Exploring the associative relationship between general practice engagement and hospitalisation in older carers to potentially reduce hospital burden

Preparation of the dataset

The study design was a cohort study with participants retrospectively identified and prospectively compared. We used the UNSW Primary and Community Health Linkage Resource, which included the Sax Institute’s 45 and Up Study in NSW. The 45 and Up Study recruited a demographically representative population of NSW totalling 267,357 participants aged ≥45 years, who completed a baseline survey questionnaire on various sociodemographic and lifestyle characteristics. These participants were recruited between 2005 and 2009, and all gave written consent for health data linkage and follow up prior to participation (Banks et al. 2008; CHeReL 2023). Further intricacies of this study have been described elsewhere (Banks et al. 2008; Bleicher et al. 2023).

Medicare claims data were provided by the Services Australia Medicare enrolment database, which encompassed almost the entire cohort, with people aged ≥80 years, and residents of rural and remote areas being oversampled. Linkage of the 45 and Up Study cohort data to these datasets was facilitated by the Sax Institute using a unique identifier and deterministic matching. The 45 and Up Study baseline data were also linked by the Centre for Health Record Linkage to hospital records from the Admitted Patient Data Collection, and death notifications from the NSW Registry of Births Deaths and Marriages. The Centre for Health Record Linkage uses a probabilistic procedure to link records, in which records with an uncertain probability of being true matches are checked by hand. Secure data access was provided through the Sax Institute’s Secure Unified Research Environment.

The analytic cohort was limited to carers using a 45 and Up Study’s baseline survey question: ‘Do you regularly care for a sick or disabled family member or friend?’. The inclusion criterion was defined as respondents who affirmed this query. Exclusion criteria included those who died during the exposure or follow-up periods, as ascertained via the Deaths Registry. All-cause mortality was delineated as any fatality that occurred within 10 years after the baseline survey (Fig. 1).

Fig. 1.

Assembly of the analytical cohort.

The follow-up time for GP interactions was set at 3 years from the baseline survey, given that the baseline data collection had been completed in December 2009. GP interactions that occurred within 3 years of recruitment were the primary focus of the study.

GP use and hospitalisation admission

GP interactions were ascertained from specific Medicare Benefits Schedule (MBS) item numbers. These item numbers pertained to various GP encounters, including short and long GP encounters, non-referred presentations, health assessments, GP care plans, after-hours GP care and medication management reviews (see Appendix 1). If a participant had an entry bearing any of these specific MBS item numbers, they were considered to have had the respective GP interaction. Accordingly, participants were classified based on their frequency of GP interactions: high GP users were those who had engaged in ≥11 GP interactions per year, a criterion that accounts for a variety of factors, including average Australian GP visits (Harris-Roxas et al. 2023). Carers who engaged in an average of <11 GP visits per year, including those who had no GP visits at all, were classed as low GP users.

For hospitalisation status, Admitted Patient Data Collection hospitalisation records were selected to include entries that had occurred within the follow-up period; between 3 and 10 years after the baseline survey (Fig. 2). Participants with a hospitalisation entry in this period were categorised as ‘hospitalised’, with the time until hospitalisation calculated. Those without a hospitalisation in the period were categorised as ‘non-hospitalised’.

Fig. 2.

Timeline of the study.

Study timeline

Because each participant completed the baseline survey at different times between 2006 and 2009, the baseline, intervention and outcome periods are all relative to this initial, participant-specific date. Thus, the period in which GP interaction data is gathered commences directly after baseline and lasts 3 years, whereas the period in which hospitalisation data is gathered commences directly after that (Fig. 2).

Data analysis

Analysis was undertaken using RStudio 2023.03.0 + 386 (Posit). Health status, risk factors and demographic characteristics were compared between high and low GP users using Chi-squared tests of Independence, and the P-values were reported.

To explore the relationship between GP use and hospitalisation, we used relative risk (RR) as the primary metric (Clark and Barr 2018). Initial analyses were conducted using a univariate Poisson regression model, which provided crude RRs and their respective 95% confidence intervals. A multivariable Poisson model was employed to adjust for key covariates, notably age and sex, and other significant variables, as identified using Chi-squares P-values <0.05. The adjusted RR, along with its 95% CI, were outputted from this model.

Ethics approval

2019/ETH01682: 45 and Up Primary and Community Health Cohort: a vehicle for enhancing integrated care in the Sydney and South Eastern Sydney Region.

Demographics and characteristics of carers

The cohort of carers included in this analysis was diverse in terms of age, sex and other demographic characteristics. Table 1 summarises the demographics of the study cohort, as collected from baseline survey.

The age distribution of the carers studied showed the majority were aged between 45 and 59 years (51.4%), with the proportion of high GP users increasing with each increasing age group. The proportion of high GP users also increased with the number of chronic conditions the carer had, with 55.4% of those with three or more chronic conditions being high GP users. In terms of sex, the ratio of male to female carers in the cohort was over 1:2, with 17,573 of the total 26,004 participants being female (67.6%). Similarly, the majority of the carers in the cohort were part-time carers (64.3%), with more full-time carers being high GP users (32%) compared with part-time carers (19.4%). Excluding those on an income less than $5000 per year, the proportion of high GP users decreased as income increased, with 41.2% of those earning between $5000 and $9999 being high GP users compared with only 8.7% of those earning above $70,000 per year. The same is largely true for the qualifications of the participants, with more educated subgroups having a lower proportion of high GP use carers (14.9% of university graduates as compared with 36.5% of those who have no school certificate). Self-reported quality of life and self-reported health showed similar trends, with groups who rated their quality of life and health as worse having a greater proportion of high GP users. The same is true for those who had been told they were depressed in the past, with 33.5% of them being high GP users compared with 21.7% of those who had not been told they were depressed.

GP use and hospitalisations among carers

As shown in Table 2, there was a statistically significant association between frequent GP users and hospitalisations among carers with a crude RR of 0.51 (95% CI: 0.48–0.55) and an adjusted RR of 0.62 (95% CI: 0.55–0.69). All of the variables (age, sex, care hours per week, whether the carer cares full time, marital status, body mass index, housing type, work status, highest qualification, income, number of conditions, self-rated quality of life, self-rated health, depression status and alcohol consumption) identified as being significantly different (P < 0.05) using the Chi-squared test (shown in Table 1) were included as covariates in the adjusted model. This indicated that, when adjusted for confounders, carers who are frequent GP users have a 38% lower risk of hospitalisation compared with carers who infrequently utilise their GP.

Interpretation of main findings

The main finding of increased GP interaction reducing the risk of hospitalisation aligns with studies undertaken across diverse patient groups and in other countries. For example, among people aged ≥45 years in China, those with primary care outpatient visits were less likely to be hospitalised (Yang et al. 2022), with a similar finding being found among German chronically ill patients (Sawicki et al. 2021). Studying a cohort of 1.5 million Australians, Bazemore et al. (2018) found that increased general practice continuity of care with primary care, a metric that takes into account frequency of visits, was associated with reduced hospitalisations. This association was also found in a 2013 British study that recommended increased continuity of GP care as a solution to reducing the burden on the wider hospital system (Hansen et al. 2013).

Possible explanations for the relationship between increased GP interaction and reduced risk of hospitalisation in the cohort could be explained by the fact that high GP users have consistent communication with their GPs, which facilitates early detection and management of potential health concerns before they escalate to the point where hospitalisation is required (Einarsdóttir et al. 2010). Comparatively, when individuals who do not see their GP as regularly experience escalating medical, emotional or social burdens, delayed intervention may culminate in a need for hospitalisation (Weissman et al. 1991; Sankey et al. 2016). Alternatively, it could be that individuals who are healthier or have greater health literacy are self-selecting to use GPs more (Shahid et al. 2022), hence confounding the association with decreased hospitalisation. Nevertheless, this limitation has been partially mitigated by including overall health and education level demographics in the adjusted model, although it is important to note that health literacy is not always correlative with education level (Shahid et al. 2022).

The characteristics of our study’s cohort also reveal much about the demographics of carers. The increase in the percentage of high GP users from 28.3% in 60–74-year-olds to 46.3% in carers aged >75 years can be explained by a higher percentage of older people visiting their GP generally (Australian Institute of Health and Welfare 2023). As shown in Table 1, the proportion of participants who were high GP users increased as self-rated health and quality of life decreased, possibly as a result of those who are more conscious of their needs seeking support more regularly. These findings suggest the potential need for tailored interventions for carers that highlight the potential advantages of engaging with GPs. Additionally, the fact that there were over twice the number of female than male carers aligns with the finding that women are more likely to take on carer responsibilities compared with men (Australian Bureau of Statistics 2018; Barr et al. 2020).

Justification of study techniques, strengths and weaknesses

The study employed a retrospective approach to analyse the relationship between GP visits and hospitalisation rates among carers, leveraging a large dataset (26,004 participants) from the 45 and Up Study. This method provided a comprehensive view of long-term outcomes and aligned with established research methodologies. The use of a longitudinal linkage resource was beneficial, as it included both self-identified carers and those not specifically self-labelled as such, enhancing the study’s inclusivity. Additionally, the study cohort was recruited from the community while healthy, allowing it to be more representative of the wider population.

A significant strength of the study was its potential to suggest causation, not just association, between frequent GP visits and reduced hospitalisations, supported by compelling relative risk figures (crude RR 0.51 and adjusted RR 0.62). The research also paid careful attention to temporality, ensuring GP interactions preceded hospitalisations, which is crucial for suggesting a causal relationship. However, the binary categorisation of GP users might have oversimplified the nuanced relationship between GP visits and hospitalisation risks. Future research could benefit from randomised controlled trials to further validate these findings (Farmer et al. 2018; Hariton and Locascio 2018).

Despite its strengths, the study faced limitations. The reliance on pre-existing cohort data meant the research was constrained by the scope of the original data collection, potentially overlooking unknown confounders. The MBS data used lacked detailed information about the GP visits, limiting the depth of analysis regarding the nature of these interactions. Additionally, although the 45 and Up Study data generally represented the demographic well, issues, such as low response rates and oversampling of certain groups (e.g. rural participants (Banks et al. 2008)), could introduce biases. The potential for participants to transition in and out of the carer role also posed a risk for nondifferential misclassification bias (Mitchell et al. 2018), suggesting that the findings should be interpreted as a cautious approximation of the actual impact of GP use on hospitalisation rates among carers.

Implications for research and practice

This study highlighted critical aspects of carer support, particularly the link between carer–GP interactions and hospitalisations. However, it left open questions about what type of these interactions confers the most benefits. It is also unclear whether the frequency, regularity or specific characteristics of GP or carer interactions are crucial in reducing hospitalisations. Future research should explore these nuances and the dynamics of carers transitioning in and out of their roles, as these periods may require tailored GP support.

Despite these uncertainties, the study’s implications for carers, GPs and policymakers are significant. Reducing hospitalisations is crucial from medical, societal and economic perspectives (Langa 2009; Productivity Commission 2011; Bastawrous 2013; Pristavec 2019). The study suggests that promoting GP engagement among carers could be an effective preventive strategy to reduce hospital burden. GPs should recognise their patients who are carers and emphasise regular check-ins to mitigate the risk of hospitalisations. Policymakers, recognising the economic importance of minimising tertiary healthcare admissions (Department of Health and Aged Care 2023), should explore strategies to encourage GP use among carers. This confers a twofold benefit for healthcare burden, whereby carers have a reduced likelihood of being hospitalised, while being able to continue their vital roles, preventing the need for institutional or hospital care for their dependents (Jiwa et al. 2002).

Interventions to enhance GP interaction with carers vary in complexity. Identifying carers within a GP’s practice has shown promise in the UK (Carduff et al. 2016). GPs acknowledge the need for specialised training to better support carers, which could improve care continuity and health follow ups (Wangler and Jansky 2021; Cronin and McGilloway 2023). Enhancing communication tools to facilitate carer access to GPs is another area with potential (Chi and Demiris 2015), although research and clinical trials in this domain are ongoing.