Utilising visual risk communication to alert hospital staff to potential hazards to patient safety in routinely collected data: a visually compelling fluid balance bar chart
Abstract
This paper presents the theoretical basis for, development, and implementation of a new risk management tool for use in hospitals. It represents a spin-off from ongoing doctoral research exploring the human factors involved in adverse events involving hospital patients.
Much of patient care comprises routine tasks designed to monitor patients physical condition and provide early warning of problems. Failure of routine procedures may account for up to 30% of the adverse events recorded annually in the NHS. One theme emerging from data collected for this study is that staff may fail to prioritise routine tasks, however important to patient safety, choosing instead to respond to (often unanticipated) hazards perceived to require more immediate action.
One such routine task, notorious for being inadequately and inaccurately carried out, is the recording and monitoring of patients' fluid input and output. Observational and interview study data indicated that, despite awareness of the serious consequences of failure to recognise early signs of deterioration, both fluid balance charts and their contents (lists of figures) were frequently ignored by both medical and nursing staff.
Consideration of ways of triggering awareness of fluid balance data, particularly cumulative totals which show trends over time, prompted design of a fluid balance bar chart which is visible even across the ward, providing 'at a glance' information about a patient's fluid balance status. Piloted in the hospital trust involved in the research, the chart proved acceptable to staff and is in process of implementation across the trust. The chart is featured on the NHS Live website and is attracting wide interest.
Keywords
risk perception, risk communication, adverse events, patient safety
1. Introduction
It is estimated that one in ten patients admitted to hospitals in developed countries will be unintentionally the victim of an error, and around 50% of these errors will be preventable (Department of Health 2006; Health Select Committee 2008). The whole issue of patient safety, medical error and adverse event reporting, once a relatively neglected area (Reason 1995), has become a high priority in health care systems in this country and across the world (Department of Health 2001). Much effort is expended by clinical management in NHS hospital trusts on trying to reduce the incidence and severity of adverse events.
The literature suggests there are multiple factors which affect people's identification of hazards to themselves or others in their environment (e.g. Turner 1976, Beck 1992, Pidgeon, Kasperson and Slovic 1993, Löfstedt R E and Boholm Å 2009), and that this is as true in health care as in other settings (e.g. Reason 1995, Hoff et al 2004, Alaszewski and Coxon 2008). This paper will argue that people's perceptions of the risks posed by hazards, and the mediation of their behaviour by those perceptions is central to understanding patient safety issues.
The doctoral research which forms the background to the development of the particular risk management tool which this paper describes, is concerned with examining the links between perceived risk and behaviour in real-life situations. Specifically, the research collects observational and interview data about people's behaviour and their perceptions of what factors motivate that behaviour. The setting for the research is one which, by its very nature, tends to present more hazards than most in everyday life, the wards of an acute hospital. The study is designed to map the hazards hospital staff perceive in their environment, both in relation to patients and to themselves, to examine what factors mediate their response to perceived hazards, and to consider the extent to which individuals' perceptions of risk may impact on, and contribute to, the aetiology of adverse events concerning patients.
It is important to note here that the content of this paper necessarily focuses on the direct contribution of individual behaviour to patient safety issues. In so doing, it does not seek to minimise the role that may be played by other, wider and more systemic factors operating at organisational and societal levels. Indeed some of these factors, as perceived by individuals within the systems, are discussed briefly within the paper. However, the research on which this paper is based will explore them in greater depth than is appropriate here.
The main purpose of this paper is to present the theoretical basis for, development, and implementation of a new risk management tool for use in hospitals. The tool represents an unexpected spin off from some early research findings about how clinical staff perceive hazards and prioritise their work in relation to them. The data suggest that staff do not always find it easy to weigh severity of consequence against immediacy of threat when deciding which hazard is most salient to them and how to prioritise action. Much of patient care comprises routine tasks designed either to provide early warning of deterioration, or to prevent mistakes. Faced with two potential hazards to patient safety, one perceived as non urgent and part of normal routine procedures and one perceived as abnormal and requiring an urgent response, consideration of the relative severity of the threats posed may often be secondary to the perceived need for action, even when the more apparently imminent hazard poses a lesser threat. Such predispositions may well contribute to the large number of incidents involving failure of routine procedures designed to protect patients from harm. These are estimated to account for up to 30% of the adverse events recorded annually in hospitals (Brennan et al 1991).
Development of the risk management tool described here arose from consideration of observation and interview data relating to a particular routine task, the recording and monitoring of patients' fluid input and output. This task is notorious for being inadequately and inaccurately carried out by hospital staff, and the data collected supported this view, indicating that, despite awareness of the serious consequences of failure to recognise early signs of deterioration, both fluid balance charts and their contents (lists of figures) were frequently ignored by both medical and nursing staff. The new tool draws on ideas about perception and risk communication to express information about a patients condition in a visually compelling format. This has the potential to instantly heighten awareness of fluid balance information and trigger an urgent behavioural response if that is required.
The paper is divided into several sections. Following this one, Section 2 begins with a brief review of the general literature on risk perception and decision making, focusing on how people identify and respond to multiple perceived hazards in their environment and in particular on what is known about how they determine which hazards are most salient and need to be prioritised. Section 3 looks at the development of patient safety as a risk issue in health care, focusing on failure of routine procedures to protect patients, and in particular on what is known about inadequate fluid balance monitoring. In Section 4 the background to and focus of the research is discussed, including the development of the project over time and a brief description of the study design and methods. The paper goes on (Section 5) to present some empirical data in support of, first, the general theme that hospital staff perceive many qualitatively different hazards in their environment; and second, that they may have difficulty in titrating severity of consequence against immediacy of threat in deciding how to prioritise action, particularly in relation to routine tasks such as fluid balance recording. Section 6 describes the development of the new risk management tool, a bar chart which converts numerical data into a visually compelling and readily understandable format. Sub sections within this relate to the various stages in the process; chart design, early consultation and discussion, piloting use of the chart on a ward, revision in response to feedback, and finally gaining management support for implementation throughout the hospital trust. The final section attempts to draw together the various themes with regard to risk perception and patient safety addressed in the paper and to suggest areas which may repay future research.
2. Risk perception and decision making in uncertainty
This section offers an explanation of the ways in which the terms 'risk' and 'hazard' are defined and used in this paper and then goes on to explore some of the literatures on risk perception and decision making in uncertainty which underpin ideas about how people identify, rate and react to threats in their environments. Towards the end of the section, the contestability and negotiability of perceptions of risk is briefly addressed, and the idea is introduced that an individual's internal debate about how to prioritise response to perceived hazards may also be seen as a process of negotiation. Finally, psychological evidence about selective attention is introduced as a way of conceptualising individuals' behaviour in the face of multiple stimuli and as a theoretical approach which has underpinned the development of the risk assessment tool described in this paper.
2.1 Defining risk and hazard
Risk may be expressed in mathematical terms of chance and probability, the statistical likelihood that a particular outcome will occur. Here the idea of risk is value free; for instance, it would be correct to say that there is a high risk of not being run over when the road is empty of traffic. The use of the word “risk” seems awkward in this context because, as cultural theorists have argued, the concept of risk has come to be associated with danger and loss (Baird and Thomas, 1985; Levitt and March, 1988; Douglas, 1990). There is thus a mismatch between decision theoretic conceptions of risk, where value free choices are based on statistical probabilities, and the perception of those making risky choices, for whom the concept of risk is inextricably associated with negatively valued outcomes of choice. As March and Shapira (1987, p. 81) note, there is:
"a persistent tension between “risk” as a measure (e.g. the variance) of the distribution of possible outcomes from a choice, and "risk" as a danger or hazard. From the former perspective, a risky choice is one with a wide range of possible outcomes. From the latter perspective, a risky choice is one that contains a threat of a very poor outcome.”
Social theorists such as Beck (1992) and Giddens (1991) argue that this concept of risk as a threat has become fundamental to the way in which people organize their world, and it is in this way, rather than as a value free mathematical concept, that the term is used here.[1]
It should be noted that, as in the quotation above, the terms 'risk' and 'hazard' are often used interchangeably. This presents a problem for a paper such as this which seeks to differentiate between something that may potentially pose a danger or threat and the perception and evaluation of that threat by an individual. For this reason, 'risk' is used here only in the second sense, i.e. as the evaluation of the degree of threat posed by something perceived as hazardous.
2.2 The perception of multiple, qualitatively different hazards
Everyone is presented, all the time, with not one but many different hazards, responses to all of which must somehow be accommodated in behaviour. Some hazards may be perceived as imminent, offering a major threat to life or wellbeing and demanding immediate action, others may be perceived as less threatening, enabling them to be temporarily, if not indefinitely, disregarded. For instance, Ulrich Beck (1986,1992), in defining his "Risk Society", suggested that everyone, is to some extent threatened by potential disaster on a global scale, mostly as a consequence of increasingly complex interventions in natural processes. Even though concerted action might reduce the threat, the effect of one individual's behaviour is small, may not be visible during an individual's life time, and may thus be relatively easy to ignore. However, as Beck (ibid) also points out, each person has in addition to deal with a set of potential hazards which are more immediately salient to them and which may be more amenable to personal influence:
"everyone is engaged in the struggle for [their own] job, income, family, little house, automobile, hobbies, vacation wishes, etc. If those are lost , then you are in a tight spot in any case – pollution or no." (Beck 1992)
Despite being addressed from many different perspectives, the twin problems of how risk messages are produced and interpreted in real life, and how to predict behavioural outcomes when people have to make decisions about simultaneously presented choices involving risk have been under investigation for more than a century and remain unresolved (Löfstedt and Frewer 1998 p5, af Wåhlberg 2001, Löfstedt and Boholm 2009 p14),.
The initial assumption, that sensible, 'rational' people choose the outcome most likely to benefit them (the Subjective Expected Utility model, Savage,1954), still forms the basis of quantitative risk assessment (QRA) and actuarial predictions. However, Simon (1955,1956), showed that people may abandon the search for the least risky outcome too soon ('satisficing'), and research in cognitive psychology under laboratory conditions has demonstrated that people frequently over- or underestimate degrees of risk by using faulty decision strategies ('heuristics' - Kahneman and Tversky, 1974). Concepts such as 'frames' (Minsky 1977; Tversky and Kahneman,1981, 1986), 'cognitive maps' (Tiemann and Tiemann, 1985) and 'mental models' (Johnson-Laird, 1983; Bostrom et al 1992, 1994: Morgan et al 2001), have been developed to explain the observation that decisions about behaviour are affected by influences other than the actual probabilities of the outcome.
Research following the psychometric paradigm developed by Fischhoff, Lichtenstein and Slovic (for an overview see Slovic, 2000), has taken a different approach, addressing the ways in which people think about 'real world' hazards, how they perceive the risks they pose and how they believe they would feel and act in response. Their data show that people answer such questions differently according to their familiarity with and degree of control over a particular hazard; and to how severe they perceive the consequences to be, should the adversity occur (Fischhoff et al 1978). These findings indicate that people's reactions to the contemplation of hazards are driven not solely by how they think, but how they feel about them, and suggest[2] that emotion, or affect, may play a role in behavioural response to risk perception (Finucaine et al, 2000: Lowenstein et al, 2001; Slovic et al, 2004, Wardman 2006, McComas 2006).
Thus the research to date amply demonstrates that people both perceive multiple hazards and (at least in theory) may respond to them in different ways. However, perhaps because the main streams of research which contribute to these insights have had divergent interests and take different approaches to examining risk (Taylor-Gooby and Zinn, 2006), there appears to have been less focus on the processes by which people simultaneously perceive and prioritise the multiple hazards they experience in their daily lives. Crucially, the circumstances under which one perceived hazard may take precedence over another, qualitatively different, threat when it is necessary to make choices about how to respond is rarely addressed. Researchers in the cognitive tradition tend to concentrate on how people choose between the same or similar hazards. Fischhoff et al (1979) do indeed note the problem:
"if hazards are dealt with one at a time, many must be neglected ...[and]… the information needed to establish priorities is not available" (p 131)
However they do not suggest a method of investigating the ways in which people select an action, when forced to do so in their everyday lives. Bostrom and Löfstedt (2003) also point this out, arguing that there is not yet sufficient understanding of the role that salience plays in moving people along the awareness-action continuum.
The survey methods used in psychometric approaches do frequently address people's perceptions of multiple and qualitatively different hazards[3], but importantly, the results only record how people say they perceive and would respond to hazards, they are not backed up by evidence of action. Cultural theories (Douglas 1966, 1982; Dake 1991, 1992) explain behaviour in response to perceived risk as being mediated by the prevalent attitudes in the social groups to which they belong, but they do not address how the varied responses to perceived risk from individuals within those groups can be explained. Analyses of disasters and accidents (Turner 1976, 1978; Reason et al 2001, Perrow and Guillen 1990, Quarantelli 1995) have usually focussed at the organisational level on multiple perceptions of risk in relation to single, not several different, hazards.
The closest that mainstream literature on risk perception comes to addressing this issue, that people's perceptions of the risks in their environment vary in their salience and hence in the extent to which they generate action, is the social amplification/attenuation of risk framework (SARF) (Kasperson et al, 1988e). The key insight of the SARF approach is the notion of amplification and attenuation of risk perception, that the salience of a hazard may wax and wane according to external circumstances. This fits very well with ideas about perceptions of risk and their effect on decision making, explored by Bennett and Ferlie (1994), and Bennett (1996), on the basis of in-depth interviews collected during empirical research on the early years of HIV/AIDS crisis. These data suggested that at any particular moment, individuals were able to identify a number of actual or potential hazards which they perceived to have salience for them and/or others. However, the perceived magnitude of the risks posed by these hazards did not remain static, even for specific individuals. A hazard which, under one set of circumstances, had come to be perceived as overwhelmingly threatening (amplification), could, under a different set of circumstances, appear less urgent than other previously ignored or unrecognised risks, forcing revision of the first response (attenuation). Examples of such processes were demonstrated in the data collected, but as Bennett (1996) points out the research was not conducted to specifically investigate risk perception and all that could be concluded was that insufficient is known about the circumstances under which one perceived risk factor may take precedence over another.