Community members and service providers residing in this remote Aboriginal community, in addition to the external partnering research institution, implemented 215 activities across three intervention years. This level of implementation demonstrates a high degree of community commitment to prevention of CVD and type 2 diabetes. With respect to the hypotheses, although the Healthy Lifestyles Project did not become more ecological over time (hypothesis 1) there was some evidence to support greater organisational partnering over the three-year project (hypothesis 2). These findings, reviewed below, should be interpreted in light of barriers experienced to implementation, and logistical and measurement challenges related to data collection and analyses.
Intervention strategies, settings and ecologicalness
Similar to other intervention studies applying the ecological coding procedure [10–12, 21] the majority of activities targeted by the Healthy Lifestyles Project were of the form IND, INT, or ORG. Community champions found activities with these targets necessary for raising the community's consciousness of the risk factors for CVD and type 2 diabetes and for motivating other community members to champion the Healthy Lifestyle Project. Indeed, experiences from other Aboriginal communities  and the community readiness literature [23, 24] suggest that activities with IND, INT and ORG targets may be necessary precursors to shifting a community from 'no awareness' to the 'pre-planning' and 'preparation' stages of readiness for intervention implementation. Future developments of the ecological coding procedure may need to consider the relative timing of implementing particular intervention strategies in relation to the level of community readiness.
Almost all participants were recruited into activities through organisational, family and community settings (via cultural networks and program champions). Integration of an ecological approach would have been strengthened by the introduction of activities that recruited participants from supra-national and societal settings. In this study, no participants were recruited through supra-national settings and few activities recruited participants though a societal setting. Examples of such actions could be local adoption of one or more recommendations from an international physical activity charter ('supranational' setting) or a state/territory policy offering community subsidies to transport fruits and vegetables to remote area grocery stores ('society' setting).
Our findings show that the Healthy Lifestyles Project did not become more ecological over time. We attribute this, in part, to the decentralised activity planning and implementation process. Academic researchers partnered with the community to screen participants for CVD risk factors and type 2 diabetes. Community research assistants collected these data in partnership with academic researchers, who were committed to strengthening community capacity. Results were returned to individual participants with full explanation (and a referral to the clinic, if necessary) and collectively to the community, with the intent of mobilising key stakeholders in an organised planning and implementation process [22, 25, 26]. Despite a cultural organisation championing the project, no central organising body with community-wide representation was able to form during the three-year intervention period. Consideration of barriers, below, illuminates some of the cultural and community factors that may have conspired against the formation of a centralised process.
Encouragingly, a greater proportion of activities involving inter-organisational collaboration was implemented in Year 3 as compared to the previous years. This was accounted for through the partnering of two organisations, one of these being the cultural organisation championing the project. Few programs involved more than two organisations in decision-making during the planning and implementation stages of a given intervention activity. Most interventions were primarily planned and implemented by a single organisation. This contrasts with the Kahnawake Schools Diabetes Prevention Project (KSDPP) in which one-half of activities involved partnering by Year 3 . In KSDPP, activities were planned and implemented through a centralised process.
A higher (or better) ecological score depends on organisations applying direct (e.g., IND) and indirect (e.g., INT, ORG) intervention strategies to reach participants in multiple settings (e.g., community, organisation). Had organisations partnered more frequently, their ecological score could potentially have been improved through the "collaborative advantage" accruing from organisations pooling their collective knowledge and resources to further the recruitment of participants across multiple settings and using their collective power to lobby for political (POL) and structural changes (COM, ORG). Such forms of changes are less attainable through the efforts of a single organisation . An excellent example of collaborative advantage comes from advocacy efforts that encouraged a very influential take-away food manager in the community to offer more healthful food options and to keep deep-fryers off until at least 11 am.
The efforts of certain change agents in the community stand out. These change agents were practitioners and cultural leaders dedicated to a holistic approach to health. They collaborated with other organisations and implemented person- and environment-centred strategies while recruiting participants from multiple settings. The organisations of these change agents received the highest ecological scores. In line with research on the ecological approach , these practitioners may have perceived themselves as having the skills to develop and/or implement environmental interventions and thus were more likely to target the environment for change than were others in the community.
Barriers to implementation of an ecological approach
Several factors may have conspired against the integration of a community-directed ecological approach and greater inter-organisational partnering in this remote Aboriginal community. These barriers represent the reflections of the study co-authors, which include two community members.
No single agency in the community was funded or had a mandate to co-ordinate prevention of CVD and type 2 diabetes. Even though a cultural organisation championed the project, it was apparent that some organisations found it difficult to work collaboratively around a shared vision whilst in competition for limited financial resources. Because organisations experienced pressure to demonstrate short-term accountability for deliverables they were less able to invest time and resources to work together to achieve longer-term outcomes required for primary prevention. Given the burden of chronic disease and issues relating to basic services, education, housing, development and social harmony, organisations tended to invest their limited resources in treatment, and respond to health and community crises. The sole organisation that partnered with the associated health research institution, which attempted to take a leadership role in facilitating centralised planning, did so while continually justifying this time investment to their funders.
Some Aboriginal stakeholders were sceptical of projects funded under western systems of accountability. Such projects were perceived not to respect cultural norms of non-interference and obligation. Community members emphasised the importance of non-Aboriginal community members working within established cultural respect frameworks . Conflicts between western and cultural systems acted as barriers to Aboriginal stakeholders forming partnerships. The high turnover of non-Aboriginal staff in organisational management positions also led to fatigue in inter-organisational engagement where partnerships were often formed on a personal rather than organisational level. Contributing to this sometimes tenuous relationship between organisations, were differences between clan groups represented within the different organisations. Further, western funding mechanisms did not consider the upfront time required to develop Aboriginal community member's program planning, management and research capacity, inclusive of understanding of how to influence policy. Capacity development, considered key to Aboriginal self-determination and program sustainability [7, 29], would have taken at least two years for the community to mount a coordinated ecological approach to community-based CVD and type 2 diabetes prevention.
Beyond community activation it can take many years for a community to build sufficient capacity to implement policy-level changes that promote program sustainability . As one example, six months after completion of the three-year evaluation of the Healthy Lifestyles Project, the Aboriginal owned store association implemented a Nutrition and Health Strategy in five communities to improve the nutritional quality of each store's food supply. This Strategy built on the three years of community-wide health promotion activity implementation associated with the Healthy Lifestyles Project. Recruiting community members into interventions implemented by the state or commonwealth ("societal") or broader supranational settings was constrained by community remoteness, the use of English as a second or third language, and the upfront time required to develop networks with external Aboriginal and non-Aboriginal stakeholders in the business, government and non-profit sectors. Moreover, the policy-making arena is largely perceived as a foreign process occurring external to the control of Aboriginal community members. Policy-making requires two-way learning whereby Aboriginal community members can share their expertise in grassroots community action and governance processes, while non-Aboriginal stakeholders can offer insights into collaborative planning to influence institutional structures and support policy implementation.
Study findings should be interpreted in light of the following logistical and measurement challenges and limitations.
Organisations in this community had previously been exposed to other participatory research projects, but were more accustomed to research focusing on outcomes, rather than process evaluation. Consequently, many organisations placed lesser value on completing activity monitoring forms, necessitating greater research assistant involvement toward ensuring data collection. Given that the community was only accessible by boat or air, considerable advance planning was required for researchers to collect data. Community remoteness also hampered researcher's efforts to develop community capacity through training and provision of ongoing mentoring for community research assistants. These logistical issues were compounded by difficulties in arranging times with implementing organisations and agencies to complete activity monitoring forms which arose, in part, from lack of a community norm for fixed schedules, and unanticipated cultural events and obligations. The persistence of the data collection effort and ongoing support of key community leaders in this study resulted in key findings likely less affected by such types of challenges.
From a measurement perspective, the extent of ecologicalness could have been under-estimated in our analysis. Organisational partnering occurring through culturally embedded collaborative planning could have occurred but may not have been identified on the monitoring forms. The ecological coding procedure may be limited in that equal statistical weighting is given to IND, INT, ORG, COM, POL and networked intervention targets (ie., ORG-ORG). It is the case, however, that significantly more time, collaborative capacity and resources are required to mount such types of activities in comparison to those with direct (HP→ IND) or indirect interpersonal (HP→ INT→ 1ND) targets. Moreover, intervention strategies that include distal POL targets (e.g., HP→ POL → IND) in which participants are recruited from societal or supranational settings carry the same weight as an intervention strategy in which clients are directly targeted (HP→ IND) through a proximal organisational setting. The coding algorithm as it stands may not adequately discriminate between and/or differentially weight such types of intervention strategies, particularly at the low end of the scale. This study has inspired further development of the ecological algorithm.