In this nationally representative study of older adults in Sweden, we found that life expectancy at age 77 continuously increased between 1992 and 2011. Overall, the number of years lived with CHP and the share of life with CHP increased for both men and women. In 2011, from age 77, men spent on average 2.15 years and women 4.01 years with CHP. For women, the added life years consisted only of years with CHP, while the added years for men were equally distributed between years with and without CHP. Fewer individuals with CHP resided in residential care in 2011 compared to one and two decades before.
The finding that there was no sign of a decrease in time spent with CHP has important implications for resource needs and funding of care services. Moreover, the composite measure of CHP is highly indicative of the need of both social care and medical care, usually involving several providers. Our results, that an increasing proportion and number of older adults with CHP live in the community, and not in residential care, underlines the importance of the development and improvement of integrated medical and social services.
The trend of an increase of years lived with CHP can mainly be attributed to an increasing prevalence of CHP in the first part of the period, which all three health dimensions contributed to. Most previous studies on time trends in health expectancies have used single health indicators and divergent trends have been shown for different indicators, countries and time-points compared [21]. Given limits in comparability, our results of an increase in years lived with complex health problems point towards a similar trend as a few previous studies that have used composite health measures. For example, expected life years spent with cognitive impairment and disability combined increased slightly in Hong Kong between 2001/02 and 2011/12 [19]. A study based on the British Cognitive Function and Ageing Studies integrated dependency in activities of daily living and geriatric conditions into a measure of dependency. Results showed a significant increase of years lived from age 65 with both low dependency (requiring help less often than daily) and high dependency (24-h care) between 1991 and 2011 [33]. As in the present study, the development was more pronounced for women than for men [33]. On the other hand, results from the Longitudinal Aging Study Amsterdam reported an increase in years lived with combined multimorbidity and disability between 1993 and 2016 for men, but no clear trend for women [22]. Studies focusing on diseases and/or physiological indicators found an increase in time spent with six or more chronic conditions and polypharmacy between 2005 and 2014 in Germany, both in absolute years and as an increasing proportion of life [32].
A study among community-dwelling people conducted in 2011 revealed substantial differences in life expectancy with frailty at age 70 across 15 European countries, with estimates ranging between 0.1 to 1.8 years for men and from 0.4 to 5.5 years for women [31]. However, to our knowledge, there is a lack of studies on time trends in life expectancy with frailty.
That women in our study at all three time points could expect to live twice as long as men with CHP reflects the male-female health-survival paradox [44], that has been observed for most other single as well as composite health measures, such as multimorbidity [32, 45], frailty [31, 46], disability [33] as well as poor physical and cognitive health combined [19, 22].
Our decomposition analyses indicated that the increase in expected life years with complex health problems among women in the period of 1992–2002 was mainly driven by change in the prevalence of CHP, rather than mortality, while the contributions of these forces were similar among men. The increase of years without CHP, on the other hand, was predominantly driven by change in mortality, rather than the prevalence complex health problems, especially during the latter part of the period (2002–2011), and among men.
Increased prevalence of diseases and chronic conditions, as well as some functional limitations, have also been widely reported in Sweden [47] and internationally [14, 25, 33], although evidence on trends in the prevalence of disability seems more favorable [47]. Earlier diagnoses, successful life-saving interventions and better disease-control probably contribute to an increase of the prevalence of chronic conditions and functional limitations. Improved survival with disease may also entail an increased risk for further diseases [48].
Some strengths and limitations of this study must be acknowledged when interpreting the results. Although the sample sizes of the three SWEOLD waves are relatively small, a main strength of the study is that prevalence estimates can be assumed to be highly representative. All survey waves comprise nationally representative samples of the population aged 77 or older with high response rates (> 85%), including people with poor cognition and those living in residential care. Identical study design and methods were used over the decades. In order to avoid non-response, and thereby an underestimation of health problems on population level, telephone interviews or proxy interviews with a close family were conducted for those individuals who were too frail and/or cognitively impaired to participate by themselves or who did not want to participate in a face-to-face interview [36].
The health indicators are mostly self-reported (except the cognitive test), which could introduce bias in the health assessments. In national population surveys, for practical reasons, clinical precision cannot be accomplished. However, crude health measures that are not clinically relevant, but allow the inclusion of proxy interviews, can be highly suitable for national surveys aimed at informing decision makers for policy and planning.
To cover the most vulnerable subset of the older population, we made our definition of complex health problems restrictive, setting high thresholds for severe problems within each domain. However, people with severe problems in only one domain could also need care from a variety of providers. Thus, rates of CHP as measured in this study, could to some extent underestimate actual needs for integrated medical and social care in the older population.
Moreover, the Sullivan method, which we used to calculate health expectancies, does not take transitions between health states into account, which could potentially introduce bias in the health estimates [41]. However, recovery from complex health problems, involving severe problems in several health domains, is likely to be limited in high age, and the estimates based on the Sullivan method have proven reliable as long as prevalence is regular and smooth over time [49, 50].