Incidence of Falls, Risk Factors for Falls, and Fall-Related Fractures in Individuals With Diabetes and a Prior Foot Ulcer

Individuals with diabetes and a history of a serious foot lesion were recruited for a clinical trial of footwear from two western Washington State health care organizations—the Department of Veterans Affairs Puget Sound Health Care System (VA) and Group Health Cooperative of Puget Sound (GHC)—between August 1997 and December 1998. Eligibility requirements included diagnosed diabetes, age 45–84 years, history of a foot ulcer, no foot deformities requiring a custom shoe, ability to walk one block and climb one flight of stairs daily, and willingness to consent to randomization and study footwear provisions. Patients were excluded from the study for a prior lower-extremity amputation of more than one digit, Charcot foot deformity, or requirement of custom shoes or nontraditional footwear for daily activities. Men were recruited from both organizations. Women came only from GHC because of the small number of female veterans. Outcome data were collected on foot ulcers, falls, and fall-related fractures.

Diabetes, demographic, health, foot, and functional status information and lower-limb sensory perception using the Semmes-Weinstein 5.07 monofilament were collected at baseline, 1 year, and 2 years. A history of six comorbid conditions (stroke, cardiovascular bypass surgery, chronic respiratory disease, heart failure, cancer, or depression) was obtained at baseline. A fall was defined as a person inadvertently ending up on the ground or at a lower level (5). Data on falls were collected through structured participant interviews at baseline, 4 weeks, and every 17 weeks thereafter for 2 years. Data collected included the circumstances of each fall, whether medical care was sought, and whether hospitalization was required. Medical records were abstracted for participants with falls requiring medical attention.

Frequency tables were constructed to describe demographic and health history characteristics of the population using SAS version 8.02 (6) and STATA version 6.0 (7). We analyzed falls by participant to determine the frequency of falls in those participants with falls. The incidence of falls in this cohort was calculated using standard person-time methods. Nonambulatory person-time was excluded. The χ² contingency tables were used to compare the proportions of participants never falling or ever falling. Multivariate logistic regression was used to investigate risk factors for at least one fall and for two or more falls. These results are reported using adjusted odds ratios and 95% CIs. Risk factors associated with fall-related fractures are reported using univariate odds ratios because of the small number of fractures.

There were 400 participants enrolled in the study: 187 men from the VA, and 122 men and 91 women from GHC. Table 1 shows that the average participant’s age was 62 years and average education was 14 years, and 67% of participants had been diagnosed with diabetes for over 6 years. Falls were significantly more common in white non-Hispanic participants and in individuals with one or more comorbid conditions. A higher percentage of participants with insensate feet (62%) reported falls compared with participants with sensate feet (52%), although this finding was not statistically significant. During the 2-year follow-up period, 252 participants (63%) reported 923 falls. Overall, participants reported that 17% of falls required medical care. However, in participants who reported more than one fall, 31% required medical care. The overall incidence of falls in this cohort was 1.25 falls/person-year (95% CI 1.17–1.33).

Table 2 shows the risk factors associated with falls. When we evaluated the risk factors for only one fall, we found that only the presence of one or more comorbid conditions increased the risk of falls. Comorbid conditions were more common in the VA population. Risk factors for multiple falls included BMI ≥30 kg/m², one or more comorbid conditions, and insensate feet. Female sex was not a significant risk factor for falls.

Table 3 shows risk factors for fractures from falls. There were 15 fall-related fractures reported by seven men (one with two fractures) and seven women. All seven men were ≥65 years old, whereas four of the seven women were ≥65 years old. Fractures locations were as follows: six ankle, three proximal humerus, one radius, one tibial plateau, one tibia and fibula, one hip, one rib, and one lumbar spine. Table 3 shows that fracture risk was significantly higher in women, in individuals who had diabetes for ≥25 years, and in study participants who wore their own footwear rather than special study shoes. The fracture rate was threefold higher in women than in men (40.2/1,000 vs. 12.4/1,000 person-years, respectively).

The frequency of falls among study participants was much higher than expected. After 1 year, 54% of our study cohort (ages 45–84 years) reported at least one fall. This result compares to the 33% of individuals aged ≥65 years with and without diabetes who fall each year, according to the National Center for Injury Prevention and Control at the Centers for Disease Control (8). The incidence for falls in our cohort was 104.2 falls per 1,000 person-months, which is 2.5-fold higher than the incidence rate of 41.4 falls per 1,000 person-months reported in an 18-month prospective study of 409 randomly selected community-dwelling individuals with and without diabetes aged ≥65 years residing in Montreal (9).

Although we expected that insensate feet (as a result of diabetic peripheral neuropathy) would be associated with an increased incidence for any fall, insensate feet significantly increased the risk only for two or more falls. Only the presence of a comorbid condition increased the overall risk of any fall. Comorbid conditions were more prevalent in the VA population. Having one fall appeared to be a random event, whereas two or more falls was not. When controlling for age, sex, diabetes duration, foot deformity, footwear, and smoking, only BMI ≥30 kg/m², comorbidity, and the presence of insensate feet increased the risk of multiple falls.

Overall, participants in this cohort sought medical care for 17% of reported total falls (153/923 falls). We observed 15 fractures in 14 participants (9%). The fracture risk was not increased in participants with insensate feet. In a study of 80 patients with type 1 diabetes, Cavanagh et al. (10) reported that individuals with diabetes and peripheral sensory neuropathy were 15 times more likely to report fall-related injuries than individuals without neuropathy.

There are several study limitations. The focus of most previously published work on incidence and risk factors for falls is on populations aged ≥60 years. This cohort of participants was between age 45 and 84 years at baseline. Few studies provide fall rates for people with and without diabetes for comparison. In addition, recall bias in some participants may have resulted in an underestimation of the number or seriousness of falls. We asked participants whether they sought medical care for falls, recorded their responses, and reviewed only those medical records. We did not uniformly record all minor injuries. Although participants may have different thresholds for seeking medical care, all participants had good access to medical care because of their affiliation with comprehensive health care systems. In addition, some individuals at high risk of serious falls may have been excluded from study participation because eligibility criteria included the ability to walk at least one block and climb one flight of stairs daily.

These results suggest that attention to falls in individuals with diabetes is needed. Clinicians need to be aware of fall risk factors in their patients. They also need to discuss with their at-risk patients the strategies to prevent fall-related injuries that can impair function and independence.

This research was supported by Rehabilitation Research and Development, the Epidemiology Research and Information Center, Department of Veterans Affairs; the National Institute of Diabetes and Digestive and Kidney Diseases, National Institute of Health; and the Centers for Disease Control and Prevention.

We thank Matthew Maciejewski for his comments on this manuscript and Jeff Rodenbaugh for programming assistance.