Secondary Analysis of Data Collected for Clinical Purposes: Lessons Learned

Tuesday, 13 July 2010

Deborah A. Saber-Moore, RN, BSN, MS1
Anne E. Norris, RN, PhD, FAAN1
Greg Trompeter, PhD2
1College of Nursing, University of Central Florida, Orlando, FL
2Kenneth G. Dixon School of Accounting, University of Central Florida, Orlando, FL

Learning Objective 1: describe four challenges that may be encountered when analyzing complex secondary clinical data.

Learning Objective 2: discuss five strategies for addressing the challenges of complex secondary clinical data analysis.

Purpose: Secondary analysis from data specifically collected for clinical purposes may reveal interesting, unexpected, and thought-provoking results. The process can be challenging because the data are not initially collected to answer specific research questions, and data management strategies are learned through trial and error. The healthcare environment is replete with clinical data that can be secondarily analyzed.  However, obstacles and problems that are encountered may lead to inaccurate analysis and findings.  The purpose of this presentation is to illustrate challenges encountered during the secondary data analysis of a clinical data set, and communicate strategies used to resolve them.

Methods: Data for secondary analysis were obtained from an organizational psychiatric consultation practice from 1978-2008 (n=1821).  Data included demographics, work history, medication profiles, psychiatric measures, cognitive ability measures, workplace environment measures, peer’s perception of individual potential to involve company in fraud, and notes from clinical interviews.  Electronic records and paper questionnaires provided qualitative and quantitative data for over 214 variables. Challenges were encountered as measures were examined

Results: Four areas that presented challenges were:  mixed electronic and paper data-base element formats, outdated clinical management software, unusual or unfamiliar measures, and inconsistent data coding and missing data.  Strategies for addressing these challenges included: (1)  identifying and working closely with the organizational gatekeeper and computer programmer, (2) reviewing the conceptual basis and pre-existing psychometric evidence for each tool, (3) comparing versions in the literature against those used by the clinicians,  (4) investigating clinician assessment practices, and (5) careful and intensive data cleaning work.  Use of these strategies will be illustrated using an analysis that involves Beck’s Hopelessness Scale.

Conclusion: Secondary data analysis can be challenging.   However, when challenges are anticipated and known techniques are employed, the processes of data management and analysis may help to reveal accurate, scientifically valued findings.