Understanding Clinical Workflow Through Direct Continuous Observation: Addressing the Unique Statistical Challenges

Abstract

Direct observation of clinicians is an effective way to study the inner workings of clinical work. Of the many observational approaches, workflow time studies generate a continuous, fine-grained record of individuals’ tasks and interactions, providing a foundation from which to explore a wide range of research questions within a quantitative framework. Although the concept of recording time-stamped tasks according to predefined categories is relatively simple, the complexity of the settings to which workflow time studies are applied often generates data that are not amenable to conventional statistical analysis methods. In this chapter, we examine some of the fundamental considerations for analysis of such data, with a particular focus on the use of nonparametric methods as a way to circumvent limitations with standard methods. This includes sampling strategies, inter-observer reliability assessment, calculating confidence intervals and performing hypothesis tests. We also discuss nonparametric possibilities for multivariate analysis. While quantitative observational studies of clinical work have great potential to help us understand clinical workflow, it is essential to apply statistical methods with care, to acknowledge their limitations and to identify areas where bespoke methodology needs to be developed. To improve the integrity of such research then requires more explicit and open discussion of quantitative methodology, and this chapter aims to initiate these discussions.