Duration of participation in continuous quality improvement: a key factor explaining improved delivery of Type 2 diabetes services

The international diabetes epidemic is expected to escalate over coming decades placing increasing economic and social pressures on health systems and communities [1]-[3]. In addition, socio-economic and ethnic inequities have been identified in provision of diabetes care [4]. Social determinants and lack of access to quality primary health care (PHC) services are important contributors to the inequity in health outcomes between Aboriginal and Torres Strait Islander communities and the Australian non-Indigenous population. Over the past decade, there has been a significant increase in prevalence of diabetes in the Aboriginal and Torres Strait Islander population and they are now three times more likely to have diabetes compared to the non-Indigenous Australian population [5].

Along with targeted prevention strategies, improvements in chronic disease care are necessary to reduce inequity, advance population health outcomes and lessen the burden on health care systems. Evidence indicates that improvements in quality of care and health outcomes can be achieved using integrated care frameworks and multifaceted improvement strategies targeting changes at all levels of the health system [6]-[9]. However, the effectiveness of local and large scale quality improvement approaches in health care remains uncertain with published studies showing considerable variation in levels of improvements achieved [10],[11].

Of the 132 participating health centres, 79% were located in very remote or remote areas, 44% had a service population of 500 or fewer, and 73% were government operated (Table 2). In all jurisdictions, on average, approximately 90% of clients audited attended the health centre within the previous six months. By location, there was a higher mean percentage of clients not attending within the previous six months in non-remote areas (25%) compared to remote (11%) and very remote areas (5%; Table 3). The mean age of patients was 52 years and 57% were female. Over 90% of records from the Northern Territory, Western Australia and South Australia were for Aboriginal or Torres Strait Islander people, compared to 69% in Queensland and 55% in Far West New South Wales health centres. Around 71% of patients had co-morbidities and 14% had one or more diabetes complications. These occurred more commonly in older patients (Table 4).

Of all health services participating in the study around one third (34%) completed three or more cycles, 42% completed one to two cycles and 24% completed baseline audit only. This does not reflect attrition of health services, as some services commenced participation at differing time periods (Table 5). Of those health centres only completing baseline audits, 75% commenced ABCD CQI participation in 2011 or 2012 (Table 1).

The distribution of remoteness of centres did not differ markedly by duration of participation in CQI, for example, 66%, 62% and 76% of health centres completing baseline, 1-2 audit cycles and 3 or more audit cycles respectively were from very remote areas. The distribution of health centres by governance showed that 47% of health centres that completed 3 or more cycles were community-controlled organisations and there was also an equal distribution of health centres by service population size that completed 3 or more cycles (Table 5).

Wide variation in delivery of Type 2 diabetes services was evident across all health centres within an audit year and over time (Figure 1a). However, for those health centres that completed at least three audit cycles (n = 45) there was a trend of decreasing variation across successive cycles of CQI (Figure 1b). There were large improvements in service delivery in lower performing health centres from less than 20% in early cycles to 40-50% in later cycles (Figure 1b).

In explaining this variation, the unadjusted logistic regression analysis showed that duration of participation in the ABCD CQI program, remoteness, regularity of patient attendance, patient age, level of co-morbidity and number of disease complications, were significantly associated with improved Type 2 diabetes service delivery (Table 6). Remoteness of health centres showed the strongest association with very remote and remote centres having 4.57 (95% CI 2.61-8.01) and 2.55 (95% CI 1.22-5.35) times the odds of being in the top quartile of Type 2 diabetes service delivery compared to non-remote centres. We tested the interaction between remoteness and duration of CQI participation and found significant improvement in service delivery with length of participation irrespective of health centre location. There was, however, a gradient of increasing odds of improvement from non-remote, to remote and very remote centres. The odds of receiving top quartile service delivery also increased with increasing age, co-morbidity and disease severity. Considering the likelihood of a correlation between patient age and disease status, we tested and found a significant interaction between these variables. All interaction terms were incorporated into the final adjusted model as shown in Table 7.

There was a significant effect of time, with a steady increase in the odds of health centres being within the top quartile of Type 2 diabetes service delivery from baseline to 2012. Health centres had approximately three times greater odds of top quartile performance in 2012 compared to 2005/2006 (95% CI 2.37-3.53) (Table 7, Model A). The effect of time diminished after adjusting for health centre and patient level factors (Table 7, Models B and C).

The PCV in Model B (Table 7) shows that the addition of health centre factors explained 37% of the variation across jurisdictions and health centres. The longer the duration of participation in the ABCD CQI program, the greater the odds of being within the top quartile of service delivery, particularly for health centres in remote and very remote locations (remote:1-2 cycles OR = 2.92, 95% CI 1.36-6.24; ≥3 cycles OR = 3.14, 95% CI 1.37-7.18; very remote: 1-2 cycles OR = 4.03, 95% CI 2.27-7.16; ≥3 cycles OR = 4.72, 95% CI 2.47-9.02). Health centres with a higher proportion of patients that attended within the previous six months also had better odds of being within the top quartile (OR = 1.40; 95% CI 1.22-1.60). The addition of patient level factors in Model C had a modest effect, accounting for a further 1% of the variation in service delivery across health centres and jurisdictions. In most age groups, patients with co-morbidities and disease complications had greater odds of receiving higher level care than those without these conditions.

The variation in the odds of top quartile delivery was reduced between jurisdictions and health centres after accounting for health centre and patient variables from MOR_STATE-HC = 4.68 (Model A; Table 7) to MOR_STATE-HC = 3.38 (Model C; Table 7). Changes in predicted odds for top quartile service delivery after adjusting for time and health centre and patient factors are shown in Figures 2 and 3 respectively. The highest performing outlier after adjusting for health centre and patient characteristics (Figure 3b) is a centre from a non-remote area that had completed one ABCD CQI Type 2 diabetes audit cycle since it joined in 2010. This health centre is renowned for its clinical leadership and model of care suited to the local Aboriginal and Torres Strait Islander population; factors which are not accounted for in this current analysis.

There were improvements over time in the level of Type 2 diabetes care across the diverse range of Aboriginal and Torres Strait Islander primary health centres participating in a wide-scale CQI project. The main factors associated with improvement were duration of participation in CQI, the location of the health centre, regularity of patient attendance and patient age, co-morbidity and disease severity.

A significant proportion of the variation across jurisdictions and centres was explained by health centre factors within the model (37%). There is an independent relationship between the duration of participation in ABCD CQI and improved documented adherence to best practice guidelines, with centres that participated for 3 or more cycles having between 1.5 to 5 times the odds of top quartile service delivery to patients depending on location (Model C; Table 7). This finding indicates that health centres in varying geographical contexts appear to obtain benefits from sustained CQI participation. Long term commitment to CQI was also associated with improvements in health centres at the lower range of performance, with an increase in the lowest levels of adherence to best practice guidelines from less than 20% to approximately 40% over three audit cycles (Figure 1b).

Health centres located in very remote (ASGC-RA 5) and remote areas (ASGC-RA 4) had increased odds of greater improvement in care processes over audit cycles compared to non-remote health centres. The ASGC-RA classification system is based on physical distance of the community to the nearest urban area and reflects access to goods and services [18]. The greater odds of a higher level of service delivery in these areas may reflect less complex service environments where the centre may be the single primary health care provider and some distance away from other providers. The availability of multiple service providers in non-remote locations provides a challenge for primary health care centres in coordinating and monitoring comprehensive delivery of care to patients with chronic disease. Yet, the quality of care shown by remote health centres has occurred in service environments characterised by high staff turnover, high use of locums and limited access to specialist services. Higher levels of service delivery in these areas may be due to the implementation of specific models of primary health care delivery (for example, hub and spoke and outreach services) designed to overcome barriers of remote geography, dispersed populations and to improve coordination of care [19]. CQI may have contributed in part to the development of these models of care as indicated in a recent evaluation of the CQI Strategy in the Northern Territory that suggested CQI processes led to positive changes in health centre practices and an increase in service delivery outputs [20].

Best practice Type 2 diabetes care dictates that patients receive certain services at regular intervals. It is unsurprising that health centres with a higher percentage of patients (≥3%) not attending in the previous six months had lower odds of providing best practice care. While non-remote health centres had on average 25% clients not attending in the previous six months compared to 11% and 5% of patients in remote and very remote areas respectively (Table 3), there was no significant statistical interaction between regularity of patient attendance and remoteness of health centres with respect to delivery of Type 2 diabetes care.

The addition of patient-level characteristics into the model explained only a small component of the variation between health centres and jurisdictions (1%). There was an increasing trend of co-morbidity and disease severity with age (Table 4) and those patients who were older, had co-morbidities and/or complications were more likely to have greater odds of receiving top quartile service delivery. Within most age groups, patients with co-morbidities and complications had greater odds of receiving more services compared to those without. This finding may reflect more opportunities for providing care or greater emphasis on managing high risk patients. Higher levels of care for high need patients along with general improvements in overall standards of care should contribute to a reduction in the Type 2 diabetes disease burden on health care systems and improved population health outcomes.

There was a consistent trend of increasing odds of top quartile service delivery over audit years compared to the baseline year 2005/6 apart from a flattening out period in 2010 (Table 7, Model A). The 2010-2011 period marked a new expansion phase for the ABCD project with new centres joining and the commencement of a nation-wide CQI service support agency (One21seventy) designed to assist with implementation of the ABCD CQI tools and processes. Also in 2010, two jurisdictions implemented a network of regional CQI facilitators to support health centres with their CQI activities. The number of health centres participating and the number of audits undertaken, including Type 2 diabetes, increased substantially from 2011 once the new support infrastructure was in place. Jurisdictional differences in the odds of health centres being within the top quartile of service delivery may be attributed to the variable state-wide level support systems as well as the staggered commencement of participation in the ABCD project. Employment of regional positions that supported collaboration and information exchange across health centres was a key explanatory factor for improved performance at health centre level over time [21]. The top two ranked jurisdictions with respect to odds of top quartile delivery, before and after adjusting for health centre and patient factors (Figures 2a and 3a), are those that participated for more years in ABCD and/or have a regional CQI facilitator network. Similar factors have been attributed to the success of a long term diabetes quality of care project in a remote community of Western Australia [22]. Learnings from jurisdictions where there is well established macro-level CQI infrastructure can support the development of similar structures in other jurisdictions [23].

The investigation of patient, health centre and jurisdictional level factors that underlie variation in adherence to best practice Type 2 diabetes care has been made possible by the availability of the extensive CQI audit dataset from health centres participating in the ABCD National Research Partnership. The voluntary nature of participation in this study by health centres, however, limits the generalisability of study findings. In addition, as data are collected from client records, service delivery may be underestimated due to poor documentation. Although a strength of this study design is the ability to correct for a range of known confounding factors, it is also recognised that the associations identified in this study, may be due, in part, to other unmeasured confounders. For example, there may be factors associated with longer term participation in CQI that may be responsible for the demonstrated improvements in adherence to best practice guidelines, such as well-functioning community health centres with good leadership and workforce stability that are both more likely to maintain CQI processes, and to achieve better results in diabetes management. The exemplary centre in Figure 3b is a case in point where their level of service delivery is amongst the best despite relatively limited participation in ABCD CQI to date. As mentioned, the health centre is renowned for its clinical leadership and model of care built around its own local CQI processes. To elucidate other potentially confounding factors enabling quality improvement, further quantitative and qualitative studies are planned to examine micro-, meso- and macro-level factors that have facilitated sustained improvement in high performing health centres [24]. It is important to note that it is through participation in the CQI initiative that data are available for the sort of analysis presented in this study. There are currently no other sources of comparable data in Australian primary care that would enable establishment of a comparison group not participating in the CQI program.

The findings of this study suggest that at the health centre level, the quality of Type 2 diabetes care for Aboriginal and Torres Strait Islander communities is associated with long term commitment to a CQI program. Sustained participation has particularly benefitted remote and very remote health centres and health centres performing at the lower range of service provision. Our findings suggest that improving regularity of patient attendance, better coordination and documentation of care within non-remote multiple service provider environments and sharing lessons from the successes of care coordination in remote areas has the potential to improve care delivery across the spectrum of health centres. Further understanding of system wide factors that underpin jurisdictional variation in delivery of care will assist health managers and policy makers to understand and develop macro-level strategies to improve quality of Type 2 diabetes care in Aboriginal and Torres Strait Islander communities.

VM performed the statistical analysis and drafted the manuscript. GS conceived the study, participated in its design, assisted with statistical analysis and drafting of the manuscript. RB conceived the study and assisted with interpretation. JM assisted with statistical analysis. CC, CK, RK, SL, EM, ST and DS were instrumental with data acquisition and interpretation. All authors read and approved the final manuscript.