The use of physical restraints has been a common practice

Transcription

1 Original Article Development and Testing of Tools to Assess Physical Restraint Use Nancy Edwards, RN, PhD, Evangeline Danseco, PhD, Kathleen Heslin, RN, MScN, Jenny Ploeg, RN, PhD, Josephine Santos, RN, MN, Melanie Stansfield, RN, MScN, Barbara Davies, RN, PhD ABSTRACT Background: To implement best practice in restraint use, healthcare providers and decision makers require current, accurate, and easily accessible information about restraint practices in their setting. There is a need for a reliable and valid instrument that is feasible for use in these settings to rapidly assess physical restraint use. Method: Two instruments to assess physical restraint use were developed and tested: an observation tool and a chart audit. The instruments were tested in complex continuing care units and rehabilitation units at two healthcare organizations. The restraint use observation tool was administered by trained observers in a series of five observations over a 2-week period. Chart audits were conducted for a sample of residents. Inter-rater agreement of the observation tool and chart audit was assessed. Point prevalence estimates were obtained for each site. The time required to use the observation tool and complete chart audits as well as the comparability of findings from the two data sources were described. Findings: Restraint use observations were completed for 289 patients and chart audits were completed for 207 patients. Prevalence and patterns of restraint use varied between sites and across time periods. Observations took an average of 2.6 (Site A) and 0.6 (Site B) minutes per patient. There was excellent inter-rater agreement for most items on the observation tool with the exception of whether or not the patient was able to release or loosen the restraint. There was significant concordance but as expected, not complete agreement between paired estimates of prevalence using the observational tool and the chart audit. Conclusion: The observation and chart audit tools are feasible to use and reliably assess physical restraint use in healthcare organizations. The patient s physical capacity to move independently, the patient s waking status, and the restraint s restriction of mobility are items that should be added to the observation tool. The tools are complementary and should be used in tandem to capture the multifaceted complexity of restraint use in health service organizations. Worldviews on Evidence-Based Nursing 2006; 3(2): Copyright 2006 Sigma Theta Tau International KEYWORDS Physical restraints, measurement tool, audit Nancy Edwards, Professor, School of Nursing, Department of Epidemiology and Community Medicine, University of Ottawa, Ontario, Canada and Director, Community Health Research Unit, University of Ottawa. Evangeline Danseco, Research Specialist, Canadian Council on Health Services Accreditation, Ottawa. Kathleen Heslin, Chief Nursing Officer, West Park Healthcare Centre, Toronto. Jenny Ploeg, Associate Professor, School of Nursing, McMaster University, Hamilton, Ontario. Josephine Santos, Regional Best Practice Coordinator Long-term Care, Ontario Ministry of Health and Long-term Care, Toronto, Melanie Stansfield, Family Nurse Practitioner, Niagara Health Services, Niagara Falls. Ontario. Barbara Davies, Associate Professor, School of Nursing, University of Ottawa. Address correspondence to Nancy Edwards, Director, Community Health Research Unit, University of Ottawa, 451 Smyth Road, Room 1118, Ottawa, Ontario, K1H 8M5 Canada; nancy.edwards@uottawa.ca Accepted February 20, 2006 INTRODUCTION The use of physical restraints has been a common practice for many years (Joanna Briggs Institute 2002). However, there is mounting evidence of patient safety risks when physical restraints are used (Evans et al. 2003; Marcy- Edwards 2005; Mott et al. 2005). In a systematic review of 12 observational studies, Evans et al. (2003) concluded that physical restraint use in acute care and residential care settings may increase the risk of death (acute care settings, pooled results from three studies OR 11.24, 95% Copyright 2006 Sigma Theta Tau International X/06 Worldviews on Evidence-Based Nursing Second Quarter

2 CI: ), and falls (residential care settings, pooled results from two studies OR 1.72, 95% CI: ). In acute care settings, pooled results from two studies indicated that restrained patients were less likely to survive to being discharged and more likely to develop nosocomial infections (Evans et al. 2003). Qualitative studies indicate the negative experiences of people who have been restrained (Evans et al. 2002a) including emotional distress, loss of dignity and independence, dehumanization, increased agitation, and depression. No studies were found that reported improvement in patient safety following physical restraint use. A decline in some types of physical restraint use has followed the introduction of either minimal restraint legislation (Brush & Capezutti 2001; Dunn 2001; Shojania et al. 2001), or education programs and procedural changes for minimization of restraints (Levine et al. 1995; Evans et al. 1997; Evans et al. 2002b; Hoelzel 2003). Changes in patterns of restraint use have most often been assessed using chart data or restraint use forms (Tinetti et al. 1992; Kirkevold & Engedal 2004). The reliability of these data has not previously been reported. From the perspective of a manager, various features of restraint use must be understood if appropriate interventions to reduce the use of restraints are to be implemented at the organizational level. In this article, a study undertaken in 2004 in two healthcare organizations providing complex continuing care and rehabilitation services for older adults in Ontario, Canada is described. The purpose of the study was to assess the utility and reliability of two restraint use instruments: an observational tool and a chart audit. In Ontario, use of restraints in healthcare organizations is restricted via legislation except as a last resort and it requires organizational policies of least restraint to encourage the use of alternative methods whenever possible (Ontario Patient Restraints Minimization Act 2001). According to nursing standards for physical restraint use in Ontario (College of Nurses of Ontario 2004), nurses cannot use any form of restraint without client or substitute decisionmaker consent, other than in some emergency situations. Healthcare organizations specify whether a physician s order is required before using physical restraints (College of Nurses of Ontario 2004). Few guidelines exist to support nurses implementing minimal restraint policies in practice settings. In 2004, the Registered Nurses Association of Ontario (RNAO) published a clinical practice guideline on care-giving strategies for older adults with dementia, delirium, and depression (Registered Nurses Association of Ontario 2004). The guideline includes recommendations regarding minimal restraint use. Our team was asked to develop evaluation tools to assess the effect of guideline implementation. The restraint tools that we adapted and tested were a result of this work. 74 Second Quarter 2006 Worldviews on Evidence-Based Nursing A review of the extant literature showed that few observational tools for physical restraint use were reported. Those developed (Levine et al. 1995; Wynn 2002; Hoelzel 2003; McMahon & Fisher 2003) were site-specific, and had not undergone reliability testing. Furthermore, in previous studies, comparisons of results obtained from observational and chart audit tools were not reported. A tool that met several of the team s requirements was a physical restraint prevalence tool developed for an acute care hospital in Ontario by Rossy and Mackey (2003). At the time of our study, this tool had been used facility-wide in an acute care teaching hospital. However, it had not undergone reliability testing. With the permission of those who developed the original tool and the input of an expert team, the tool was revised for use with patients in complex continuing care and rehabilitation. METHODS Definition of Physical Restraint The operational definition of a physical restraint used in this study was drawn from the restraint definition of the minimum data set (MDS). The MDS is one of several components of care regularly assessed for residents in complex continuing care (Hirdes et al. 1998; Canadian Institute of Health Information 2002; Center for Medicare and Medicaid Services 2005). The MDS algorithm requires staff to identify whether residents can easily remove the device, material, or equipment, and whether the device restricts the resident s freedom of movement. Table 1 shows the algorithm that was used to determine whether a device was a physical restraint. Development of the Instruments: Restraint Use Observation Tool and Chart Audit Rossy and Mackey s (2003) tool consists of two sections. The first section requires an observer to record the number of patients in the unit, the number of patients with physical and chemical restraints, and whether the restraint is properly applied. The types of physical restraint devices that observers are instructed to note include mitts, wheelchair belts, soft waist belts, pelvic supports, limb holders, and four-point restraints. The second section of the tool requires the observer to review the charts of those with physical restraints for the 24-hour period before the time of observation. Items in this second section include: whether the restraint meets Ontario legislation requirements; whether a documentation record form of the hospital was in the chart; whether the nurse s physical restraint order was documented and if the order was specific to the device; evidence of patient assessment and exploration of alternatives; and evidence of consent, monitoring requirements, and patient and family education.

3 TABLE 1 Algorithm to determine whether a device is a physical restraint ITEM YES NO CANNOT BE DETERMINED Is there any manual method or physical or mechanical device, A potential restraint Not a restraint Not applicable material, or equipment attached or adjacent to the resident s body? Can resident easily remove or loosen the device? Not a restraint Definitely a restraint A potential restraint Does device restrict freedom of movement or normal access to one s body? Definitely a restraint Not a restraint A potential restraint The algorithm is based on the minimum data set (MDS) 2.0 Canadian Version (Canadian Institute of Health Information 2002). With the assistance of an expert group, several changes were made to Rossy and Mackey s (2003) tool. Table 2 indicates a summary of these changes. To assess content validity, eight clinicians with expertise in restraints and gerontology reviewed the revised tools for clarity of items and instructions, and comprehensiveness of topics covered. Their suggestions were incorporated in a final refinement of the tool. Participants In the fall of 2004, participants were recruited from two healthcare organizations selected by the RNAO as part of a formal request for proposal process. Site A is a hospital within a multisite healthcare network that has over 1,000 beds. The site has over 120 rehabilitation and complex continuing care beds. Site B is a healthcare facility providing regional rehabilitation and complex continuing care for over 450 patients. Nurses in both sites worked both 8- and 12-hour shifts. Most patients in both sites are age 65 or older. Procedures The ethics board of the University of Ottawa as well as the research ethics boards of both participating sites approved this study. Posters and flyers describing the study were posted in the participating units. A research coordinator spoke individually to the patient, family, or substitute decision maker; and to staff in the participating units to explain the study before data collection began. Participants provided verbal consent for the observations. Patients and family members were informed that chart audits would be done. These procedures were consistent with the requirements of the Tri-Council Policy Statement on ethical standards (Canadian Institutes of Health Research, Natural Sciences and Engineering Research Council of Canada, Social Sciences and Humanities Research Council of Canada 1998) for quality assurance studies. The research coordinators trained two data collectors per site on the use of the tools. Training took 2 6 hours per site. Data collectors at one site were third-year nursing students who were familiar with the site but not currently practising there. Data collectors at the second site were experienced health professionals, one a nursing manager, the other a member of the research department who did not provide direct care to patients in the participating units. Observations were made over a 2-week period, between 0700 and 2200 hours. Five randomly timed observations of restraint use were conducted so that variations in restraint use could be assessed. Random numbers were generated to determine the day of the week and the hour of the day when observations would be completed. The order for observing units was also randomized. Table 3 shows the schedule of observations for each of the sites. In Site A, the observations were limited to patients aged 65 and older, while in Site B all adult patients were observed. Demographic information was obtained from the daily census at Site A and from the chart audits in Site B. The research coordinators at each site randomly assigned data collectors to observe patients on half of the participating units. Data collectors noted the time they started their observations, including time spent locating patients who were not physically on the unit during the observation period. Ability to release the restraint was not assessed if patients were asleep or resting in bed. Data collectors at Site A conducted inter-rater observations on a random sample of 10% of the patients on 3 of the 5 observation days. Independent observations were made by each rater within a 10-minute span to decrease the possibility of a change in patient activity. Data collectors at Site B conducted the inter-rater agreement of the observation tool for all patients on the 1st and 5th days. Data collectors at the second site independently made their observations at the same time. Completed observation tools were mailed to the research team. Chart audits were conducted by the same people who completed the observational tool, about 3 weeks after the last day of the observation period in both sites. At Site A, chart audits were conducted on all patients for whom a physical restraint was noted during at least one observation period, and a random sample of 25% of patients for whom restraint use was not observed during any of the five observation periods. For Site B, chart audits were conducted for all patients in the participating units. Retrospective chart Worldviews on Evidence-Based Nursing Second Quarter

4 TABLE 2 Summary of modifications to Rossy & Mackey restraint prevalence tool ORIGINAL VERSION MODIFICATIONS AND RATIONALE ELEMENT OF TOOL (Rossy & Mackey 2003) FOR CHANGE Chart Audit Tool 1. Timing of chart audit Concurrent with observation, conducted Retrospective chart audit. during the sweep 2. Period reviewed 24 hours prior to the time of observation 12 hours before and 12 hours after the observation 3. Types of physical restraint devices 4. Documentation on reasons for restraint use 5. Documentation on alternatives and interventions attempted Observation Tool 1. Observation of physical restraint 2. Ability to release/loosen restraint Physical restraint devices coded as: mitts, limb holder, wheelchair belts, soft waist belt, pelvic support, and 4-point restraint. Reasons for restraint use coded as: impaired mobility; pulling out tubes/drains; agitation; combative; unable to follow instructions; disorientation/confusion; memory deficit; movement disorder; family request; and other. Alternatives and interventions coded as: regular toileting; occupational/physiotherapist; sensory aids; room change; family at bedside; sitter at bedside; diversional activities; increase mobility/activities of daily living (ADLs); positioning; posey alternatives (e.g., arm splint, floor mat, bed alarm); observation; explain procedure/treatment; pain management; medication review; nourishment; orientation/reassurance; other. Part of the data collection form, conducted immediately after chart review. Not included period Physical restraint devices coded as: limb restraints (including mittens), lap belts, trunk restraints, (4-point restraint excluded)bed rails (full and other types of bed rails), chair that prevents rising, and elevated wheelchair foot pedals. Codes were revised to reflect minimum data set (MDS) codes and the inventory of physical restraint devices available at the participating sites. Elevated wheelchair foot pedals were included because these can potentially reduce the mobility of patients (Simmons et al. 1995). Reasons for restraint use coded as: patient with impaired mobility; patient who tampers with medical devices; agitated or violent patient; cognitively impaired patient; patient at risk of falling; patient who wanders; patient undergoing medical treatments; patient is suicidal; and other. Alternatives and interventions coded as: toileting and continence; physiological alternatives; environmental changes; safety in bed; activities and programs; seating and position support; alterations to nursing care; psychological alternatives; alarms; other.codes were based on Sorrentino (2004) and the Joanna Briggs Institute guidelines on physical restraint (2002). Separate stand-alone tool, conducted before the chart review. Patient s ability to release/loosen restraint coded as yes, no, or unable to assess. Presence of physical restraint device is positive if patients cannot release/loosen the device. 3. Current activity Not included Current activity included: sleeping or resting in bed, in therapy or treatments, in group activities, in personal activities (e.g., in the washroom), eating, outdoors, moving in wheelchair, and leave of absence. 4. Start and end time of observation Not included Duration of observation period noted. audits were conducted for all 5 days that patients had been observed. Completed chart audit data were mailed to the research team. The inter-rater agreement of the chart audit tool was assessed in Site B where both data collectors independently audited all charts from a randomly selected unit. 76 Second Quarter 2006 Worldviews on Evidence-Based Nursing

5 TABLE 3 Schedule of random observations by site DAY OF START ORDER OF THE OBSERVATIONS a TIME UNITS OBSERVED Site A Day 1: Tuesday of week C, A, D, B Day 2: Wednesday of week D, A, C, B Day 3: Friday of week A, C, D, B Day 4: Sunday of week A, D, B, C Day 5: Thursday of week C, D, B, A Site B Day 1: Friday of week I, H, E, F, J, G Day 2: Sunday of week G, J, I, H, E, F Day 3: Monday of week J, G, I, F, E, H Day 4: Wednesday of week J, H, E, F, I, G Day 5: Monday of week F, J, H, I, G, E a For Site A, the 1st day of the 2-week period began on a Sunday. For Site B, the 1st day of the 2-week period began on a Wednesday. Analytic Strategy Data from the observations and chart audits were coded and entered into SPSS-PC version 12. Prevalence estimates of restraint use from observational data were derived from the scoring algorithm that allowed scorers to take into account the presence of a physical restraint device and the inability of a patient to remove or loosen the device unaided (see Table 1). Therefore, two prevalence estimates were calculated for each observation period. The first estimates definite restraint use, while the second estimates the combination of definite and potential restraint use. Henceforth, the latter estimate will be referred to as definite + potential restraint use. The number of patients observed on the units was used as the denominator for prevalence estimates. Thus, eligible patients who were on the unit census but away from the unit at the time of the observation period were not included in the prevalence estimate. Pearson chi-square statistics were computed to test for significant differences in point prevalence estimates of definite restraint use, and in definite + potential restraint use, among the 5 observation days for Site A and the 3 observation days for Site B. Site B had only 3 days of observation since data for the other 2 days were used to calculate inter-rater agreement. Prevalence estimates obtained for chart audits were calculated using documentation of any physical restraint use during the 12 hours before and 12 hours after the observation period as the numerator. To assess the frequency with which items on the chart audit tool were documented, the cumulative number of observation days across all patients was counted. For example, a patient who was on the ward for all 5 days of observations, provided 5 observation days to this total. The percentage of observation days for which a physical, or physical and chemical restraint was recorded was then estimated. Similarly, frequencies were counted for the number of observation days for which the following were recorded: reasons for restraint use, assessment of physical restraints for proper application, physical restraint ordered by a registered nurse, and patient consent to use of the physical restraint. For measuring inter-rater agreement, authors have noted inaccuracies in interpreting the kappa statistic when tables are asymmetric (Thompson & Walter 1988; Feinstein & Cicchetti 1990; Byrt et al. 1993), which was the case for our data. Hence, we calculated the percentage of overall agreement between the two raters in each site for items on both the observation and chart audit tools. Inter-rater data for the observation tool at Site A were collected on 3 separate days. However, this data was pooled because the subsample of patients on whom the inter-rater observations were made was small (n = 35). For the observation tool at Site B, inter-rater agreement was computed separately for all observations on days 1 and 5. To examine concordance between prevalence estimates of restraint use from the observation tool and the chart data, these two sources of information were matched for each patient. Concordance for this paired data was examined using the phi coefficient. The phi coefficient is a measure of association that is interpreted as a correlation coefficient (Fleiss et al. 2003). The total time taken to complete all observations on the ward was tallied from the start and finish time recorded by the data collectors. The average time was then computed based on the number of patients for whom observations were recorded. The average time to complete chart audits was computed based on the total time required to extract data from each chart and the number of charts reviewed. The 95% confidence intervals on these average times were then calculated. RESULTS Participants Patients observations were collected on nine complex continuing care and rehabilitation units in the two sites. At Site A, 127 of the 128 patients were observed one or more times during the observation period. At Site B, 144 of the 161 patients were observed one or more times during the observation period. At Site A, % of eligible patients were observed each day. At Site B, the proportion observed on a daily basis was lower, ranging from 61.6% to 81.8%. As shown in Table 4, all patients from Site A were aged 65 and older, while 64% from Site B were aged 65 and Worldviews on Evidence-Based Nursing Second Quarter

6 TABLE 4 Patient characteristics by site SITE A SITE B PATIENT CHARACTERISTICS N PERCENTAGE N PERCENTAGE Gender Male Female Missing Total Age (years) Mean (SD) 80.1(7.37) 66.2 (16.4) Range N older. Age data were missing for 23 patients from Site B as this information was extracted during chart audits and, as per study procedures, not all patients charts were audited. At Site A, 68 chart audits were conducted. At Site B, retrospective chart reviews for all 161 patients were attempted. However, only 139 charts could be retrieved. Restraint Prevalence Observation Tool Table 5 shows the estimates for restraint prevalence across the different observation days, by site. For Site A, estimates for all 5 days are reported. An average of 86 patients were observed per day at Site A, and 127 unique patients. For Site B, the estimates for 3 days only have been reported TABLE 5 Physical restraint use and estimates of restraint prevalence by site based on the observation tool TOTAL PATIENTS OBSERVATIONS DAY 1 (n) DAY2(n) DAY3(n) DAY4(n) DAY5(n) (UNIQUE CASES) (n) Site A Time of observation Number of eligible patients Number of patients observed per day Number of patients identified with one or more definite physical restraints a Number of patients identified with one or more definite+potential physical restraints b Prevalence of definite restraints a 22.1% 30.6% 18.8% 32.9% 7.8% χ 2 = 20.5, df =4 Prevalence of definite+potential restraints b 29.1% 31.8% 22.4% 37.6% 30.0% χ 2 = 4.9, df =4 Site B Time of observation Number of eligible patients Number of patients observed per day NA c NA c 144 Number of patients identified with one or more NA c NA c 109 definite physical restraints a Number of patients identified with one or more NA c NA c 121 definite+potential physical restraints b Prevalence of definite restraints a 37.8% 78.5% 40.2% χ 2 = 15.0, df =2 Prevalence of definite+potential restraints b 76.7% 82.6% 60.7% χ 2 = 47.4, df =2 a Definite physical restraint was defined as the presence of a device and patients could not remove or loosen the device. b Definite + potential physical restraint was defined including cases of definite restraint plus cases where a device was present but the rater could not determine whether the resident could easily remove or loosen the device and/or could not determine whether the device restricted freedom of movement. c Observations for Days 1 and 5 at Site B were used to obtain inter-rater agreement and are excluded from the analysis of prevalence estimates. p < Second Quarter 2006 Worldviews on Evidence-Based Nursing

7 TABLE 6 Types of physical restraint devices by site based on the observation tool OBSERVATIONS DAY 1 (n) DAY2(n) DAY3(n) DAY4(n) DAY5(n) Site A (restraint type a ) Lap belts Trunk restraint Limb restraint Full bed rails Other type of bed rails Chair that prevents rising Elevated wheelchair foot pedals Other Site B (restraint type a ) Lap belts Trunk restraint Limb restraint Full bed rails Other type of bed rails Chair that prevents rising Elevated wheelchair foot pedals Other a The types of physical restraint devices reported here refer to definite physical restraints (defined as the presence of a device and patients could not remove or loosen the device). as all observations for the first and last days were used for evaluating inter-rater agreement. An average of 108 patients were observed per day across the 3 observation days, and 144 unique patients. Pearson chi-square tests showed that point prevalence estimates of definite restraints varied significantly among the 5 observation days in Site A, and point prevalence estimates of both definite and definite + potential restraint use varied significantly among the 3 observation days in Site B. The types of physical restraint devices observed are shown in Table 6. Full bed rails were the most frequently used restraint device, followed by other types of bed rails and lap belts. Reasons for Restraint Use As shown in Table 7 data for a total of 616 patient observation days across the two sites were collected. The documented use of physical restraints was 40.7% in site A and 88.2% in site B. For nearly half of these patients, a combination of physical and chemical restraints had been used. The most frequently documented reasons for physical restraint use at both sites were patients risk for falling and impaired mobility. In nearly 90% of cases, the chart record showed that the physical restraint had been assessed for proper application. The sites differed substantially concerning whether a registered nurse had ordered the restraint (65.9% of cases in site A, 0% of cases in site B). The patient s consent for the use of the physical restraint was rarely indicated, either because consent had not been obtained (the majority of cases in both sites) or because data regarding patient consent was missing from the chart (one third of cases in Site A). Reliability Table 8 shows the inter-rater agreement for each item on the observation tool. Inter-rater agreement varied for different types of devices. It ranged from 60% to 100% in site A and from 97% to 100% for day 5 in site B. In site A, interrater agreement was lowest for all types of bed rails. In Site B, a similar pattern was seen but additional sources of discordance were noted for chairs that prevent rising and use of elevated wheelchair foot pedals. In Site B, an improvement in levels of inter-rater agreement was noted between day 1 and day 5 observations. These improvements were most notable for the categories other potential restraint and patient able to release/loosen restraint. Inter-rater agreement for items on the chart audit was conducted on 25 charts from Site B. With the exception of two items on reason for restraint use (cognitively impaired patient and patient at risk of falling), there was % agreement on all items. The documented use of restraints was compared for the observation tool and the chart audit for matched pairs of data (see Table 9). The number of patients for whom both chart audit and observation data were collected ranged from 37 to 47 per day in site A, and from 99 to 109 per day in site B. Using definite restraint use on the observation tool as the basis for comparison, the chart audit data indicated higher rates of restraint use for all observation periods in Worldviews on Evidence-Based Nursing Second Quarter

8 TABLE 7 Frequencies of items in the chart audit tool on restraint prevalence across observation days SITE A SITE B TOTAL PERCENT OF TOTAL TOTAL PERCENT OF TOTAL ITEMS NUMBER OBSERVATION DAYS (%) NUMBER OBSERVATION DAYS (%) Cumulative total number of observation days a Patient days with physical restraint Patient days with physical and chemical restraint Total with physical restraints Reason for physical restraint use: b No reason given Cognitively impaired patient Patient at risk of falling Patient who wanders Patient who tampers with medical devices Agitated or violent patient Patient with impaired mobility Patient undergoing medical treatments Patient is Suicidal Other Assess restraint for proper application c Yes No Missing Physical restraint ordered by registered nurse c Yes No Missing Patient consented to the use of physical restraint c Yes No Missing a The cumulative total number of observation days was obtained by adding number of observations for Day 1 through Day 5 for Site A, and Days 2, 3 and 4 for Site B. b The frequencies for this variable are confined to those patients for whom documentation for physical restraint only was present. More than one reason was recorded for some patients. c The frequencies for these variables are confined to patients for whom any physical restraint was documented (including patients with both chemical and physical restraints). both sites. The difference in estimated prevalence rates between chart audit and observation data ranged from 2.4% to 29.7% at Site A, and from 14.7% to 52.4% at Site B. Phi coefficients, to assess concordance between chart and observational data for definite restraint use, ranged from.45 to.95 for Site A and from.25 to.42 for Site B. All phi coefficients were statistically significant with the exception of agreement between definite + potential physical restraint use from observations and chart audits for site A on day 5. Time Required to Complete Tools Observations took an average of 2.6 minutes per patient in Site A and 0.6 minutes per patient in Site B. Data collectors spent more time conducting the observations when patients had to be located outside of their room. At Site B, the patients demographic data were entered on the observation tool before data collectors observed the patients. This step saved data collectors time. At Site review of patients charts took an average of 16 minutes (95% CI: 3 29). Less time was required for chart reviews at Site B, an average of 4.3 minutes (95% CI: 1.6 7). At Site B, the research coordinator had completed information available from the observation sheet (e.g., patient ID, date, and time of observation) in advance of the chart audit. This might have reduced time required to complete the chart audits. DISCUSSION The use of restraints is complex and multifaceted. From this study, it is evident that using only one approach 80 Second Quarter 2006 Worldviews on Evidence-Based Nursing

9 TABLE 8 Inter-rater agreement for the restraint prevalence observation tool SITE A, 3 DAYS OF OBSERVATIONS SITE B, DAY 1 SITE B, DAY 5 AGREEMENT (n = 35) AGREEMENT (n = 169) AGREEMENT (n = 163) Item n a % n a % n a % Potential physical restraint (unscored) b Lap belts Trunk restraint Limb restraint Full bed rails Other types of bed rails Chair that prevents rising Elevated wheelchair foot pedals Other Patient able to release/loosen restraint c (n = 14) (n = 87) (n = 72) Proper application of restraint c (n = 14) (n = 90) (n = 72) a n = number of cases where ratings were concordant. Numbers in columns differ because of missing cases. b This variable is the unscored physical restraint, i.e., based on the presence of a device and not considering whether the patient can remove or loosen the device. c Number of cases in brackets refers to number of cases with potential restraint devices. This excludes cases where both raters indicated no potential restraint. (observation tools or chart audits) provides an insufficient evaluation of restraint utilization within an organization. Complete concordance between the chart audit and the observation tool data was not expected because the chart audit provided restraint use data over a 24-hour period, whereas the observation tool provided restraint use data for one time point during this same 24-hour period. What the observation tool uniquely captured, however, was whether the patient could release a restraint. When the observation tool was used at different times of the day, it also indicated variations in patterns of restraint use that might have been related to factors such as types of patient activity and staffing patterns. On the other hand, the chart audit tool allowed capturing information about documented patient consent, who ordered the restraint, reasons for restraint use, and whether a chemical restraint was also being used. Thus, the observation and chart audit tools are complementary and should be used together. This study was undertaken in settings where alternatives to restraint use may be particularly difficult to identify because of the characteristics of the patient population. In both rehabilitation and complex continuing care settings, many patients will have mobility and cognition limitations. Thus, determining whether a device meets all of the criteria for the definition of a restraint is particularly challenging. Documentation of a restraint in a chart assumes that all three elements for the classification of physical restraint have been verified and that a restraint device is in place: manual method or physical or mechanical device adjacent to the body; cannot be removed or loosened by the patient; and restricts freedom of movement. However, unless nurses consistently apply the restraint use algorithm as a basis for documentation, chart audits may overestimate restraint use. In both study settings, several care strategies were listed on the charts that matched those listed on the audit form as alternatives to restraints. However, in the absence of a specific progress note to indicate that these care strategies were being used as alternatives to restraints, they could not be coded as such. At both sites significant gaps in documentation were present in relation to required elements of charting for individuals in restraints. This might indicate lack of clarity among staff regarding the core elements of what constitutes a restraint. Also, these gaps in charting likely led to an overestimation of the prevalence of restraint use from the chart audits. The observation tool included two of the three required components based on MDS to determine whether a restraint device was used and whether the patient could remove the device. However, to identify whether the patient could remove the device required that the resident demonstrate or describe their capacity to do so. This assumes that the patient comprehends the question appropriately or accurately. An alternative would be to obtain the information from a care provider rather than identifying a patient s ability to remove or loosen the device. The third element of restraint use concerning whether the device restricted freedom was not adequately answered by either the audit or observation tool. This gap introduces the potential for misclassification. For example, a lap Worldviews on Evidence-Based Nursing Second Quarter

10 TABLE 9 Agreement on restraint prevalence estimates using two methods DAY1(n) DAY2(n) DAY3(n) DAY4(n) DAY5(n) Site A Number of patients with both chart audit and observation data Number of patients identified with definite physical restraints based on observation 2. Number of patients identified with definite + potential restraints based on observation 3. Number of patients identified with physical restraints based on chart audit information Concordant pairs of definite physical restraint in observation and chart audit (#1 and #3) Phi coefficient Concordant pairs of definite + potential physical restraint in observation and chart audit (#2 and #3) Phi coefficient Site B Number of patients with both chart audit and observation data Number of patients identified with definite physical restraints based on observation 2. Number of patients identified with definite + probable restraints based on observation 3. Number of patients identified with physical restraints based on chart audit information Concordant pairs of definite physical restraint in observation and chart audit (#1 and #3) Phi coefficient Concordant pairs of definite + potential physical restraint in observation and chart audit (#2 and #3) Phi coefficient p <.05, p <.01, p <.001. = not significant. belt that protects a patient from falling out of a wheelchair when that patient does not have the physical capacity to walk could be viewed as facilitating movement about the facility by wheelchair rather than restricting the individual. While the observation tool indicated excellent interobserver agreement on most items, more discrepancies existed between raters for patients abilities to remove or loosen the device. In part, this might have been because of the fact that observers were asked not to wake patients to ask whether they could loosen or remove a restraint. Another aspect of the MDS definition of restraint that was not included in either the audit or the observation tool was the patient s physical capacity to move. If the patient does not have physical capacity to move, bedrails are not considered a restraint. Further, if a patient has no voluntary or involuntary movement, a geri-chair is not considered a restraint. Clearly defining and specifying criteria for determining the existence of restraints as it relates to specific clinical populations could increase the instruments utility in rehabilitation settings. Based on these findings, we suggest that the following items be added to the observation tool: (a) the patient s waking status, (b) whether the device promotes or restricts mobility, (c) patient s capacity to move independently and (d) patient s location at the time of the assessment (in bed, chair, etc.). The utility of the observation tool was particularly evident. Observers were able to assess most of the patients on the unit during each observation period. The average time required for documenting an observation ranged from 0.6 to 2.6 minutes at the two sites. The tool indicated important fluctuations in the use of restraints at different times of the day. However, more observations would be needed to describe variations in restraint use. Furthermore, the timing of observations would need to be purposefully selected to reflect differing patterns of staffing, differing staffing ratios, and different ward schedules for patient care activities. The chart audit took more time to complete than did the observational tool. However, each chart indicated information about the five observation periods. Finding the progress notes for the date the restraint was initiated, 82 Second Quarter 2006 Worldviews on Evidence-Based Nursing

11 especially in the absence of a physician s order, was particularly difficult. Chart abstraction would be facilitated by use of a restraint flow sheet. These flow sheets should contain information such as the type of device, date and time of application, and alternatives attempted and in place. Use of the chart audit tool would need to be coordinated with the use of organizational restraint flow sheets. Despite the time involved in observing patients and extracting data from charts, the information obtained is highly relevant to managers who are implementing restraint minimization guidelines. Use of this data can inform answers to time-sensitive administrative questions such as the following: Do patterns of physical restraint use vary over the course of the day or the week? Are these patterns related to staffing ratios or related to changes in patients activities? Who is providing orders for the use of physical restraints? Is patient consent for restraint use being documented? Limitations Data were collected before the initiation of clinical practice guidelines, which include evidence-based recommendations to reduce restraint use. The charts audited, likely reflect real world examples of restraint documentation patterns. In addition to reducing the use of restraints, managers should be concerned with appropriate documentation of restraint use. While the preintervention timing of this study might have reduced inter-rater agreement between chart audit and observation data, it shows the importance of using both data sources for decision making by managers. Validity of the restraint use tools remains elusive. A gold standard was not used to determine, for example, whether information from the charts reflected what had actually transpired in interactions between patients and nurses with respect to obtaining consent. IMPLICATIONS FOR PRACTICE Using the observation tool required, several hours of training on restraints and how to conduct observations. The tool covers a diverse range of physical restraints. Institution-specific estimates of prevalence can be derived easily by determining what types of restraint use are of interest and using the appropriate numerator data to reflect that. The main advantage of the chart audit data is the potential to show details regarding who ordered the restraint and the reason for physical restraint use. Bedrails were the most prevalent type of restraint used, which is consistent with reports in several recent studies on utilization of restraints (Capezuti et al. 2002; Hamers et al. 2004). The differences in prevalence estimates between definite and definite + potential estimates indicate that it might be most difficult to assess whether patients can release bedrails. While the chart audits show a 24-hour record on the use of physical restraints, they do not indicate insights into the patterns of physical restraint use over the day. Physical restraint use was highest when bedrails were being used, suggesting that this is a particular facet of nursing care that must be addressed in continuing education on restraint use. CONCLUSIONS The observation tool can be used efficiently. It provides just in time data required for critical management decisions. The differing patterns of restraint use across days of the week and hours of the day indicate that the tool should be used at different time points. Training raters and assessing inter-rater reliability is an important part of the process required to use the tool. Chart audits were time consuming. However, data from chart audits provide insights regarding which disciplines are ordering the restraints and reasons for physical restraint use. A combination of observation tools and chart audits are needed to know the complexity of restraint use in health service organizations. Acknowledgments The authors thank Dianne Rossy and Marlene Mackey for permission to adapt their restraint prevalence tool; project teams at participating sites (June Duesburry-Porter, Abel Cheng, Tim Pauley, Barb Cassel); other members of our evaluation team (Pat Griffin, Diane Buchanan, Denyse Pharand, Lucie Brosseau); other coinvestigators (Jennifer Skelly, Tazim Virani); project staff (Viren Bharti, Valerie Cronin, Andrea Perrier, Elana Ptack) and a graduate student (Sylvia Ralphs-Thibodeau). Funding for this project was provided by the Ontario Ministry of Health and Long-term Care. Dr. Edwards holds a Nursing Chair funded by the Canadian Health Services Research Foundation, the Canadian Institutes of Health Research and the Ontario Ministry of Health and Long-term Care. Dr. Davies holds a Career Scientist award funded by the Ontario Ministry of Health and Long-term Care. Dr. Ploeg holds a career investigator award from the Canadian Institutes of Health Research and St. Joseph Healthcare, Hamilton. References Brush B.L. & Capezutti E. (2001). Historical analysis of siderail use in American hospitals. Journal of Nursing Scholarship, 33(4), Worldviews on Evidence-Based Nursing Second Quarter

12 Byrt T., Bishop J. & Carlin J.B. (1993). Bias, prevalence and kappa. Journal of Clinical Epidemiology, 46(5), Canadian Institutes of Health Research, Natural Sciences and Engineering Research Council of Canada, Social Sciences and Humanities Research Council of Canada. (1998, with 2002 amendments). Tri-council policy statement: Ethical conduct for research involving humans. Canadian Institute for Health Information (2002). Minimum data set (MDS) 2.0 Canadian version. Ottawa: Canadian Institute for Health Information. Capezuti E., Maislin G., Strumpf N., & Evans L.K. (2002). Side rail use and bed-related fall outcomes among nursing home residents. Journal of the American Geriatric Society, 50(1), Center for Medicare & Medicaid Services (2005). Revised long-term care facility resident assessment instrument user s manual version 2.0. Baltimore, MD: U.S. Department of Health and Human Services. College of Nurses of Ontario (2004). Practice standard: Restraints. Toronto, Ontario, Canada. Dunn K.S. (2001). The effect of physical restraints on fall rates in older adults who are institutionalized. Journal of Gerontological Nursing, 27(10), Evans D., Strumpf N.E., Allen-Taylor S.L., Capeuzuti E., Maislin G. & Jacobsen B. (1997) A clinical trial to reduce restraints in nursing homes. Journal of the American Geriatric Society, 45(6), Evans D., Wood J., Lambert L. & Fitzgerald M. (2002a). Physical Restraint in Acute and Residential Care: A Systematic Review. The Joanna Briggs Institute, Adelaide, Australia[on-line] Evans D., Wood J. & Lambert L. (2002b) A review of physical restraint minimization in the acute and residential care settings. Journal of Advanced Nursing, 40(6), Evans D., Wood J. & Lambert L. (2003). Patient injury and physical restraint devices: A systemic review. Journal of Advanced Nursing, 41(3), Feinstein A.R. & Cicchetti D.V. (1990). High agreement but low kappa: I. The problems of two paradoxes. Journal of Clinical Epidemiology, 43(6), Fleiss J.L., Levin B. & Paik M.C. (2003). Statistical methods for rates and proportions, 3rd ed. New York: John Wiley & Sons. Hamers J.P.H., Gulpers M.J.M. & Stirk W. (2004). Use of physical restraints with cognitively impaired nursing home residents. Journal of Advanced Nursing, 45(3), Hirdes J., Zimmerman D., Hallman K. & Soucie P. (1998). Use of the MDS quality indicators to assess quality of care in institutional settings. Canadian Journal of Quality in Health Care, 14(2), Hoelzel S.L. (2003). Reducing use of restraints on a rehabilitation unit. Rehabilitation Nursing, 28(3), Joanna Briggs Institute (2002). Physical restraint - Part 2: Minimization in acute and residential care facilities.best Practice, 6(4), 1 6. Blackwell Publishing Asia, Australia. Kirkevold O. & Engedal K. (2004). A study into the use of restraint in nursing homes in Norway. British Journal of Nursing, 13(15), Levine J.M., Marchello V. & Totolos E. (1995). Progress toward a restraint-free environment in a large academic nursing facility. Journal of the American Geriatric Society, 43, Marcy-Edwards D. (2005) Bedrails: Is there an upside? Canadian Nurse, 101(1), McMahon M. & Fisher L. (2003). Achieve emergency department restraint reduction. Dimensions of Critical Care Nursing, 22(5), Mott S., Poole J. & Kendrick M. (2005). Physical and chemical restraints in acute care: Their potential impact on the rehabilitation of older people. International Journal of Nursing Practice, 11, Registered Nurses Association of Ontario (2004). Caregiving strategies for older adults with delirium, dementia and depression among older adults. Toronto, Canada: Registered Nurses Association of Ontario. [Available online] guidelines/bpg Guide C4 caregiving elders ddd.asp. Rossy D. & Mackey M. (2003). The 3 A s to alternatives & the least restraint, last report program at the Ottawa hospital: A self-directed resource guide. Ottawa, ON: The Ottawa Hospital. Shojania K.G., Duncan B.W., McDonald K.M., Wachter R.M. & Markowitz A.J. eds. (2001). Making health care safer: A critical analysis of patient safety practices. Evidence/Report/Technology Assessment No. 43 [Prepared by the University of California at San Francisco- Stanford Evidence-based Practice Center under Contract No ), AHRQ Publication No. 01- E058, Rockville, MD: Agency for Healthcare Research and Quality. Simmons S.F., Schnelle J.F., MacRae P.G. & Ouslander J.G. (1995). Wheelchairs as mobility restraints: Predictors of wheelchair activity in nonambulatory nursing home residents. Journal of the American Geriatric Society, 43, Sorrentino A. (2004). Chemical restraints for the agitated, 84 Second Quarter 2006 Worldviews on Evidence-Based Nursing