Psychiatric Characteristics of Homicide Defendants

We coded psychiatric evaluation files (N=278) and available matching court case files (N=208) for unselected defendants charged with criminal homicide in Allegheny County, Pennsylvania, between 2001 and 2005, excluding cases of vehicular homicide. During the sampling time frame, local court rules mandated that all individuals arrested on a charge of criminal homicide receive a psychiatric evaluation within 48 hours of arrest to assess competence to stand trial and acute treatment needs. Behavior Assessment Unit staff psychiatrists and forensic psychiatry fellows conducted these clinical interviews, which took approximately 30 minutes.

In approximately 70% of the sample, staff social workers completed a template form to collect collateral information; in the remainder, the collateral contact could not be reached or refused to cooperate. In some cases, the evaluators also had access to affidavits of probable cause, records from a state hospital, and copies of prior assessments for the court (including reports of prior arrests). Official criminal records and results of psychological testing were not routinely available.

A standard reporting format was followed in the resulting three- to six-page narrative assessments. These contained demographic data, circumstances of the current arrest, legal history, mental health history, medical history, social history, a mental status examination, and any diagnoses established on axes I, II, and III. Opinions regarding competence to stand trial were explicitly stated, as well as disposition recommendations, if relevant. Structured diagnostic instruments were not used.

All records were the property of the court and were not part of the individual’s medical record. Their use in this study was authorized by the court’s president judge. The University of Pittsburgh Institutional Review Board reviewed and approved a waiver of informed consent for examination of these court records.

Two individuals trained and supervised by the study’s senior investigators coded the records. The investigators and coders held consensus meetings biweekly to review the coding process and to resolve ambiguities. Coders reviewed data for consistency when coding rules were revised.

Stata/SE 10 and 12 (StataCorp, College Station, Tex.) were used for data analyses. Chi-square or Fisher’s exact tests were performed as appropriate for cross-tabulated categorical data. Continuous variables were compared with t tests or between-group analysis of variance. Given low outcome base rates and small sample sizes, logistic models were examined using a penalized log likelihood adjustment for complete or partial separation to correct bias (20, 21). Likelihood ratio chi-square tests were performed to explore differences between nested groups of variables in penalized logistic models, after examining patterns of missing data.

This examination of all homicide defendants in a large U.S. urban jurisdiction over a 5-year period highlights notable differences from previous work that are likely attributable to sampling. To our knowledge, this is the first study to provide a view of the relationship between psychiatric disorders and homicide in a comprehensive U.S.-based sample unaffected by referral bias. While similar studies have been conducted in Europe, social and cultural differences as well as lower per capita homicide rates limit the applicability of those findings to the United States. A study by Frierson and Finkenbine (17) most closely resembles this study in its aims, but the defendants in that investigation had been referred for assessment because of concerns regarding competency to stand trial or criminal responsibility.

We found a lower overall rate of mental disorders (58%) than Frierson and Finkenbine (91%) (17). The observed lower prevalence of axis I diagnoses in general and of psychotic disorders in particular is reasonable, given the unselected nature of our sample. Our observed rate of psychotic disorders (4%) was below the rate of approximately 6.5% observed in some other studies (19, 22). The prevalence rate of mental disorders (58%) in this sample of defendants, however, still greatly exceeded the rate of 25% found in the general population in the United States (23).

We found a higher rate of personality disorders than Frierson and Finkenbine (24% compared with 13%) (17). Personality disorders may not manifest with the same type of overt symptomatology as axis I disorders, and defendants with personality disorders may not be referred for psychiatric assessments as often as those with other diagnoses. The low rate of identified antisocial personality disorder (5%) seen in this study may reflect the lack of a structured personality disorder instrument, and clinicians may be reluctant to assign such a potentially pejorative diagnosis prior to a defendant’s court appearance.

Many of the patterns observed in this study align with those seen in previous samples of homicide defendants. First, males were overrepresented, comprising 93% of the defendants. Second, African American defendants were overrepresented at 79% of the overall sample but only 13% of the local population. Moreover, African American defendants were more likely than non-African American defendants to have used a firearm and to have killed a man rather than a woman. We speculate that these differences reflect higher rates of drug- and gang-related homicides associated with younger African American males in this locale and that jurisdictions with different community characteristics and crime patterns would show different results. Third, the rates of substance dependence or abuse in this study are consistent with those reported in previous studies.

It is notable that clinical variables, such as axis I diagnoses, were not associated with offense characteristics or case outcomes when demographic and historical characteristics of the cases were included in the models. In particular, while age and race were significantly related to the use of a firearm, the addition of clinical variables to demographic and historical variables did not improve model fit. Furthermore, a model including demographic/historical and clinical variables did not significantly predict a guilty verdict, suggesting that case-specific factors were more salient in these determinations. Of interest, though, is that agreeing to a guilty plea was positively associated with a past history of mental illness but inversely related to a past history of violent crime. Sentences for defendants with an axis I diagnosis, particularly of a substance use disorder, tended to be longer. These findings raise questions regarding potential biases toward mentally ill individuals in plea offers and sentencing practices.

A higher prevalence of non-substance use axis I disorders and mood disorders in defendants over age 40 in this study warrants further exploration. The patterns noted here could indicate risk factors associated with homicide related to specific clinical issues in older as distinguished from younger persons. If additional studies were to reveal that older defendants had a prior history of treatment and distinctive victim characteristics, this could have implications for more refined risk assessment and preventive treatment interventions.

The rate of previous treatment observed in this sample raises issues relevant to mental health policy. Although 53% of the sample were diagnosed with an axis I diagnosis (including substance use disorders), less than half of these individuals had ever been hospitalized. Also, among those with an axis I diagnosis, only 8% had received any treatment in the 3 months preceding the homicide offense. Moreover, this low frequency of recent psychiatric treatment differed markedly by race; only 3% of African American defendants with at least one axis I diagnosis received treatment in the 3 months preceding the offense, compared with 28% of non-African American defendants. Widespread disparities in access to care and cultural differences regarding help seeking are likely explanations for this difference. The low rate of treatment in the months preceding the offense, however, highlights the need for enhanced engagement of high-risk individuals (especially during times of emotional crisis) if mental health care providers expect to have an impact on serious violence.

A number of limitations of this study should be kept in mind. Although the lack of referral bias and a relatively large sample size were distinct advantages, a lack of highly structured diagnostic assessments and recording of case information introduced unknown biases in the determination of diagnoses. Our impression is that the procedures used identified major psychiatric disorders (e.g., schizophrenia, depression) with sufficient accuracy and substance use disorders at a slightly reduced rate of accuracy. As a result, we have more confidence in results regarding broad diagnostic groups or the presence of a disorder.

Variability in the narrative reports also produced incomplete information on some variables. Although the data in models appear to be missing at random, possible biases in estimates of effects could still exist. In addition, the size of the sample is small for some of the logistic regression models. A larger sample size might have allowed detection of other relevant main effects or interactions. Finally, much of the information and the diagnostic assessment were based on the report of defendants who were interviewed immediately after arrest in a pretrial evaluation, and these reports may have been biased by perceived self-interest. A structured prospective study of these and other factors might produce a different picture.

Studies distinguishing defendants accused of homicides associated with drug or gang activity from other homicides would be helpful in identifying what, if any, clinical factors are related to lethal violence, the targets of such violence, and possible methods of prevention. Clinical factors appear to play a limited role in homicide in general, but their importance in a minority of specific cases has yet to be fully explored in systematic investigations. Clarification of these dynamics would improve our ability to educate courts and juries regarding the potential role of mental illness in homicide.

The authors thank Judge Donna Jo McDaniel (president judge of Court of Common Pleas, Allegheny County), Judge Jeffrey Manning (administrative judge of Court of Common Pleas, Criminal Division, Allegheny County), the staff of the Allegheny County Behavior Assessment Unit, and the staff of the Department of Records, Allegheny County, for their assistance and support. They also thank Jonathan Yabes, Ph.D. (from the Center for Research on Health Care Data Center at the University of Pittsburgh School of Medicine) and Barbara Hanusa, Ph.D., for statistical advice and Angela Martone, Molly Delaney, Koraleigh Gritz, and Adam Ligas, M.D., for their coding work.

The views expressed in this article are those of the authors and do not necessarily reflect the views of the Department of Veterans Affairs or the U.S. government.