Family Practice Vol. 17, No. 6, 490-496
© Oxford University Press 2000
The construction of a patient record-based risk model for recurrent falls among elderly people living in the community
Department of General Practice and
a Department of Medical Sociology, Maastricht University, Postbox 616, 6200 MD Maastricht and
b Institute for Rehabilitation Research, Hoensbroek, The Netherlands
Stalenhoef PA, Diederiks JPM, Knottnerus JA, de Witte LP and Crebolder HFJM. The construction of a patient record-based risk model for recurrent falls among elderly people living in the community. Family Practice 2000; 17: 490–496.
Received 2 May 2000; Accepted 17 July 2000.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Background. Predictive models of fall risk in the elderly living in the community may contribute to the identification of elderly at risk for recurrent falling.
Objectives. Our aim was to investigate occurrence, determinants and health consequences of falls in a community-dwelling elderly population and the contribution of data from patient records to a risk model of recurrent falls.
Methods. A population survey was carried out using a postal questionnaire. The questionnaire on occurrence, determinants and health consequences of falls was sent to 2744 elderly persons of 70 years and over, registered in four general practices (n = 27 000). Data were analysed by bivariate techniques and logistic regression.
Results. A total of 1660 (60%) responded. Falls (
1 fall) in the previous year were reported by 44%: one-off falls by 25% and recurrent falls (2 falls) by 19%. Women had significantly more falls than men. Major injury was reported by 8% of the fallers; minor injury by 49%. Treatment of injuries was by the GP in 67% of cases. From logistic regression, a risk model for recurrent falls, consisting of the risk factors female gender, age 80 years or over, presence of a chronic neurological disorder, use of antidepressants, problems of balance and sense organs and complaints of muscles and joints was developed. The model predicted recurrent falls with a sensitivity of 64%, a specificity of 71%, a positive predictive value of 42% and a negative predictive value of 86%.
Conclusion. A risk model consisting of six variables usually known to the GP from the patient records may be a useful tool in the identification of elderly people living in the community at risk for recurrent falls.
Keywords. Elderly, falls, general practice, patient records, risk model.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Falls in the elderly belong to the major problems in health care, because of their high incidence and severe consequences.1–4 From earlier studies, it is known that 25–40% of community-dwelling elderly people of 65 years and over have at least one fall each year, and half of them fall twice or more.4–6 The incidence of falls increases with age and reaches 40% at the age of 80 years or over, and in institutionalized elderly persons even 100%.4,7 Women fall more often than men, but the sex difference in fall incidence disappears in the higher age categories.4,5,7 Falls may have important consequences for health and functioning.8–10 Fractures are the most common serious injuries resulting from falls in elderly persons. Hip fractures have an incidence of 0.5–1% a year and a mortality of about 30% in the first 3 months.4,11–13 Other consequences are head traumas and wrist or hand fractures, which occur in about another 5%.14 The inability to get up after a fall may result in serious morbidity and a high mortality.15,16 Psychological and social consequences, such as fear of falls and social isolation, may also occur, sometimes resulting in a ‘post-fall syndrome’.17–19
Many determinants and risk factors are associated with falls in the elderly.4,5,11,20–23 Falls are usually classified on the basis of the type of risk factors such as intrinsic (individual), extrinsic (environmental) and risk factors related to hazardous behaviour.4,24 The predominant role of intrinsic risk factors in the aetiology of falls is generally accepted.25 The most important intrinsic risk factors are mobility impairment, cognitive deterioration, use of medication, especially sedatives, quantity of medication taken regularly, a history of stroke and orthostatic hypotension.4,5,11,20–23 However, there is a great diversity in the number and kind of risk factors.4,5 In an earlier study, physical health indicators, such as the number of complaints and disorders, appeared to have strong associations with falls and recurrent falls.26 In some studies, the difference in the risk profile between one-time and recurrent fallers is emphasized.2,11 For the GP, the question when a single fall occurs should be: is it an occasional event or the beginning of a series of falls leading to general deterioration and impairment?
Based on the results of a systematic review, risk factors can be classified, apart from the classification into intrinsic and extrinsic risk factors, into four main areas:5 (i) demographic determinants; (ii) physical and (iii) mental health indicators; and (iv) risk factors concerning and influencing mobility, including environmental hazards. Most of these risk factors can be deduced from the medical history, patient records and assessments at the disposal of the GP. A risk model based on data easily and directly available from patient records could be very helpful to the GP.21 We therefore set up a community survey to study the incidence, determinants and health consequences of falls and recurrent falls in a community-dwelling elderly population. In addition, we evaluated the contribution of data to a risk model for recurrent falls.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Participants
All elderly persons of 70 years or over (n = 2946), registered in three primary health care centres and one dual practice (n = 27 000), were selected for this study. Of these, 202 individuals were excluded by the GPs. Exclusion criteria were wheelchair dependence, advanced cognitive impairment, illiteracy and severe somatic or psychiatric disease, which would hamper participation. Furthermore, persons living in or planned for admission into a hospital or nursing clinic were excluded, but those living independently in residential homes and apartments for the elderly were included. Practices are members of the Registration Network Family Practices (RNH).27
The study design was approved by the Ethical Review Board of the University Hospital Maastricht.
Measurements
A questionnaire was sent to the selected population (n = 2744), including the following fall-related items:
- (i) Socio-demographic data: age, gender, educational level, marital status and the presence of stairs in the dwelling.
- (ii) Fall history and injuries due to falls: falls and near-falls in the previous year, perceived causes of falls, injuries due to falls, treatment and fear of falling.
- (iii) Physical health status: complaints and disorders, the number of medicines taken, the regular use of prescribed medication being considered to increase the risk of falling, perceived health status and alcohol consumption.28,29
- (iv) Mental health status: the Psychological Autonomy and Communication and the Emotional Behaviour scales from the Sickness Impact Profile (SIP68).30–32
- (v) Mobility impairment: a 5-point rating scale regarding perceived gait problems and the Mobility Control (MC) scale from (SIP68).30–32
- (vi) Dependence, activities and household activities of daily life (ADL/HDL): a 5-point rating scale on need for professional aid and sum scores of six 3-point scales on ADL/HDL for the determination of the level of independent functioning.33
- (ii) Fall history and injuries due to falls: falls and near-falls in the previous year, perceived causes of falls, injuries due to falls, treatment and fear of falling.
In the questionnaire, a definition of a fall is given as any unintentional coming down on the ground or to a lower level. ‘Ever fallers’ were defined as participants with at least one fall in the previous year; ‘one-time fallers’ with only one fall and ‘recurrent fallers’ with two falls or more in the previous year.
Data analysis
To detect significant differences of sociodemographic characteristics between the study population and the up-dated RNH practice population, a chi-square test was performed. Potential determinants were analysed by odds ratios (OR) with 95% confidence intervals (95% CI). Three outcome variables concerning falls in the previous year were defined: ‘ever fallers’ (1 fall), ‘one-time fallers’ and ‘recurrent fallers’ (2 falls), all contrasted to ‘non-fallers’.
Age was dichotomized into two categories: <80 years and 80 years. Cut-off values for dichotomization of other variables were used as recommended in the literature or, in case of the SIP68 scores, median values were used.
Factor analysis (principle components analysis, followed by varimax rotation) for complaints was performed to obtain a reduction in the number of complaints and to determine coherence between variables. Three relevant factors were extracted: factor 1 related to complaints of mobility such as pain and stiffness of joints and muscles, backache and foot problems; factor 2 related to problems of balance, dizziness and impairment of vision and hearing; and factor 3 related to chest complaints such as chest pain, palpitations and dyspnoea.
Forward stepwise logistic regression analysis of age, gender, variables reaching significance in the bivariate analysis and factors 1 and 2 from the factor analysis of complaints was performed. Variables were selected that had independent significant associations with outcome, taking their contribution into account simultaneously. A risk model containing the strongest independent risk factors was derived. The threshold for significance was 
= 0.05. The fit of the logistic model was assessed using the goodness-of-fit test according to Hosmer Lemeshow.34
A receiver operator characteristic (ROC) analysis was performed to describe the relationship between sensitivity and specificity at different cut-off values and to determine the maximal sum of sensitivity and specificity as the statistically optimal cut-off value. The predictive power of the logistic model was determined using the area under the ROC curve (AUC).35,36 Finally, positive and negative predictive values (PV+ and PV–) of the model were calculated.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Subjects
After one postal reminder, completed questionnaires were received from 1660 persons, 642 men (39%) and 1018 women (61%) (response rate: 60%). Table 1
gives the main sociodemographic characteristics of the study population (n = 1660) compared with those of the registered 70+ population of the RNH practices (n = 2170). Responders were similar to the RNH 70+ population regarding gender, but not regarding age: in the study population, the percentage of subjects of 80 years or over was lower than in the RNH population. Percentages of widowed and single persons were similar in both populations.
|
In the study population, 89% were living independently. The level of education was subdivided into low (63%), medium (20%), high (13%) and not classifiable (4%). The presence of stairs was reported by 57%.
Falls and injuries
Seven hundred and twenty-three subjects (44%) reported at least one fall in the previous year. One fall was reported by 413 (25%) and recurrent falls (2 falls) by 310 persons (19%). Women had significantly more one-off falls (29 versus 19%) and recurrent falls (22 versus 12%) than men. Indoor falls were reported by 41%, outdoor falls by 42%, and both indoor and outdoor falls by 17%. Near-falls were reported by 37% and fear of falling by 48% of the respondents. Generally, fallers have a higher mean age than non-fallers: mean age of non-fallers was 75.4 years (SD 4.8), of one-time fallers 76.4 years (SD 5.4) and of recurrent fallers 77.4 years (SD 5.3).
Table 2 presents major and minor injuries due to falls in the previous year with type of injury, the proportion of subjects needing treatment and the location of treatment.
|
Determinants
Table 3
lists prevalences of and odds ratios for potential determinants of falls (1 fall), one-off falls and recurrent falls (2 falls). Comparison of the odds ratios in the three outcome groups reveals a spectrum of increasing odds ratios for one-time fallers, ever fallers and recurrent fallers. The strongest associations are shown in the recurrent faller group. It has to be noted that the prevalence of chronic neurological disorders and the use of antidepressants is low: 3 and 2%, respectively. About 73% of all respondents regularly used prescribed medication, being considered as risk factors for falls. Antidepressants were significantly associated with falls in all fall groups; hypnotics and diuretics only in the ever and the recurrent falls groups.
|
Risk model
Age, gender and variables showing significant associations with recurrent falls in the bivariate analysis, including two factors from the factor analysis, were retained in a logistic regression analysis. Six variables appeared to make an independent statistically significant contribution to the prediction of recurrent falls. Table 4
reveals that female gender, the presence of a chronic neurological disease and the use of antidepressants are the strongest independent risk factors for recurrent falls in the model. In a calculation of probabilities for recurrent falls, three examples of the application of the model are given.
|
The results of a Hosmer Lemeshow goodness-of-fit test gave a P = 0.6. The model had satisfactory scores, i.e. there is no significant difference between expected and observed frequency of recurrent fallers (chi-square: 5.53). The AUC (0.73) is reasonably high.
Table 5 shows that the statistically optimal cut-off value of the model is reached at 0.26, representing the maximum sum of the sensitivity and the specificity. The positive predictive value (PV+) at this cut-off is rather low (42%), but the PV– value is high (86%).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Main results
The reported incidences of falls and mean ages of the fallers, one-time, recurrent fallers and major and minor injuries are comparable with those found in other retrospective and prospective studies.2,4,5,6,8,11,20–23 More than two-thirds of all fall-related injuries were treated by the GP and about one-third in an emergency ward of a hospital, which emphasizes the important contribution of the GP to the management of fall-related injuries in community-dwelling elderly.37
The use of sedatives, hypnotics and antidepressants is associated with ever and recurrent falls, but in the regression analysis only the use of antidepressants had a significant, independent contribution to the risk of ever falls and recurrent falls.8
The logistic regression analysis yields a risk model for recurrent falls consisting of six variables, which are usually known to the GP from the patients' medical records.
With regard to the PV+ and PV–, the model is apparently more appropriate to predict no falls in the absence of risk factors than the occurrence of falls in the presence of them.
Methodological considerations
A response rate of 60% may not seem very high, but is comparable with other reported rates of postal surveys in the elderly published in medical journals.38 The study population had a similar gender distribution as the community population of persons of 70 years and over, but the very old (80+) are somewhat under-represented. In general, we think that the study population represents a relatively ‘healthy’ population of the elderly.
Recall bias, due to short-term memory loss, was avoided by pre-structuring the questions on falls into two separate, sequential periods of a half year each.39,40 Data on incidence from our study are comparable with those of prospective studies, supporting the reliability of the outcomes.4,11,20–23
As the study had a cross-sectional design, causal relationships cannot, in principle, be established. However, on the basis of prior knowledge, common sense and literature, the plausibility of the assumed causal relations of the factors of the model seems satisfactory.
Implications for general practice
Information on all six significant risk factors for recurrent falls in the risk model (age, gender, pain and/or stiffness in muscles and joints, problems with balance and/or senses, the presence of a chronic neurological disorder and the use of antidepressants) is generally known to the GP or easy to extract from the patients' medical records. Subsequently, elderly persons identified by data from the patient record as being at risk for recurrent falls can be assessed by a concise physical examination, particularly by mobility performance tests, to identify the main preventable risk factors.
Recurrent falls can be predicted with a moderate predictive value. The sensitivity value, however, is satisfactory from a preventive point of view.
Our conclusion is that a useful risk model to predict recurrent falls can be based on a limited number of parameters which are generally known to the GP or directly derivable from the patient records.
|
Acknowledgments |
|---|
This study was supported by a grant from ZON (ZorgOnderzoek Nederland; Care Research, The Netherlands.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
1 Kellogg International Work Group on the Prevention of falls by the elderly. The prevention of falls in later life. Dan Med Bull 1987; 34: 1–24.[Medline]
2
Campbell AJ, Reinken J, Allan BC, Martinex GS. Falls in old age: a study of frequency and related clinical factors. Age Ageing 1981; 10: 264–270.
3 van Weel C, Vermeulen H, van den Bosch W. Falls, a community care perspective. Lancet 1995; 345: 1549–1551.[ISI][Medline]
4 Tinetti ME, Speechley M, Ginter SF. Risk factors for falls among elderly persons living in the community. N Engl J Med 1988; 319: 1701–1707.[Abstract]
5
Stalenhoef PA, Crebolder HFJM, Knottnerus JA, Horst van der FGEM. Incidence, risk factors and consequences of falls among elderly subjects living in the community: a criteria-based analysis. Eur J Publ Health 1997; 7: 328–334.
6 Prudham D, Grimley Evans J. Factors associated with falls in the elderly: a community study. Age Ageing 1981; 10: 141–146.
7 Rubinstein LZ, Robbins AS, Schulman BL, Rosado J, Osterweil D, Josephson KR. Falls and instability in the elderly. J Am Geriatr Soc 1988; 36: 266–278.[ISI][Medline]
8
Blake AJ, Morgan K, Bendall MJ et al. Falls by elderly people at home: prevalence and associated factors. Age Ageing 1988; 17: 365–372.
9 Wild D, Nayak USL, Isaacs B. How dangerous are falls in old people at home? Br Med J 1981; 282: 266–268.
10 Tinetti ME. Prevention of falls and fall injuries in elderly persons: a research agenda. Prev Med 1994; 23: 756–762.[ISI][Medline]
11 Nevitt CN, Cummings SR, Kidd S, Black D. Risk factors for recurrent nonsyncopal falls. A prospective study. J Am Med Assoc 1989; 261: 2663–2668.[Abstract]
12 Grisso JA, Kelsey JL, Strom BL et al. Risk factors for falls as a cause of hip fracture in women. N Engl J Med 1991; 324: 1326–1331.[Abstract]
13 Boereboom FTJ, Raymakers JA, Faber JAJ, Duursma SA. Survival following hip fractures and total hip arthroplasty for osteoarthritis of the hip, in The Netherlands. In Hip fractures: a public health problem. PhD Thesis, Rijksuniversiteit Utrecht, The Netherlands, 1991.
14
Sattin RW, Lambert Huber DA, DeVito CA et al. The incidence of fall injury events among the elderly in a defined population. Am J Epidemiol 1990; 131: 1028–1037.
15 Tinetti ME, Liu AL, Claus E. Predictors and prognosis of inability to get up after falls among elderly persons. J Am Med Assoc 1993; 265: 269–270.
16 Ryynänen O-P, Kivelä S-L, Honkanen R, Laippala P. Falls and lying helpless in the elderly. Zeitschr Gerontol 1992; 25: 278–282.
17 Tinetti ME, Richman D, Powell L. Falls efficacy as a measure of fear of falling. J Gerontol Psyciatr Sci 1990; 45: 239–243.
18
Vellas BJ, Wayne SJ, Romero LJ. Fear of falling and restriction of mobility in elderly fallers. Age Ageing 1997; 26: 189–193.
19 Murphy J, Isaacs B. The post-fall syndrome: a study of 36 elderly patients. Gerontology 1982; 28: 265–270.[ISI][Medline]
20
O'Loughlin JL, Robitaille Y, Boivin J-F, Suissa S. Incidence of and risk factors for falls and injurious falls among the community-dwelling elderly. Am J Epidemiol 1993; 137: 342–354.
21
Graafmans WC, Ooms ME, Hofstee HMA, Bezemer PD, Bouter LM, Lips P. Falls in the elderly: a prospective study on risk factors and risk profiles. Am J Epidemiol 1996; 143: 1129–1136.
22 Campbell AJ, Borrie MJ, Spears GF. Risk factors for falls in a community-based prospective study of people 70 years and older. J Gerontol Med Sci 1989; 44: 112–117.
23
Campbell AJ, Borrie MJ, Spears GF et al. Circumstances and consequences of falls experienced by a community population 70 years and over during a prospective study. Age Ageing 1990; 19: 136–141.
24 Lach HW, Reed TR, Arfken CL et al. Falls in the elderly: reliability of a classification system. J Am Geriatr Soc 1991; 39: 197–202.[ISI][Medline]
25 Tinetti ME, Speechley M. Prevention of falls among the elderly. N Engl J Med 1989; 320: 1055–1059.[ISI][Medline]
26 Stalenhoef PA, Diederiks JPM, Witte de LP, Schiricke KH, Crebolder HFJM. The impact of gait problems and falls on functioning in independent living persons of 55 years and over: a community survey. Patient Educ Couns 1999; 36: 23–31.[ISI][Medline]
27 Metsemakers JFM, Höppener P, Knottnerus JA, Kocken RJJ, Limonard CBG. The Registration Network Family Practices: a computerized health information system in the Netherlands. Br J Gen Pract 1992; 42: 102–106.[ISI][Medline]
28 Bos van den GAM. Zorgen van en voor chronisch zieken (Care of and for chronic ill persons). Thesis. Bohn, Scheltema en Holkema, Amsterdam, 1989.
29 Meyboom-de Jong B. Bejaarde patienten: een onderzoek in twaalf huisartspraktijken (Elderly patients: a study in twelve general practices). Thesis, 1989. Meditekst, Lelystad, The Netherlands.
30 Bergner M, Bobbitt RA, Carter WB, Gilson BS. The Sickness Impact Profile: development and final revision of a health status measure. Med Care 1981; 19: 787–805.[ISI][Medline]
31 Bruin de AF, Diederiks JPM, Witte de LP, Stevens FCJ, Philipsen H. The development of a short generic version of the Sickness Impact Profile. J Clin Epidemiol 1994; 47: 407–418.[ISI][Medline]
32 Bruin de AF, Diederiks JPM, Witte de LP, Stevens FCJ, Philipsen H. Assessing the responsiveness of a functional status measure: the Sickness Impact Profile versus SIP68. J Clin Epidemiol 1997; 50: 529–540.[ISI][Medline]
33 Meenan RF, Gertman PM, Mason JH. Measuring health status in arthritis: the arthritis impact measurement scales. Arthritis Rheum 1980; 23: 146–152.[ISI][Medline]
34 Hosmer DW, Lemeshow S. Assessing the fit of the model. In Applied Logistic Regression. New York: John Wiley and Sons, 1989.
35
Hanley JA, McNeil BJ. The meaning and the use of the area under a receiver operator characteristic curve. Radiology 1982; 143: 29–36.
36 Knottnerus JA. Diagnostic prediction rules: principles, requirements, and pitfalls. Med Decision Making 1995; 22: 341–363.
37 Stalenhoef PA, Busse van der S, Crebolder HFJM. Falls in the elderly: the involvement of the general practitioner. Huisarts Wetensch 1998; 41: 241–245.
38 Asch DA, Jedrziewski MK, Christakis NA. Response rates to mail surveys published in medical journals. J Clin Epidemiol 1997; 50: 1129–1136.[ISI][Medline]
39 Cummings SR, Nevitt MC, Kidd S. Forgetting falls. The limited accuracy of recall of falls in the elderly. J Am Geriatr Soc 1988; 36: 613–616.[ISI][Medline]
40 Hale WA, Delaney MJ, Cable T. Accuracy of patient recall and chart documentation of falls. J Am Board Fam Pract 1993; 6: 239–242.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. C. M. van Haastregt, G. A. R. Zijlstra, E. van Rossum, J. Th. M. van Eijk, and G. I. J. M. Kempen Feelings of Anxiety and Symptoms of Depression in Community-Living Older Persons Who Avoid Activity for Fear of Falling Am J Geriatr Psychiatry, March 1, 2008; 16(3): 186 - 193. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. Melzer, I. Kurz, D. Shahar, M. Levi, and L. Oddsson Application of the voluntary step execution test to identify elderly fallers Age Ageing, September 1, 2007; 36(5): 532 - 537. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S.-T. Liaw, N. Sulaiman, C. Pearce, J. Sims, K. Hill, H. Grain, J. Tse, and C.-K. Ng Falls Prevention within the Australian General Practice Data Model: Methodology, Information Model, and Terminology Issues J. Am. Med. Inform. Assoc., September 1, 2003; 10(5): 425 - 432. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||