A total of 1,093 EDPLH patients were included in the study. Patients that completed the study (1,001) had a mean age of 83.7 years (±6.8 sd.), 66.8% were female (729). 290 patients died during the one year follow-up (28.9%).
As reported in Table , patients lost to follow-up (92) had similar characteristics to patients who were followed up (1,001), with the exception of the former receiving proportionally more rehabilitation at home and having a higher average number of days temporarily admitted to nursing homes.
Differences between patients lost to follow-up and followed-up patients
Regarding personal characteristics, health status and life-styles of patients included in the study, Table shows there are some significant differences between those who died and those who survived until the end of the study. Indeed, female gender is a relevant variable in the study associated to survival both in the whole population as well as in the hospitalized subgroup. Similarly, individuals who survived in both groups showed lower comorbidity, higher levels of autonomy, less risk of pressure ulcers and lower prevalence of pressure ulcers. In the entire sample –in contrast to what was observed in the sub-group of patients admitted to hospital–, survival was associated to less cognitive impairment and higher score in physical and mental self-perceived health status.
Patients’ characteristics, health status and lifestyles
Furthermore, patients who died during follow-up received a higher proportion of informal care, largely provided by women carers, and these informal carers reported a higher burden of care as measured by the Zarit test (see Table ). In contrast, informal care was not found to be a significant variable in explaining survival among patients that were hospitalised the year before they entered the home care program.
Informal carer characteristics and overburden at the basal assessment
With respect to health and social services utilisation (Tables and ), patients dying during the follow-up year received largely the same amount of services in the year before their death as those who survived, with the exception of a lower use of tele-assistance services (8.3% vs. 13.6%) and hospital at home services (1.4% vs. 1.6%), and a higher use of health centre emergency services (17.2% vs. 10.3%), emergency community services (24.8%, 15.3%), and in-patient care (1.6% vs. 0.9%). In the sub-group of 226 patients admitted into hospital before entering the home care program, no differences were found in health and social services utilisation (Tables and ) except for a higher use of centralized after-hours emergency community services among patients who died.
Services received the year before the basal assessment
Type of services received by EDPLH patients during the one year follow-up period
Variables found to be independently associated to the risk of dying during follow-up in this EDPLH group (N=1,001) were male gender, comorbidity (as measured by the Charlson test), the number of hospital admissions the year before, and both the existence of and degree of pressure ulcers (see Table ). When considering the EDPLH group with an informal carer (N=821) we found the same variables associated to risk of dying as in the total sample with the exception of gender, and two additional variables, namely self assessed health status (as measured by SF-12) and burden on the carer (as measured by the Zarit test) (see Table ).
Independent risks of dying during the following year (N=1,001) (logistic regression analysis)
Independent risks of dying during the following year in population with informal carer (N = 821) (logistic regression analysis)
Variables that predict the risk of dying among those patients that were hospitalised before entering the home care program were male gender, comorbidity (as measured by the Charlson test), having received home care rehabilitation, and both the existence and degree of pressure ulcers. The resulting predictive model is represented by the formula below. The Homer-Lemeshow goodness of fit test is p=0.4837. The probability of dying during the next year can be calculated by the following equation:
The area below the ROC curve is 0.754 (95%CI=0.689-0.820) (see Figure ). The sensitivity and specificity values are 65.12% and 71.11%, respectively. True positive predictive value is 0.59 and true negative predictive value is 0.76.
Predictive model of death in EDPLH patients admitted to hospital the year before entering the home care program: ROC curve.