Do the Positives Through Life Decrease Later-life Frailty?: Positive Psychosocial Resources and Health Status in Older Adult, a Prospective Cohort Study

Background: Frailty is a known predictor of poorer outcomes for hospitalised older adults, but does not account for all variation in outcomes. Health Assets, which include positive psychosocial factors, have been associated with improved outcomes in the hospital setting. Methods: A prospective cohort study from adults aged 70 and older with an unplanned admission to general medical, orthogeriatric and subacute wards of two hospitals in Australia. 298 participants were recruited with an average age of 84.7. The Health Assets Index (HAI), frailty, functional status and covariates were measured at the time of recruitment. Outcomes were mortality at 30 days and functional decline at the time of discharge. Results: odds but of health assets where frailty was associated with increased mortality. Conclusions: Health and for entering hospital.


Page 3/19
been identi ed across multiple domains including biological (eg. high activity) functional (eg. ability to undertake community activities) and subjective (eg. a sense of wellbeing). 10 In community studies, health assets can act individually and cumulatively to improve survival and health status. 9,11 Although the protective effect of health assets in the community has been established, their impact in the hospital setting is less well characterised. 12,13 Systematic reviews have demonstrated that individual health assets can improve outcomes. 11,13 Health Assets have previously been found to moderate the impact of frailty associated outcomes. 4 An example is that two frail individuals might have different outcomes if one is able to access the community versus remaining at home. 4 A Health Assets Index (HAI) was developed to measure health assets in the acute health care setting. The aims of this study were : 1. To examine the relationship between frailty and Health Assets 2. To determine whether health assets moderate frailty associated outcomes in the acute hospital setting.

Setting and Sample
The study protocol has been previously published. 14 In brief, study recruitment took place in general medical, orthogeriatric and subacute wards of two hospitals in Victoria and Queensland, Australia. Eligible participants were adults aged 70 and older who had an unplanned admission to hospital. Participants could be recruited at any time during hospital admission. Participants were excluded if they had cognitive impairment with no available next of kin, non-English speaking, severe psychiatric disturbance, too medically unwell or receiving terminal care.
Researchers screened for cognitive impairment by discussion with medical and nursing staff. Researchers also used clinical judgment when seeking consent, in case of new onset of delirium. If researchers or medical staff identi ed cognitive impairment, the next of kin was asked for formal consent. Participants deemed not able to consent were still asked about their willingness to participate to ensure they did not object to being involved.

Measures
The Health Assets Index (HAI) (See supplementary data S2) was developed based on variables identi ed by systematic review and a secondary analysis of a large, Australian inpatient dataset. 4,13 The following criteria were chosen for variables to be included in the HAI: Associated with positive health outcomes.
Not included in the frailty index.
Not present or absent in greater than 95% of patients.
As a group, the candidate assets must cover a range of domains Assets must be age appropriate The proposed scoring system was that each asset would be scored between 0-1 with a higher score corresponding to a higher number of health assets.

HAI and covariates
Measures were obtained using patient reported data, observation and medical records. At the time of recruitment, health assets were recorded on the HAI by trained assessors. Covariates measured included frailty, illness severity, sex and usual place of residence. Frailty was measured based on premorbid function at two weeks before admission using an FI (see supplementary data S1), which has been previously validated in the inpatient setting. Although the timing of recruitment from admission was variable, the FI has previously been used at different times during the admission, and remains valid. Illness severity was measured with the Modi ed Early Warning System (MEWS), which uses routine observations at the time of admission. 15 Katz Activities of daily living (ADLs) 16 and Instrumental Activities of Daily Living (iADLs) at the participant's reported baseline were recorded with the same scoring used for men and women.

Primary Outcomes
The primary outcomes were: 1. Mortality at 30 days after discharge, which was identi ed on hospital records.
2. Functional decline at the time of discharge from hospital which was de ned as a decreased score on ADLs compared to baseline or new discharge to residential aged care Secondary Outcomes 1. Total length of stay including acute and subacute wards 2. Readmission within 30 days, identi ed on hospital record or phone call 3. Functional decline at 30 days post discharge home, which was de ned as a decreased composite score for ADLs and IADLs at 30 days post discharge. Participants who were readmitted to hospital were excluded from this analysis as the interceding illness or injury that precipitated readmission would provide a confounder for functional decline.
Participants admitted from residential care were excluded from analyses of functional decline at 30 days post discharge as IADLs are largely not applicable to this population. Those discharged to a new residential aged care were assumed to have persistent functional decline.

Analysis
Frequency distributions were used to describe cohort characteristics, including each variable included in the HAI. For statistical analysis, multivariate logistic regression models were used to estimate the effect of health assets and frailty on 30 day mortality. An interaction term for health assets and frailty was included, allowing the effect of one to vary at different levels of the other.
For functional decline and the secondary outcomes, regression models estimate the effect of health assets and frailty. A negative binomial regression was used for length of stay to account for right skew of the data.
All analyses were adjusted for age, sex and MEWS.

Results
A total of 312 participants were screened, and 15 declined to participate. 298 participants recruited and 250 had a completed HAI. The average age was 84.7 and 193(66%) were women (table 2) 127(43%) lived alone, 137(46%) lived ADLs mean(SD); median (IQR) 4.9 (1.6) 6 IQR(4,6) iADLs mean(SD); median (IQR) 5.1(2.9) 6 IQR (2,8) with others and 31(11%) lived in residential aged care. 80 (27%) were admitted to an orthogeriatric unit at the time of admission, 146(49%) were admitted to a general medical unit and 70 (27%) were admitted to subacute care. 238 (80.1%), had a frailty index of greater than 0.25with a population mean score of 0.38 (SD 0.12) (table 2). The mean HAI score was 10.86 (SD 2.87) with a minimum of 5.5 and a maximum of 15. Table 2 describes the distribution of   assets (table 1). There was a signi cant inverse relationship between a higher number of health assets and frailty, OR 0.36(95%CI 0.19-0.68), with a higher number of health assets protecting against frailty. In analysis examining only frailty or the HAI there was no signi cant association with length of stay or in hospital functional decline (Table 3). After further consultation with a statistician, a logistic regression model was used that accounted for interaction. This showed a signi cant interaction between frailty and health assets for mortality, p=0.011 (95%CI 1.10-2.20) (see supplementary gure S1). A marginal plot demonstrates that at the lowest levels of frailty, a higher number of health assets was protective against mortality, and at the highest levels of frailty, a higher number of health assets was associated with an increased risk of mortality (see gure 1).
A higher number of health assets did not mitigate the risk of functional decline on leaving hospital, but in univariate analysis it was protective against persistent functional decline after discharge and readmission (table 3). This effect was not present when the interaction with frailty was accounted for.

## interaction term *P<0.05
Discussion This is the rst study to examine the use of a health assets index in a hospital population. Hospitalisation is a critical juncture for older adults and as frailty does not explain all variation in outcomes, improving prognostication could have great bene ts to individuals and hospital systems. This study demonstrates a signi cant interaction between health assets and frailty and provides insights into both the development and management of frailty. Among these older inpatients, a higher score on the HAI was associated with an improved baseline health status, as evidenced by the lower likelihood of frailty. The HAI alone was not predictive of mortality or functional decline, but in a model that accounted for the interaction with frailty it had differing effects for more robust compared to more frail older adults.
This suggest that health assets are likely to moderate the development of frailty and to mitigate adverse outcomes for more robust older adults.
In studies of older adults, frailty alone is a better predictor of mortality than age alone. 17 As not all older adults become frail at the same rate, with population ageing, it is important to develop a better understanding of factors that in uence the development of frailty. 18 The inverse proportional relationship between a higher number of health assets and frailty is in keeping with other studies indicating that psychosocial factors through life affect health status in older age. 19 This highlights the importance of taking a life course approach to understanding ageing.
One of the strengths of the frailty index is that it utilizes a multidimensional approach to frailty, which incorporates physical, cognitive and functional components, which are all components of a comprehensive geriatric assessment. 20 The concept of health assets was rst developed and explored in the community setting in longitudinal studies. 9,11 The lack of impact of protective factors for those who were already frail is consistent with ndings in the community and supports that once an individual is frail, protective factors do not improve survival. 12 The Canadian Study of Health and Aging has also demonstrated that for adults aged 65 and older who were t, a higher level of self-rated health, which is a Health Asset, protected against mortality. 21 The different effect of Health Assets for those who are robust and frail is likely related to the underlying physiological differences between these groups. It may be that when someone is extremely frail, the allostatic load leads to a critical loss of physiological reserve, so that any biological impact of protective factors is negated. 21 It is not clear why the higher number of health assets would be associated with a higher mortality for people who are very frail. Those with a higher number of health assets may be better supported in the community, and only present to hospital with a more signi cant illness. Conversely it is also possible that this frail group are particularly dependent on their assets, such as carers and emotional support, and are at greater risk when they cannot access these.
The lack of a clear proportional relationship between health assets and mortality contrasts with other studies in the hospital setting. 13 Although many studies have identi ed an association between individual assets and improved outcomes, most of these studies did not include a measure of frailty. Multiple studies investigated mortality and functional decline up to a year after discharge from hospital, and so it may be that health assets have more impact over the months following discharge. 22 This differing effect of health assets on older adults depending on frailty status indicates a challenge with measuring health assets. Although individual assets may have varying effects for individuals, the advantage of measuring multiple assets is that it the higher overall score is associated is associated with a lower level of frailty, somewhat mitigating individual variation.
Although frailty is mostly de ned in physiological terms as a loss of homeostatic reserve, and is characterized by a stochastic accumulation of subcellular de cits, the impact of psychosocial factors on biology needs to be considered as part of the pathogenesis. In longitudinal community studies, a higher number of social supports are protective for older adults, when accounting for frailty and co-morbidity. 10,23 Mechanistically there is evidence that negative psychosocial factors are associated with higher levels of in ammation, which is a proposed mechanism of accelerated biological ageing and the development of frailty. 24,25 It may be that a higher number of Health Assets can buffer these changes, which is why it is associated with a better health status.
In this cohort, frailty was not predictive of mortality, functional decline or length of stay, which contrasts with previous studies in the hospital setting. 1,5 A recent review identi ed that 25% of studies in the hospital setting, increasing frailty is not predictive of mortality. 26 This is in contrast studies where frailty is measured in the community setting. 17,27 This highlights the need for further implementation work in the clinical setting to better determine how this score improves risk prediction for individuals. It also highlights the importance of utilizing data that has been collected in the community setting for hospital based risk prognostication.
Sarcopenia is a loss of muscle strength and function. This is highly prevalent in populations of older adults admitted to hospital, with prevalence at around 35%. 28 Sarcopenia is strongly associated with frailty and limitation in ADLs. 28 Sarcopenia was not measured in this cohort, but it is likely to provide an important contribution to limitations in ADLs identi ed.
This study also highlights one of the di culties in measurement of health assets. When a biological measure, such as optimal creatinine, is identi ed, a laboratory cut off range is chosen by identifying a range that will cover most of the population. For items like social connection, it is not only the frequency, but also the quality of connections that impact health. It may be more appropriate to take a subjective and individualised approach to these items.
The study has certain strengths: very few patients refused participation, and due to the use of routine data, there was a high rate of follow-up for the primary outcomes. There were also important limitations: the relationship between frailty and health assets was only measured at one point in time, so causation cannot be inferred. It is possible that mortality was under-reported as hospital record data was used, along with phone calls to individuals who had returned home, although not all could be contacted. The decision to not include people receiving terminal care so as not to cause undue burden and distress from participation in research likely meant that mortality was lower than expected.
The small sample size meant that there may have been insu cient power to detect a statistically signi cant result. Due to limited numbers of research personnel, not all possible participants could be approached. The statistical model that utilised an interaction was not pre-speci ed, so this should be interpreted with caution. People who did not speak English were not included due to the lack of resources for interpreters, which limits generalizability in a multicultural setting. The follow-up was limited to a maximum of 30 days after hospital discharge, but it is possible that over a longer duration of time after discharge, health assets may have an impact on survival. The study took place in an Australian setting, so the HAI may not be valid in other countries.
Further qualitative research speci c to older adults could help determine which factors this age group think are desirable and have an immediate impact on their own wellbeing.

Conclusion
The interaction between frailty and health assets highlights the complex interplay between social, psychological and biological factors on individual rates of ageing, as de ned by the measurement of frailty. This study further supports the need to consider psychosocial factors in the development of frailty. It is not yet clear whether health assets are effective due to behavioural mechanisms, or if there is also an underlying physiological effect. The interaction between health assets and frailty has intriguing implications for health at the broader population level to identify strategies to improve long-term outcomes and immediate quality of life.  How many children do you have? 0 zero 0.5 for one to two 1 for three or more Social engagement 10 Can you count on anyone to provide you with emotional support eg talking over a problem, or helping with a decision? 0 no 1 yes 11 How many times a week do you see or talk to a family member or friend who does not live with you? 0 never 0.5 less that once a week 1 once a week or more 12

List Of Abbreviations
In the 3 days prior to the onset of the illness precipitating admission, number of days went out of the house or building in which he/she resides  Fi gu re s Figure 1 Effects of Health Assets and Frailty on Mortality. This demonstrates a differential effect of health assets, depending on frailty status. For those who were not frail, a higher number of health assets decreased the risk of mortality. For the most frail, this relationship was reversed and a higher number of health assets was associated with increased mortality.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. supptables.docx