An active research agenda exists with a focus on prevention [2]. Cognitive impairment, low body mass index and certain medications such as benzodiazepines have been consistently associated with severe injuries from falls [5,9-12]. Data on the proportion of fall-related injuries attributable to each of these factors is sparse and it is likely that their relative contribution varies from one setting to another.

Most studies of risk factors for fractures due to falls have been carried out in developed countries, although the size of the elderly population is increasing fast in middle income countries. In Brazil the proportion of people over 60 doubled from 4.1% in 1940 to 8.6% in the year 2000, and it is expected to reach 14% in 2025 [13-15]. Little is known about the frequency, circumstances, risk factors and consequences of falls in Brazil. Perracini & Ramos [16] examined risk factors for any kind of fall in a cohort of elderly community residents and Rozenfeld et al [17] did a cross-sectional investigation in a recreational facility for the elderly. Coutinho & Silva [18] carried out a case-control study with hospital controls investigating the association between drugs used in the previous 24 hours and the risk of severe fracture after falling.

Seventy percent of the falls resulting in severe fracture occurred between 6:00 am and 6:00 pm, with a similar proportion in the morning and in the afternoon. Most falls took place at home (67%), and this proportion increased with age. The most commonly fractured bone was the femur (72%) followed by arm/forearm (19%). Two cases had vertebral fracture (2.7%) and eleven (4.4%) had more than one bone fractured. Ninety nine per cent of the cases had to undergo surgical procedures.

Health related factors with a level of significance less than 0.25 in their univariate association with fracture (matched analysis) are presented in table ​table2.2. Increased odds ratios were observed for low BMI, low blood pressure, dizziness, diabetes, cognitive impairment, history of stroke, lack of urine control, poor vision, limit in carrying activities of daily living (ADL), fall in the previous 12 months, use of antidepressants, benzodiazepines, muscle relaxants and cerebral vasodilators while reduced odds ratios were observed for poor health status, regular use of alcohol, calcium supplement and calcium channel blockers. Most benzodiazepines were long acting, and the most frequently prescribed was bromazepan. Almost all prescribed muscle relaxants were carisoprodol.

Osteoporosis, Parkinson disease, epilepsy, high blood pressure, use of angiotensin converting enzyme (ACE) inhibitors, antihistamines, analgesics, antiacids, alpha and beta-adrenergic blockers, nitrates, non steroidal anti-inflamatory drugs, digoxin, calcium and vitamin D supplements did not reached the pre-defined level of 0.25 significance [see Additional file 2].

The average total number of drugs were 2.2 (sd = 1.47) in cases and 2.1 (sd = 1.48) among the controls. The difference was not statistically significant (p = 0.43).

When variables presented in table ​table22 were entered in a multivariate conditional logistic model, diabetes, high blood pressure, rheumatism, poor vision and use of diuretics had significance levels over 0.05 and were dropped from the multivariate model. Table ​Table33 presents the final model. Body mass index equal or less than 20 kg/m2, cognitive impairment, previous stroke and lack of urine control were associated with increased incidence of severe fall related fractures while use of alcohol at least once a week was associated with reduced incidence. Concerning the use of drugs in the previous 24 hours, benzodiazepines and muscle relaxants were related to an increased risk of severe fractures while calcium channel blockers were associated with a reduced risk. The highest odds ratio (approximately 5 fold) was observed for use of muscle relaxants and history of stroke, although confidence intervals were large. No effect modification was observed for the variables included in the final model.

Our study in a middle-income country setting agrees with some previously reported associations with fracture due to fall observed in developed countries: low body index mass [9,22,23], cognitive impairment [5,22], stroke [9,10], lack of urine control [24-26].

In the present study regular users of alcohol ("at least once a week") were at a reduced risk of severe fractures. Findings for alcohol have not been consistent. Peel et al [27] found a reduced risk of fall-related fracture for moderate alcohol intake while other authors have found the reverse: higher risk of falls leading to fracture associated with higher use of alcohol [10,28]. "J" shape patterns in alcohol use have been observed for outcomes such as cardiovascular disease, with non drinkers or those with very low intakes having higher risks compared to those with mild or moderate drinking patterns. In our study, those who reported using alcohol at least once a week rarely made use of alcohol more than twice a week. It is possible that people drinking at least once a week were healthier than those not drinking. Although we attempted to control for this by the inclusion of other health related variables in the model, we cannot exclude the possibility of residual confounding.

As in our investigation, previous studies have identified an effect of benzodiazepines on the risk of fall-related fractures [12]. Hartikainen et al [29] carried out a systematic review of 29 studies that reported the association between the use of medicines and the risk of falls or fall-related fractures among people aged 60 or more. Nine of them were case-control studies matched by sex and age like our investigation. The authors concluded that central nervous system drugs, mainly psychotropic drugs, were associated with an increased risk of these accidents,

Many drugs commonly used by elderly people have not been systematically studied as risk factor for falls [29]. An important and novel result from our study was the association between the use of muscle relaxants in the last 24 hours and severe fracture due to falling. The odds ratio was very high (OR = 4.42) although the 95% confidence interval was wide (1.02–19.21). To the best of our knowledge, there is only one study that reported this empirical association among the elderly. French et al [30] used database information to investigate the relationship between registered primary diagnosis of fracture and previous use of some drugs. The authors found that those registered with fracture were prescribed muscle relaxants 1.4 times more than controls (those with non-specific chest pain). This value is much lower than the one we found, but it is difficult to compare these findings as the study designs were quite different.

That muscle relaxants can cause falls is biologically plausible: these drugs are recognized to cause weakness, drowsiness, sedation and anticholinergic effects [31]. Data on the use of muscle relaxants by elderly people, especially for extended periods, are limited but such studies that have been done reported usage by: 3% of the 60 and over population in Rio de Janeiro-Brazil [32], 0.77% of the 60 and over population in the USA [33], and 1.2% of the 75 and over age group in Finland [34]. Although muscle relaxants are recommended for short-term treatment of back pain, Dillon et al [33] reported a mean length of use of 2.1 years in the USA; 44.5% of users referred use for more than a year. Although it is generally acknowledged that the use of muscle relaxant may be inappropriate and hazardous in the elderly [31,35], the figures quoted above show that their use and their long term use remains a problem. It is likely that usage figures will be higher in places where there is easy access to medications over the counter, commonly in low and medium income countries such as Brazil. The 2002 criteria for potentially inappropriate medication use in older adults [31] does not mention explicitly the risk of falling and suffering a fracture in its evaluation of miorelaxants; the only group of drugs for which concern with falls is mentioned is long acting benzodiazepines

In contrast to previous studies we did not find a significant association between visual impairment [9,16,36,37] and diabetes [28,38] with fall related fracture. We did not measure visual acuity but relied on self report and there may have been under-reporting leading to dilution of effect. In the case of diabetes, finding an association with falling may be influenced by the proportion of those with neurological and foot problems. Ottenbacher et al [38] found that the association between diabetes and hip fracture particularly for those taking insulin.

Our study showed an unexpected inverse association between the use of calcium channel blockers (CCB) and the occurrence of severe fall related fracture. Two systematic reviews [29,39] did not find any association between these variables. We cannot exclude the possibility that this finding was due to residual confounding of self-reported health status. CCB and angiotensin converting enzyme inhibitors (ACEI) were the most reported antihypertensives. Although the proportion of controls taking CCB and ACE was the same (16%), the average total number of drugs referred by the first group was 3.0 while by the second group it was 3.5 (p = 0.02). This suggests that users of CCB could be healthier than those to which ACEI were prescribed.

The study had some limitations. Most variables were self reported and, in some of the interviews, information was provided or added by relatives that were in the hospital (for cases) or at home (for controls). This could lead to an unknown degree of misclassification of exposures. Moreover, cognitive impairment was evaluated after the fall, and we cannot be sure about the influence of the accident on mental state.