An Algorithm for Creating Alternate CES-D Composite Scores: Advancing Research through Methodological Clarity

Abstract

to contribute to health research literature dealing with psychological assessment by primarily arguing that radical clarity in the procedures undertaken for creating data are necessary if research is to detect a more probable truth as it distances itself from absolute uncertainty. In contribution towards these goals, the project presents and makes available a novel statistical program (originated by author without the use of any external sources) to help data managers create five different variables that can help secondary data user understand the withinperson level of in-completeness for the full CES-D scale and its four sub-scales separately. BackgroundAt the most fundamental level, investigating human health through statistical techniques (e.g., using regression equations) requires numerically coding their behaviors, perceptions, physical attributes, and environments. In order for the legitimacy of statistics to be employed in analyses, human behavior must be converted into a numerical existence-i.e., the abstraction of the physical world must occur with the use of numbers. Numbers attributed with meaning are then explored in a world of equations where results from samples are used to infer population characteristics or where statistical significance is inferred typically using frequentist views on probabilities. Unfortunately, a study participant could be born into the world of numbers through ambiguous mechanisms and careless documentation. Paying attention to how information is handled at the most early stages of data creation is crucial to the value of any subsequent work because the quality of data influences the potential value of results. AbstractThe Center for Epidemiologic Studies Depression Scale (CES-D) is a widely used screening test for depression symptomatology in population studies. Publications have dealt with the psychometric properties of the instrument. Until now, no published work has examined the how questionnaire items are handled in data processing protocols. This technical report introduces the "CESD-Flag" macro algorithm using SAS ® 9.3 syntax. The main contribution of the statistical program in this report is to provide data managers with a program that creates five CES-D flag variables to help identify the level and location of missingness in CES-D related data. The paper highlights implications from how data editing algorithms handle missing or inconclusive responses to CES-D questions. If absolute transparency should be encouraged as a desirable goal, then the methodology used to "clean" CES-D data should be provided along descriptive statistics for the five flag variables (measure of item incompleteness) created by the CESD-Flag program. The assignment of responses through non-evidence based algorithms merits further attention as the use of CES-D continues to proliferate internationally.

The incidence and predictors of depressive and anxiety symptoms in older adults with vision impairment: a longitudinal prospective cohort study

“…Instead we explored the final model for these two variables by adding them to the final model with only the participants of T2 and T4 (when those two variables were measured). Studies show that missing data in items of the CES-D scale appear to be ignorable, 40 80% of the 20 questions need to have logical value for a person to be assigned a CES-D composite score, 41 and multiple imputations are complex and do not increase precision in the estimated rate of change in the end point in linear mixed modelling. 42,43 Therefore, we chose to use zero imputation for the few items that were missing in the CES-D (n T0 = 26, n T1 = 29, n T2 = 26, n T3 = 24, n T4 = 40), HADS-A (n T0 = 0, n T1 = 5, n T2 = 10, n T3 = 4, n T4 = 6), LVQOL (n = 61) and AVL (n = 19) if there was at least an 80% completion.…”

A critical analysis of the internal logic in the Life-Space Assessment (LSA) composite score and suggested solutions

“…The core criticism in this paper is that the implicit logic of the LSA produces a data-editing algorithm that uses actual and created responses. Under the fundamental view that creating data should be avoided in health research, the main argument in this critique is that using data editing algorithms 20 in the computation of any composite score should be avoided. The report offers a solution by providing readers with an easy-to-use statistical program that estimates a Non-Data-Edited LSA Composite Score (NDE-LSA-CS).…”

Number of Allocations as a Measure of “Within-Person Item-Incompleteness” (WPII): 2005 to 2010 Data from the American Community Survey

“…The specific aim of this study is to use ACS flag variables to sum the number of allocations (referred by some as imputations) within an individual in order to determine if and how levels of data edits vary as a function of basic demographic factors. Please note the paper is not directly concern with "item nonresponse" as data editing algorithms have the ability to change recorded responses (Siordia, 2014b). In this study, "allocations" include changes to both responses and nonresponses-where high within-person level of allocation may signal a "high level of questionnaire incompleteness".…”