Incidence of Home Care and Nursing Home entry
In this eight-year cohort study, a follow-up of more than 95% participants was obtained, with 19% entering home care programmes, and 8% geriatric nursing homes, outcomes that are generally associated to advanced frailty, especially when it is not addressed early and adequately. Regarding HC, although it is difficult to compare different countries and health services, similar rates were observed between homebound incidence in Japan39 (32.1 per 1000 individuals/year) and that of our study (30.6 per individual/year). With respect to NH entry, we found lower rates than those reported by the USA40, 16.1% in two years, but closer to Germany, with rates of 4.7% in a three-year follow-up41. This could be explained by the fact that in southern European cultures the involvement of the family in the care of the elderly is considerable, whether for cultural or economic reasons 42,43.
Age and sex
In our results, increasing age was the main predisposing factor associated with frailty and both HC and NH placement. Although women had a higher incidence of NH inclusion it was not a statistically significant predictor. The higher life expectancy of women, and the greater percentage of their living alone, could explain this trend 44.
In our study it was observed that IADL deterioration was associated with HC and NH. The ability to perform instrumental activities of daily life autonomously is essential to live at home independently. It is, therefore, a relevant measure to take into account when predicting the path to functional decline and dependence. We found that even mild dependence in the IADL was associated with future HC, and moderate to severe dependence with NH entry. IADL impairment has been described as a potential marker of frailty45, implying losses in different functioning domains 46. There is, however, controversy with respect to disability and its inclusion in the definition of frailty 47. Nevertheless, early stages of IADL impairment could be useful in detecting individuals at risk, it is an easy measure to collect and has a long-established tradition in PHC settings. Our sample included very few subjects with dependence in basic activities of daily living as it was composed of community-dwelling, independent individuals. BADL thus had no impact on the prediction models.
Mobility was measured in our study with the TUGT. It has been shown to have high sensitivity for identifying frailty48, moreover, as it is a simple test requiring little equipment and space the TUGT is a valuable tool in a clinical setting. Savva et al49 found that a cut-off point of more than 16 seconds was optimum to identify the frail population. Our results concur, we observed a mean score of 16.2 seconds for subjects entering the HC programme, and 15.5 seconds for the NH one. Due to the fact that the TUGT has been used as a proxy measure of frailty50 and subsequent functional decline, it is a relevant factor in our HC prediction model.
Polypharmacy is a measure of medication-associated frailty, irrespective of the number of comorbidities and their severity51. It is associated with increased rates of falls52 and hospitalization, disability, and mortality53. In our sample, polypharmacy, collected from the electronic primary healthcare records, was a prevailing factor. Present in 52% of the participants, it was higher than in other studies which reported a prevalence based on health surveys of between 26 and 40%54. It was, however, closer to those authors employing electronic healthcare records55 who observed over 50%. In our study, polypharmacy predicted NH entry, a fact that might be related to multimorborbidity in addition to adverse drug reactions/interactions, and greater risk of falls, and negative health outcomes56.
Particular emphasis should be placed on psychoactive drugs as 50% of those entering an NH were taking them. Moreover, inappropriate polypharmacy is a key issue to address in order to improve outcomes in the elderly57 by means of active medication review and deprescription processes with tools such as STOPP START criteria and other available strategies58.
We observed that whilst living alone had no effect on the need for HC it did influence NH entry (sHR = 2.83), a finding that has been already described in other countries59–62. Employing an exhaustive socio-family situation measure, the SFRSE scale, we found a strong association between greater social risk and a higher institutionalization rate, irrespective of functional status or comorbidities. It appears as one of the predictors in the NH entry equation highlighting the importance of social support and environment in maintaining the capacity to live in the community in one’s home. Social and caregiver networks could help circumvent institutionalization, as has been observed in various studies that only take either living arrangements or caregiver networks into account63,64. The need to assess the social sphere of the frail elderly is evident65. Indeed, as the issue of social frailty is increasingly conceptualized66, the design of interventions to improve social support resources and promote inclusion of the elderly will become essential in granting their preferences for living in the community, and thus improving quality of life.
Cognitive impairment and dementia are factors classically described as being related to NH placement67. Whilst we observed a bivariate association between cognitive status and adverse events it was not included in the final model. This was due to the low prevalence of dementia in our free-living, community-dwelling population. As we lacked a longitudinal measure of the incidence of cognitive impairment in our sample, we could not test the association with enough statistical power. The same pattern of bivariate association was also reported for depression, nutritional risk, and urinary incontinence. The latter is additionally usually found as a strong gender-specific predictor68, negatively affecting daily life although our final model did not include it.
Strengths and limitations
Few longitudinal studies can be found in the literature analysing the transition of the frail elderly from the commencement of their requiring home care to later nursing home placement, both outcomes related to functional decline and loss of autonomy. Our cohort had an excellent follow-up rate, up to 95% of the sample, and was representative of the elderly patients attended in primary healthcare in Catalonia, around 12.6% of the total patient population. Although our external validity was limited to those who sought medical assistance at the PHC it should be noted that this was not a health survey aimed at representing all the elderly population in this region. Moreover, most of these individuals in Spain seek medical assistance in the public health sector.
Despite the fact that our models included the main CGA variables there might have been other factors influencing HC/NH admission. Nevertheless, the CGA is comprised of the most important known dimensions, and they were measured using standardised, validated questionnaires and scales. Finally, although an extensive follow-up was performed, changes in baseline variables during follow-up were not analysed as it was a prediction model based on the initial situation of the sample.
Prognostic models established with comprehensive geriatric assessments can predict the commencement of the need for HC and subsequent NH entry in community-dwelling, older adults. Our findings underline the necessity to measure functional capacity, mobility, inappropriate prescriptions, and social aspects of the elderly in primary care settings where they can be offered holistic, longitudinal assessments and tailored interventions.
Such models could also be useful for the risk classification of the frail elderly and in the planning of health care policies.