Congress Abstracts 2005

5

IMPROVING RESEARCH DATA INTEGRITY: APPLYING NOVEL TOOLS IN A LONGITUDINAL BREAST CANCER CLINICAL TRIAL. Patrick McNees, PhD, Applied Health Science, Orlando, FL; Karen Hassey-Dow, PhD, Sreeramen Ramaswamysanthanam, MS, Ganesh Subramanian, MS, and Victoria Wochna-Loerzel, RN, MSN, OCN^®, University of Central Florida, Orlando, FL.

Improving data integrity is a process and not an event. Yet methodologies that systematically improve most research processes and produce higher quality research data have not been specified or systematically evaluated. Longitudinal clinical trials present particularly salient challenges that can limit analyses, threaten interpretation and/or conclusions drawn from the data.

The objectives of this paper are to (a) describe the application of engineering quality improvement techniques to a longitudinal quality of life (QOL) clinical trial, and (b) determine the impact of engineering techniques to maintain and improve data integrity.

Deming’s quality improvement framework and principles of statistical process control form the theoretical underpinnings for this work.

The investigators are conducting an ongoing randomized QOL clinical trial that will accrue 250 subjects. Subjects have either 6 or 7 monthly data accrual points. Based on initial results of the first quality audit of 50 subjects’ data, the investigators identified improvements needed, and designed and implemented a novel and systematic approach to full quality improvement. This process included application of engineering techniques such as: statistical process control, item sampling, data review, quality audit, and feedback control.

A behavioral observational model was paired with statistical process analyses for both informing the research processes and performing analyses. The specific formula for estimating reliability was r.coeffecient = [(agreements)/(agreements + disagreements). Other data were tabulated from data entry records.

The techniques used in this study resulted in incremental improvements including: greater inter-rater reliability, decreased error in missing data, improved data entry, enhanced data flow coordination, and reduced person hours involved in data management. Baseline reliability was 0.9676. While relatively high, analysis of first 110 subjects, reflect a 41% reduction in data errors from baseline. Thus, applying quality improvement engineering techniques and focusing on controllable sources of variability resulted in significantly fewer errors and improved data quality and integrity. Improving quality or data integrity is not an event, but a process. As such, application of engineering quality control techniques can result in improvement towards error-free data, while simultaneously providing an ongoing system for continuing to improve future research projects.