Original article
Risk of falls for hospitalized patients: A predictive model based on routinely available data

https://doi.org/10.1016/S0895-4356(01)00406-1Get rights and content

Abstract

The incidence rate of falls is often used as an indicator of nursing care outcome. Comparing outcome between different settings should, however, make allowance for case mix. To measure the incidence of falls, describe their circumstances and develop a prediction model based on routinely collected data to allow comparison between hospital settings with different case mix. A dynamic population of patients hospitalized over a year in which a case was defined as any accidental fall systematically reported on an ad hoc form. A Swiss university hospital of 800 beds; 634 falls were reported for 26,643 hospitalizations over 236,307 hospitalization days. First fall rates were analyzed using a Poisson regression model with routinely computerized discharge data as independent variables. The incidence rate of first falls was 2.2 per 1000 patient-days. For subsequent falls the rates of incidence increased with the number of falls. Five independent variables played a significant role: age, gender, morbidity predisposition, surgical procedure and length of stay. Two of the interactions between these variables were significant and remained in the model (length of stay with age, morbidity with age). The model offers good medical plausibility and satisfactory predictive performance. The proposed model can be used by national health agencies to compute expected first fall rates accounting for case mix. Hospitals can use these rates for evaluation. Recommendations for measuring, monitoring and assessing fall rates are also given.

Introduction

Many hospitals routinely report inpatient falls. As a matter of fact, falls are the most frequently noted incidents 1, 2. Quality assurance programs insist upon the importance of monitoring adverse events and often use fall rates as an indicator of nursing outcome 3, 4, 5. Wide variations in fall rates have been reported among institutions for the elderly (0.2–3.6 per bed per year) [6]. Studies in acute-care hospitals are less numerous, but also show notable variations: 2–15% of inpatients experience at least one fall 1, 7, 8, 9, 10, the range of published incidence rates of falls is wide (0.3–19 for 1000 patient-days) 1, 7, 8, 10, 11, 12, 13, 14, 15. Comparing fall rates among various institutions may be helpful but raises concerns. The lack of an accurate definition of the occurrence indicator compromises data comparability 16, 17. The unit of measurement is often the number of events per patient-days but multiple falls by the same individuals are variably reported, and falls that do not result in visible injuries are sometimes not reported at all. What is more important is that, to clarify the issue of nursing care quality, factors outside the direct influence of hospital policy should be controlled for. Interpreting an outcome must therefore account for variations linked to case mix. Although falls are a traditional target of risk management, comparative data are too scant to set a benchmark determining an acceptable level of falls [18].

Controlling for patient risk is required to evaluate quality differences between care settings. Most research to date has focused on elderly persons living in the community or in nursing homes. A number of researchers have investigated symptoms that may contribute to a fall in these environments: gait and balance disorders 13, 19, 20, 21, dizziness or vertigo 21, 22, visual deficit [23], incontinence [24], cognitive impairment and sedation 6, 19, 24, 25. Selected risk factors have been used to derive relatively simple assessment tools identifying patients liable to be targeted for preventive strategies 13, 24, 26. Because these data have to be extracted from medical or nursing records, they are not readily available for comparison across care units or hospitals. Furthermore, many of those variables escape detection because they are not regularly recorded in medical records [25]. It has been shown that in acute-care settings a few variables could identify patients prone to falls: falls during hospitalization are more common in confused patients and those with greater comorbidity [25]. Several diseases have been shown to increase the risk of fall such as Alzheimer's disease [27], Parkinson's disease [28] and stroke [10]. These findings suggest that risk measures relying only on routinely collected data could perform quite well for hospitalized patients and favorably compare with methods requiring additional record abstracting.

A further concern is the necessity to adopt a probabilistic approach, due to the fact that it is usually difficult to determine whether a fall might have been avoided on a case-by-case basis. Data about the proportion of potentially avoidable falls are scarce; a systematic medical records review of adverse events in hospitalized patients judged that over half of 200 falls with injuries resulted from substandard care [29]. Few falls can be clearly prevented, for instance by suppressing a form of medication or an environmental hazard. Some are probably unavoidable, such as in a unit trying to improve the independent living skills of its patients [30]. Most patients fall because factors related to patients and to environmental hazards interact. In fact, only multifactorial interventions targeting both intrinsic and environmental risk factors have been shown to significantly reduce the number of falls in high-risk populations 31, 32, 33.

The objectives of the study were: (1) to measure the incidence of falls and to describe the circumstances under which they occur, and (2) to use routinely collected information to develop a prediction model, making it possible to compare observed rates accounting for the identifiable patients' risks.

Hospital characteristics, such as the type of care unit and the patient:nurse ratio, were intentionally not taken into account in order to use only case mix variables clearly beyond the hospital's control.

Section snippets

Methods

The Centre Hospitalier Universitaire Vaudois (CHUV) offers an appropriate setting to observe falls, document their circumstances and measure potential risk factors. It generates intense activity (more than 200,000 hospitalization days per annum) for a comprehensive panel of patients—all surgical and medical services provided to patients of all ages with all specialized areas of medicine represented except ophthalmology and psychiatry. Thus, the CHUV presents a large spectrum of predictive

Description of falls

Six hundred thirty-four falls were reported during 236,307 hospitalization days, corresponding to a global rate of 2.7 falls per 1000 days. The proportion of inpatients falling was 1.8% (488/26,643). The incidence rate was 2.2 first falls per 1000 days. Twenty percent of patients with a first fall relapsed (100/488) during the follow-up period. The incidence rate of a second fall by patients who had fallen previously was 12.7 per 1000 days (100/7893). The incidence rate of a subsequent fall

Comments

The multivariate model offers good predictive performance. Poisson modeling is adequate and data do not present over-dispersion. The range of predicted rates is large, varying from a low of 0.32 per 1000 days to 9.87 in the high- risk group. The age variable is the most significant factor, which comes as no surprise from a medical point of view. The fact that morbidity predisposition and length of stay show similar contributions suggests that AP-DRGs do not sufficiently account for the severity

References (51)

  • M Casas

    Issues for comparability of DRG statistics in Europeresults from EURODRG

    Health Policy

    (1991)
  • V.R Morgan et al.

    Hospital fallsa persistent problem

    Am J Public Health

    (1985)
  • J.C Sutton et al.

    Patient accidents in hospitalincidence, documentation and significance

    Br J Clin Pract

    (1994)
  • A Avillion et al.

    Developing a nursing quality assurance programassessment and planning

  • M.C Ruckstuhl et al.

    Patients fallsan outcome indicator

    J Nurs Care Qual

    (1991)
  • Nursing quality indicatorsdefinitions and indications

    (1996)
  • L.Z Rubenstein et al.

    Falls in the nursing home

    Ann Intern Med

    (1994)
  • A Huda et al.

    Evolution of compliance within a fall prevention program

    J Nurs Care Qual

    (1998)
  • C Favaretti et al.

    The quality improvement system in the hospitals of Padua (Italy)

    Qual Assurance Health Care

    (1992)
  • E.C Poster et al.

    A retrospective cohort study of falls in a psychiatric inpatient setting

    Hosp Community Psychiatry

    (1991)
  • J.A Tutuarima et al.

    Risk factors for falls of hospitalized stroke patients

    Stroke

    (1997)
  • L Nyberg et al.

    Incidence of falls in three different types of geriatric carea Swedish prospective study

    Scand J Soc Med

    (1997)
  • J.M Morse et al.

    A retrospective analysis of patient falls

    Can J Public Health

    (1995)
  • M.L Aisen et al.

    Falls on a neurorehabilitation unitreassessment of a prevention program

    J Am Paraplegia Soc

    (1994)
  • P Sehested et al.

    Falls by hospitalized elderly patientscauses, prevention

    Geriatrics

    (1977)
  • Cited by (159)

    View all citing articles on Scopus
    View full text