Health Services Research & Development

Veterans Crisis Line Badge
Go to the ORD website
Go to the QUERI website

2007 HSR&D National Meeting Abstract

Printable View

National Meeting 2007

3010 — Temporal Knowledge Discovery and Data Visualization for Clinical Process Improvement

Calvitti A (School of Medicine, University of California San Diego)

Objectives:
To develop Knowledge Discovery in Databases (KDD) and Data Visualization (Viz) methods to measure time-critical clinical processes, via observational studies of secondary data from CPRS and other databases. Then to identify, in conjunction with process modeling of selected use-cases, material and information bottlenecks and other inefficiencies that impede delivery of high quality care.

Methods:
KDD+Viz is a computational methodology that aims to extract informative patterns from high-dimensional, heterogeneous datasets for exploratory analysis. It is complementary to, but distinct from, classical model-driven statistical analysis. KDD+Viz reveals, via timeline or other statistical graphics, detailed relationships latent in data, taking advantage of our visual system’s capacity to parse complex patterns. However, KDD+Viz is not an “out of the box” solution. It requires iterated algorithmic development (querying, cleaning, structuring, aggregation, and graphing). Further, ethnographic studies of clinical processes are needed to relate normative schemas to ground truth, e.g. consistency of logged time-stamps versus actual transaction times.

Results:
We have developed KDD+Viz code libraries in Mathematica (Wolfram Research) for visualization of interruptions in system processes for an experimental wireless client/server system (http://wiisard.org) used for management of mass casualty events. By linking measures of clinical performance and information system performance in a single timeline, we were able to identify critical process faults in this complex and dynamic health care delivery system. We are currently working to generalize the code libraries from this project to parse and display more complex clinical datasets from CPRS/VISTA.

Implications:
From a systems perspective, clinical processes are dynamic and nonlinear. KDD+Viz may be better suited than conventional statistical models to exploratory analysis of temporal data generated by such processes. It is feasible to use the KDD+Viz approach to provide an integrated view into healthcare systems that combine clinical and information technology processes. These views can reveal critical process inefficiencies and bottlenecks.

Impacts:
KDD+Viz approach may help the VA better leverage its investments in information technology. Traditional analysis tools face limitations when applied to complex systems, limiting quality improvement efforts. KDD+Viz along with process modeling can be applied to quantitatively measure process factors leading to bottlenecks and delays, to optimize performance in VAHS facilities.