The individual-level data were obtained from a randomised controlled trial concerning a geriatric rehabilitation programme designed for frail elderly persons 2122, which was conducted during 2002–2007 (Age-Study). The inclusion criteria were age of 65+ years, progressively decreasing functional ability, and risk of institutionalization within 2 years. The definition of frailty is based on the Pensioners’ Care Allowance benefit granted by the Finnish Social Insurance Institution (SII). This definition covers biological, physiological, social, and environmental changes. Our analysis is based on a sample of 732 frail elderly persons living in 41 municipalities. The sample characteristics are presented in Table 1.

* GDS: Geriatric Depression Scale, max 15, values 0–6 indicate non-depression.

† MMSE: Mini Mental State Examination, max 30, values under 24 indicate existence of dementia.

‡ 15D: Health-related quality of life (HRQoL), range 0–1, 1 indicates the best imaginable health.

§ FIM: Functional Independence Measure, max 126, three subscales (Self Care 8 items, Mobility 5 items, Cognition 5 items) were formed from 18 items (range: 1 = total assistance – 7 = complete independence).

The present study was approved by the Ethical Committees of the SII and Turku University Hospital. All of the study participants gave their written consent to the study. The individual-level data were linked with data from the national databases of the Care Registers of the National Institute for Health and Welfare (THL) and the Social Insurance Institution (SII). The registry data aren’t publicly available.

The utilization of health care services covered both the public and private sector care. Data on inpatient care and day case surgeries were drawn from the National Care Registers of the THL. Data on the use of outpatient care services within the private sector and the use of medicines were obtained from the SII registers. A self-reported questionnaire was used to collect information about the number of visits to the public sector outpatient care during the preceding year.

The utilization of social services covered institutional care and formal home-help services provided by municipalities. For those living in residential homes and sheltered housing, services such as home help, washing and cleaning were included in the analysis. For those who were living at home, formal home care, home nursing, and support services were included. Data on the utilization of social care services were obtained from questionnaires. We asked the municipal social and health care officials to collect data on service use from their clients’ individual care and service plans. The data derived from questionnaires were cross-sectional both at baseline and 12-month follow-up. For those cases where changes occurred in the use of services during the follow-up, the annual data consisted 6 months of services used at the baseline and 6 months of services used at the follow-up.

The total expenditure of health and social services utilization was determined by multiplying the use of services by their standard unit costs. For the monetary valuation of health and social care services, we used Finnish standard costs information 23. Because we used standard costs, any variation in the costs variables resulted from differences in the utilization of services.

To estimate the effects of independent characteristics on the costs of service utilization, we used the following three models:

Model 1: The costs of health care services and medicines (continuous variable, logarithm adjusted euro/year)

Model 2: The use of social care services (categorized into two groups; ‘yes’ if service use was observed and ‘no’ if otherwise)

Model 3: The costs of social care services used (categorized into three groups; under 1500, 1500–6000, or over 6000 euro/year)

Due to the fact that informal care is a common way to organize care for elderly, there were 240 subjects how did not use social services at all (‘No’ in MODEL 2). For those who used social services, living in sheltered housing causes stairwise effect to costs. Because the distribution was rising in steps and the assumptions of continuity did not hold, we categorized the costs of social care (MODEL 3).

Functional ability was assessed by the Functional Independence Measure (FIM), with the total score ranging from 18 (the lowest) to 126 (the highest level of independence) 24, and by the subject’s ability to perform the instrumental activities of daily living as measured by the IADL index, which varies between 7 and 21 (the higher the score, the more difficulties). Mood was measured by the Geriatric Depression Scale (GDS), where the maximum value is 15 and values 0–6 indicate non-depressiveness 25. Cognitive capacity was measured by the Mini Mental State Examination (MMSE), with the maximum value being 30 and values under 24 indicating dementia 26. The assessments were carried out by three accredited examiners.

Health-related quality of life (HRQoL) was evaluated by using the 15D score, ranging 0–1 (‘1’ indicates the best imaginable health) 2728. To estimate the impact of recent changes in health status, the subjects were asked “How has your health status changed during the last year?” Three class variables were created from the five alternative responses: worse (“much worse”, “worse”), same (“no change”), and better (“better”, “much better”). The questionnaires were checked by the examiner and any incomplete sections were completed by interviewing the participants.

To assess the living conditions, the subjects were asked “Is your residential environment rural or urban?” (yes/no) and “Are you living alone?” (yes/no). The dwelling circumstances were categorized into three categories: home without formal home-help, home with formal home-help, and residential home or sheltered housing.

The municipal social and health care officials were asked to collect information on informal support and assistance from their clients’ personal care and service plans. They reported "yes" if the person received informal care that was considered as supplemental to formal care, and "no" if otherwise.

A self-reported questionnaire was presented to the subjects to collect information on the availability and extent of informal care. Respondents were asked to answer the question “Is there any person who helps you when needed?” The answers were categorized into four categories; not available, occasionally, when needed shortly, as long as needed. Respondents were also asked whether they had been granted a Home Care Allowance by their local municipality.

For the purposes of the Age Study, the subjects were randomly assigned to either an in-patient rehabilitation programme or standard social and health care. In this study, to standardize the effect of the rehabilitation programme, we included rehabilitation as an explanatory variable in all our models.

At municipal level, we investigated if municipalities had integrated the administration of their health care and social services (joint budgets and management). We asked the municipal social and health care officials: “Is your local health and social care administration integrated or not?” (yes/no).

A hierarchical analysis with fixed (patient level) and random (municipality level) effects was used to estimate the effects of explanatory variables on the utilization and costs of health and social services. We used the SAS PROC GLIMMIX procedure to fit the models 2930.

First, bivariate correlations between independent variables were examined to check for correlations. All correlations were low. The independent variables were included in the multivariate model using the entrance value of P < 0.05 in the univariate analysis. All multivariate analyses were adjusted for rehabilitation in order to standardize the effect of the rehabilitation in the randomized trial setting. The effect sizes as a result of linear and logistic analyses were expressed as estimates (β) or odds ratios (OR) or cumulative odds ratios (COR) with 95% confidence intervals (CL) and the corresponding p-values.

For quantitative variables, skewness of distribution was analysed. Skewness was detected in health care costs, and a logarithm transformation was calculated to correct the distribution. In the analysis, we divided the variables concerning functional needs (FIM, GDS, MMSE, 15D, IADL) by their standard deviation to make them commensurable and to make it easier to compare the odds ratios. Age was divided by five in order to highlight the impact of every five-year increment in age.

At the baseline, the mean age of the subjects was 78 years (range 65–96 years). In the older age group (over 75-year-old subjects), the number of people living alone was higher and impaired cognitive capacity was found more often. Informal care and hospital admissions were also more common in the over 75-year-olds (Table 1). Due to these differences in need-specific factors, the over 75-year-olds were the largest client group for home care services.

The mean total costs of the utilization of health and social services were 10300 euro during one year (median 5400 euro) (Table 2). The proportion of health care costs was, on average, 48% of the total costs. Altogether 492 subjects had used social care services during the preceding year, and the mean costs of their social service utilization were 8700 euro (median 4200 euro) per year.

An analysis was carried out in two steps. First, an intercept-only model without any explanatory variables was applied. The intraclass correlation of the total costs of health and social care services was 0.06, meaning that roughly 6% of the variance is attributable to the municipality level and 94% to the patient level. When costs were examined for the health care and social care sectors separately, the variance in the costs of health care services was totally explained by individual differences in health; the covariance estimate for the municipality variable was zero. In the use of social services, a null model with two levels showed that 9% of the variability in the social services costs was explained by the municipality level and 91% by the patient level.

In the multivariate model, the health-related quality of life (15D score) appeared to be a powerful indicator for utilization of health care services (p = 0.0004) (Table 3). A statistically significant decline in HRQol 15D (0.02 units) leads to 1.5 folds increase in costs. The association between other explanatory factors and health care utilization was examined by fitting the multivariate model without 15D score. Independent living and better functional ability (FIM) decreased health care utilization (not tabulated).

*= P<0.05.

^† Living conditions is not included in the models of social care utilization because it is logically too close to the despondence variable.

¹ Coefficient corresponding change of standard error. The higher IADL scores indicates more difficulties.

² Coefficient corresponding change of standard error. The higher GDS scores indicates more more depressiveness.

³ Coefficient corresponding change of standard error. The higher MMSE scores indicates less demented.

⁴ Coefficient corresponding change of standard error. The higher 15D scores indicates better health.

⁵ Coefficient corresponding change of standard error. The higher FIM scores indicates highest level of independence.

Costs of the utilization of health and social services.

A logistic analysis was conducted to explain the use of social care services (Table 3). In the multivariate analysis, improvement in health status during the preceding year was associated with social services utilization. Instrumental activities of daily living (IADL), mood (GDS), dementia (MMSE) and functional ability (FIM) had a major impact on the use of social care services. Location of residence and health-related quality of life were not significantly related to the presence of social services utilization in the multivariate model. Every five-year increment in age and living alone increased the probability of social services utilization. Availability of informal care reduced the probability.

The costs of received social care services were categorized into three groups: annual costs under 1500, 1500–6000, and over 6000 euro (Table 3). When controlled for the confounding effect of the explanatory variables with a multivariate model, the caregiver’s financial support was most strongly associated with lower costs of social service use. Receiving informal care was also associated with a lower amount of used services. Poor ability in instrumental activities of daily living and low level of functional performance were associated with higher costs of service use. Living in a residential home or sheltered housing and living alone were associated with higher costs. The entrance value to multivariate analysis was P < 0.05, but we conducted the analysis also with the integrated social and health care sector variable included. The integrated social and health care sector was not significantly related to costs in the multivariate setting (not tabulated).

In the multivariate models, we introduced explanatory variables mostly at the patient level. At the municipality level, we focused on the question of integrated services. We measured integration with one question concerning the administrative integration, and this may be insufficient to describe the multidimensional phenomena of integration.

No other specific variables – such as proportion of elderly people, economical situation, or population size – were introduced at the municipality level, because we did not aim to explain differences between municipalities. Such variables should, however, be taken into consideration in future studies.

While frail aged people with deteriorated health must be provided care as soon as any need occurs, the results of this study cannot necessarily be generalized to apply to the aged population as a whole.

In lack of national registers, data on municipal primary health care and social care services were collected through questionnaires that were completed by the elderly participants and by the representatives of the local social and health care units. In Finland, municipalities are the main provider of services for the elderly, and therefore, municipal records on service use are comprehensive and reliable. Furthermore, a vast majority of the costs were calculated from registry data, which in Finland are regarded to be very reliable 38.