The objective of this paper is to understand what characteristics and features of clinical data influence physician’s decision about ordering laboratory tests or prescribing medications the most. We conduct our analysis on data and decisions extracted from electronic health records of 4486 post-surgical cardiac patients. The summary statistics for 335 different lab order decisions and 407 medication decisions are reported. We show that in many cases, physician’s lab-order and medication decisions are predicted well by simple patterns such as last value of a single test result, time since a certain lab test was ordered or time since certain procedure was executed. Keywords:

Most existing algorithms for clinical risk stratification rely on labeled training data. Collecting this data is challenging for clinical conditions where only a small percentage of patients experience adverse outcomes. We propose an unsupervised anomaly detection approach to risk stratify patients without the need of positively and negatively labeled training examples. High-risk patients are identified without any expert knowledge using a minimum enclosing ball to find cases that lie in sparse regions of the feature space. When evaluated on data from patients admitted with acute coronary syndrome and on patients undergoing inpatient surgical procedures, our approach successfully identified individuals at increased risk of adverse endpoints in both populations. In some cases, unsupervised anomaly detection outperformed other machine learning methods that used additional knowledge in the form of labeled examples. 1.

Abstract. In this work we develop and test a novel hierarchical frame-work for modeling and learning multivariate clinical time series data. Our framework combines two modeling approaches: Linear Dynamical Systems (LDS) and Gaussian Processes (GP), and is capable to model and work with time series of varied length and with irregularly sampled observations. We test our framework on the problem of learning clinical time series data from the complete blood count panel, and show that our framework outperforms alternative time series models in terms of its predictive accuracy.

Developing methods to identify unusual clinical actions may be useful in the development of automated clinical alerting systems. We developed and evaluated a data-driven approach for identifying clinical actions such as omissions of medication orders or laboratory orders in the intensive care unit (ICU) that are unusual with respect to past patient care. We generated 250 medication-omission alerts and 150 laboratory-omission alerts using a database of 24,658 ICU patient admissions. These alerts were evaluated by a group of intensive care physicians. Overall, the true positive alert rate was 0.52, which we view as quite promising.

The objective of this paper is to understand what characteristics and features of clinical data influence physician’s decision about ordering laboratory tests or prescribing medications the most. We conduct our analysis on data and decisions extracted from electronic health records of 4486 post-surgical cardiac patients. The summary statistics for 335 different lab order decisions and 407 medication decisions are reported. We show that in many cases, physician’s lab-order and medication decisions are predicted well by simple patterns such as last value of a single test result, time since a certain lab test was ordered or time since certain procedure was executed. Keywords:

In this work, we describe a comprehensive framework for knowledge discovery from medical records called SDM-Miner. The records are created before, during and after pancreatic islet cell transplantation1 on a group of diabetic patients. The knowledge discovery focuses on selecting the most relevant variables for predicting the outcome of islet cell transplants temporally, and supporting the medical understanding of the variable relationships that would lead to insulin-free outcome of a transplant with machine learning models. The challenges of knowledge discovery lie in the temporally sparse nature of medical records and the large number of variables which make the traditional statistical analyses ineffective. Our approach to overcome the challenges is to combine data-driven computationally intensive modeling with statistical modeling. The framework incorporates this approach during three phases of knowledge discovery including (1) statistical data-preprocessing, (2) pattern search based dimensionality reduction, and (3) association rule based and conditional probability based data-driven modeling. We evaluate the framework by cross validating the models (of machine learning) using prediction errors and uncertainty of rule discovery. In order to demonstrate the novelty of the framework and the improved performance in knowledge discovery, we report results using real and synthetic datasets. Experimental results on synthetic data act as a sanity check in order to verify the effectiveness of our models in the absence of standard test results. The evaluation results show that our framework led to smaller mean error with the decreasing number of variable samples, higher robustness to Gaussian noise, and higher confidence and support of association rules than the previous methods. Furthermore, we evaluate our proposed technique using existing machine learning algorithms such using the Weka toolkit and show the improved performance of our work as compared to previous approaches.

Feature importance analysis for patient management decisions

The emergence of large-scale datasets in health care that record large amounts of in-formation about the patients, their diseases and treatments and provide us with an op-portunity to understand better the dynamics of the disease, efficacy of treatments, and various influences affecting the well-being of a patient. The development of computer