A simple pooling method for variable selection in multiply imputed datasets outperformed complex methods.

BACKGROUND: For the development of prognostic models, after multiple imputation, variable selection is advised to be applied from the pooled model. The aim of this study is to evaluate by using a simulation study and practical data example the performance of four different pooling methods for variable selection in multiple imputed datasets. These methods are the D1, D2, D3 and recently extended Median-P-Rule (MPR) for categorical, dichotomous, and continuous variables in logistic regression models. METHODS: Four datasets (n = 200 and n = 500), with 9 variables and correlations of respectively 0.2 and 0.6 between these variables, were simulated. These datasets included 2 categorical and 2 continuous variables with 20% missing at random data. Multiple Imputation (m = 5) was applied, and the four methods were compared with selection from the full model (without missing data). The same analyzes were repeated in five multiply imputed real-world datasets (NHANES) (m = 5, p = 0.05, N = 250/300/400/500/1000). RESULTS: In the simulated datasets, the differences between the pooling methods were most evident in the smaller datasets. The MPR performed equal to all other pooling methods for the selection frequency, as well as for the P-values of the continuous and dichotomous variables, however the MPR performed consistently better for pooling and selecting categorical variables in multiply imputed datasets and also regarding the stability of the selected prognostic models. Analyzes in the NHANES-dataset showed that all methods mostly selected the same models. Compared to each other however, the D2-method seemed to be the least sensitive and the MPR the most sensitive, most simple, and easy method to apply. CONCLUSIONS: Considering that MPR is the most simple and easy pooling method to use for epidemiologists and applied researchers, we carefully recommend using the MPR-method to pool categorical variables with more than two levels after Multiple Imputation in combination with Backward Selection-procedures (BWS). Because MPR never performed worse than the other methods in continuous and dichotomous variables we also advice to use MPR in these types of variables.

Kinesiophobia is not required to predict chronic low back pain in workers: a decision curve analysis.

BACKGROUND: Currently used performance measures for discrimination were not informative to determine the clinical benefit of predictor variables. The purpose was to evaluate if a former relevant predictor, kinesiophobia, remained clinically relevant to predict chronic occupational low back pain (LBP) in the light of a novel discriminative performance measure, Decision Curve Analysis (DCA), using the Net Benefit (NB). METHODS: Prospective cohort data (n = 170) of two merged randomized trials with workers with LBP on sickleave, treated with Usual Care (UC) were used for the analyses. An existing prediction model for chronic LBP with the variables 'a clinically relevant change in pain intensity and disability status in the first 3 months', 'baseline measured pain intensity' and 'kinesiophobia' was compared with the same model without the variable 'kinesiophobia' using the NB and DCA. RESULTS: Both prediction models showed an equal performance according to the DCA and NB. Between 10 and 95% probability thresholds of chronic LBP risk, both models were of clinically benefit. There were virtually no differences between both models in the improved classification of true positive (TP) patients. CONCLUSIONS: This study showed that the variable kinesiophobia, which was originally included in a prediction model for chronic LBP, was not informative to predict chronic LBP by using DCA. DCA and NB have to be used more often to develop clinically beneficial prediction models in workers because they are more sensitive to evaluate the discriminate ability of prediction models.

CLINICAL PROGNOSTIC FACTORS FOR PATIENTS WITH ANTERIOR KNEE PAIN IN PHYSICAL THERAPY; A SYSTEMATIC REVIEW.

BACKGROUND: Although many authors have studied the prognostic factors that may contribute to anterior knee pain, synthesis of the existing evidence has not been performed. PURPOSE: The purpose of this systematic review is to summarize and examine existing prognostic models in patients with anterior knee pain that first present to physical therapists (primary care setting). DESIGN: Systematic review. METHOD: For this review Pubmed, Embase and Cinahl databases were searched and published papers that reported prognostic models for patients with anterior knee pain that first present to physical therapists (primary care setting) were selected. The authors extracted and summarized the univariate and multivariate predictors and evaluated which predictors consistently appeared to be relevant to pain, function, or recovery. RESULTS: Nine studies were included. The quality scores of these studies ranged from 9 to 17 positive items out of 21 items included in the assessment for quality. None of the prognostic models were validated internally or externally. Four studies were considered to be of sufficient quality. The authors of these four studies found 14 different predictors significantly related to pain intensity of which seven with limited evidence. Fifteen different predictors were found that were related to function of which seven with limited evidence. Furthermore, strong evidence was found that baseline pain intensity, pain coping and kinesiophobia are of no predictive value for pain, and activity related pain, pain coping and kinesiophobia are of no predictive value for function at follow up. CONCLUSIONS: Because of the low quality of a number of studies and the heterogeneity of the examined variables and outcome measures of most of the studies, only limited evidence for seven predictors related to pain and seven predictors related to function in patients with anterior knee pain in a primary care setting was found. LEVEL OF EVIDENCE: 1b.