Using implementation science to close the gap between the optimal and typical practice of quantitative methods in clinical science

Abstract

Quantitative methods remain the fundamental approach for hypothesis testing, but in approaches to data analysis there is substantial evidence of a gap between what is optimal and what is typical. It is clear that diffusion and dissemination alone are not maximally effective at improving data analytic practices in clinical psychological science. Amid declines in quantitative psychology training, and growing demand for advanced quantitative methods, applied researchers are increasingly called upon to conduct and evaluate research using methods in which they lack expertise. This "research-to-practice" gap in which rigorously developed and empirically supported quantitative methods are not applied in practice has received little attention. In this article, we describe how implementation science, which aims to reduce the research-to-practice gap in health care, offers a promising set of methods for closing the gap for quantitative methods. By identifying determinants of practice (i.e., barriers and facilitators of change), implementation strategies can be selected to increase adoption and high-fidelity application of new quantitative methods to improve scientific inferences and policy and practice decisions in clinical psychological science. General scientific summary:Making studies more replicable will require more effective use of statistics in research, but there is a large gap between how statistics are applied in psychological research and how they should be applied. The current article describes how the lessons of clinical implementation science, which has focused on getting evidence based treatments into community practice settings, may be applied to improve research in clinical psychology.

Interpreting interaction effects in generalized linear models of nonlinear probabilities and counts

“…Further, there remains an ongoing and pressing need to improve the accessibility of these approaches for use among methodological non-experts (K. M. King, Pullmann, Lyon, Dorsey, & Lewis, 2019;Sharpe, 2013). We provide computational and inferential solutions developed in this paper through open-source R code.…”

Potentially harmful therapies: A meta-scientific review of evidential value.

“…Indeed, many of the same factors argued to underlie low replicability in other areas of psychology (e.g., low statistical power; Maxwell, 2004) have a long history of being acknowledged—and ignored—in clinical psychological science (Cohen, 1962). The replicability of clinical research might also be challenged by unique subdisciplinary factors, such as the burden of recruiting samples for psychotherapy trials and tightly accredited training curricula often leaving little in the way of time or encouragement for trainees to invest in deeper levels of methodological and analytical acumen (King, Pullmann, Lyon, Dorsey, & Lewis, 2018). The possibility, if not likelihood, of unreplicable findings in clinical research is all the more concerning in light of clinical science being brought to bear on the everyday lives of consumers, health-care providers, managed care administrators, and policy makers seeking evidence-based treatments.…”