The use of conditional cutoffs in a forward selection procedure

Information for evaluating the statistical significance of stepwise regression models developed with a forward selection procedure is presented. Cumulative distributions of the adjusted coefficient of determination ([Formula: see text]) under the null hypothesis of no relationship between the dependent variable and m potential independent variables are derived from a Monté Carlo simulation study. The study design included sample sizes of 25, 50, and 100, available independent variables of 10, 20, and 40, and three criteria for including variables in the regression model. The results reveal that the biases involved in testing statistical significance by two well-known rules are very large, thus demonstrating the desirability of using the Monté Carlo cumulative [Formula: see text] distributions developed by the authors. Although the results were derived under the assumption of uncorrelated predictors, the authors show that the results continue to be useful for the correlated predictor case.

Download Full-text

Shunt Failure—The First 30 Days

Neurosurgery ◽

10.1093/neuros/nyz379 ◽

2019 ◽

Vol 87 (1) ◽

pp. 123-129

Author(s):

Ryan P Lee ◽

Sonia Ajmera ◽

Fridtjof Thomas ◽

Pooja Dave ◽

Jock C Lillard ◽

...

Keyword(s):

Selection Procedure ◽

Research Database ◽

Shunt Failure ◽

Shunt Surgery ◽

Forward Selection ◽

Multivariable Logistic Regression Model ◽

Shunt System ◽

Index Model ◽

Surgery Research ◽

Significant Covariates

Abstract BACKGROUND Incontrovertible predictors of shunt malfunction remain elusive. OBJECTIVE To determine predictors of shunt failure within 30 d of index surgery. METHODS This was a single-center retrospective cohort study from January 2010 through November 2016. Using a ventricular shunt surgery research database, clinical and procedural variables were procured. An “index surgery” was defined as implantation of a new shunt or revision or augmentation of an existing shunt system. The primary outcome was shunt failure of any kind within the first 30 days of index surgery. Bivariate models were created, followed by a final multivariable logistic regression model using a backward-forward selection procedure. RESULTS Our dataset contained 655 unique patients with a total of 1206 operations. The median age for the cohort at the time of first shunt surgery was 4.6 yr (range, 0-28; first and third quartile, .37 and 11.8, respectively). The 30-day failure rates were 12.4% when analyzing the first-index operation only (81/655), and 15.7% when analyzing all-index operations (189/1206). Small or slit ventricles at the time of index surgery and prior ventricular shunt operations were found to be significant covariates in both the “first-index” (P < .01 and P = .05, respectively) and “all-index” (P = .02 and P < .01, respectively) multivariable models. Intraventricular hemorrhage at the time of index surgery was an additional predictor in the all-index model (P = .01). CONCLUSION This study demonstrates that only 3 variables are predictive of 30-day shunt failure when following established variable selection procedures, 2 of which are potentially under direct control of the surgeon.

Download Full-text

Conditional p-values for the F-statistic in a forward selection procedure

Computational Statistics & Data Analysis ◽

10.1016/0167-9473(94)00047-m ◽

1995 ◽

Vol 20 (3) ◽

pp. 239-263 ◽

Cited By ~ 8

Author(s):

Eugene Grechanovsky ◽

Ilia Pinsker

Keyword(s):

Selection Procedure ◽

Forward Selection ◽

P Values

Download Full-text

Factors and Predictors of Length of stay in Offenders diagnosed with schizophrenia- A Machine-learning-based approach

10.21203/rs.3.rs-15817/v3 ◽

2020 ◽

Author(s):

Johannes Kirchebner ◽

Moritz Günther ◽

Martina Sonnweber ◽

Alice King ◽

Steffen Lau

Keyword(s):

Machine Learning ◽

Length Of Stay ◽

Learning Algorithms ◽

Evaluation Criteria ◽

Selection Procedure ◽

Legal Aspects ◽

Influential Factors ◽

Machine Learning Algorithms ◽

Forward Selection ◽

Factors Affecting

Abstract Background: Prolonged forensic psychiatric hospitalizations have raised ethical, economic, and clinical concerns. Due to the confounded nature of factors affecting length of stay of psychiatric offender patients, prior research has called for the application of a new statistical methodology better accommodating this data structure. The present study attempts to investigate factors contributing to long-term hospitalization of schizophrenic offenders referred to a Swiss forensic institution, using machine learning algorithms that are better suited than conventional methods to detect nonlinear dependencies between variables. Methods: In this retrospective file and registry study, multidisciplinary notes of 143 schizophrenic offenders were reviewed using a structured protocol on patients’ characteristics, criminal and medical history and course of treatment. Via a forward selection procedure, the most influential factors for length of stay were preselected. Machine learning algorithms then identified the most efficient model for predicting length-of-stay. Results: Two factors have been identified as being particularly influential for a prolonged forensic hospital stay, both of which are related to aspects of the index offense, namely (attempted) homicide and the extent of the victim's injury. The results are discussed in light of previous research on this topic. Conclusions: In this study, length of stay was determined by legal considerations, but not by factors that can be influenced therapeutically. Results emphasize that forensic risk assessments should be based on different evaluation criteria and not merely on legal aspects.

Download Full-text

Prognosis of hepatocellular carcinoma (HCC): Comparison of four staging systems in two French clinical trials

Journal of Clinical Oncology ◽

10.1200/jco.2007.25.18_suppl.4589 ◽

2007 ◽

Vol 25 (18_suppl) ◽

pp. 4589-4589 ◽

Cited By ~ 2

Author(s):

S. Collette ◽

F. Bonnetain ◽

X. Paoletti ◽

M. Doffoel ◽

O. Bouche ◽

...

Keyword(s):

Predictive Accuracy ◽

Cox Model ◽

External Validation ◽

Selection Procedure ◽

Alcoholic Cirrhosis ◽

Information Criteria ◽

Carcinome Hépatocellulaire ◽

Forward Selection ◽

Prognostic Information ◽

Staging Systems

4589 Background: The aims of our study were to compare performances of 4 staging systems and to explore how to improve prognostic classification among French patients with HCC whose main aetiology is alcoholic cirrhosis. Methods: We have pooled 2 RCTs in palliative condition from Federation Francophone de Cancerologie Digestive (FFCD): - FFCD 9403 comparing tamoxifen vs symptomatic treatment and - FFCD 9402 comparing chemoembolization + tamoxifen vs tamoxifen alone. They had respectively included 416 and 122 patients. Performance of Okuda, Cancer of the Liver Italian Program (CLIP), Barcelona Clinic Liver Cancer group (BCLC) and GRoupe d’Etude et de Traitement du Carcinome Hépatocellulaire scores have been compared using: Akaike information criteria (AIC), discriminatory ability (Harrell’s c and the Royston’s D statistics), monocity of gradients and predictive accuracy (Schemper statistics Vs). To explore how to improve classifications univariate and multivariate Cox model were performed. Variables with univariate p< 0.10 have been retained for multivariate analyses. A forward selection procedure has then been implemented. Bootstraps validation was performed to test the robustness of our results. Analyses were done for each trial and for the pooled database with trial stratification. Results: Median OS was 5,3 months (IC 95%: [4,6; 6,2]), 402 patients had (75%) an alcoholic cirrhosis aetiology . As shown in Table 1 , CLIP staging had the best properties, followed by Okuda and BCLC. Performances of all staging systems were rather disappointing. WHO staging for CLIP or alphafetoprotein for BCLC allowed a significant improvement of prognostic information. Conclusions: Our results suggest that CLIP staging seems to be most adapted to french patients, it could be better by associating WHO PS. An external validation of our result will be performed on another trial in palliative condition. [Table: see text] No significant financial relationships to disclose.

Download Full-text

Factors and predictors of length of stay in offenders diagnosed with schizophrenia- A machine-learning-based approach

10.21203/rs.3.rs-15817/v2 ◽

2020 ◽

Author(s):

Johannes Kirchebner ◽

Moritz Günther ◽

Martina Sonnweber ◽

Alice King ◽

Steffen Lau

Keyword(s):

Machine Learning ◽

Length Of Stay ◽

Learning Algorithms ◽

Evaluation Criteria ◽

Selection Procedure ◽

Legal Aspects ◽

Influential Factors ◽

Machine Learning Algorithms ◽

Forward Selection ◽

Factors Affecting

Abstract Background: Prolonged forensic psychiatric hospitalizations have raised ethical, economic, and clinical concerns. Due to the confounded nature of factors affecting length of stay of psychiatric offender patients, prior research has called for the application of a new statistical methodology better accommodating this data structure. The present study attempts to investigate factors contributing to long-term hospitalization of schizophrenic offenders referred to a Swiss forensic institution, using machine learning algorithms that are better suited than conventional methods to detect nonlinear dependencies between variables. Methods: In this retrospective file and registry study, multidisciplinary notes of 143 schizophrenic offenders were reviewed using a structured protocol on patients’ characteristics, criminal and medical history and course of treatment. Via a forward selection procedure, the most influential factors for length of stay were preselected. Machine learning algorithms then identified the most efficient model for predicting length-of-stay. Results: Two factors have been identified as being particularly influential for a prolonged forensic hospital stay, both of which are related to aspects of the index offense, namely (attempted) homicide and the extent of the victim's injury. The results are discussed in light of previous research on this topic. Conclusions: In this study, length of stay was determined by legal considerations, but not by factors that can be influenced therapeutically. Results emphasize that forensic risk assessments should be based on different evaluation criteria and not merely on legal aspects.

Download Full-text

The predictors of information needs and different views of patients and relatives on disclosure and treatment participation: A multicentric survey study

Journal of Clinical Oncology ◽

10.1200/jco.2009.27.15_suppl.e20526 ◽

2009 ◽

Vol 27 (15_suppl) ◽

pp. e20526-e20526

Author(s):

M. Ozdogan ◽

H. Bozcuk ◽

O. Er ◽

H. Abali ◽

H. S. Coskun ◽

...

Keyword(s):

Cancer Patients ◽

Information Needs ◽

Multinomial Logistic Regression ◽

Selection Procedure ◽

Survey Study ◽

Forward Selection ◽

Patient Age ◽

Treatment Participation ◽

Patient Âgé ◽

Disclosure Of Diagnosis

e20526 Background: We wanted to explain if the views of patients and their relatives differed on disclosure and treatment participation or not, and if so, to evaluate the predictors of this difference. Methods: The survey aimed to unveil the attitude of patients and their relatives in 3 domains; disclosure of diagnosis, prognosis and participation to treatment. Multinomial logistic regression models with forward selection procedure were constructed for the multivariate analysis to explain the origins of discordance in these domains. Results: A total of 1,052 consecutive cases (526 patients and their 526 relatives) were interviewed. The relatives, when they were asked to imagine themselves with a new diagnosis of cancer, opted for disclosure of diagnosis, prognosis, and participation to treatment in 92.4%, 84.6%, and 86.3% of cases. When patients were asked for their information needs in disclosure of diagnosis, prognosis, and participation to treatment, 83.8%, 70.2%, and 70% wanted disclosure, whereas, their relatives wanted disclosure for their patients in these 3 domains in 32.9%, 40.5%, and 60.3% of cases, respectively. The multivariate predictors of discordance for disclosure of diagnosis between patients and relatives were patient age, social insurance, and oncology centers (P=0.003, 0.007, and <0.001, respectively). The associates of discordance for disclosure of prognosis were again oncology centers and relatives’ relationship with the patients (P<0.001 and 0.012). Likewise, the correlates of discordance for treatment participation were again oncology centers and patient age (P<0.001 and 0.016). Conclusions: Information needs of patients and their relatives, when they imagine themselves as cancer patients, are quite similar. However, relatives are not in favor of disclosure to their patients of diagnosis and prognosis, moreover, in spite of this, they want their patients to actively participate in treatment decisions. We believe this paradox represents a desire of the relatives to escape from responsibility and emotional burden of the care of the cancer patients. This paradox, in addition, is a threat to a healthy patient and physician communication. No significant financial relationships to disclose.

Download Full-text

Groupwise Non-Rigid Registration with Deep Learning: An Affordable Solution Applied to 2D Cardiac Cine MRI Reconstruction

Entropy ◽

10.3390/e22060687 ◽

2020 ◽

Vol 22 (6) ◽

pp. 687

Author(s):

Elena Martín-González ◽

Teresa Sevilla ◽

Ana Revilla-Orodea ◽

Pablo Casaseca-de-la-Higuera ◽

Carlos Alberola-López

Keyword(s):

Deep Learning ◽

Selection Procedure ◽

Structural Similarity ◽

The Other ◽

Cine Mri ◽

Forward Selection ◽

Cardiac Cine ◽

Template Image ◽

Mri Reconstruction ◽

Computationally Expensive

Groupwise image (GW) registration is customarily used for subsequent processing in medical imaging. However, it is computationally expensive due to repeated calculation of transformations and gradients. In this paper, we propose a deep learning (DL) architecture that achieves GW elastic registration of a 2D dynamic sequence on an affordable average GPU. Our solution, referred to as dGW, is a simplified version of the well-known U-net. In our GW solution, the image that the other images are registered to, referred to in the paper as template image, is iteratively obtained together with the registered images. Design and evaluation have been carried out using 2D cine cardiac MR slices from 2 databases respectively consisting of 89 and 41 subjects. The first database was used for training and validation with 66.6–33.3% split. The second one was used for validation (50%) and testing (50%). Additional network hyperparameters, which are—in essence—those that control the transformation smoothness degree, are obtained by means of a forward selection procedure. Our results show a 9-fold runtime reduction with respect to an optimization-based implementation; in addition, making use of the well-known structural similarity (SSIM) index we have obtained significative differences with dGW with respect to an alternative DL solution based on Voxelmorph.

Download Full-text

Assessment of genetic risk for myocardial infarction

Thrombosis and Haemostasis ◽

10.1160/th06-02-0117 ◽

2006 ◽

Vol 96 (08) ◽

pp. 220-227 ◽

Cited By ~ 29

Author(s):

Hitoshi Matsuo ◽

Tomonori Segawa ◽

Sachiro Watanabe ◽

Kimihiko Kato ◽

Takeshi Hibino ◽

...

Keyword(s):

Myocardial Infarction ◽

Genetic Risk ◽

Large Scale ◽

Selection Procedure ◽

Epidemiological Studies ◽

Multivariable Logistic Regression Analysis ◽

Forward Selection ◽

Suspension Array ◽

Reliable Assessment ◽

Oligonucleotide Ligation Assay

SummaryAlthough lifestyle and environmental factors influence the prevalence of myocardial infarction, genetic epidemiological studies have suggested that several genetic variants increase the risk for this condition. We have performeda large-scale association study to identify gene polymorphisms for reliable assessment of the genetic risk of myocardial infarction. The study population comprised 3,483 unrelated Japanese individuals (1,913 men; 1,570 women), including 1,192 subjects with myocardial infarction and 2,291 controls. The genotypes for 164 polymorphisms of 137 candidate genes were determined with an oligonucleotide ligation assay based on analysis of fluorescent microspheres with suspension array technology. Multivariable logistic regression analysis with adjustment for age, sex, body mass index, and the prevalence of smoking, hypertension, diabetes mellitus, and hypercholesterolemia revealed that the 677C→T (Ala222Val) polymorphism of MTHFR, the 1595C→G (Ser447Stop) polymorphism of LPL, and the –108/3G→4G polymorphism of IPF1 were significantly associated with the prevalence of myocardial infarction. A stepwise forward selection procedure demonstrated that IPF1, MTHFR, and LPL genotypes significantly affected the prevalence of myocardial infarction. Combined genotype analysis of these polymorphisms yieldeda maximum odds ratio of 2.54 for the combined genotype of TT for MTHFR,CC for LPL,and 3G3G for IPF1.The genotypes for MTHFR, LPL, and IPF1 may prove reliable for assessment of genetic risk for myocardial infarction. Determination of the combined genotype for these genes may contribute to primary, personalized prevention of this condition.

Download Full-text