Search

2042 Background: Lung cancer patients commonly need unplanned visits to ED. Many of these visits could be potentially avoidable if it were possible to identify patients at risk when the previous scheduled visit takes place. At that moment, it would be possible to perform elective actions to manage patients at risk to consult the ED in the near future. Methods: Unplanned visits of patients in active cancer therapy (i.e. chemo or immunotherapy) are attended in our own ED facilities. Our Electronic Health Record (EHR) includes specific modules for first visit, scheduled visits and unplanned visits. Lung cancer patients with at least two visits were eligible. The event of interest was patient visit to ED within 21 or 28 days (d) from previous visit. Free text data collected in the three modules were obtained from EHR in order to generate a feature vector composed of the word frequencies for each visit. We evaluate five different machine learning algorithms to predict the event of interest. Area under the ROC curve (AUC), F1 (harmonic mean of precision and recall), True Positive Rate (TPR) and True Negative Rate (TNR) were assessed using 10-fold cross validation. Results: 2,682 lung cancer patients treated between March 2009 and October 2019 were included from which 819 patients were attended at ED. There were 2,237 first visits, 47,465 scheduled visits (per patient: range 1-174; median 12) and 2,125 unplanned visits (per patient: range 1-20; median 2). Mean age at diagnosis was 64 years. The majority of patients had late stage disease (34.24 % III, 51.56 % IV). The Adaptive Boosting Model yields the best results for both 21 d or 28 d prediction. Conclusions: Using unstructured data from real-world EHR enables the possibility to build an accurate predictive model of unplanned visit to an ED within the 21 or 28 following d after a scheduled visit. Such utility would be very useful in order to prevent ED visits related with cancer symptoms and to improve patients care. [Table: see text]

Background: Extracting relevant information from microarray data is a very complex task due to the characteristics of the data sets, as they comprise a large number of features while few samples are generally available. In this sense, feature selection is a very important aspect of the analysis helping in the tasks of identifying relevant genes and also for maximizing predictive information. Methods: Due to its simplicity and speed, Stepwise Forward Selection (SFS) is a widely used feature selection technique. In this work, we carry a comparative study of SFS and Genetic Algorithms (GA) as general frameworks for the analysis of microarray data with the aim of identifying group of genes with high predictive capability and biological relevance. Six standard and machine learning-based techniques (Linear Discriminant Analysis (LDA), Support Vector Machines (SVM), Naive Bayes (NB), C-MANTEC Constructive Neural Network, K-Nearest Neighbors (kNN) and Multilayer perceptron (MLP)) are used within both frameworks using six free-public datasets for the task of predicting cancer outcome. Results: Better cancer outcome prediction results were obtained using the GA framework noting that this approach, in comparison to the SFS one, leads to a larger selection set, uses a large number of comparison between genetic profiles and thus it is computationally more intensive. Also the GA framework permitted to obtain a set of genes that can be considered to be more biologically relevant. Regarding the different classifiers used standard feedforward neural networks (MLP), LDA and SVM lead to similar and best results, while C-MANTEC and k-NN followed closely but with a lower accuracy. Further, C-MANTEC, MLP and LDA permitted to obtain a more limited set of genes in comparison to SVM, NB and kNN, and in particular C-MANTEC resulted in the most robust classifier in terms of changes in the parameter settings. Conclusions: This study shows that if prediction accuracy is the objective, the GAbased approach lead to better results respect to the SFS approach, independently of the classifier used. Regarding classifiers, even if C-MANTEC did not achieve the best overall results, the performance was competitive with a very robust behaviour in terms of the parameters of the algorithm, and thus it can be considered as a candidate technique for future studies.

Purposes This paper presents the experience on the design and implementation of a user-centered Oncology Information System developed for the Medical Oncology Department at the “Hospital Universitario Virgen de la Victoria”, in Malaga, Spain. The project focused on the aspects considered in the literature as critical factors for a successful deployment and usage of a health information system. Methods System usability, adequate technology, integration of clinical routines, real-time statistical analysis of data, information confidentiality and standard protocol-based external interconnection were the key aspects considered. Results The developed system is based on a web application with a modular and layered architecture accounting for usability, ease of maintenance and further system development. Evaluation of system usability was carried at three and fifteen months after system deployment to analyze the advantages/disadvantages experienced by the end-users. Conclusions A thorough prior analysis of clinical activities and workflows, the use of the adequate technology, and the availability of data analysis tools will almost guarantee success in the deployment of an Oncology Information System.