Frequencies of falls in Swiss hospitals: Concordance between nurses’ estimates and fall incident reports

Available online at www.sciencedirect.com

International Journal of Nursing Studies 46 (2009) 164–171 www.elsevier.com/ijns

Frequencies of falls in Swiss hospitals: Concordance between nurses’ estimates and fall incident reports Barbara Cina-Tschumi a, Maria Schubert a, Reto W. Kressig b, Sabina De Geest a, Rene´ Schwendimann a,* a

Institute of Nursing Science, University of Basel, Bernoullistrasse 28, 4056 Basel, Switzerland b Geriatric University Clinic, University of Basel, Basel, Switzerland

Received 28 March 2008; received in revised form 11 September 2008; accepted 12 September 2008

Abstract Background: Patient falls are frequent incidents in hospitals, and various measurement methods are described in the literature to assess in-patient fall rates. However, the literature includes no validation of nurses’ estimates of fall frequencies, which are the preferred assessment method in multi-centre surveys. Objectives: To explore the concordance of nurses’ estimated fall frequencies with continuously collected data. Design: Cross-sectional, correlational secondary data analysis. Sample/Setting: Patient fall data from 21 wards in 2 Swiss acute care hospitals participating in the RICH Nursing Study. Methods: Registered nurses’ (N = 233) estimated fall frequencies, assessed by the International Hospital Outcome Study questionnaire in absolute number of falls over the last month, and, using a four-point Likert scale (never = 1; frequently = 4), over the last year, were compared to standardized hospital fall incident reports compiled over the same periods. Fall incident reports for the last month were assessed in absolute numbers and were calculated as fall rates per 1000 patient days, with data computed at the ward level. The concordance with nurses’ estimates was then tested using Spearman’s rho and Kendall’s tau correlations. Results: The mean last-year fall frequencies estimated by nurses on the four-point Likert scale ranged from 1.4 to 3.1 for noninjurious falls and from 1.0 to 2.6 for injurious falls per ward. The fall rates assessed using fall incident reports over the same period ranged from 0.1 to 3.8 non-injurious falls per 1000 patient days and from 0.1 to 2.6 injurious falls per ward. Nurses’ estimates and fall incident reports correlated significantly regarding the last year, both for injurious falls (r = 0.685, p = 0.014) and non-injurious falls (r = 0.630, p = 0.028), although no statistically significant correlations were found regarding the 1 month estimates. Conclusions: Nurses’ long-term estimates of patient incidents are concordant with continuously and systematically assessed data, and offer valid data where other measurement methods are unavailable. # 2008 Elsevier Ltd. All rights reserved. Keywords: Accidental falls; Correlations; Fall frequencies; Hospital incident reporting; Nurses’ estimates

What is already known about the topic?

* Corresponding author. Tel.: +41 61 267 09 19; fax +41 61 267 09 55. E-mail address: [email protected] (R. Schwendimann).

 Various methods are used to assess fall frequencies in hospitals.  Nurses’ estimated fall frequencies are the preferred assessment method when no continuously collected fall data are available.

0020-7489/$ – see front matter # 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijnurstu.2008.09.008

B. Cina-Tschumi et al. / International Journal of Nursing Studies 46 (2009) 164–171

What this paper adds  This study substantiates the concordance of nurses’ estimated fall frequencies in comparison to fall incident reports over a 1-year period.  Evidence is presented that nurses’ perceptions are a valuable source of fall incident data when other measurement methods are unavailable. 1. Background In hospitals, one of the most cited indicators of patient safety is the rate of adverse events. As assessed in the UK’s 2001/2002 National Patient Safety Agency multi-centre study, slips, trips and falls accounted for 41% of that period’s 28,998 voluntarily reported incidents (Shaw et al., 2005). This confirmed Sutton et al. (1994)’s finding that falls represented the most frequent accidents among hospitalised patients (69.6%). In contrast, in the Harvard Medical Practice Study, falls accounted only for 2.1% (Brennan et al., 1991) of adverse patient events. These discrepancies in frequencies of adverse events may be explained by different methodological approaches and the focus of what was assessed. The widely accepted definition of a fall as ‘‘an unexpected event in which the person comes to rest on the ground, floor, or lower level,’’ as proposed by the Profane Group (Hauer et al., 2006) was not always applied in these studies. Differing fall rates in hospitals – varying from 1.3 falls per 1000 patient days (PDs) in a general University hospital (Tan et al., 2005) to 19.2 falls per 1000 PDs in an acute care hospital’s geriatric department (Hanger et al., 1999) – are due to definitional inconsistencies as well as diverse medical disciplines. The higher proportion of old and frail patient population are generally considered to result in higher fall rates (Krause, 2005; von Renteln-Kruse and Krause, 2004). Each fall is one to much in the light of its consequences as physical injuries, negative psychological outcomes for the faller as well as financial implications with prolonged lengths of stay, higher degrees of dependence and therefore increased care costs making them a burden for the entire healthcare system (Cho et al., 2003; Krause, 2005; Krauss et al., 2005). Furthermore, even though different clinical departments in hospitals indicate unequal fall frequencies, the clinical discipline alone does not explain fall rates’ variability. Other influences include fall prevention interventions (Oliver et al., 2007), variations in patient and organisational characteristics, as well as the measurement methods used to assess falls. Various recording methods are in use, including fall incident reports (Krause, 2005; Mark and Burleson, 1995; Schwendimann, 1998), fall registration in a hospital’s central adverse event database (Krauss et al., 2005), review of various medical, nursing or administrative records (Grenier-Sennelier et al., 2002a; Krauss et al., 2005; Murff et al., 2003; Sutton et al., 1994a,b,c), nurses’ estimates

165

(Blegen et al., 2004; Schubert et al., 2008) and patient interviews (Krauss et al., 2005; Sutton et al., 1994a,b,c). Each method has its strengths and weaknesses, and the accuracy is influenced by the length of observation. For example, interviewing patients to identify fallers is labourintensive (and therefore expensive) or may only represent a small sample and/or a limited period; chart reviews are time consuming, but can only indicate falls that have been registered; administrative databases are prone to coding errors. Yet, whatever other weaknesses a measuring method may have, the most difficult to correct is simple underreporting. Whereas von Renteln-Kruse and Krause (2004) assumed that underreporting could be neglected, other investigators estimate that 20–50% of hospital inpatient falls may go unreported (Blegen et al., 2004; Grenier-Sennelier et al., 2002a; Sutton et al., 1994a,b,c). In particular, Blegen et al. (2004) and Evans et al. (2006) indicated that falls that did not result in injuries were more often unreported than those that did. Underreporting remains a critical issue in all kinds of investigations and settings. Nuckols et al. (2007), for example, observed that in a voluntary hospital reporting system higher number of falls or drug errors were reported than those involving surgery, concluding that incident reporting systems inherent important limitations. In paediatric inpatient settings, voluntary reporting was also significantly higher: compared to figures obtained from a nurse survey, only about 30% of medication errors resulted in incident reports (Antonow et al., 2000). For the most accurate figures, however, the patients themselves must be consulted. An examination of three reporting methods in nursing homes showed that the highest numbers of falls were obtained by residents’ self-reports, whereas the highest agreement on falls was between chart reviews and incident reports (Kanten et al., 1993). In noninstitutional community settings, however, fall self-reports, was more likely to involve underreporting than calendar reports (Mackenzie et al., 2006). While difficult to obtain, however, accurate fall frequency data are essential indicators of quality of nurse sensitive care and patient safety (Dugan et al., 1996; McGillis Hall et al., 2004; Sovie and Jawad, 2001). Although manual chart review has been considered an accurate method for identifying fall frequency, as stated above, it is time consuming and reflects only documented falls. Furthermore, since systematic fall incident registers are not always employed, and may be neither accessible nor comparable between hospitals in multi-site studies, or even between units in the same institution, the use of nurses’ estimates for falls and other adverse events has been suggested as a cost-effective alternative (Aiken et al., 2002; McCusker et al., 2004; Sochalski, 2004). This method offers quick data access for extensive surveys and allows comparability in multi-centre studies. To this end, nurses were asked to assess the quality of care on their unit using a four-point scale. This method has also been used to estimate rates of medication errors, complaints from patients, work-related

166

B. Cina-Tschumi et al. / International Journal of Nursing Studies 46 (2009) 164–171

Table 1 Nurses’ characteristics.

N Female gender Age group in years, median (IQRa) Work time per month at work, median (IQR) Years worked at this ward, median (IQR) a

Total

Hospital A

Hospital B

233 208 (89.1%) 30–40 (20–30 to 30–40) 90% (70–100%) 3.4 (1.6–7.3)

103 88 (85%) 30–40 (20–30 to 40–50) 80% (60–100%) 3.5 (1.8–9.3)

130 120 (92.2%) 30–40 (20–30 to 30–40) 90% (70–100%) 3.8 (1.3–5.4)

IQR: interquartile range.

injuries to employees, ventilator-associated pneumonia and catheter-associated sepsis (Manojlovich and DeCicco, 2007; McCusker et al., 2004; Sochalski, 2004). Authors using nurse-rated indicators consider them reliable and valid, but also acknowledge that they could be subject to respondent bias and recall bias, both of which are generally associated with self-reporting (Manojlovich and DeCicco, 2007; Sochalski, 2004). For example, the definitions of ‘excellent’ quality of nursing care or ‘frequent’ falls could vary considerably across respondents, as well as carrying a potential for raising or lowering estimations of outcomes. Even though nurses’ estimated fall frequencies have not yet been used very frequently, it showed usefulness in broad investigations (such as in Sochalski, 2004). However, their use has not been validated in association with continuously collected fall frequency data. Therefore, it is relevant to compare fall frequencies estimated by nurses to other methods of fall registry such as systematic, continuous fall incident reporting. The objectives of this study were therefore to explore the concordance of nurses’ estimated fall frequencies from the RICH Nursing Study (Schubert et al., 2008) and fall frequencies captured by hospital fall incident reports. The following research questions guided this study: 1. What are the nurses’ estimated fall frequencies? 2. What number of falls and what fall rates were assessed using fall incident reports? 3. Is there an association between nurses’ estimated fall frequencies and fall frequencies assessed by fall incident reports? 2. Methods 2.1. Design In a secondary data analysis, the fall frequency data of the cross-sectional, multi-centre Rationing in Nursing Care in Switzerland (RICH) Study (Schubert et al., 2008) were compared with the number of falls from available hospital fall incident reporting systems to determine their concordance. The RICH Nursing Study (described elsewhere in detail) aimed to assess the relationship between implicit rationing

of nursing care in Swiss acute care hospitals and selected patient and nurse outcomes (Schubert et al., 2008). It was conducted with the approval of all relevant ethics committees. As no personal data on patients’ identities were sought from the fall incident registers, no further ethical approval was needed for the use of these data. Data of the participating hospitals were handled confidentially. 2.2. Sample/Setting/Sampling method Fall frequencies estimated by nurses (N = 233) for the RICH Nursing Study were used for the present study. In the eight acute care hospitals that participated in the RICH Nursing Study, all registered nurses holding a Swiss diploma or foreign equivalent were invited to participate. As further inclusion criteria, they were required to have worked in direct patient care at their hospitals for at least 3 months, including at least 1 month on their current unit.1 Three of these eight hospitals maintained fall incident registries during the period under study, and were therefore approached for participation in the current study. Agreement to participate and fall data were obtained from two of these three hospitals (one with over 300 beds, one with over 800 beds). In both participating hospitals, all wards where fall incident registers were available for the full year preceding the RICH Nursing Survey were included. In total, 21 wards (6 medical, 14 surgical and 1 gynaecological) (Table 1) and 233 nurses were incorporated for further analysis. 2.3. Variables and measurement 2.3.1. Nurses’ estimates The nurses’ estimated fall frequencies were assessed using three questions originally used in the International Hospital Outcome Study (IHOS) questionnaire (Aiken et al., 2001). For two questions, nurses were asked to rate the occurrence of: (a) non-injurious patient falls; and (b) injurious patient falls during the last year, on a four-point Likert scale including never (1), rarely (2), and sometimes (3) and frequently (4). Respondents also scored an additional question asking about absolute numbers of falls occurring during the last month, again distinguishing between falls with and 1

More details are given elsewhere.

B. Cina-Tschumi et al. / International Journal of Nursing Studies 46 (2009) 164–171

without injuries. No further definitions were given regarding what constitutes a fall or what is considered an injury. 2.3.2. Fall incident reporting system Fall incident reports were characterized by the obligation, mainly of registered nurses, to report every fall on a comprehensive register. Both hospitals had implemented their fall registers in 2002, but differed in the way individual falls and essential characteristics were entered into their electronic data processing systems. The current study included the following items from the fall incident registers: patients’ age and sex, unit, date and hour when the fall occurred, whether the fall was or was not directly observed, and the consequences of each fall. ‘Non-injurious falls’ were categorized as falls resulting in no consequences beyond fright of the patient, whereas ‘injurious falls’ categorized all falls with physical consequences such as pain, bruises, lacerations, distortions and fractures. Fall frequencies were summed per ward for the last month, while fall rates per 1000 patient days were computed over the entire year under investigation (total number of falls/number of patient days  1000). 2.4. Data collection The data regarding nurses’ responses to falls was collected in Hospital A in May 2004 and in Hospital B in March 2004. The questionnaires were distributed on a defined day to all nurses who met the inclusion criteria. Then, for the following 4 weeks, completed questionnaires were collected in a closed box on each participating ward. The fall incident register data were collected continuously, comprising the 12 months preceding the nurses’ survey, and were obtained retrospectively from the hospitals.

167

2.5. Statistics Frequencies, medians and interquartile ranges (IQR) or means and standard deviations (S.D.s) were calculated, where applicable, for the nurse estimate data sets and fall incident registers for each ward, for each hospital and over the total sample, distinguishing between falls with and without injury. Wards were chosen as the level for further analysis of the data final month’s data. For each question, mean values of the nurses’ answers were computed for each ward. For the data analysis of the entire year, ward clusters, which pooled ward data, had to be formed according to clinical disciplines, e.g., medicine and surgery. This approach was preferable because the numbers of patient days (as denominators for the fall rate calculations) in Hospital B (and in one case in Hospital A) were not accessible for the single wards but only for clinics. Appropriateness of data pooling was examined by assessing response agreement using intraclass correlation—an indicator of how much of the nurses’ fall estimate variability was shared by others working in the same ward or discipline. Reliable measurement requires that estimates within wards/disciplines correlate more closely than between wards/disciplines. The fall frequencies estimated by nurses and the fall frequencies registered by fall incident reports were compared using Spearman’s rho correlation test and Kendall’s tau correlation test. Falls without injuries were tested separately from falls with injuries. The tests were chosen based on the data characteristics of two small dependent samples and no normal distribution. The Spearman rank correlation coefficient quantifies the degree of linear association between the rankings of nurses’ estimates and those of fall incident reports. Kendall’s tau is recommended if the sample

Table 2 Patients’ characteristics and characteristics of falls in fall incident reports. Total (N = 477) Patients’ characteristics Gender Female Male

in age groups 59 60–79 80+ Missing data (N) Characteristics of falls Observed falls (% of all falls) a

QR: interquartile range.

Hospital B (N = 152)

136 (44.0%) 173 (56.0%)

Missing data (N) Age in years, median (IQRa)

Hospital A (N = 325)

16

152

79 (68–84)

79 (71–85)

71 (63.5–81.5)

9.7% 43.2% 46.9%

9.8% 40.3% 49.8%

18.1% 54.5% 27.3%

84

20

64

29.2%

28.9%

30.5%

168

B. Cina-Tschumi et al. / International Journal of Nursing Studies 46 (2009) 164–171

Table 3 Results of fall incident reports. Total

Falls Without injury With injury Missing data Fall rate (per 1000 PDsa) Without injury With injury a

Hospital A

Hospital B

Last year

Last month

Last year

Last month

Last year

Last month

477 300 (62.9%) 173 (36.3%) 4

26 18 8 0

325 204 (62.8%) 121 (37.2%) 0

15 10 5 0

152 96 (63.2%) 52 (34.2%) 4

11 8 3 0

3.0 1.9 1.1

3.0 1.9 1.1

2.9 1.8 1.0

PDs: patient days (total: N = 159,543; Hospital A: N = 107,052; Hospital B: N = 53,491).

contains a considerable number of tied ranks. The alpha value was set at 0.05. All statistical testing was performed using SPSS version 14 (SPSS Inc., Chicago, IL, USA).

(S.D. 0.68); Hospital B mean: 0.14 (S.D. 0.38) (missing data: 60)). Intraclass correlations for the yearly fall estimates were 37% (without injury) and 29% (with injury); for the monthly estimates, the corresponding figures were 5% and 9%, respectively.

3. Results 3.2. Fall incident reports 3.1. Nurses’ estimated fall frequencies On the wards studied, of 395 distributed questionnaires, 233 were returned (response rate: 59%). Characteristics of the participating nurses are shown in Table 2. Over the entire year, the mean fall frequencies estimated by nurses on the four-point Likert-type scale (never = 1; frequently = 4) ranged from 1.4 to 3.1 for falls without injuries per ward (Hospital A mean: 2.63 (S.D. 0.88); Hospital B mean: 2.38 (S.D. 0.73) (missing data: 8)) and from 1.0 to 2.6 for falls with injuries (Hospital A mean: 2.04 (S.D. 0.80) and Hospital B mean: 1.66 (S.D. 0.67) (missing data: 30)). The mean fall frequencies for the nurses’ estimates of the preceding month ranged from 0 to 2 non-injurious falls per ward (Hospital A mean: 1.07 falls (S.D. 1.55); Hospital B mean: 0.83 falls (S.D. 1.24) (missing answers: 44)) and from 0 to 0.7 injurious falls per ward (Hospital A mean: 0.41 falls

In total, 477 falls (Hospital A: 325; Hospital B: 152) were registered over the preceding year. The ratios of injurious to non-injurious falls were similar for both hospitals. The number of reported falls per ward varied considerably, ranging from 1 to 49 for non-injurious falls and 0 to 34 for falls resulting in injuries. The fall rate over the entire year ranged from 0.1 to 3.8 non-injurious falls per 1000 patient days and from 0.1 to 2.6 injurious falls per 1000 patient days per ward or ward cluster. Characteristics of the fallers and the fall incident reports are given in Table 3 and overall fall frequencies in Table 4. A total of 26 falls (16 without injuries, 8 with injuries) were reported on the fall incident register for the last month preceding the RICH Nursing Survey. The overall number of falls ranged from 0 to 5 non-injurious falls and 0 to 2 injurious falls per ward. The proportion of falls without injuries was similar

Table 4 Fall frequencies on last year (aggregated according to clinical discipline). Wards (N)

Non-injurious patient falls Nurses’ estimates

a

N

Mean (S.D.)

Hospital A Medical (3) Surgical (5) Gynaecological (1)

43 47 7

2.55 (0.69) 2.91 (0.92) 1.43 (0.79)

Hospital B Surgical (9) Medical (3)

91 37

2.35 (0.78) 2.43 (0.60)

Injurious patient falls FIR falls per 1000 PDs

Nurses’ estimatesa

FIR falls per 1000 PDs

N

Mean (S.D.)

2.78 1.73 0.13

42 43 7

1.91 (0.81) 2.36 (0.66) 1.00 (0.00)

1.71 0.95 0.13

1.89 1.48

79 32

1.67 (0.67) 1.63 (0.66)

0.90 1.48

FIR: fall incident reports; PD: patient day; S.D. = standard deviation. a Rated on a four-point Likert scale ranging from 1 = never to 4 = frequent.

B. Cina-Tschumi et al. / International Journal of Nursing Studies 46 (2009) 164–171

169

Table 5 Fall frequencies over last month (aggregated according to clinical discipline). Wards (N)

Non-injurious patient falls Nurses’ estimates

a

N

Mean (S.D.)

Hospital A Medical (3) Surgical (5) Gynaecological (1)

36 39 7

1.00 (1.36) 1.31 (1.83) 0.29 (0.49)

Hospital B Surgical (9) Medical (3)

78 29

0.86 (1.36) 0.76 (0.87)

Injurious patient falls FIR range

Nurses’ estimates a

FIR range

N

Mean (S.D.)

0–5 0–2 0

31 38 7

0.34 (0.67) 0.58 (0.72) 0.00 (0.00)

0–2 0–1 1

0–5 0–1

70 27

0.16 (0.40) 0.11 (0.32)

0–1 0–1

S.D. = standard deviation; FIR: fall incident reports. a Rating in absolute numbers.

between the hospitals. In Hospital A the number of falls reported during the last month (15 falls) differed considerably from the mean number of falls per month (27 falls); in Hospital B this difference was less pronounced (11 falls over the last month; mean number of falls per month: 13). 3.3. Correlations between nurses’ estimates and fall incident reports Nurses’ estimated fall frequencies correlated significantly with fall rates based on incident reports for the last year, both for non-injurious falls (Spearman’s rho: N = 12, r = 0.685, p = 0.014; Kendall’s tau: N = 12, r = 0.545, p = 0.014) and for those resulting in injuries (Spearman’s rho: N = 12, r = 0.630, p = 0.028; Kendall’s tau: N = 12, r = 0.473, p = 0.033). In contrast, for the last-month data, nurses’ estimates showed no significant correlations with fall incident reports, either for non-injurious falls (Spearman’s rho: N = 21, r = 0.324, p = 0.151; Kendall’s tau: N = 21, r = 0.259, p = 0.150) or for injurious falls (Spearman’s rho: N = 21, r = 0.286, p = 0.209; Kendall’s tau: N = 21, r = 0.247, p = 0.191) (Table 5).

4. Discussion This is the first study to explore the concordance of nurses’ estimated fall frequencies with fall registries. Our findings suggest that, where no standardized fall incident reporting system is available, nurses’ estimates over the last year are a valuable data source for fall evaluation. This study’s observed fall rates are comparable to those recorded in the literature, and the overall rate yielded by the fall incident register – 3.0 falls per 1000 patient days – is in line with the 2.2–3.29 patient falls per 1000 patient days given here (Halfon et al., 2001; Krauss et al., 2005). As surgical wards comprised two-thirds of our sample, our findings are comparable to 2.1 falls per 1000 patient days

on those wards (Milisen et al., 2007), considerably lower than in samples consisting of 58% and 66% medical or geriatric patients, which yielded, respectively, 7.3 and 8.9 falls per 1000 patient days (Milisen et al., 2007; Schwendimann et al., 2006). The normal tendency toward higher fall frequencies on medical wards than on surgical or even obstetric wards could also be observed in our sample. Comparing nurses’ estimated fall frequencies to fall incident reports, our analyses showed that nurses’ estimated fall frequencies over the period of 1 year were more accurate than over 1 month. The Likert scale rating over the last year appears to have simplified the task, as only four answering possibilities were given and no exact number had to be recalled. In contrast, nurses’ estimates of the exact number of falls over the shorter period of 1 month showed low concordance, with no differences shown between non-injurious and injurious falls. These findings are in line with HillWestmoreland and Gruber-Baldini (2005) who found a 75% concordance between Minimum Data Set (MDS) fall variables and falls abstracted from resident charts over a 180-day period, whereas, for a 30-day period, concordance was only 65%. Our findings contrast further with physical restraint episodes and assaultive/aggressive episodes in in-patient psychiatry, where the majority of nurses were able to recall these events within a 20% margin over periods of 2 and 4 weeks (Gerolamo, 2008). Furthermore, the tendency for injurious accidents in hospital to be more frequently reported (Blegen et al., 2004) – and therefore more easily recalled – was not shown in our sample, as the correlation between nurse recollections and injurious falls was actually slightly lower than for non-injurious ones. There are several possible explanations why the 1-month estimates were more difficult. As known from diagnostic tests, accuracy decreases with low prevalence and is further challenged by a small sample size (Bachmann et al., 2006; Obuchowski and Zhou, 2002). As the number of non-injurious or injurious falls in 1 month varied from 0 to 5 or from 0 to 2 falls recorded in the fall incident reports, recalling such small ranges may be very challenging. The small intraclass

170

B. Cina-Tschumi et al. / International Journal of Nursing Studies 46 (2009) 164–171

correlations supported the view that nurses had difficulties reliably estimating the exact number of falls over this relatively short time period. Also, there could be an association between the nurses’ time at work and the number of falls they were able to recall. Considering that half of the nurses surveyed worked 80% or less, some bias is possible, as parttime nurses’ estimates would not reflect fall frequencies for periods when they were not present. For a Likert scale rating, as used for the 1-year estimates, this bias may be minimized, as the nurses are more likely to rate the frequencies of falls in relation of their time present at work. Certain other limitations also have to be kept in mind when interpreting our findings. One is that the IHOS questionnaire provides no definitions to differentiate between non-injurious and injurious falls; therefore, nurses had to answer according to their own differentiations, which may have differed from those governing the fall incident register. A further drawback of this secondary data analysis is the limited sample size, since only a small proportion of the hospitals and wards of the RICH Nursing Study conducted a fall incident register for the period covered by the nurses’ survey. Our analysis lost further power due to data gaps: for the last month, between 19% and 25% of answers were omitted, respectively, for non-injurious and injurious falls. Finally, regarding the fall incident registers, the extent of underreporting can only be estimated, but may account for 20–50% (Blegen et al., 2004). This suggests an effective overall fall rate of 3.6–4.5 per 1000 patient days. One issue concerning the external validity of this study is the potential effect of an established fall incident register on increasing nurses’ awareness of patient falls, which has to be considered when our findings are applied to hospitals not employing fall incident registers. Overall, as we found promising indices that, in the absence of a consecutive fall incident register, nurses’ estimates are a valuable assessment method for frequencies of patient falls over 1 year, these findings will have to be confirmed using a larger sample. This assessment method could also be successfully applied to other patient incident types (e.g., pressure ulcers or medication errors). For future investigations addressing the accuracy of nurses’ estimates, we suggest standardizing fall incident reports and variables to increase comparability between multiple sites. It might also be useful to assess which nurses estimate fall frequencies most accurately. In order to have a reliable estimate of falls occurring in hospital, a systematic fall incident register currently provides the most precise information possible in view of absolute numbers, including types of injuries following a fall. However, when fall incident registers are unavailable, nurses’ estimates may be used as general indicators.

5. Conclusions Nurses’ estimated fall frequencies show significant concordance with fall rates registered in fall incident reports

over the period of 1 year. This association was confirmed for a Likert scale rating that offered only a rather crude estimate of the frequencies of falls. However, recalling absolute numbers appears to be more challenging, as few individual nurses learn of each fall that happens on their wards. As an assessment method, nurse estimates appear reliable for larger surveys such as the RICH Nursing Study, but further investigation is recommended to strengthen evidence. Especially when no continuously collected data is available, nurses’ estimates may be a valuable data source for quantification of falls.

Funding source There was no funding to realize this study. Data of the nurses’ estimates are a secondary data analysis of the RICH Nursing Study which has been approved by all relevant ethics committees. The use of hospital fall incident data (no data on patients identity) was approved by the hospital management.

Acknowledgments Special thanks goes to Kris Denhaerynck, PhD, RN, and Tracy Glass, MSc, for their statistical advices and to Chris Shultis for editing the manuscript.

Conflict of interest None declared.

References Aiken, L.H., Clarke, S.P., Sloane, D.M., 2002. Hospital staffing, organization, and quality of care: cross-national findings. Nurs. Outlook 50 (5), 187–194. Aiken, L.H., Clarke, S.P., Sloane, D.M., Sochalski, J.A., Busse, R., Clarke, H., Giovannetti, P., Hunt, J., Rafferty, A.M., Shamian, J., 2001. Nurses’ reports on hospital care in five countries. Nurses Rep. 20 (3), 43–53. Antonow, J.A., Smith, A.B., Silver, M.P., 2000. Medication error reporting: a survey of nursing staff. J. Nurs. Care Qual. 15 (1), 42–48. Bachmann, L.M., Puhan, M.A., ter Riet, G., Bossuyt, P.M., 2006. Sample sizes of studies on diagnostic accuracy: literature survey. BMJ 332 (7550), 1127–1129. Blegen, M.A., Vaughn, T., Pepper, G., Vojir, C., Stratton, K., Boyd, M., Armstrong, G., 2004. Patient and staff safety: voluntary reporting. Am. J. Med. Qual. 19 (2), 67–74. Brennan, T.A., Leape, L.L., Laird, N.M., Hebert, L., Localio, A.R., Lawthers, A.G., Newhouse, J.P., Weiler, P.C., Hiatt, H.H., 1991. Incidence of adverse events and negligence in hospitalized patients. Results of the Harvard Medical Practice Study I. N. Engl. J. Med. 324 (6), 370–376.

B. Cina-Tschumi et al. / International Journal of Nursing Studies 46 (2009) 164–171 Cho, S.H., Ketefian, S., Barkauskas, V.H., Smith, D.G., 2003. The effects of nurse staffing on adverse events, morbidity, mortality, and medical costs. Nurs. Res. 52 (2), 71–79. Dugan, J., Lauer, E., Bouquot, Z., Dutro, B.K., Smith, M., Widmeyer, G., 1996. Stressful nurses: the effect on patient outcomes. J. Nurs. Care Qual. 10 (3), 46–58. Evans, S.M., Berry, J.G., Smith, B.J., Esterman, A., Selim, P., O’Shaughnessy, J., DeWit, M., 2006. Attitudes and barriers to incident reporting: a collaborative hospital study. Qual. Saf. Health Care 15 (1), 39–43. Gerolamo, A.M., 2008. Measuring adverse outcomes in inpatient psychiatry: the reliability of nurse recall. Arch. Psychiatr. Nurs. 22 (2), 95–103. Grenier-Sennelier, C., Lombard, I., Jeny-Loeper, C., Maillet-Gouret, M.-C., Minvielle, E., 2002a. Designing adverse event prevention programs using quality management methods: the case of falls in hospitals. Int. J. Qual. Health Care 14 (5), 419–426. Halfon, P., Eggli, Y., Van Melle, G., Vagnair, A., 2001. Risk of falls for hospitalized patients: a predictive model based on routinely available data. J. Clin. Epidemiol. 54 (12), 1258–1266. Hanger, H.C., Ball, M.C., Wood, L.A., 1999. An analysis of falls in the hospital: can we do without bedrails? J. Am. Geriatr. Soc. 47 (5), 529–531. Hauer, K., Lamb, S.E., Jorstad, E.C., Todd, C., Becker, C., 2006. Systematic review of definitions and methods of measuring falls in randomised controlled fall prevention trials. Age Ageing 35 (1), 5–10. Hill-Westmoreland, E.E., Gruber-Baldini, A.L., 2005. Falls documentation in nursing homes: agreement between the minimum data set and chart abstractions of medical and nursing documentation. J. Am. Geriatr. Soc. 53 (2), 268–273. Kanten, D.N., Mulrow, C.D., Gerety, M.B., Lichtenstein, M.J., Aguilar, C., Cornell, J.E., 1993. Falls: an examination of three reporting methods in nursing homes. J. Am. Geriatr. Soc. 41 (6), 662–666. Krause, T., 2005. Sturzfolgen bei geriatrischen Krankenhauspatienten. Pflege 18 (1), 39–42. Krauss, M.J., Evanoff, B., Hitcho, E., Ngugi, K.E., Dunagan, W.C., Fischer, I., Birge, S., Johnson, S., Costantinou, E., Fraser, V.J., 2005. A case–control study of patient, medication, and care-related risk factors for inpatient falls. J. Gen. Intern. Med. 20 (2), 116–122. Mackenzie, L., Byles, J., D’Este, C., 2006. Validation of selfreported fall events in intervention studies. Clin. Rehabil. 20 (4), 331–339. Manojlovich, M., DeCicco, B., 2007. Healthy work environments, nurse-physician communication, and patients’ outcomes. Am. J. Crit. Care 16 (6), 536–543. Mark, B.A., Burleson, D.L., 1995. Measurement of patient outcomes: data availability and consistency across hospitals. J. Nurs. Adm. 25 (4), 52–59. McCusker, J., Dendukuri, N., Cardinal, L., Laplante, J., Bambonye, L., 2004. Nursing work environment and quality of care: differences between units at the same hospital. Int. J. Health Care Qual. Assur. Inc. Leadersh Health Serv. 17 (6), 313–322. McGillis Hall, L., Doran, D., Pink, G.H., 2004. Nurse staffing models, nursing hours, and patient safety outcomes. J. Nurs. Adm. 34 (1), 41–45.

171

Milisen, K., Staelens, N., Schwendimann, R., De Paepe, L., Verhaeghe, J., Braes, T., Boonen, S., Pelemans, W., Kressig, R.W., Dejaeger, E., 2007. Fall prediction in inpatients by bedside nurses using the St. Thomas’s Risk Assessment Tool in Falling Elderly Inpatients (STRATIFY) Instrument: a multicenter study. J. Am. Geriatr. Soc. 55 (5), 725–733. Murff, H.J., Patel, V.L., Hripcsak, G., Bates, D.W., 2003. Detecting adverse events for patient safety research: a review of current methodologies. J. Biomed. Inform. 36 (1–2), 131–143. Nuckols, T.K., Bell, D.S., Liu, H., Paddock, S.M., Hilborne, L.H., 2007. Rates and types of events reported to established incident reporting systems in two US hospitals. Qual. Saf. Health Care 16 (3), 164–168. Obuchowski, N.A., Zhou, X.H., 2002. Prospective studies of diagnostic test accuracy when disease prevalence is low. Biostatistics 3 (4), 477–492. Oliver, D., Connelly, J.B., Victor, C.R., Shaw, F.E., Whitehead, A., Genc, Y., Vanoli, A., Martin, F.C., Gosney, M.A., 2007. Strategies to prevent falls and fractures in hospitals and care homes and effect of cognitive impairment: systematic review and metaanalyses. BMJ 334 (7584), 82. Schubert, M., Glass, T.R., Clarke, S.P., Aiken, L.H., SchaffertWitvliet, B., Sloane, D.M., De Geest, S., 2008. Rationing of nursing care and its relationship to patient outcomes: the Swiss extension of the International Hospital Outcomes Study. Int. J. Qual. Health Care. Schwendimann, R., 1998. Ha¨ufigkeit und Umsta¨nde von Sturzereignissen im Aktuspital: Eine Pilotstudie. Pflege 11 (6), 335– 341. Schwendimann, R., Buhler, H., De Geest, S., Milisen, K., 2006. Falls and consequent injuries in hospitalized patients: effects of an interdisciplinary falls prevention program. BMC Health Serv. Res. 6, 69. Shaw, R., Drever, F., Hughes, H., Osborn, S., Williams, S., 2005. Adverse events and near miss reporting in the NHS. Qual. Saf. Health Care 14, 279–283. Sochalski, J., 2004. Is more better?: the relationship between nurse staffing and the quality of nursing care in hospitals. Med. Care 42 (2 Suppl.), II67–II73. Sovie, M.D., Jawad, A.F., 2001. Hospital restructuring and its impact on outcomes. J. Nurs. Adm. 31 (12), 588–600. Sutton, J., Standen, P., Wallace, A., 1994a. Accidents to patients in hospital: a comparative study. Nurs. Times 90 (36), 52–54. Sutton, J., Standen, P., Wallace, A., 1994b. Incidence and documentation of patient accidents in hospital. Nurs. Times 90 (30), 29–35. Sutton, J., Standen, P., Wallace, A., 1994c. Unreported accidents to patients in hospital. Nurs. Times 90 (39), 46–49. Tan, K.M., Austin, B., Shaughnassy, M., Higgins, C., McDonald, M., Mulkerrin, E.C., O’Keeffe, S.T., 2005. Falls in an acute hospital and their relationship to restraint use. Irish J. Med. Sci. 174 (3), 28–31. von Renteln-Kruse, W., Krause, T., 2004. Sturzereignisse stationa¨rer geriatrischer Patienten-Ergebnisse einer 3-ja¨hrigen prospektiven Erfassung. Zeitschrift fu¨r Gerontologie und Geriatrie 37 (1), 9–14.