Health-related quality of life and associated factors after hip fracture. Results from a six-month prospective cohort study

Study design

EMAAge is a multicenter, prospective, observational cohort study focusing on emergency patients with hip fractures. It is based on standardized interviews with patients or proxies at baseline and a follow-up interview supplemented by clinical data on diagnoses and treatment. The observational period was six months.

Due to the observatory and exploratory character of the study, the sample size was calculated based on feasibility considerations and a generalized power calculation. Given the average number of hip fracture cases in the recruitment area, the target sample size was set at 350.

The study is part of the research network EMANET—Emergency and Acute Medicine Network for Health Care Research, which is funded by the Federal Ministry of Education and Research (BMBF). It involves the eight Emergency Departments (ED) in the central district of Berlin, Germany (Berlin-Mitte) (Schmiedhofer et al., 2018). Two EDs are part of a university hospital, six of general hospitals.

The study protocol was registered in the German Clinical Trials Register (DRKS00014273). The ethics committee of Charité—Universitätsmedizin Berlin (EA1/362/16) approved of this study.

Setting and participants

Participants were consecutively recruited between June 2017 and June 2019 in six of eight hospitals of the network which provide orthopedic & trauma surgery. All patients admitted to one of the ED with “hip or pelvic pain” and “fall” were identified and assessed for inclusion. Patients were eligible for enrollment if (1) 18 years or older and (2) having an ED diagnosis of hip fracture based on the International Classification of Diseases, 10th revision (ICD-10): S72.0, S72.1 and S72.2.

The exclusion criteria were (1) refusal of the patient or their proxy to participate, (2) life-threating conditions during hospitalization, (3) limited proficiency in one of the questionnaire languages (German, English, Arabic, and Turkish).

Perioperative care was provided in general EDs, Intensive Care Units (ICUs) and orthopedic & trauma wards without special focus on geriatric care.

Patient recruitment

Eligible patients were approached by trained study nurses during the course of their in-hospital stay and after initial treatment of the fracture. After education and given written consent, study nurses carried out a standardized, tablet-based, face-to-face bedside interview with participating patients. Responses were entered by the interviewer and were directly transferred to a secure database server. When patients were unable to consent to participate and to provide reliable information due to cognitive impairment or dementia, a proxy was asked to give consent and information on behalf of the patient. This was either a close relative or a legal guardian. The latter were provided with a short version of the questionnaire excluding those items only a person itself or a close relative could assess. Proxy versions of survey instruments were used if available.

Clinical data on ED and in-hospital care and diagnoses from hospital information systems were assessed. Hospital data were systematically collected after discharge and entered into a standardized electronic clinical report form (eCRF). Manual double data entry of clinical data was performed to ensure reliability and accuracy. The follow-up survey with patients or their proxies was conducted six months later, preferably via telephone or alternatively postal questionnaires.

Data

For the multivariable analysis of factors associated with six-month HRQOL, we considered patient and health care characteristics that have been under investigation in previous studies. In detail, we assumed that the outcomes are influenced by the previous burden of disease and disability, by typical aspects of the geriatric syndrome such as dementia, malnutrition, and mental health. Furthermore, we assumed that complications during the initial hospital stay that led to ICU treatment and the provision of follow-up rehabilitation might have an effect on the recovery of HRQOL. We also sought to examine the role of social factors such as the existence of resources of social support, educational resources, and migration background.

In addition, we adjusted for sex, age, type of hip fracture, type of surgery, study center, the baseline levels of EQ-5D-5L and the baseline hip functionality.

Basic patient characteristics such as sex and age, as well as diagnoses including all diagnosed comorbidities, surgery procedures, and hospital processes such as ICU care or discharge destination were extracted from hospital information systems and medical files. We used self-designed questionnaire items to assess fracture etiology, history of falls, and patients’ living situation.

We assessed the burden of comorbidities by reviewing all documented diagnoses according to the Charlson Comorbidity Index (CCI) (Charlson et al., 1987), a well-established index weighting and classifying the number and seriousness of comorbidities. We used the unadjusted index version of the CCI, and four categories. As a second indicator for the general disease severity and its consequences, we analyzed the number of medications and categorized five and more drug substances as polypharmacy.

Pre-fracture care dependency was determined according to the German long-term care insurance. People with a dependency in their activities of daily living (ADL) are entitled to financial benefits. After a thorough assessment by medical officials, applicants are assigned to a category that describes their degree of dependency and according benefits (Schnitzer et al., 2017). Up until 2016, there were three basic levels, after a law reform five degrees of care dependency were defined—each with higher values representing higher need for support and higher benefits. This transition coincided with the study period, so both systems had to be applied and set into relation. We used the self-reported care dependency status and crosschecked with eCRF data.

Pre-fracture hip functioning was assessed by using the Oxford Hip Score (OHS), a self-report questionnaire developed for patients undergoing total hip replacement (Murray et al., 2007). Its validity and reliability have been confirmed, including for the German version (Naal et al., 2009). The instrument consists of 12 items with responses on a five level Likert scale. The sum score ranges from 0 to 48 (best functionality). It has been recommended as a disease-specific outcome measure in hip fracture patients (Hutchings, Fox & Chesser, 2011).

Signs of malnutrition were screened by using the Short Nutritional Assessment Questionnaire (SNAQ) (Kruizenga et al., 2005). Symptoms of depression and anxiety were assessed by the short screener from the Patient Health Questionnaire (PHQ), the PHQ-4 (Kroenke et al., 2009). The score ranges from 0 to 12 (severe), and moderate or severe symptoms are assumed from a value of 6.

For social support, we adopted one item from the Oslo-3-Items-Social-Support Scale assessing the number of people one can rely on (Dalgard, 1996), and a 3-items-questionnaire from the German Ageing Survey (DEAS) assessing need for more help and support (Engstler & Schmiade, 2013).

For educational and vocational attainment, we used standard questionnaire categories and the CASMIN Educational Classification (Comparative Analysis of Social Mobility in Industrial Nations) to classify responses (Brauns, Scherer & Steinmann, 2003). We stratified educational attainment into: basic (1a –1c: primary and low secondary), intermediate (2a –2c: intermediate and high secondary) and high (3a and 3b: tertiary education).

A basic set of indicators was used to record migration background (Schenk et al., 2006). In accordance with the definition of the Federal Statistical Office of Germany, a migration background was assumed when the participant or at least one parent was born outside Germany. People from former German territories in Eastern Europe who had to leave their home during or after World War II were also included in this group although being born “German” since they had experienced an episode of migration.

The study center variable was dichotomized into university vs. general hospital due to confidentiality reasons and due to uneven distributions of case numbers.

Outcome measures

Health-related quality of life was determined by the generic and widely used EQ-5D questionnaire which has been recommended for studies examining hip fracture outcomes (Parsons et al., 2014). It consists of two components: the first one indicates the health states in five dimensions, including mobility, self-care, usual activities, pain, and anxiety & depression. For each dimension, participants are asked to rank their personal level. The EQ-5D-5L extends its previous version from three to five possible levels ranging from “no problems” (1) to “extreme problems/unable to” (5) (Herdman et al., 2011). Answers to the five dimensions form an individual health profile that can be assigned a summary index score based on societal preference weights for the health state (EuroQol, 2019). The German value set for the EQ-5D-5L was used to calculate the EQ-5D Index value scoring from −0.661 to 1, with values below zero indicating a health state worse than death and 1 indicating full health (Ludwig, von der Schulenburg & Greiner, 2018).

The second component of the EQ-5D is the EQ Visual Analogue Scale (VAS) recording participants’ self-rated health on a scale from 0 (“worst imaginable health”) to 100 (“best imaginable health state”). At baseline, items were modified to assess patients’ unimpaired HRQOL: participants were asked to rate their state of health during the four weeks before the fracture. For non-German speaking participants, the English, Turkish, and Arabic version of the EQ-5D-5L, and for proxy interviews, the EQ-5D-5L proxy version was used.

Cases of death between baseline and follow-up were ascertained by inquiries with the local residents’ registration offices, which hold complete records of all decedents.

Data analysis

Descriptive analyses were performed for patient characteristics and outcome measures, stratified for women and men. For the outcome measures, mean differences between baseline and follow-up values were calculated, both for the individual dimensions of the EQ-5D-5L, the Index value, and the EQ-VAS scores.

For EQ-5D index and EQ-VAS scores multivariable linear regressions were performed. The dependent variables were the follow-up values of these scores. The independent variables were: Baseline value of the respective EQ-5D score, sex, age, proxy-interview (cognitive impairment), educational status, migration background, living situation, comorbidities (CCI), polypharmacy, pre-fracture dependency, pre-fracture hip functionality (OHS), malnutrition, symptoms of depression and anxiety (PHQ-4), social support, subjective need for more support, type of fracture, type of surgery, ICU episode, referral to a rehabilitation facility. Data comes from six different sites, two university and four non-university hospitals. As the number of patients from two sites is very low (n = 4), a mixed model with site as random effect was not applicable. Therefore, the dichotomous variable “study center” was added to account for the data structure.

Regarding the model assumptions, the Variance Inflation Factor and the distributions of the independent variables were analyzed.

For these analyses a multiple imputation was performed. The imputation model includes all independent variables of the analysis model as well as baseline and follow-up value of the EQ-5D index and EQ-VAS scores. The multiple imputation resulted in 10 imputed datasets and the number of iterations was restricted to 20. Results were pooled using Rubin’s rules. Standard errors are reported for estimates and 95% confidence intervals for r-squared values.

The analyses included all surviving patients until follow-up, including living patients that only participated in the baseline interview.

Sensitivity analyses were done for the subset of patients being 65 years old and older, and for a cohort including only patients who participated in the follow-up interview.

All multivariable analyses were done using R version 4.1.2, for imputation the mice package was used (van Buuren & Groothuis-Oudshoorn, 2011).

Due to the exploratory character of the models, we did not apply a certain p-value level to assume variables to be worth considering as relevant for HRQOL at six months after a hip fracture.

Baseline characteristics

Patient and clinical characteristics, stratified for women and men, are displayed in Table 1. The majority of patients were women (67%), mean age of the cohort at time of ED admission was 75.80 years (SD: 12.16), with women (77.8) being six years older than men (71.7). A total of 45.4% had a femoral neck fracture (ICD-10 S72.0), 46.9% a trochanteric fracture (ICD-10 S72.1), and 6.1% a subtrochanteric fracture (ICD-10 S72.2). Five cases, initially diagnosed as usual hip fractures, eventually turned out to be periprosthetic hip fractures.

The vast majority was able to participate in the study by themselves, in 16.5% interviews had to be conducted with a proxy due to cognitive impairments. Approximately 13% had a migration background or migration experience. The proportion of participants with a higher education was larger in men than in women (30.1% vs 16.1%, p = 0.013). The majority lived independently in their own apartment or house, mostly alone, with more women living by themselves than men (51.4% vs 42.9%, p = 0.053). Participants had a high burden of comorbidities (49.7% with a CCI>=2) and limitations in ADL (36.2%). Women were more often dependent in their ADL than men (42.8% vs 26.0%). More than one in five showed signs of malnutrition (22.7%), and 14.2% showed at least moderate symptoms of depression and anxiety.

Pre-fracture hip functioning according to the OHS was good (Median: 45 of 48), however, this only applies to self-answering participants, as the OHS instrument is unsuitable for proxy interviews.

One in four participants had experienced another fall before suffering from the hip fracture, and 29.8% had fallen before in the last six months. In addition, almost one in four patients had visited an ED during the six months before the fracture.

The majority received internal fixation (62.6%). One in three patients had an ICU episode during their hospital stay, and 61.7% were transferred directly to a rehabilitation facility afterwards.

During the six months follow-up 40 (12.0%) patients died. There were no substantial differences between men and women regarding mortality.

Health-related quality of life

At baseline, the majority had no problems in the EQ-5D dimensions self-care (56.3%) and anxiety (55.8%). Patients had at least some problems with general pain (45.0%) and mobility (41.8%). Women had greater problems with pain (61.6% at least some problems vs 52.9%) and activities (55.6% at least some problems vs 37.9%). Overall, the mean EQ-5D index according to the German tariff was 0.70 (SD: 0.32). Women had a lower index value than men (0.66 vs 0.78). The mean self-rated health on the EQ-VAS was 69.9 (SD: 22.9), with no difference between women and men.

At six months, all EQ-5D health dimensions were lower than before the fracture. The largest losses were observed in activities (mean difference: −0.70), mobility (mean difference: −0.55), and self-care (mean difference: −0.42) (see Files 1–5). The mean difference regarding the EQ-5D index was −0.09 (SD: 0.31) points (mean EQ-5D index at six months: 0.63, SD: 0.32) (see Fig. 2). Both women and men rated their health 11 points (SD: 25.13) below their baseline level (mean EQ VAS at six months: 59.38, SD: 22.93) (see Fig. 3). Almost two in three patients reported a worse health state compared to their pre-fracture situation.

In the regression model for the six-month EQ-5D index, three factors appeared to be influential: depressive and anxiety symptoms (β =−0.19, p = 0.001) and a pre-fracture dependency in ADL (β =-0-17, p =0.011). Patients who were transferred to a rehabilitation facility had a better outcome (β =0.08, p =0.051) (see Table 2).

For the six-month EQ VAS model, age was removed as an independent variable. The linear regression model including age would have suggested an artificial association indicating higher EQ VAS values with increasing age (see Table 6). Graphical analysis of the association between age and six-month EQ VAS revealed that there was no linear or quadratic association between these two variables. Apart from the intercept, the estimates of the other independent variables did not change considerably with removing age from the model. In the final model, EQ VAS at six months appeared to be positively associated by better pre-fracture hip functioning (β =0.69, p =0.045), a direct transferal to a rehabilitation facility (β =8.05, p = 0.01), and treatment in a university hospital (parameter for general hospital: β =−6.3, p = 0.034). Negative impact was seen for migration background (β =−9.25, p = 0.014), polypharmacy (β =−6.48, p = 0.083), symptoms of depression and anxiety (β =−9.89, p =0.068), and living in an institution (−8.47, p = 0.104) (see Table 3). Our sensitivity analyses including only patients older than 64 years, and only patients with a complete follow-up interview showed no serious divergences in the model parameters (see Tables 7 & 8).

Strengths and limitations

To our knowledge, this is the first multicenter study on quality of life in hip fracture patients in Germany. One of its strengths lies in the study’s inclusiveness. As a patient-centered study in health services research, we sought to capture real-life data of as many affected patients as possible. Therefore, ability to consent and actively participate was not a prerequisite, and both community dwelling and nursing home residents were included. Further, we tried to include patients with migration background by providing questionnaires in the most prevalent immigrant languages in this area. However, the use of questionnaires in foreign languages remained limited. We did not set age limits to include both very old patients and younger adults who can be affected by osteoporotic fractures as well. Another strength is the use of the EQ-5D-5L version which holds better discriminatory properties than the widely used EQ-5D-3L version (Buchholz et al., 2018). We also included important dimensions such as migration background, malnutrition, or polypharmacy in our study that have hardly been investigated in hip fracture studies before. Nevertheless, the list of analyzed factors does not claim to be exhaustive.

The inclusive real-life approach, however, is accompanied by limitations: Our sample is comparatively heterogeneous which can be criticized. Another limitation lies in the fact that a considerable number of participants was lost to follow-up. Due to data protection rules, participants who had moved during the six months after enrollment without letting us know were unable to reach. Among those, there might be relevant number of institutionalized people. Previous studies have reported institutionalization rates of 10–15% within six months after hip fracture (Dyer et al., 2016; Rapp et al., 2015). Clinical data were only available for the initial hospital stay. Therefore, information on the follow-up health care remains limited. This includes the exact volume and type of rehabilitation care. Lastly, our results cannot be easily transferred to other populations. They originate from an inner-city district of a metropolitan city in Germany with specific living and health care conditions.