Empirical Validation of the Structure of an Information Quality Model

Quality of information is a critical issue in today’s interconnected society. Yet while most information quality studies posit or depend on a model consisting of various information attributes and their relationships with information quality and each other, few validate the simultaneous theoretical relationships inherent in the overall model structure. Assumptions underlying the major information quality models preclude the use of structural equation modeling via tools such as LISREL and AMOS. This study addresses both issues, empirically validating the structure of an information quality model with partial least squares applied to data from professionals who process patient claims information for health care providers. The structure of the model showed reasonable fit with the data sample, and predicted 59% of the variance in perceived overall IQ for information they received. Although participants rated all model attributes as important to information quality, two well-accepted information quality attributes, Accessibility and Relevance, did not adequately contribute to its prediction in the model. The results support and illustrate a formative approach to modeling information quality and the applicability of the tested model to patient claims information used by health care providers. They also raise questions about the ability of importance ratings to adequately identify information attributes that explain information quality perceptions, and they reinforce the context or domain-specificity of even fundamental information quality attributes. INTRODUCTION & BACKGROUND Quality of information is a critical issue in today’s interconnected society. Poor quality information has caused political controversy and high-profile disasters [16], clinical accidents [12,23], lost productivity [13,25], and failed enterprise [19,29] with estimated costs in the millions of US dollars [13]. Many studies of information quality employ models that posit or depend on theorized relationships between information attributes and overall information quality. For example, the FASB [15] models information usefulness as hierarchically dependent on attributes such as Relevance, Reliability, and Timeliness (Figure 1). Wang and Strong’s factor model [29], widely used in management information systems (MIS), shows information quality as dependent on dimensions such as Accessibility and attributes such as Interpretability, Accuracy, Completeness and Timeliness (Figure 2). Many information quality models (and studies) derive from or depend on one or both of these models, presenting a conceptual hierarchy of elements (dimensions or attributes) leading to IQ. However, few such overall models have had the simultaneous relationships in their structure validated. The FASB model is explanatory and, although Wang and Strong’s measurement model was developed through factor analysis, the model structure was not validated. Similarly, although Lee, et al. [22] assessed perceptions of information actually used, they were unable to assess the structural relationships of the model. Software tools for covariance-based structural equation modeling (SEM) abound (e.g. LISREL [20] and AMOS [1]), and the factor analytic development of the latent constructs of Wang and Strong’s popular model makes this an alluring choice for validating IQ model structure. Chae and Kim [6], for example, used LISREL to empirically test attributes of IQ in a user satisfaction model for mobile internet services. Although they reported robust fit indices, there are several reasons to suggest that such application of covariance-based SEM for IQ modeling is inappropriate. Proceedings of the Ninth International Conference on Information Quality (ICIQ-04)