Prediction Model Performance With Different Imputation Strategies: A Simulation Study Using a North American ICU Registry

Steif, Jonathan; Brant, Rollin; Sreepada, Rama Syamala; West, Nicholas; Murthy, Srinivas; Görges, Matthias

doi:10.1097/pcc.0000000000002835

Cited by 8 publications

(4 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also performed and reported the performance measure from a complete-case analysis. Even though the results of the complete-case analysis were different from those of the MICE approaches, we believe this finding had no strong clinical implication, as a growing body of evidence has found that complete-case analysis often leads to errors, result misinterpretations, and impaired generalizability [ 28 ]. Second, a large proportion of eligible patients were excluded due to incomplete or unverifiable data on relevant dates.…”

Section: Discussionmentioning

confidence: 94%

External Validation of the Individualized Prediction of Breast Cancer Survival (IPBS) Model for Estimating Survival after Surgery for Patients with Breast Cancer in Northern Thailand

Charumporn

Jarupanich

Rinthapon

et al. 2022

Cancers

View full text Add to dashboard Cite

The individualized prediction of breast cancer survival (IPBS) model was recently developed. Although the model showed acceptable performance during derivation, its external performance remained unknown. This study aimed to validate the IPBS model using the data of breast cancer patients in Northern Thailand. An external validation study was conducted based on female patients with breast cancer who underwent surgery at Maharaj Nakorn Chiang Mai hospital from 2005 to 2015. Data on IPBS predictors were collected. The endpoints were 5-year overall survival (OS) and disease-free survival (DFS). The model performance was evaluated in terms of discrimination and calibration. Missing data were handled with multiple imputation. Of all 3581 eligible patients, 1868 were included. The 5-year OS and DFS were 85.2% and 81.9%. The IPBS model showed acceptable discrimination: C-statistics 0.706 to 0.728 for OS and 0.675 to 0.689 for DFS at 5 years. However, the IPBS model minimally overestimated both OS and DFS predictions. These overestimations were corrected after model recalibration. In this external validation study, the IPBS model exhibited good discriminative ability. Although it may provide minimal overestimation, recalibrating the model to the local context is a practical solution to improve the model calibration.

show abstract

Section: Discussionmentioning

confidence: 94%

External Validation of the Individualized Prediction of Breast Cancer Survival (IPBS) Model for Estimating Survival after Surgery for Patients with Breast Cancer in Northern Thailand

Charumporn

Jarupanich

Rinthapon

et al. 2022

Cancers

View full text Add to dashboard Cite

show abstract

“…Dealing with missing data Multiple imputation methods were undertaken to estimate missing data over 50 iterations to generate 20 complete datasets for model development [33]. To determine whether imputation led to radically different results, a sensitivity analysis was performed by comparing the outcomes of the imputed data set with complete cases only.…”

Section: Discussionmentioning

confidence: 99%

Development and internal validation of two prediction models of walking dependence and restricted walking speed in individuals with subacute stroke

Fukumoto,

Oyake,

Kurobe

et al. 2023

Preprint

View full text Add to dashboard Cite

Background The prognosis of the walking ability of individuals with stroke affects the choice of rehabilitation program. Identifying patients who will need assistance with ambulation at discharge allows clinicians to deliver rehabilitation programs focused on educating caregivers and adjusting the patient’s environment to allow safe transfer within the home. The primary objective of this study was to develop and internally validate a prediction model of walking dependence for patients with stroke admitted to a rehabilitation facility. The secondary objective was to establish a prediction model of restricted walking speed. Methods This retrospective cohort study included 476 individuals with subacute stroke who were admitted to a rehabilitation facility. Model 1 was developed to predict the probability of walking dependence. Model 2 was developed to predict restricted walking speed. Walking dependence was defined as a functional independence measure walk score of 5 or less. Restricted walking speed was defined as the ability to walk at 0.93 m/s or less. Potential predictors, including age, leg strength of the affected side, sitting balance, cognitive function, and urinary function, were selected based on the literature and analyzed using logistic regression analysis. Models were internally validated using the bootstrap method. Model performance was assessed by discrimination (area under the receiver operating characteristic curve (AUC)) and calibration (Hosmer–Lemeshow (H–L) goodness-of-fit test and calibration plots). Results A total of 236 patients (49.6%) walked dependently at discharge. Of the 240 individuals who achieved walking independence, 121 (50.4%) had restricted walking speed. In model 1, older age, poor leg strength, sitting balance, cognitive function, and urinary incontinence were significantly associated with walking dependence at discharge. The AUCs of models 1 and 2 were 0.93 (95% confidence interval (CI) = .90–.95) and 0.69 (95%CI = .62–.76), respectively. Both models had good calibration confirmed by the H-L test. Conclusions The internally validated prediction model of walking dependence had good discrimination and calibration, while the prediction model of restricted walking speed had poor discrimination. The prediction model for walking dependence developed in this study may be useful for planning rehabilitation strategies and setting realistic goals for patients.

show abstract

“…32 Data were assumed to be missing at random and potential predictors of missing variables were included for imputation. 33,34 Convergence and observed versus imputed value plots were visually checked for quality control. 32 A penalized linear regression method, least absolute shrinkage and selection operator (LASSO), was used for variable selection.…”

Section: Discussionmentioning

confidence: 99%

“…32 Data were assumed to be missing at random and potential predictors of missing variables were included for imputation. 33 34 Convergence and observed versus imputed value plots were visually checked for quality control. 32…”

Section: Methodsmentioning

confidence: 99%

Patient-Specific Factors Associated with Dexmedetomidine Dose Requirements in Critically Ill Children

Crisamore

Empey

Pelletier

et al. 2022

J Pediatr Intensive Care

View full text Add to dashboard Cite

The objective of this study was to evaluate patient-specific factors associated with dexmedetomidine dose requirements during continuous infusion. A retrospective cross-sectional analysis of electronic health record-derived data spanning 10 years for patients admitted with a primary respiratory diagnosis at a quaternary children's hospital and who received a dexmedetomidine continuous infusion (n = 346 patients) was conducted. Penalized regression was used to select demographic, clinical, and medication characteristics associated with a median daily dexmedetomidine dose. Identified characteristics were included in multivariable linear regression models and sensitivity analyses. Critically ill children had a median hourly dexmedetomidine dose of 0.5 mcg/kg/h (range: 0.1–1.8), median daily dose of 6.7 mcg/kg/d (range: 0.9–38.4), and median infusion duration of 1.6 days (range: 0.25–5.0). Of 26 variables tested, 15 were selected in the final model with days of dexmedetomidine infusion (β: 1.9; 95% confidence interval [CI]: 1.6, 2.3), median daily morphine milligram equivalents dosing (mg/kg/d) (β: 0.3; 95% CI: 0.1, 0.5), median daily ketamine dosing (mg/kg/d) (β: 0.2; 95% CI: 0.1, 0.3), male sex (β: −1.1; 95% CI: −2.0, −0.2), and non-Black reported race (β: −1.2; 95% CI: −2.3, −0.08) significantly associated with median daily dexmedetomidine dose. Approximately 56% of dose variability was explained by the model. Readily obtainable information such as demographics, concomitant medications, and duration of infusion accounts for over half the variability in dexmedetomidine dosing. Identified factors, as well as additional environmental and genetic factors, warrant investigation in future studies to inform precision dosing strategies.

show abstract

Prediction Model Performance With Different Imputation Strategies: A Simulation Study Using a North American ICU Registry

Cited by 8 publications

References 46 publications

External Validation of the Individualized Prediction of Breast Cancer Survival (IPBS) Model for Estimating Survival after Surgery for Patients with Breast Cancer in Northern Thailand

External Validation of the Individualized Prediction of Breast Cancer Survival (IPBS) Model for Estimating Survival after Surgery for Patients with Breast Cancer in Northern Thailand

Development and internal validation of two prediction models of walking dependence and restricted walking speed in individuals with subacute stroke

Patient-Specific Factors Associated with Dexmedetomidine Dose Requirements in Critically Ill Children

Contact Info

Product

Resources

About