Predicting septic shock outcomes in a database with missing data using fuzzy modeling: Influence of pre-processing techniques on real-world data-based classification

Abstract

Real-world databases often contain missing data and existing correction algorithms deliver varying performance. Also, most modeling techniques are not suitable to deal with them automatically. In this study we examine different approaches to predicting septic shock in the presence of missing data. Some preprocessing techniques for managing missing data include disregarding data, or replacing it with information that by design introduces bias. In this study, we show that predictive performance improves by employing a minimum pre-processing technique, the Zero-Order-Hold (ZOH) method, by applying a Fuzzy C-Means clustering technique based on the partial distance calculation strategy (FCM-PDS) and by computing the final classification regarding the samples from each patient. Performance improvements continue to occur where up to approximately 60% of the data is missing, though for higher percentage the classification performance still is statistically improved. We further validate this approach by making comparisons with previous studies.

Septic Shock Prediction for Patients with Missing Data

“…The model required at least 10 of the 12 variables to be present and could only correctly classify ∼70% of the test data with a sensitivity of 15.01%. Another study proposed the use of a modified Fuzzy C-Means algorithm with Partial Distance Strategy (FCM-PDS) that does not require any imputation of the missing values by means of product-space clustering [Pereira et al 2011]. The authors also suggested the combination of Zero-Order-Hold (ZOH), which holds the measurement value until a new observation is available, to deal with incomplete data and FCM-PDS to predict abdominal septic shock.…”

Linguistic summaries of categorical time series for septic shock patient data

“…This process allows all variable samples to be available at the same point in time as the template variable. The template variable chosen is the heart rate, since it is the most frequently measured variable (in average one sample every 60 minutes) and thus, the one introducing fewer artifacts in the data [24].…”

“…When it is possible to prove that a variable was not measured during a certain period of time because of an intentional reason (e.g. ventilator parameters when a patient is extubated), this missing segment is considered as non-recoverable [24]. In this work, these non-recoverable missing segments are deleted.…”

“…On the other hand, if the variable is supposed to exist, but for some unintentional reason (e.g. sensor malfunction) it is missing, this absent segment is considered recoverable and thus, proper imputation techniques can be applied [24]. Missing values are not a problem for the derivation of linguistic summaries or the calculation of quality measures.…”

A Review of Predictive Analytics Solutions for Sepsis Patients

“…Count of methods used among analyzed studies ¼ 2),18,33 to help train and validate their tools. The MIMIC-III data set contains 53,423 distinct hospital admissions for adult patients (aged 16 years or above) admitted to critical care units between 2001 and 2012, while the MEDAN data set contains data from 71 German ICUs from 1998 to 2002.…”