We found that increasing age, female sex, lower pre-tax household income, not having a partner, not being in paid work, Indigenous background and living in a regional or remote location were key influences on use of HACC services. People who were born outside Australia or spoke a language other than English at home were significantly less likely to use these services. The higher use of HACC with lower income, after adjusting for other need-related variables, suggests that fees for HACC services are not a major barrier to their use.
Rates of HACC service use increased with increasing remoteness regardless of other need-related factors. This reflects targeting of services to special needs groups, as well as perhaps facilitation of access to services in rural areas through closer community networks  and variations in assessment practices . Similar factors, as well as higher rates of disability, may contribute to higher rates of HACC service use among Indigenous people.
People born overseas and speaking a language other than English at home were around 20% less likely to use HACC services than other individuals. Studies in other Australian states have found that the proportion of HACC clients speaking a language other than English at home was similar to the population proportion [8, 24], but they received fewer hours of service overall, and made less use of delivered meals and social activity groups . Our findings regarding home-delivered meals were similar, but we found that HACC clients who spoke languages other than English at home were more likely than others to use centre-based day care, most likely reflecting the availability of culturally specific services in NSW.
Our study was the first to examine associations between a range of lifestyle-related factors and HACC service use. We found a U-shaped association with body weight: both underweight and obese individuals were significantly more likely to be HACC clients than people who were overweight or of normal bodyweight. Both these conditions reflect opportunities for improved nutrition among HACC clients. Underweight and malnutrition are particular problems among the elderly and are major contributors to hospitalisation and physical decline . Consistent with our findings regarding bodyweight, eating no or few daily serves of fruits and vegetables was also more common among HACC clients, again identifying opportunity and needs for improving nutrition for this client group.
Rates of HACC service use were 2 to 3 times higher among people who were sedentary compared to other individuals. There was also a particularly strong association between HACC service use and number of falls in the past year; taken together these findings demonstrate the considerable potential for implementing programs to increase levels of physical activity, with a focus on improving strength and balance, in the HACC context. Good practice guidelines for such programs are already available .
Current smoking was also higher among HACC clients, indicating the potential to engage this client group in targeted smoking prevention programs. Conversely, HACC clients were more likely than other participants to abstain from drinking alcohol. This finding cannot be taken to indicate that alcohol consumption has beneficial effects, because it is possible that participants may have given up drinking as a result of poor health.
Consistent with earlier Australian studies [10–12], and as expected given that people receive HACC services because they have a disability, we found strong gradients of increasing HACC service use according to decreasing level of physical functioning, and poorer self-ratings of eyesight and memory. Reflecting the disabling nature of many chronic health conditions, we found also that HACC clients were more likely than non-clients to report having conditions including Parkinson's disease, stroke or diabetes, anxiety and depression, cancer, heart attack or angina, blood clotting problems, asthma and osteoarthritis. More detailed analysis to explore the service types used by these clients is required to assess the potential for implementing intensive chronic disease management programs through HACC services.
Important strengths of our study include its large population-based sample, the detailed data about socio-demographic, lifestyle and health-related factors that were available though the 45 and Up Study questionnaire, and the independent ascertainment of HACC service use through data linkage. It allowed us to characterise the health of HACC clients in a very comprehensive way.
It is possible that HACC clients in the 45 and Up Study were not representative of the broader HACC client population. In keeping with other similar large-scale population-based cohort studies, its response rate was 18% . A comparative analysis found that the prevalence of many factors in the 45 and Up Study, including country of birth, educational attainment, fruit consumption, body-mass-index and falls, was similar to the NSW Population Health Survey (PHS), a population-based survey which has a response rate of around 60%. However, 45 and Up participants tended to have higher incomes, and had lower prevalence of smoking, high psychological distress, hypertension, diabetes and asthma . This suggests that 45 and Up Study participants are in general "healthier" than the overall population. However, importantly, we have reported relative measures of effect (RRs) calculated from internal comparisons within the 45 and Study, which will be valid provided there is sufficient heterogeneity within the predictor variables . Moreover, empirical data demonstrate that RRs for a wide range of exposure-outcome relationships in the 45 and Up Study are very similar to those calculated using 'representative' PHS data . Any bias resulting from an absence from the 45 and Up Study of the sickest, most dependent, HACC clients would generally cause underestimation of the associations between health risk factors, health conditions and HACC use. Misclassification relating to the incomplete coverage of the HACC MDS is also likely to bias findings towards the null.
All of our predictor variables were based on self-report. In general, people tend to under-report lifestyle risk factors , and to under-report many but not all health conditions . Again, any resulting bias would most often result in more conservative RR estimates in relation to HACC use.
As a cross-sectional study, ours could not identify temporal sequences of events, nor identify personal and service-related factors that can potentially prevent avoidable admissions of home care clients to hospital or residential aged care [31, 32]. Longitudinal analyses needed to address questions of this type will become possible as the 45 and Up Study progresses.