Standing in the Shadows of Theory

As the field of biomedical informatics matures, investigators and systems implementers increasingly rely on theories, models, and frameworks to guide their work. Of these three, theory provides the greatest value to any discipline, because theory provides well-validated pathways to link observed phenomena with foundational knowledge, thus enhancing efficiency and generalizability. Theories are sets of propositional statements that explain, predict, or prescribe the relationship between concepts capable of being tested, verified, or refuted through empirical work. Theories serve to illuminate concepts deemed essential to understanding complex situations, such as why people with AIDS accept an electronic coaching system, ¹ or predict future states, such as how quickly physicians will adopt electronic health records. ² Reliance on theory demonstrates a level of sophistication in any discipline, and, as such, this increase in attention to theory-driven research and practice is much welcomed in biomedical informatics.

So long as an applicable theory or theories exist, the necessary first step in any investigation or implementation should be to select the theory that will guide the endeavor. This is a significant right and responsibility. Each time a researcher or systems implementer selects a theory to guide his or her work, he or she is purposely attending to certain aspects of a given situation, and leaving other aspects in the shadows. This is not to say that the shadowed dimensions of a situation are unimportant, but rather that the particular story that the investigator wishes to tell, or the specific implementation process the implementers want to address, is best characterized by the theory selected.

The article by Callen and colleagues ³ highlights the importance and value of employing theory when implementing and evaluating clinical information systems (CIS). Resting their arguments on critical contributions from previous investigations of successful and unsuccessful implementations of clinical information systems, the authors argue for careful, systematic considerations of the context within which clinical information systems are implemented, calling for attention to user characteristics, clinical department, and organizational features that influence the process of adoption of clinical information systems.

As aptly documented by Callen and colleagues, many theories have proved useful to guiding the design and implementation of clinical information systems and predicting or explaining their consequences. Like health informatics standards, the good thing about theories for CIS implementation is that there are so many to choose from! Some direct CIS implementers select theoretical underpinnings that attend to key actors and early adopters; others choose theories that emphasize critical work flows or institutional policies that inhibit or advance the adoption of clinical information systems applications.

Theory serves many purposes in biomedical informatics. Some theories are structural, identifying key elements or players in a given situation. Others are process-focused, highlighting the actions and consequences expected when particular events initiate a sequence of actions. Theories serve to identify key elements of a critical situation, such as a clinical practice environment, which highlight where attention should be paid. Some theories not only address the critical elements of a situation, but also explicate the relationship between predictors and outcomes, for example, the contribution of computerized provider-order entry systems to patient safety goals. The summary provided by Callen and colleagues illustrates the wide range of theories, models, frameworks, and guiding principles that abound in the literature attempting to prescribe, guide, and predict successful CIS implementation. They criticize existing theories as being insufficient to guide CIS implementation, having left too many key ideas in the shadows. Basing their assessment on observations gleaned through interaction with clinicians actually using a special-purpose CIS, Callen and colleagues complete and posit expanding existing sociotechnical theories with attention to characteristics of clinical environment (diversity and differentiation) currently left in the shadows by existing theories.

Theories best serve biomedical informatics investigators and clinical information systems implementers when the theories originate from a plausible set of premises, there is a logical congruence between elements, and, most importantly, when they are validated empirically. The utility of any theory applied to CIS implementation is bounded by the scope of the theory itself. While it is tempting to seek an omnibus, complex theory to examine all possible components of a situation, as Callen and colleagues did, it is also worthwhile to select more focused theories that highlight specific phenomena and processes. Selection of theory to guide a study or a systems implementation rests on both scholarly reasoning and artful insight. But, always, theories leave some aspects of the actual unaddressed and unattended to—i.e., standing in the shadows. These ignored aspects still continue to exert influence on the situation, but because of the specific theory’s perspective are not considered germane to the understanding of the phenomena of interest. This ‘shadow play’ does not negate the value of the theory, but rather serves as a reminder of the responsibilities for those who select theories that other forces may be at play that interest their phenomena.

Biomedical informatics investigators and systems implementers are urged to strengthen the rigor and potential value of their works though judicious selection of theories. Selection of theory to guide a study or a systems implementation rests on both scholarly reasoning and artful insight. Selection rests on the congruence between what the theory highlights, as well as what is left in the shadows. Most importantly, theories that are well validated with empirical work and scholarly review hold the greatest value.

While theory-driven research has significant value both for the individual project as well as for extending knowledge in the field, sometimes no suitable theory can be found. In those cases, investigators and implementers may rely on theories external to the biomedical informatics domain (e.g., sociotechnical system theory) or use their project to generate the foundations for new theories that may prove valuable to the field as a whole. We look forward to greater specification and validation of Callen and colleagues’ work, demonstrating the contributions of the ideas of diversity and differences to explicating good ways to implement CIS.