The Internal Struggle Between the Wish to Die and the Wish to Live: A Risk Factor for Suicide

The study group was drawn from a group of 6,891 patients who were consecutively evaluated at the psychiatric outpatient clinic of the University of Pennsylvania between Jan. 1, 1975, and Dec. 31, 1995, and who were followed prospectively to ascertain the occurrence of death. Detailed information about this group and the procedures have been previously described (3), but only 5,814 patients (84%) with complete information on selected variables were chosen for the present analyses. Of 44 suicides, 42 (95%) were identified as prior to 1993. For the 44 suicides (0.8%) and 5,770 comparison subjects (99.2%), respectively, data were as follows:

After obtaining written informed consent from the patients, consecutive editions of the Structured Clinical Interviews for DSM-III or DSM-III-R Axis I and Axis II Disorders (4) and the Scale for Suicide Ideation (5) were administered by a doctoral-level clinician. The Scale for Suicide Ideation includes two items that assess the wish to live that is rated by using “moderate to strong” (score=0), “weak” (score=1), and “none” (score=2) and the wish to die that is rated by using “none” (score=0), “weak” (score=1), and “moderate to strong” (score=2). A wish to die/wish to live difference score was created by reversing the ratings for the wish to live item and subtracting them from the ratings for the wish to die item. This index score ranged from –2 to 2, with higher scores indicating a stronger orientation toward death. A revised Scale for Suicide Ideation total score was created by subtracting the wish to live and wish to die item ratings from the total score.

As described elsewhere (3), deaths in this group were ascertained prospectively by using the National Death Index (6), a centralized computer database of death record information compiled from data submitted annually by individual state vital statistics offices. Death certificates were obtained from state vital statistics offices to determine the cause of death. The median length of follow-up was 10 years (ranging from <1 to 20). The mean length of time from the intake interview until suicide was 4.07 years (SD=3.96, range=2 weeks to 12 years).

Nonparametric survival analyses were conducted with Cox’s proportional hazard regression model (7). The dependent variable was the occurrence of a suicide. The length of follow-up for each patient who died from suicide is represented by the number of days between the date of evaluation and the date of death. Data from patients who died were censored at the date of death. Data from patients who survived were censored as of Dec. 31, 1995.

The wish to die/wish to live index was found to be a risk factor for suicide (univariate hazard ratio=2.44, 95% confidence interval [CI]=1.60–3.72). Next, the wish to die/wish to live index was dichotomized by using cutoff scores ranging from –2 to 2. The results indicated that there were no significant hazard ratios for predicting suicide with cutoff scores <0. However, a cutoff score of 1 yielded a hazard ratio of 2.68 (95% CI=1.46–4.87), and cutoff score of 2 afforded a higher hazard ratio of 6.51 (95% CI=2.89–14.64).

A multivariate survival analysis was conducted to evaluate the unique contribution of the wish to die/wish to live index for predicting suicide after we controlled for the seven variables that Brown et al. (3) had found to be unique risk factors for suicide: age (years), hospitalization for a psychiatric disorder, previous suicide attempt, bipolar disorder, major depressive disorder, unemployment status, and suicidal ideation. When the revised Scale for Suicide Ideation total scores and the wish to die/wish to live index were simultaneously entered into the model with the other six variables, the revised Scale for Suicide Ideation was no longer significant (hazard ratio=1.04, 95% CI=0.99–1.09) and was excluded from a subsequent multivariate analysis. Table 1 indicates that the dichotomized wish to die/wish to live index was retained in the model; the index was a significant and unique risk factor for suicide even after control for other risk factors. In addition, the following baseline variables were found to be associated with the wish to die/wish to live index: previous suicide attempts (odds ratio=4.98, 95% CI=2.45–10.01), psychiatric hospitalizations (odds ratio=4.98, 95% CI=2.45–10.11), major depression (odds ratio=3.14, 95% CI=1.51–6.56), and female gender (odds ratio=2.34, 95% CI=1.10–4.98).

In addition, 113 deaths due to natural causes (excluding suicidal, homicidal, or accidental deaths) were identified during the follow-up period. A Cox regression survival analysis was conducted (113 deaths and 5,657 comparison subjects) to evaluate the risk of the wish to die/wish to live index for natural death. The results indicated that the index was unassociated with natural death (χ²=1.40, df=1, p=0.24; odds ratio=1.65, 95% CI=0.72–3.75).

The present study found that the assessment of the difference between clinical ratings of a patient’s wish to die and a wish to live constitutes a risk factor for suicide in a group of psychiatric outpatients. These findings support the internal struggle hypothesis that proposes that the severity of suicidal behavior is shaped by the interplay of conflicting motivations. When the orientation toward living predominates, psychiatric patients may be less likely to attempt suicide. But when there is a moderate to strong desire to die and not even a weak desire to live, a higher risk for suicide occurs.

There are several limitations of the present study. First, given that the study group was predominantly composed of educated Caucasian Americans, caution should be used in generalizing the conclusions of this study to minority psychiatric populations. Second, there is the possibility of underreporting of deaths by suicide (8), and it is likely that additional suicides may have occurred after the observation period. Third, the prediction of suicide for individual patients remains problematic because of the low base rate for this behavior. The present study, however, suggests that the wish to die/wish to live index may be useful in assessing suicidal risk on repeated administrations over time (9). It is recommended that patients who score high on this index receive a more comprehensive assessment for the risk of suicide.

Received May 27, 2004; revision received Jan. 19, 2005; accepted Feb. 27, 2005. From the Department of Psychiatry, University of Pennsylvania; the Department of Psychiatry, School of Osteopathic Medicine, University of Medicine and Dentistry of New Jersey, Stratford, N.J.; and the Department of Graduate Psychology, James Madison University, Harrisonburg, Va. Address correspondence and reprint requests to Dr. Brown, Department of Psychiatry, University of Pennsylvania, 3535 Market St., Rm. 2030, Philadelphia, PA 19104–3309; [email protected] (e-mail). Supported by an NIMH grant (R37 MH-47383).