Health and health service use of Australian female live-in caregivers at multiple caregiver trajectory timepoints: a population-based dataset with time-matched controls Authors: Leigh Tooth, Richard Hockey, Gita D Mishra Published: 2026 ABSTRACT AbstractObjectiveTo compare health and health service use in female live-in caregivers along multiple timepoints with population-based controls over a 20-year period.MethodsData from 6000 women aged 50–77 caring for someone who was ill, disabled or frail, were collected over eight waves of Australian Longitudinal Study on Women’s Health 1946–51 cohort (from 2001 to 2022). Physical functioning (PF), general mental health (MH), depressive symptoms, stress and general practitioner (GP) and medical specialist visits were compared between caregivers and time-matched controls, before, during and after caregiving. Fully adjusted mixed linear models or logistic regression, including sensitivity analyses by caregiving intensity, were conducted.ResultsIn adjusted results, caregivers and controls had similar initial levels of PF, but caregivers had significantly poorer PF over time including during [2.39 (−3.68, −1.10)] and after stopping caregiving [−2.19 (−3.51, −0.87)]. Caregivers had poorer MH on all three measures [eg, −5.56 (−6.61, −4.51) on SF-36 MH] and more GP visits across the study than controls, with the poorest scores in the period before stopping caregiving. Accessing medical specialists did not differ between caregivers and controls until after stopping caregiving, when caregivers had higher odds [1.22 (1.06, 1.40)]. Associations were stronger in caregivers providing higher intensity care.ConclusionsCaregiver physical and MH was poorest and GP visits highest prior to stopping caregiving, while specialist visits increased after stopping caregiving. Interventions to assist caregivers manage their own health needs, particularly in the period before stopping caregiving, and assisting them maintain health service use throughout and following caregiving cessation should be tested.Keywords: caregiver, health, services, physical, mental, older people Key Points Caregiver mental health trajectories were consistently poorer than controls, even prior to caregiving. Caregiver mental health trajectories were consistently poorer than controls, even prior to caregiving. Caregiver physical functioning trajectories, while initially similar to controls, worsened over time and remained poorer. Caregiver physical functioning trajectories, while initially similar to controls, worsened over time and remained poorer. Mental health and physical functioning outcomes were poorest in the time period before caregiving ended. Mental health and physical functioning outcomes were poorest in the time period before caregiving ended. Caregiver general practitioner use was higher than controls, even after ceasing caregiving. Caregiver general practitioner use was higher than controls, even after ceasing caregiving. Specialist doctor use was equivalent in caregivers and controls, only increasing for caregivers once caregiving ceased. Specialist doctor use was equivalent in caregivers and controls, only increasing for caregivers once caregiving ceased. Introduction In Australia in 2022, around 13% of women and 11% of men provided unpaid, informal care to older people or those with a disability [1]. While men were more likely to provide informal care at younger and older ages (65 years plus), women provided more care throughout mid-life and provided the most intensive (or primary) informal care compared to men at all ages. These prevalences and patterns of informal care are similar to those in the USA [2], the UK [3], Canada [4] and countries in Europe [5, 6]. Both the negative and positive outcomes of providing informal care are well documented. Negative associations include poorer mental health [7–12], physical health [11, 13–15], quality of life [16], and life satisfaction [9], and higher stress [11, 17, 18] and health service use [19–21]. Positive associations include resilience [22], higher life satisfaction [16] and better mental health (MH) [7, 16]. It is also known that caring for someone with dementia, versus other conditions, may be associated with higher stress and burden, due to increased care hours and behavioural challenges [23], and restrictions on caregiver social participation [24]. When caregiving results are analysed by sex, women and men often show differing associations [12, 25–28]. Women generally have poorer outcomes, possibly reflecting the more intensive caregiving and higher burden they experience as primary or live-in caregivers [14, 29–31]. However, negative associations between caregiving and caregiver’s health may reflect the study research design characteristics rather than from caregiving itself [11]. This may also be evidenced in papers showing no differences between caregivers and non-caregivers in outcomes like self-rated health [10], physical function [10, 32] or cognition [10]. Roth et al. [11] recommended caregiving research include large, population-based samples of individuals with health and well-being data collected before and after caregiving, with time-matched control groups from the same population, ensuring caregiving samples are those with sufficient exposure to potential caregiving stressors and capturing individual differences in caregiving experiences like type of care provided and care recipient conditions. Another limitation of caregiving research is only capturing one or two timepoints, typically after starting and/or after stopping caregiving, over a few years or less [8, 11, 18, 27, 33]. To truly understand associations of caregiving with caregiver health, looking beyond just starting and stopping is important. For example, while assessment before starting and after stopping might show caregiver health gradually declined, it would fail to document individual fluctuations throughout caregiving. Identifying the timing of individual fluctuations and associated factors may enable more optimised decision making about the timing of interventions to support caregivers maintain their health. Expanding the range of caregiver outcomes to include physical health and health service use is also important to gain a fuller picture of the association of caregiving with outcomes over time and potentially inform better health service provision. Our aim was to compare the health and health service use in caregivers of people with long-term illness, disability or frailty with a time (survey) matched control group of non-caregivers from the same population. The groups were compared at specific points along the caregiver trajectory, namely before starting, during and after stopping caregiving. We utilised a longitudinal dataset with over 20 years of data and focussed only on live-in caregivers, as they experience higher caregiver burden and have poorer outcomes [7, 9, 31, 34–37]. Materials and methods Participants Data come from the population-based Australian Longitudinal Study on Women’s Health (ALSWH). In 1996, self-reported health, health service use and socio-demographic data were collected from over 41 500 women in three cohorts: those born 1973–78 (aged 18–23 in 1996), those born 1946–51 (aged 45–50) and those born 1921–26 (aged 70–75). The participants were selected randomly from the Medicare Australia database, which covers all citizens and permanent residents of Australia. Women living in rural and remote areas were sampled at twice the rate of women living in urban areas to ensure continued representation of women living outside major centres over the study. ALSWH data are linked to Medicare Benefits Schedule (MBS) data which include healthcare services subsidised under Medicare (Australia’s universal health insurance system). Details of recruitment and estimated initial response rates are published elsewhere [38]. Since 1996, each cohort has been re-surveyed approximately every 3 years. Participants provided informed consent at each survey, including to linked MBS data, with ethical clearances obtained from the Human Research Ethics Committees at the University of Queensland (approval number 2004000224) and the University of Newcastle (H-076-0795). The present study includes the 1946–51 cohort who have been surveyed 10 times since 1996. The sample size in Survey 3 was N = 11,226 and in Survey 10 was 7159 (see Appendix Figure 1 for participant numbers and response rates at each survey). Measures Caregiving exposure Women were asked ‘Do you regularly provide care or assistance (e.g. personal care, transport) to any other person because of their long-term illness, disability or frailty?’ at each survey. For this analysis, data on caregiving were used from Surveys 3 to 10 because Survey 1 did not ask about live-in caregiving, and the data were not complete in Survey 2. Response options to the caregiving question in Surveys 3 to 10 were ‘Yes, for someone who lives with me’; ‘Yes, for someone who lives elsewhere’ or ‘No, I do not provide care’. For this analysis, data were coded at each survey as (1) the women provided live-in caregiving, and (2) they did not provide caregiving. Women providing caregiving for someone who did not live with them were excluded from the analysis as previous research shows this group are very heterogenous with similar health outcomes to non-caregivers [36]. Care intensity A care intensity index that captured key issues underpinning caregiving context [39] was created by combining responses to caregiving frequency (from daily to every few weeks) and duration (from all day/night to about an hour at a time) [36]. Resultant index categories were low (caregiving once/week for 1 h at a time; or caregiving every few weeks for several hours at a time), medium (caregiving several times/week for several hours at a time; or caregiving every day for about an hour) or high (caregiving all day, several times a week or caregiving every day for several hours at a time). Health-related quality of life and health service use outcomes Physical health-related quality of life was measured using the physical functioning (PF) subscale from the Short Form-36 (SF-36) [40]. A weighted sum of responses to items was calculated ranging from 0 (worst physical health-related quality of life) to 100 (best physical health-related quality of life). MH was measured using two scales. General MH-related quality of life was measured with the MH subscale from the SF-36. The MH subscale has 5-items measuring anxiety, depression, emotional control and psychological health with similar scoring (0–100) as described above. Depressive symptoms were measured using the Centre for Epidemiological Studies Depression scale (CESD-10) [41] with higher scores indicating more depressive symptoms and a cut-off score of ≥10 indicating possible depressive illness. Stress about own health, health of family members, work, living arrangements, study, money and relationships was measured with the Perceived Stress Scale, which has been validated with the ALSWH 1946–51 cohort [42]. Scores range from 0 (not at all stressed) to 4 (extremely stressed). Health service use data were number of general practitioner (GP) and medical specialist (e.g. cardiologist, psychiatrist) visits in the previous 12 months, obtained from linked MBS data. Covariates Demographic, socioeconomic status and health covariates were chosen a priori [14, 43] and included relationship status, country of birth, managing on income, education and chronic conditions (Table 1 shows their measurements). Table 1Descriptive characteristics of women in the 1946–51 cohort of the Australian Longitudinal Study on Women’s Health at Survey 3 (in 2001) and at Survey 10 (2022) by caregiving statusSurvey 3 (aged 50–55)Survey 10 (aged 71–77)Characteristic, n (%)Provide live-in care (N = 722)Do not provide live-in care (N = 8411)Provide live-in care (N = 822)Do not provide live-in care (N = 5518) Demographic Relationship status Married/defacto578 (80.2)6846 (81.8)704 (86.6)3527 (64.3) Single/never married37 (5.1)231 (2.8)11 (1.4)144 (2.6) Widowed21 (2.9)244 (2.9)53 (6.5)984 (18.0) Divorced/separated85 (11.8)1043 (12.5)45 (5.5)827 (15.1) Ability to manage on available income Easy71 (9.9)1555 (18.8)168 (20.6)1562 (28.5) Not too bad261 (36.4)3630 (43.8)396 (48.6)2696 (49.2) Sometimes difficult261 (36.4)2202 (26.6)167 (20.5)857 (15.6) Always difficult/impossible125 (17.4)903 (10.9)84 (10.3)361 (6.6) Highest educational qualification Year 12 or below493 (68.8)5429 (65.1)467 (56.8)2834 (51.4) Trade/certificate/diploma138 (19.2)1626 (19.5)204 (24.8)1355 (24.6) Degree/higher degree86 (12.0)1284 (15.4)151 (18.4)1325 (24.0) Country of birth Australia553 (77.9)6346 (76.4)631 (77.6)4270 (77.9) Other English speaking88 (12.4)1203 (14.5)121 (14.9)798 (14.6) Other69 (9.7)760 (9.2)61 (7.5)413 (7.5) Number of chronic conditions 0351 (48.6)6063 (57.9)201 (24.5)1555 (25.4) 1261 (36.2)3328 (31.8)417 (50.8)3015 (49.2) 2+110 (15.2)1073 (10.2)203 (24.7)1554 (25.4) Study outcomes SF-36 physical functioning score, mean (SD)78.3 (21.1)82.2 (19.7)69.1 (24.4)71.1 (24.1) SF-36 general mental health score, mean (SD)69.2 (19.3)74.2 (18.0)73.8 (16.8)78.0 (16.2) CESD-10 depressive symptoms score, n (%) ≥10206 (30.4)1688 (21.0)166 (20.5)866 (16.0) <10473 (69.7)6342 (79.0)642 (79.5)4545 (84.0) Perceived stress scale score, mean (SD)0.93 (0.68)0.68 (0.57)0.74 (0.58)0.49 (0.49) Number of general practitioner visits in last 12 months, mean (SD)6.8 (6.4)5.5 (5.8)12.4 (8.2)10.9 (6.9) Any medical specialist visit in last 12 months, n (%)306 (44.9)4019 (40.5)487 (62.6)3747 (63.9) Note. SF-36, Short Form-36 Item General Health Survey; CESD-10, Centre for Epidemiological Studies Depression scale; Other Country of birth category included Asia, Europe, other; Chronic conditions included diabetes, heart disease, hypertension, stroke, asthma and breast cancer. Descriptive characteristics of women in the 1946–51 cohort of the Australian Longitudinal Study on Women’s Health at Survey 3 (in 2001) and at Survey 10 (2022) by caregiving status Note. SF-36, Short Form-36 Item General Health Survey; CESD-10, Centre for Epidemiological Studies Depression scale; Other Country of birth category included Asia, Europe, other; Chronic conditions included diabetes, heart disease, hypertension, stroke, asthma and breast cancer. Statistical analysis Caregiver trajectories were determined deterministically using all available data from each survey between Surveys 3 and 10 the women completed. The following caregiving time points along these trajectories were determined: One to three years before starting caregiving (data from survey before they reported caregiving) One to three years before starting caregiving (data from survey before they reported caregiving) During caregiving (data from the first and last surveys they reported caregiving) During caregiving (data from the first and last surveys they reported caregiving) One to three years after stopping caregiving (data from survey they reported not caregiving). One to three years after stopping caregiving (data from survey they reported not caregiving). A control group of women who never reported caregiving from Surveys 3 to 10 were randomly matched to caregivers at the survey time points, using a 2:1 ratio of controls:caregivers. Each survey covers a narrow age range (5 years), so additional matching on age was unnecessary. To address the aim, both within- and between-group analyses of caregivers and controls were conducted. For within-group analyses of PF, MH, CESD-10, stress and GP visits, repeated measures mixed linear models were used to compare longitudinal changes before starting to during (after starting) and from during (before stopping) to after stopping caregiving for each outcome separately for caregivers and controls. For medical specialist visits a binary variable (no visits; any visits in a 12-month period) was created as around 50% of women had no visits. A GEE repeated measures model was used to obtain adjusted ORs comparing before and after entering or leaving caregiving for caregivers and controls. In sensitivity analyses, the above analyses were conducted stratified by care intensity (low/medium/high). For between-group analyses, differences between caregivers and controls at each of the time points were determined using fully adjusted mixed models. All mixed model analyses used covariate data collected at the same timepoints described above for the caregiving exposure. In further sensitivity analyses, average outcome scores across the number of surveys women completed before starting and after stopping caregiving, rather than just at the survey before or after, were examined. SAS 9.4 was used for all statistical analyses (see Appendix for expanded methodological details). Results Table 1 shows the demographic, socioeconomic and health characteristics of ALSWH women at Surveys 3 and 10. Caregiving prevalence ranged from 6% in 2001 to 11% in 2022. Most women were partnered, Australian born and had a high school level education or below. The ability to manage on their income was poorer at Survey 3 but improved by Survey 10. SF-36 PF scores declined with age while scores on the MH scales showed improvement or stability with age (Table 1). Use of GPs and medical specialists increased with age. Within-group analyses—caregivers and controls For both caregivers and controls, Figure 1 shows adjusted mean SF-36 PF (1a) and GP visits (1b) and odds of medical specialist visits (1c). Figure 2 shows mean SF-36 MH (2a), CESD-10 depressive symptoms (2b) and perceived stress (2c). Table 2 shows the adjusted within-group results of the change from before to during (after starting) and during (before stopping) to after stopping caregiving for each outcome (Appendix Table 1 and Appendix Fig. 2a–f show the unadjusted results). Physical functioning and health service use outcomes for caregivers and matched controls: (a) adjusted mean SF-36 physical functioning scores (95% CI); (b) adjusted mean general practitioner (GP) visits (95% CI); and (c) odds of medical specialist visits (95% CI). Mental health outcomes for caregivers and matched controls (a) adjusted mean SF-36 mental health scores (95% CI); (b) adjusted mean CESD-10 depressive symptom scores (95% CI); and (c) adjusted mean perceived stress scores (95% CI). Table 2Fully adjusted mixed-effect model within-group results of caregiving trajectory time points on health outcomes, with results also presented for randomly matched controlsOutcomeCaregiving statusChange from before to during caregiving (b, 95% CI, P) (N = 6171)Change from during to after stopping caregiving (b, 95% CI, P) (N = 4410)Physical functioningCaregiver−2.31 (−3.13, −1.50)****−1.55 (−2.55, −0.56)**Control−2.13 (−2.75, −1.51)****−1.75 (−2.46, −1.04)****General mental healthCaregiver−3.41 (−4.13, −2.68)****2.76 (1.87, 3.66)****Control−0.67 (−1.22, −0.13)*−0.52 (−1.15, 0.11)Depressive symptomsCaregiver0.96 (0.75, 1.17)****−0.67 (−0.93, −0.40)****Control0.28 (0.12, 0.44)***0.31 (0.12, 0.50)**Perceived stressCaregiver0.18 (0.15, 0.20)****−0.13 (−0.16, −0.10)****Control0.01 (−0.01, 0.03)0.02 (−0.01, 0.04)Number of general practitioner visitsCaregiver0.58 (0.29, 0.85)****0.30 (−0.03, 0.64)Control0.44 (0.23, 0.66)****0.29 (0.05, 0.52)*Any medical specialist visit (OR)Caregiver1.05 (0.93, 1.18)1.46 (1.27, 1.68)****Control1.09 (1.01, 1.19)*1.19 (1.08, 1.32)*** Note. All analyses adjusted for manage on income, highest educational qualification, country of birth, relationship status, number of chronic conditions and survey time; Linear mixed models used for physical functioning, all mental health measures and general practitioner visits; Logistic regression with OR (odds ratio) used for medical specialists. * P < .05; **P < .01; ***P < .001; ****P < .0001. Fully adjusted mixed-effect model within-group results of caregiving trajectory time points on health outcomes, with results also presented for randomly matched controls Note. All analyses adjusted for manage on income, highest educational qualification, country of birth, relationship status, number of chronic conditions and survey time; Linear mixed models used for physical functioning, all mental health measures and general practitioner visits; Logistic regression with OR (odds ratio) used for medical specialists. * P < .05; **P < .01; ***P < .001; ****P < .0001. Caregivers had significantly poorer PF and general MH, more depressive symptoms and higher perceived stress from before to during caregiving as well as significantly improved general MH, less depressive symptoms and less perceived stress from during to after stopping caregiving (Table 2). Figure 2 shows the clear trend of better MH during the 1–3 years after stopping caregiving. From before to during caregiving, women showed a higher number of GP visits, with no further change from during to after stopping caregiving (Table 2). Women did not show any change in the number of medical specialist visits from before to during caregiving, but odds of visits increased from during to after stopping caregiving (Table 2). The first sensitivity analysis confirmed that all the associations found above were generally strongest in caregivers providing high or medium intensity care (Appendix Table 2). Controls showed better general MH, depressive symptoms and perceived stress across the study period (Fig. 2). They showed a similar initial decline in PF, but unlike caregivers, showed an improvement in PF over time before showing another decline. Controls showed a similar increase in pattern of visits to GPs and medical specialists across the study to caregivers. The second sensitivity analyses that used average outcome scores across the number of surveys caregivers and controls completed over the study period showed very similar results to those in Table 2 (data on request). Between-group analyses In adjusted results, caregivers and controls had similar initial levels of PF, but caregivers had significantly poorer PF over time including during [−2.39 (−3.68, −1.10)] and after stopping caregiving [−2.19 (−3.51, −0.87)] (Table 3). Caregivers had poorer MH on all three measures with the poorest scores shown in the period before stopping caregiving. Caregivers had higher odds of GP visits across the study period than controls. There were no differences in odds of medical specialist visits between caregivers and controls until after stopping caregiving, when the caregivers had 22% higher odds (Table 3). Table 3Fully adjusted mixed-effect model between-group results of caregivers versus matched controls on health outcomes at the caregiving trajectory time pointsOutcomeBefore starting caregiving (mean difference, 95% CI, P) (N = 6171)During caregivingAfter stopping caregiving (mean difference, 95% CI, P) (N = 4410)Survey after starting caregiving (mean difference, 95% CI, P) (N = 6171)Survey before stopping caregiving (mean difference, 95% CI, P) (N = 4410)Physical functioning0.37 (−0.74, 1.48)0.18 (−0.93, 1.30)−2.39 (−3.68, −1.10)***−2.19 (−3.51, −0.87)***General mental health−1.24 (−2.14, −0.33)**−3.97 (−4.88, −3.06)****−5.56 (−6.61, −4.51)****−2.27 (−3.36, −1.19)****Depressive symptoms0.44 (0.18, 0.71)***1.13 (0.86, 1.39)****1.78 (1.47, 2.09)****0.80 (0.48, 1.12)****Perceived stress0.12 (0.09, 0.15)****0.28 (0.25, 0.31)****0.32 (0.28, 0.35)****0.17 (0.13, 0.21)****Number of general practitioner visits0.26 (−0.08, 0.61)*0.39 (0.05, 0.74)*0.78 (0.37, 1.18)****0.79 (0.38, 1.20)***Any medical specialist visit (OR)0.94 (0.84, 1.06)0.90 (0.80, 1.01)0.99 (0.87, 1.13)1.22 (1.06, 1.40)** Note. Reference group = controls; All analyses adjusted for manage on income, highest educational qualification, country of birth, marital status, number of chronic conditions and survey time; Linear mixed models used for physical functioning, all mental health measures and general practitioner visits; Logistic regression with OR (odds ratio) used for medical specialists. * P < .05; **P < .01; ***P < .001; ****P < .0001. Fully adjusted mixed-effect model between-group results of caregivers versus matched controls on health outcomes at the caregiving trajectory time points Note. Reference group = controls; All analyses adjusted for manage on income, highest educational qualification, country of birth, marital status, number of chronic conditions and survey time; Linear mixed models used for physical functioning, all mental health measures and general practitioner visits; Logistic regression with OR (odds ratio) used for medical specialists. * P < .05; **P < .01; ***P < .001; ****P < .0001. Discussion The innovative approach of this study was to examine caregiver health as a function of transitioning into and out of caregiving, including in the period before caregiving stopped. On all health and health service use outcomes, caregivers had poorer outcomes towards the end of the caregiving period compared with controls, with the poorest results for caregivers providing higher intensity care. While other authors have examined starting and stopping caregiving, they could not show fluctuating health outcome patterns over time as they did not measure multiple time points [8, 11, 13, 15, 17, 31, 37]. Our study yielded several novel findings. First, that the caregiver general MH, depressive symptoms and stress declined compared to controls, with the poorest MH outcomes in the period before stopping caregiving. Further, while the decline in MH lessened only after the women exited caregiving, it still remained poorer than controls. Some authors have shown MH to improve following caregiving [8, 17, 25], while others have shown MH to remain poorer [30]. Our finding of the distinctly poorer MH prior to stopping caregiving may reflect our unique sample, including that the women had poorer general MH, more depressive symptoms and higher stress than controls, even before caregiving [29, 30]. It may also reflect the relationship between the caregiver and care recipient and reason for stopping caregiving [44]. Second, while caregivers and controls initially had similar PF, caregivers had poorer PF during caregiving, which persisted after stopping caregiving. This may indicate the PF decline occurred greater than that expected by age alone [11]. While others have shown that caregivers exhibit a faster or greater decline in physical health or functioning versus controls [13, 15, 31], these studies did not document whether the decline persisted after stopping caregiving. Third, compared with controls, caregivers had more GP visits throughout caregiving, but particularly before and after stopping. In contrast, caregiver access to medical specialists only increased after stopping caregiving (22% higher odds). It is challenging to compare these results as few papers using population-based samples of caregivers and controls have examined use of GP or medical specialist services. One study [21] found increased total Medicare expenditure by female primary caregivers up to 2 years following their spouse’s death. Comparison with our study is difficult as we did not capture whether ceasing caregiving was due to bereavement. Our results showed the poorest MH and PF scores matched the period when GP visits were highest, possibly reflecting that caregivers were accessing the GP services they needed throughout and after caregiving due to their poorer mental and physical health [45]. Caregivers may have delayed accessing medical specialists until after caregiving ended for a few reasons. They may have overlooked specific health or preventive care needs requiring specialist attention due to caregiving demands or prioritising the health of the care recipient, resulting in the worsening of their own health [46]. They may also have been under financial stress from caregiving with less resources to access medical specialists [47]. In Australia, while GP visits are largely bulk billed, ensuring no or minimal cost to patients, medical specialist visits are not usually bulk billed resulting in potentially large out-of-pocket expenses [48]. The provision of targeted support programmes or interventions, e.g. including education, counselling and physical activity and focussing on caregiver mental and physical health have been proposed to assist caregivers manage their health during and following caregiving. Yet, such interventions, e.g. using telephone or web-based formats, have shown largely null or small effects on mental and physical health and functioning [49, 50] or health service use [51]. Other options could include government payments to assist caregivers access health services to maintain optimal health [47], with relevance to the economic burden of caregiving and out-of-pocket costs of medical specialists. Research should focus on better understanding the needs of caregivers across the caregiving trajectory and identifying the best type of interventions to assist caregivers’ management of their own health and health service needs, particularly in the period before and after stopping caregiving. Strengths of the study were the large, population-based sample with the caregivers and controls from the same population, including multiple caregiving trajectory time points, exploring differences by caregiving intensity and obtaining health service data from linked administrative databases. The 20-year study period enabled examination of the findings against naturally occurring changes in women’s health when aged from their 50s to late 70s. Limitations include that as ALSWH is not a specific study of caregiving some relevant factors, such as detail about care recipients, or caregiver choice, were not fully captured. It is known that caring for people with different conditions, e.g. dementia versus cancer, may be associated with differing caregiver health outcomes [14, 17, 23]. Care recipient conditions were only captured halfway through the study, precluding their use in sensitivity analyses. While missingness due to overall attrition does impact the validity of the findings, effective retention was 82%. Finally, while a population-based study, the ALSWH 1946–51 baseline cohort had slightly more women with a tertiary level education and who were Australian born. Overtime, the women have become slightly less representative of the wider population due to migration [38]. Conclusion Caregivers showed differing trends of poorer physical and mental health and health service use before, during and after caregiving versus controls. The examination of more caregiver trajectory time points, including the period before stopping caregiving, yielded insights into caregiver health and health service use outcomes above those occurring by age alone. Interventions to assist caregivers manage their own health needs, particularly in the period before stopping caregiving, and assisting them maintain health service use throughout and following caregiving cessation should be tested. Supplementary Material