The determinants of healthcare utilisation in regional, rural and remote South Australia: A cross‐sectional study

Abstract Accessibility of health services outside metropolitan centres in Australia is sub‐optimal. Recognising the barriers and enablers of healthcare access in rural, remote and regional settings is necessary to improving health service access in these disadvantaged populations. Accordingly, this study aimed to examine the determinants of healthcare use in rural, remote and regional South Australia. Cross‐sectional survey data were collected from adults living outside metropolitan Adelaide in South Australia between April 2017 and March 2018. Using a multi‐modal recruitment campaign, eligible adults were invited to complete the 44‐item consumer utilisation, expectations and experiences of healthcare instrument. Independent predictors of health service utilisation (dependent variable) were determined using negative binomial regression. The questionnaire was completed by 3926 predominantly female (52.5%) adults aged ≥50 years (56.7%), residing in regional South Australia (84.5%). Fifteen independent variables were significantly associated with health service utilisation using univariate analyses. Using negative binomial regression analysis, two predisposing factors (sex, remoteness), three enabling factors (income, health literacy, employment), two need factors (health rating, multimorbidity) and two personal health practices (alcohol, diet) were independently and significantly associated with healthcare use. Female sex (OR = 1.436, p < 0.001), good/excellent health rating (OR = 0.589, p < 0.001) and high multimorbidity (OR = 1.408, p < 0.001) were the strongest predictors of health service use. These findings will help inform the development of targeted health promotion and service engagement strategies for regional populations, which in addition to addressing workforce shortages, may help address inequity in health outcomes, particularly for groups engaging with regional healthcare services infrequently.

Where populations are exposed to multiple structural, environmental, and consumer-related barriers to healthcare access, there is likely to be a reduction in service utilisation, and subsequently, considerable unmet healthcare need (Corscadden et al., 2017;Russell et al., 2013). This is certainly apparent in rural, remote and regional locations, where geographical isolation, stoicism and healthcare avoidance behaviours are evident (Bourke et al., 2012), and availability, approachability, acceptability, affordability, appropriateness and utilisation of health services are often less than optimal (Leach et al., 2021;Thomas et al., 2015).
An important first step to improving access to healthcare in rural, remote and regional populations (hereon referred to as 'regional') is recognising the factors that facilitate and hinder healthcare access/utilisation in these populations. These factors-referred to as determinants of healthcare utilisation-are not only numerous and complex but also vary in their conceptualisation and operationalisation. One framework that has been well used over the past five decades, across diverse healthcare settings including regional settings (Li et al., 2016;Shah et al., 2014), and is clear and concise in its conceptualisation of these determinants, is Andersen's behavioural model of healthcare utilisation (Andersen & Davidson, 2007).
Andersen 's model posits that healthcare access/utilisation is influenced by four types of factors (domains): predisposing (e.g. education, region of residence), enabling (e.g. income, employment status), need (e.g. health status, morbidity) and personal health practice (e.g. alcohol consumption, smoking status) factors (Andersen & Davidson, 2007). Using Andersen's model to understand the determinants of healthcare utilisation in regional populations may help facilitate improvements in health service access/use within these regions, by informing the development of appropriate and targeted health services planning activities, health policies and needs-based resource allocation. Given that regional communities are not homogenous, each community is likely to encounter different healthcare needs, challenges and opportunities. Accordingly, any health workforce/services planning activities conducted in these regions should be undertaken at a level that is likely to generate findings that are meaningful and translatable to these communities. Such activity should also take into account constraints associated with jurisdictional boundaries (e.g. health system borders). Taking these factors into consideration, we focussed our attention on a group of communities that were serviced by the same state-based health system and governed by the same health policies. We also selected a region where healthcare access/utilisation was problematic (i.e. there was considerable spatial isolation), there was known to be a high level of unmet healthcare needs, and the factors impacting healthcare access/utilisation had not been rigorously examined.
Regional South Australia represents one of the least densely populated regions in the world (Population of Australia, 2019), where rates of chronic disease, co-morbidity, and psychological distress are among the highest of any Australian State or Territory (Australian Institute of Health and Welfare, 2016). And yet, the predictors of healthcare use in this region have received little attention to date. To address this knowledge gap, we aimed to examine the determinants of healthcare utilisation in regional South Australia.
What is known about this topic?
• Myriad factors havebeen shown to impact access to, and utilisation of health care services.
• Understandingand managing these determinants is critical to addressing the unmet health careneeds of communities.
• The determinantsof health care access/utilisation in regional South Australia is currently poorlyunderstood.

What this paper adds?
• Our studyidentified myriad predisposing, enabling and need factors, and personal healthpractices that predicted health service utilisation in this population.
• The range offactors impacting health service use indicate this phenomenon is more complexthan previously anticipated.
• Understanding this complexity is critical todeveloping effective policies and strategies to improve health care access/utilisationin this population.

| Study design
The Regional South Australia Health (RESONATE) project was a cross-sectional study designed to explore the healthcare needs and behaviours of the remote, rural and regional South Australian population. The study was developed to address eight primary objectives, which are described in detail elsewhere . This paper responds to the eighth objective of RESONATE: to identify the determinants of health service utilisation among adults living in regional South Australia. The specific question that this research intended to answer was: What predisposing, enabling, need and personal health practice factors predict health service use in a sample of regional South Australians?
Health service use (dependent variable) was defined as the sum number of visits to all healthcare providers in the preceding 12 months. This included visits to a nursing/midwifery (e.g. nurse, nurse practitioner, midwife), complementary medicine (e.g. acupuncturist, aromatherapist, Ayurvedic practitioner, chiropractor, herbalist, homeopath, kinesiologist, massage therapist, naturopath, osteopath, reflexologist, reiki practitioner, shiatsu practitioner, tai chi instructor, traditional Chinese medicine practitioner, yoga instructor), allied health (e.g. dietician/nutritionist, exercise physiologist, optician/optometrist, occupational therapist, pharmacist, physiotherapist, podiatrist, psychologist/counsellor, social worker, speech pathologist) or medical (e.g. general practitioner, medical specialist, dentist) professional. We examined healthcare use across all health providers, rather than by particular disciplines or professional groupings, to capture a 'whole of system' perspective of the phenomenon.

| Participants
Adult residents of regional, rural, remote or very remote South Australia, who had used any healthcare service or health intervention in the preceding 12 months, and had internet access or a fixed address (to enable the questionnaire to be completed online or to be posted, respectively) were eligible for inclusion in the study.
Residents who had severe cognitive or visual impairment, or were unable to provide informed consent, were ineligible to participate.
Based on a ±3% margin of error, 99% confidence level, and a target population of 290,290 adults (Australian Bureau of Statistics, 2018a), we required a sample size of at least 1832 people to identify health service use at any prevalence rate (i.e. frequency of the outcome in the population = 50%).

| Data collection
Participants were recruited using non-probability sampling and an extensive, multi-modal marketing campaign. Recruitment was undertaken between April 2017 and March 2018 and included the use of print media (e.g. newspaper articles, residential mailbox drop across regional South Australia), online media (e.g. organisational email blasts, development of a dedicated project website), social media (e.g. Instagram and Twitter posts, Facebook advertisements), broadcast media (e.g. radio interviews, television classified advertisements) and community engagement activities (e.g. attendance at community/public events, meetings/presentations with regional stakeholder groups). A detailed description of the recruitment campaign is reported elsewhere . All recruitment activities directed participants to either call a toll-free telephone number (to acquire a print-copy of the questionnaire) or visit the project website (where they could complete the questionnaire online, via the SurveyMonkey™ platform). Questionnaires completed in print-copy were returned by reply-paid mail, and on receipt, were manually entered into the online survey platform by the research team.

| Questionnaire
Participants were invited to complete the consumer utilisation, expectations, and experiences of healthcare instrument (CONVERSATIONS)-a questionnaire specifically designed to measure healthcare need, health service utilisation, and the determinants of healthcare use . This selfadministered, multidimensional questionnaire comprises 44 items, divided into five segments: lifestyle and health status (10 items), preferences concerning health service mix (1 item), use/attitude/ experience/satisfaction of conventional health services (8 items), use/attitude/experience/satisfaction of complementary health/ self-prescribed services (9 items) and demographic characteristics (16 items). Psychometric evaluation of CONVERSATIONS has shown the instrument to have acceptable internal reliability, good test-retest reliability, a high degree of acceptability and good content validity. A more detailed narrative of the development and validation of CONVERSATIONS is included in the study protocol .

| Exposures
The four exposures in this study reflect the four determinants of healthcare use described in Andersen's behavioural model of healthcare utilisation (Andersen & Davidson, 2007): • Predisposing factors: These preceding conditions predispose an individual to use a health service  • Enabling factors: These factors foster or impede health service use , including: gross annual household income (i.e. AU$0-$29,999, AU$30,000-$59,999, AU$60,000-$89,999, AU$90,000 or more), employment status (i.e. unemployed/retired vs. employed), English language proficiency (i.e. not at all/not well vs. well/very well) and health literacy (i.e. needs assistance none of the time vs. needs assistance some/most/all of the time).
• Need factors: These factors represent a potential or actual need for healthcare . These need factors were mea-

| Data analysis
Data were analysed using IBM® SPSS® Statistics v.25.0. Konstan et al. (2005) deduplication procedure for online surveys was used to identify and manage duplicate entries. Any data omitted from non-forced response items (i.e. religion, household income) were reported as 'missing'. To account for the non-probability sampling strategy, we adjusted the survey sample distribution by applying weights to the sex, age (i.e. by 5-year age group), and location distribution (i.e. by ABS statistical area 3 level) of the regional SA population. Weights were derived from 2016 Australian population census data (Australian Bureau of Statistics, 2019). The dependent variable (i.e. frequency of healthcare utilisation over the previous 12 months) was log-transformed to ensure data followed a near-normal distribution. Univariate analyses (using independent samples t-tests for comparisons between two sub-groups, and ANOVA for comparisons between three or more sub-groups) were performed to identify differences in healthcare utilisation between sub-groups (e.g. female vs. male) of each variable (e.g. sex). Variables demonstrating a statistically significant (p < 0.05) difference in healthcare utilisation were included in the regression models. We used negative binomial regression to examine the association between the dependent variable-healthcare utilisation (i.e. number of visits with any healthcare provider)-and the independent variables listed within each of the four domains of Andersen's model (i.e. predisposing, enabling, need and personal health practice factors). A two-sided p-value <0.05 was considered statistically significant. To ensure the assumptions for negative binomial regression were not violated, we tested for normality of distribution (using probability-probability plots), homoscedasticity (using scatterplots), and multicollinearity (using variance inflation factors).

| Ethics
This study was reviewed and approved by the Human Research Ethics Committee of the University of South Australia (Protocol ID: 0000034611). All participants were provided with a copy of the study information sheet and were required to provide informed consent before commencing the survey. Participants were anonymous, and they could choose to opt-in/opt-out of the study at their discretion, without consequence.

| RE SULTS
The questionnaire was completed by 3926 adults. The effective sample size (after adjusting for sex, age and location distribution) was 3743. Calculation of the survey response rate was not possible as the denominator (i.e. population reached) was not known.

| Predisposing factors
Univariate analyses (of weighted cases) found five of the nine predisposing factors were statistically significantly associated with healthcare use (Table 2). Significantly higher rates of healthcare utilisation were evident among female participants, as well as those that had completed none/primary/secondary education, identified as religious, were born in Australia and resided in inner regional locations.

| Enabling factors
Three of the four enabling factors were shown to be statistically significantly associated with healthcare use ( Table 2). Participants that were unemployed or retired, had suboptimal health literacy (i.e. required assistance with understanding health information some/ most/all the time) and reported an annual gross household income of less than AU$60,000 had significantly higher levels of healthcare utilisation than their subgroup counterparts.

| Need factors
Univariate analyses found both need factors to be statistically significantly associated with healthcare use ( Table 2). The frequency of healthcare provider visits over the past 12 months was significantly greater among participants living with multimorbidity (i.e. two or more chronic conditions), and those rating their overall health as fair or poor.

| Personal health practices
Of the five personal health practices examined, all were shown to be statistically significantly associated with healthcare use (Table 2).
Significantly higher rates of healthcare utilisation were found for smokers, those reporting a lower level of compliance with WHO physical activity guidelines, as well as those demonstrating a higher level of compliance with NHMRC alcohol consumption guidelines and Australian dietary guidelines, and those with no risk of lifestyleassociated all-cause mortality.

| Predictors of health service utilisation
The 15 independent variables that were significantly associated with health service utilisation were entered into a negative binomial regression model (

| DISCUSS ION
This study set out to identify the determinants of health service utilisation among adults living in regional South Australia. Eight factors were found to predict health service utilisation in this rural, regional and remote sample; including two predisposing factors, two enabling factors, two need factors and two personal health practices.
The range of factors impacting health service use in this sample indicate that health service utilisation in regional communities is a complex issue.
Significantly higher rates of healthcare utilisation were evident among female participants, as well as those who resided in inner regional (as opposed to outer regional, remote or very remote)

Determinant Sub-group
Weighted (n = 3743) Unweighted (n = 3926) b Difference in the mean number of healthcare provider visits between subgroups (using independent-sample t-tests of log 10 transformed weighted data for comparisons between two groups, and ANOVA for comparisons between three or more groups).

Statistically significant p values are in bold.
Test statistics are in italics. locations. Female gender has been widely found to be predictive of health service utilisation both in Australia (Mills et al., 2012;Parslow et al., 2004) and overseas (Fleury et al., 2012), as has rurality (Mills et al., 2012). These findings suggest that careful attention needs to be paid to ensuring that health services appropriately engage with, and are acceptable and accessible to, individuals living in regional settings, especially men. This is particularly important given that men are significantly more likely to die from preventable illness than women (Bots et al., 2017;Judd et al., 2008;Perez et al., 2020).
An interesting observation in this study was the absence of a statistically significant association between age and healthcare utilisation. Although this finding is contrary to the results of other research (where age and healthcare use have been shown to be positively correlated; Harrison et al., 2019;Recchia et al., 2022), the finding is not unique, with many studies reporting similarly divergent results (Agyemang-Duah et al., 2020;Rana et al., 2019;Thompson et al., 2016). A possible explanation for our finding may relate to the sample itself. First, our sample comprised only of regional South Australians-a population shown in previous research to have high rates of stoicism and perceived need for self-relianceboth of which have been identified as significant barriers to health service use in regional populations (Hull et al., 2017). Second, our sample (as reported elsewhere; Leach et al., 2020) had a high prevalence of mental health disorders relative to urban South Australia (not dissimilar to that reported internationally; Fu et al., 2018); these disorders are consistently reported as barriers to health service use (Leyva et al., 2020;Thompson et al., 2016). It is possible that these factors combined may have attenuated the association between age and healthcare use, further highlighting the complexity of healthcare utilisation in regional populations.
In terms of enabling factors, participants who were unemployed or retired or had an annual gross household income of greater than AU$90,000, reported significantly higher levels of healthcare utilisation than their subgroup counterparts. These findings are consistent with that reported in other Australian studies (Collie et al., 2021;Schlichthorst et al., 2016). Although these findings may appear contradictory, they can be elucidated to some extent through the lens of affordability. For instance, unemployed or retired persons in Australia may be eligible to receive national healthcare concessions to help reduce the costs of health services and medicines. By contrast, persons earning a high income may have the financial means to either pay for health services out-of-pocket or access private services covered by private health insurance (Carpenter et al., 2015;Gao et al., 2022). In either case, affordability appears to be an important predictor of healthcare access and utilisation.
Another determinant of healthcare utilisation is health need.
Regional populations are disproportionally affected by chronic disease, and as such, have a high level of health need relative to urban populations (Hetzel et al., 2015;Leach et al., 2020). Indeed, our analyses revealed that multimorbidity (i.e. living with two or more chronic conditions) and fair/poor overall health were significant pre-  Glynn et al., 2011;Jardine et al., 2011). Although an integrative and multidisciplinary approach is preferential for the management of multimorbidity (Leach & Steel, 2018;Palmer et al., 2018;Smith et al., 2016), this approach may be somewhat problematic to implement in regional communities where access to diverse allied health services is generally poor. Thus, failing to address health workforce maldistribution in regional settings may essentially deny regional communities access to best practice care.
Personal health practices (including lifestyle behaviours) are closely correlated with the development and progression of chronic disease and multimorbidity (Freisling et al., 2020), as well as increased health service use, including hospitalisations (Arcury et al., 2005;Haapanen-Niemi et al., 1999;Luben et al., 2020). While this was supported in univariate analyses, our regression analyses did not find lifestyle behaviours, such as smoking status and compliance with physical activity guidelines, to be significant predictors of healthcare utilisation in this regional South Australian sample.
One explanation may relate to how some data were collected in the CONVERSATIONS instrument. For example, the smoking status item did not differentiate non-smokers from former smokers. Given that former smokers are significantly more likely to use health services than non-smokers and smokers (Kahende et al., 2009) (Collins et al., 2011;Daya et al., 2020;Haapanen-Niemi et al., 1999). Although it is not entirely clear why patient adherence to alcohol and dietary guidelines was associated with higher health service use, it may be that regular health service users were more exposed to contemporary health information on risks associated with alcohol consumption and poor dietary behaviour. Another possible explanation is that in light of the high prevalence of preexisting chronic disease and multimorbidity in the study sample, these health practices may simply reflect adaptive behaviours (i.e. consequential changes due to the diagnosis of a disease) rather than primary prevention behaviours. Either way, the findings point to several key areas requiring additional investigation; the findings of which may help further our understanding of the impact of health service access on the health practices of individuals living in regional communities.
The strengths of this study are the large sample size, the inclusion of participants from diverse remoteness categories, and the representativeness of the sample compared with regional South factors impacting health-seeking behaviour in this regional community. Thus, future work may need to consider looking beyond the variables measured in this study; for example, attitudinal influences on regional help-seeking behaviour that may operate separately to health literacy (Dollman et al., 2021).

| CON CLUS ION
This study highlights a range of predisposing, enabling and need factors, and personal health practices that predict health service utilisation in rural, regional and remote South Australia. The findings have also alluded to several groups that may be susceptible to engaging less frequently with health services, including males, and those living in an outer regional, remote or very remote location. However, this phenomenon appears to be more complex than previously anticipated, with a wide range of factors contributing to health service utilisation in this population. Thus, further research is required to explore what the other predictors of health service use might be in this and other regional populations.
At the very least, the findings from this research will help inform the development of policies and targeted strategies aimed at improving healthcare access, and reducing health inequities, in regional Australia.

AUTH O R CO NTR I B UTI O N S
ML conceptualised the project. ML drafted the introduction, methods and results. KM, KG and ML drafted the discussion. All authors reviewed and edited the manuscript and approved the final manuscript.

ACK N OWLED G EM ENTS
The authors wish to thank the residents of regional South Australia who kindly participated in the study and the community organi-

FU N D I N G I N FO R M ATI O N
This study was supported by infrastructure provided by the Department of Rural Health, University of South Australia.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.