Estimation of a Predictor’s Importance by Random Forests When There Is Missing Data: RISK Prediction in Liver Surgery using Laboratory Data

Abstract

In the last few decades, new developments in liver surgery have led to an expanded applicability and an improved safety. However, liver surgery is still associated with postoperative morbidity and mortality, especially in extended resections. We analyzed a large liver surgery database to investigate whether laboratory parameters like haemoglobin, leucocytes, bilirubin, haematocrit and lactate might be relevant preoperative predictors. It is not uncommon to observe missing values in such data. This also holds for many other data sources and research fields. For analysis, one can make use of imputation methods or approaches that are able to deal with missing values in the predictor variables. A representative of the latter are Random Forests which also provide variable importance measures to assess a variable's relevance for prediction. Applied to the liver surgery data, we observed divergent results for the laboratory parameters, depending on the method used to cope with missing values. We therefore performed an extensive simulation study to investigate the properties of each approach. Findings and recommendations: Complete case analysis should not be used as it distorts the relevance of completely observed variables in an undesirable way. The estimation of a variable's importance by a self-contained measure that can deal with missing values appropriately reflects the decreased relevance of variables with missing values. It can therefore be used to obtain insight into Random Forests which are commonly fit without preprocessing of missing values in the data. By contrast, multiple imputation allows for the assessment of a variable's relevance one would potentially observe in complete-data situations, if imputation performs well. For the laboratory data, lactate and bilirubin seem to be associated with the risk of liver failure and postoperative complications. These relations should be investigated by future studies in more detail. However, it is important to carefully consider the method used for analysis when there are missing values in the predictor variables.

Optical Cloud Pixel Recovery via Machine Learning

“…For example, RF has been used to improve the prediction of missing values using laboratory generated medical data [16] and also to classify salt marsh vegetation [34]. In general, RF is a decision tree based method for classification and regression.…”

“…The importance of data recovery in numerous fields such as medical [15,16], neuro-computation [17], and climate science [18,19] has long been realized. Several approaches have been applied to recover the value of missing data.…”

Accessing habitat suitability and connectivity for the westernmost population of Asian black bear (Ursus thibetanus gedrosianus, Blanford, 1877) based on climate changes scenarios in Iran

“…In this model, the importance of the variables entered into the model is measured based on two criteria, mean decrease in accuracy and mean decrease in Gini [114]. This feature of the model has led to its use in many studies to measure the importance of variables [115][116][117][118][119]. There are other advantages that have led to the increase in the usefulness of this model.…”

Random forest robustness, variable importance, and tree aggregation

“…Measuring variable importance becomes increasingly complicated when data are missing. Hapfelmeier et al (2012Hapfelmeier et al ( , 2014a introduce an approach for estimating variable importance in such settings. This approach is intended to capture the value of the information provided by the data present, and variables with large amounts of missing data are often ranked as less important than they would be if complete data were available.…”

“…Manual-setting up, and understanding random forests V4.0. Hapfelmeier, A., Hothorn, T., Riediger, C., and Ulm, K. (2014a). Estimation of a predictor's importance by random forests when there is missing data: risk prediction in liver surgery using laboratory data.…”