Years gained, health lost? A population-based cohort study of sex differences in healthy life expectancy among people living with HIV

Abstract

Objectives

To assess sex differences in three measures of healthy life expectancy (HLE) among people living with HIV (PLWH) in British Columbia (BC), Canada, between 1996 and 2020.

Design

Retrospective population-based cohort study.

Setting

BC, Canada, using the Comparative Outcomes and Service Utilisation Trends (COAST) cohort, derived from linked population-based administrative health datasets within a universal healthcare system.

Participants

All known PLWH identified in COAST between 1996 and 2020. Of these, 2511 (17.8%) females and 11 624 (82.2%) males met inclusion criteria of age ≥20 years with ≥1 day of follow-up.

Outcome measures

HLE at ages 20, 40 and 55, estimated using Sullivan’s abridged life table approach, integrating age-specific mortality rates with health condition prevalence. HLE was measured in three ways: age-related condition-free life expectancy (LE), multimorbidity-free (≥3 conditions) LE and mental health condition-free LE.

Results

All three HLE measures increased over time but remained higher among males than females across nearly all comparisons. Years lived with ≥1 condition, multimorbidity and mental health conditions also increased over time. In the most recent period (2012–2020), age-related condition-free LE was higher among males than females by 4.2, 2.4 and 0.2 years; multimorbidity-free LE by 7.2, 5.6 and 3.6 years; and mental health condition-free LE by 10.0, 6.8 and 4.2 years at ages 20, 40 and 55, respectively.

Conclusions

Among PLWH within a universal healthcare system, gains in longevity have been accompanied by an increasing burden of chronic and mental health conditions. Females experience shorter HLE than males and lose more years to age-related multimorbidity and mental health conditions. These findings highlight the need for sex-responsive strategies to prevent and manage chronic and mental health conditions to ensure gains in LE translate into equitable healthy ageing among PLWH.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• We used the Comparative Outcomes and Service Utilisation Trends study, a population-based cohort including virtually all known people living with HIV in British Columbia, Canada.

• The study sample is identified through administrative health data from 1996 to 2020, with linkage across multiple population-level databases enabling comprehensive ascertainment of sex stratified mortality and health conditions over time.

• Healthy life expectancy was estimated using Sullivan’s abridged life table approach, integrating age-specific mortality with condition prevalence across three different treatment eras.

• Health conditions and disease prevalence was estimated using administrative data and may be subject to misclassification, changes in coding practices over time and related biases such as survivor bias.

Introduction

Life expectancy (LE) among people living with HIV (PLWH) has substantially increased over the last 30 years, largely due to the introduction of combination antiretroviral therapy (cART) and treatment as prevention strategies.^1–3 However, females living with HIV experience smaller LE gains than males, and this gap has widened over time.^{2 4} Between 2012 and 2020 in British Columbia (BC), Canada, LE at age 20 among PLWH was 48.0 years for males and 40.9 years for females, reflecting an increase of 23.5 and 18.8 years, respectively, compared with the period 1996–2001.² These findings reveal persistent sex differences in survival gains among PLWH in a setting with universal access to HIV care and treatment. While LE is an important indicator of population health, it provides limited insight into the population morbidity burden. It remains unclear whether gains in longevity have translated into equivalent gains in healthy LE (HLE) among PLWH, whether this differs by sex and how this has changed over time.

Among PLWH, females experience higher rates of several non-AIDS defining comorbidities, mortality and hospitalisation than males.^5–7 In BC, females have a 1.7-fold higher hospitalisation rate and more than twofold higher hazard of mortality from renal, liver and chronic respiratory disease compared with males, even after adjusting for several socio-structural and clinical factors.^{2 7} These disparities in risk profiles may partly reflect biological differences in sex hormones, reproductive ageing and immune regulation.^{5 8} For example, females show higher systemic immune activation and elevated monocyte activation markers linked to cardiovascular risk.⁸ Broader socio-structural factors such as unstable housing, poverty, stigma and barriers to care also likely contribute to sex differences in health outcomes among PLWH: these factors can hinder access to care and increase stress, raising risks of depression, anxiety and other comorbidities.^{5 9} Sex differences in the prevalence of substance use among PLWH may further shape comorbidity patterns and health outcomes.^{10 11}

Despite observed sex differences in LE and morbidity, less is known about how long females and males living with HIV can expect to live in good health, or how these patterns have changed in the modern treatment era in a setting with universal healthcare. HLE quantifies the number of years an individual can expect to live in good health by considering both the length of life and the quality of health during those years.^{12 13} By estimating HLE for females and males at ages 20, 40 and 55 across three cART eras (1996–2001, 2002–2011, 2012–2020), this study captures sex differences in healthy ageing over time among PLWH in BC. We assessed three HLE concepts: (1) age-related condition-free LE, (2) multimorbidity ($\ge 3$ age-related conditions) free LE and (3) mental health condition-free LE. We hypothesised that females living with HIV would have shorter HLE and spend more of their lives in an unhealthy state than males, with the latter increasing over time.

Methods

Study design

The Comparative Outcomes And Service Utilisation Trends (COAST) study is a population-based cohort study, designed to characterise and compare health outcomes and healthcare utilisation of PLWH and the general population in BC.¹⁴ COAST consists of a linkage between BC Centre for Excellence in HIV/AIDS’s Drug Treatment Program (DTP) registry and administrative healthcare data provided by Population Data BC. The DTP centrally manages distribution of antiretroviral therapy (ART) to all medically eligible known PLWH in BC.¹⁵ Through a linkage to laboratory data, individuals with a detectable HIV plasma load (and no known ART exposure) are also included in the DTP database, which contains data on ART and HIV disease progression. Population Data BC is a provincial organisation that provides academic researchers with access to a provincial repository of population-wide individual-level administrative data. The following datasets were used in this study: the Consolidation File, containing basic demographics, geographical information and census data; Medical Service Plan Payment Information (MSP) dataset, containing billing information on medically necessary services provided by healthcare practitioners; Discharge Abstract Database (DAD), containing administrative clinical and demographic information capturing hospital discharges, transfers and deaths at acute care hospitals; PharmaNet, containing data on prescription drug dispensations from community and outpatient pharmacies; BC Cancer Registry, including cancer diagnoses; and the Vital Events and Statistics Deaths database, including all deaths registered in BC (https://my.popdata.bc.ca/project_listings/18-223/).¹⁶

Study population

We included all PLWH in COAST aged ≥20 years with at least 1 day of follow-up from 1 April 1996 until 31 March 2020. Baseline was defined as the latest of: (1) first record of HIV positive status (either first ART dispensation for HIV treatment or first detectable HIV plasma viral load measurement), (2) 20th birthday and (3) 1 April 1996. Individuals were followed until the first of: (1) death, (2) end of follow-up (latter of last date of registration in BC’s universal insurance programme or last healthcare contact date) and (3) 31 March 2020.

To capture key changes in ART treatment availability and guidelines over time we distinguished three time periods: the early cART era (1 April 1996–31 December 2001 (1996–2001)), characterised by availability of the first highly active ART; modern cART era (1 January 2002–31 December 2011 (2002–2011)), characterised by growing availability of better tolerated antiretroviral drugs, including the first one-pill regimen, but also by guidelines recommending more restrictive treatment initiation; and recent cART era (1 January 2012–31 March 2020 (2012–2020)), during which guidelines recommended cART initiation for all PLWH regardless of CD4+ T cell count (CD4 count) and treatment as prevention became the standard of care throughout BC: province-wide expansion of the Seek and Treat for Optimal Prevention of HIV/AIDS Programme resulted in intensified HIV testing, treatment and engagement in care.^{3 17 18}

Descriptive variables

The study population was stratified by sex, a time-fixed variable derived from the DTP database. While we recognise that both factors related to sex, a biological construct and gender, a social construct, may underlie observed differences in health and HLE, we were only able to assess differences between males and females as gender data were not systematically collected.¹⁹

History of injection drug use (IDU), historical CD4 counts and ART history were derived from the DTP database. Residence location was derived using a combination of address information from the DTP database and location information from the Consolidation File. Rurality was defined using the Statistical Area Classification Type classification of census subdivisions.²⁰ Residence in the Downtown Eastside, a neighbourhood in Vancouver’s inner-city characterised by complex problems including a high prevalence of drug use and housing instability, was defined as a postal code beginning with ‘V6A’.²¹ Area income quintile, adjusted for household size, was derived from census data as a measure of socio-economic status aggregated at the neighbourhood-block level. CD4 count (immune status) at ART initiation (±90 days) was categorised as <200, 200–349, 350–499, ≥500 cells/mm³; those who never initiated ART were classified as missing. All sociodemographic and clinical variables were assessed at baseline.

Healthy life expectancy

HLE estimates the number of healthy years an individual is expected to live at a certain age, if current age-specific mortality rates and disease prevalence rates remain constant over the individual’s remaining lifetime.^{12 13} HLE was estimated using Sullivan’s abridged life table method, which integrates age-specific mortality and disease prevalence rates.

Sex-specific standard period life tables were constructed for each of the three calendar intervals using age-specific mortality rates aggregated into 5-year age intervals. We estimated remaining LE at ages 20, 40 and 55 years for each sex and time period. These estimates represent the average number of years an individual is expected to live at a certain age, if current age-specific mortality rates remain constant.

Age-specific disease prevalence was calculated at the midpoint of each calendar interval (15 February 1999 for 1996–2001, 31 December 2006 for 2002–2011 and 14 February 2016 for 2012–2020), stratified by sex and 5-year age interval, assuming midpoint prevalence represented the entire interval. The population in the cohort, alive and with a 5-year lookback window applied at the midpoint was used as the denominator. This 5-year lookback window was applied when ascertaining health conditions and this window length was chosen to balance sensitivity and specificity while minimising differential misclassification.²² Prevalence estimates were then extrapolated to the entire population contributing person-years to that calendar period.¹³

Sullivan’s abridged life table method, which combines mortality and disease prevalence data, was then used to partition total LE into years lived with and without disease.¹²

Both sex-stratified mortality data and disease prevalence rates were aggregated into 5-year age intervals, starting at age 20 and ending just before the last, open-ended interval (75 years and older). Intervals with fewer than 100 person-years were combined with the adjacent interval containing the smallest number of person-years.²³

HLE estimates

We considered three estimates of HLE: (1) age-related condition-free LE, (2) age-related multimorbidity-free LE and (3) mental health condition-free LE. We pragmatically identified common age-related and mental health conditions for which standardised case-finding algorithms have been published in Canadian settings, primarily those used by BC Chronic Disease Registries (online supplemental appendix table 1).²⁴ Non-AIDS defining cancers were derived from the BC Cancer Registry, and all other conditions were identified by querying healthcare records (healthcare practitioner billings (MSP), hospitalisations (DAD) and medications (PharmaNet) databases) for condition-related International Classification of Diseases diagnostic codes, surgical procedure codes and/or medication dispensations, applying previously published case-finding algorithms and a 5-year lookback window (online supplemental appendix table 1).

Age-related condition-free LE was defined as LE without any of the following age-related conditions: cardiovascular disease (including ischaemic heart disease, heart failure and hospitalised stroke), hypertension, chronic kidney disease, chronic liver disease, chronic obstructive pulmonary disease, non-AIDS defining cancers (excluding non-melanoma skin cancers), diabetes mellitus, osteoarthritis, osteoporosis, hypertension and Alzheimer’s disease and other dementia.

Age-related multimorbidity-free LE was defined as having less than three age-related conditions. Mental health condition-free LE was defined as LE without any mental health condition; we considered substance use disorder, mood and anxiety disorder and schizophrenia and delusional disorder, without distinguishing severity (online supplemental appendix table 1).

Midpoint prevalence estimates of having at least one age-related or mental health condition, used to calculate HLE, are presented for each interval in online supplemental appendix table 2, stratified by age group and sex. For each HLE measure, we estimated HLE with a 95% CI at age 20, 40 and 55 for the calendar intervals 1996–2001, 2002–2011 and 2012–2020, stratified by sex. We used Chiang’s variance formula to estimate 95% CIs.²⁵

We descriptively estimated expected number of years lived in an unhealthy state and expected proportion of time lived in an unhealthy state, defined as the ratio of (LE-HLE)/LE, at age 20, 40 and 55 stratified by sex and for each calendar interval. Relative expansion of the unhealthy state occurs when this ratio increases, and relative compression when this ratio decreases.

Analyses were conducted using SAS V.9.4 (SAS Institute, Cary, North Carolina, USA). Sex differences in HLE were quantified by calculating the HLE gap (male minus female HLE) and corresponding 95% CIs for each age and calendar period. Z tests were used to compare HLE between males and females across several ages and periods; multiple comparisons were not adjusted for, and results should be interpreted in the context of overall patterns. All tests were two-sided with p<0.05 considered statistically significant.

Sensitivity analyses

Hypertension could arguably be considered as one of several cardiovascular risk factors rather than a disease itself.²⁶ Therefore, we conducted a sensitivity analysis excluding hypertension from the conditions considered in age-related disease-free LE. We also conducted a sensitivity analysis restricting to individuals with a first record of HIV infection in or after 1996 only, as this marks the introduction of highly active ART. To further focus on individuals more engaged in care, we conducted a sensitivity analysis excluding those who were never on ART.

Role of the funding source

The funders had no role in the data collection, analysis, interpretation, writing of the manuscript or decision to submit for publication.

Patient and public involvement statement

Patients or the public were not involved in the design, conduct and reporting of this analysis.

Results

As shown in figure 1, COAST included a total of 17 119 individuals, of whom 14 135 (82.6%) were included in the analysis. Among these, 11 624 (82.2%) were males and 2511 (17.8%) were females. The median follow-up period between April 1996 and March 2020 was 10 person-years for both sexes (25th–75th percentile (Q1–Q3): males, 4–17 person-years; females, 4–16 person-years). Baseline characteristics are summarised in table 1. Median age at baseline among females was 34 years (Q1: 28, Q3: 42) and 38 years (Q1: 32, Q3: 46) among males. A history of IDU was more common among females (43.8%) than males (26.2%).

Figure 1.

Selection of study participants. BC, British Columbia; COAST, Comparative Outcomes and Service Utilisation Trends.

Table 1.

Characteristics of the study population at baseline

	Females with HIV	Males with HIV
	N=2511	N=11 624
	Median (Q1, Q3) or N (%)	Median (Q1, Q3) or N (%)
Age, years	34 (28–42)	38 (32–46)
Fiscal year of first record of HIV infection
<1 April 1992	66 (2.6)	511 (4.4)
1 April 1992–31 March 1996	166 (6.6)	1346 (11.6)
1 April 1996–31 March 2002	934 (37.2)	3865 (33.3)
1 April 2002–31 March 2012	913 (36.4)	3651 (31.4)
1 April 2012–31 March 2020	432 (17.2)	2251 (19.3)
History of IDU
Yes	1101 (43.8)	3044 (26.2)
No indication	1410 (56.2)	8580 (73.8)
Area-level income (quintiles)
Q1 (lowest)	978 (47.4)	3627 (36.1)
Q2	428 (20.7)	2096 (20.9)
Q3	250 (12.1)	1852 (18.4)
Q4	212 (10.3)	1305 (13.0)
Q5 (highest)	196 (9.5)	1166 (11.6)
Missing	447*	1578*
Living in a rural area
Yes	155 (6.4)	421 (3.7)
No	2281 (93.6)	10 856 (96.3)
Missing	75*	347*
Residence in Vancouver’s Downtown Eastside
Yes	311 (12.8)	830 (7.4)
No	2118 (87.2)	10 434 (92.6)
Missing	82*	360*
Initiated ART (ever)
Yes	2215 (88.2)	10 730 (92.3)
No	296 (11.8)	894 (7.7)
CD4+ T-cell count at start of ART, cells/mm3
≥500	455 (22.6)	2369 (24.0)
350–499	364 (18.1)	1956 (19.8)
200–349	529 (26.3)	2538 (25.7)
<200	662 (32.9)	3013 (30.5)
Missing	501*	1748*

*Percentages are calculated excluding missing values.

ART, antiretroviral therapy; IDU, injection drug use; Q1–Q5, first to fifth quintile.

Age-related condition-free LE

Age-related condition-free LE was higher among males than females at age 20 and 40; differences were not significant at age 55 (figure 2A, online supplemental appendix table 3). In 1996–2001 males had higher age-related condition-free LE by 2.7 (95% CI 1.9 to 3.5), 3.1 (95% CI 1.8 to 4.3) and 1.6 (95% CI −0.8 to 4.0) years at ages 20, 40 and 55, respectively. In 2002–2011 the corresponding differences were 4.5 (95% CI 2.4 to 6.6), 2.2 (95% CI 0.7 to 3.8) and −0.4 (95% CI −3.0 to 2.2) years; and in 2012–2020, 4.2 (95% CI 2.6 to 5.8), 2.4 (95% CI 0.8 to 3.8) and 0.2 (95% CI −1.7 to 2.2) years (online supplemental appendix table 4).

Figure 2.

Healthy life expectancy and years spent in an unhealthy state at age 20, 40 and 55, stratified by sex and time period. (A) Age-related condition-free life expectancy and years spent with at least one chronic age-related condition; (B) age-related multimorbidity-free life expectancy and years spent with age-related multimorbidity; (C) mental health condition-free life expectancy and years spent with at least one mental health condition.

Years lost to chronic age-related conditions were similar in the periods 1996–2001 and 2002–2011; however, in 2012–2020, males lost more years to age-related conditions than females at all ages (online supplemental appendix tables 3 and 4). The proportion of expected remaining life with at least one age-related disease was similar between sexes, and increased over time (figure 3A).

Figure 3.

Proportion of remaining life expectancy spent in an unhealthy state at age 20, 40 and 55, stratified by sex and time period. (A) Proportion of remaining life expectancy spent with at least one age-related chronic; (B) proportion of remaining life expectancy spent with age-related multimorbidity; (C) proportion of remaining life expectancy spent with at least one chronic age-related condition.

Multimorbidity-free LE

Multimorbidity-free LE was consistently higher among males than females, except at age 55 in 2002–2011 (figure 2B, online supplemental appendix table 3). In 1996–2001, males had higher multimorbidity-free LE by 2.3 (95% CI 2.1 to 2.4), 2.8 (95% CI 2.6 to 3.0) and 2.7 (95% CI 2.3 to 3.1) years at ages 20, 40 and 55, respectively. In 2002–2011 the corresponding differences were 4.4 (95% CI 3.5 to 5.3), 3.5 (95% CI 2.3 to 4.7) and 0.0 (95% CI −2.1 to 2.1) years; and in 2012–2020, 7.2 (95% CI 4.9 to 9.4), 5.6 (95% CI 3.9 to 7.3) and 3.6 (95% CI 1.3 to 6.0) years (online supplemental appendix table 4).

Sex differences in years lost to age-related multimorbidity were small. However, despite females having a lower LE in the most recent period (2012–2020), they were expected to spend more years with multimorbidity than males (online supplemental appendix tables 3 and 4). Thus, the proportion of expected years lived with multimorbidity was higher among females and increased over time in both sexes (figure 3B).

Mental health condition-free LE

Mental health condition-free LE was higher among males than females, except at age 55 in calendar periods 1996–2001 and 2002–2011 (figure 2C, online supplemental appendix table 3). In 1996–2001, males had higher mental health condition-free LE by 6.5 (95% CI 4.1 to 8.8), 4.1 (95% CI 2.9 to 5.2) and 1.6 (95% CI −0.8 to 4.1) years at ages 20, 40 and 55, respectively. In 2002–2011 the corresponding differences were 9.1 (95% CI 6.8 to 11.4), 4.8 (95% CI 3.2 to 6.4) and 1.8 (95% CI −0.8 to 4.5) years; and in 2012–2020, 10.0 (95% CI 6.9 to 13.0), 6.8 (95% CI 4.9 to 8.6) and 4.2 (95% CI 1.8 to 6.5) years (online supplemental appendix table 4).

Females consistently spent more years, except at age 55 in 1996–2001, and a higher proportion of their life with mental health conditions than males (figure 3C). Expected years spent with at least one mental health condition increased over time in both sexes (online supplemental appendix table 3).

Sensitivity analyses

Sensitivity analyses were consistent with the main findings. Excluding hypertension reduced estimated years with at least one age-related condition by 1–3 years, but sex differences persisted (online supplemental appendix table 3). Restricting to individuals with a first record of HIV infection in or after 1996 only (n=12 255; 81% males, 19% females) showed similar patterns across all HLE measures (online supplemental appendix table 5). Further excluding individuals who were never on ART (n=11 204; 82% male, 18% female) yielded higher HLE, particularly multimorbidity-free LE in periods 1 and 2 (online supplemental appendix table 6).

Discussion

In BC, Canada, a high-income setting with universal healthcare, LE among females living with HIV has persistently remained lower than that of males, with the gap increasing over time. In this context, we examined HLE among males and females living with HIV in BC between 1996 and 2020.^{2 4} Regardless of how HLE was operationalised—whether as age-related condition-free LE, multimorbidity-free LE or mental health condition-free LE—both HLE and LE increased over time. Overall, we find that even in the modern cART treatment era, PLWH are expected to spend not only a growing absolute number of life years in an unhealthy state, but also an increasing proportion of their remaining lives in an unhealthy state.

HLE was lower among females than males, across age groupings and time periods and females were generally expected to spend a greater proportion of life in an unhealthy state. In the most recent period (2012–2020), females at exact ages 20 and 55 years were expected to spend 2.9 and 2.6 fewer years than males with age-related conditions, 0.1 and 0.9 more years with multimorbidity and 2.9 and 1.4 more years with mental health conditions. Although females were expected to spend a similar proportion of their remaining life with at least one age-related condition than males, they were expected to spend a larger proportion with mental health conditions and multimorbidity. These sex disparities persisted even among PLWH with demonstrated access to HIV treatment.

Age-related conditions and multimorbidity

The number of years lived with at least one age-related condition increased substantially for both sexes, consistent with patterns reported in prior literature.²² However, in the most recent time period (2012–2020), we found that females spent more years and a higher proportion of their remaining LE with multimorbidity than males.

Multimorbidity is more common among females than males in the general population, although the individual conditions contributing to the burden differ by sex.²⁷ This pattern may be exacerbated among PLWH. A large US study found people with or at risk for HIV reported higher multimorbidity prevalence among females, particularly those with HIV.^{5 28} Substance use, which in Canada is generally higher among females than males with HIV, and biological factors such as ongoing inflammation and immune dysregulation may contribute to this sex difference among PLWH.^{5 27 28} Research shows that females persistently have higher levels of immune activation and inflammation than males with suppressed HIV viremia.²⁹ In addition, females living with HIV may face unique intersecting socio-structural challenges, including poverty, unstable housing, stigma and other barriers to care, which can exacerbate the development and management of comorbidities, contributing to greater multimorbidity in this population.⁵

Mental health

Mental health condition-free LE improved over time for both sexes, with males consistently having higher mental health condition-free LE than females. Expected years spent with mental health conditions, both absolutely and proportionately, were consistently higher among females (except at age 55 in the period 1996–2001), aligning with prior literature.³⁰

Over time, both females and males could expect to live more absolute years with at least one mental health condition. However, the proportion of remaining LE spent with these conditions decreased across the three-time intervals. This proportional decline is likely driven by gains in overall LE; however, the uptake of some integrated mental health services into routine HIV care may also play a role in containing further expansion of the mental health burden.^{9 17} Mental health outcomes are closely linked to poorer HIV and other longer-term health outcomes.^{30 31} The consistently higher number of expected years lived with at least one mental health condition among females than males highlights the need to better understand the underlying drivers of these sex disparities.

Implications

Our findings indicate that PLWH, especially females, are spending more of their lives with chronic age-related conditions, multimorbidity and mental health conditions. A large part of this sex disparity likely stems from socio-structural factors unique to females living with HIV. In our study sample, 44% of females and 26% of males had a known history of IDU, a marker of substance use burden and underlying socio-structural disadvantage.¹⁰ Along with other socio-structural factors, IDU and broader substance use may contribute directly to the development of mental health and age-related conditions or indirectly by compounding social exclusion, instability and barriers to accessing care.

Although the proportion of expected time lived with mental health conditions slightly decreased among both sexes, the number of years lived with these conditions continues to rise. People on ART in our study had increased HLE, underscoring the role of antiretrovirals in extending LE and improving the quality of years of life gained. However, the gap in HLE remained by sex in this group. Overall, our research underscores the need for targeted interventions to address age-related and mental health conditions among PLWH, especially females, who are disproportionately impacted. This should include scaling up integrated models of care that combine HIV treatment, chronic disease management, mental healthcare and harm reduction, with particular attention to the needs of women and individuals with histories of substance use.

Limitations

Readers should be cautious in interpreting our findings. Disease prevalence was calculated over a subset of the population with a 5-year lookback window and may underestimate the prevalence of comorbidities and particularly of long-standing but stable conditions in older adults and especially among females who may be less likely to access care. Age-related and mental health conditions were identified in administrative health data and may be subject to potential misclassification, changes in coding or diagnostic practices over time, and sex differences in healthcare seeking behaviour and diagnosis. Left truncation and survivor bias may also influence HLE estimates. Liver disease, included in the composite measure of age-related disease, captures long-term consequences of viral hepatitis which may be considered independent from the ageing process. Cancer prevalence was defined using the BC Cancer Registry, limiting to diagnoses within the 5 years preceding midpoint in each calendar period.³² The registry’s completeness exceeds 95% but it may include potentially less clinically relevant conditions and therefore overestimate prevalence. The aggregation of small age groups and the use of an open-ended interval (75 years and older) may obscure heterogeneity and mask age-related differences in health outcomes at advanced ages.

When evaluating trends in health LE, the Sullivan method can generate biased estimates as it combines mortality incidence with disease prevalence, in particular in settings of rapidly changing mortality or morbidity.³³ However, it is widely used in population health research for estimating HLE and enables comparisons across populations and time periods. More advanced approaches, such as multistate or Markov models, could explicitly model transitions between health states and death.³⁴ Future studies applying these methods would require more complex longitudinal data on these transitions but may provide a more detailed understanding of health trajectories and ageing among PLWH.

Conclusion

In this population-based study among PLWH in BC, Canada, we found that HLE has improved over time for both males and females but has not kept pace with increasing LE. As a result, the number of years expected to be spent in an unhealthy state has increased. Females consistently had lower HLE than males and despite having a lower overall LE, in recent years they are expected to spend more years and a greater proportion of their remaining LE with multimorbidity and mental health conditions. Addressing age-related and mental health conditions is critical to closing the growing gap between LE and HLE among PLWH. Overall, our findings underscore the need to prioritise the health of females living with or at risk for HIV to achieve health equity in this population.

Data availability statement

Data may be obtained from a third party and are not publicly available. Access to data provided by the Data Stewards is subject to approval but can be requested for research projects through the Data Stewards or their designated service providers. The following datasets were used in this study: Consolidation File, Medical Service Plan, Discharge Abstracts Database, PharmaNet, BC Cancer Registry and Vital Events and Statistics – Deaths. Further information regarding these datasets can be found on the Population Data BC project webpage at: https://my.popdata.bc.ca/project_listings/18-223/. All inferences, opinions and conclusions drawn in this publication are those of the authors, and do not reflect the opinions or policies of the Data Stewards. The Population Data BC Data Access Unit can be contacted through dataaccess@popdata.bc.ca.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

COAST has received approval from the University of BC/Providence Health Care Research Ethics Board and Simon Fraser University Office of Research Ethics (#H22-02875). The study complies with the BC Freedom of Information and Protection of Privacy Act and does not require informed consent as it is conducted retrospectively for research and statistical purposes only using anonymised data.