SEARCH

SEARCH BY CITATION

Keywords:

  • bone/musculo-skeletal;
  • elderly;
  • falls;
  • geriatric medicine;
  • internal medicine;
  • polypharmacy

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure statement
  9. References

Objective:  To investigate the predictors of falls, such as comorbidity and medication, in geriatric outpatients in a longitudinal observational study.

Methods:  A total of 172 outpatients (45 men and 126 women, mean age 76.9 ± 7.0 years) were evaluated. Physical examination, clinical history and medication profile were obtained from each patient at baseline. These patients were followed for up to 2 years and falls were self-reported to their physicians. The factors associated with falls were analyzed statistically.

Results:  A total of 32 patients experienced falls within 2 years. On univariate analysis, older age, osteoporosis, number of comorbid conditions and number of drugs were significantly associated with falls within 2 years. On multiple logistic regression analysis, the number of drugs was associated with falls, independent of age, sex, number of comorbid conditions and other factors that were significantly associated in univariate analysis. A receiver–operator curve evaluating the optimal cut-off value for the number of drugs showed that taking five or more drugs was a significant risk.

Conclusion:  In geriatric outpatients, polypharmacy is associated with falls. Intervention studies are needed to clarify the causal relationship between polypharmacy, comorbidity and falls. Geriatr Gerontol Int 2012; 12: 425–430.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure statement
  9. References

Previous studies have assessed the risk factors for falls in community-dwelling elderly,1–3 but not in geriatric outpatients, and history of falls, physical ability and living environment were found to be predictors of falls. Outpatients have different characteristics from community-dwelling elderly, and previous studies have not assessed whether medical comorbidity and therapeutic drugs might be risk factors for falls. Falls in patients on medication are complicated, because some drugs, such as aspirin, can cause serious bleeding when they have injurious falls, and others, such as antihypertensive4 and hypoglycemic5,6 agents, can cause falls.

Previously, we reported that polypharmacy was associated with the tendency for falls using four indices of fall tendency in a cross-sectional setting in geriatric outpatients,7 though that study did not evaluate fall occurrences, and also not in a longitudinal manner. Therefore, we aimed at investigating whether polypharmacy was predictive of fall occurrences in a prospective fashion. For this purpose, we followed geriatric outpatients for up to 2 years, and assessed whether polypharmacy is a risk for fall occurrence, together with other risks.

The validity of two novel indices of fall tendency, the 22 items fall risk index8 and the 13 points simple screening test,3 which were used in our previous study, have been confirmed in community-dwelling elderly, but not in geriatric outpatients. Therefore, in the present investigation, the association of these two indices with falls was also evaluated to confirm their validity in geriatric outpatients in a longitudinal study.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure statement
  9. References

Patients

From 2006 to 2007, a total of 190 consecutive patients aged 65 years or older who were receiving treatment for chronic diseases, such as hypertension, dyslipidemia, diabetes and osteoporosis, who were seen every 2–4 weeks at the outpatient clinic of the Research Institute of Aging Science, Tokyo, were enrolled. All the patients were able to walk independently and their condition was stable. Patients who had acute illness or overt dementia were excluded. Anthropometric and medical information including past history of stroke, myocardial infarction, malignancy and prescribed drugs was obtained from each patient at baseline from the medical chart recorded by the physician in charge. However, 18 patients were excluded, because they were lost to follow up soon after enrolment and the medical information was not fully obtained. All prescribed drugs had not been changed in the included patients for at least 2 months before enrolment. The patients were followed up for 2 years.

Occurrence of falls

During the follow-up period, the patients and their family members responded to the annual questionnaire asking about the occurrence of falls within the past year. The questionnaire was repeated for 2 years.

Indices of fall tendency

After enrolment, the patients were examined for two indices to investigate the fall tendency. These were (i) a questionnaire of the 22 items portable fall risk index;8 and (ii) the 13 points simple screening test to assess the fall tendency.3

Ethical consideration

The present study was approved by the Institutional Review Board of the Research Institute of Aging Science. We obtained written consent from all participants and/or their guardians.

Data analysis and statistical methods

Values are expressed as mean ± standard deviation. In order to analyze the relationship between falls and comorbidity or drugs, variables were compared using Student's t-test or χ2-test as appropriate. Significant factors found in univariate analysis were included in multivariate logistic regression analysis to determine the association of falls with other variables. Receiver–operating curve (ROC) analysis was carried out to identify the optimal cut-off value of the number of drugs for predicting falls within 2 years. The value with the highest sum of sensitivity and specificity was used as the optimal cut-off value. Logistic regression analysis was carried out to assess the validity of the two indices of fall tendency, adjusted by age and sex. P-values <0.05 were considered statistically significant. Data were analyzed using JMP version 8.0.1 (SAS Institute, Cary, North Carolina, USA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure statement
  9. References

Baseline medical information and two indices of fall tendency were evaluated in 172 patients (Table 1). Drugs prescribed in less than 5% of the patients are not shown. Because only patients who were in a stable condition and were able to walk independently were included, patients with Parkinson's disease, severe paresis or painful arthralgia were not included. Calcium channel blockers prescribed in the present study were all long-acting agents, and the prescribed aspirin dosage was 100 mg in all cases. Only a few patients were receiving insulin therapy, sulfonylureas, angiotensin converting enzyme inhibitors, β-blockers, α-blockers, non-steroidal anti-inflammatory drugs or anticoagulants. No patients were taking neuroleptics or antiparkinsonian drugs.

Table 1.  Characteristics and univariate analysis of association with fallers and non-fallers within 2 years and risk factors
Total Non-fallers (n = 133)Fallers (n = 32)P-value (Fallers vs. Non-fallers)
  1. Values are expressed as mean ± SD (n = 165).

Age (years)77.0 ± 7.076.3 ± 6.980.0 ± 6.90.007
Body mass index (kg/cm2)22.7 ± 3.222.7 ± 3.322.7 ± 3.10.98
No. comorbid conditions1.9 ± 1.11.8 ± 1.12.3 ± 0.90.009
No. drugs3.2 ± 2.82.8 ± 2.74.9 ± 2.5<0.0001
Female (n = 122)72.9%78.1%0.66
Hypertension (n = 106)62.4%71.8%0.41
Dyslipidemia (n = 76)47.3%40.6%0.56
Diabetes (n = 23)12.8%18.8%0.40
Osteoporosis (n = 59)30.8%56.3%0.01
History of stroke (n = 6)2.3%9.4%0.09
History of myocardial infarction (n = 3)0.8%6.3%0.10
History of cancer (n = 8)5.3%3.1%0.99
Calcium channel blocker (n = 59)33.3%46.9%0.16
Angiotensin II receptor blocker (n = 56)33.3%37.5%0.68
Statin (n = 40)23.5%28.1%0.65
Aspirin (n = 31)19.0%24.1%0.61
Bisphosphonate (n = 9)4.6%9.4%0.38
H2-blocker (n = 9)3.8%12.1%0.80
Proton pump inhibitor (n = 11)5.3%12.1%0.23
Hypnotic (n = 31)16.7%28.1%0.14

After 1 year, all patients, except for one who died of congestive heart failure, were followed up (n = 171, follow-up rate 99.4%). Falls occurred in 22 patients. Only a higher age was associated with falls within 1 year on univariate analysis (non-fallers: 76.4 ± 6.8 years, fallers: 81.0 ± 6.9 years, P = 0.004).

After another year (2 years after enrolment), one patient had died of lung cancer, and five patients were lost to follow up. A total of 165 patients were evaluated (follow-up rate 95.9%), and 10 patients had fallen during the second year; thus a total of 32 patients had fallen within 2 years. As shown in Table 2, higher age, osteoporosis, number of comorbid conditions and number of drugs were significant factors associated with falls. To determine the association of falls with these significant factors, multivariate logistic regression analysis was carried out, and as shown in Table 2, the number of drugs was the only factor that was significantly associated with falls within 2 years.

Table 2.  Logistic regression analysis of association of falls within 2 years with age, sex, other significant factors found in univariate analysis, and polypharmacy
 Unadjusted odds ratio (95% CI)Adjusted odds ratio (95% CI)Adjusted odds ratio (95% CI)
  1. *P < 0.05, P < 0.005, P < 0.0005. CI, confidence interval.

Age (/1 year)1.08 (1.03–1.13)1.06 (0.99–1.13)1.06 (0.99–1.13)
Sex (male = 0, female = 1)1.39 (0.56–3.48)0.98 (0.29–3.23)0.75 (0.23–2.38)
Osteoporosis (n = 0, Y = 1)3.12 (1.43–6.84)2.76 (0.92–7.38)3.02 (0.96–6.15)
No. comorbid conditions (/disease)1.63 (1.14–2.32)*0.90 (0.55–1.47)0.99 (0.62–1.56)
No. drugs (/drug)1.29 (1.12–1.48)1.30 (1.08–1.57)*
Five or more drugs (n = 0, Y = 1)5.04 (2.25–11.3)4.50 (1.66–12.2)

As polypharmacy was assumed to be a risk for falls within 2 years, the cut-off of the number of the drugs was analyzed. Figure 1 shows the ROC curves to define the optimal cut-off point in relation to falls within 2 years: the area under the ROC was 0.731, and the optimal cut-off value of the number of drugs was five (sensitivity 0.576, specificity 0.788). Logistic regression analysis showed that taking five or more drugs was significantly associated with an increased risk of falls (odds ratio 4.5, 95% CI 1.7–12.2) after adjustment for age, sex, osteoporosis and number of comorbid conditions (Table 2).

image

Figure 1. Receiver–operating curves to define optimal cut-off value of number of drugs at baseline in relation to falls within 2 years. Area under the curve was 0.731, optimal cut-off value of the number of drugs was five (sensitivity = 57.6%, specificity = 78.8%).

Download figure to PowerPoint

Also, the association between falls and two indices of fall tendency was evaluated to confirm the validity of each index in geriatric outpatients. As both indices included the questionnaire asking whether patients were “taking five or more drugs,” the number of drugs was excluded from this analysis because of duplication in the statistical model. As shown in Table 3, the 22 items fall risk index showed a tendency towards an association with falls within 2 years, odds ratio 1.12 (95% CI 1.00–1.26; P = 0.05), whereas the 13 points screening test was significantly associated with falls after adjustment for age, sex and other factors significantly associated in the univariate analysis. Therefore, these indices are considered to be good predictors of falls in geriatric outpatients, as has been shown in community-dwelling elderly subjects.

Table 3.  Logistic regression analysis of association between 2-year fall occurrences with two indices of fall tendency; 22 items fall risk index and 13 points simple screening test
 Unadjusted odds ratio (95% CI)Adjusted odds ratio (95% CI)Adjusted odds ratio (95% CI)
  1. *P < 0.05, **P < 0.005, ***P < 0.0005. CI, confidence interval.

Age (/year)1.08 (1.03–1.15)**1.06 (0.99–1.13)1.06 (1.00–1.13)
Sex (male = 0, female = 1)1.39 (0.56–3.48)0.75 (0.23–2.43)0.79 (0.24–2.56)
Osteoporosis (n = 0, Y = 1)3.12 (1.43–6.84)**2.56 (0.96–6.82)2.61 (0.98–6.95)
No. comorbid conditions (/disease)1.63 (1.14–2.32)*1.24 (0.83–1.86)1.32 (0.88–1.97)
Fall risk index (/item)1.23 (1.11–1.37)***1.12 (1.00–1.26)
Simple screening test (/point)1.19 (1.06–1.33)**1.14 (1.01–1.29)*

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure statement
  9. References

The risk of falls has been assessed in community-dwelling elderly, and history of falls, physical ability and living environment were found to be predictors of falls. Also, in nursing home residents, cognitive function, gait disturbance and urinary incontinence are reported to be risk factors for falls,9,10 and length of stay, disease condition, surgical procedures and some specific drugs are reported to be risk factors in hospital inpatients.11,12

Nevertheless, the risks in geriatric outpatients have not been sufficiently assessed, although assessment of fall risk in geriatric outpatients is important; their medical conditions or drugs might cause falls, and drugs, such as antiplatelet agents or anticoagulants, might cause critical bleeding after a fall. Also, physicians could prevent falls in their patients by giving advice during regular consultations, if risk factors are identified.

In our previous cross-sectional study assessing geriatric outpatients, polypharmacy was significantly correlated with indices of fall tendency, and the present follow-up study of geriatric outpatients showed the impact of polypharmacy on falls within 2 years. Statistical analyses showed that polypharmacy was a risk factor for falls, independent of age, sex and comorbidity.

Besides polypharmacy, several medications and comorbid conditions have been reported as risks for falls.13–22 Among these, diabetes,5,6 insomnia,13 hypnotics,13–15 antiarrhythmics22 and antihypertensive agents14 were not significantly associated with fall risk in the present study. Just 11 patients (45.9% of diabetic patients) were prescribed hypoglycemic agents, such as a sulfonylurea (n = 8) or insulin (n = 3), and the relatively low rate of prescription of hypoglycemic agents might have affected our result. Neither hypnotics nor antihypertensives were associated with falls. This result might be a result of the small sample size. Anti-arrhythmics were taken by just three patients (digoxin: n = 2, class IA anti-arrhythmic drug: n = 1). Other drugs, such as major tranquillizers,14 antidepressants17,18 and antiparkinsonian agents,19,22 might increase fall risk; however, no patient used these drugs in the present study. In the present study, most of the patients were in a stable condition throughout the 2 years, though their drugs were changed gradually according to their medical conditions during the observation period. We only used the number of drugs at baseline for statistical analysis; however, the number of drugs increased from 3.2 ± 2.8 to 3.9 ± 3.0 during the 2 years. There were 17 patients whose number of drugs had been decreased, 70 patients not changed and 78 patients increased. The number of drugs after 2 years was also associated with falls (P < 0.0005). The optimal cut-off point for the number of drugs was again five (area under ROC curve 0.780, sensitivity 0.576, specificity 0.788). Furthermore, the changes in number of drugs were also associated with falls (P < 0.05), and the optimal cut-off point for the change in number of drugs was +1 (area under ROC curve 0.649, sensitivity 0.727, specificity 0.409). Consequently, polypharmacy, especially taking five or more drugs, should be considered a risk for falls.

There were several limitations of the present study. First, the falls were self-reported by the patients. Although all the patients had no overt dementia, they might have forgotten the incident of falling. We attempted to count the total fall occurrences in each patient; however, we could not differentiate the repeated falls in the second year from the fall occurrence in the first year. In fact, we asked 22 patients who reported falls in the first year about fall occurrence during the second year, but they did not accurately recall whether they experienced falls in the first or second year. Second, five patients were lost to follow up at 2 years for unknown reasons. The follow-up ratio was acceptable, although some of the patients might have fallen, have been no longer able to come to the clinic and moved to nursing homes. This might have slightly influenced the result. Also, the cause of falls in polypharmacy patients is not explained. Potentially inappropriate medications, which could cause adverse drug reactions, are usually seen in patients with polypharmacy, and falls might be the consequence of adverse drug reactions, such as dizziness, instability and light-headedness. Pathophysiological assessments and drug-reducing interventions are expected to elucidate the causal relationship.

Additionally, we showed that the 22-item fall risk index and its simple screening test were useful to predict falls in geriatric outpatients. Although both indices have been validated in community-dwelling elderly people, the present finding also showed their association with fall risk among geriatric outpatients. The difference of statistical significance between fall risk index and simple screening test might be a result of small sample size or the difference in the contribution of each item to total scores between the two indices. “Taking five or more drugs” accounts for only one item out of the 22-item fall risk index; in contrast, the same questionnaire accounts two points in the 13-point simple screening test. Because polypharmacy was a strong risk factor of falls in elderly outpatients in the present study, the proportion of polypharmacy in the scores might have caused the discrepancy. Taken together, it is likely that 13-point screening test was more suitable to our subjects who were taking several medicines.

In summary, the present study showed that geriatric outpatients with polypharmacy were at a high risk of falls, especially those receiving five or more drugs. Our finding might add new information for pharmacotherapy and geriatric research in elderly patients with chronic diseases. Intervention studies examining the effect of drug reduction for the prevention of falls are required in the future.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure statement
  9. References

We thank Ms Fumie Tanaka for her excellent technical assistance. This study was financially supported by grants from the Ministry of Health, Labour and Welfare of Japan (H21-Chouju-Ippan-005, H22-Chouju-Shitei-009).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure statement
  9. References