Clinical Prediction Guide
A 6-point risk score predicted which elderly patients would fall in hospital
ACP J Club. 1998 May-June;128:81. doi:10.7326/ACPJC-1998-128-3-081
Oliver D, Britton M, Seed P, Martin FC, Hopper AH. Development and evaluation of evidence based risk assessment tool (STRATIFY) to predict which elderly inpatients will fall: case-control and cohort studies. BMJ. 1997 Oct 25;315:1049-53.
To develop and evaluate a risk-assessment tool that uses data from risk factors easily collected by nurses in routine hospital care to predict which elderly patients will fall while in the hospital.
A case-control study to develop the tool, and 2 cohort studies to validate it.
A teaching hospital (development and local validation) and a district general hospital (remote validation) in the United Kingdom.
For development of the tool, 116 patients (mean age 85 y) in a geriatric unit who had fallen (involuntarily coming to rest on the ground or surface lower than their original station) were matched with the patient in the next bed who had not fallen. Local validation was done with 217 patients (395 assessments and 71 falls) and remote validation was done with 331 patients (446 assessments and 79 falls).
Description of prediction guide
For each case- and control-patient, 21 variables were collected on age, sex, function (including the Barthel Index), mental status, medication use, medical history, and nursing assessment of current clinical characteristics. Assessors were not blinded to fall status of the patients. 5 clinical factors (presented with fall or has fallen since admission, agitation, visual impairment, frequent toileting needs, and poor transfer ability and mobility) were associated with falling. For validation, each patient was assessed weekly (up to 8 weeks) using scores from 0 to 5. Risk scores for those who had fallen were compared with those who had not fallen for a given week.
Main outcome measures
Sensitivity and specificity of scores 0 to 5 for predicting falls in the local and remote validation studies.
A score of ≥ 2 had high sensitivity and a score of ≥ 3 had high specificity for predicting elderly persons who were at risk for falling (Table).
A risk score that predicted which elderly patients were at risk for falling was developed, validated, and found to be accurate.
Source of funding: No external funding.
For article reprint: Dr. D. Oliver, Department of Elderly Care (Division of Medicine), United Medical and Dental Schools, St. Thomas's Hospital, London SE1 7EH, England, UK. FAX 44-171-928-2339.
Table. Validation of a 0- to 5-point risk-assessment score to predict which elderly patients are at risk for falling in hospital
|Validation||Score||Sensitivity (95% CI)||Specificity (CI)||+LR*||-LR*|
|Local||≥ 2||93% (84% to 98%)||88% (84% to 91%)||7.6||0.08|
|Local||≥ 3||69% (57% to 80%)||96% (94% to 98%)||18.6||0.3|
|Remote||≥ 2||92% (84% to 97%)||68% (63% to 73%)||2.9||0.1|
|Remote||≥ 3||54% (43% to 66%)||88% (84% to 91%)||4.4||0.5|
*+LR = likelihood ratio for the presence of condition (falls) if the test is positive; -LR = likelihood ratio if the test is negative. Both calculated from data in article.
Interventions targeted to decrease the incidence of falls in the elderly are important in any setting. The appeal of this study by Oliver and colleagues is that it proposes an instrument to identify patients in the hospital who are at risk for falling, using readily available clinical data that can be rapidly summarized in a score by nursing staff.
The instrument had the greatest predictive value in the population for which it had been derived (teaching hospital), but it also performed fairly well when tested in a different setting (community hospital). Implementation of preventive measures, which are supported by observational studies cited by the authors, can be accomplished simultaneously with calculation of a high-risk score.
A limitation of the study is that no randomized trials have convincingly shown the effectiveness of preventive measures in hospitalized geriatric patients. An identification bracelet designed to increase awareness in high-risk patients of their potential for falling failed to produce positive results in 1 study (1), and use of a bed alarm system only resulted in a trend toward a decrease in falls out of bed in another (2). Also, no attempt has been made to differentiate risk for injurious falls from that of noninjurious falls. Even though noninjurious falls have substantial morbidity (e.g., fear of falling, immobility, anxiety, and depression), morbidity and mortality primarily result from hip fractures. The same measures may not be equally effective for the prevention of both. Interventions that have proven to be effective in preventing fractures in nursing homes (3) should receive a high priority in future studies of risk reduction in elderly hospitalized patients.
Because of the variability in patient population and nursing practices (a point stressed by the authors), the scale may need to be validated locally.
Claudia Beghe, MD
University of South FloridaTampa, Florida, USA