Healthcare systems are vastly complex and set in an environment that is equally (if not more) complex [3, 4]. Besides consisting of a multitude of individual components (e.g. doctors and nurses, technological artefacts, regulatory pressures), systems of such complexity also possess subsystems (e.g. anaesthesiology team, general surgery team) that are working to achieve their own goals [31]. These goals are not always compatible, however, resulting in conflicts that need managing. Those involved would have to make decisions based on the situation and some of these decisions might require the sacrificing/trade-off of safety to achieve a particular production goal or to live up to other duties [37, 38]. Typically, this trade-off does not yield any immediate negative consequences [39]. Therefore, those involved would be misled into assuming that the trade-off is acceptable and it becomes part of the normal process. When another conflict emerges and another trade-off is made with no adverse results, this second trade-off might be once again be assumed to be acceptable and becomes part of the normal process. This process (known as normalisation of deviance) will repeat itself, slowly nudging the system towards greater risks until an adverse event takes place.

Despite the risks involved, those within the system are unlikely to be aware of this drift to failure as signs are typically only noticed by those outside of the system (e.g. accident investigators) after an accident has occurred [24]. To those within the system, seemingly poor decisions in hindsight are actually rational, given the contemporaneous circumstances [31]. While seemingly a bad phenomenon, the drift away from safety is not necessarily a negative indicator of an organisation’s performance [24]. Rather, it is simply a by-product of a complex system adapting to the challenges from both within itself and the environment. The challenge is to ensure that the clinicians involved understand the role and importance of these trade-offs (i.e. clinical sensemaking) [40].

So what are the elements that contribute to a system drifting towards failure? At present, it is theorised that at least five factors are involved, namely (a) scarcity and competition, (b) decrementalism, (c) sensitivity to initial conditions, (d) unruly technology, and (e) contribution of protective structure [24].

Scarcity and competition refer to an organisation experiencing a lack of resources, and facing intense competition [24]. Rasmussen suggested that a typical organisation has to work within three boundaries, the first being economic, the second being safety, and the third being workload [41]. Working beyond the economic boundary means that the organisation would not be able to maintain itself financially, while crossing the safety boundary means that the organisation’s operation is highly dangerous (e.g. patient’s well-being may be endangered). Lastly, exceeding the workload boundary means that the people and/or the technologies within the organisation are no longer capable of carrying out their work. As mentioned earlier, organisations generally drift away from the safety boundary to satisfy production pressure since the loss of safety is rarely felt while the reaching (or not reaching) of production pressure is tangible [37].

Decrementalism means that an organisation moves to the edges of the safety boundary over a series of small steps (instead of instantaneously), as it attempts to meet production pressure, as explained earlier [24]. This should not be confused with normalisation of deviance, which refers to trade-offs made in response to abnormal situations (e.g. high demands) being seen as the new norm.

Sensitivity to initial conditions (otherwise known as the butterfly effect) essentially argues that seemingly small factors in a system’s starting conditions can lead to large failures, as these factors interact in novel ways to give birth to unintended consequences, pushing an organisation towards the edge of the safety boundary [24]. Unruly technology refers to the gap that exists between how designers of a technology think it will work, and how the technology actually works when exposed to the environment [24, 42]. For instance, the introduction of poorly designed health information technology in some hospitals has been argued to cause issues such as (a) making it difficult for physicians to gain a proper understanding of a patient’s condition, and (b) producing reports that lack information value, due to the technology’s insistence of using standard phrases [43].

The last factor is the contribution of protective structure, which suggests that the protective structure that was deliberately created to keep the operation safe can end up contributing to a drift towards failure [24]. One example is a safety or governance department that, through its generation of many different layers of defence and guidelines, actually contributes to complexity, thereby rendering real sources of risk less visible to the sharp end users.

Despite the potential for drift to result in unwanted consequences, a definitive solution to reduce an organisation’s potential for drift does not appear to exist. Nonetheless, this section will be devoted to the exploration of some of the ideas in the hopes that some would find it useful.

As suggested earlier, signs of drift are not always obvious to those within the organisation [24]. Therefore, one plausible approach of reducing an organisation’s potential for drift is to study how decision makers make sense of the information environment (e.g. why they take in certain bits of information and ignore others) as well as how they make and rationalise their decisions [44]. However, this may not be a fruitful endeavour since an organisation’s drift into failure is usually only known after an accident has occurred and any knowledge gleaned might be specific to that accident and have little applicability in other contexts.

Arguably, a decision maker must pay attention to multiple sources of information and invite doubt to make the best possible decisions [45]. But this may be an idealistic notion as decision makers may be bombarded with an enormous amount of information, which would require a long time to process, and immense cognitive resources [24]. Furthermore, tell-tale signs of drift may be weak or unbelievable, and hence go unnoticed [37].

Another potential approach would be to move the organisation away from the safety boundary, reducing the likelihood that it will be crossed and produce an accident [41]. Examples include reducing production pressure or investing in proven safety methods. However as with the above, expecting an organisation to reduce production pressure might be wishful thinking. Even if an organisation chooses to invest in proven safety methods, it is highly likely that production pressure will follow this increase as staffs would be expected to produce a greater output with the same resources (i.e. be more efficient) [37].

In sum, while there has been several suggestions on ways to diminish an organisation’s potential for drift, these suggestions each come with their own caveat . Nevertheless, this does not mean that it is impossible to reduce an organisation’s drift potential since there may be other solutions that have yet to be explored. For example, Rochlin and his colleagues have observed that the various subsystems on board a naval aircraft carrier were able to balance multiple constraints and pressures to consistently produce smooth performances [5]. Perhaps an in-depth study on how these subsystems co-operate and negotiate with one another might yield some useful information.

As argued earlier, drift is not an indicator of an organisation’s failing, but a sign of it adapting [24]. It can appear in many forms, such as procedural violation (also known as workarounds). Workarounds appear to be frowned upon as it deviates from rules and regulations, which some consider sacred [46]. Such a viewpoint may have its merits, for deviations from rules and regulations have resulted in unwanted results. For instance, it was argued that non-compliance with rules and regulations contributed to an incident where the wrong patient was given an invasive procedure.

However, it might be a mistake to assume that all forms of procedural violations are bad. For example, one form of medical guidelines in the USA specified the use of levofloxacin for community-acquired pneumonia [47]. But others have suggested that a physician should not always follow these guidelines as levofloxacin is an expensive form of antibiotics that not all patients can afford, and not having antibiotics could lead to patients’ conditions worsening [48]. To avoid this outcome, physicians need to deviate from the rules and regulations and prescribe a different and more affordable form of antibiotics. Furthermore, each patient has their own unique co-morbidities and medical history, making it near impossible to create a set of guidelines to address each case. Under such circumstances, physicians should be allowed to act as they see fit instead of being penalised for not complying with procedures. In other words , procedural violation may not always be a bad thing as it captures the local wisdom of the providers.