Baseline characteristics of the cohort and assessment of frailty dimensions are shown in Tables 1 and 2. The mean baseline age was 76.4 years, 55.5% were women, 73.5% did not have complete primary education, and 22% lived alone (32% women vs 9.6% men). Mean BMI was 28.5 and 4.5% smoked. Health was rated as good by 47.4% and bad by 6.7%. Patients on average took 5 drugs a day, and 32% consumed psychotropic ones. At the functional level, the mean Barthel index score was 96.5, the majority were independent for the BADL (62.5%), and only 15.3% had moderate/severe dependence. The IADL, measured with the Lawton index, showed that 76.7% were independent, and only 6.5% presented moderate/severe dependence. Mobility, measured with TUGT, was 13 seconds on average, and 14.4% of the participants led a sedentary life. At the mental level, the average MCE score was 27/30 points, and the GDS scores showed 19.8% probable depression.
During follow-up 19.2% (n=118) of the 616 participants entered a HC programme (30.6 incidence per 1000 person-years), while 8.2% (n=51) were admitted to a NH one (13.1 incidence per 1000 person-years). Of those admitted to a NH, 31.5% (n=17) had previously been in a HC. Of the 616, mortality during follow-up was 15.4% (n=95) for participants presenting no event, and 46.2% (n=78) for those who were in either a HC or NH programme. During follow-up 4.5% (n=28) was lost with a greater proportion of men (64%, p<0.05). However, there were no statistically significant differences in the rest of the main variables between those who completed the study and those who did not (Figure 1).
The median follow-up was 91.8 months (IQ: 58.1-97.7) and 92.3 months (IQ: 59.7-97.9) for the HC and NH subjects, respectively.
Table 1 shows the bivariate sHRs of admission to HC and NC programmes during the eight-year follow-up period according to baseline variables.
Comparing those who entered a HC programme with those who did not, HC incidence increased with age (78.9 years versus 75.2 years, sHR = 1.1), sedentary life style (22.9% versus 11.3%, sHR = 2.4), and poor self-perceived health (12.7% versus 6.7%, sHR = 2.3).
Table 2 shows the bivariate sHRs of admission to HC and NC programmes during the eight-year follow-up period according to the geriatric assessment.
Comparing those who entered a HC programme with those who did not, HC incidence increased with worse functional status, Barthel index (94.1 versus 97.5, sHR = 1.0), and Lawton and Brody index, in light (22.2% versus 13.4% , sHR = 1.7) and moderate dependence (15.4% versus 3.2%, sHR = 4.8). It also augmented in individuals with worse mobility (mean TUGT 16.2 versus 11.8, sHR = 1.1); worse cognitive scores (mean score 26.2 versus 27.4, sHR = 0.9); worse affective state (mean 4.3 versus 3.6, sHR = 1.1); urinary incontinence (59.3% versus 40.5%, sHR = 1.8); and higher social risk (mean 9.4 versus 8.7, sHR = 1.1).
Admission to a NH programme was associated with age (mean 78.7 versus 75.8, sHR = 1.1); living alone (47.1% versus 19.4%, sHR = 2.8); and greater drug consumption (mean 6.6 versus 4.6, sHR = 1.2).
Functional impairment in BADL was related to NH and the Barthel index (mean 93.1 versus 97.1, sHR = 1.0). Unlike a HC, however, NH entry was only associated with the highest degree of dependence in the Lawton and Brody index (moderate/severe dependence, with 15.7% versus 4.9%, sHR = 4.5). It was also related to worse mobility (longer TUGT time, mean 15.5 versus 12.4, sHR = 1.1); worse cognitive scores (mean 25.6 versus 27.3, sHR = 0.9) and risk of depression (mean 5.1 versus 3.6, sHR = 1.1).
We found a higher risk of NH admission for urinary incontinence (66.74% versus 42.4%, sHR = 2.5); higher number of specific morbidities (mean 1.0 versus 0.7, sHR = 1.3); and social risk (higher score on the socio-familial assessment scale of the elderly, mean 10.7 versus 8.7, sHR = 1.2).
A complementary, bivariate sub-analysis was carried out between the participant’s social risk and the occurrence of any events (NH, HC; data not shown). We observed a dose-response association between increasing social risk and being first admitted to a HC programme; with a still higher baseline social risk, entering a NH one: Finally, the highest baseline social risk was associated with first a HC programme and later a NH facility.
The multivariate adjusted model showed that the incidence risk of a HC entry was associated with older age, dependence on the IADL (moderate/severe dependence), and slow gait measured by TUGT. There was a significant association between the risk of being admitted to a NH programme and older age, dependence on the IADL (moderate/severe), more prescriptions, and the presence of social risk, see Table 3.
Based on the results of the multivariate analyses, we constructed two prediction models according to the risk of inclusion in a HC/NH. Subjects with a HC risk were classified into three groups: lower risk with a prognostic index (PI) between 7.5 and 8.3; medium risk, between 8.4 and 8.9; and higher risk > 9.0. The NH model was also categorized into three groups: lower risk between 7.9 and 9.5; medium risk between 9.6 and 10.2; and higher risk PI > 10.3, see Table 3.
The calibration plot showed that both models (HC and NH) presented a good calibration for predicting risk outcomes. In addition, discrimination was good for a HC (C-index=0.7) and moderate for a NH (C-index=0.7), see Figure 2.
Figure 3 depicts the cumulative incidence for each of these prognostic groups for the two events. For a HC in the lower risk group the cumulative incidence was 10.0 per 1000 patients-year; for the medium risk one it was 25.6 per 1000 patients-year; and for the higher risk one 63.1 per 1000 patients-year (p <0.001). For a NH in the lower risk group the incidence was 1.9 per 1000 patients-year; the medium risk one was 4.2 per 1000 patients-year; and the higher risk one was 32.9 per 1000 patients-year (p <0.001).