Abstract
Objective
To evaluate the prevalence, correlates and subgroups at highest risk for suicidal ideation among adults with arthritis.
Methods
We used data on U.S. adults with arthritis, aged ≥40, participating in the 2007–2008 NHANES survey. Suicidal ideation was assessed by item 9 of the Patient Health Questionnaire-9 (PHQ-9). Socio-demographic factors, health behaviors and comorbid conditions were examined as potential correlates. Depression was measured by the PHQ-8 score (range 1–24). We used random forests to identify subgroups at highest risk for suicidal ideation. To determine if any correlates were unique to arthritis, we compared results to those for persons with diabetes mellitus and cancer.
Results
The prevalence (± standard error) of suicidal ideation was 5.6% ± 0.8% among persons with arthritis and 2.4% ± 0.4% among those without. The most important correlates for suicidal ideation in adults with arthritis were depression, anxiety, duration of arthritis, age, income/poverty ratio, number of close friends, pain, alcohol, excessive daytime sleepiness and comorbidities. Eleven of 16 most important contributors for suicidal ideation among adults with arthritis were also important for people with diabetes and cancer. Among persons with arthritis, subgroups at highest risk for suicidal ideation were those with PHQ-8 between 18 and 24 and less than 4.5 years of arthritis (96.5%), and those with PHQ-8 between 7 and 17, ≥1.24 days of binges/month and either income ≥$45,000/year (85.4%) or income <$45,000/year and >3 comorbidities (70.8%).
Conclusion
Depression, short duration of arthritis, binge drinking, income, and >3 comorbidities identified subgroups of adults with arthritis at greatest risk for suicidal ideation.
Keywords: Suicidal ideation, arthritis, correlates, depression
Suicide is one of the highest public health priorities worldwide (1). The World Health Organization objectives for suicide prevention emphasize identification of high-risk groups (2). Suicidal ideation, referring to wishes that one’s life would end or thoughts of harming or killing oneself, represents an important phase in the suicidal process and often precedes suicidal attempts or completed suicide (3, 4). Patients with chronic medical illnesses and especially those suffering from pain are more likely to report suicidal thoughts (5, 6). As a chronic condition frequently associated with pain, arthritis may increase the risk of suicidal ideation. Coping with the effects of arthritis on a daily basis can have detrimental consequences for the mental health of those affected.
Little is known about the occurrence of suicide ideation among persons with arthritis (7, 8). One study showed that almost 11% of outpatients with rheumatoid arthritis (RA) experienced suicidal thoughts at the time of the study (7). However, the sample was small and was focused on RA, which represents only part of the arthritis population. Another study using a survey-based sample showed that suicidal ideation was approximately twice as common among adults with arthritis than those without (8). However, the association between arthritis and suicidal ideation in the general population, and whether arthritis differs from other chronic disorders in this association, remains poorly characterized. Most importantly, the subgroups of adults with arthritis at greatest risk for suicidal thoughts are not known. Suicidal ideation is the result of a complex set of interactions between predisposing factors, buffers, and more acute situational events. Focus on subgroups at high risk can help improve suicide prevention and intervention strategies in persons with arthritis.
To address these questions, we used data from the National Health and Nutrition Examination Survey (NHANES) for 2007–2008, the largest U.S. survey that included questions about both suicidal thoughts and arthritis and related conditions such as pain. The purpose of this study was to determine the prevalence of suicidal ideation among adults with arthritis, and to identify subgroups of individuals with arthritis at greatest risk for suicidal thoughts. Additionally, to evaluate if the most important correlates among persons with arthritis differ from those of people with other chronic diseases,.we compared results with those of people with diabetes mellitus and cancer.
Methods
Data source
We examined publicly-available data from NHANES 2007–2008, a cross-sectional population-based survey of the U.S noninstitutionalized civilian population (9). Participants underwent a home interview that included information on demographic characteristics, health behaviors, medical conditions, depression, and social situation. Depression and suicidal thoughts were assessed in participants aged 20 years or older who participated in the examination component. Because measures of social support were only obtained in those aged 40 or older, and because most persons with arthritis were also older, we limited our analyses to subjects aged 40 or older. The sample included 3863 persons (response rate 69.1 %). Of these, 7 persons had missing data on arthritis, and 371 had missing data on the entire PHQ-9 questionnaire, either because they had proxy respondents, ran out of time, or refused or didn’t provide an answer.
Subjects with Arthritis, Diabetes, or Cancer
We selected subjects with arthritis based on self-reported physician-diagnosed arthritis, as ascertained by the question “Has a doctor or other health professional ever told you that you had arthritis?”. Self reports of arthritis are considered to have sufficient validity for surveillance purposes (10). For comparison, we also examined two other representative chronic diseases, diabetes mellitus and cancer (excluding non-melanoma skin cancers), which were ascertained through questions similar to that above. Since a recent diagnosis of cancer is more likely to be associated with depression or suicidal thoughts (11, 12), we limited the analysis of this group to those diagnosed in the 2 years preceding the survey.
Outcome
We used suicidal ideation, assessed by item 9 of the Patient Health Questionnaire-9 (PHQ-9), as the outcome variable. The PHQ-9 is a 9-item screening instrument with high reliability and validity in the general population that evaluates the frequency with which depressive symptoms are present over the prior 2 weeks (13). Item 9 asks “Over the last two weeks how often have you been bothered by the following problem: thoughts that you would be better off dead, or of hurting yourself in someway?” Possible responses were: “not at all”, “several days”, “more than half the days” or “nearly every day.” For analysis, we categorized the responses as not present (0) or present at any frequency (1).
Item 9 of the PHQ-9 has been used by many studies assessing suicidal ideation, however, it does not ask specifically about thoughts of attempting suicide (14, 15). Use of a more specific item, such as the suicidal intent item of the Beck Depression Inventory (16), is more likely to lead to non-response.
Correlates
Socio-demographic factors
Based on a review of the literature, the following socio-demographic factors were examined as potentially predisposing to suicidal ideation: age (continuous variable), gender, ethnic origin (White, Black, Hispanic, other), nativity (U.S. versus other), education level (less than high school diploma, high school diploma, some college without degree, college degree or higher), marital status (never married, married, divorced/separated, widowed), social support, annual household income, ratio of income to the federal poverty level (continuous variable), food security, presence of health insurance, presence of a usual source of healthcare, and mental healthcare visit in the past year.
Food security was used as an additional measure of poverty and evaluated by a validated 10 item scale (17). We examined three items of social support: emotional support (count on anyone to provide you with emotional support), instrumental support (count on anyone to provide you with financial support) and social network (number of close friends or relatives). Responses to the first two items were coded as present or absent.
Health behaviors and comorbid conditions
Smoking, alcohol use and illicit drugs use are among the health behaviors more frequently associated with suicidal thoughts (1). Smoking status was classified as never, former or current smoker. For current smokers, the number of cigarettes per day was assessed. Alcohol use was evaluated by two questions: days of drinking per month (“In the past 12 months, how many days per month did you drink?”), and days of binges per month (“How many days per month did you have 5 or more drinks in a single day?”). Illicit drug use was assessed by two variables examining current use of marijuana and current use of other illicit drugs (cocaine, heroin or methamphetamine).
We included as comorbid conditions heart disease (coronary disease, myocardial infarction, angina or chronic heart failure), stroke, hypertension, pulmonary disease (chronic bronchitis, emphysema or asthma), liver disease, diabetes mellitus and cancer. The number of comorbid conditions (0–7) was also evaluated.
Depression, anxiety and sleep disturbances were also included because they have been associated with suicidal ideation (1). Depression was evaluated by the PHQ-8 score (possible range 0–24), removing the question pertaining to suicidal ideation from the PHQ-9 (18). A score of 10 or higher was used for the diagnosis of depression. Anxiety was assessed by a question on “days felt worried, tense or anxious during the past month”. We used 3 measures of sleep disturbance (19): physician-diagnosed insomnia, short sleep duration (< 6 hours of sleep per night), and excessive daytime sleepiness (never; rarely = 1 time a month; sometimes = 2–4 times a month; often = 5–15 times a month; almost always = 16–30 times a month).
Pain was assessed by the following question: “During the past 30 days how many days did pain make it hard for you to do your usual activities?” Physical disability was assessed by questions examining limitations in four activities of daily living (ADL): getting in or out of bed, dressing, eating and walking, and one question addressing the use of special equipment because of a health problem. Responses to questions on limitations in ADL were coded as no difficulty = 1, some difficulty = 2, much difficulty = 3, unable to do = 4. We also constructed a dichotomous measure of “unable to do” at least one of four ADLs.
Statistical analysis
For descriptive analyses, we used methods to account for the complex sampling strategy of NHANES 2007–2008. We also used sample weights to adjust standard errors for selection and non-response, and to provide nationally representative results from the surveyed samples. Analyses were based on samples of 1545 persons with arthritis, 672 with diabetes, and 127 with cancer diagnosed in the preceding two years. Persons with missing data were included as a sub-population so their influence on standard errors could also be captured. Descriptive analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC).
We used random forests to identify and validate the most influential correlates for suicidal ideation. A random forest is an ensemble classification method that combines the results of multiple classification trees (20). A classification tree is a non-parametric, non-model based, hierarchical classification procedure that uses recursive partitioning to identify the variables that best subset patients into groups with higher and lower prevalence of the outcome. The classification tree procedure first identifies the independent variable (i.e. the predisposing variable noted above) that best separates those with high from those with lower prevalence of suicidal ideation. The procedure is then repeated for each resulting subgroup until subgroups of sufficient purity are obtained. Variables higher in the tree are more influential that those lower in the tree. Because subgroups are sequentially derived, the importance of a single variable in correctly classifying a group of people is conditional on earlier variables in the tree. In this way, classification trees naturally identify interactions among correlates.
Random forests build multiple classification trees, using a random sample of subjects (bootstrap sample with replacement) and a random subset of predisposing variables eligible for consideration at each node of the tree. The accuracy of each tree is then determined by whether the tree correctly predicts the outcome for the subjects omitted from the development of the tree, and provides validation. The proportion misclassified, averaged over all the trees, is an unbiased estimate of classification error. The relative predictive importance of each variable is determined by the difference in classification between the original tree and a tree in which the values for one independent variable were intentionally jumbled (to remove its predictive power). A large difference in classification indicates the variable was an important predictor, while a small difference indicates the variable was not very important for classification. Variable importance, averaged over all trees, provides a relative ranking of correlates in their ability to classify subjects correctly with respect to suicidal ideation.
In a complex process like suicidal ideation, we chose to use classification trees to identify important, interacting factors that would identify subgroups of persons at increased risk, rather than using regression-based methods to identify whether individual predictors were independently associated with suicidal ideation. Identification of subgroups at high risk emphasizes clinical applicability of the results, whereas identification of independent predictors is focused on etiology. To determine if any correlates in persons with arthritis were shared with persons with other chronic diseases, we ran random forests separately for persons with arthritis, diabetes, and cancer. In each group, results were based on forests of 500 trees each. Analyses were based on weighted samples to accommodate the survey design, and were done using R software (version 2.11.1) (21) and a fast parallelized implementation of the random forest algorithm called Random Jungle (version 1.2.362, http://www.randomjungle.org) (22).
Results
Descriptive data of the sample
The mean (± standard error of the mean) age of persons with arthritis was 61.5 ±0.4 years (Table 1). Those with arthritis were more commonly women (60.4 %), whites (78.9%), and had attended at least some college (26.7%). Suicidal ideation was present in 5.6% (standard error ± 0.8%) of people with arthritis. Similar prevalence was found among those with diabetes (6.8% ± 1.4%) and cancer (5.1% ± 2.3%). The prevalence among those without arthritis was 2.4% ± 0.4%. Socio-demographic characteristics, health behaviors and comorbid conditions were generally similar among those with arthritis, diabetes, and cancer, with the exception of higher proportions of men in the diabetes and cancer groups (Table 1).
Table 1.
Characteristics of persons age 40 or older with arthritis, diabetes and cancer participated in the National Health and Nutrition Examination Survey for 2007–2008
| Arthritis | Diabetes | Cancer within 2 years |
|
|---|---|---|---|
| Suicidal ideation (%) | 5.6 | 6.8 | 5.1 |
|
Demographic and socio-economic factors |
|||
| Age, years(± SEM) | 61.5 (0.4) | 62.3 (0.2) | 68.4 (1.0) |
| Male (%) | 39.6 | 47.8 | 57.9 |
| Ethnicity (%) | |||
| White | 78.9 | 64.1 | 86 |
| Black | 10.1 | 18.2 | 6.8 |
| Hispanic | 7.1 | 11.9 | 3.5 |
| Other | 3.9 | 5.8 | 3.7 |
| Born in U.S. (%) | 91.7 | 86 | 92.9 |
| Education level (%) | |||
| Less than high school | 24.7 | 33.5 | 21.1 |
| High school graduate | 28 | 27.9 | 25.5 |
| Some college | 26.7 | 24 | 31.6 |
| College graduate | 20.6 | 14.6 | 21.8 |
| Marital status (%) | |||
| Single | 5.1 | 6 | 1.9 |
| Married | 64.8 | 59.7 | 69 |
| Divorced/separated | 15.2 | 18.8 | 65 |
| Widowed | 14.9 | 15.5 | 22.6 |
| Social support | |||
| Emotional support (%) | 93.2 | 89.1 | 95.1 |
| Financial support (%) | 73.4 | 68.4 | 63.5 |
| Number of close friends (± SEM) | 7.2 (0.3) | 6.6 (1.3) | 8.2 (1.1) |
| Household income (median), $ | 35,000 – 45,000 | 35,000 – 45,000 | 35,000 – 45,000 |
| Income to poverty ratio | 2.9 (0.1) | 2.6 (0.1) | 3.1 (0.1) |
| Food security (%) | |||
| Full | 83.7 | 81.2 | 92 |
| Marginal | 6.7 | 6.5 | 3.8 |
| Low | 5.6 | 7.3 | 1.8 |
| Very low | 4 | 5 | 2.4 |
| Insured (%) | 91.8 | 89.8 | 96.3 |
| Presence of usual healthcare source (%) |
96.3 | 97.8 | 96.8 |
| Mental health care visit past year (%) | 8.2 | 6.4 | 4.3 |
|
Health behaviors and comorbid conditions |
|||
| Smoking status | |||
| Never (%) | 44.5 | 47.5 | 43.8 |
| Former (%) | 37.2 | 36.6 | 43.1 |
| Current (%) | 18.3 | 15.9 | 13.1 |
| Cigarettes per day (± SEM) | 3.3 (0.6) | 3.1 (0.5) | 2.8 (1.0) |
| Alcohol status | |||
| Drinking days per month (± SEM) | 3.9 (0.6) | 1.9 (0.2) | 5.1 (0.8) |
| Binge days per month (± SEM) | 0.7 (0.1) | 0.4 (0.1) | 0.2 (0.08) |
| Illicit drug use (%) | |||
| Current marijuana use | 2.8 | 1.7 | 0 |
| Current other illicit drugs use | 0.2 | 0.3 | 0 |
| Heart disease (%) | 15.6 | 27.5 | 21.1 |
| Stroke (%) | 8.1 | 12.6 | 12.6 |
| Hypertension (%) | 55.4 | 71.5 | 55.2 |
| Pulmonary disease (%) | 25.3 | 26.6 | 21.6 |
| Liver disease (%) | 5.5 | 3.7 | 5.5 |
| Arthritis (%) | 100 | 52.4 | 47.3 |
| Diabetes (%) | 17.5 | 100 | 18.5 |
| Cancer (%) | 17.9 | 18.9 | 100 |
| Number of comorbidities | |||
| 0 | 24 | 12.2 | 16 |
| 1 | 22 | 24 | 30 |
| 2 | 25.4 | 25.6 | 26.3 |
| 3 | 11.9 | 22.4 | 18.8 |
| 4 | 3.7 | 9.6 | 4 |
| 5 | 1.7 | 5.5 | 2.7 |
| 6 | 0.3 | 0.7 | 2.2 |
| 7 | 0 | 0 | 0 |
| Depression (%) | 4.1 (0.l2) | 4.0 (0.1) | 3.1 (0.5) |
| Anxiety days past month (± SEM) | 6.6 (0.5) | 5.7 (0.5) | 4.5 (0.7) |
| Sleep disturbances | |||
| Insomnia (%) | 1.8 | 2 | 0.7 |
| Sleep duration < 6 hours/ night (%) | 18.9 | 17.9 | 14.2 |
| Excessive daytime sleepiness (%) | |||
| Never | 28.4 | 38.8 | 35.9 |
| Rarely | 19.4 | 14.3 | 17.4 |
| Sometimes | 29.6 | 25 | 23.8 |
| Often | 14.5 | 11.7 | 16.6 |
| Almost always | 8.1 | 10.2 | 6.3 |
| Pain days past month (± SEM) | 7.8 (0.6) | 7.2 (0.9) | 6.8 (0.9) |
| Limitation in ADL | |||
| In/out of bed difficulty (%) | |||
| None | 82.3 | 79 | 82.4 |
| Some | 13.6 | 14.5 | 13.6 |
| Much | 3.5 | 5.6 | 3.2 |
| Unable | 0.6 | 0.9 | 0.8 |
| Dressing difficulty (%) | |||
| None | 86.1 | 84 | 90 |
| Some | 11 | 10.6 | 7.9 |
| Much | 2.5 | 4.4 | 2.3 |
| Unable | 0.4 | 1 | 0.8 |
| Eating difficulty (%) | |||
| None | 93.1 | 91.5 | 96 |
| Some | 6.2 | 6.5 | 3.1 |
| Much | 0.6 | 1.8 | 0.9 |
| Unable | 0.1 | 0.2 | 0 |
| Walking difficulty (%) | |||
| None | 89.6 | 83.3 | 87.2 |
| Some | 7.4 | 11.5 | 8.3 |
| Much | 1.7 | 2.9 | 1.5 |
| Unable to do | 1.3 | 2.3 | 3 |
| Unable in at least one ADL (%) | 2 | 4 | 3 |
| Special equipment use (%) | 17.3 | 24.9 | 20 |
SEM = standard error of the mean; ADL = activities of daily living
The characteristics of those with and without suicidal ideation among adults with arthritis are shown in table 2. Persons with suicidal ideation were more likely non-white, had lower education levels, were poorer, and had higher frequency of mental health care visits in the past year, current smoking, binge days per month, anxiety days and pain days during the past month. They also had more comorbidities, sleep disturbances, depression, and ADL limitations.
Table 2.
Characteristics of adults with and without suicidal ideation among those with arthritis
| Present | Absent | P | |
|---|---|---|---|
|
Demographic and socio-economic factors |
|||
| Age, years(± SEM) | 59.2 (2.0) | 61.4 (0.3) | 0.26 |
| Male (%) | 37.7 | 40.1 | 0.72 |
| Ethnicity (%) | |||
| White | 67.7 | 80.4 | 0.03 |
| Black | 11.4 | 9.8 | |
| Hispanic | 16.5 | 6.1 | |
| Other | 4.4 | 3.7 | |
| Born in U.S. (%) | 86.6 | 93.4 | 0.005 |
| Education level (%) | |||
| Less than high school | 42.5 | 23.0 | 0.0002 |
| High school graduate | 23.1 | 29.0 | |
| Some college | 27.2 | 26.6 | |
| College graduate | 7.2 | 21.4 | |
| Marital status (%) | |||
| Single | 8.7 | 5.0 | 0.15 |
| Married | 53.3 | 65.3 | |
| Divorced/separated | 21.5 | 15.4 | |
| Widowed | 16.5 | 14.3 | |
| Social support | |||
| Emotional support (%) | 82.7 | 93.9 | <0.0001 |
| Financial support (%) | 60.2 | 73.6 | 0.02 |
| Number of close friends (± SEM) | 5.5 (0.7) | 7.4 (0.3) | 0.02 |
| Annual household income, $ | |||
| < 20,000 | 39.2 | 21.3 | <0.0001 |
| 20,000 – 44,999 | 51.0 | 31.4 | |
| 45,000 – 64,999 | 4.9 | 14.6 | |
| ≥ 65,000 | 4.9 | 32.7 | |
| Income to poverty ratio | 1.7 (0.2) | 3.0 (0.1) | <0.0001 |
| Food security (%) | |||
| Full | 58.1 | 84.8 | <0.0001 |
| Marginal | 13.0 | 6.3 | |
| Low | 15.6 | 5.2 | |
| Very low | 13.3 | 3.7 | |
| Insured (%) | 80.6 | 92.1 | 0.004 |
| Presence of usual healthcare source (%) |
97.1 | 96.5 | 0.73 |
| Mental health care visit past year (%) |
28.0 | 7.3 | <0.0001 |
|
Health behaviors and comorbid conditions |
|||
| Smoking status | |||
| Never (%) | 32.0 | 44.2 | <0.0001 |
| Former (%) | 25.2 | 38.9 | |
| Current (%) | 42.8 | 16.9 | |
| Cigarettes per day (± SEM) | 9.0 (1.4) | 3.0 (0.6) | <0.0001 |
| Alcohol status | |||
| Drinking days per month (± SEM) | 2.1 (0.9) | 4.1 (0.6) | 0.003 |
| Binge days per month (± SEM) | 1.6 (0.8) | 0.6 (0.2) | 0.28 |
| Illicit drug use (%) | |||
| Current marijuana use | 6.7 | 2.8 | 0.23 |
| Current other illicit drugs use | 0.7 | 0.2 | 0.42 |
| Heart disease (%) | 21.6 | 14.7 | 0.06 |
| Stroke (%) | 17.9 | 7.3 | 0.001 |
| Hypertension (%) | 74.5 | 54.7 | 0.04 |
| Pulmonary disease (%) | 34.5 | 25.4 | 0.11 |
| Liver disease (%) | 9.2 | 5.4 | 0.19 |
| Arthritis (%) | 15.0 (1.6) | 13.6 (0.7) | 0.43 |
| Diabetes (%) | 25.3 | 16.5 | 0.003 |
| Cancer (%) | 18.5 | 17.6 | 0.84 |
| Number of comorbidities | |||
| 0 | 0.8 | 25.5 | <0.0001 |
| 1 | 39.1 | 32.2 | |
| 2 | 33.4 | 25.2 | |
| 3 | 16.2 | 11.9 | |
| 4 | 7.5 | 3.3 | |
| 5 | 1.8 | 1.6 | |
| 6 or 7 | 1.2 | 0.3 | |
| Depression (%) | 11.7 (0.8) | 3.7 (0.2) | <0.0001 |
| Anxiety days past month (± SEM) | 19.4 (1.2) | 5.9 (0.4) | <0.0001 |
| Sleep disturbances | |||
| Insomnia (%) | 7.4 | 1.4 | 0.001 |
| Sleep duration < 6 hours/ night (%) | 36.2 | 17.9 | 0.006 |
| Excessive daytime sleepiness (%) | |||
| Never | 9.0 | 28.1 | <0.0001 |
| Rarely | 13.3 | 19.5 | |
| Sometimes | 26.9 | 31.2 | |
| Often | 27.1 | 14.1 | |
| Almost always | 23.7 | 7.1 | |
| Pain days past month (± SEM) | 17.5 (1.3) | 7.2 (.6) | <0.0001 |
| Limitation in ADL | |||
| In/out of bed difficulty (%) | |||
| None | 48.2 | 84.3 | <0.0001 |
| Some | 32.6 | 12.8 | |
| Much | 17.8 | 2.3 | |
| Unable | 1.4 | 0.6 | |
| Dressing difficulty (%) | |||
| None | 72.7 | 87.4 | <0.0001 |
| Some | 18.0 | 10.4 | |
| Much | 6.9 | 2.0 | |
| Unable | 2.3 | 0.2 | |
| Eating difficulty (%) | |||
| None | 76.4 | 94.2 | <0.0001 |
| Some | 17.9 | 5.5 | |
| Much | 4.6 | 0.3 | |
| Unable | 1.1 | 0 | |
| Walking difficulty (%) | |||
| None | 71.6 | 91.5 | <0.0001 |
| Some | 18.7 | 6.2 | |
| Much | 7.7 | 1.2 | |
| Unable to do | 2.0 | 1.1 | |
| Unable in at least one ADL (%) | 6.7 | 1.6 | 0.04 |
| Special equipment use (%) | 37.6 | 14.9 | <0.0001 |
SEM = standard error of the mean; ADL = activities of daily living
Correlates for suicidal ideation
Figure 1 presents the variable importance plots of correlates for suicidal ideation, based on the random forest. The most important correlates for suicidal ideation in individuals with arthritis were the following: depression evaluated by PHQ-8, anxiety days per month, age, duration of arthritis, income to poverty ratio, number of close friends or relatives, pain, alcohol use, and excessive daytime sleepiness. Number of comorbidities was also important and ranked higher than any specific comorbid condition. Other important factors were marital status, education level, cigarettes per day, and difficulty in getting in or out of bed. Other measures, including country of birth, current use of marijuana or other illicit drugs, and the presence of a usual source of healthcare were much less important in distinguishing those at high risk of suicidal ideation.
Figure 1.
Variable importance plots of correlates for suicidal ideation based on the random forest in adults with arthritis (A), diabetes (B) and cancer (C).
The prevalence of suicidal ideation by the top 16 correlates is shown in Table 3. The prevalence of suicidal thoughts increased with PHQ-8 scores, number of days with anxiety, pain, excessive daytime sleepiness, more comorbid conditions, heavier smoking, binge drinking, and more difficulty transferring from bed. Suicidal ideation was generally lower among those with longer duration of arthritis, older age, higher income and income to poverty ratio, more close friends, more frequent light drinking, and more education.
Table 3.
Prevalence of suicidal ideation by the top 16 correlates in persons with arthritis
| Correlates | Suicide ideation (%) | |
|---|---|---|
| PHQ-8 score | < 10 | 2.9 |
| 10–16 | 16.9 | |
| 17–24 | 51.0 | |
| Anxiety days past month | 0–5 | 1.6 |
| 6–14 | 7.6 | |
| 15–30 | 19.1 | |
| Duration of arthritis, years | 0–2 | 7.1 |
| 2.1–5.0 | 4.1 | |
| 5.1–10 | 4.5 | |
| 10.1+ | 5.8 | |
| Age, years | 40–54 | 7.5 |
| 55–64 | 4.7 | |
| 65+ | 4.6 | |
| Income to poverty ratio | <0.5 | 9.7 |
| 0.5–0.99 | 12.5 | |
| 1.0–1.99 | 8.2 | |
| 2.0+ | 2.7 | |
| Pain days past month | 0–5 | 2.6 |
| 6–14 | 6.1 | |
| 15–30 | 13.7 | |
| Number of close friends | 0–2 | 8.5 |
| 3–6 | 5.8 | |
| 7+ | 3.4 | |
| Household income (median), $ | < 45,000 | 9.2 |
| 45,000–75,000 | 1.9 | |
| >75,000 | 1.1 | |
| Binge days per month | None | 4.6 |
| Any | 6.1 | |
| Excessive daytime sleepiness | Never | 1.8 |
| Rarely | 3.9 | |
| Sometimes | 4.9 | |
| Often | 10.2 | |
| Almost always | 16.4 | |
| Number of comorbidities | 0 | 0.2 |
| 1 | 6.7 | |
| 2 | 7.3 | |
| 3 | 7.4 | |
| 4 | 11.9 | |
| 5 | 6.3 | |
| 6 | 14.5 | |
| 7 | 100 | |
| Drinking days per month | 0–5 | 6.3 |
| 6–14 | 4.0 | |
| 15–30 | 2.1 | |
| Marital status | Single | 9.2 |
| Married | 4.6 | |
| Divorced/separated | 7.6 | |
| Widowed | 6.4 | |
| Education level | Less than high school | 9.8 |
| High school graduate | 4.5 | |
| Some college | 5.7 | |
| College graduate | 1.9 | |
| Number of cigarettes per day | 0 | 3.9 |
| 1–19 | 7.6 | |
| 20+ | 18.7 | |
| In/out of bed difficulty | None | 3.2 |
| Some | 13.1 | |
| Much | 31.0 | |
| Unable | 12.9 |
To assess the commonality of the most important factors for suicidal ideation among persons with arthritis, diabetes and cancer, we used the top 16 variables which were distinctly more predictive in arthritis group. Eleven of the top 16 correlates in people with arthritis were also among the top correlates in people with diabetes and cancer, including PHQ-8, anxiety, age, income to poverty ratio, social network, pain, binge drinking and number of comorbidities. No correlates were unique to persons with arthritis.
The random forests were highly accurate. The cross-validated test set misclassification error for each of the random forests was 0, indicating that the set of correlates separated those with and without suicide ideation very well in each of the arthritis, diabetes, and cancer groups.
Subgroups of persons with arthritis at high risk for suicidal ideation
A single classification tree analysis was performed using the top 16 correlates for suicidal ideation in adults with arthritis. This analysis demonstrates how individual correlates interact to identify particular subgroups of persons with arthritis at greatest risk for suicidal ideation. The tree fit the data well, with a 10-fold cross-validated test set misclassification rate of 3.56 %. The risk for suicidal ideation varied most by PHQ-8 scores, duration of arthritis, binge drinking, income, and number of comorbidities (Figure 2). Among persons with arthritis, subgroups most highly predisposed to suicidal ideation were those with PHQ-8 between18 and 24 and less than 4.5 years of arthritis (96.5%), and those with PHQ-8 between 7 and 17, at least 1.24 days of binges per month and either annual income of $45,000 or higher (85.4%) or less than $45,000 and more than 3 comorbidities (70.8%). People with arthritis and PHQ-8 between 0 and 6 had a low predisposition to suicidal ideation (1.1%).
Figure 2.
Identification of subgroups at highest risk for suicidal ideation among adults with arthritis using classification tree analysis.
Discussion
Suicidal ideation among adults with arthritis is of growing relevance to public health given the increase in projected prevalence of arthritis in U.S. adults (23). Our study, using data from a nationally representative survey, showed that the prevalence of suicidal ideation in adults with arthritis was higher than that of persons without arthritis. Depression, anxiety, duration of arthritis, age, income to poverty ratio, social network, pain, alcohol use, excessive daytime sleepiness and comorbidities were the most important correlates for suicidal thoughts among individuals with arthritis. Those with PHQ-8 score between 18 and 24 and short duration of arthritis (<4.5 years) were at highest risk for suicidal ideation.
Our findings were consistent with two previous studies showing that suicidal ideation was prevalent among adults with arthritis. One study examined 123 hospital outpatients with RA without including any control group, and the other used data from the Canadian Community Health Survey 2000–2001 (7, 8). A similar prevalence of suicidal thoughts among persons with arthritis, diabetes and cancer was identified in our study suggesting that the chronic nature of illness can impair the psychological well-being of people with these disorders (5, 6).
Despite continuing research, it remains difficult to predict which individuals are most vulnerable for suicidal behavior which is never the consequence of a single cause (1). Mood disorders such as anxiety and particularly depression significantly increase the risk of suicidal thoughts in the general population and several medical conditions (24–26). Depression was also the main risk factor for suicidal ideation and suicide completion in studies of RA patients (7, 27). In our study, depression was the most important factor identifying subsets of individuals with arthritis affected with suicidal ideation. However, it is well established in the psychiatric literature that not only depression but also other mental illnesses (psychotic and personality disorders and alcohol/substance use), socio-demographic factors, and physical health problems contribute to suicidal ideation, each being interrelated (1, 28–30).
In our study, the risk for suicidal ideation was higher among adults aged 40–54 years compared to older people, those with limited number of close friends or relatives, and those with low income to poverty ratio (29, 31). Although suicide rates in elderly people have fallen in many countries, those in younger people have risen. (1). Weak ties and low social support from friends or relatives have been significantly associated with suicidal thoughts in the literature (31, 32). Furthermore, data from psychiatric epidemiology surveys showed increased rates of suicidal ideation and suicidal attempts among adults with low income and income to poverty ratio, which remained unchanged after adjusting for the presence of mental disorders (33, 34).
Regarding health behaviors and comorbidities, the results of our study were consistent with those of previous studies reporting that alcohol, smoking and coexistence of arthritis with other chronic diseases were important correlates for suicidal thoughts (1, 35, 36). The screening for alcohol abuse is highly feasible in clinical practice and its detection should increase the suspicion for suicidal ideation in persons with depression. The presence of a general medical illness has been associated with an increased risk of both suicidal ideation and suicide attempts, and having more than one illness conferred a particularly high risk in the general population (37, 38, 39).
Persistent pain has been associated with elevated rates of suicidal behavior (40, 41). However, studies of chronic pain conditions have not adequately elucidated whether the increased risk of suicidal behavior is associated with pain per se or with the physical disability that pain produces (8, 42, 43). A recent review also suggested that sleep problems might be an indirect mechanism by which chronic pain is associated with suicidal thoughts (41). The present study indicates that mainly pain but also limitations in ADL and excessive daytime sleepiness were among the most important contributors for suicidal ideation in adults with arthritis, cancer and diabetes. Excessive daytime sleepiness was a more important correlate than insomnia, possibly because it was much more prevalent than insomnia and therefore had a higher likelihood of demonstrating associations, but also because it may be a consequence of insomnia or a manifestation of depression.
Commonalities among the major correlates for suicidal ideation in individuals with arthritis, cancer and diabetes include depression and anxiety, age, income to poverty ratio, social network, pain, binge drinking, and comorbidities. Eleven of the top 16 most important contributors for suicidal thoughts were common among arthritis and the other two chronic disease groups. This finding supports the general role of these factors on suicidal behavior among chronic diseases regardless of the disease itself.
The results of the current study suggest that individuals with arthritis at high risk for suicidal thoughts should receive close monitoring and early intervention (44, 45). Self- report measures such as PHQ-9 can be used as screening tools of depressive disturbances in outpatient clinics (44). Recognition and effective treatment of depression, anxiety and comorbidities, enhancement of social support resources, and management of pain and physical disability are fundamentally important in reducing suicidal behavior (46–48). The strengths of this study include the large nationally representative sample, and the wide range of socio-demographic, physical and mental health factors examined. We also included pain and physical limitations, often present in persons with arthritis. We provided new information on the prevalence of suicidal ideation among individuals with arthritis, and on subgroups of people with arthritis most vulnerable for suicidal thoughts. The use of classification tree analysis helped to assess multiple correlates and their interactions to identify subgroups at greatest risk.
This study is limited in that the survey did not include some factors that may be associated with suicidal thoughts such as other psychiatric (e.g. personality disorders) or chronic pain diseases (abdominal pain, migraine). In addition, since NHANES used a self-completed screening measure rather than an interview, we had to rely on the patient’s interpretation of the question. However, PHQ-9 is a screening instrument with high reliability and validity in the general population and various medical settings such as primary care population, general hospital inpatients and patients with rheumatic diseases (RA, osteoarthritis, fibromyalgia) (43, 49). Data on physician-diagnosed arthritis were also self-reported, but self-reports have been validated previously for surveillance purposes (10). Self-reported arthritis likely includes many types of arthritis, which are represented in the study in proportion to their prevalence. An additional limitation is that our study cannot identify correlates of suicidal ideation that may be unique or specific to a given type of arthritis, and may not reflect correlates of suicidal ideation in less prevalent types of arthritis, such as rheumatoid arthritis. Data on medications and on prior suicide attempts were not available in the survey. Finally, the cross-sectional nature of study precluded determination of causality.
In conclusion, suicidal ideation is prevalent among adults with arthritis and should be carefully evaluated by physicians. Depression and short duration of arthritis, binge drinking, income, and the presence of more than three comorbidities identified subgroups at greatest risk for suicidal ideation that should be the focus of prevention and intervention approaches.
Significance and innovation.
First population-based U.S. study of suicidal ideation in adults with arthritis.
Identification of subgroups of people with arthritis at highest risk for suicidal ideation.
Main correlate was depression assessed by PHQ-8.
Similar risk factors in diabetes and cancer.
Acknowledgement
This work was supported in part by the Intramural Research Program, National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health.
References
- 1.Hawton K, van Heeringen K. Suicide. Lancet. 2009;373:1372–1381. doi: 10.1016/S0140-6736(09)60372-X. [DOI] [PubMed] [Google Scholar]
- 2.WHO. World report on violence and health. Geneva: World Health Organization; 2002. [Google Scholar]
- 3.Nock MK, Borges G, Bromet EJ, Cha CB, Kessler RC, Lee S. Suicide and suicidal behavior. Epidemiol Rev. 2008;30:133–154. doi: 10.1093/epirev/mxn002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Mann JJ, Waternaux C, Haas GL, Malone KM. Toward a clinical model of suicidal behavior in psychiatric patients. Am J Psychiatry. 1999;156:181–189. doi: 10.1176/ajp.156.2.181. [DOI] [PubMed] [Google Scholar]
- 5.Beck AT, Weishaar ME. Suicide risk assessment and prediction. Crisis. 1990;11:22–30. [PubMed] [Google Scholar]
- 6.Harris EC, Barraclough BM. Suicide as an outcome for medical disorders. Medicine. 1994;73:281–296. doi: 10.1097/00005792-199411000-00001. [DOI] [PubMed] [Google Scholar]
- 7.Treharne GJ, Lyons AC, Kitas GD. Suicidal ideation in patients with rheumatoid arthritis. Research may help identify patients at high risk. BMJ. 321:1290. 200018. [PMC free article] [PubMed] [Google Scholar]
- 8.Fuller-Thomson E, Shaked Y. Factors associated with depression and suicidal ideation among individuals with arthritis or rheumatism: findings from a representative community survey. Arthritis Rheum. 2009;61:944–950. doi: 10.1002/art.24615. [DOI] [PubMed] [Google Scholar]
- 9.Questionnaires, datasses, and related documentation. [Accessed December 17, 2009];National Health and Nutrition Examination Survey. http://www.cdc.gov/nchs/nhanes/nhanes_questionnaires.htm.
- 10.Sacks JJ, Harrold LR, Helmick CG, Gurwitz JH, Emani S, Yood RA. Validation of a surveillance case definition of arthritis. J Rheumatol. 2005;32:340–347. [PubMed] [Google Scholar]
- 11.Gonçalves V, Jayson G, Tarrier N. A longitudinal investigation of psychological morbidity in patients with ovarian cancer. Br J Cancer. 2008;99:1794–1801. doi: 10.1038/sj.bjc.6604770. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Partridge A, Adloff K, Blood E, Dees EC, Kaelin C, Golshan M, et al. Risk perceptions and psychosocial outcomes of women with ductal carcinoma in situ: longitudinal results from a cohort study. J Natl Cancer Inst. 2008;100:228–229. doi: 10.1093/jnci/djn010. [DOI] [PubMed] [Google Scholar]
- 13.Martin A, Rief W, Klaiberg A, Braehler E. Validity of the Brief Patient Health Questionnaire Mood Scale (PHQ-9) in the general population. Gen Hosp Psychiatry. 2006;28:71–77. doi: 10.1016/j.genhosppsych.2005.07.003. [DOI] [PubMed] [Google Scholar]
- 14.Thompson D, Goebert D, Takeshita J. A program for reducing depressive symptoms and suicidal ideation in medical students. Acad Med. 2010;85:1635–1639. doi: 10.1097/ACM.0b013e3181f0b49c. [DOI] [PubMed] [Google Scholar]
- 15.Walker J, Waters RA, Murray G, Swanson H, Hibberd CJ, Rush RW, et al. Better off dead: suicidal thoughts in cancer patients. J Clin Oncol. 2008;26:4725–4730. doi: 10.1200/JCO.2007.11.8844. [DOI] [PubMed] [Google Scholar]
- 16.Feightner JW, Worrall G. Early detection of depression by primary care physicians. CMAJ. 1990;142:1215–1220. [PMC free article] [PubMed] [Google Scholar]
- 17.Nord M. Briefing rooms. [Accessed November 15, 2006];Food security in the United States: hunger and food security. http://www.ers.usda.gov/Briefing/FoodSecurity/labels.htm.
- 18.Kroenke K, Strine T, Spitzer R, Williams J, Berry J, Mokdad A. The PHQ-8 as a measure of current depression in the general population. J Affect Disord. 2009;114:163–173. doi: 10.1016/j.jad.2008.06.026. [DOI] [PubMed] [Google Scholar]
- 19.Louie GH, Tektonidou MG, Caban-Martinez AJ, Ward MM. Sleep disturbances in adults with arthritis: Prevalence, mediators, and subgroups at greatest risk. Arthritis Care Res (Hoboken) 2011;63:247–260. doi: 10.1002/acr.20362. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Breiman L. Random Forests. Machine Learning. 2001;45:5–32. [Google Scholar]
- 21.R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2010. ISBN 3-900051-07-0, http://www.R-project.org. [Google Scholar]
- 22.Schwarz DF, König IR, Zeigler A. On safari to Random Jungle: a fast implementation of Random Forests for high-dimensional data. Bioinformatics. 2010;26(14):1752–1758. doi: 10.1093/bioinformatics/btq257. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hootman JM, Helmick CG. Projections of US prevalence of arthritis and associated activity limitations. Arthritis Rheum. 2006;54:226–229. doi: 10.1002/art.21562. [DOI] [PubMed] [Google Scholar]
- 24.Nock MK, Hwang I, Sampson NA, Kessler RC. Mental disorders, comorbidity and suicidal behavior: results from the National Comorbidity Survey Replication. Molecular Psychiatry. 2010;15:868–876. doi: 10.1038/mp.2009.29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Hintikka J, Koivumaa-Honkanen H, Lehto SM, Tolmunen T, Honkalampi K, Haatainen K, et al. Are factors associated with suicidal ideation true risk factors? A 3-year prospective follow-up study in a general population. Soc Psychiatry Psychiatr Epidemiol. 2009;44:29–33. doi: 10.1007/s00127-008-0401-6. [DOI] [PubMed] [Google Scholar]
- 26.Cavanagh JT, Carson AJ, Sharpe M, Lawrie SM. Psychological autopsy studies of suicide: a systematic review. Psychol Med. 2003;33:395–405. doi: 10.1017/s0033291702006943. [DOI] [PubMed] [Google Scholar]
- 27.Timonen M, Viilo K, Hakko H, Särkioja T, Ylikulju M, Meyer-Rochow VB, et al. Suicides in persons suffering from rheumatoid arthritis. Rheumatology (Oxford) 2003;42:287–291. doi: 10.1093/rheumatology/keg082. [DOI] [PubMed] [Google Scholar]
- 28.Paykel ES, Myers JK, Lindenthal JJ, Tanner J. Suicidal feelings in the general population: a prevalence study. Br J Psychiatry. 1974;124:460–469. doi: 10.1192/bjp.124.5.460. [DOI] [PubMed] [Google Scholar]
- 29.Casey PR, Dunn G, Kelly BD, Birkbeck G, Dalgaard OS, Lehtinen V, et al. on behalf of the ODIN group. Factors associated with suicidal ideation in the general population. Five-centre analysis from the ODIN study. Br J Psychiatry. 2006;189:410–415. doi: 10.1192/bjp.bp.105.017368. [DOI] [PubMed] [Google Scholar]
- 30.ten Have M, de Graaf R, van Dorsselaer S, Verdurmen J, van 't Land H, Vollebergh W, et al. Incidence and course of suicidal ideation and suicide attempts in the general population. Can J Psychiatry. 2009;54:824–833. doi: 10.1177/070674370905401205. [DOI] [PubMed] [Google Scholar]
- 31.Gunnell D, Harbord R, Singleton N, Jenkins R, Lewis G. Factors influencing the development and amelioration of suicidal thoughts in the general population. Cohort study. Br J Psychiatry. 2004;185:385–393. doi: 10.1192/bjp.185.5.385. [DOI] [PubMed] [Google Scholar]
- 32.Park SM, Cho SI, Moon SS. Factors associated with suicidal ideation: role of emotional and instrumental support. J Psychosom Res. 2010;69:389–397. doi: 10.1016/j.jpsychores.2010.03.002. [DOI] [PubMed] [Google Scholar]
- 33.Kalist DE, Molinari NA, Siahaan F. Income, employment and suicidal behavior. J Ment Health Policy Econ. 2007;10:177–187. [PubMed] [Google Scholar]
- 34.Appleby L, Cooper J, Amos T, Faragher B. Psychological autopsy studies of suicide by people aged under 35. Br J Psychiatry. 1999;175:168–174. doi: 10.1192/bjp.175.2.168. [DOI] [PubMed] [Google Scholar]
- 35.Tanskanen A, Viinamaki H, Hintikka J, Koivumaa-Honkanen H-T, Lehtonen J. Smoking and suicidality among psychiatric patients. Am J Psychiatry. 1998;155:129–130. doi: 10.1176/ajp.155.1.129. [DOI] [PubMed] [Google Scholar]
- 36.Wilcox HC, Conner KR, Caine ED. Association of alcohol and drug use disorders and completed suicide: an empirical review of cohort studies. Drug Alcohol Depend. 2004;76 Suppl:S11–S19. doi: 10.1016/j.drugalcdep.2004.08.003. [DOI] [PubMed] [Google Scholar]
- 37.Wells KB, Golding JM, Burnam MA. Psychiatric disorder in a sample of the general population with and without chronic medical conditions. Am J Psychiatry. 1988;145:976–981. doi: 10.1176/ajp.145.8.976. [DOI] [PubMed] [Google Scholar]
- 38.Ratcliffe GE, Enns MW, Belik SL, Sareen J. Chronic pain conditions and suicidal ideation and suicide attempts: an epidemiologic perspective. Clin J Pain. 2008;24:204–210. doi: 10.1097/AJP.0b013e31815ca2a3. [DOI] [PubMed] [Google Scholar]
- 39.Scott KM, Hwang I, Chiu WT, Kessler RC, Sampson NA, Angermeyer M, et al. Chronic physical conditions and their association with first onset of suicidal behavior in the world mental health surveys. Psychosom Med. 2010;72:712–719. doi: 10.1097/PSY.0b013e3181e3333d. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Ilgen MA, Zivin K, McCammon RJ, Valenstein M. Pain and suicidal thoughts, plans and attempts in the United States. Gen Hosp Psychiatry. 2008;30:521–527. doi: 10.1016/j.genhosppsych.2008.09.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Tang NK, Crane C. Suicidality in chronic pain: a review of the prevalence, risk factors and psychological links. Psychol Med. 2006;36:575–586. doi: 10.1017/S0033291705006859. [DOI] [PubMed] [Google Scholar]
- 42.Lépine JP, Briley M. The epidemiology of pain in depression. Hum Psychopharmacol. 2004;19 Suppl 1:S3–S7. doi: 10.1002/hup.618. [DOI] [PubMed] [Google Scholar]
- 43.Margaretten M, Yelin E, Imboden J, Graf J, Barton J, Katz P, et al. Predictors of depression in a multiethnic cohort of patients with rheumatoid arthritis. Arthritis Rheum. 2009;61:1586–1591. doi: 10.1002/art.24822. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Nicassio PM. The Problem of Detecting and Managing Depression in the Rheumatology Clinic. Arthritis Rheum. 2008;58:155–158. doi: 10.1002/art.23348. [DOI] [PubMed] [Google Scholar]
- 45.Mann JJ, Apter A, Bertolote J, Beautrais A, Currier D, Haas A, et al. Suicide prevention strategies. A systematic review. JAMA. 2005;94:2064–2074. doi: 10.1001/jama.294.16.2064. [DOI] [PubMed] [Google Scholar]
- 46.Gettings L. Psychological well-being in rheumatoid arthritis: a review of the literature. Musculoskeletal Care. 2010;8:99–106. doi: 10.1002/msc.171. [DOI] [PubMed] [Google Scholar]
- 47.Sleath B, Chewning B, de Vellis BM, Weinberger M, de Vellis RF, Tudor G, et al. Communication about depression during rheumatoid arthritis patient visits. Arthritis Rheum. 2008;59:186–191. doi: 10.1002/art.23347. [DOI] [PubMed] [Google Scholar]
- 48.Phillips RS. Preventing depression: a program for African American elders with chronic pain. Fam Comm Health. 2000;22:57–65. [Google Scholar]
- 49.Rosemann T, Backenstrass M, Joest K, Rosemann A, Szecsenyi J, Laux G. Predictors of depression in a sample of 1,021 primary care patients with osteoarthritis. Arthritis Rheum. 2007;57:415–422. doi: 10.1002/art.22624. [DOI] [PubMed] [Google Scholar]