Article Text
Abstract
Objective: The authors used data from a population based prospective cohort study to determine if depressive symptoms predicted incidence of unintentional injury.
Methods: The Keokuk County Rural Health Study, based in Iowa, is a prospective cohort study of health status that includes injury outcomes. Depressive symptoms were measured using the 11-item Center for Epidemiologic Studies Depression Scale at the beginning of the study on 1493 participants. Quarterly follow up phone calls were made to measure injury incidence.
Results: 471 injuries were reported for an overall injury rate of 9.8 per 100 person-years. Crude injury rates were significantly higher for those with depressive symptoms (p = 0.0003). Those with depressive symptoms had a 41% increased risk for injury after controlling for antidepressant medication use, gender, prior injury, income, and sleepiness (RR = 1.41, 95% CI 1.10 to 1.80). Depressive symptoms remained a risk factor for injury regardless of current antidepressant medication use (no medication use, RR = 1.43, 95% CI 1.09 to 1.88; medication use, RR = 1.31, 95% CI 0.76 to 2.26).
Conclusions: Depressive symptoms were found to be risk factors for unintentional injury. Medical practitioners should consider talking about safety with their patients, especially those reporting symptoms of depression, and recognize that an increased risk for injury remains until the depressive symptoms subside.
- CES-D, Center for Epidemiologic Studies Depression Scale
- KCRHS, Keokuk County Rural Health Study
- depression
- rural population
- prospective studies
Statistics from Altmetric.com
Post-traumatic stress disorders, anxiety syndromes, and depression commonly follow events such as unintentional falls,1,2 motor vehicle crashes,3,4 workplace injuries,5 and traumatic brain injuries.6–8 While many studies have examined depression as an outcome of injury, depression is rarely examined as a risk factor for injury, even though mechanisms exist to support this hypothesis. Medications used to treat depression have been linked with increases in a variety of occupational injuries, as well as accidental falls in the elderly.9–12 Additionally, untreated depression has been correlated with symptoms that are risk factors for unintentional injury such as a lack of concentration and daytime drowsiness. The few studies that considered such psychological risk factors for injury have been cross sectional, making it difficult to determine which occurred first—the injury or the psychological symptoms.13–15
Depression affects approximately 121 million people worldwide and is among the world’s leading causes of disability.16 Furthermore, each year, injuries kill more than 5 million people worldwide and by 2020, injuries will be the third leading cause of death and disability worldwide.17 Although depression and injury remain global public health problems, both are treatable and preventable. Depression can easily be diagnosed and treated, while minor modifications such as seatbelts, helmets, or smoke alarms often have significant effects on injury risk and mortality.
Here, we analyze data from a population based prospective cohort study to determine if depressive symptoms are a predictor of subsequent injury. The Keokuk County Rural Health Study provides a unique opportunity to examine depression, both treated and untreated, as a risk factor for unintentional injury in a longitudinal manner. To the best of our knowledge, this is the first prospective analysis to examine depression as a risk factor for unintentional injury in a large cohort.
METHODS
Study design and population
Our study is nested in the Keokuk County Rural Health Study (KCRHS). The KCRHS is a population based, 20 year prospective cohort study of health status and injury of a large stratified random sample of households located in a single rural Iowa county. The sample was randomly selected from a compiled list of all county households.18 Of the 3749 households invited, 273 were not eligible, 1204 could not be reached by phone, and 1262 declined to participate. One thousand two households, with 1582 individual participants older than 18 years of age, participated in the first round of data collection.18 This analysis used the 1493 participants, aged 18 years or older, participating in the second round of data collection. For a detailed description of the methods see Merchant et al.18 The University of Iowa institutional review board approved the study protocol and data collection instruments.
Each adult participant received a medical and mental health screening and was interviewed in the clinic by trained staff members to investigate injury and disease incidence in relation to occupational, agricultural, and other environmental exposures. The first round of cohort data collection occurred between June 1994 and February 1998, with the second round between March 1999 and April 2004. This analysis used data collected during the second round. In addition to data collected during the clinic interviews, families were followed prospectively using quarterly phone calls to collect information about injury incidence.
Outcome
Injuries were defined as traumatic events that restricted normal activities for at least four hours, that caused any loss of awareness or memory for any length of time, or that required professional medical care. After the clinic interview, participating households were contacted an average of every three months by a trained interviewer. One adult was asked to recall the accidents and injuries experienced by each family member. Out of the 471 injuries included in this analysis, 127 (27%) of the phone respondents were a family member of the injured person and not the injured person themselves. Information collected on the injury included: date of injury, injury location, body part(s) affected, type of injury, lost school/work days, activity at time of injury, medical care received, and recovery progress. Injuries included in this analysis were those occurring any time after the second round clinic interview until the end of the follow up period in June 2004.
Primary risk factor
The primary risk factor was the presence of depressive symptoms measured by the abbreviated 11-item Center for Epidemiologic Studies Depression Scale (CES-D scale) at the time of the clinic interview. Participants were considered to have depressive symptoms if their total score was greater than or equal to eight.19,20 The 11-item CES-D scale is a shortened version of the original CES-D; a self-report index of depressive symptoms with high internal consistency reliability.19 The abbreviated CES-D tap the same dimensions as the original and little reliability is sacrificed.19
Information on antidepressant medication use was determined from prescription medications that participants were asked to bring to their clinic interview. Medications included in this analysis were selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressants (TCAs), heterocyclic antidepressants, and monoamine oxidase inhibitors (MAOIs). Anti-anxiety medications were also included as they have side effects similar to those brought on by medications used to treat depression such as nervousness, lightheadedness, excitement, and insomnia. The presence of any one of these medications was considered medication use and coded as “yes” or “no”. Length of time spent on the medication was not measured throughout the cohort study.
Other risk factors
We examined the following potential confounders chosen through prior research findings as being related to depression and injury risk: major demographics, sleepiness, prior injury, and potential alcohol problems. At the clinic interview, participants were asked a set of sleep questions that included frequency of snoring, frequency of waking up in the night, frequency of waking too early in the morning, hours of sleep per night, feelings of tiredness during the day, and difficulty getting to sleep. Number of sleeping hours was categorized into 7.5 hours or less and more than 7.5 hours a night. Participants were asked if they had an injury, as defined above, in the 12 months before the clinic interview. Prior injury was coded as “yes” or “no”. Potential alcohol problems were measured using the CAGE questionnaire.21 Adults who answered “yes” to two or more of these questions were classified as having potential alcohol problems.21
Statistical analysis
Injury rates per person-year were calculated and compared between different strata of demographic and potential confounding variables. Frequencies and cross tabulations were used to describe the potential confounders, injuries, and depressive symptoms. Pearson χ2 tests were used for categorical data, t test for continuous variables, and Wilcoxon non-parametric tests for continuous variables with a non-normal distribution.
Rate ratios (RR) and 95% confidence intervals (95% CI) were used to describe associations between depressive symptoms and subsequent injury adjusting simultaneously for the potential confounding variables of age, gender, marital status, education level, sleepiness, and potential alcohol problems. The effects of covariates in the multivariate model were tested by likelihood ratio tests. RRs were derived using the multivariate Poisson regression model in SAS with injury count as the outcome variable, adjusted for follow up time (Version 9.2, SAS Institute Inc, Cary, NC, USA). Person-time was accrued for each participant from the date of their clinic interview until the occurrence of an injury or the end of the follow up on 30 June 2004. Since individuals were nested within families, clustering was accounted for by using the generalized estimation equation method through proc genmod in SAS.22
RESULTS
KCRHS cohort demographics and statistics
The cohort used for analysis included 838 females (56%) with a mean age of 55.1 years (SD 16.0) and 655 males (44%) with a mean age of 55.9 years (SD 16.4). Seventy eight percent of our sample were married and 92% had at least a high school education. Two hundred and fifty nine people had depressive symptoms according to the CES-D (17%) and 127 participants (9%) were on at least one antidepressant medication at the time of the clinic interview.
Injury rates
Four hundred and seventy one injuries were reported over a mean follow up time of 3.2 years (SD 1.4 years) for an overall injury rate of 9.8 per 100 person-years. Crude injury rates were significantly higher among those with depressive symptoms, those on an antidepressant medication, males, those with an injury before the clinic interview, those earning less than $20,000 a year, and those who slept less than seven hours a night (table 1). Also, there was a trend indicating those with more education and those with CAGE scores indicative of alcohol problems had higher injury rates than their counterparts, but these trends were not statistically significant.
Bivariate and multivariate analysis
Table 2 presents the bivariate rate ratios and their confidence intervals for the association between the various risk factors and injury. Those with depressive symptoms had a significantly increased risk for injury (RR = 1.54, 95% CI 1.22 to 1.95). Being on an antidepressant medication was also associated with an increased risk for injury (RR = 1.61, 95% CI 1.19 to 2.18). Other significant risk factors for injury in the bivariate analysis included male gender (RR = 1.28, 95% CI 1.04 to 1.51), not being married (RR = 1.31, 95% CI 1.03 to 1.66), making less than $20,000 a year (RR = 1.46, 95% CI 1.09 to 1.96), having a prior injury (RR = 1.43, 95% CI 1.14 to 1.79), and getting less than seven hours of sleep (RR = 1.29, 95% CI 1.05 to 1.60).
The final model was chosen by entering all variables significant in the bivariate analysis in a full model and using likelihood ratio tests to choose the best overall model. The correlation between depressive symptoms and antidepressant medication use was measured and found to be only moderate (Φ = 0.17); therefore, both variables were included in the final model. Depressive symptoms, as measured by the CES-D, increased the risk of injury 41% after controlling for antidepressant medication use, gender, prior injury, income, and sleepiness (RR = 1.41, 95% CI 1.10 to 1.80). Additionally, males, those on an antidepressant, those with a prior injury, those making less than $20,000 a year, and those getting less than seven hours of sleep a night were at an increased risk for injury.
To examine whether antidepressant medication modified the association between depressive symptoms and injury, the full model containing an interaction term between antidepressant medication (yes/no) and depressive symptoms (yes/no) was tested against a reduced model without this interaction term by means of likelihood ratio test. The interaction term was not significant. A variety of other interactions such as depression and gender, gender and antidepressant medication, and income and depression were tested and none were found to be significant. The full model was then stratified by antidepressant medication use (table 3). The presence of depressive symptoms remained a risk factor for injury regardless of the use of antidepressant medication. The RR for those on antidepressant medications did not reach statistical significance because of small sample sizes, however the trend was apparent (RR = 1.31, 95% CI 0.76 to 2.26).
Injury characteristics
There were minor differences in the injuries experienced by those with depressive symptoms and those without (table 4). Four hundred and sixty nine of the injuries had corresponding injury data; 357 (76%) occurred to those without depressive symptoms and 112 (24%) occurred to those with depressive symptoms. Those without depressive symptoms were slightly more likely to have had the injury occur in a work area than those with depressive symptoms (p = 0.054). Those with depressive symptoms were significantly less likely to see themselves in complete recovery after the injury event than those without depressive symptoms (p = 0.01). There were no significant differences in the body part injured, the cause of injury, or the average missed days of work or school; however, those without depressive symptoms had a higher average missed days or work or school and were less likely to have had experienced a fall.
DISCUSSION
The Keokuk County Rural Health Study provided a unique opportunity to prospectively examine depressive symptoms as a risk factor for injury while controlling for a variety of known confounders, including antidepressant medication use and prior injury. Previous cross sectional research has pointed towards an association between injury and depression, but was unable to establish the direction of causality. Our prospective study found depression to be a risk factor for subsequent injury after controlling for gender, prior injury, income, and potential alcohol problems as measured by the CAGE questionnaire.
Sprince et al found that current depressive symptoms were associated with a farm work related injury in the prior year in a multivariate analysis (OR = 1.65, 95% CI 1.06 to 2.56).15 Peele used a case control study to examine how psychological factors influence occupational injury.23 She found that women who had depression symptoms were 6.2 times more likely to have had an occupational injury than those without depressive symptoms.23 Finally, Poole et al used a case control study to compare psychosocial factors from a structured interview between victims of intentional trauma, unintentional trauma, and patients undergoing elective surgery at a trauma center.14 Logistic regression analysis identified younger age, lower intelligence, antisocial personality, mental retardation, depression, and low income to be associated with an increased risk for trauma. Our prospective study builds upon this prior work by establishing depressive symptoms, measured at baseline, as a risk factor for subsequent injury.
While few studies have considered depression as a risk factor for injury, numerous studies have examined antidepressant medication use, specifically for falls in the elderly. All of the large, landmark studies in this area show antidepressant medication use to be a risk factor for falls; however, while most studies control for dementia, cognitive functioning, or functional ability, none control for depression or depressive symptoms.24–28 While this body of research is extensive, it has not been clarified whether it is the psychotropic effects of the drugs, the symptoms of the depressive state, or an interaction between the two that puts an elderly person at an increased risk for a fall. This study demonstrates that depressive symptoms, in and of themselves, are a risk factor for injury, regardless of the use of antidepressants.
Our study has several limitations. Information on the injuries was obtained from self-report without validation with medical records. This raises the possibility of differential recall bias since it is possible that depressed and non-depressed people recall and perceive health related events differently.29 However, as respondents were reporting on tangible, acute events, we believe this bias to be minimized. Also, a single informant provided the data on the injuries for all family members and they may not have been aware of an injury, or know of the specific details. This occurred in 127 of the 471 injuries (27%). This may lower the total number of injuries reported for the cohort.
Finally, it is possible that depressive symptoms measured at the clinic interview would not persist into the follow up period. We re-analyzed the data and only included injuries occurring 24 months after the clinic interview and risk estimates were not affected (data not shown). Also, current research demonstrates that mild depression can be chronic and recurrent for as long as six years in many people.30
Our study also has unique strengths. To our knowledge, it is the only study examining depression as a risk factor for injury in a prospective and multivariate manner. This is a stable cohort whose members were comfortable with the study protocol and much effort was made to collect injury information by following participants over a long time period. In addition to using a reliable and valid screening tool to measure depressive symptoms, we were able to examine and control for the use of antidepressant medications.
Even though this study was located in a rural area of the United States, results from this study are generalizable to urban areas, as well as to areas outside of the US. Although urban and rural people may experience different types of injuries, we have no reason to believe that people in urban areas or in areas outside of the US experience depression differently from rural people.
This research suggests that a multifaceted approach is necessary when considering the relation between injury and mental health. We believe that depression and depressive symptoms have a unique effect on injury risk, independent from that of the risk associated with antidepressant medication use. Further work that would allow for the examination of subclasses of antidepressants, degrees of depressive symptoms, as well as interactions between the levels of depression and types of treatments, would help to further decipher these findings. These results should be validated in other populations and their causal mechanisms investigated. Medical practitioners should consider talking about safety with their patients, especially those reporting symptoms of depression, and recognize that an increased risk for injury remains until the depressive symptoms subside.
Key points
-
Cross sectional studies have demonstrated an association between injury and depression, but this is the first study to indicate that depressive symptoms are a risk factor for injury in a prospective manner.
-
This study found depressive symptoms increased one’s risk for subsequent injury 41% after controlling for antidepressant medication use, gender, prior injury, income, and sleepiness.
-
Depressive symptoms remained a risk factor for injury when stratified by antidepressant medication use, although not reaching statistical significance due to small sample sizes.
-
In this study, depressive symptoms were a distinct risk factor for injury, independent of antidepressant medication use.
Acknowledgments
Support for this project was provided by the CDC/NCIPC funded University of Iowa Injury Prevention Research Center (CCR 703640), the CDC/NIOSH funded Heartland Center for Occupational Health and Safety Occupational Injury Prevention Program (T42/CCT717547), and the CDC/NIOSH funded Great Plains Center for Agricultural Health (U07/CCU706145), all housed in The University of Iowa Department of Occupational and Environmental Health. We appreciate support from Drs James Merchant, Kevin Kelly, and other researchers involved in the Keokuk County Rural Health Study.
REFERENCES
Footnotes
-
Competing interests: none.