When making life and death decisions about when to hospitalize an in- dividual who may be at risk for suicide, a clinicians' choice often relies more on 'best guesses' than on solid, evidence based knowledge. While there is extensive evidence about long term risk factors that predict lifetime risk but not necessarily immediate danger to self, far less is known about sub-acute risk factors, predictors of serious suicidal behaviors within weeks to months. Even less is known about acute risk factors, predictors of serious suicidal behaviors within hours to days. We examined the prevalence of acute risk factors (suicide warning signs) in a high-risk population of veterans. We also examined agreement between self-report (a self-administered survey), and psychiatrist report, of the suicide warning signs of suicidal ideation (SI) and behaviors. We then looked at whether any single or cluster of suicide warning signs differentiated veterans who were hospitalized for a suicide attempt (SA) or severe SI the day of the survey, from all other veterans in the study. We also looked at whether suicide warning signs or other characteristics were associated with an SA or hospitalization for severe SI in the following 12 months. Participants were 430 men and 52 women who presented to the walk-in Psychiatric Emergency Clinic at the Veterans Administration Health Care System in San Diego, California between January and May 2010. In the prevalence portion of the study we found that more than half (52%) of participants reported SI or suicidal behaviors in the past week. Other suicide warning signs were also highly prevalent with the lowest prevalence, 19%, for 'hurt self' or 'reckless behavior'. Prevalences of suicide warning signs were significantly higher for veterans with a major depressive episode (MDE by PHQ-9 score), positive screening for post-traumatic stress disorder (PTSD), or both, compared to none. Agreement between self-report and psychiatrists' report of SI and behavior was generally low with veterans rating SI and suicidal behaviors as significantly more severe than psychiatrists. Veteran ratings were consistently more severe than psychiatrists ratings even when veterans with characteristics that may have led to over-reporting (e.g. homelessness) were removed from the analysis. In univariate analysis, many of the individual suicide warning signs and the two clusters of warning signs formed from factor analysis, were strongly associated with immediate hospitalization for a suicidal crisis (SA or severe SI). In addition, a smaller number of warning signs and clusters were also associated with a suicidal crisis in the following 12 months albeit at a much lower magnitude in both univariate and multivariate analysis. In multivariate analysis, veterans who were hospitalized immediately had 13 fold higher odds of self-reporting feeling out of control compared to veterans not hospitalized immediately. The clinical utility of warning signs is attenuated by the high prevalence of all of the warning signs in veterans who were not hospitalized immediately or in the 12 month follow-up study. Moreover, only 3.1% of veterans (n = 15) did not endorse any warning signs and the mean number of signs endorsed was 7.6. When history of SA was included with the suicide warning signs, only 2.3% (n = 11) did not make any endorsements. Overall the frequency of the suicide warning signs was highest in veterans hospitalized for SI or SA the day of the survey, intermediate for veterans hospitalized during the 12 month follow-up and lowest for veterans with no SA or hospitalization during the study period.