To reduce the number of falls which result in serious injury and ensure effective treatment and rehabilitation for those who have fallen.
Author: Geoff Cooper is senior clinical risk manager, Winchester and Eastleigh Healthcare NHS Trust, Winchester, Hampshire
The National Service Framework for Older People (Department of Health, 2001a) has the stated aim: ‘To reduce the number of falls which result in serious injury and ensure effective treatment and rehabilitation for those who have fallen.’
However, when clinical risk management unit staff came to review our data on patient falls we found that adverse incidents involving patient falls were being reported either as accidents or as clinical incidents, dependent on how they were perceived by the staff involved. Furthermore differences in the way the incident reporting systems were managed (now resolved) meant that local clinical risk co-ordinators were not informed of all patient injuries.
An additional concern was that there was often no risk assessment, or care plan, documented in the clinical record.
Determining a strategy
We knew from the literature that falls among older inpatients are common and result in morbidity, loss of confidence, and higher health-care costs (Oliver et al, 1997). Falls result in serious injuries and are the leading cause of death from injury among people over 75 in the UK (Oakley et al, 1996).
We wanted to comply with the National Patient Safety Agency (NPSA) requirements by learning the appropriate lessons from adverse patient incidents and, where appropriate, changing practice in order to improve the safety and quality of care for our patients (DoH/NPSA, 2001b).
Uden et al (1999) suggest that several measures should be implemented in order to care properly for the patient at risk of falling. These should include the development of a strategy for fall prevention, the identification of patients at risk and their specific needs and the implementation of an appropriate preventive strategy.
Various clinical characteristics have been shown to be associated with an increased incidence of falls. A strategy that has proved to be successful in the prevention of pressure sores is to select patients who are at high risk and target preventive strategies (Waterlow, 1985). We determined to use a similar approach to first identify patients at risk of falling and, second, to develop preventive strategies based on patients’ individual needs.
Another element of this strategy was to ensure that staff documented their risk assessments and care plans. In 1999 Uden et al also reported problems in relation to documenting falls-risk assessments despite the fact that Swedish law made care planning obligatory. We felt that clear local guidance was essential to remind staff of Nursing and Midwifery Council requirements to provide clear evidence of the care planned, the decisions made, the care delivered and the information shared (NMC, 2002).
Finding an assessment tool
Like Moore et al (1996) we felt that the identification of the patient at risk is the first step in falls prevention. Risk assessment assists staff in targeting patients at risk and in choosing appropriate individualised preventive strategies. Our initial objective therefore was to review the literature and identify an appropriate risk assessment tool that had good predictive value and which was validated for use in an acute trust.
Moore et al (1996) reported that, although more than a dozen tools have been published, few have undergone extensive testing of reliability and validity. Furthermore the processes for establishing the weighting of items and a cut-off point for high-risk status are frequently not described in the literature.
Oliver et al (1997) did not comment on the nature of available risk assessment tools but instead reported that they had found over 400 clinical characteristics associated with an increased incidence of falls. From this systematic review they collected 21 separate pieces of information and identified five risk factors ‘significantly and independently’ associated with falling which they incorporated into their risk assessment tool STRATIFY (St Thomas’s Risk Assessment Tool In Falling elderlY inpatients) (Oliver et al, 1997) (Box 1).
In 1999 Conley et al selected four articles that had addressed risk assessment scales in acute care patient populations. However, they then discarded each of these for various reasons, noting that, at the time of their study, there was not a published falls-risk assessment scale that reflected the characteristics of the complex acutely ill patient. They subsequently decided to develop a scale themselves and identified 10 statistically significant indicators in the literature, which they reduced to six variables in their final published scale (Box 2).
Unfortunately, as far as we were concerned, only two risk factors, a history of falls and agitation, were common to both risk assessment tools.
At this stage we realised that our objective of finding a predictive falls-risk assessment tool validated for use in an acute trust was optimistic. We did not feel we had the resources to develop and validate our own risk assessment tool as Conley et al (1999) has done. Stetler et al (1999) also went on to develop a resource-intensive falls prevention programme after having noted that the literature did not contain a valid patient fall tool. Although we noted that Simmons (2001) chose Oliver et al’s STRATIFY, primarily on the grounds that it scored high for both sensitivity and specificity, we felt we needed to further review the validity issues raised by Oliver et al, Conley et al and Stetler et al before reaching a decision.
Validity and reliability in risk assessment tools
There is much published literature on the difficulties of developing valid risk assessment scales for pressure sores, which seems highly relevant to the development of a falls scale.
In her published work on ‘risk calculators in pressure sore prevention’ Edwards (1994) notes that, according to the National Pressure Ulcer Advisory Panel (NPUAP, 1989), a good instrument should have good predictive value, high sensitivity (ability to predict risk), specificity (ability to predict the absence of risk) and be easy to use. She cites Larson (1986) when stating that sensitivity and specificity, together with the predictive value of positive and negative tests, are held the measures of the validity of the tool.
However, Waterlow (1995) makes the point that the effect of preventive interventions on the performance of risk calculators raises questions as to whether sensitivity, specificity and predictive values are indeed valid measures of the performance of a risk calculator.
Waterlow argues that if, for example, only two patients from an at-risk group of 40 develop pressure sores then this is not, as had been suggested, a disastrous failure of the risk calculator but rather demonstrates the mediating effect of preventive interventions and intercurrent illnesses. Moore et al (1996) suggested that it is difficult to separate falls prediction from intervention and nurses who identify patients at risk will intervene in some way, thus lowering the apparent predictive accuracy of the scale.
Oliver et al (1997) observed that nurses aware of the predictive power of the score might have altered their care of ‘high risk’ patients (something they were asked not to do) thus preventing some falls (the Hawthorne effect).
Waterlow expresses the view that the performance of the scale can be assessed only in this way:
- If there is no change in the patient’s condition during the period of the survey
- If no intervention at all is allowed, even for patients in a very high risk group (which she felt would be unethical) and
- If compensating negative scores are attributed to all interventions, and changes in patients’ condition are in some way catered for.
In conclusion, Waterlow argues that risk calculators are designed to warn of risk so that intelligent judgements can be made as to what steps need to be taken in relation to deciding on appropriate nursing interventions. In her view they are not, and do not profess to be, scientific instruments for predicting the inevitable development of sores, yet feels this is how academics judge their performance.
Oppenheim (1992) defines reliability ‘as meaning consistency’. Edwards (1994) cites Polit and Hungler (1991) who stated that the reliability of an instrument is the property of the instrument when administered to a certain sample under certain conditions. The increasing sophistication of risk calculators in attempting to capture the determinants of pressure sore formation has been an attempt to increase the validity of the instruments. However Polit and Hungler (1991) noted: ‘a measuring device that is unreliable cannot possibly be valid’.
When discussing the problem of reliability Conley et al (1999) cites Schroeder (1995) who found that benchmarking, or comparing, fall rates among hospitals and other nursing facilities is difficult and often inaccurate due to the lack of a consistent, standardised definition of a fall. This point is illustrated further by Conley et al’s (1999) own work where they defined a fall as ‘an unintentional event that resulted in a patient coming to rest on the floor’. Oliver et al (1997) had earlier defined a fall as ‘an incident in which a patient suddenly and involuntarily came to rest upon the ground or surface lower than original station’.
Conley et al (1999) also raises the issue that consistency throughout a patient episode is problematic, noting that the ability of nurses to identify some risk factors (for example ‘lack of safety awareness’) may be limited at admission and improve as the nurse becomes familiar with the patient’s behaviours.
Stetler et al (1999) provided evidence of lack of consistency, reporting nurses’ lack of knowledge regarding the operational definition of some risk factors and noting that a nurse might list diuretics as an ‘other’ factor when such a risk should have been recognised under the standardised criteria of ‘altered elimination’.
Reviewing the evidence
Having identified a number of published works on falls-risk assessments we wanted to critically appraise them for validity and predictability. Fortunately for us almost all of the authors addressed the issue of validity in their reflective discussions and many were very frank about their perceptions of the limitations of their own works.
Moore et al (1996) stated that the relatively low predictive values of their results suggested that neither the risk assessment tool nor the clinical judgement of experienced nurses alone adequately predicted the risk of falling. She argued that there is no method that will predict all falls, citing Morse et al (1987) who found that around 20% of falls are accidents or unanticipated physiological falls.
Oliver et al (1997) reported sensitivity scores of 93% for their local cohort and 92% for their remote cohort and specificity scores of 92% and 68% respectively. They described this as ‘clinically useful’ but identified some relevant explanations for the reduced power of the risk assessment tool in its remote validation study, suggesting that the predictive power of risk factors is likely to be specific to one unit or patient group.
Conley et al (1999) provides some detail on the reasons why they discarded four different scales, noting concerns about risk factors used, lack of clarity of definitions, length of the assessments and lack of validation in the acute care population. Conley et al (1999) addressed the issue of validity in the development and analysis of their own scale and reported that they felt the sensitivity (71%) and specificity (59%) of the scale were ‘acceptable’, considering it was completed only on admission. In their summary they observed that the Conley scale is ‘a quick and easy scale for use in a busy hospital setting’.
Stetler et al (1999) found that 16% of patients who fell had been assessed as low risk for falls and reported that the falls-risk assessment tool used displayed ‘reasonable’ validity, observing that one potential issue may have been that the staff’s initial assessment might not have been accurate.
In addition to their general observations on validity some of the authors also commented on their tool’s validity in different clinical settings mostly, it has to be said, to urge caution with regard to the likely generalisability of their assessment tool.
Stetler et al (1999) reported that the highest risk factor varied across units, for example, a history of falls was most recurrent on orthopaedics and neuroscience units, whereas a patient’s poor judgement was the most common risk factor on the high-risk obstetric unit.
Oliver et al (1997) suggest that a prospective validation may be necessary in any hospital unit before use of their risk assessment tool since case mix, ward design, and nursing philosophy and skills vary widely.
Conley et al (1999) observed that the psychometric properties of their scale needed further evaluation as its use is expanded to other patient populations and when it is completed more frequently than just on admission.
Ease of use
While validity and predictability are key elements in the design of a risk assessment so too is the ability to use it in the clinical area. Conley et al (1999) rejected a very detailed tool on the grounds that it was too lengthy. Moore et al (1996) suggested that nurses frequently see the tools as being just ‘another piece of paper’, offer little advantage over their own clinical judgement and, understandably, are concerned about having less time available to deliver patient care.
In selecting the factors to construct the risk assessment tool, Oliver et al (1997) used the factors that were readily accessible by ward nurses based on a day-to-day observation of patients and that could be performed shortly after admission to hospital. This conferred the advantage of generating a pragmatic risk assessment tool, taking about one minute per patient per week and requiring no formal measurements, additional training or equipment.
Conley et al (1999) identified that, with shortened hospital stays, one of the criteria for an assessment tool was that it should be easy to use. In reviewing the use of another assessment tool within the same organisation, to predict skin breakdown, they found that the scale was completed on admission but was rarely used during the remainder of the patient stay. They therefore determined that a fall-risk assessment scale that was predictive on admission would be the most feasible solution.
Simmons (2001) saw the user friendliness of a tool as an important factor in successful implementation of such an assessment and referred to the simple design and minimal time for completion when selecting Oliver et al’s STRATIFY.
Determining an evidence-based strategy for risk assessment and falls prevention is also not helped by the current level of scientific evidence with regard to the effectiveness of preventive interventions. Oliver et al (1997) and Stetler et al (1999) raise questions about the evidence base for the falls-prevention interventions. Oliver et al (1997) argues that further study is needed to determine whether the falls of inpatients identified as high risk can be prevented by a targeted intervention. Stetler et al (1999) comments that ‘initial interventions gleaned from the falls literature were not well substantiated by science’.
Clinical risk management unit staff had defined three strategic objectives:
- Development of a strategy for fall prevention
- The identification of patient risk and needs
- The implementation of appropriate preventive measures.
Our new policy document ‘Guidelines to identify patients at risk of falling’ identifies a strategy for falls prevention based on a clear, evidence-based, context. This is that falls are unwelcome but are often the inevitable consequence of encouraging patients to regain their mobility and reduce their dependence after an acute illness (Oliver et al, 1997). The policy further informs staff that, while they may see an increase in the numbers of reported falls, the incidence of serious injury will reduce if the suggested preventive strategies are followed (Stetler et al, 1999). All of this is firmly within the context of the National Service Framework for Older People (Department of Health, 2001a), which aims to ‘reduce the number of falls which result in serious injury’.
The policy also makes explicit statements about the various responsibilities for informing staff about the falls-risk management protocol, for providing appropriate training, for carrying out the risk assessments, for implementing preventive strategies and for documenting the process.
Identifying patient risk and needs
We decided to use Conley’s risk assessment tool on the basis of its development with complex acutely ill patients, and its ease of use. However, we accepted the argument that risk assessment tools are not scientific instruments for predicting the inevitable development of the risk in question. Consequently we decided not to score the risk factors but instead stated that ‘the risk of a further fall increases with the number of positive responses in the assessment’.
The policy makes clear that, as a previous history of falling is a predictor of an increased risk of further falls, an initial risk assessment should be carried out on every patient on admission.
We also used other evidence, based on input from colleagues in our multidisciplinary falls project, that:
- In some instances it may be appropriate to seek the advice from physiotherapy colleagues for assessment of gait patterns
- All older patients should be checked for signs of postural hypotension as this will increase the risk of falling (Cwikel and Fried, 1992)
- The assessment tool should be used in conjunction with clinical judgement of each patient as required by the UKCC’s Guidelines for Professional Practice (1996) (now the Nursing and Midwifery Council)
- Each patient should be assessed as an individual.
Appropriate preventive measures
Our policy statement takes account of the uncertainty in the published literature about the effectiveness of preventive interventions yet seeks to provide nurses with a useful falls intervention resource. It provides a list of strategies that have been shown to minimise the risk of serious injury through falling from Oliver et al (1997) and Uden et al (1999) and groups these interventions under four main headings:
- Information and education
- Promotion of patient participation
- Structuring of the environment
- General care.
The policy makes it clear that these strategies are a resource and should be used in conjunction with clinical judgement and routine physical assessment.
Our review of the literature identified that many authors have attempted to produce a validated falls-risk assessment tool but the result is a proliferation of different scales based on varying criteria and different definitions of a fall.
In addition, these scales have been developed in different clinical areas and the authors themselves have suggested that further validation may be necessary in other patient populations and have noted their predictive accuracy as ‘relatively low’, ‘acceptable’ or at best ‘clinically useful’.
There are also good arguments in the literature to support the view that risk calculators are not scientific instruments for predicting the inevitable development of risk and that sensitivity and predictability may not be the best criteria to use to assess them. Indeed, taking the argument to its logical conclusion, a risk calculator which in use proved to be 100% sensitive would have to reflect the fact that any preventive interventions must have been 100% ineffective.
A further limitation to our aim to identify appropriate interventions for identified risks was the recorded concerns of several authors in relation to the scientific basis of the evidence base for falls-prevention interventions.
However, we were impressed by the simple, pragmatic approach to risk assessment taken by several authors. This was best summed up in some earlier work by the National Pressure Ulcer Advisory Panel (NPUAP, 1989) that ‘the ultimate choice of a risk calculator which has proved to be valid and reliable should probably be made on the basis of its ‘user friendliness’.’ We felt that simplicity of use was important, not only to ensure it would be used in the wards but also to ensure that it did not require a significant ongoing training input.
In the light of this work we felt that we had not found in the literature the validated, predictive risk assessment tool we were looking for. However, we felt we had been able to develop an evidence-based approach to falls-risk management, a simple clinically useful risk assessment tool and a useful resource for staff which identified strategies that would minimise the risk of serious injury through falling.
Uden et al (1999) argue that their study established that nurses could operate one instrument in everyday nursing care practice and justified the systematic recording of admission of an initial risk assessment.
The clinical risk management team hope that this will be the experience in our trust but, in any case, our incident-reporting systems will enable us to monitor the number of patient falls and the subsequent severity of injuries. The resultant data collected as a result of this systematic approach will provide the opportunity for professional, managerial and clinical audit functions to assess the efficiency and effectiveness of the new strategy over time.
Conley, D., Schultz, A., Selvin, R. (1999) The challenge of predicting patients at risk of falling: development of the Conley scale. MEDSURG Nursing 8: 6, 348-354.
Cwikel, J., Fried, A.V. (1992) The social epidemiology of falls among community-dwelling elderly: guidelines for prevention. Disability and Rehabilitation 14: 3, 113-121.
Department of Health. (2001a) National Service Framework for Older People. London: The Stationery Office.
Department of Health/ National Patient Safety Agency. (2001b) Doing Less Harm: Improving the quality of care through reporting, analysing and learning from adverse incidents involving NHS patients - key requirements for health-care providers. London: DoH/NPSA.
Edwards, M. (1994) The rationale for the use of risk calculators in pressure sore prevention, and the evidence of the reliability and validity of published scales. Journal of Advanced Nursing 20: 288-296.
Larson, E. (1986) Evaluating the validity of screening tests. Nursing Research 35: 3, 186-188.
Moore, T., Martin, J., Stonehouse, J. (1996) Predicting falls: risk assessment tool versus clinical judgement. Perspectives 20: 1, 8-11.
Morse, J., Tylko, S., Dixon, H. (1987) Characteristics of the fall-prone patient. Gerontologist 27: 516-522.
National Pressure Ulcer Advisory Panel. (1989) Pressure ulcers: incidence, economics and risk assessment. Care: Science and Practice 7: 4, 96-99.
Nursing and Midwifery Council. (2002) The Code of Professional Conduct. London: NMC
Oakley, A,, Dawson, M., Holland, J. et al. (1996) Preventing falls and subsequent injury in older people. Quality in Health Care 5: 243-249.
Oliver, D., Britton, M., Martin, F., Hopper, A. (1997) Development and evaluation of evidence-based risk assessment tool (STRATIFY) to predict which elderly inpatients will fall: case-control and cohort studies. British Medical Journal 315: 1049-1053.
Oppenheim, A.N. (1992) Questionnaire Design, Interviewing and Attitude Measurement (2nd edn). London: Pinter.
Polit, D., Hungler, D. (1991) Nursing Research: Principles and methods (4th edn). Philadelphia, Pa: Lippincott.
Schroeder, P. (1995) Benchmarking patient falls. Nursing Quality Connection 4: 5, 5.
Simmons, M. (2001) Implementation of a patient falls risk-management strategy. Professional Nurse 17: 3, 168-171.
Stetler, C., Corrigan, C., Sander-Buscemi, K., Burns, M. (1999)Integration of evidence into practice and the change process: fall prevention program as a model. Outcomes Management for Nursing Practice 3: 3, 102-111.
Uden, G., Ehnfors, M., Sjostrom, K. (1999) Use of initial risk assessment and recording as the main nursing intervention in identifying risk of falls. Journal of Advanced Nursing 29: 1, 145-152.
UKCC. (1996) Guidelines for Professional Practice. London: UKCC (now Nursing and Midwifery Council).
Waterlow, J.A. (1985) Pressure sores: a risk assessment card. Nursing Times 81: 48, 49-55.
Waterlow, J.A. (1995) Reliability of the Waterlow score. Journal of Wound Care. 4: 10, 474.