BACKGROUND

A fall is defined as ‘an unexpected event in which the person comes to rest on the ground, floor or a lower level’, irrespective of whether an injury occurs (Kellogg International Work Group, 1987). Falls are highly prevalent in health and social care settings, especially in older people (Sharif et al., 2018). More than 50% of people older than 80 years fall at least once a year (Bor et al., 2017).

Falls can lead to serious consequences, such as injuries, increased care dependency, or even death (Alamgir et al., 2012; Evans et al., 2015). More than 20% of falls in older adults require a hospital visit (Alamgir et al., 2012; Milat et al., 2011). In addition, the increased utilization of health care associated with falls, such as surgery or hospitalization, increases costs associated with falls (Alekna et al., 2015).

In the international literature, many studies have been conducted to examine the prevalence of institution-acquired falls (IAFs) and risk factors in different countries. In hospital settings, an IAF prevalence of 1.9%–17.1% was reported (Rao et al., 2018). In nursing home settings, IAF prevalence ranged from 13.0% to 92.5% (Bor et al., 2017; Zhang et al., 2018). These large ranges can be explained using different definitions, instruments, and measurement methods to assess IAFs. Most studies have been conducted with specific groups of patients and residents (e.g. stroke patients [Cox et al., 2017] or residents with dementia [Kosse et al., 2015]), whereas some studies have been carried out on older people. The most frequently investigated IAF risk factors were the use of specific medications (Pelaez et al., 2015; Westerlind et al., 2019) (e.g. benzodiazepines, anticonvulsants), followed by age (Bor et al., 2017; Mazur et al., 2016). Having a fall history (Mazur et al., 2016), impairments in mobility (Mazur et al., 2016; Zhang et al., 2018) as well as high care dependency and functional limitations (O'Neil et al., 2018) are also frequently mentioned IAF risk factors.

The literature provides a wealth of knowledge on both IAF prevalence and risk factors in acute care and long-term care settings, but certain research gaps still exist. First, no recent studies have analyzed international data on the IAF prevalence and risk factors using the same standardized method. Second, no studies provide an insight into IAF prevalence in different healthcare settings (i.e., both hospitals and nursing homes). Third, most studies have been carried out on specific patient or resident groups. A comprehensive, general overview of IAF prevalence and risk factors for hospital patients and nursing home residents is still lacking.

The current study was conducted to provide an overview of IAF prevalence and IAF risk factors in hospital patients and nursing home residents in Austria (AT), Switzerland (CH), the Netherlands (NL), Turkey (TR), and the United Kingdom (UK).

METHODS Design

This study involved a secondary data analysis of data collected in the International Prevalence Measurement of Care Quality (Landelijke Prevalentiemeting Zorgkwaliteit, LPZ study) (van Nie-Visser et al., 2013), which is a cross-sectional study carried out to measure care quality in AT, NL, CH, TR, and the UK. Data were collected each year on one specific day (14 November 2017 and 13 November 2018) in all countries.

In the LPZ study, the prevalence, preventive measures, and treatment options associated with nursing care problems, such as pressure injuries, continence, malnutrition, falls, restraints, and pain are measured. Participating hospitals and nursing homes in AT, TR, and the UK as well as hospitals in CH are required to measure falls, while hospitals and nursing homes in NL and nursing homes in CH can choose which nursing care problems they want to measure.

Each country has a national coordinator who is responsible for the study design, training, and adaptation of the questionnaires, as part of the International LPZ Research Group.

Setting and sample

National coordinators annually invite health and social care institutions to participate in the study. In CH, all acute-care hospitals are required to participate in the measurement. In all other countries, participation is voluntary. Long-term care settings are called either nursing homes or care homes in these countries, but all meet the internationally accepted definition of the nursing home (Sanford et al., 2015). Data gathered from all types of hospitals (e.g., general or academic) (AT, CH, NL, TR) and nursing homes (AT, CH, NL, TR, UK) in 2017 and 2018 were included in the study.

Data collection

Participating hospitals and nursing homes collected data for this study. These were invited to training sessions organized by the national coordinator, at which they were provided with comprehensive training materials, standardized questionnaires, and informed consent forms.

Patients and residents were considered eligible for study inclusion if they were present in the respective ward on the morning of the measurement and their oral informed consent (CH, UK, NL) or written informed consent (AT, TR) could be obtained. In the UK, the LPZ study is conducted as part of service improvement and development work and, therefore, is an exempt from formal consent procedures. In CH, the Canton of Bern ethics committee declared that the LPZ measurement was not subject to the Swiss Human Research Act, and approval was deemed unnecessary. In the remaining jurisdictions (AT, NL, and TR) the ethical committees gave their approval for the study protocol annually.

Data collection was performed by pairs of trained nurses working in the participating organizations. One nurse worked on the ward of the patient/resident, and one nurse worked on another ward. If the nurses disagreed about an answer on the questionnaire, the nurse from the other ward made the final decision. To ensure comparable data collection procedures in all participating hospitals and nursing homes in all countries, the data collection procedure was standardized (e.g., the same manual with explanations for the questionnaire was used in each country).

Variables in the data collection process

The original Dutch LPZ questionnaire was developed based on the literature, guidelines, and expert opinions. It was forward and back-translated into the native languages of the participating countries (English, German, Italian, French, Turkish), and feasibility tests were performed. The questionnaire includes recent, evidence-based guidelines and is regularly updated by the International LPZ Research Group to foster face validity (van Nie-Visser et al., 2013).

Several studies on the reliability and validity of the included scales as well as on specific questions, such as those on pressure ulcers and falls (Bours et al., 1999; Halfens et al., 2011; van Nie-Visser et al., 2013) have been performed. For example, the interrater reliability in terms of Cohen's Kappa of the pressure ulcer grading system was reported to vary between different health care institutions from 0.49 to 0.97 (Bours et al., 1999). Overall, the inter-rater reliability of the questionnaire was 0.87 and found to be good (Meijers et al., 2009), which is a good support for adequate psychometric properties.

The questionnaire used in the current study includes general questions on age, sex, surgery within the last two weeks (yes/no), medical diagnosis according to ICD-10, as well as additional diagnoses such as dementia (yes/no) and care dependency. Care dependency is described using the Care Dependency Scale (CDS) (Dijkstra et al., 1996). The CDS comprises 15 items covering basic Activities of Daily Living (ADL), which are scored on a five-point Likert scale (1 = completely dependent to 5 = completely independent). Lower scores indicate higher care dependency levels. Studies demonstrated good levels of reliability for the CDS, with Cronbach's alpha of 0.98 for geriatric clinics, 0.94 for home care and 0.96 for the nursing home setting (Dijkstra et al., 2012). The German version of the CDS in the hospital setting showed a Cronbach's alpha coefficient of 0.98, constituting a high degree of internal consistency (Lohrmann et al., 2003a,2003b). Moreover, the scores of the CDS correlates with the scores of two other instruments, indicating criterion-related validity (Lohrmann et al., 2003a,2003b). In the present study, the Cronbach's alpha for the CDS items range between 0.96 in CH, and 0.98 in AT hospitals, with an overall score of 0.97. Whereas the overall Cronbach's alpha score for the CDS in the nursing home setting was 0.97 ranging from 0.96 in UK up to 0.99 in TR.

The LPZ falls module includes a question about falls that have occurred in the host institution in the last 30 days (IAF) (yes/no/unknown); these answers were examined to determine the primary outcome of this data analysis. An additional question asks whether the person has experienced a fall in the last 12 months (falls history), which was treated as an independent variable in our analysis (yes/no). More specifically, a fall is defined as ‘an unexpected event in which the participant comes to rest on the ground, floor or a lower level’, irrespective of whether an injury occurs (Kellogg International Work Group, 1987).

Data analysis

Questionnaire responses collected from the five countries in 2017 and 2018 were merged into one data file. Data analyses were performed using SPSS version 26.0. Since the main outcome had three answer possibilities (yes/no/unknown), the answer option ‘unknown’ was treated as missing data. The level of significance was set at p < 0.05. For categorical variables, percentages and chi-square tests were used. For non-parametric continuous variables, medians and interquartile range (IQR) and Kruskal-Wallis tests were performed. Data were analyzed to explore relationships between baseline exposures and IAF, using a three-step, separate approach for hospitals and nursing homes. First, a bivariate analysis using chi-square tests and Mann-Whitney U tests was conducted to identify significant IAF risk factors. Second, IAF-associated variables in the bivariate analysis which did not show collinearity based on a variance inflating factor (VIF) less than four were tested against an outcome of falls using a univariable logistic regression. Third, significant variables in univariable logistic regression were included in the generalized estimating equation (GEE) models. The GEE model is useful to predict binary outcomes (IAF yes/no) when clustered data are used (IBM Corp. Released, 2019). We assumed that patients and residents within a country were more similar to each other than patients and residents in different countries, due to e.g. different health care systems, so the country was used as a cluster (inner-subjects variable). The GEE model used a binary logistic model due to the binary outcome and exchangeable order. An exchangeable order for the working correlation matrix should be used to cluster data with no chronological order (Wakefield, 2009).

The following criteria were used to choose the model: (1) it had to converge to be reliable; (2) there had to be a low quasi-likelihood under the independence model criterion (QIC) to select the adequate mean model; and (3) non-significant variables in the GEE model were excluded.

RESULTS

Of 70,940 patients and residents who were admitted at the time of the measurement, 58,319 participated in the study (response rate: 82.2%). Of these, 34,361 (58.9%) were admitted to hospitals and 23,958 (41.1%) resided in nursing homes (Table 1).

TABLE 1. Sample characteristics (overall and by country) from hospitals and nursing homes Hospitals Austria (n = 6337) Switzerland (n = 26,603) The Netherlands (n = 251) Turkey (n = 1170) United Kingdom (n = 0) Total sample (N = 34,361) Sex % (n)* Male 48.3 (3063) 50.7 (13,478) 47.8 (120) 50.3 (589) – 50.2 (17,250) Female 51.7 (3274) 49.3 (13,125) 52.2 (131) 49.7 (581) – 49.8 (17,111) Median age in years (IQR)* 69 (23) 70 (23) 73 (12) 65 (23) – 70 (24) Median number of medical diagnosis (IQR)* 2 (3) 3 (2) 3 (2) 2 (2) – 3 (3) Median care dependency (IQR)* 74 (11) 70 (14) 72 (11) 72 (25) – 71 (14) Nursing homes Austria (n = 666) Switzerland (n = 221) Netherlands (n = 19,563) Turkey (n = 1830) United Kingdom (n = 1678) Total sample (N = 23,958) Sex % (n)* Male 26.6 (177) 40.7 (90) 30.5 (5975) 55.4 (1014) 30.7 (515) 32.4 (7771) Female 73.4 (489) 59.3 (131) 69.5 (13,588) 44.6 (816) 69.3 (1163) 67.6 (16,187) Median age in years (IQR)* 86 (11) 83 (14) 85 (11) 74 (17) 86 (11) 85 (12) Median number of medical diagnosis (IQR)* 5 (3) 4 (3) 2 (3) 3 (2) 2 (2) 2 (3) Median care dependency (IQR)* 44 (28) 51 (34.5) 46 (28) 56 (46) 36 (29) 45 (30) Abbreviation: IQR, interquartile range.

The median ages of hospital patients and nursing home residents, respectively, were 70 and 85 years. Hospital patients had a median care dependency of 71, indicating that they were completely care independent, while nursing home residents had a median care dependency of 45, indicating that they were partially care dependent.

Hospital patients’ ages differed significantly among the countries, with patients in NL being the oldest. In addition, we identified significant differences in care dependency levels among the hospital patients in most countries (the difference between TR and NL was insignificant). Hospital patients in CH had the highest care dependency levels.

Nursing homes residents also differed significantly with regard to their ages and care dependency levels in all countries. Residents in TR were the youngest (i.e., 74 years on average), and residents in the UK were the most care dependent.

IAF prevalence

IAF prevalence was 3.9% in hospitals and 11.9% in nursing homes (Table 2). IAF prevalence ranged from 2.8% to 7.7% in hospitals and between 6.0% and 15.8% in nursing homes with significant differences identified. The highest IAF prevalence rate for hospitals was found in TR (7.7%) followed by AT (4.0%). The highest IAF prevalence in nursing homes was found in CH a (15.8%) followed by AT (15.5%).

TABLE 2. Institution-acquired fall (IAF) prevalence for each country and the total sample separated into hospital patients and nursing home residents Hospitals % (n) Austria (n = 6326) Switzerland (n = 26,484) The Netherlands (n = 251) Turkey (n = 1163) United Kingdom (n = 0) Total sample (N = 34,224)a IAF % (n)* 4.0 (250) 3.7 (974) 2.8 (7) 7.7 (90) – 3.9 (1321) Nursing homes % (n) Austria (n = 665) Switzerland (n = 221) The Netherlands (n = 8483) Turkey (n = 1828) United Kingdom (n = 1608) Total sample (N = 12,805)a IAF % (n)* 15.5 (103) 15.8 (35) 12.3 (1046) 6.0 (110) 14.2 (228) 11.9 (1522) Risk factors for IAF

The bivariate analysis of hospital patient data showed that sex and suffering from a motor disorder were not significantly associated with IAF; therefore, these factors were excluded from further testing (Table 3). Eleven variables for hospitals and ten variables for nursing homes were identified as statistically significant; therefore, these were included in the GEE models (Table 4).

TABLE 3. Results of bivariate analysis, multicollinearity test, and univariable logistic regression of IAF risk factors in the hospitals and nursing homes Hospitals Nursing homes Bivariate analysis VIF Univariable logistic regression Bivariate analysis VIF Univariable logistic regression p-value n p-value p-value n p-value Country 0.00 34,224 <4 0.00 0.00 12,805 <4 0.00 Age 0.00 <4 0.00 0.00 <4 0.00 Care dependency sum score 0.00 <4 0.00 0.00 <4 0.00 No. of medical diagnoses 0.00 <4 0.00 0.0 <4 0.00 Sex 0.16 – – 0.53 – – Surgery 0.00 <4 0.00 0.01 <4 0.02 Cancer 0.00 <4 0.00 0.08 – – Cardiovascular diseases 0.00 <4 0.00 0.09 – – Dementia 0.00 <4 0.00 0.00 <4 0.00 Motor disorders 0.35 – – 0.01 <4 0.01 Diseases of the nervous system 0.00 <4 0.00 0.00 <4 0.00 Stroke 0.00 <4 0.00 0.03 <4 0.03 Fall historya 0.00 32,761 <4 0.00 0.00 11,868 <4 0.00 Abbreviation: IAF, institution-acquired fall; VIF, variance inflating factor. TABLE 4. GEE models on IAF for hospitals and nursing homes Hospitals (N = 32,761) Nursing homes (N = 11,868) Odds ratio 95% CI Odds ratio 95% CI Country Austria 0.526 0.401–0.688 0.658 0.488–0.886 Switzerland 0.402 0.312–0.517 0.743 0.451–1.225 The Netherlands 0.432 0.195–0.960 0.895 0.749–1.069 Turkey (reference hospital) 1 – 0.768 0.568–1.038 United Kingdom (reference nursing homes)a – – 1 – Age 1.010 1.006–1.015 1.005 0.998–1.012 Care dependency sum score 0.976 0.972–0.979 0.992 0.989–0.996 No. of medical diagnoses 1.107 1.077–1.138 0.993 0.954–1.034 Surgery Yes 0.705 0.612–0.813 1.646 0.973–2.785 No (reference) 1 – 1 – Diseases of the nervous system Yes – – 1.309 1.103–1.553 No (reference) 1 – 1 – Fall history Yes 2.814 2.472–3.202 26.026 21.681–31.243 No (reference) 1 – 1 – Abbreviation: IAF, institution-acquired fall.

Different risk factors for hospitals and nursing homes contributed to a higher risk of an IAF. Outcomes of the GEE models showed that hospital patients in AT, CH, and NL had a lower IAF risk as compared to patients in TR. In addition, having a fall history increased the risk two-fold (OR 2.814) of experiencing an IAF in this setting. The analysis of nursing home data indicated that a higher care dependency score (i.e., a lower care dependency) was associated with an IAF risk (OR 0.992). Furthermore, having a fall history increased the risk of an IAF in the nursing homes by 26 times (OR 26.026).

DISCUSSION

This study was carried out to explore prevalence and risk factors associated with IAFs in hospital and nursing home settings in several European countries. We found that hospitals in TR had the highest prevalence of IAFs (7.7%), whereas hospitals in NL had the lowest prevalence (2.8%). The highest prevalence of IAFs in nursing homes was seen in CH (15.8%) and the lowest, in TR (6.0%). The application of our model revealed several significant IAF risk factors, whereby the history of a fall in the last year was identified as the most important risk factor in both settings.

The literature includes a range of IAF prevalence in hospitals (1.9%–17.1%) and nursing homes (13.0%–92.5%) (Carryer et al., 2017; Pelaez et al., 2015). Our study results for hospitals agree with those presented in the literature. Even though we found statistical significance between the countries, our samples are comparable in terms of age and care dependency. More specifically as an example, TR had median CDS score of 72, which lies in the range of the other countries (70–74).

However, our results from nursing homes (6.0% and 15.8%) were lower. This difference might be due to the fact that TR residents who took part in the study were relatively young and had low care dependency levels or that different definitions and measurements methods were used in our study as opposed to previous studies.

These study results revealed a significant difference in IAF prevalence between the countries, but these differences cannot exclusively be explained by country-specific differences in demographic variables. The GEE models used treats countries as clusters. Because differences in IAF prevalence between the countries were not affected by country-specific adjustments of data or the inclusion of other influencing factors such as age or care dependency, they must be explained by other factors.

One possible explanation for these differences may be that CH hospitals only measure pressure ulcer and falls modules, and these two problems have been the focus of attention for years. This may have led to an increased awareness and to more interventions to prevent falls. In contrast, AT hospitals and nursing homes measure all six nursing care problems and can independently choose areas in which they want to initiate quality improvement projects.

Another explanation could be that the countries differ with regard to their hospital structure, design, equipment, and furnishings (e.g., light intensity). For example, a study showed a significantly increased fall incidence in a hospital with single rooms as compared to a facility with multi-bedded rooms (Singh et al., 2015). Another study showed that the nursing skill mix and registered nurse staffing ratio play important roles (Lucero et al., 2019). However, these aspects were not included in the LPZ measurement.

The GEE model results revealed a fall within the last year to be the most important IAF risk factor, with 2.824 individuals in hospitals and 26.026 in nursing homes identified as being at increased risk of falling. Various other studies have also showed an association between a positive fall history and the IAF prevalence but to lesser extent. As in our study, another study determined that hospital patients with a fall history had an approximately 2.5-higher chance of falling (Mazur et al., 2016). These results suggest that a simple question about fall occurrence in the last 12 months should be included in every comprehensive nursing/geriatric assessment, either when a patient is admitted to hospital or regularly in a nursing home. Our results also support those of a systematic review, which showed that risk assessment tools may be helpful but no more helpful than an experienced nurse's clinical view (Haines et al., 2007). Evidence-based guidelines also do not recommend the general use of a fall risk assessment tool (National Institute for Health and Care Excellence [NICE], 2013). Our results strongly support the addition of at least one question about the patient's or resident's existing fall history, as this can help nurses identify risk groups more quickly and subsequently prevent IAFs.

This study is one of the first to describe IAF risk factors in different countries based on a large sample. One main strength of this study is that the data collection procedure was standardized; i.e. data were comparable between countries. This