Performance of logistic regression modeling: beyond the number of events per variable, the role of data structure

Courvoisier, Delphine S.; Combescure, Christophe; Agoritsas, Thomas; Gayet-Ageron, A; Perneger, Thomas

doi:10.1016/j.jclinepi.2011.06.013

Cited by 53 publications

(64 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…First, we selected the most strongly associated predictors from each subgroup of variables, using multivariable logistic regression with backward stepwise selection based on a likelihood-ratio test with a P value of 0.1. 17 Second, the final model was selected from this set of variables and validated with backward stepwise selection in multivariable logistic regression. One thousand bootstrap samples were drawn from the original sample, estimating the overfitting-corrected regression coefficients from the final model and the overfittingcorrected measures of the model performance.…”

Section: Discussionmentioning

confidence: 99%

In-Hospital Risk Prediction for Post-stroke Depression

Ginkel

Hafsteinsdøttir²,

Lindeman³

et al. 2013

Stroke

View full text Add to dashboard Cite

Background and Purpose— The timely detection of post-stroke depression is complicated by a decreasing length of hospital stay. Therefore, the Post-stroke Depression Prediction Scale was developed and validated. The Post-stroke Depression Prediction Scale is a clinical prediction model for the early identification of stroke patients at increased risk for post-stroke depression. Methods— The study included 410 consecutive stroke patients who were able to communicate adequately. Predictors were collected within the first week after stroke. Between 6 to 8 weeks after stroke, major depressive disorder was diagnosed using the Composite International Diagnostic Interview. Multivariable logistic regression models were fitted. A bootstrap-backward selection process resulted in a reduced model. Performance of the model was expressed by discrimination, calibration, and accuracy. Results— The model included a medical history of depression or other psychiatric disorders, hypertension, angina pectoris, and the Barthel Index item dressing. The model had acceptable discrimination, based on an area under the receiver operating characteristic curve of 0.78 (0.72–0.85), and calibration ( P value of the U- statistic, 0.96). Transforming the model to an easy-to-use risk-assessment table, the lowest risk category (sum score, <−10) showed a 2% risk of depression, which increased to 82% in the highest category (sum score, >21). Conclusions— The clinical prediction model enables clinicians to estimate the degree of the depression risk for an individual patient within the first week after stroke.

show abstract

Section: Discussionmentioning

confidence: 99%

In-Hospital Risk Prediction for Post-stroke Depression

Ginkel

Hafsteinsdøttir²,

Lindeman³

et al. 2013

Stroke

View full text Add to dashboard Cite

show abstract

“…In this study, even in the multivariable model, the EPV was !10, indicating acceptable accuracy and precision of regression coefficients (Peduzzi et al, 1996). However, it should be mentioned that there is no confirmed consensus on number of EPVs and some authors have criticized the issue (Courvoisier et al, 2011).…”

Section: Discussionmentioning

confidence: 99%

Gender differences in the relationship between serum zinc concentration and metabolic syndrome

Ghasemi¹,

Zahediasl²,

Hosseini-Esfahani³

et al. 2014

Annals of Human Biology

View full text Add to dashboard Cite

show abstract

“…In the latter case, controlling too many variables by conventional means can lead to or aggravate two closely related problems: (a) data sparsity, in which full control results in too few subjects at crucial combinations of the variables, with consequent inflation of estimates (59,76,116,134), and (b) multicollinearity, by which we mean high multiple correlation (or more generally, high association) of the controlled variables with study exposures (116). In particular, if we include covariates that together are highly predictive of an exposure but are not all necessary to control confounding, the resulting effect estimate may be inflated or have unnecessarily wide confidence intervals (15,26,116). These problems increase as the ratio of number of covariates to sample size increases, motivating strategies to reduce the number of modeled covariates (116; S. Greenland & N. Pearce, unpublished manuscript, "Modeling Strategies for Observational Epidemiology").…”

Section: Why Not Adjust For Every Available Covariate?mentioning

confidence: 99%

Statistical Foundations for Model-Based Adjustments

Greenland

Pearce²

2015

Annu. Rev. Public Health

216

198

View full text Add to dashboard Cite

Most epidemiology textbooks that discuss models are vague on details of model selection. This lack of detail may be understandable since selection should be strongly influenced by features of the particular study, including contextual (prior) information about covariates that may confound, modify, or mediate the effect under study. It is thus important that authors document their modeling goals and strategies and understand the contextual interpretation of model parameters and model selection criteria. To illustrate this point, we review several established strategies for selecting model covariates, describe their shortcomings, and point to refinements, assuming that the main goal is to derive the most accurate effect estimates obtainable from the data and available resources. This goal shifts the focus to prediction of exposure or potential outcomes (or both) to adjust for confounding; it thus differs from the goal of ordinary statistical modeling, which is to passively predict outcomes. Nonetheless, methods and software for passive prediction can be used for causal inference as well, provided that the target parameters are shifted appropriately.

show abstract

Performance of logistic regression modeling: beyond the number of events per variable, the role of data structure

Cited by 53 publications

References 5 publications

In-Hospital Risk Prediction for Post-stroke Depression

In-Hospital Risk Prediction for Post-stroke Depression

Gender differences in the relationship between serum zinc concentration and metabolic syndrome

Statistical Foundations for Model-Based Adjustments

Contact Info

Product

Resources

About