Deming and the Knowledge of Variation (Harnessing Variability Part 2)

Introduction

This issue of The Human Stream is the fourth instalment in our five part series on ‘reliability thinking’ in patient safety. This week we look at W Edwards Deming’s ideas on the topic of variation. Deming is widely (and rightly) regarded as one of the most influential business theorists of the 20th century.

Understanding variation (or rather ‘knowledge of variation’) was a key tenet of Deming’s management philosophy and he wrote extensively on the need for organisations to do better at understanding the sources of variation in process and quality. Deming directed much attention in his later writings (such as in ‘The New Economics’) at the American manufacturing sector, which he felt was in steep decline in the 70s and 80s due to its neglect of quality and tolerance for poor management practices.

Within his critiques and exhortations for change, Deming clearly emphasised the usefulness of scientific (statistical) methods to study variation in processes but he did so with a far more nuanced, systems-oriented approach than is typically employed in the use of these tools today.

While statistical process control (SPC) charts have become commonplace in healthcare quality and patient safety (for instance, as tools employed by institutions to model and monitor adverse event rates), we stand to gain tremendously by situating these methods closer to the systems improvement philosophy within which Deming promoted their use.

In this short issue of The Human Stream, we pick out some salient features of Deming’s ideas on improvement and management as they apply to understanding and managing variation.

Appreciation of the System

Deming’s System of Profound Knowledge (SPK), which in many ways represented the culmination of his life’s work in quality management, placed an “Appreciation of the system” (the first of four tenets) ahead of a “knowledge of variation”, signalling the importance he placed on developing a fulsome understanding of how a system works, as a prerequisite to attempts at improving it.

Deming was particularly critical of efforts to modify processes in response to variation when management had not yet developed a profound understanding of the systems of production - likening it to “tampering”.

This raises a key question in the application of SPC methods in the context of patient safety. For the rates that we track using these methods, do we have a sufficient understanding of underpinning systems (and how they shape outcomes) to effectively act on signals?

Distinguishing common from special causes of variation

Deming advocated for the use control charts as a means of helping organisations distinguish between the normal, everyday variation that is typical of the systems they lead (‘common’ causes) and the variations in process and outcomes that stem from other sources (‘special’ causes)¹ . However, Deming promoted this distinction in a way that can seem counterintuitive to those unschooled in his philosophies.

Deming observed that many organisational leaders tended to reflexively view sources of variation as ‘special’ (such as a specific human failure, a external cause, the fault of suppliers or unexpected breakdowns of equipment) as a way of avoiding responsibility for fundamental shortcomings in the design and management of their systems.

Deming observed that this was a reflection of late 20th century management approaches in the West where deep systemic causes of unwarranted variation in outcomes tended to be framed as problems created by workers when in fact they were inherent to the design of the organisation (and therefore a responsibility of management rather than workers).

Deming frequently railed against the incentives, metrics, quotas and corrective measures that were inflicted on frontline workers to solve problems that stemmed from the design of the system, “robbing them of their pride in workmanship”. While Deming placed much emphasis on high quality data, he was deeply critical of prevalent management thinking that elevated data (performance measurement) above local knowledge and the valuable insights of frontline workers.