Objectives To develop an understanding of the factors that influence patient safety-related behaviours by nurses, doctors and allied health staff employed by Queensland Health, using a theory-driven behavioural model.
Design Cross-sectional survey analysed with multiple logistic regression.
Setting Metropolitan, regional and rural public hospitals in Queensland, Australia.
Participants 5294 clinical and managerial staff.
Main outcome measures The Theory of Planned Behaviour was used to develop behavioural models for high-level Patient Safety Behavioural Intent (PSBI) of senior and junior doctors, senior and junior nurses, and allied health professionals. Multiple logistic regression analysis was used to identify factors that significantly influenced PSBI between the five professional groups.
Results The factors that influence high-level PSBI give rise to unique predictive models for each professional group. Two factors stand out as influencing high-level PSBI for all healthcare workers (HCWs): (1) Preventive Action Beliefs (adjusted OR 2.38), HCWs' belief that engaging in the target behaviours will lead to improved patient safety; and (2) Professional Peer Behaviour (adjusted OR 1.79), perceptions about the patient safety-related behaviours of one's professional colleagues.
Conclusions Professional peer-modelling behaviours and individuals' beliefs about the value of those behaviours in improving patient safety are important predictors of HCWs' patient safety behaviour. These findings may help explain the limitations of current knowledge-based educational approaches to patient safety reform. Use of the behavioural models developed in this study when designing future patient safety improvement initiatives may prove more effective in driving the behavioural change necessary for improved patient safety.
- Patient safety
- safety culture
- safety climate
- theory of planned behaviour
Statistics from Altmetric.com
Patient safety problem
Since the publication of the landmark report To Err Is Human: Building a Safer Health System by the Institute of Medicine,1 Patient Safety (preventing patient harm) has become a major healthcare reform issue, both in Australia and in other industrialised countries.
Australian data suggest that approximately one in six (16.6%) patients in Australian public hospitals, suffer adverse or harmful events as a result of their hospitalisation,2 with up to one-half of these adverse events considered preventable. Adverse events are estimated to cost the Australian healthcare system $2 billion per annum3 in direct costs alone.
What is safety culture?
Understanding the shared attitudes, beliefs and values of healthcare workers (HCWs), and the degree to which these influence safety behaviours, is critical to sustainable improvement in patient safety.4 ,5 Some authors6–8 emphasise the difference between safety climate (individuals' attitudes towards safety) and safety culture (the shared beliefs and convictions underpinning individuals attitudes; the prevailing values of the social group). Despite this, the overall trend appears to be in favour of the term safety culture.9 ,10 There are many definitions of safety culture,10–12 but the one which is most commonly used is the phrase: ‘the way we do things around here.’
Why is safety culture important in safety performance?
High-hazard industries such as aviation and petrochemicals have long believed safety culture to be an important predictor of safety performance. This has resulted in research focussed on understanding and influencing safety culture in these so-called high-reliability organisations.13–16 In contrast, the healthcare industry has only recently begun to describe the extent to which organisational culture impacts on patient safety.
Limitations of current safety culture measurement instruments
Two recent reviews noted significant weaknesses in psychometric properties and variation in common dimensions of culture, in commonly used patient safety culture survey instruments.26 ,27 The authors noted that instruments rarely have an explicit theoretical basis. While efforts have been made to establish content validity (ie, dimensions which are representative of safety culture), exploratory factor analysis was evident in only one of 12 instruments reviewed. Criterion validity (ie, the degree of correlation of the climate scores with outcome data) in terms of patient injury has not been established for any currently available instrument. Despite known cultural differences between professional groups,28 some instruments measure and report culture only in terms of generic HCWs.18 Most of the available survey instruments were developed and validated in the USA. Given the importance of the broader influence of national and organisational cultures,8 ,29 it cannot be assumed that the validity of such instruments is generalisable to other countries. None of the current instruments allow for behavioural modelling to establish the factors that predict and influence patient safety behaviours.
Behavioural theory applied to patient safety behaviours
Many theories have been developed to explain human behaviour. Behavioural modelling has been used successfully to assist in identifying factors that predict the adoption of favoured healthcare behaviours by populations.30 The Theory of Reasoned Action31 and the modified Theory of Planned Behaviour (TPB)32 have been used extensively in relation to public health interventions such as safe sex,33 hand hygiene34 and immunisation.35 The TPB was chosen for this study because of its prior use in explaining the behaviour of healthcare workers; its use in explaining covert behaviours (behaviours that are not readily observable such as voting and seatbelt wearing); and in addition to attitude, the influence of peer pressure and behavioural control upon behaviour. As patient safety behaviours (eg, reporting incidents and speaking up when a colleague makes an error) cannot easily be measured, behavioural intent is used as the proxy in the model.
In the TPB (figure 1), behavioural intention is influenced by the following:
Attitudes towards patient safety behaviour which is determined by behavioural beliefs (whether the individual believes that the behaviour will improve patient safety) and evaluation of behavioural outcomes (whether the individual has experienced the improved patient safety resulting from the behaviour).
Subjective norms which are determined by normative beliefs (a person's belief about how the people around them think they should behave in order to keep patients safe) and motivation to comply (the degree to which the individual is motivated to comply with the wishes of the people around them).
Perceived behavioural control which is determined by control beliefs (the degree to which an individual believes that their own contribution can lead to the improvement of patient safety) and their perceived power (an individual's perception of whether they have the power to engage in the patient safety behaviour).
The application of behavioural modelling to patient safety has not previously been undertaken.
Since its inception in 2005, The Queensland Health Patient Safety Centre has taken a lead role in planning, implementing, managing and evaluating system-wide patient safety initiatives in a large decentralised public healthcare system.36 Establishing a baseline patient safety culture in the organisation was considered essential to subsequent evaluation of a range of planned improvement strategies. Our study had three main objectives:
To develop and validate in an Australian setting an instrument to effectively measure patient safety culture and model factors that influence patient safety behaviours.
To develop a better understanding of the factors that influence patient safety-related behaviours by nurses, doctors and allied healthcare workers employed by Queensland Health, using a theory-driven behavioural model.
To measure the baseline patient safety culture in Queensland Health prior to the introduction of a systematic approach to improving patient safety.
This paper will report on objectives (1) and (2).
Items to be included in the questionnaire were identified from literature review, an existing survey tool37 and focus discussion groups, to ensure cultural generalisability of the literature in the Australian healthcare setting. Focus groups were held in metropolitan, regional and rural public hospitals, involving doctors, nurses, allied health staff and managers. Focus-group discussions were prompted using items from the literature about patient safety and safety culture. Key themes identified from the discussion groups are listed in table 1.
Items included in the questionnaire were grouped and worded according to the TPB constructs (see figure 1) and included additional constructs identified during the focus discussion groups. The questionnaire was pilot tested for clarity on 145 HCWs who were medical (13%), nursing (66%) and allied health (19%) staff from rural (19%), regional (30%) and metropolitan sites (48%). Analysis of the pilot data led to minor changes in the final survey instrument. Changes included modified wording of several questions to improve clarity as well as the removal of questions to maximise the internal consistency of the instrument.
Patient Safety Behavioural Intent (PSBI) was the target construct (dependent variable) and was the sum of eight questions relating to incident reporting behaviour, speaking out, or intervening when an error was witnessed (table 2).
Twelve other constructs (independent variables) that might predict PSBI were identified from the safety literature and focus groups, and grouped according to the TPB. Each construct was made up of questions describing twelve themes (table 3). Each construct question was worded as a statement and responses given on a seven-point Likert Scale. The final questionnaire (see online only appendix 1) comprised 136 questions, which measured responses to one dependent variable, nine demographic questions and 12 behavioural constructs (see table 3). Responses were scored seven for extremely agree and one for extremely disagree. Scores were summed and the constructs tested for internal consistency using Cronbach alpha correlation. Correlations of the constructs were high, ranging from r=0.71 to r=0.94.
The survey questionnaire and covering letter (cosigned by relevant district managers) was administered to 22 557 staff, in one acute public hospital facility in each of the 37 health service districts in Queensland. Distribution took place in January and February 2006. The survey was distributed by hand in the workplace, using existing networks of patient safety officers and relevant line managers. Respondents returned their completed surveys using a reply-paid envelope, independent of Queensland Health.
Data were entered onto Excel (Microsoft, Seattle, Washington), and descriptive statistics were performed using EpiInfo V.6.0 (CDC, Atlanta, GA, USA) and multiple logistic regression modelling was performed in SPSS V.15.0 (SPSS, Chicago, Illinois). A total of 5294 completed surveys (23.5%) were received, representing approximately 10% of the total staff of Queensland Health. Respondents comprised 45.2% (n=2391) junior nurses, 23.7% (n=1259) senior nurses, 15.2% (n=807) allied health, 6.5% (n=342) senior doctors, 4.5% (n=237) junior doctors, 1.9% (n=98) executive managers and 3.0% (n=160) other. Senior doctors included consultants and career medical officers; junior doctors included training grades at or below senior registrar and prevocational junior doctors; senior nurses were defined by their role as nurse manager, clinical nurse consultant and clinical nurse; junior nurses included registered nurse, enrolled nurse and assistants in nursing. Allied health staff included physiotherapists, occupational and speech therapists, and pharmacists.
Responses to the 12 independent constructs were summed in accordance with the TPB and then divided by the total items to regain a score within the original scale 1–7.
The continuous outcome variable, PSBI, was reclassified into high (responses of ≥6) and low (<6). The proportions of clinicians with high-level of PSBI were examined, and ORs for each clinical group were compared against the clinical group with the lowest score.
The TPB model tested 12 constructs to identify significant predictors of high-level PSBI. A single model for all clinicians, n=5036, and individual models for each of the five clinical subgroups were developed. The 12 independent variables were entered into a backwards (nonconditional) stepwise multiple logistic regression model to identify significant predictors of high-level PSBI. Alpha was set at the 5% level. To compare the crude ORs of each significant predictor, by the amount of improvement in high-level PSBI that could be achieved for every one-unit improvement along the Likert-type Scale, the β coefficient was adjusted for the interquartile range of each predictor. Adjusted ORs (AORs) in the final model were calculated by multiplying the β coefficients by the interquartile range of the scale. The AORs were equivalent to crude ORs where the interquartile range was one.
Between 39% and 59% of the clinical subgroups were employed in metropolitan hospitals, 29–41% worked in the provincial district hospitals, and 10–20% worked in a rural location. The primary professional qualification for 91% of allied health staff, 89% of junior nurses, 88.7% of senior nurses, 65.5% of junior doctors and 68.4% of senior doctors was obtained in Australia.
Training workshops, known as Human Error and Patient Safety, had been attended by 56.7% of executive respondents, 22.9% of senior doctors, 20.1% of senior nurses, 14.5% of allied health staff, 8.8% of junior nurses and 4.2% of junior doctors.
Behavioural models as predictors of high-level PSBI for all healthcare workers
The aggregated HCW model (figure 2) identified Preventive Action Beliefs (AOR 1.82, 95% CI 1.66 to 1.99, p<0.0001) and Professional Peer Behaviour (AOR 1.68, 95% CI 1.57 to 1.80, p<0.0001) as the two strongest predictors of high-level PSBI. Management Responsiveness and Work Satisfaction were less important predictors of PSBI, while a belief in fatigue and poor communications (Personal Causes of Errors) and a belief that the hospital did not support patient safety (Hospital Support) were significantly less likely to hold high PSBI. The seven predictive constructs of the aggregated PSBI model explained only 21% of the variation in behaviour, and so separate models were developed.
Level of PSBI
The proportion of each clinical subgroup having high-level PSBI was compared with that of junior doctors, and up to a sixfold difference between the clinical groups was identified (table 4). Senior doctors were 1.5 times (38.1%, 95% CI 1.01 to 2.1) more likely than junior doctors to have a high PSBI, allied health staff were 2.7 times more likely (53.2%, 95% CI 1.9 to 3.7), junior nurses were 3.9 times more likely (61.9%, 95% CI 2.8 to 5.3) and senior nurses were 6.0 times more likely (73.6%, 95% CI 4.8 to 9.2).
Medical respondents: The strongest predictor of high-level PSBI for both junior and senior doctors was Professional Peer Behaviour (AOR 2.47, p<0.0001 and AOR 3.21, p=0.003, respectively). The models differed thereafter with junior doctors influenced by beliefs in System Causes of Errors (AOR 1.61, p=0.026) and their Work Satisfaction (AOR 1.25, p=0.029). Strong predictors for senior doctors included Attitudes (AOR 2.38, p<0.0001) towards high PSBI and Preventive Action Beliefs (AOR 2.33, p=0.001), while those who believed in the Personal Causes of Errors were less likely to hold high PSBI (AOR 0.56, p=0.007).
Nurse respondents: Predictors for junior and senior nurses included: Preventive Action Beliefs (AOR 1.71, p<0.0001 and AOR 1.96, p<0.0001 respectively) and Professional Peer Behaviour (AOR 1.64, p<0.0001 and AOR 1.58, p<0.0001). Junior nurses were significantly more likely (AOR 1.22, p<0.0001) to have high PSBI if they held high Work Satisfaction, while those who held a belief in Personal Causes of Errors were less likely to hold high PSBI (AOR 0.80, p=0.001). Senior nurses were less likely (AOR 0.89, p=0.032) to hold high PSBI if they held belief in poor Hospital Support.
Allied health respondents: The allied health respondents' model was closer to that of the junior nurses with allied health more likely to hold high PSBI if they held strong Professional Peer Behaviour (AOR 2.01, p<0.0001) and Preventive Action Beliefs (AOR 1.66, p<0.0001), while they were less likely to hold high PSBI if they believed in Personal Cause of Errors (AOR 0.69, p<0.0001).
This is the first study to develop predictive models for patient safety behaviours of HCWs. The findings of this study go beyond merely describing aspects of patient safety culture in a healthcare organisation, to an improved understanding of the factors that influence HCWs' intention to engage in behaviours, traditionally known to be associated with organisations with high reliability and safety.38 ,39
Relationship with previous research
Patient safety culture has not been comprehensively studied in Australia. International studies (mainly from the USA) have modelled respondents as a single group of healthcare workers using factor analysis.40 ,41 Factors associated with patient safety included teamwork climate, safety climate, perceptions of management, job satisfaction, working conditions and stress recognition. Common themes emerging from the literature that impact negatively upon patient safety culture include: fear of reprisal after reporting adverse events; the impact of production pressure; leadership; and senior management disconnect.23 ,42–44 No previous research could be identified that applied behavioural theory to the measurement of patient safety culture.
Factors influencing intention to engage in safety behaviours
Table 5 summarises the behavioural models for the individual clinical subgroups. Each professional group has a unique behavioural model of factors that influence their intention to engage in patient safety behaviours. Further work is needed to determine the robustness of these models in the wider healthcare community. However, we believe it reflects existing evidence of powerful professional subcultures within healthcare.28 Moreover, the application of these unique models in design and implementation of patient safety improvement strategies may be more likely to lead to the behavioural changes which underpin sustainable safety improvement.
Differences in respondent level of intention to engage in safety behaviours
It is perhaps not surprising that the study findings suggest that nurses have significantly higher levels of intended engagement in patient safety behaviours than doctors. This is consistent with the experience of the authors and is most evident in clinical incident reporting, which is predominantly nursing-driven.45 However, the study findings of a sixfold difference between reported intention to engage in safety behaviours between senior nurses and junior medical officers is of great concern, cannot be explained through differences in reporting behaviours alone and warrants further examination.
Several studies confirm our findings that doctors, as a group, and especially junior doctors, appear to be disconnected from the system in which they work.46 ,47 When compared with the other respondents, they are more likely to perceive blame as a result of an incident, and less likely to report positively about management, or speak up when an error is made. Existing medical culture is characterised by individual concepts of clinical work, clinical purism and opaque accountability.28 Junior doctors are quickly socialised to these values and behaviours,48 and this remains a significant challenge for patient safety improvement, which by its nature is concerned with systematised concepts of clinical work and transparent accountability. Without appropriate modelling of patient safety behaviours by senior medical staff in the workplace, it is unlikely that junior doctors will value or adopt such behaviours as part of their practice, even if they graduate with these concepts firmly in mind.49
Limitations of the study
Survey data are potentially biased by non-responders. The most important limitation of our study was the relatively low response rate. Despite this obvious weakness, analysis of the demographic data collected suggests our sample was representative for professional background, location, age, seniority and country of primary qualification. Our sample represents almost 10% of the estimated total health service workforce of 60 000 staff.
The degree to which the factors identified contribute to the target behaviour (PSBI) varies from an R2 value of 0.21 (factor predicts 21% of the behaviour) in the all respondent group to 0.44 (44%) for the senior medical respondent model. Despite extensive literature review, focus discussion groups and pilot testing, other factors that predict safety behaviours remain unexplained.
Survey responses are limited to measurement of the surface features of safety culture (ie, safety climate) at a point in time. Caution should be exercised in reliance on survey methods alone to measure safety culture. Where possible, findings from surveys should be confirmed with well-designed ethnographic studies. However, surveys remain an important method in measuring culture when behaviours are covert and difficult to measure by direct observation.
Lessons for future patient safety reform
Despite the fact that there has been major international investment in patient safety reform since the landmark Institute of Medicine report To Err Is Human, in 2000, little change is evident.50 This study clearly demonstrates that two key factors influence the safety behaviours of all HCWs: observed behaviour of professional peers (Professional Peer Behaviour) and a genuine belief in the safety outcomes of the behaviours (Preventive Action Beliefs). Despite this, much of the focus of current national and international safety reform strategy appears to be based on a flawed assumption that change will occur as a result of educating individual HCWs to improve knowledge of safety.
It follows from these findings that the key to effective and broad-based behaviour change strategies to improve patient safety is the influence of credible, clinical leaders that both believe in, and are prepared to model, patient safety behaviours in the workplace. This will not happen by accident. It will require a preparedness by healthcare organisations to invest in developing, supporting and rewarding clinicians as leaders and managers of safety reform. A desirable first step would be to engage and develop key clinician leaders at work unit and facility levels, with specific roles in modelling and teaching patient safety.
Many of the current patient safety initiatives being implemented at state and national levels are met with scepticism and resistance from healthcare workers, particularly doctors. Conventional change management methodologies do not work well in complex healthcare organisations,51 where there is significant variation in practice, individual practitioner autonomy and little regulation of safety. While behavioural theory has been successfully applied in many key public health interventions, it remains greatly underutilised in patient safety. Understanding and applying such methodologies to patient safety improvement may be the key to more effective interventions. We hope that our study will prompt those involved in patient safety reform, to explore and consider the application of behavioural approaches, as well as stimulating further research aimed at gaining a better understanding of the factors that positively influence patient safety behaviours in HCWs.
What is already known on this topic
There are known cultural differences between healthcare organisations and other high-hazard industries, with excellent safety performance.
It is only recently that research has been undertaken to measure patient safety culture in healthcare. Work to date has been limited to describing the features of patient safety culture.
What this study adds
Professional peer-modelling behaviours and individual beliefs about the value of those behaviours in improving patient safety are the most important predictors of HCWs' patient safety behaviour.
Use of the behavioural models from this study in designing patient safety improvement initiatives may lead to more effective and sustainable behavioural change and safety improvement outcomes.
Key findings from the Queensland patient safety culture survey
Two factors stand out as influencing high-level Patient Safety Behavioural Intent for all healthcare workers. These are: (a) Preventive Action Beliefs (an individual's belief about whether engaging in the target behaviours will lead to improved patient safety); and (b) Professional Peer Behaviour (perceptions about one's own professional colleagues' patient safety behaviour).
The factors that influence high-level Patient Safety Behavioural Intent vary for each professional subgroup and give rise to unique predictive models.
There is a sixfold difference in the level of Patient Safety Behavioural Intent from junior doctors (lowest) to senior nurses (highest).
We are grateful to all the staff across Queensland who took the time to participate in this survey. Without their generous support, this study would not have been possible. Thanks also goes to the staff at the Patient Safety Centre, Patient Safety Officers and other district staff who were involved in the distribution of the survey.
Funding Queensland Health funded the study under the auspices of the Queensland Health Patient Safety Centre, Queensland Centre for Healthcare Improvement, 147–163 Charlotte Street, Brisbane, Queensland, Australia.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.