Search

Objective: To compare how different imputation methods affect the estimates and performance of a prediction model for premature mortality. Study Design and Setting: Sex-specific Weibull accelerated failure time survival models were run on four separate datasets using complete case, mode, single and multiple imputation to impute missing values. Six performance measures were compared to access predictive accuracy (Nagelkerke R2, integrated brier score), discrimination (Harrell's c-index, discrimination slope) and calibration (calibration in the large, calibration slope). Results: The highest proportion of missingness for a single variable was 10.86% for the female model and 8.24% for the male model. Comparing the performance measures for complete case, mode, single and multiple imputation: the Nagelkerke R2 values for the female model was 0.1084, 0.1116, 0.1120 and 0.111–0.1120 with the male model exhibited similar variation of 0.1050, 0.1078, 0.1078 and 0.1078–0.1081. Harrell's c-index also demonstrated small variation with values of 0.8666, 0.8719, 0.8719 and 0.8711–0.8719 for the female model and 0.8549, 0.8548, 0.8550 and 0.8550–0.8553 for the male model. Conclusion: In the scenarios examined in this study, mode imputation performed well when using a population health survey compared to single and multiple imputation when predictive performance measures is the main model goal. To generate unbiased hazard ratios, multiple imputation methods were superior. This study shows the need to consider the best imputation approach for a predictive model development given the conditions of missing data and the goals of the analysis. [ABSTRACT FROM AUTHOR]

Objective: Healthcare is increasingly digitized, yet remote and automated machine learning (ML) triage prediction systems for virtual urgent care use remain limited. The Canadian Triage and Acuity Scale (CTAS) is the gold standard triage tool for in-person care in Canada. The current work describes the development of a ML-based acuity score modelled after the CTAS system. Methods: The ML-based acuity score model was developed using 2,460,109 de-identified patient-level encounter records from three large healthcare organizations (Ontario, Canada). Data included presenting complaint, clinical modifiers, age, sex, and self-reported pain. 2,041,987 records were high acuity (CTAS 1–3) and 416,870 records were low acuity (CTAS 4–5). Five models were trained: decision tree, k-nearest neighbors, random forest, gradient boosting regressor, and neural net. The outcome variable of interest was the acuity score predicted by the ML system compared to the CTAS score assigned by the triage nurse. Results: Gradient boosting regressor demonstrated the greatest prediction accuracy. This final model was tuned toward up triaging to minimize patient risk if adopted into the clinical context. The algorithm predicted the same score in 47.4% of cases, and the same or more acute score in 95.0% of cases. Conclusions: The ML algorithm shows reasonable predictive accuracy and high predictive safety and was developed using the largest dataset of its kind to date. Future work will involve conducting a pilot study to validate and prospectively assess reliability of the ML algorithm to assign acuity scores remotely. [ABSTRACT FROM AUTHOR]

Approximately 3.5% of deliveries in Canada result in potentially preventable neonatal readmission, often times due to preventable morbidities. With complexities in hospital discharge planning, health care providers may benefit in identifying infants at risk of readmission for additional monitoring. To develop and validate models for predicting 7-day neonatal readmission following vaginal or cesarean births. All liveborn term singleton infants without congenital anomalies in the province of Alberta who were not admitted to the NICU were identified using perinatal and hospitalization databases. A temporal split-sample was used for model development (2012–2014, vaginal n = 63,378; cesarean n = 21,225) and external validation (2014–2015, vaginal n = 21,583, cesarean n = 7,477). Multivariable logistic regression models using backward stepwise selection were used to identify predictors of 7-day readmission. We evaluated predictors of maternal age, Apgar score, length-of-stay, birthweight, gestational age, parity, residence, and sex. Hosmer-Lemeshow test and c-statistics were used to estimate calibration and discrimination. The rate of readmission was 3.3% (95% CI 3.1%, 3.4%) and 2.1% (95% CI 1.9%, 2.3%) following vaginal and cesarean births in the development dataset. Prediction model following vaginal birth, excluding predictors of length-of-stay and birthweight, had sub-optimal performance in development (c-statistics 0.69) and validation data (c-statistics 0.68). Prediction model following cesarean birth, excluding predictors of maternal age, birthweight, and residence, had sub-optimal performance in development (c-statistics 0.62) and validation data (c-statistics 0.64). Readmission was observed in 7.9% (95% CI 7.1%, 8.8%) and 4.9% (95% CI 3.9%, 6.1%) of infants of vaginal and cesarean births, respectively, in the top quintile for the risk of 7-day readmission. Using routinely collected administrative data, we developed and validated prediction models for neonatal readmission following vaginal and cesarean births. Presently the model is sub-optimal for use in risk assessment and planning at discharge, however, additional information may improve the predictive performance. [ABSTRACT FROM AUTHOR]

Type 1 diabetes mellitus (T1DM) patients are a significant threat to chronic kidney disease (CKD) development during their life. However, there is always a high chance of delay in CKD detection because CKD can be asymptomatic, and T1DM patients bypass traditional CKD tests during their routine checkups. This study aims to develop and validate a prediction model and nomogram of CKD in T1DM patients using readily available routine checkup data for early CKD detection. This research utilized 1375 T1DM patients' sixteen years of longitudinal data from multi-center Epidemiology of Diabetes Interventions and Complications (EDIC) clinical trials conducted at 28 sites in the USA and Canada and considered 17 routinely available features. Three feature ranking algorithms, extreme gradient boosting (XGB), random forest (RF), and extremely randomized trees classifier (ERT), were applied to create three feature ranking lists, and logistic regression analyses were performed to develop CKD prediction models using these ranked feature lists to identify the best performing top-ranked features combination. Finally, the most significant features were selected to develop a multivariate logistic regression-based CKD prediction model for T1DM patients. This model was evaluated using sensitivity, specificity, accuracy, precision, and F1 score on train and test data. A nomogram of the final model was further generated for easy application in clinical practices. Hypertension, duration of diabetes, drinking habit, triglycerides, ACE inhibitors, low-density lipoprotein (LDL) cholesterol, age, and smoking habit were the top-8 features ranked by the XGB model and identified as the most important features for predicting CKD in T1DM patients. These eight features were selected to develop the final prediction model using multivariate logistic regression, which showed 90.04% and 88.59% accuracy in internal and test data validation. The proposed model showed excellent performance and can be used for CKD identification in T1DM patients during routine checkups. [ABSTRACT FROM AUTHOR]

As the importance of eco-friendly transportation increases, providing an efficient approach for marine vessel operation is essential. Methods for status monitoring with consideration to the weather condition and forecasting with the use of in-service data from ships requires accurate and complete models for predicting the energy efficiency of a ship. The models need to effectively process all the operational data in real-time. This paper presents models that can predict fuel consumption using in-service data collected from a passenger ship. Statistical and domain-knowledge methods were used to select the proper input variables for the models. These methods prevent over-fitting, missing data, and multicollinearity while providing practical applicability. Prediction models that were investigated include multiple linear regression (MLR), decision tree approach (DT), an artificial neural network (ANN), and ensemble methods. The best predictive performance was from a model developed using the XGboost technique which is a boosting ensemble approach. Our code is available on GitHub at https://github.com/pagand/model%5foptimze%5fvessel/tree/OE for future research. • Comparison of fuel consumption models for west coast Canada vessel • Feature engineering utilized domain knowledge to identify key variables • Parametric and non-parametric models used for prediction model development • Evaluation included running time, mean error, and future step prediction error. [ABSTRACT FROM AUTHOR]