Evidence table 2Risk factors: multivariate analysis

(please refer to Evidence table 1 for further details)

Falls history
Community dwelling: statistically significant results
StudyResultsQualityComments
Northridge 1996Baseline status for all analysis included one fall prior to baseline.
OR (95%CI)
One non-environmental fall at follow-up
1.15 (1.01-1.31)
One environmental fall at follow-up
1.20 (1.05-1.36)
Two non-environmental falls
1.19 (1.05-1.36)
Two environmental falls
1.15 (1.00-1.32).
MediumOnly previous fallers were recruited.
Subjective baseline measurement of risk factors Analysis of two falls at follow-up n= less than 50.
OR are adjusted for all other variables.
Covinsky 2001OR (95% CI)
2.42 (1.49-3.93).
MediumRetrospective falls history at follow-up.
Subjective baseline measurement of risk factors.
Three models were computed and each adjusted for falls history.
All risk factors significant at p<0.05 were retained in multivariate analysis.
Tromp 2001OR (95%CI)
Single fallers
2.6 (2.0-3.3)
Recurrent fallers
3.1 (2.2-4.4).
HighAll risk factors were adjusted for the others and all were adjusted for age, gender. Recurrent falls and fractures.
Friedman 2002OR (95% CI)
2.51(2.04-3.09)
HighLogistical regression. Adjusted for other variables in the model. (Please refer to Evidence table 5 for further details).
Stalenhoef 2002OR (95%CI)
3.1 (1.5-6.7).
HighVariables meeting an OR of two or more in bivariate analysis were entered into multivariate analysis. Stratification included age and sex were also entered.
Adjustment reported but unclear.
Stenbacka 2002RR(95%CI)
Earlier injuries: men >60years
2.48(1.19-5.13).
Medium
Wood 2002OR (95%CI)
4.0 (1.3-12.1.
LowVariables significant at p<0.1 were entered in logistic regression analysis. No adjustment for covariates reported.
Small sample n=74 fallers.
Parkinson's disease only.
Extended care: statistically significant results
StudyResultsQualityComments
Thapa 1996IDR (95%CI)
Non-ambulatory
2.23 (1.14-4.37).
HighMultivariate model included factors with a significance of p=≤0.10.
Separate analysis was conducted for the non-ambulatory and ambulatory participants.
Each variable was adjusted for other variables with exception of falls history.
Falls history was assessed in a separate model.
Bueno-Cavanillas 2000DR (95%CI)
Intrinsic falls
1.9 (1.3-2.9).
LowAdjusted density ratios referred to but no details.
Small sample n=106 falls.
Kallin 2002OR (95%CI
4.65 (1.48-14.60).
LowSmall sample multivariate analysis.
No adjustment for confounding.
Extended care: statistically non-significant results
StudyComments
O Loughlin 1993Falls history not included in pooled logistical regression for other factors.
Secondary analysis including falls history in the model (IRR= 2.1 (1.4-3.3). Poor methods of reporting.
Tinetti 1995Adjusted RR (95%CI)
1.2 (0.9-1.5)
Thapa 1996Adjusted IDR (95%CI)
Ambulatory
1.22(0.73-2.04)
Koski 1998Measured but not reported.
Tromp 1998Previous falls established by history of fracture.
Cesari 2002Unsure if measured at baseline. MDS at baseline. Not reported as significant in results.
Stenbacka 2002Adjusted for age
RR(95%CI)
Earlier injuries: Women>60 years
1.21(0.76-1.92)
Muscle weakness
Statistically non-significant results (ns)
StudyComments
Bueno-Cavanillas 2000DR (95%CI) Adjusted but unclear reporting
Intrinsic fall
Poor muscle tone in hand:
1.4 (0.9-2.4)
Extrinsic fall
Poor muscle tone in hand:
1.3 (0.7-2.3).
Koski 1998Ns in multivariate analysis.
Gait deficit
Gait, mobility and balance described separately but some overlap may be present due to some tests examining both aspects
Community dwelling: statistically significant results
StudyResultsQualityComments
Koski 1998OR (95%CI)
Incomplete step continuity
2.2 (1.11-4.17).
HighLogistic regression with adjustment for age and gender.
Cesari 2002OR (95%CI)
Gait problems
2.13 (1.81-2.51).
MediumLogistic regression with adjustment for age and gender.
Northridge 1996OR (95%CI)
Tandem walk performance: non-environmental single fall
1.96(1.44-2.68).
MediumOnly previous fallers were recruited.
Analysis of two falls at follow-up n= less than 50.
OR are adjusted for all other variables.
Extended care: statistically non-significant results
StudyComments
Bueno-Cavanillas 2000Gait disorders were examined in this study but categorised into twelve domains. Adjusted density ratios referred to but no details
Small sample n=106 falls:
Multivariate analysis: sitting down incorrectly: (? Not specific enough) DR=3.4 (1.5-7.6)
Kallin 2002Ns in logistic regression.
Community dwelling: non-significant results
Northridge 1996Adjusted for all variables.
Tandem walk performance: environmental single fall 1.24 (0.91-1.69)

Non-environmental and environmental second fall both ns in multivariate analysis (no data).
Stalenhoef 2002TUGT: Ns in logistical regression.
Wood 2002Parkinson's disease only.
Gait measured at baseline, ns in multivariate analysis (no data).
Tinetti 1995Gait speed: Ns Multivariate analysis.
Balance (including dizziness)
Community dwelling: statistically significant results
StudyResultsQualityComments
O Loughlin 1993IRR(95%CI) (Adjusted)
Dizziness= 2.0(1.3-2.8).
MediumPooled logistical regression, with all ns risk factors that were not retained in the model were entered one by one to identify potential confounders.
Stalenhoef 2002OR(95%CI)
Abnormal postural sway
3.9 (1.3-12.1).
HIghVariables meeting an OR of two or more in bivariate analysis were entered into multivariate analysis. Stratification included age and sex were also entered.
Adjustment reported but unclear.
Covinsky 2001OR(95%CI)
Unbalanced or dizzy:
Model 2 adjusted for falls history= 1.96(1.25-3.07)
Model 3 included falls history= 1.83(1.16-2.89).
MediumMultivariate logistic regression.
Extended care: statistically significant results
Bueno-Cavanillas 2000Eight aspects of balance examined and analysis according to intrinsic or extrinsic fall. All ns in multivariate analyses with exception of Romberg incorrect: DR=4.0 (1.2-13.3)Low small sampleCox regression analysis no adjustment variables reported.
Extended care: statistically non-significant results (ns)
StudyComments
O Loughlin 1993NS in pooled logistical regression, with all ns risk factors that were not retained in the model were entered one by one to identify potential confounders.
Tinetti 1995Balance and Gait score= ns in multivariate analysis.
Northridge 1996Balance on one leg
Multivariate analysis with adjustment for all other variables
First fall
Environmental= ns
Non-environmental = ns.
Second fall
Environmental= OR 1.12(0.94-1.32)
Non-environmental = OR 0.71(0.55-0.93).
Koski 1998Unsteady standing
NS in multivariate analysis.
Wood 2001Balance score ns in multivariate analysis. Small sample n= 69 fallers/ 32 non fallers.
Stalenhoef 2002Trendelenburg test (abnormal), bending down test, functional reach test.
All ns in logistic regression.
Bueno-Cavanillas 2000Eight aspects of balance examined and analysis according to intrinsic or extrinsic fall. All ns in multivariate analyses with exception of Romberg incorrect test as above.
Kallin 2002Functional reach: ns in multivariate analysis.
Mobility impairment
Community dwelling: statistically significant results
StudyResultsQualityComments
O Loughlin 1993IRR(95%CI) (adjusted)
Trouble walking 400m=
1.6(1.2-2.4)
Trouble bending down=
1.4(1.0-2.0).
MediumPooled logistical regression, with all ns risk factors that were not retained in the model were entered one by one to identify potential confounders.
Covinsky 2001OR(95%CI)
Impaired mobility:
Model 2 adjusted for falls history= 3.06(1.93-4.86)
Model 3 included falls history= 2.64(1.64-4.26).
MediumMultivariate logistic regression.
Community dwelling: statistically non-significant results
StudyComments
Bueno-Cavanillas et al (2000)See gait and balance.
Kallin et al (2002)User of walking aid ns in logistic regression.
Cesari et al (2002)Unsure if measured at baseline. MDS at baseline. Not reported as significant in results.
Stalenhoef et al (2002)Mobility was assessed with balance and gait tests. SIP68 MC also utilised within the mobility domain= ns in multivariate analysis. Bivariate= 2.6(1.3-5.3).
Fear of falling
Community dwelling: statistically significant results
StudyResultsQualityComments
Arfken 1994
USA
OR (95%CI)
1 fall
A= 1.52 (1.06-2.17)
B= 2.49(1.48-4.20)

Recurrent falls
A=1.71(1.01-2.89)
B=3.12(1.61-6.06).
HighStatistical methods:
Logistic regression adjusted for age, gender.

(Please refer to Evidence table 5 for further details).
Cumming 2000Adjusted hazard ratio (95%CI)
Falls efficacy scale ≤ 75 (n= 88) =2.09 (1.31-3.33).
HighLinear regression with adjustment for other related variables.
Friedman 2002OR(95%CI)
Fear of falling at baseline/ falls at follow-up= 1.78(1.41-2.24)
Fear of falling at baseline and follow-up= 5.40(4.23-6.91)
Fear of falling at baseline with no history of falling= 1.79(1.33-2.42).
MediumThis study explored the temporal relationship between falls and the fear of falling.
Logistic regression analysis with all other factors entered into the model.
Community dwelling: statistically non-significant results
StudyComments
Tromp 2001Ns in multivariate analysis.
Visual deficit
Community dwelling: statistically significant results
StudyResultsQualityComments
Northridge 1996OR(95%CI) adjusted
Second fall: non-environmental
Corrected visual acuity (5 units worse)
1.18(1.00-1.39)
Environmental
1.22(1.02-1.46).
LowMultivariate logistic regression. Each variable adjusted for others.
Koski 1998OR (95%CI)
Poor distant visual acuity
2.3(1.18-4.63).
HighLogistic regression with adjustment for age and gender.
Extended care: statistically significant results
Kallin 2002OR (95%CI)
Impaired vision
5.85(1.14-30.08).
Low - small sampleLogistic regression. No adjustment for confounding reported.
Extended care: statistically non-significant results (ns)
StudyComments
Tinetti 1995Visual impairment
Ns in multivariate analysis.
Northridge 1996OR(95%CI)
Depth perception score
Multivariate analysis with adjustment for all other variables
First fall
Environmental= 0.81(0.70-0.94)
Non-environmental = 1.04(0.92-1.18).
Tromp 1998Multivariate analysis adjusted for age and gender, and recurrent falls:
Vision problems:
OR 1.7(0.9-3.0).
Cesari 2002Visual impairment
Ns in multivariate analysis.
Stalenhoef 2002Distant vision
Ns in multivariate analysis.
Wood 2002Visual acuity
Ns in multivariate analysis.
Thapa 1996Visual impairment measured but ns in multivariate analysis. No data reported.
Bueno-Cavanillas 2000Many aspects of vision were measured. Ns in multivariate analysis.
Cognitive impairment
Community dwelling: statistically significant results
StudyResultsQualityComments
Tinetti 1995Adjusted OR (95%CI)
Serious injurious fall (entire cohort)
MMSE<26 = 2.2(1.5-3.2)
Serious injurious fall (those who fell at least once
MMSE<26 =2.4(1.6-3.4).
HighPooled logistic regression adjusted for housing stratum, moth of follow-up, history of fall, at least two chronic conditions, Balance and gait scores female gender, body mass index.
van Schoor 2002Adjusted OR (95%CI)
*RCPM and adjusted variable
Age and education=1.03(1.00-1.07)

**CT


Age and education=1.02(1.00-1.04).
MediumLogistic regression with adjustment for age, sex, depression, education level and stroke.
Extended care: statistically significant results
Bueno-Cavanillas 2000DR (95%CI)
Intrinsic fall / dementia
6.2(1.7-23.3).
Low - small sampleCox regression analysis no adjustment variables reported.
Community dwelling and extended care: statistically non-significant results (ns)
StudyComments
Tinetti 1995MMSE <20 ns in multivariate analysis.
Northridge 1996Mental status test ns in multivariate analysis.
Tromp 1998Cognitive impairment ns in multivariate analysis.
Cesari 2002Cognitive performance scale ns in multivariate analysis.
Stalenhoef 2002MMSE<24 ns in multivariate analysis.
van Schoor 2002RCPM and adjusted variable
Age= 1.02(0.98-1.05)
Age and depression and education= 1.03(0.99-1.07)
MMSE and adjusted variable
Age= 1.03(0.99-1.07)
Age and depression= 1.02(0.97-1.06)
Age and depression and education= 1.03(0.99-1.08)
CT and adjusted variable
Age= 1.00(0.99-1.02)
Age and depression and education= 1.02(0.99-1.04).
Thapa 1996Adjusted IDR(95%CI)
Cognitive impairment / moderate
1.49(0.89-2.50)
Cognitive impairment / severe
1.59(0.78-3.26)
Adjusted for all other variables.
Kallin 2002MMSE ns in multivariate analysis.
Wood 2002MMSE ns in multivariate analysis.
MMSE= mini mental state examination
Urinary incontinence
Community dwelling: statistically significant results
StudyResultsQualityComments
Tromp 1998Adjusted OR(95%CI)
1.8(1.2-2.7).
HighLogistic regression adjusted for age, gender recurrent falls
Brown 2000Adjusted OR(95%CI)
Urge incontinence
1.26(1.14-1.40).
HighMultivariate model with adjustment for all factors.
Statistically non-significant results
StudyComments
Tinetti 1995Ns in multivariate analysis.
Koski 1998Urinary incontinence ns in multivariate analysis.
Brown 2000Adjusted OR(95%CI)
Stress incontinence
1.06(0.95-1.19).
Cesari 2002Adjusted OR(95%CI)
1.06(0.93-1.20)
Adjusted for age and gender.
Thapa 1996Ns in multivariate analysis.
Home hazards
Community dwelling: statistically significant results
StudyResultsQualityComments
Cesari 2002Adjusted OR(95%CI)
1.51(1.43-1.69).
MediumLogistic regression with adjustment for age and gender.
Gill 2000Proportional Hazards ratio (95%CI)
Loose rugs, mats etc=
5.87(1.42-24.2)
Carpet fold or tripping hazard=
3.45(1.29-9.27).
HighAdjusted for age, gender and housing type.

Many potential hazards were assessed in this study. Only significant in adjusted results reported here.
*

RCPM = non-verbal, visual test to measure a persons ability of nonverbal and abstract reasoning.

**

CT= coding task

From: Appendix E, Clinical effectiveness evidence table

Cover of Clinical Practice Guideline for the Assessment and Prevention of Falls in Older People
Clinical Practice Guideline for the Assessment and Prevention of Falls in Older People.
NICE Clinical Guidelines, No. 21.
National Collaborating Centre for Nursing and Supportive Care (UK).
Copyright © 2005, Royal College of Nursing.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means electronic,mechanical, photocopying, recording or otherwise,without prior permission of the Publishers or a licence permitting restricted copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1T 4LP.This publication may not be lent, resold, hired out or otherwise disposed of by ways of trade in any form of binding or cover other than that in which it is published,without the prior consent of the Publishers.

PubMed Health. A service of the National Library of Medicine, National Institutes of Health.