Effect of an osteoporotic fracture prevention program on fracture incidence in routine care: a cluster-randomized trial

The implementation of the program took place in 47 administrative districts in five federal states (i.e., Baden-Württemberg, Bavaria, Hesse, Lower Saxony, Rhineland-Palatinate). These federal states cover a large percentage of the whole area of Germany. All administrative districts of the five federal states were randomly assigned to either the implementation or control group (cluster randomization; 47 intervention districts, 139 control districts). Randomization was stratified by federal state. The control districts did not receive any intervention within the program (Fig. 1A).

The program was offered as a new component of routine care to (a) all community-living women and men aged 70 to 85 years with a history of a fragility fracture (femur, spine, forearm, shoulder/upper arm, lower leg, or pelvis) in the previous 5 years and (b) all community-living women aged 75 to 80 years who lived in the intervention districts and were insured by the SVLFG (complete sampling). The choice of these two subgroups and their age ranges followed different considerations: (a) to meet the recommendations of the German osteoporosis guideline for bone health evaluation (https://​dv-osteologie.​org/​osteoporose-leitlinien), (b) to capture people with a high imminent fracture risk, and (c) to combine the complete sampling approach with the staff capacities of the health care fund. Insured individuals were excluded if they lived in a care home or if their formal home care need for the basic activities of daily living was 120 min or more according to the categorization of the German long-term care insurance (level of care: grade 2 or 3). Individuals of the intervention and control districts were identified from the health insurance records in August 2015 and then subsequently recruited and entered into the study and contacted about the interventions over a 2-year period (1 October 2015–30 September 2017). If one of the exclusion criteria or death occurred before study entry, the selected individual was excluded. The staggered study entry was needed to manage the resource available to provide the intervention. Insured individuals aged 70 to 85 years who were not selected in August 2015 but had an incident fragility fracture during the recruitment period were additionally included. Inclusion and exclusion procedures were handled identically for individuals from the intervention and control districts.

The program had 3 components: 1)  “Trittsicher”-mobility and falls prevention exercise classes, 2)  the examination of bone health by a bone mineral density measurement (dual-energy X-ray absorptiometry (DXA) scan), and 3) a consultation about “safety in the living environment.” Details are described in Table 1 and in the study protocol [11].

The selected individuals in the intervention districts were informed about the program by a letter by the health care fund and visited one to four weeks later at home by a “prevention manager” who motivated them to attend an exercise class or to make use of a DXA measurement. If required, recommendations to increase safety in the nearby living environment were given. The prevention manager contacted the insurance telecentre and informed them of the individual’s willingness to attend exercise sessions. The telecentre referred individuals who were willing to attend exercise sessions to a class taking place near their home and informed them about DXA devices in the surrounding area. Individuals received a final report in which their preferences regarding the components of the program, dates, and arrangements were summarized and information sheets for the GP were included.

One of the objectives of OFRA was to build up an infrastructure of “Trittsicher”-mobility and falls prevention exercise classes. About 700 exercise instructors received a mandatory training course in the year before the start of recruitment. The exercise classes were generally free of charge and also open to older people who were not selected for the scientific program. During the 2-year recruitment and 1-year follow-up period, 2339 “Trittsicher”-mobility and falls prevention exercise classes with a mean number of 11.6 participants per class were implemented in the intervention districts [15].

The primary outcome was a measure of incident fragility fractures (ICD-10 S12, S22, S32, S42, S52, S72, S82) requiring hospitalization and recorded as main diagnosis in the routine health claims database of the insurance company after study entry. Not à priori defined explorative endpoints were the fractures of each entity separately (femur (S72), spine (S12, S22.0, S22.1, S22.2, S22.3, S22.4, S22.5, S22.8, S22.9, S32.0, S32.7, S32.82), forearm (S52), shoulder/upper arm (S42), lower leg (S82), pelvis (S32.1, S32.2, S32.3, S32.4, S32.5, S32.81, S32.83, S32.89)), death, and nursing home admission. Uptake of a “Trittsicher”-mobility and falls prevention exercise class, reimbursed DXA measurements, new prescriptions of specific anti-osteoporotic drugs, and advice given about measures to increase “safety in the living environment” served as process parameters.

The study database was merged with the routine data collection system of the health insurance company to gain data on incident fractures, reimbursed DXA measurements, new prescriptions of specific anti-osteoporotic drugs (bisphosphonates and denosumab; ATC-codes: M05BA, M05BB, M05BX) prescribed after discharge and dispensed by community pharmacists, nursing home admission, and death. Information about uptake of a “Trittsicher”-mobility and falls prevention exercise class was derived from a telephone interview of a random subsample of 780 selected individuals of the intervention districts since this information was not available in the routine data collection system [16]. The information on advice given to individuals on “safety in the living environment” was extracted from the records of feedback from the prevention managers to the telecentres.

The date of study entry of the individuals of the intervention group was defined by the date of a first information letter sent by the health insurance company. Since the letter was usually followed by a visit of a prevention manager, a recruitment period of 2 years was needed. To align data collection periods between the intervention and control groups and to receive a “study entry date” for the control-persons, for each individual in the intervention arm we identified people in the control districts to provide data for the equivalent time period. Federal state and target group (women and men 70 to 85 years with prior fracture; women 75 to 80 years) were used as matching criteria, resulting in 10 strata. Within each stratum, for each individual of the intervention districts, an average of three individuals from the control districts (according to the randomization balance) was matched randomly with replacement. This procedure was performed for individuals selected in August 2015 and for those with an incident fragility fracture during the recruitment period in the same way. The SAS-PROC SURVEYSELECT was used to perform the matching procedure.

The time-period between the information letter and the start of an exercise class was usually 2 months or more. Therefore, follow-up began 2 months after the date of the letter in the individuals of the intervention districts (independently if they participated in an exercise class or not) and 2 months after the assigned study entry in the matched individuals of the control districts and was 12 months or until censoring due to a fracture death or study ending. In an additional sensitivity analysis, the maximal available follow-up time of each included individual was used which is one up to 3 years (in median 1.7 years).

All selected individuals from the intervention districts and the matched individuals from the control districts were included in the analyses. This means that individuals from the intervention districts who did not receive any component of the intervention were also part of the analyses (modified intention to treat approach). Cox frailty analyses were used to calculate hazard ratios (HR) and 95% confidence intervals (95% CI). All models included federal state in the strata statement to account for the stratified randomization. To account for the cluster-randomized design of the study, a random effect-statement including the administrative districts was applied. No further adjustment was performed. Since the intervention did not influence mortality, death was not considered as competing risk. The proportional hazard assumption was checked applying proportionality tests and by inspection of Schoenfeld-residuals and found to be fulfilled for all models. Rates per 1000 person-years were calculated for the total study population and stratified by sex, age, and a history of fracture. All calculations were performed using SAS 9.4.

The study protocol was published in 2016 [11]. The study was registered at the German Clinical Trials Register (DRKS-ID: 00009000) and approved by the ethics committee of Ulm University (proposal 120/15).