Seite - 4 - in Applied Interdisciplinary Theory in Health Informatics - Knowledge Base for Practitioners

Box 2. The importance of theory in aeronautical engineering: a sobering example In the 1950s there was a disastrous series of 13 aircraft accidents in which the world’s first passenger jet airplane, the de Havilland Comet, exploded in midair or on the runway, with a total loss of 426 lives [8]. After some delays, the cause of most of these Comet accidents was traced to cracks in the fuselage of planes, which by then were a few years old. Materials scientists hypothesised that the cracks were caused by metal fatigue starting at the corners of the square fuselage windows and spreading during multiple pressurization cycles. This clear, generic, testable and enduring theory was confirmed by metallurgists examining fragments from crashed planes and subjecting new fuselage segments to multiple compression cycles in a water bath. Fuselage metal fatigue as a cause of aircraft failure is now eliminated by rounded window corners. This is a compelling example of how a theory was identified, tested, then applied universally to make a very complex device more reliable. A long series of such events has led to aeronautical engineering becoming a theory-based discipline in a way that health informatics sadly cannot yet claim to be. 2. Another perspective on “theory”, where do relevant theories originate, and which theories are useful? Although Nilsen defines theory as a set of analytical principles designed to structure our observations, understanding and explanation of a phenomenon in the world, I would advance a slightly nuanced perspective on what definition is needed for “theory” in health informatics: “a concise, testable predictive principle that can guide the design, development or implementation of clinical information systems”. Such predictive principles can be derived from an existing theory or can be the basis for a new theory after sufficient testing in multiple settings. Predictive principles derived from theories originate in many disciplines, including psychology; management, implementation or computer science; or healthcare [10]. Some examples: • Psychology: theories of information design, behavior change or self-efficacy • Management science: innovation theory, organization theory, marketing theory • Implementation science: active implementation of guideline recommendations, informed by a study to elicit relevant barriers and enablers, is more effective than simple dissemination • Computer science: software engineering theories, human computer interaction theories, persuasive technology theory • Healthcare: investigation is more efficient when test-treatment threshold is considered; prescriptions are safer when drug allergies, interactions and disordered drug metabolism or excretion are considered. Once we know where we can locate potentially useful theories, we need to understand which predictive principles derived from these existing theories are most likely to be relevant to health informatics, and to be useful so that we can select the most promising. Or we can define and test new predictive principles to help develop new theories. My theory (!) is that, to help our discipline, a predictive principle needs to be clear, predictive, testable, generic but relevant and enduring [10]. Table 1 explains these terms, gives examples and some counter examples of hypothetical predictive principles that would violate each criterion. J.C.Wyatt /TheNeed forTheory to InformClinical InformationSystems4