimation in R. Results based on both Akaike’s Information Criterion and the McFadden’s R2 value revealed that the model formulated in presence of mediation performed better compared to that without mediation.

Anti-PLA2R Antibody Levels and Clinical Risk Factors for Treatment Nonresponse in Membranous Nephropathy.

The 2021 Kidney Disease Improving Global Outcomes (KDIGO) guidelines recommend following anti-phospholipase A2 receptor (PLA2R) antibody levels as a marker of treatment response in membranous nephropathy; however, the optimal timing to evaluate antibody levels and how to combine them with other clinical variables are currently unknown. We used a cohort of 85 patients from the Membranous Nephropathy Trial Of Rituximab (MENTOR) with anti-PLA2R antibodies ≥14 RU/ml to identify risk factors for not experiencing proteinuria remission after 12 months of treatment with cyclosporine or rituximab. Three landmark times were considered: at baseline and after 3 and 6 months of treatment. Logistic regression model performance was evaluated using C-statistics and model fit (Akaike information criterion [AIC], R 2 ). The model at baseline that best predicted no remission included anti-PLA2R antibodies >323 RU/ml and creatinine clearance; the best model after 3 months included the change from baseline in both antibody and albumin levels; and the best model after 6 months included antibody levels >14 RU/ml, creatinine clearance, and the change from baseline in albumin. Compared with the model at baseline, the model at 3 months had better model fit (AIC 70.9 versus 96.4, R 2 51.8% versus 30.1%) and higher C-statistic (0.93 versus 0.83, P = 0.008). The model at 6 months had no difference in performance compared with the model at 3 months (AIC 68.6, R 2 53.0%, C-statistic 0.94, P = 0.67). In patients with membranous nephropathy treated with cyclosporine or rituximab in the MENTOR trial, we found that the optimal method to evaluate risk factors for the probability of treatment response was to use anti-PLA2R antibody levels combined with albumin levels after 3 months of treatment, which was significantly better than using antibody levels alone or risk factor evaluation at baseline, with no added benefit of waiting until 6 months of treatment. This article contains a podcast at https://dts.podtrac.com/redirect.mp3/www.asn-online.org/media/podcast/CJASN/2023_10_09_CJN0000000000000237.mp3.

Paper 42: A Novel Machine Learning (ML) Algorithm to Predict Outcomes after Revision ACLR (rACLR) in the Multicenter Anterior Cruciate Ligament Reconstruction Study (MARS) Cohort

Objectives: The multicenter anterior cruciate ligament reconstruction study (MARS) Group has implicated several factors that influence revision anterior cruciate ligament (rACLR) outcomes. As machine learning (ML) becomes increasingly utilized in the orthopaedic literature, the application of ML methodology to MARS cohort data presents a valuable opportunity to translate data into patient-specific insights. This study sought to apply novel ML methodology to the MARS cohort data in order to determine 1) an optimal predictive model of rACLR graft failure and 2) features that are most predictive of rACLR graft failure in context of the predictive model. Methods: The MARS Group of 83 surgeons and 52 sites prospectively enrolled a cohort of patients undergoing rACLR. Patients were followed up at 2 and 6 years postoperatively regarding patient-reported outcomes (PROs), additional surgeries, and incidence of graft failure. Surgeon reported intraoperative findings and preoperative radiographic measures were obtained. Data was preprocessed and 381 demographic, clinical, and surgical features were used to build five machine learning models predictive of graft failure at 6 years postoperatively. Models used included traditional logistic regression (LR), XGBoost, Gradient Boosting, Random Forest, and a validated ensemble algorithm, AutoPrognosis (AP). Validated performance metrics for binary outcome prediction models were used to determine discriminative power and calibration. Individual feature importance was calculated for the highest performing models using partial dependence and perturbation-based feature importance. Results: The cohort included 831 patients who completed six year follow up, and 5.8% (n=48) of whom experienced graft failure. While all models had moderate to good concordance, AP demonstrated the highest discriminative power compared to other models (Model: AUROC | AP: 0.722 | Random Forest: 0.621 | Gradient Boosting: 0.654 | XGBoost: 0.690 | Logistic Regression: 0.630). The AP model was well-calibrated, with calibration scores similar to the other studied models (Model: Brier Score | AP: 0.054 | Random Forest: 0.054 | Gradient Boosting: 0.059 | XGBoost: 0.059 | Logistic Regression: 0.102). Features deemed important for AP differ from those for LR model performance. For AP, partial dependence feature importance calculations demonstrated the following top five contributors to model predictive ability: surgeon years of experience, prior femoral tunnel position measured via preoperative radiograph, compromised prior ACLR femoral tunnel position and size, and baseline patient age. Similarly, AP perturbation-based calculations showed the following top five features that contribute most to the AUROC performance: baseline patient age, current ACLR graft type, years since previous ACLR, prior ACLR femoral tunnel position, and prior tibial tunnel position on sagittal view preoperative radiographs. Perturbation-based calculations showed the following top five features that contribute most to the AUPRC performance: current ACLR graft type, baseline patient age, prior tibial tunnel position on sagittal view preoperative radiographs, previous ipsilateral medial meniscus repair, and baseline SF-36 subscale scores for patient vitality. Other leading contributors to model predictive ability included baseline PROs including SF-36 subscale scores for physical function, mental health and composite mental health, KOOS quality of life, WOMAC pain, and MARX activity rating scores. Conclusions: Of the studied models in this preliminary analysis, AP appears to most accurately predict rACLR graft failure. These findings build on prior studies, identifying key surgical, clinical, radiographic, and patient-reported factors contributing to the model’s ability to predict rACLR graft failure at 6 years postoperatively. Further work includes creation of a clinical risk calculator using important feature inputs, and the use of survival modeling to determine risk scores at various postoperative time points.

The role of clinical phenotypes in decisions to limit life-sustaining treatment for very old patients in the ICU

BackgroundLimiting life-sustaining treatment (LST) in the intensive care unit (ICU) by withholding or withdrawing interventional therapies is considered appropriate if there is no expectation of beneficial outcome. Prognostication for very old patients is challenging due to the substantial biological and functional heterogeneity in that group. We have previously identified seven phenotypes in that cohort with distinct patterns of acute and geriatric characteristics. This study investigates the relationship between these phenotypes and decisions to limit LST in the ICU.MethodsThis study is a post hoc analysis of the prospective observational VIP2 study in patients aged 80 years or older admitted to ICUs in 22 countries. The VIP2 study documented demographic, acute and geriatric characteristics as well as organ support and decisions to limit LST in the ICU. Phenotypes were identified by clustering analysis of admission characteristics. Patients who were assigned to one of seven phenotypes (n = 1268) were analysed with regard to limitations of LST.ResultsThe incidence of decisions to withhold or withdraw LST was 26.5% and 8.1%, respectively. The two phenotypes describing patients with prominent geriatric features and a phenotype representing the oldest old patients with low severity of the critical condition had the largest odds for withholding decisions. The discriminatory performance of logistic regression models in predicting limitations of LST after admission to the ICU was the best after combining phenotype, ventilatory support and country as independent variables.ConclusionsClinical phenotypes on ICU admission predict limitations of LST in the context of cultural norms (country). These findings can guide further research into biases and preferences involved in the decision-making about LST.Trial registration Clinical Trials NCT03370692 registered on 12 December 2017.

Predicting lymph node metastasis from primary tumor histology and clinicopathologic factors in colorectal cancer using deep learning

BackgroundPresence of lymph node metastasis (LNM) influences prognosis and clinical decision-making in colorectal cancer. However, detection of LNM is variable and depends on a number of external factors. Deep learning has shown success in computational pathology, but has struggled to boost performance when combined with known predictors.MethodsMachine-learned features are created by clustering deep learning embeddings of small patches of tumor in colorectal cancer via k-means, and then selecting the top clusters that add predictive value to a logistic regression model when combined with known baseline clinicopathological variables. We then analyze performance of logistic regression models trained with and without these machine-learned features in combination with the baseline variables.ResultsThe machine-learned extracted features provide independent signal for the presence of LNM (AUROC: 0.638, 95% CI: [0.590, 0.683]). Furthermore, the machine-learned features add predictive value to the set of 6 clinicopathologic variables in an external validation set (likelihood ratio test, p < 0.00032; AUROC: 0.740, 95% CI: [0.701, 0.780]). A model incorporating these features can also further risk-stratify patients with and without identified metastasis (p < 0.001 for both stage II and stage III).ConclusionThis work demonstrates an effective approach to combine deep learning with established clinicopathologic factors in order to identify independently informative features associated with LNM. Further work building on these specific results may have important impact in prognostication and therapeutic decision making for LNM. Additionally, this general computational approach may prove useful in other contexts.

Application of Artificial Intelligence in Assessing the Self-Management Practices of Patients with Type 2 Diabetes.

The use of Artificial intelligence in healthcare has evolved substantially in recent years. In medical diagnosis, Artificial intelligence algorithms are used to forecast or diagnose a variety of life-threatening illnesses, including breast cancer, diabetes, heart disease, etc. The main objective of this study is to assess self-management practices among patients with type 2 diabetes in rural areas of Pakistan using Artificial intelligence and machine learning algorithms. Of particular note is the assessment of the factors associated with poor self-management activities, such as non-adhering to medications, poor eating habits, lack of physical activities, and poor glycemic control (HbA1c %). The sample of 200 participants was purposefully recruited from the medical clinics in rural areas of Pakistan. The artificial neural network algorithm and logistic regression classification algorithms were used to assess diabetes self-management activities. The diabetes dataset was split 80:20 between training and testing; 80% (160) instances were used for training purposes and 20% (40) instances were used for testing purposes, while the algorithms' overall performance was measured using a confusion matrix. The current study found that self-management efforts and glycemic control were poor among diabetes patients in rural areas of Pakistan. The logistic regression model performance was evaluated based on the confusion matrix. The accuracy of the training set was 98%, while the test set's accuracy was 97.5%; each set had a recall rate of 79% and 75%, respectively. The output of the confusion matrix showed that only 11 out of 200 patients were correctly assessed/classified as meeting diabetes self-management targets based on the values of HbA1c < 7%. We added a wide range of neurons (32 to 128) in the hidden layers to train the artificial neural network models. The results showed that the model with three hidden layers and Adam's optimisation function achieved 98% accuracy on the validation set. This study has assessed the factors associated with poor self-management activities among patients with type 2 diabetes in rural areas of Pakistan. The use of a wide range of neurons in the hidden layers to train the artificial neural network models improved outcomes, confirming the model's effectiveness and efficiency in assessing diabetes self-management activities from the required data attributes.

A Comparative Study of Shallow Machine Learning Models and Deep Learning Models for Landslide Susceptibility Assessment Based on Imbalanced Data

A landslide is a type of geological disaster that poses a threat to human lives and property. Landslide susceptibility assessment (LSA) is a crucial tool for landslide prevention. This paper’s primary objective is to compare the performances of conventional shallow machine learning methods and deep learning methods in LSA based on imbalanced data to evaluate the applicability of the two types of LSA models when class-weighted strategies are applied. In this article, logistic regression (LR), random forest (RF), deep fully connected neural network (DFCNN), and long short-term memory (LSTM) neural networks were employed for modeling in the Zigui-Badong area of the Three Gorges Reservoir area, China. Eighteen landslide influence factors were introduced to compare the performance of four models under a class balanced strategy versus a class imbalanced strategy. The Spearman rank correlation coefficient (SRCC) was applied for factor correlation analysis. The results reveal that the elevation and distance to rivers play a dominant role in LSA tasks. It was observed that DFCNN (AUC = 0.87, F1-score = 0.60) and LSTM (AUC = 0.89, F1-score = 0.61) significantly outperformed LR (AUC = 0.89, F1-score = 0.50) and RF (AUC = 0.88, F1-score = 0.50) under the class imbalanced strategy. The RF model achieved comparable outcomes (AUC = 0.90, F1-score = 0.61) to deep learning models under the class balanced strategy and ran at a faster training speed (up to 63 times faster than deep learning models). The LR model performance was inferior to that of the other three models under the balanced strategy. Meanwhile, the deep learning models and the shallow machine learning models showed significant differences in susceptibility spatial patterns. This paper’s findings will aid researchers in selecting appropriate LSA models. It is also valuable for land management policy making and disaster prevention and mitigation.

Detection of signs of disease in external photographs of the eyes via deep learning.

Retinal fundus photographs can be used to detect a range of retinal conditions. Here we show that deep-learning models trained instead on external photographs of the eyes can be used to detect diabetic retinopathy (DR), diabetic macular oedema and poor blood glucose control. We developed the models using eye photographs from 145,832 patients with diabetes from 301 DR screening sites and evaluated the models on four tasks and four validation datasets with a total of 48,644 patients from 198 additional screening sites. For all four tasks, the predictive performance of the deep-learning models was significantly higher than the performance of logistic regression models using self-reported demographic and medical history data, and the predictions generalized to patients with dilated pupils, to patients from a different DR screening programme and to a general eye care programme that included diabetics and non-diabetics. We also explored the use of the deep-learning models for the detection of elevated lipid levels. The utility of external eye photographs for the diagnosis and management of diseases should be further validated with images from different cameras and patient populations.

Predictions of attrition among US Marine Corps: Comparison of four predictive methods

ABSTRACT The present study compared the performance of logistic regression models with that of machine learning classification models (classification trees and random forests) in the context of predicting training attrition from the Delayed Enlistment Program in the United States Marine Corps (USMC) with scores from the Tailored Adaptive Personality Assessment System (TAPAS). Performance was assessed according to the type of misclassification error and across a variety of different reasons for attrition. The base rate of attrition was low, which impeded the training process, but the machine learning models outperformed logistic regression in predicting voluntary attrition in a stratified 50% attrition sample.

Prediction of Major Adverse Events After Endovascular Aneurysm Repair Using a Machine Learning Model

Predicting major adverse events (MAE), including type I and III endoleaks (ELs), after endovascular aneurysm repair (EVAR) is essential for clinical decision-making. No clinical decision instruments can reliably predict for MAE after EVAR. We developed a machine learning model (MLM) for predicting MAE within 3 years after elective EVAR using preoperative demographic information and aneurysm geometry. MLMs infer complex interaction without prespecification, possibly increasing the prediction accuracy compared with logistic regression models (LRMs). Patients with intact infrarenal abdominal aortic aneurysms who had undergone EVAR with preoperative M2S reconstruction from 2010 to 2017 at a single institution were analyzed. The MAEs included aneurysm- or intervention-related death within 30 days, aneurysm rupture, the development of type I or III ELs, greater than three reinterventions, or EVAR explantation. Patients with <3 years of follow-up were excluded. Seventy-four preoperative predictors were evaluated. Of these, 30 were geometric factors derived from M2S models. The remaining factors were demographic. The performance of LRMs, random forest models (RFMs), and gradient-boosted models (GBMs) were evaluated using a 75% and 25% training and testing split, fivefold cross-validation, and receiver operating characteristic performance metric. A total of 250 patients were included (mean age, 71.9 ± 8.6 years; 86% male). The abdominal aortic aneurysm diameter was 56.0 ± 10.6 mm, and the infrarenal neck length was 24.6 ± 4.5 mm. Of the 250 patients, 40 (16%) had experienced MAE: 11 had died within 30 days, 3 had experienced aneurysm rupture, 10 had developed a type I EL, 10 had developed a type III EL, 10 had required graft explantation, and 4 had undergone four or more reinterventions. The most significant 45 predictive variables are shown in the Fig. All discrimination and calibration metrics favored the MLM over the LRM. The area under the curve for the LRM was 0.70 compared with 0.83 for the RFM and 0.77 for the GBM. The LRM classification accuracy was 68% compared with 84% for the RFM and GBM. The GBM showed superior calibration performance (Hosmer-Lemeshow P = .201). The removal of the geometric parameters significantly degraded the performance for all the models (likelihood ratio P < .001). The MLM displayed superior discrimination and calibration compared with the LRM in predicting MAE for the first 3 years after elective EVAR, possibly owing to the complex interaction terms in the predictive factors. The anatomic characteristics were more influential than the demographic data in predicting for MAE. Our MLM might predict for MAE after EVAR and thereby aid in selecting patients for EVAR, nonoperative observation, or open repair.

High Cell-Free DNA Integrity Is Associated with Poor Breast Cancer Survival.

Simple SummaryA recent point of focus in breast cancer (BC) research has been the utilization of cell-free DNA and its concentration (cfDConc) and integrity (cfDI) as potential biomarkers. Though the association of cfDConc and BC survival is already recognized, studies on the prognostic value of cfDI have had contradictory results. The aim of this study was to investigate the prognostic potential of cfDConc and cfDI in Eastern Finnish BC cases with a non-metastatic disease. While the prognostic value of cfDConc remained non-significant in our analyses, high cfDI was an independent prognostic factor for poor overall survival (OS) and breast cancer-specific survival (BCSS). Inclusion of cfDI in the multivariate logistic regression model improved the predictive performance of the model, thus suggesting that the combined use of traditional tumor features and liquid biopsy could help to discriminate BC patients with poor OS and BCSS more accurately at the time of diagnosis.Background: A recent point of focus in breast cancer (BC) research has been the utilization of cell-free DNA (cfDNA) and its concentration (cfDConc) and integrity (cfDI) as potential biomarkers. Though the association of cfDConc and poor survival is already recognized, studies on the prognostic value of cfDI have had contradictory results. Here, we provide further evidence to support the use of cfDI as a potential biomarker. Methods: We selected 204 Eastern Finnish BC cases with non-metastatic disease and isolated cfDNA from the serum collected at the time of diagnosis before any treatment was given. The cfDConc and cfDI were measured with a fluorometer and electrophoresis and analyzed with 25 years of survival data. Results: High cfDConc was not an independent prognostic factor in our analyses while high cfDI was found to be an independent prognostic factor for poor OS (p = 0.020, hazard ratio (HR) = 1.57, 95% confidence interval (CI) 1.07–2.29, Cox) and BCSS (p = 0.006, HR = 1.93, 95% CI 1.21–3.08)). Inclusion of cfDI in the multivariate logistic regression model improved the predictive performance. Conclusions: Our results show high cfDI is an independent prognostic factor for poor OS and BCSS and improves the predictive performance of logistic regression models, thus supporting its prognostic potential.

P120. Risk factors for failure to achieve minimal clinically important difference following cervical disc replacement

<h3>BACKGROUND CONTEXT</h3> Patient reported outcome measures (PROMs) and minimal clinically importance difference (MCID) thresholds have emerged as effective, reliable parameters to assess improvement and surgical success. <h3>PURPOSE</h3> To investigate the risk factors associated with failure to achieve MCID following cervical disc replacement (CDR). <h3>STUDY DESIGN/SETTING</h3> Retrospective review of prospectively collected data. <h3>PATIENT SAMPLE</h3> A total of 107 patients who had undergone primary, one-or-two level CDR by three surgeons at a single academic institution. <h3>OUTCOME MEASURES</h3> Patient-reported outcome measures - Rates of MCID achievement in neck disability index (NDI), VAS (visual analogue score) neck, and VAS-arm were calculated with the following threshold values, based on previously published studies: NDI: -15.0, VAS-neck: -2.5, and VAS-arm: -2.5. <h3>METHODS</h3> Risk factors for failure to achieve MCID were identified using multivariate logistic regression models with alternating backward stepwise elimination and forward entry of removed variables (entry: 0.05, removal: 0.10). Receiver operating characteristic (ROC) curves were generated to assess the performance of each model. <h3>RESULTS</h3> A total of 107 patients met inclusion criteria (mean age: 42.99±9. Thirteen years, mean BMI: 26.02±4.01 kg/m2). There were 99 patients who presented for a follow-up visit within three months of their operation, of which 82 (89.1%) had achieved MCID in NDI, VAS-neck, or VAS-arm. Specific to the NDI, BMI≥25 kg/m2 (OR:153.75, p=0.002), depression (OR:50.32, p=0.016), myelopathy (OR:51.44, p=0.022), degenerative disc disease/spondylosis (OR:21.54, p=0.001), and two-level procedures (OR:6.59, p=0.027) were identified as risk factors for failure to achieve MCID within the first three months. Preoperative narcotic use (OR:0.11, p=0.027) was associated with a lower risk for failure to achieve MCID in NDI within the first three months. The area under the ROC curve (AUROC) assessing the logistic regression model's performance was 0.873 (p<0.001), suggesting strong predictive ability. Overall, 103 patients had valid MCID calculations, of which 93 (90.3%) achieved MCID in NDI, VAS-neck, or VAS-arm at any time point (2 weeks, 6 weeks, 3 months, 6 months, or 1 year) following CDR. Depression (OR:18.21, p=0.022), myelopathy (OR:18.81, p=0.037), and two-level procedures were associated with failure to achieve MCID in NDI by the 1-year time point. The AUROC assessing model performance was 0.840 (p<0.001), demonstrating strong predictive ability. No significant risk factors were identified for failure to achieve MCID in VAS-neck and VAS-arm. <h3>CONCLUSIONS</h3> The large majority of patients had achieved MCID in at least one of the three PROMs. Clinical depression, myelopathy and two-level procedures were associated with failure to achieve MCID in NDI by one year and also within the early postoperative period (first three months) following CDR. BMI≥ 25 kg/m2 and spondylosis were also significant factors associated with failure to achieve MCID in NDI within the early postoperative period. Although preoperative narcotic use was associated with achievement of MCID in NDI, these patients likely presented initially with more chronic and severe symptoms, potentiating greater improvements in PROMs. Future studies with larger cohorts may provide additional insight on the true impact of the identified risk factors on outcomes following CDR. <h3>FDA DEVICE/DRUG STATUS</h3> This abstract does not discuss or include any applicable devices or drugs.

Vertebral MRI-based radiomics model to differentiate multiple myeloma from metastases: influence of features number on logistic regression model performance.

This study aimed to use the most frequent features to establish a vertebral MRI-based radiomics model that could differentiate multiple myeloma (MM) from metastases and compare the model performance with different features number. We retrospectively analyzed conventional MRI (T1WI and fat-suppression T2WI) of 103 MM patients and 138 patients with metastases. The feature selection process included four steps. The first three steps defined as conventional feature selection (CFS), carried out 50 times (ten times with 5-fold cross-validation), included variance threshold, SelectKBest, and least absolute shrinkage and selection operator. The most frequent fixed features were selected for modeling during the last step. The number of events per independent variable (EPV) is the number of patients in a smaller subgroup divided by the number of radiomics features considered in developing the prediction model. The EPV values considered were 5, 10, 15, and 20. Therefore, we constructed four models using the top 16, 8, 6, and 4 most frequent features, respectively. The models constructed with features selected by CFS were also compared. The AUCs of 20EPV-Model, 15EPV-Model, and CSF-Model (AUC = 0.71, 0.81, and 0.78) were poor than 10EPV-Model (AUC = 0.84, p < 0.001). The AUC of 10EPV-Model was comparable with 5EPV-Model (AUC = 0.85, p = 0.480). The radiomics model constructed with an appropriate small number of the most frequent features could well distinguish metastases from MM based on conventional vertebral MRI. Based on our results, we recommend following the 10 EPV as the rule of thumb for feature selection. • The developed radiomics model could distinguish metastases from multiple myeloma based on conventional vertebral MRI. • An accurate model based on just a handful of the most frequent features could be constructed by utilizing multiple feature reduction techniques. • An event per independent variable value of 10 is recommended as a rule of thumb for modeling feature selection.

Perihematomal Edema and Clinical Outcome in Intracerebral Hemorrhage Related to Different Oral Anticoagulants.

Background: There is a need to examine the effects of different types of oral anticoagulant-associated intracerebral hemorrhage (OAC-ICH) on perihematomal edema (PHE), which is gaining considerable appeal as a biomarker for secondary brain injury and clinical outcome. Methods: In a large multicenter approach, computed tomography-derived imaging markers for PHE (absolute PHE, relative PHE (rPHE), edema expansion distance (EED)) were calculated for patients with OAC-ICH and NON-OAC-ICH. Exploratory analysis for non-vitamin-K-antagonist OAC (NOAC) and vitamin-K-antagonists (VKA) was performed. The predictive performance of logistic regression models, employing predictors of poor functional outcome (modified Rankin scale 4–6), was explored. Results: Of 811 retrospectively enrolled patients, 212 (26.14%) had an OAC-ICH. Mean rPHE and mean EED were significantly lower in patients with OAC-ICH compared to NON-OAC-ICH, p-value 0.001 and 0.007; whereas, mean absolute PHE did not differ, p-value 0.091. Mean EED was also significantly lower in NOAC compared to NON-OAC-ICH, p-value 0.05. Absolute PHE was an independent predictor of poor clinical outcome in NON-OAC-ICH (OR 1.02; 95%CI 1.002–1.028; p-value 0.027), but not in OAC-ICH (p-value 0.45). Conclusion: Quantitative markers of early PHE (rPHE and EED) were lower in patients with OAC-ICH compared to those with NON-OAC-ICH, with significantly lower levels of EED in NOAC compared to NON-OAC-ICH. Increase of early PHE volume did not increase the likelihood of poor outcome in OAC-ICH, but was independently associated with poor outcome in NON-OAC-ICH. The results underline the importance of etiology-specific treatment strategies. Further prospective studies are needed.

Mortality Prediction of the CHA2DS2-VASc Score, the HAS-BLED Score, and Their Combination in Anticoagulated Patients with Atrial Fibrillation.

Background and Objectives: Atrial fibrillation (AF) is associated with increased mortality, predictors of which are poorly characterized. We investigated the predictive power of the commonly used CHA2DS2-VASc score (congestive heart failure, hypertension, age ≥ 75 years [doubled], diabetes, stroke/transient ischemic attack/thromboembolism [doubled], vascular disease [prior myocardial infarction, peripheral artery disease, or aortic plaque], age 65–75 years, sex category [female]), the HAS-BLED score (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio [INR], elderly [age ≥ 65 years], drugs/alcohol concomitantly), and their combination for mortality in AF patients. Methods: The PREvention oF thromboembolic events—European Registry in Atrial Fibrillation (PREFER in AF) was a prospective registry including AF patients across seven European countries. We used logistic regression to analyze the relationship between the CHA2DS2-VASc and HAS-BLED scores and outcomes, including mortality, at one year. We evaluated the performance of logistic regression models by discrimination measures (C-index and DeLong test) and calibration measures (Hosmer and Lemeshow goodness-of-fit and integrated discrimination improvement (IDI), with bootstrap techniques for internal validation. Results: In 5209 AF patients with complete information on both scores, average one-year mortality was 3.1%. We found strong gradients between stroke/systemic embolic events (SSE), major bleeding and—specifically—mortality for both CHA2DS2-VASc and HAS-BLED scores, with a similar C-statistic for event prediction. The predictive power of the models with both scores combined, removing overlapping components, was significantly enhanced (p < 0.01) compared to models including either CHA2DS2-VASc or HAS-BLED alone: for mortality, C-statistic: 0.740, compared to 0.707 for CHA2DS2-VASc or 0.646 for HAS-BLED alone. IDI analyses supported the significant improvement for the combined score model compared to separate score models for all outcomes. Conclusions: Both the CHA2DS2-VASc and the HAS-BLED scores predict mortality similarly in patients with AF, and a combination of their components increases prediction significantly. Such combination may be useful for investigational and—possibly—also clinical purposes.