Background

Many elderly will experience a reduction in physical function, leading to more falls and injuries. This leads to a loss of independence, hospitalisation, long-term nursing home care as well as premature death [1, 2]. In 2013, 2.63 million people in Germany were in need of nursing care. Roughly two third of them received ambulatory nursing care and one third received care in a nursing home. From the years 2011 to 2013, people receiving nursing care increased by 5% [3]. In 2014, 27% of the population aged 65–79 received either ambulatory or institutionalised nursing care [4]. Given the projected demographic changes in Germany, the population in the age group ≥65 years will increase from 21% (17.3 million) in the year 2015 to 27% (21.8 million) in the year 2030 [5]. Long term nursing care and institutionalisation in a nursing home can be associated with a significant reduction of quality of life, mainly due to loss of autonomy and social contacts [6]. The increased need of long term nursing care poses a major economic challenge [7]. Thus measures to prevent, minimise or delay long term nursing care for elderly are urgently needed.

Germany has mandatory nursing care insurance attached to statutory health insurance since 1995. Elderly with disabilities or dementia can apply for a nursing care level [Pflegestufe]. To identify elderly in need, a basic geriatric assessment performed in general practice was introduced in 2005 [8]. Although elderly are entitled to rehabilitative services as codified in the Book V (SGB V) and XI (SGB XI) of the German social code general practitioners have only limited access to specialised rehabilitative services for geriatric patients. While the number of hospitals providing geriatric services is increasing, the demand for ambulatory services is not met [9]. Elderly patients with a need for rehabilitation prefer to stay close to their home and relatives, maintaining their everyday life [10]. Consequently, ambulatory rehabilitation is preferred. Rehabilitative services were and are still mainly available after hospitalisation e.g. for stroke or fracture after a fall.

Preventive ambulatory geriatric rehabilitation (AGR [Ambulante Geriatrische Komplexbehandlung]) was introduced in 2008 within the legal frame (§ 140 Book V of the social code) of selective contracts for integrated care. Therefore it is only available as a model intervention in some areas for holders of specific statutory health insurances (e.g. AOK Nordost). AGR is not part of regular health care.

It is intended as a community based outpatient intervention to improve patient’s physical function, increase patient’s safety and quality of life as well as to prevent falls and injuries, to avoid and delay hospitalisation, the progression of nursing care level and admission to nursing home.

A systematic review of controlled trials of ambulatory and hospital interventions to improve physical function and maintain independent living in elderly people concluded that they were effective to achieve the goal [1]. This review comprised only one German trial which included geriatric patient post hospital discharge [11]. However, AGR is primarily intended to prevent hospitalisation and not as a post-discharge rehabilitation. The effectiveness of German AGR programs has not been evaluated rigorously yet. Previous studies have relied on uncontrolled study designs [12].

Objectives and hypotheses

The aim of our study is to evaluate the effectiveness of AGR regarding patient’s progression to higher nursing care levels, incident fractures, admission to nursing home, hospital admissions as well as health care costs. For this purpose we compare patients receiving AGR with patients receiving care as usual. We will estimate average treatment effects based on a cohort design using propensity score techniques to match cases and controls. The follow-up period will be up to 2 years.

Our primary hypotheses are:

  1. 1.

    AGR reduces and delays progression to higher nursing care levels.

  2. 2.

    AGR reduces and delays nursing home admissions.

  3. 3.

    AGR reduces the risk of incident fractures.

Our secondary hypotheses are:

  1. 1.

    AGR decreases and delays hospital admissions and reduces the days spent in hospital during follow-up time.

  2. 2.

    AGR decreases and delays ambulatory care sensitive hospital admissions.

  3. 3.

    AGR decreases total health care costs from the statutory health insurance perspective.

On an exploratory basis we will investigate the effect on drug prescriptions.

Methods

Study design

The conduct of a randomised controlled trial to assess the effectiveness of AGR is currently limited due to logistic and ethical reasons. Therefore we will conduct a matched cohort study using claims data. Anonymised data will be provided by the statutory health insurance AOK Nordost which comprises basic demographic data, data on nursing care level, admission to a nursing home, billing data for ambulatory services (EBM position numbers) and for hospital services (DRG-codes/OPS-codes), as well as diagnoses (ICD-10 codes), and all health care costs, including costs for hospitalisation, remedies and aids, ambulatory costs and costs for medication. To balance potential confounders, we will apply a propensity score matching. Controls will be matched patients insured by the AOK Nordost.

The observation period comprises 4 billing periods (each 3 months, corresponding to 1 year) prior to the intervention, the intervention (index) billing period, and up to 8 billing periods post intervention, resulting in a total observation period of up to 13 billing periods. The billing period covering the most days of the intervention will be considered as the index billing period. Participants in the intervention group received a 4 weeks AGR in between the years 2009 and 2013.

Description of the AGR intervention and the setting

AGR is a multimodal intervention consisting of physiotherapy, ergotherapy, speech therapy, occupational therapy, social support by qualified social workers, psychological counselling and counselling regarding aids and care. Patients who are deemed suitable for AGR by their general practitioners can be referred to special rehabilitation centre for a geriatric assessment. If they fulfil eligibility criteria for AGR and agree to participate they receive the intervention. The intervention is tailored to the patients’ needs and delivered in individual and group sessions. Patients are commonly treated for a total of 20 days with two to three 30 min therapy units per day. Included are meals and a pick up and return service for the elderly every day. During the intervention period AGR was available in three locations in Mecklenburg-Vorpommern (Trassenheide, Ueckermünde, Waren) (Fig. 1).

Fig. 1
figure 1

Map of Mecklenburg-Vorpommern with the three locations which provided AGR

Full size image

Eligibility criteria for AGR

Due to selective contracts only patients insured by the AOK Nordost are entitled to receive AGR. The geriatric assessment comprises the activity of daily living scale (Barthel-Index), instrumental activity of daily living, Timed “Up & Go”, Chair-Rising-Test, Tandem Stance, Berg-Balance-Scale, and handgrip strength. Eligibility criteria are:

  1. 1.

    Aged 70 and older

  2. 2.

    at least two geriatric multimorbidity listed in Table 1

  3. 3.

    impairment or handicap with functional deficits

  4. 4.

    at least one of the health conditions listed in Table 2

Table 1 Geriatric multimorbidity
Full size table
Table 2 Eligibility criteria for participation in AGR
Full size table

Patients are not eligible for AGR if the assessment indicates a need for hospital admission, if they are not able to participate due to a poor health status or if they are unable to provide informed consent.

Exclusion criteria from the analyses

We aim to quantify treatment effects for typical AGR patients. Therefore, we will exclude participants with rare conditions which are likely to affect the health course beyond the effects which can be reasonably assumed from AGR (e.g. organ transplantation, dialysis, chemotherapy) or participants with extremely high health care costs indicating severe conditions. Prespecified criteria used to exclude AGR participants from our analyses are shown in Table 3. Exclusion criteria for AGR participants also apply to potential controls.

Table 3 Criteria for the exclusion of AGR participants for study analysis
Full size table

Sample size

The intervention group consists of all AGR participants, receiving the AGR in Mecklenburg-Vorpommern during the years 2009–2013. This will comprise approximately 700 patients. Controls will be chosen from a pool of around 250.000 members of the AOK Nordost aged 70 years and older. Approximately 2800 controls will be selected from the pool.

Outcome measures and data collection

Claims data will comprise the billing periods from 01.01.2008 until 31.12.2014, thus ensuring a minimal observation period of 12 months prior and after the index billing period. All variables are provided for each billing period during the observation period.

The primary outcomes are ‘progression of nursing care level’, ‘admission to nursing home’ and ‘incident fractures’.

Four nursing care levels (0–3) are defined by the German social code XI (SGB XI). The nursing care level needs to be approved by the Medical Review Board of the statutory health insurances (MDK) in a standardised procedure. The definition for nursing care level is described in Table 4 [13]. Elderly who do not reach the requirements for nursing care level I, but need help for daily living (SGB XI § 45a) receive care for their needs. This is referred to as nursing care level 0 [13].

Table 4 Definitions of the nursing care levels [13]
Full size table

The outcomes ‘admission to nursing home’ and ‘incident fractures’ are both coded as binary variables, stating whether the patients experienced the event after the index billing period or not.

Secondary outcome variables are ‘any hospital admission’ (yes/no), ‘days spent in the hospital’, ‘ambulatory care sensitive hospital admissions’ (yes/no) and ‘total health care costs from the statutory health insurance perspective’ during the entire follow-up period and during each billing period in the follow-up period. Ambulatory care sensitive hospital admissions are defined as potentially preventable hospital admissions by interventions in primary care and are displayed in Table 5. Conditions were chosen from two studies [14, 15].

Table 5 Ambulatory care sensitive conditions used to define hospital admissions to be prevented by AGR (according to [14, 15])
Full size table

‘Total health care costs’ comprises expenditures for hospitalisation, remedies and medical aids, ambulatory costs, and medication. The variables concerning costs refer to costs excluding out-of-pocket spending except for remedies and medical aids where the available data does not allow for any separate analysis of out-of-pocket spending.

Matching and statistical analyses

To balance the distribution of potential confounders among cases and controls we will conduct a propensity score matching using a many-to-one matching [16]. Variables for the estimation of the propensity score will be selected based on their expected importance to predict AGR participation as well as the outcomes of interest.

We will perform a two-step matching process. In the first step, for all patients who received AGR in a determined billing period, we plan to match controls with similar morbidity and cost characteristics during a.) the four billing periods prior to the index period in which the intervention took place, and b.) by additionally including the index period in the matching. This first step allows for the definition of an index period in controls. Using variables during the index billing period is complicated by the fact that no information on the temporal sequence of events within a billing period is available due to legally obliged data protection agreements. However, ignoring the index period may lead to a systematic ignorance of events leading to AGR which might have taken place in the index period. Therefore results from both analysis scenarios (a., b.) will be systematically compared. The first-step will be conducted as an exact match (pre matching) on selected variables listed in Table 6. These variables were selected due to their high expected conceptual importance for the estimation of treatment effects. Should no adequate numbers of controls be found, a more lenient matching may be employed. Controls will be drawn with repetition across billing periods. We plan to match up to 100 controls per case at this stage.

Table 6 Matching criteria
Full size table

Subsequently, propensity scores will be calculated using a logistic regression model using coded morbidity and costs under the scenarios a.) and b.), based on all variables listed in Table 6, taking statistical interactions and nonlinear associations into account. Up to four controls may be assigned to one case without repetition.

Appropriate regression models (e.g. time-event models, mixed models, two-part models) will be applied to study effects on our primary and secondary outcomes. Because treatment might also affect censoring due to mortality, competing risks models (using the Fine-Gray approach) will be applied [17]. Patients dying during the follow-up period will not be excluded from the analyses.

Discussion

AGR is currently only available in few areas for elderly people from selected statutory health insurances which opted to offer AGR to their beneficiaries. Evaluation of this intervention was previously limited to uncontrolled study designs [12]. To the best of our knowledge, this will be the first study to evaluate AGR using a quasi-experimental design. This study may provide an estimation of the effectiveness of AGR on progression to higher nursing care levels and hospitalisation and other endpoints of clinical relevance. Our results will be important for providers of AGR, policy makers and stakeholders to make informed decisions on whether to continue, modify or even expand AGR to other areas in Germany. Additionally, our results might help to optimise AGR by identifying subgroups of patients who are more likely to benefit from AGR.

A limitation will be the restriction to claims data. Clinical measurements from instruments used for the geriatric assessment are not available for the control group. This might impair the quality of the matching as potential imbalance between clinical data and individual motivation to participate in AGR cannot be ruled out even if a high balance on claims data is achieved. Therefore, residual confounding may still be an issue after our matching. Final decisions on the applied methods need to account for the precise properties of the data.