This is a longitudinal one-year follow-up study of a cohort of elderly patients living at home (EDPLH). The methodology of the study has been published elsewhere [6, 18]. This study was conducted in Catalonia, a region in Spain with a National Health Service that provides health care to every citizen free of charge at the point of delivery. Primary care in the region is provided by 340 Primary Care Teams (PCT), with the participation of family doctors and community nurses, each of them covering an area ranging from 5 to 25 thousand inhabitants. Each individual PCT delivers a home care programme addressing the needs of patients who cannot pay a regular visit to a primary care centre. This is done in addition to the regular services provided within primary care facilities, which consists of basic medical assistance and nursing care as well as an emergency service, both at the primary care centre and at the patients’ home. In addition, other centralised 24h emergency services are available, activated by patients’ telephone calls.
The research protocol of this study was approved by the Ethics and Research Committee in the Jordi Gol i Gurina Primary Care Research Institute.
Criteria for inclusion required patients to be non-institutionalised, chronically ill , 65 years of age and over, who were unable to autonomously seek care in a primary health care centre. Patients selected for the study were already included in a home care programme. The recruitment period was between the 1st July and the 31st December 2005. Patients who had been admitted to hospital care before the baseline assessment were considered as a subgroup and analysed separately.
Patient exclusion criteria included refusal to participate, transitory patients (followed by the primary care team for less than 9 months/year), patients with a life expectancy below 4 months, patients receiving transitory post-surgical care, and when both the patient and the carer had been diagnosed with dementia.
Informal carers are defined in the study as those looking after the basic needs of the patient, receiving no specific remuneration for that task and having a close relationship (next-to-kin or friend) with the patient.
Regarding patient recruitment, each of the 378 primary health care professionals who agreed to cooperate in the study selected, on average, 3 patients fulfilling the inclusion criteria. Participating professionals work throughout the Catalan territory, including rural and urban areas. They collaborated in the study on a voluntary basis and no randomisation was performed. They were asked to include in the study at least the first three patients that were found eligible under the agreed inclusion and exclusion criteria. Each of these professionals acted as collaborative researcher and a member of the research team trained them in their tasks. Health professionals were responsible for the administration of the questionnaires used in this study and the gathering of the necessary patient data from clinical visits, clinical records and reports from other health care providers.
The following data were collected from each patient at baseline and after one year follow-up: patient clinical characteristics, comorbidity level (Charlson test) , functional status (Barthel test) , cognitive status (Pfeiffer test) , existence of decubitus ulcers, risk of appearance of ulcers (Braden test) , subjective health status (SF-12) , social risk (Gijon Test)  and carer burden (Zarit test) . All the questionnaires mentioned above have been validated for use in a Spanish context.
In addition we looked into variables that measured the utilisation of health and social services including hospital admissions (defined as the patient remaining in acute hospital care for over 24 h), emergency room visits, home emergency visits data, primary care and community services activity data. The utilisation of social services included home help, tele-assistance, meals on wheels, volunteer care, social work visits, day centre and nursing homes, among others. These variables were collected for all patients one year before the baseline assessment and during the one year follow up period.
The utilisation of health and social services the year before the baseline assessment was extracted from patient’s clinical records. Collaborative researchers in the study also collected the same utilisation variables prospectively during follow up.
Researchers participating in the study received training for data collection standardisation and collected data were analysed centrally. A continuous data quality assurance process was in place in order to minimize errors and information losses. A research team member audited and validated 10% of the data gathered by collaborative researchers from patients’ original clinical records. The Kappa index and interclass correlation index were used to test the concordance among registers . The results showed an adequate concordance.
A descriptive analysis is provided including average and standard deviation for all continuous variables. We analysed baseline differences between individuals who died and those who did not die during the study period by means of the Mann–Whitney-Wilcoxon non-parametric test. Categorical variables were analysed using the chi-square test and Fisher’s exact test. We agreed on a p-value significance level of p < 0.05.
Following a bivariate analysis we included significant variables in a logistic regression model, in which the dichotomous dependent variable was vital status. We used stepwise techniques for inclusion and exclusion of variables in the model, testing the significance of each individual variable by means of a deviance analysis. We ran two separate analyses, one including all patients in the sample, and a second analyses including only patients with an informal carer. Accordingly, carer-related variables were excluded from the first analysis since not all patients had an informal carer. The statistical results are expressed in terms of odds ratios (OR) and 95% confidence intervals (CI). The statistical package used was R .
We fitted a second logistic regression model in order to identify those variables that best explained the event of death among those patients that were hospitalised the year before entering the home care program. The resulting predictive model shows the probability of death among home care patients that were hospitalised before entering the program, in terms of patient’s characteristics, carer’s and service variables. We used the Hosmer-Lemeshow test to report the goodness of fit and ROC curve analysis to assess the discriminatory power of the model.