Can a continuous quality improvement program create culturally safe emergency departments for Aboriginal people in Australia? A multiple baseline study

Background Providing culturally safe health care can contribute to improved health among Aboriginal people. However, little is known about how to make hospitals culturally safe for Aboriginal people. This study assessed the impact of an emergency department (ED)-based continuous quality improvement program on: the accuracy of recording of Aboriginal status in ED information systems; incomplete ED visits among Aboriginal patients; and the cultural appropriateness of ED systems and environments. Methods Between 2012 and 2014, the Aboriginal Identification in Hospitals Quality Improvement Program (AIHQIP) was implemented in eight EDs in NSW, Australia. A multiple baseline design and analysis of linked administrative data were used to assess program impact on the proportion of Aboriginal patients correctly identified as Aboriginal in ED information systems and incomplete ED visits in Aboriginal patients. Key informant interviews and document review were used to explore organisational changes. Results In all EDs combined, the AIHQIP was not associated with a reduction in incomplete ED visits in Aboriginal people, nor did it influence the proportion of ED visits made by Aboriginal people that had an accurate recording of Aboriginal status. However, in two EDs it was associated with an increase in the trend of accurate recording of Aboriginality from baseline to the intervention period (odds ratio (OR) 1.31, p < 0.001 in ED 4 and OR 1.15, p = 0.020 in ED 5). In other words, the accuracy of recording of Aboriginality increased from 61.4 to 70% in ED 4 and from 72.6 to 73.9% in ED 5. If the program were not implemented, only a marginal increase would have occurred in ED 4 (from 61.4 to 64%) and, in ED 5, the accuracy of reporting would have decreased (from 72.6 to 71.1%). Organisational changes were achieved across EDs, including modifications to waiting areas and improved processes for identifying Aboriginal patients and managing incomplete visits. Conclusions The AIHQIP did not have an overall effect on the accuracy of recording of Aboriginal status or on levels of incomplete ED visits in Aboriginal patients. However, important organisational changes were achieved. Further research investigating the effectiveness of interventions to improve Aboriginal cultural safety is warranted. Electronic supplementary material The online version of this article (10.1186/s12913-019-4049-6) contains supplementary material, which is available to authorized users.


Background
In Australia, Aboriginal people 1 have poorer health than non-Aboriginal people [1]. Although improvements have been made in recent decades, Aboriginal people continue to experience higher rates of infant mortality, injury, non-communicable diseases, preventable illnesses and nutritional disorders than other Australians [2,3]. Several social factors contribute to the gap in health outcomes between Aboriginal and non-Aboriginal people, such as higher levels of unemployment, social exclusion and experiences of racism and suboptimal use of health care among Aboriginal people [3][4][5]. Aboriginal people are more likely than non-Aboriginal people to delay seeking care until their illness is advanced and/or comorbid [6].
Aboriginal peoples' engagement with health care is influenced by the physical accessibility, affordability and cultural safety of health service provision [7]. Culturally safe care is characterised by a genuine partnership between patient and health care provider in which: power is shared; the life experiences, views and beliefsespecially cultural beliefsof the patient are respected; and Aboriginal histories and associated social impacts are acknowledged [8][9][10][11]. Some Aboriginal people feel culturally unsafe when using mainstream health services in Australia and are therefore reluctant to seek care in these settings [12]. Some factors that undermine the cultural safety of Aboriginal patients include: culturally insensitive health staff; institutionalised racism; a shortage of Aboriginal health workers; distrust of the health system among some Aboriginal patients; and unwelcoming waiting areas in key settings like hospital emergency departments (ED) [13,14].
Aboriginal people are overrepresented among ED patients, relative to population size [15]. Additionally, EDs are the first point of contact with the health system for many Aboriginal people. EDs are therefore uniquely positioned to improve the health of this group [16]. However, EDs can feel unwelcoming and culturally unsafe for some Aboriginal people [13,14]. Common aspects of the ED visit experience, like variable and sometimes long wait times and a lack of information about ED processes, contribute to a perceived sense of powerlessness or lack of control among some Aboriginal people [13]. Further, rates of incomplete ED visits (where the patient either left the ED before receiving a medical assessment or left the ED after a medical assessment but before completion of care) are elevated among Aboriginal people compared to non-Aboriginal people (9.7% vs. 7.1%), providing indirect evidence of higher levels of dissatisfaction with ED care in Aboriginal people than in other Australians [17][18][19].
Improving the cultural safety of health services can contribute to improved health among Aboriginal people and is a health priority in Australia [20,21]. However, little is known about how to effectively intervene, including in ED settings. Recent reviews have found a predominance of descriptive studies, a lack of high quality intervention studies and a need for more rigorous innovation testing and translational research [10,22,23]. Continuous quality improvement (CQI) seeks to improve service quality through on-going cycles of reflection and refinement. Existing organisational processes are examined and modifications are developed, tested and, if effective, adopted [24]. CQI approaches have shown initial promise in improving the cultural safety of Aboriginal people in some primary care settingsespecially in Aboriginal community-controlled health servicesand warrant further investigation in secondary and tertiary care settings [10,24].
We evaluated the Aboriginal Identification in Hospitals Quality Improvement Program (AIHQIP), a CQI initiative aiming to improve the cultural safety of Aboriginal patients in eight EDs in New South Wales (NSW), Australia. The study aimed to investigate whether the program: increased the proportion of Aboriginal patients correctly identified as Aboriginal in ED information systems; reduced the proportion of Aboriginal patients who had an incomplete ED visit; and improved the cultural appropriateness of ED systems and environments.

Intervention
Each participating ED implemented a CQI project with a focus on working with Aboriginal people to improve the cultural safety of ED services for Aboriginal patients. Each ED employed a project officer and established a working group to lead and guide project implementation. Working group membership typically included key hospital staff, staff of Aboriginal community-controlled organisations and local Aboriginal community members.
The research team supported EDs to implement their CQI projects in the following ways: 1. Provision of a nine-step "Plan, Do, Study, Act" CQI framework with a unique emphasis on genuine partnership with Aboriginal communities. The framework included an implementation toolkit with supportive resources, such as an action plan template and examples of activities to aid each framework step. The nine steps of the framework are shown in Fig. 1 Each ED selected objectives for their CQI project from a predetermined list of objectives aligning with the overarching aims of the AIHQIP. Common project objectives included to: encourage Aboriginal patients to identify as Aboriginal in the ED; improve the Aboriginal cultural competence of ED staff and other hospital staff; improve collaboration between the ED and Aboriginal community-controlled organisations; and reduce incomplete ED visits among Aboriginal patients. Tailored strategies were developed, implemented and refined to meet these objectives.

Setting
The AIHQIP was implemented in eight public EDs in NSW, one of eight States and Territories in Australia. About 7.5 million people live in NSW, of whom about 3% (n~216,000 persons) are Aboriginal people [26]. There are about 2.7 million visits to NSW EDs annually, with about 6% (n~462,000) of visits made by Aboriginal people [15].
In NSW, public EDs are operated by autonomous, geography-based health corporations called Local Health Districts, while the NSW Ministry of Health monitors and manages ED performancefor example, levels of re-presentation to the same ED within 48 h among Aboriginal patients. Several policies and procedures are in place to improve Aboriginal peoples' experiences of emergency care in NSW, including mandatory Aboriginal cultural competency training for all hospital staff, mandatory recording of the Aboriginal status of hospital patients in information systems and a strategy for increasing and building the capacity of the Aboriginal health workforce in public hospitals, including EDs. Additionally, some EDs employ Aboriginal Liaison Officers (ALO) to provide emotional, social and cultural support to Aboriginal patients and their families when in hospital.

Study sites
Site selection was non-random. EDs were chosen to include a mix of rural and metropolitan EDs across multiple Local Health Districts. Additionally, executive support for implementing the AIHQIP and employment of an ALO were essential criterion for study participation, as these characteristics were considered fundamental to implementing the program as intended. The characteristics of participating EDs are included in Table 1.
A multiple baseline design, qualitative interviews with program stakeholders and document review were used to evaluate the AIHQIP.

Multiple baseline design
A multiple baseline design [27] and secondary analysis of linked administrative health data were used to assess quantitative outcomes of the AIHQIP.

Data source
Data were obtained from the NSW Admitted Patient, Emergency Department Attendance and Deaths Register (APEDDR). The APEDDR is a statutory public health and disease register established under the NSW Public Health Act 2010 to support, among other things, evaluations of public health interventions. It includes linked records of the following datasets: These data sources were linked by the Centre for Health Record Linkage using probabilistic record linkage methods.

Study population
The study population included Aboriginal people who attended one of the eight participating EDs between 1st January 2010 and 31st March 2015.

Outcomes
The two outcomes investigated were the proportion of Aboriginal patients correctly identified as Aboriginal in ED information systems and rates of incomplete ED visits. Both are considered indirect indicators of the cultural safety of ED service provision [23].
In NSW, staff in public EDs are required to ask every ED patient if they are Aboriginal 2 although adherence to this requirement is variable within and across EDs. Health staff are more likely to ask this question of patients if they are culturally competent and work in an environment in which Aboriginal cultural competency standards are integrated into policies and procedures [28]. An Aboriginal person may choose not to identify as Aboriginal if they feel culturally unsafe in the ED, especially if they think that identifying as Aboriginal will negatively influence the quality of care they receive [29,30]. The proportion of Aboriginal patients correctly identified as Aboriginal in ED information systems was assessed using the Enhanced Reporting of Aboriginality (ERA) variable in the APEDDR. The different datasets in the APEDDR and records of the NSW Perinatal Data Collection 3 (all of which hold information on the Aboriginal status of the patient/person) were linked and an algorithm applied to identify ED records for which the patient could reasonably be considered an Aboriginal person for analysis purposes. The algorithm was as follows: if the person had three or more independent units of information (for example, a birth record, a hospital separation record and an ED separation record) and at least two of these indicated that the person was Aboriginal, then the person was considered Aboriginal; or if the person had one or two independent units of information and at least one of these indicated that the person was Aboriginal, then the person was considered Aboriginal. More on the ERA variable can be found elsewhere [31]. The accuracy of recording of Aboriginality was calculated as the observed number of ED visits for which the patient was recorded as Aboriginal in the ED information system divided by the "expected" number of ED visits by Aboriginal people, as determined by the ERA calculation.
Incomplete ED visits included visits for which the patient either left the ED before receiving a medical assessment or left the ED after a medical assessment but before completion of care or ED discharge. Feeling culturally unsafe is a main reason why Aboriginal people leave the ED early [14,32]. Incomplete ED visits provide indirect evidence of patient dissatisfaction with the ED experience [19]. Rates of incomplete ED visits among Aboriginal people were calculated by dividing the number of incomplete visits by the total number of ED visits, using the Indigenous status variable in the NSW Emergency Department Data Collection.

Procedure
The eight EDs were allocated to three clusters. The clustering of EDs was non-random and informed by practical considerations, such as readiness to start implementing the AIHQIP. Clusters were then randomly assigned an implementation order of first, second or third. It was intended that program implementation in each cluster would be separated by three months. However, some EDs experienced delays in implementing the AIHQIP such that the period separating program commencement in EDs ranged from one month to four months. The first cluster of EDs started implementation in August 2012, with the implementation period at each site ranging in duration from 10 months to 15 months (Table 1).
Monthly data on the two outcomes were obtained for each ED and partitioned temporally into: before the intervention was implemented (control period: January 1st 2010 until the first site visit by the research team d4 ) and during and immediately after program implementation (intervention period: from the date of the first site visit until March 31st 2015). Each ED had at least 17 control period time points and 19 intervention period time points.
The multiple baseline design acknowledges that EDs differ by catchment population, staffing arrangements, administrative procedures and other characteristics. This variation, combined with the random staggering of intervention commencement across participating EDs (described above), increases the likelihood that any detected improvement in outcomes is due to the intervention, rather than external factors [27].

Statistical analysis
In determining program effectiveness, the main consideration was whether the trend in each outcome changed significantly, and in the desired direction, following program implementation. Change in intercept (that is, immediate change in the outcome at the start of the intervention period) was considered less relevant given the organisational changes the AIHQIP aimed to achieve can take a long time to establish. Data were aggregated by month of ED presentation, hospital and patient sex, age group and Aboriginal status. Baseline patient characteristics were calculated for each site. Two modelling phases were undertaken for each outcome. First, preliminary logistic regression models were prepared for each study ED to investigate facility-level intervention effects. Following this, generalised linear mixed models were prepared to investigate the average intervention effect among all study EDs. Preliminary analyses identified a need to adjust for varying intercepts, trends and intervention effects between study sites. In all models, the key parameter of interest was the change in the linear trend from the control period to the intervention period. Parameter estimates were calculated, along with odds ratios, 95% confidence intervals and corresponding p-values. Data management and aggregations were performed in SAS version 9.3. Modelling was conducted in R version 3.2.1. An additional file provides a complete description of the statistical analysis [see Additional file 1].

Qualitative interviews with program stakeholders and document review
Following implementation of the AIHQIP, qualitative interviews of 15-60 min duration were conducted with participating ED staff and other hospital staff (n = 23) exploring factors influencing project implementation and perceived achievements. Interviewees were AIHQIP project officers (n = 8), hospital managerial staff (n = 8), a hospital executive, an LHD Deputy Director of Aboriginal Health, and ALOs (n = 5). Six of the 23 interviewees were Aboriginal people.
Interviews were conducted in a quiet room in participating EDs (three were done by phone) by a non-Aboriginal researcher with experience conducting qualitative interviews with Aboriginal people. Detailed notes were taken during the interviews and circulated to interviewees for comment and verification. The notes were analysed using qualitative thematic analysis and a data mining approach. Responses were grouped under the questions that were asked during interviews. Following this, responses were repeatedly read to identify common themes and key achievements, both within individual EDs and across all sites.
Documents relating to the AIHQIP (n = 50) were also reviewed to describe changes to participating ED systems and environments relating to the cultural safety of Aboriginal patients. The main documents reviewed were site visit reports. Following each site visit, the research team and local working groups compiled a report using a standardised template. Reports described: the hospital context; the CQI support provided; implementation progress and challenges; and organisational changes perceived to be related to the AIHQIP. Documents were analysed using qualitative thematic analysis and a data mining approach (as described above), and findings triangulated with key informant interview findings. Several other methods were used as part of a comprehensive process evaluation of the AIHQIP. However, these are not the focus of this paper.

Results
During the control period, the average number of ED visits per month in the eight study EDs combined was 27,096 ( Table 2). The average age of the presenting population was 28.3 years. However, average age varied among participating EDs (ranging from 25.2 years to  Table 2). During the control period, the proportion of ED visits made by Aboriginal people that had an accurate recording of Aboriginal status varied among participating EDs (ranging from 45.5 to 87.2%; avg. 76%). The proportion of ED visits made by Aboriginal people that were incomplete also differed between study sites (ranging from 6.5 to 24.9%; avg. 11.5%) ( Table 2).

Recording of Aboriginal status in ED information systems
During the control period, the proportion of ED visits made by Aboriginal people that had an accurate recording of Aboriginal status was significantly increasing over time in six of the eight participating EDs, while the accuracy of recording was decreasing significantly in one ED. A non-significant increase was observed in one ED ( Fig. 2 and Table 3).
In two EDs, the AIHQIP was associated with a statistically significant increase in the trend of accurate recording of Aboriginality from the control period to the intervention period (Table 3) (the key parameter to consider in the table is the change in slope).
In ED 4, during the control period the odds of an ED visit made by an Aboriginal person having an accurate recording of Aboriginal status were increasing by a factor of 1.12 per year. During the intervention period, the odds were increasing by a factor of 1.47 per year (p < 0.001). In other words, if the AIHQIP was not implemented the accuracy of recording would have increased from 61.4% in the control period to 64% 12 months after the program was first implemented. However, program implementation was associated with a  In ED 5, during the control period the odds of an ED visit made by an Aboriginal person having an accurate recording of Aboriginal status were decreasing by a factor of 0.93 per year. During the intervention period, the odds were increasing by a factor of 1.07 per year (p = 0.020) ( Table 3). In other words, if the AIHQIP was not implemented the accuracy of recording would have decreased from 72.6% in the control period to 71.1% 12 months after the program was first implemented. However, program implementation was associated with an increase in the accuracy of recording during this periodfrom 72.6 to 73.9%.
In all EDs combined, the AIHQIP was not associated with a change in the trend of accurate recording of Aboriginality from the control period to the intervention period (Table 4).

Incomplete ED visits
During the control period, the proportion of ED visits made by Aboriginal people that were incomplete remained steady over time in seven of the eight participating EDs and decreased significantly in one ED (Fig. 3 and Table 5).
The AIHQIP was not associated with a decrease in the trend of incomplete ED visits among Aboriginal people from the control period to the intervention period in any of the eight study EDs (Table 5), nor was an overall program effect detected (Table 6). Although decreases in the trend were observed in some sites, these were not statistically significant (Table 5). the strengthening of existing referral mechanisms between the ED and local Aboriginal communitycontrolled health services; the establishment of a critical incident response procedure for all incomplete ED visits among Aboriginal patients, which aimed to learn from the incident, prevent similar incidents in the future and ensure follow up and care of the affected patient; and the establishment of a performance indicator dashboard on the ED care of Aboriginal patients, which aimed to monitor and guide ED and hospital practice.

Organisational changes
Key informant interviewees identified several factors that they felt enabled program implementation, such as supportive hospital executive staff and a high level of engagement of local Aboriginal community-controlled health services and ALOs in project design and implementation. A more complete description of the factors influencing implementation is included in Additional file 2.

Discussion
To our knowledge, this is the first study to comprehensively investigate the effectiveness of a CQI program in making EDs more culturally safe for Aboriginal people in Australia. Our complementary use of a quasi-experimental design investigating primary outcomes and qualitative methods exploring organisational changes and program implementation is particularly novel in this area. In all study EDs combined, the AIHQIP was not associated with a reduction in incomplete ED visits in Aboriginal people, nor did it influence the proportion of ED visits made by Aboriginal people that had an accurate recording of Aboriginal status. However, in two EDs the AIHQIP was associated with an increase in the trend of accurate recording of Aboriginality from the control period to the intervention period.
There are four main factors that might explain this limited evidence of an intervention effect. First, the intervention may have lacked the intensity to influence the primary outcomes investigated. Second, in some participating EDs recording of Aboriginal patients' Aboriginality was fairly accurate and trending up at baseline. As such, improvements in this outcome may not have been detected, or indeed possible, in these EDs. The same issue applied to the incomplete ED visits outcome. Third, it is possible that the outcomes of interest were assessed prematurely, as the system changes that the AIHQIP sought to achieve can take considerable time to institutionalise [21]. Premature outcome assessment can occur in programs for Aboriginal people, especially if implementation is in complex systems where cause and effect and action and outcome relationships are poorly understood [33]. A longer intervention period may have increased the statistical power to detect intervention effects and provided more time for the program to produce change in the primary outcomes. This was not possible, however, due to constraints of time and resources. Similar multiple baseline studies might employ a longer intervention period. Lastly, we did not conduct power calculations prior to study implementation. Consequently, our study may have been under-powered to detect subtle changes associated with the intervention. That said, we analysed all ED records that were reasonably available to us during the five year study period. The linked administrative health data used in this study have several strengths. They provide populationlevel insights, they are easy and inexpensive to obtain relative to primary data collection methods, and they draw on multiple health datasets and the ERA algorithm to provide as accurate a description of the study population as possible. However, the two outcomes used -Aboriginal patients correctly identified as Aboriginal in patient information systems and incomplete ED visits in Aboriginal peoplemay not be the most sensitive measures of culturally safe care. Complementary use of other quantitative    Improve collaboration between the ED and Aboriginal community-controlled organisations 1-8 • A formal partnership agreement was established between the ED/hospital and the local Aboriginal community-controlled health service, which emphasised joint planning and service delivery and included provision for staff exchanges (n = 3). • Existing referral mechanisms between the ED and local Aboriginal communitycontrolled health services were refined (n = 1).

Reduce incomplete ED visits among
Aboriginal patients 1-8 • An ED critical incident response procedure was established for incomplete visits in Aboriginal patients, which sought to learn from the incident, prevent similar incidents in the future and ensure follow up and care of the affected patient (n = 1). • Leaflets and electronic messaging were implemented in EDs explaining triage and administrative processes (n = 4). • A process for informing patients of wait times was embedded into routine practice, such as the shift handover procedure (n = 4). methods, such as patient satisfaction surveys, may have provided a more sensitive and comprehensive way of measuring the impacts of the AIHQIP [34,35]. As part of a comprehensive program of surveys, the Bureau of Health Information implements regular surveys of patients' experiences of public hospital and emergency care in NSW, with some surveys oversampling Aboriginal patients [36]. These surveys provide a powerful tool for monitoring and evaluating the extent to which health services are meeting the needs of Aboriginal patients. Our finding that the AIHQIP reoriented ED environments and systems towards the provision of culturally safe care for Aboriginal patients is encouraging. Organisations with cultural competency standards integrated into policies and practices have been found to foster culturally competent behaviours among health professionals [37]. It is therefore likely that the organisational changes achieved in this study, if sustained, will reinforce culturally safe care in participating EDs in the future.
A strength of our study was that the AIHQIP aligned with the available evidence on building cultural competency in health organisations. Although few studies have rigorously examined the effectiveness of cultural competency interventions, there is growing recognition of the need for sustained, multifaceted organisational-level approaches [10,38]. Interventions that combine organisational change strategies with efforts to develop individual competency are more likely to be effective than stand-alone low intensity interventions [39,40]. Similarly, the strong engagement of Aboriginal communities and staff in our study reflect best practice principles for working with Aboriginal people to improve health outcomes [20]. Our study reinforces the utility of CQI approaches as a tool for engaging Aboriginal communities in system change interventions. CQI should therefore form the basis of policy responses in this area.

Limitations
There are three main limitations of our study. First, two participating EDs were delayed in commencing their CQI projects and therefore had condensed implementation periods, which may have reduced the likelihood of achieving program objectives and influencing the study outcomes. Second, the routinely collected data used in the multiple baseline design have some limitations relating to quality and completeness. In public EDs in NSW, practices relating to the entry of patient information in ED information systems, such as timeliness and quality checking, may vary among staff within EDs as well as between EDs, which introduces the potential of information bias in the findings. Third, the researcher who conducted the key informant interviews also conducted site visits in some participating EDs. This may have influenced some interviewees' willingness to provide honest and fulsome responses during interviews.

Conclusions
The AIHQIP did not have an overall effect on the accuracy of recording of Aboriginality in ED information systems or on levels of incomplete ED visits in Aboriginal patients. Still, key organisational changes were achieved, which over time may lead to more welcoming and culturally safe ED environments for Aboriginal patients. Further research investigating the effectiveness of interventions to improve Aboriginal cultural safety is warranted. Endnotes 1 In this paper, Aboriginal and Torres Strait Islander people are referred to as Aboriginal people in recognition that Aboriginal people are the original inhabitants of the Australian state of NSW, where this study was conducted.
2 NSW Health clinical staff are required to ask the following question of all patients: "Are you (is the person) of Aboriginal or Torres Strait Islander origin?" 3 The NSW Perinatal Data Collection holds records of all births in NSW, regardless of the birth setting. 4 Due to data quality problems, the control period for ED 8 was from April 1st 2012 until the first site visit by the research team (September 1st 2013).