Gunther Schadow

The Regenstrief Medical Record System: a quarter century experience

Entrusted with the records for more than 1.5 million patients, the Regenstrief Medical Record System (RMRS) has evolved into a fast and comprehensive data repository used extensively at three hospitals on the Indiana University Medical Center campus and more than 30 Indianapolis clinics. The RMRS routinely captures laboratory results, narrative reports, orders, medications, radiology reports, registration information, nursing assessments, vital signs, EKGs and other clinical data. In this paper, we describe the RMRS data model, file structures and architecture, as well as recent necessary changes to these as we coordinate a collaborative effort among all major Indianapolis hospital systems, improving patient care by capturing city-wide laboratory and encounter data. We believe that our success represents persistent efforts to build interfaces directly to multiple independent instruments and other data collection systems, using medical standards such as HL7, LOINC, and DICOM. Inpatient and outpatient order entry systems, instruments for visit notes and on-line questionnaires that replace hardcopy forms, and intelligent use of coded data entry supplement the RMRS. Physicians happily enter orders, problems, allergies, visit notes, and discharge summaries into our locally developed Gopher order entry system, as we provide them with convenient output forms, choice lists, defaults, templates, reminders, drug interaction information, charge information, and on-line articles and textbooks. To prepare for the future, we have begun wrapping our system in Web browser technology, testing voice dictation and understanding, and employing wireless technology.

Privacy-preserving data integration and sharing

Integrating data from multiple sources has been a longstanding challenge in the database community. Techniques such as privacy-preserving data mining promises privacy, but assume data has integration has been accomplished. Data integration methods are seriously hampered by inability to share the data to be integrated. This paper lays out a privacy framework for data integration. Challenges for data integration in the context of this framework are discussed, in the context of existing accomplishments in data integration. Many of these challenges are opportunities for the data mining community.

Protein quantification in label-free LC-MS experiments.

The goal of many LC-MS proteomic investigations is to quantify and compare the abundance of proteins in complex biological mixtures. However, the output of an LC-MS experiment is not a list of proteins, but a list of quantified spectral features. To make protein-level conclusions, researchers typically apply ad hoc rules, or take an average of feature abundance to obtain a single protein-level quantity for each sample. We argue that these two approaches are inadequate. We discuss two statistical models, namely, fixed and mixed effects Analysis of Variance (ANOVA), which views individual features as replicate measurements of a proteins abundance, and explicitly account for this redundancy. We demonstrate, using a spike-in and a clinical data set, that the proposed models improve the sensitivity and specificity of testing, improve the accuracy of patient-specific protein quantifications, and are more robust in the presence of missing data.

Conceptual alignment of electronic health record data with guideline and workflow knowledge

Even though computerized practice guidelines and workflow management (WfM) are proven effective techniques to improving quality of care and reducing costs, they are not widely deployed today. One reason for this is the impedance mismatch between guideline systems and electronic health record (EHR) systems. This paper presents the Unified Service Action Model (USAM) that has been developed for the HL7 Reference Information Model (RIM) and that conceptually integrates guidelines and WfM in the EHR. We argue that the information items recorded in the EHR are logically similar to elements of guideline and WfM definitions. Therefore, the USAM suggests that guidelines and EHR reuse the same information structures. This reuse is possible through a technique borrowed from natural language grammar and modal logic. The conceptual alignment of guidelines, WfM and the EHR could facilitate the sharing and deployment of guidelines in routine health care.

An outline for a vocabulary of nominal properties and examinations - basic and general concepts and associated terms

Abstract Scientists of disciplines in clinical laboratory sciences have long recognized the need for a common language for efficient and safe request of investigations, reporting of results, and communication of experience and scientific achievements. Widening the scope, most scientific disciplines, not only clinical laboratory sciences, rely to some extent on various nominal examinations, in addition to measurements. The ‘International vocabulary of metrology – Basic and general concepts and associated terms’ (VIM) is designed for metrology, science of measurement. The aim of the proposed vocabulary is to suggest definitions and explanations of concepts and terms related to nominal properties, i.e., properties that can be compared for identity with other properties of the same kind-of-property, but that have no magnitude. Clin Chem Lab Med 2010;48:1553–66.

Decision support from local data

OBJECTIVE Reducing care variability through guidelines has significantly benefited patients. Nonetheless, guideline-based Clinical Decision Support (CDS) systems are not widely implemented or used, are frequently out-of-date, and cannot address complex care for which guidelines do not exist. Here, we develop and evaluate a complementary approach - using Bayesian Network (BN) learning to generate adaptive, context-specific treatment menus based on local order-entry data. These menus can be used as a draft for expert review, in order to minimize development time for local decision support content. This is in keeping with the vision outlined in the US Health Information Technology Strategic Plan, which describes a healthcare system that learns from itself. MATERIALS AND METHODS We used the Greedy Equivalence Search algorithm to learn four 50-node domain-specific BNs from 11,344 encounters: abdominal pain in the emergency department, inpatient pregnancy, hypertension in the Urgent Visit Clinic, and altered mental state in the intensive care unit. We developed a system to produce situation-specific, rank-ordered treatment menus from these networks. We evaluated this system with a hospital-simulation methodology and computed Area Under the Receiver-Operator Curve (AUC) and average menu position at time of selection. We also compared this system with a similar association-rule-mining approach. RESULTS A short order menu on average contained the next order (weighted average length 3.91-5.83 items). Overall predictive ability was good: average AUC above 0.9 for 25% of order types and overall average AUC .714-.844 (depending on domain). However, AUC had high variance (.50-.99). Higher AUC correlated with tighter clusters and more connections in the graphs, indicating importance of appropriate contextual data. Comparison with an Association Rule Mining approach showed similar performance for only the most common orders with dramatic divergence as orders are less frequent. DISCUSSION AND CONCLUSION This study demonstrates that local clinical knowledge can be extracted from treatment data for decision support. This approach is appealing because: it reflects local standards; it uses data already being captured; and it produces human-readable treatment-diagnosis networks that could be curated by a human expert to reduce workload in developing localized CDS content. The BN methodology captured transitive associations and co-varying relationships, which existing approaches do not. It also performs better as orders become less frequent and require more context. This system is a step forward in harnessing local, empirical data to enhance decision support.

Clinical laboratory sciences data transmission : the NPU coding system

In health care services, technology requires that correct information be duly available to professionals, citizens and authorities, worldwide. Thus, clinical laboratory sciences require standardized electronic exchanges for results of laboratory examinations. The NPU (Nomenclature, Properties and Units) coding system provides a terminology for identification of result values (property values). It is structured according to BIPM, ISO, IUPAC and IFCC recommendations. It uses standard terms for established concepts and structured definitions describing: which part of the universe is examined, which component of relevance in that part, which kind-of-property is relevant. Unit and specifications can be added where relevant [System(spec)-Component(spec); kind-of-property(spec) = ? unit]. The English version of this terminology is freely accessible at http://dior.imt.liu.se/cnpu/ and http://www.labterm.dk, directly or through the IFCC and IUPAC websites. It has been nationally used for more than 10 years in Denmark and Sweden and has been translated into 6 other languages. The NPU coding system provides a terminology for dedicated kinds-of-property following the international recommendations. It fits well in the health network and is freely accessible. Clinical laboratory professionals worldwide will find many advantages in using the NPU coding system, notably with regards to an accreditation process.

Statistical Design and Analysis of Label-free LC-MS Proteomic Experiments: A Case Study of Coronary Artery Disease

This chapter presents a case study, which applies statistical design and analysis to an LC-MS-based -investigation of subjects with coronary artery disease. First, we discuss the principles of statistical -experimental design, and the specification of an Analysis of Variance (ANOVA) model that describes the major sources of variation in the data. Second, we discuss procedures for detecting differentially abundant proteins, estimating protein abundance in individual samples, testing predefined groups of proteins for enrichment in differential abundance, and calculating sample size for a future experiment. The discussion is accompanied by examples of computer code implemented in the open-source statistical software R, which can be followed for an independent implementation of a similar investigation.

Querying the National Drug File Reference Terminology (NDFRT) to assign drugs to decision support categories.

INTRODUCTION The accurate categorization of drugs is a prerequisite for decision support rules. The manual process of creating drug classes can be laborious and error-prone. METHODS All 142 drug classes currently used at Regenstrief Institute for drug interaction alerts were extracted. These drug classes were replicated as fully-defined concepts in our local instance of the NDFRT knowledge base. The performance of these two strategies (manual classification vs. NDFRT-based queries) was compared, and the sensitivity and specificity of each was calculated. RESULTS Compared to existing manual classifications, NDFRT-based queries made a greater number of correct class-drug assignments: 1528 vs. 1266. NDFRT queries have greater sensitivity (74.9% vs. 62.1%) to classify drugs. However, they have less specificity (85.6% vs. 99.8%). CONCLUSION The NDFRT knowledge base shows promise for use in an automated strategy to improve the creation and update of drug classes. The chief disadvantage of our NDFRT-based approach was a greater number of false positive assignments due to the inclusion of non-systemic doseforms.