Modeling and analysis of uncertainty in clinical laboratory measurement processes

Varun Ramamohan, Purdue University

Abstract

A statement of uncertainty associated with the result of a measurement describes the quality of the measurement procedure. Information about the quality of the measurement result is particularly important in the case of clinical laboratory measurement processes since clinical laboratory test results inform every stage of medical decision-making, from diagnosis and prognostic assessment to determining drug dosage prescriptions. This study presents a methodology that enables estimation of the uncertainty associated with a clinical laboratory measurement system via the development of a simulation model of the measurement system. The methodology involves the development of a physics-based mathematical model of clinical laboratory measurement processes, and utilizes the Monte Carlo method to characterize the long-term behavior of such a model with uncertainty in its parameters and variables. This study was carried out in collaboration with Roche Diagnostics Corporation based in Indianapolis, IN (USA) and the Mayo Clinic at Rochester, MN (USA). In the first part of this thesis, general guidelines for the development of the mathematical (uncertainty) model are presented, and the implementation of these guidelines is illustrated by building a model of the serum cholesterol laboratory test. The uses of this methodology to estimate the contribution of the various sources of uncertainty to the net system uncertainty, and to optimize quality control policies and calibration policies followed in the laboratory, are illustrated with examples. The second part of the thesis describes the use of uncertainty models to quantify uncertainty in instrument calibration on the measurement result distribution. Further, the use of the model to identify optimal combinations of calibrator concentration levels that minimize the uncertainty at medical decision points is explored. The third part of this thesis describes the use of this methodology to model time-drift in the performance of the instrument used to perform the laboratory tests. Recalibration of the instrument is postulated as a method to nullify the effect of instrument drift, and the model is used to estimate the time interval between successive calibrations of the instrument. Finally, we conclude the thesis by describing the development of a category-specific mathematical framework for building uncertainty models of laboratory tests that belong to a given category.

Degree

Ph.D.

Advisors

Yih, Purdue University.

Subject Area

Biomedical engineering|Industrial engineering|Operations research

COinS