Risk factors for fall occurrence in hospitalized adult patients: a case-control study

ABSTRACT Objective: to identify risk factors for falls in hospitalized adult patients. Methods: a matched case-control study (one control for each case). A quantitative study conducted in clinical and surgical units of a teaching hospital in Southern Brazil. The sample comprised 358 patients. Data were collected over 18 months between 2013-2014. Data analysis was performed with descriptive statistics and conditional logistic regression using Microsoft Excel and SPSS version 18.0. Results: risk factors identified were: disorientation/confusion [OR 4.25 (1.99 to 9.08), p<0.001]; frequent urination [OR 4.50 (1.86 to 10.87), p=0.001]; walking limitation [OR 4.34 (2.05 to 9.14), p<0.001]; absence of caregiver [OR 0.37 (0.22 to 0.63), p<0.001]; postoperative period [OR 0.50 (0.26 to 0.94), p=0.03]; and number of medications administered within 72 hours prior the fall [OR 1.20 (1.04 to 1.39) p=0.01]. Conclusion: risk for falls is multifactorial. However, understanding these factors provides support to clinical decision-making and positively influences patient safety.


Introduction
According to the World Health Organization (WHO), fall is defined as "inadvertently coming to rest on the ground, floor or other lower level, excluding intentional change in position to rest in furniture, wall or other objects" (1) .
In hospitalized patients, incidence rates of falls are responsible for two in five adverse events, and their frequency varies from 1.3 to 13.0 per 1,000 patients per day (1)(2) .
A recent study showed that in the United States of America (USA) the prevalence of falls increased from 28.2% to 36.3% in 2010 (3) . In England and  (4) . In Austria, of the 3,648 patients investigated in hospitals, 38.5% suffered injuries due to falls.
Similar results were found in Switzerland, where of the 10,098 patients, the prevalence of falls reached 34.7% (5) . This can be a result, possibly, of an increase Observational studies investigating these risk factors in hospitalized patients presented some possible bias, such as sample constituted exclusively of patients aged 65 or more (7)(8)(9)(10) ; investigation of events just within the first week of admission (8) ; definition as exclusion criteria: patients with dementia, delirium or memory change (11) ; and absence of data collection on Sundays and Holidays (7) .
In this context, the hypothesis of this study was that the identification of risk factors for falls in hospitalized adult patients facilitates a more accurate measurement of the risk for falls, and has a positive impact on patient safety. Therefore, the objective of this study was to identify predictors for falls in hospitalized adult patients.

Method
This is an observational case-control study (one control for each case) with matching. Patients were matched regarding sex, unit and date of admission.
The outcome was the incidence or not of fall(s). First, patients who have suffered falls were selected (cases).
Next, subjects who have not suffered falls were selected (controls).
The study setting comprised 12 clinical and surgical units of an 843-beds hospital, connected academically to an university in Southern Brazil, which was recently accredited by the Joint Commission International (JCI) (12) . In this institution, nurses report hospital inpatient falls in the electronic health record. This notification creates an e-mail that is sent to the multi-professional team responsible for risk management and patient safety. During the study data collection, the investigators received this same e-mail and conducted an active search in the units during all weekdays, covering all shifts, in order to identify the incidence of falls.
The sample consisted of 358 clinical and surgical patients. Patients included were 18 years old or older, both sexes, controls with the same admission date as the cases, or subsequent dates. Exclusion criteria were: patients without clinical conditions (torpor or coma) to participate in the study, those who did not have caregiver in the time of data collection, patients under palliative care, those whose falls occurred outside the units of study and those whose falls occurred for the second time (or more).
The study protocol specified no more than 72 hours after the fall for including patients in the study.
Data were collected over 18 months, between 2013-2014, by the researcher, four registered nurses and one Nursing student, and they received specific training before data collection. The training comprised theoretical (threehour-long meetings) and theoretical-practical classes (daily supervision by the principal investigator on the logistic of the research assistants, and in the field, between April and July 2013). The evaluation of the event, data collection technique and documentation were carried out in conjunction. After these three months, the research assistants were considered capable of collecting the data individually.
Data were collected directly from patients, from the electronic health record, from the fall risk scale adopted in the hospital (Morse Fall Scale) (13)(14) , and from the institutional instrument for describing falls. This instrument is composed of the factors that trigger the fall and patient clinical conditions before the event.
The variables (risk factors) of the study were selected from a previous study (15) and included in a data collection manual. Conceptual and operational definitions were constructed for the included variables ( Figure 1 Present in the patient chart next to the notes. Previous fall (last three months) If the patient has been hospitalized or has a recent history (up to three months) of falls due to physiological causes (13)(14) .
Information asked the patient or his relative. It can also be verified in the patient chart.

Drowsiness
Patient is drowsy and/or experiencing difficulty in waking up when asked.
Reported by the patient and/or observed in the patient chart.
Walking limitation Walking ability that may be limited. Evaluated by the Time up and go test (16) . Patients who perform the test in more than 10 seconds, who are unable to lift from the chair or bed, who use a walking stick, walker and/or wheelchair and who cannot perform these activities due to physiological reasons, such as lower limb injuries/fractures and altered visual acuity (e.g., diplopia and amaurosis).
Disorientation and/or confusion If one or more of the operational questions is not answered correctly (17) , in this study, the patient will be considered disoriented and/or confused.
Asked the patient and their answers verified: What is your name? Where are you (in which city or hospital name)? What year are we in? (17) It can also be verified in the nursing notes.
Frequent urination and urinary/ intestinal urgency Urination more frequent than usual, urgency to urinate and/or presence of diarrhea (18) .
The presence of these changes were asked the patient/ family in the current or previous shift. It can also be verified in the nursing notes.

Agitation
Excessive motor activity associated with a subjective tension experience (19) .
Present in the patient chart along with the notes.
Absence of caregiver Absence of the patient's caregiver at the time of the event.
Asked the unit nurse or the patient.
Absence of the nursing diagnosis Risk for Falls (20) Patient without the nursing diagnosis Risk for falls listed in the electronic health record.
Verified if the nursing diagnosis Risk for falls was open in the patient chart.

Name and number of medications
The administration of medications such as benzodiazepines, opioids, barbiturates, antipsychotics, antidepressants, antihypertensives, laxatives, diuretics, antihistamines, sedatives and anticonvulsants may increase the possibility of falls (15) .
The last administered dose (up to three days) was counted and considered as a risk period based on the elimination half-life of the prescribed medications. The sample calculation was performed according to Chang and et al (9) , from the therapy of narcotics, with odds ratio (OR)=2.13 and a prevalence of falls of 13.9%. It was considered a statistical power of 80% and a significance level of 0.05, with 20% of possible losses that could occur during the study.
The Research Ethics Committee of the hospital approved this study (protocol #130012).

Results
The sample consisted of 54% (n=204) of male

Discussion
This study presented the largest casuistry with for controls. On the other hand, in this study, age greater than 60 years did not appear as a statistically significant variable.
The incidence of the event was higher in male patients (57%). However, there is no consensus in the literature on the association between sex and an increased risk for falls (22)(23)(24) . For this reason, this variable was one of the criteria adopted for the matching in this study.
The descriptive data presented in Table 1 showed that the length of hospital stay was one of the variables significantly related to the event. The median of the length of hospital stay was similar in both cases and controls, as this variable was investigated until the incidence of the fall. As it was a case-control study, no follow-up of these patients was performed after the event.
This has a direct relationship with a better clinical profile of these patients, what corroborates a lower number of patients in postoperative period in the sample (cases=22.9% and controls=32.4%).
In general, in hospital clinical practice, non-surgical patients had a higher rate of falls when compared to surgical patients (considered in this study with a history of surgery(s) in the current hospitalization), as the former had a longer length of hospital stay, a higher incidence of comorbidities and a greater demand for health care (9,12,(21)(22) .
In this study, the postoperative period served as   that different studies confirmed the postoperative period as an important risk factor for falls (16,18,27) . However, the complexity of non-surgical patients could have The use of anticonvulsants medications and benzodiazepines was also investigated using the Hendrich II Fall Risk Model (29) . The administration of medications of different classes (tranquilizers/sedatives, diuretics, antihypertensives, antiparkinsonians, antidepressants and others) is also part of the Downton Fall Risk Index, which has been not fully tested and disseminated across studies (15,30) .
In an integrative review that aimed to find risk factors for falls, as this study, antidiabetic agents were found in only two observational studies (15) . Therefore, from the researchers' point of view, there was not enough evidence to associate them with the outcome and they were not included in this study.
In addition, when considering medications as predictors, the researchers point out that the association between different medications of the same class or the combination of different classes may produce or potentiate clinical conditions of hypotension, confusion, dizziness, attention deficit, drowsiness, and other.
Furthermore, the researchers report that polypharmacy use should be supervised by health professionals, in order to identify factors that may contribute to the incidence of falls (31)(32) .  (33) . It is known that among the factors that independently correlate with an increased risk of falls are walking limitation, frequent urination and change in mental status (e.g. disorientation/confusion and drowsiness) (15,(33)(34) . These items are evaluated by the most relevant predictive models (29)(30) .
Regarding the alteration in urinary and/or intestinal eliminations, the variable frequent urination was found as a predictive factor. This variable is so important that the Hendrich II Fall Risk Model (29) includes the presence of urinary and/or intestinal alteration between its evaluated items. The same is true of the Risk Assessment Tool in Falling Elderly Inpatients (STRATIFY) (35) , which evaluates the frequency that patients go to the toilet. One explanation is that a more frequent need of to urinate is related to a greater need to go to the toilette, which exposes the patients to a greater risk of falling (24,33,36) .
Environmental risk factors were not evaluated in the controls due to the limitations inherent to the casecontrol studies, in which patients are evaluated before the incidence of the event.
The results in Table 2 showed that 35.2% (n=63) of the cases were followed by a caregiver at the time of the fall and 59.3% (n=106) of the controls. In some situations, the family members were present, but were not able to interfere in the event, for example, when they were sleeping or when they were walking at the side of the patient, but were unable to hold them. Perhaps, in this study, if the incidence of the event was also considered as the absence of caregiver at the time of the fall, the behavior of this variable could be different. asking the assistance of their caregiver because they are sleeping (23,36) .
Nonetheless, these situations reflect the reality of many health institutions, where there is a push in stimulating the participation of family members in patient care (36)(37) , in addition to an increased number of patients per nurse and an increased demand at work (7,11) . which is not a reflection of the clinical practice reality (25) .
In another study, using clinical practice and carried out in the same institution of this study, a prevalence of It is highlighted as limitations of this study that it was carried out in a single center, with secondary use of the data from the electronic chart and from the instrument of notification of falls of the institution. In addition to these, there is a risk of bias inherent of retrospective studies, for example, when patients were asked to remember the information prior to the event, which means that the evaluation was biased by the memory of patients.

Conclusions
The In research, the support of a statistical and epidemiological reference can stimulate the development of future research and the establishment of new hypothesis, whose main outcome is patient safety.