Manuel Filipe Santos

Rating organ failure via adverse events using data mining in the intensive care unit

OBJECTIVE The main intensive care unit (ICU) goal is to avoid or reverse the organ failure process by adopting a timely intervention. Within this context, early identification of organ impairment is a key issue. The sequential organ failure assessment (SOFA) is an expert-driven score that is widely used in European ICUs to quantify organ disorder. This work proposes a complementary data-driven approach based on adverse events, defined from commonly monitored biometrics. The aim is to study the impact of these events when predicting the risk of ICU organ failure. MATERIALS AND METHODS A large database was considered, with a total of 25,215 daily records taken from 4425 patients and 42 European ICUs. The input variables include the case mix (i.e. age, diagnosis, admission type and admission from) and adverse events defined from four bedside physiologic variables (i.e. systolic blood pressure, heart rate, pulse oximeter oxygen saturation and urine output). The output target is the organ status (i.e. normal, dysfunction or failure) of six organ systems (respiratory, coagulation, hepatic, cardiovascular, neurological and renal), as measured by the SOFA score. Two data mining (DM) methods were compared: multinomial logistic regression (MLR) and artificial neural networks (ANNs). These methods were tested in the R statistical environment, using 20 runs of a 5-fold cross-validation scheme. The area under the receiver operator characteristic (ROC) curve and Brier score were used as the discrimination and calibration measures. RESULTS The best performance was obtained by the ANNs, outperforming the MLR in both discrimination and calibration criteria. The ANNs obtained an average (over all organs) area under the ROC curve of 64, 69 and 74% and Brier scores of 0.18, 0.16 and 0.09 for the dysfunction, normal and failure organ conditions, respectively. In particular, very good results were achieved when predicting renal failure (ROC curve area of 76% and Brier score of 0.06). CONCLUSION Adverse events, taken from bedside monitored data, are important intermediate outcomes, contributing to a timely recognition of organ dysfunction and failure during ICU length of stay. The obtained results show that it is possible to use DM methods to get knowledge from easy obtainable data, thus making room for the development of intelligent clinical alarm monitoring.

Mortality assessment in intensive care units via adverse events using artificial neural networks

OBJECTIVE This work presents a novel approach for the prediction of mortality in intensive care units (ICUs) based on the use of adverse events, which are defined from four bedside alarms, and artificial neural networks (ANNs). This approach is compared with two logistic regression (LR) models: the prognostic model used in most of the European ICUs, based on the simplified acute physiology score (SAPS II), and a LR that uses the same input variables of the ANN model. MATERIALS AND METHODS A large dataset was considered, encompassing forty two ICUs of nine European countries. The recorded features of each patient include the final outcome, the case mix (e.g. age) and the intermediate outcomes, defined as the daily averages of the out of range values of four biometrics (e.g. heart rate). The SAPS II score requires 17 static variables (e.g. serum sodium), which are collected within the first day of the patients admission. A nonlinear least squares method was used to calibrate the LR models while the ANNs are made up of multilayer perceptrons trained by the RPROP algorithm. A total of 13,164 adult patients were randomly divided into training (66%) and test (33%) sets. The two methods were evaluated in terms of receiver operator characteristic (ROC) curves. RESULTS The event based models predicted the outcome more accurately than the currently used SAPS II model (P<0.05), with ROC areas within the ranges 83.9-87.1% (ANN) and 82.6-85.2% (LR) versus 80% (LR SAPS II). When using the same inputs, the ANNs outperform the LR (improvement of 1.3-2%). CONCLUSION Better prognostic models can be achieved by adopting low cost and real-time intermediate outcomes rather than static data.

Pervasive and Intelligent Decision Support in Intensive Medicine – The Complete Picture

In the Intensive Care Units (ICU) it is notorious the high number of data sources available. This situation brings more complexity to the way of how a professional makes a decision based on information provided by those data sources. Normally, the decisions are based on empirical knowledge and common sense. Often, they don’t make use of the information provided by the ICU data sources, due to the difficulty in understanding them. To overcome these constraints an integrated and pervasive system called INTCare has been deployed. This paper is focused in presenting the system architecture and the knowledge obtained by each one of the decision modules: Patient Vital Signs, Critical Events, ICU Medical Scores and Ensemble Data Mining. This system is able to make hourly predictions in terms of organ failure and outcome. High values of sensitivity where reached, e.g. 97.95% for the cardiovascular system, 99.77% for the outcome. In addition, the system is prepared for tracking patients’ critical events and for evaluating medical scores automatically and in real-time.

Intelligence in interoperability with AIDA

Healthcare systems have to be addressed in terms of a wide variety of heterogeneous, distributed and ubiquitous systems speaking different languages, integrating medical equipments and customized by different entities, which in turn were set by different people aiming at different goals. Demands of information within the healthcare sector range from clinically valuable patient-specific information to a variety of aggregation levels for follow-up and statistical and/or quantifiable reporting. The main goal is to gathering this information and present it in a readable way to physicians. In this work we show how to achieve interoperability in healthcare institutions using AIDA, an interoperability platform developed by researchers from the University of Minho and being used in some major Portuguese hospitals.

INTCare : a knowledge discovery based intelligent decision support system for intensive care medicine

This paper introduces the INTCare system, an intelligent information system based on a completely automated Knowledge Discovery process and on the Agents paradigm. The system was designed to work in Hospital Intensive Care Units, supporting the physicians’ decisions by means of prognostic Data Mining models. In particular, these techniques were used to predict organ failure and mortality assessment. The main intention is to change the current reactive behaviour to a pro-active one, enhancing the quality of service. Current applications and experimentations, the functional and structural aspects, and technological options are presented.

Enabling a Pervasive Approach for Intelligent Decision Support in Critical Health Care

The creation of a pervasive and intelligent environment makes possible the remote work with good results in a great range of applications. However, the critical health care is one of the most difficult areas to implement it. In particular Intensive Care Units represent a new challenge for this field, bringing new requirements and demanding for new features that should be satisfied if we want to succeed. This paper presents a framework to evaluate future developments in order to efficiently adapt an Intelligent Decision Support System to a pervasive approach in the area of critical health (INTCare research project).

New Models for Strength and Deformability Parameter Calculation in Rock Masses Using Data-Mining Techniques

Due to the inherent geological complexity and characterization difficulties in rock formations, the evaluation of geomechanical parameters is very complex, mostly in the initial project stages and in small-scale geotechnical works, where information is scarce for the definition of an accurate geotechnical model. However, in large geotechnical projects, a great amount of data are produced and used to establish near-homogeneous geotechnical zones. If properly analyzed, these data can provide valuable information that can be used in situations where knowledge of the rock mass is limited. Yet, this implies the organization of geotechnical data in formats for proper analysis using advanced tools which is not normally done. Data-mining techniques have been successfully used in many fields but scarcely in geotechnics. They seem to be adequate as an advanced technique for analyzing large and complex databases that can be built with geotechnical information within the framework of an overall process of knowledge d...

The next generation of interoperability agents in healthcare

Interoperability in health information systems is increasingly a requirement rather than an option. Standards and technologies, such as multi-agent systems, have proven to be powerful tools in interoperability issues. In the last few years, the authors have worked on developing the Agency for Integration, Diffusion and Archive of Medical Information (AIDA), which is an intelligent, agent-based platform to ensure interoperability in healthcare units. It is increasingly important to ensure the high availability and reliability of systems. The functions provided by the systems that treat interoperability cannot fail. This paper shows the importance of monitoring and controlling intelligent agents as a tool to anticipate problems in health information systems. The interaction between humans and agents through an interface that allows the user to create new agents easily and to monitor their activities in real time is also an important feature, as health systems evolve by adopting more features and solving new problems. A module was installed in Centro Hospitalar do Porto, increasing the functionality and the overall usability of AIDA.

A pervasive approach to a real-time intelligent decision support system in intensive medicine

The decision on the most appropriate procedure to provide to the patients the best healthcare possible is a critical and complex task in Intensive Care Units (ICU). Clinical Decision Support Systems (CDSS) should deal with huge amounts of data and online monitoring, analyzing numerous parameters and providing outputs in a short real-time. Although the advances attained in this area of knowledge new challenges should be taken into account in future CDSS developments, principally in ICUs environments. The next generation of CDSS will be pervasive and ubiquitous providing the doctors with the appropriate services and information in order to support decisions regardless the time or the local where they are. Consequently new requirements arise namely the privacy of data and the security in data access. This paper will present a pervasive perspective of the decision making process in the context of INTCare system, an intelligent decision support system for intensive medicine. Three scenarios are explored using data mining models continuously assessed and optimized. Some preliminary results are depicted and discussed.

Multiple Organ Failure Diagnosis Using Adverse Events and Neural Networks

In the past years, the Clinical Data Mining arena has suffered a remarkable development, where intelligent data analysis tools, such as Neural Networks, have been successfully applied in the design of medical systems. In this work, Neural Networks are applied to the prediction of organ dysfunction in Intensive Care Units. The novelty of this approach comes from the use of adverse events, which are triggered from four bedside alarms, being achieved an overall predictive accuracy of 70%.