Findings from three methods to identify falls in hospitals: Results from the Ambient Intelligent Geriatric Management system fall prevention trial

To (a) compare characteristics of patients who fall with those of patients who did not fall; and (b) characterise falls (time, injury severity and location) through three fall reporting methods (incident system reports, medical notes and clinician reports).


| INTRODUCTION
Falls are the second leading cause of unintentional death as a result of injury, with rates highest amongst those older than 60 years. 1 In Australian public hospitals, it has been reported that during 2015-2016, four inpatient falls cause harm per 10,000 hospitalisations, with patients who had fallen staying 18.8 days longer than those who had not fallen, potentially resulting in $38,991 in extra costs. 2 Clinicians are required to report all falls to incident management systems. 3This is to encourage learning from past incidents, support the resolution of incidents and identify preventative strategies. 3Falls are hospitalacquired complications, and fall prevention is a national safety and quality imperative. 46][7] An Australian fall prevention study in 2010 cautioned that the hospital incident management systems captured only three-quarters of falls (or 78% of injurious falls). 5A key recommendation arising from that study was that in-hospital falls research should not depend purely on incident reports.An older Japanese study in 2015 similarly reported that 25% of falls recorded in the medical notes were not captured by incident reports. 8It is unknown whether the degree of underreporting remains as large.
Leveraging data from a wearable sensor-based fall prevention trial, 9 the aim of this brief report was to (a) compare the characteristics of patients who had fallen with those of patients who had not fallen and (b) characterise falls (time, injury severity and location) through three different methods of falls reporting (incident system reports, medical notes and clinician staff reports).

| Setting and intervention
The study was conducted in three wards in two hospitals in two states in Australia.The South Australian (SA) ward was a Geriatric Evaluation and Management Unit (GEMU).In Western Australia (WA), a GEMU and general medicine ward were involved.All wards spent 25 weeks in control (first wedge), and then after a 3-week period, successive wards moved into a 25-week intervention wedge over the next 75 weeks in the following order: (1) SA GEMU; (2) WA GEMU; and (3) WA General Medicine.

| Participants
All patients aged 65 years or older were eligible for the primary trial.Those requiring palliative care and those previously enrolled in this study within the same admission were excluded.Data from 3239 trial participants, including 273 patients who had fallen, were investigated as part of this analysis. 9

| Measurements
Patient demographic details were recorded.In the trial, fall was an 'event which results in a person coming to rest inadvertently on the ground or floor or other lower Conclusions: This study reaffirms the need to improve reporting falls in incident systems and at clinical handover to the team leader.Research should continue to use more than one method of identifying falls, but include data from medical records.Many falls cause injury, resulting in poor health outcomes.

K E Y W O R D S
accidental falls, aged, aged 80 and over, inpatients, patient safety, risk management

Practice Impact
As inpatient falls were under-reported in incident management systems, with a higher proportion of injurious falls reported, clinicians should be encouraged to report all falls, including noninjurious falls in incident reporting systems, to improve reliable data for system learning and research.Future studies should not rely solely on falls data sourced from incident reports.level'. 12Falls data were from three different sources: (1) electronic incident management systems, (2) case notes' review at a minimum of twice-weekly and (3) daily inquiry of the team leader and other clinicians by research staff.Injury in this study was classified as bruising, laceration, fracture, loss of consciousness or if the patient reported persistent pain. 13,14Where fractures were recorded, radiological confirmation was required. 15In terms of the incident reporting system, responsible officers at the hospitals produced reports for the research team.

| Program fidelity
Two weeks before the trial commenced, in-service programs occurred with ward staff to ensure that staff were familiar with the definition of a fall and with best practices.Staff were reminded on how falls should be recorded in the hospital incident reporting system as well as patient medical records.

| Statistical analysis
In this study, 'day' was defined as 0700-1959 h (13 h) whilst 'night' was 2000-0659 h (11 h).Descriptive statistics (numbers and percentages, or median and interquartile range [IQR]) described the characteristics of patients who had fallen and their falls.Patient characteristics of those who had not fallen and those who had fallen were compared using chi-squared tests, two-sample t-tests and median tests.Associations between patient/fall characteristics and whether a fall was recorded in an incident report were analysed using bivariate logistic regressions using generalised estimating equations (GEE) with the use of incident reports as the dependent variable.Analyses of fall characteristics included all falls, accounted for clustering of multiple falls per patient and used robust standard errors.Categorical patient/fall characteristics were entered as factors with one level specified as the reference category.
The Venn diagram (Figure 1) demonstrates that there was a total of 371 falls, with 346 (93%) captured through F I G U R E 1 Falls events captured through medical notes, incident reports and clinicians.medical notes, 254 (68%) incident reports and 265 via clinician reports (71%).Of the 25 falls not documented in the medical notes, 20 had been reported to the incident reporting system, but five others were known to clinicians and were not documented or reported.
Based on medical notes (Table 1), 34% of falls were injurious, with the majority (84%) of these classified as moderate to severe.The proportion of injurious falls was 34%-40% across reporting methods, with 84%-89% of these classified as moderate to severe.
Where the time of fall was recorded (n = 301), almost half (n = 128; 42%) of falls occurred at night.Almost a quarter of falls occurred in the bathroom (13%) or toilet (10%), with the majority of falls (70%) in the patient room.Similar patterns were noted no matter which method was used to record the falls.
Falls (Table S2) were less likely to be incident reported in females (odds ratio [OR] 0.56; 95% confidence interval [CI] 0.36-0.90;p = 0.014) compared to men.Falls occurring in patients where the primary reason recorded for their admission was falls without fracture were less likely to be reported in the incident report (OR 0.42; 95% CI 0.24-0.73;p = 0.002) when compared to patients admitted for other reasons.Injurious falls were more likely to be incident reported (OR 3.02; 95% CI 1.79-5.10;p < 0.001) than non-injurious falls, and falls of a moderate-to-severe nature were more likely to be incident reported (OR 4.34; 95% CI 2.31-8.18;p < 0.001) compared to falls without injury.There were no other significant associations noted in relation to patient characteristics (age, co-morbidities of delirium or dementia, falls or fracture and polypharmacy) or time and location of falls.

| DISCUSSION
This study confirmed that falls continue to be underreported to hospital incident reporting systems.Compared to previous studies where one quarter of falls were not recorded in the incident reports, 5 almost one-third of falls were missed when relying on incident reports only in this study.Of concern also was that 25 falls were not documented in the medical notes, which was similar (7% vs. 8%) to that previously reported. 5Additionally, this research found that falls may be under-reported during clinical handovers.These were results obtained during a clinical trial where the Hawthorne effect, which is the modification of behaviour when being watched, was likely at play. 17These issues of under-reporting and failing to document are likely higher in real-world settings.In study of older inpatients, patients who had fallen were more likely to have been admitted following a fall (with or without fracture) and to have cognitive impairment (dementia or delirium) as a co-morbidity.A recent study identified that factors such as age, history of falls and mental status in addition to Morse Fall Scale score, quality of gait, hospital unit type and the number of high fall risk-increasing drugs were important factors when it came to predicting inpatient falls risk. 18In keeping with findings from other studies, 2 patients who had fallen experienced poor health outcomes when compared to patients who had not fallen, including a lower proportion discharged directly home, higher mortality and increased length of stay.
The same patterns are noted no matter which method is used to document falls.One-third of falls resulted in injuries.Similar to findings from Hitcho et al., a quarter of falls occurred in wet areas (in the toilet or shower), with the majority in the patient's room.Falls without injuries are less likely to be reported in incident reports, a finding similar to that seen in the study by Hill and colleagues. 5 limitation of this study was that the reference method relied on was falls recorded in the medical records.In this study, at least 26 falls were not recorded in this way (Figure 1).Whilst outside the scope of this study, it may be possible to investigate the number of falls not identified by any of the methods through analysis of the sensor data in the AmbIGeM study, an area for further exploration.

| CONCLUSIONS
Quality improvement initiatives ought not to rely only on incident reports as the one source of truth, and ongoing training to ensure clinicians report all falls to incident management systems is still necessary.This study reaffirms the recommendation that research studies (including quality improvement initiatives) focussed on in-hospital falls should utilise more than one method when recording falls. 16
Characteristics of falls by reporting method.