Assessing Predictions 1 Assessing Predictions of Domestic Violence: The Pre-employment Psychological Evaluation Stephen F. Curran Abstract: Domestic violence emerges as one of the characteristics that can be isolated by a statistical technique known as receiver-operating characteristic (ROC) analysis. This paper also describes the psychometric problems associated with pre-employment predictions as well as the recently revised Pre-employment Screening Standards developed by the Police Psychological Services Section of the International Association of Chiefs of Police. KEY WORDS: screening, pre-employment evaluation, psychological evaluation, guidelines, violence prediction. Address correspondence concerning this article to Stephen F. Curran, Ph.D., Atlantic OccuPsych, Post Office Box 4071, Timorium, Maryland 21094-4071. Assessing Predictions 2 ASSESSING PREDICTIONS OF DOMESTIC VIOLENCE: THE PRE-EMPLOYMENT PSYCHOLOGICAL EVALUATION The selection of law enforcement personnel who exhibit stable emotional and behavioral functioning has become critical to policing agencies. Excessive force complaints, community policing initiatives, police suicides, violence against intimates and allegations of negligent hiring have demonstrated that hiring decisions must take into account the psychological characteristics of an applicant. During the past 20 years, the field of police psychological screening has expanded dramatically (Curran, 1996). In 1986, the Police Psychological Services Section of the International Association of Chiefs of Police (IACP) adopted its first set of guidelines to assist police chiefs and psychologists conducting programs to evaluate police applicants (Solomon, 1992). These guidelines provided a generic approach to defining elements of an evaluation program facilitating the valid and reliable selection of psychologically suitable police officers. Since then, the majority of police agencies now conduct pre-employment psychological evaluation of applicants. A number of surveys have assessed the methodologies employed by psychologists. Scrivner’s (1994) report on the use of excessive force emerged as the most notable effort. Perhaps the most significant development since the 1986 IACP guidelines involved the enactment on July 26, 1990, of the Americans With Disabilities Act (ADA, 1990). As of 1992 the ADA applied to most law enforcement agencies. This federal law, designed to protect individuals with disabilities, posed many challenges to administrators of police agencies and psychologists conducting pre-employment evaluations. While the initial Assessing Predictions 3 fears about active drug users and psychotics succeeding in gaining employment have proven unfounded, the law has profoundly affected pre-employment procedures. The increased tendency to view pre-employment psychological screening as a Apass or fail” evaluation has emerged as an issue. This approach alters the threshold of suitability and, consequently prediction accuracy. The Police Psychological Services Section of the IACP came into existence in 1984. In his history of the Section, Shaw (1992) noted that the formation resulted from police psychologists’ search for an organization within which direct, rapid, two-way consultation between police chiefs and psychologists could occur. The IACP commenced the process of revising its pre-employment psychological guidelines in 1994. This partially addressed aspects of evaluations affected by the ADA and also incorporated generally accepted standards of practice. The revised guidelines received approval from Section members during the spring of 1998. The psychologist engaging in pre-employment psychological evaluations must know the issues inherent to police screening. PRE-EMPLOYMENT GUIDELINES A brief review of three of the twenty-two guidelines (See IACP paper at end for complete guidelines) relating to the discussion of predicting domestic violence follows. The introductory section of the guidelines defines public safety positions as the type of employment to which they apply. While the guidelines prominently address applicants for positions with the authority to make arrests, they apply to related law enforcement positions as well (e.g., correctional officers, police aides, and emergency dispatchers). Further, these guidelines constitute recommended rather than mandatory Assessing Predictions 4 policy. Guideline item numbers 16 and 17 cover two important considerations when evaluating an applicant; they concern the need to focus on essential job functions rather than diagnostic issues. 16. While a clinical assessment of overall psychological suitability may be made, clinical diagnoses or psychiatric labeling of candidates should be avoided when the goal of the assessment is to identify candidates whose psychological traits may adversely affect specific job performance. In all cases, the screening should be focused on an individual candidate’s ability to perform the essential functions of the position under consideration. 17. Specific cut-off scores should be avoided, unless there is clear statistical evidence that such scores are valid and have been cross-validated in research studies by the test developer and/or in the agency where they will be used. If cutoff scores are utilized, the report should acknowledge their use and the basis for using the specific cut-off level. Conclusions concerning a candidate’s qualifications should be based on consistencies across data sources rather than on a single source. The guidelines recommend a description of psychological traits identified from more than one data source. Linking these traits to potential deficits in behavioral functioning (e.g., impulse control, motor vehicle violations, and job absences), emotional stability (e.g., excessive sick leave, sleep and appetite changes) and social judgement attributes make up the primary factors for evaluation. Guideline number 14 states, interviews should be scheduled to allow for sufficient Assessing Predictions 5 time to cover appropriate background and test results verification. Evaluators who have conducted five to ten minute interviews of applicants have precipitated a number of incidents including the initiation of criminal investigations. Because of such concerns, many federal, state, and local agencies require documentation of 45-minute interviews. This guideline recommends interviews of at least 30 minutes. Some interviews may require 90 minutes to adequately assess the behavioral history of an applicant in conjunction with psychological testing. Havassy (1996) recommends a semi-structured interview. Interviews should have a standard format to reduce interviewer bias and allow for probing in areas related to essential job functions. Psychological testing results comprise an integral component of the interview. Emotional, behavioral, and social psychological characteristics remain essential to successful performance of law enforcement tasks and form the main focus of an interview. Emotional factors include stability in impulse and anger control, flexibility, maturity, self-esteem, and management of stress. Behavioral stability elements include an applicant’s capacity to follow rules and regulations, response to trauma, prior substance use patterns, and job adjustment history. Social factors relate to work and family relationships, interpersonal communication skills, assertiveness, empathy, and cultural sensitivity. These domains comprise the breadth of interview elements designed to elicit, in conjunction with testing data, an adequate representation of an applicant’s capacity to perform the varied duties of a law enforcement officer. PREDICTIONS OF FUTURE BEHAVIOR: MOVING BEYOND WIZARDRY Mental health professionals have liability concerns. To maintain evasiveness, they Assessing Predictions 6 often choose to hide behind the protective cloak of the law, which cannot predict future behavior. This stance contradicts available data. We can predict future behavior. The question really concerns the confidence levels we want to apply to those predictions. This paper cannot include a complete discussion of quantifying prediction accuracy. Interested readers can refer to Mossman (1994) for a detailed description of issues in predictions of behavior. The central point of this paper focuses on testing and interview data as potent predictors of subsequent performance. The pre-employment evaluators using this data for employment recommendations need to consider the tradeoffs (sensitivity and specificity). Some relevant examples of successes serve to illuminate behavior predictions follow. Jones, Beidleman, and Fowler (1981) used multivariate procedures to correctly classify 72.9% of inmates who would commit violent acts and 80.6% of inmates who would not commit violent acts while incarcerated based on scales of the Minnesota Multiphasic Personality Inventory (MMPI) alone. Lidz, Mulvey, and Gardner (1993), in a prospective study of emergency room psychiatric patients, found mental health clinicians accurately predicted future violence in 63% of male patients. However, their predictions about women and subsequent violence failed to do better than chance. Sandberg, McNiel, and Binder (1998) reported on data of particular interest to the discussion of domestic violence. These researchers examined psychiatric inpatients who stalk, harass or threaten hospital staff after discharge. They found these patients to have the diagnosis of personality disorder or paranoid disorder; and a pre-admission history of fear-inducing or physically assaultive behavior. Inwald, Resko, and Trainer (1997), using the Inwald Personality Inventory, conducted a study of 21 known violent Assessing Predictions 7 police officers for which they had access to pre-employment testing data. A discriminant analysis correctly identified 85.5% as either violent or nonviolent. Specificity and sensitivity continue to comprise the primary issue related to predicting domestic violence in police officer applicants. Specificity relates to the number of true negatives (those applicants not violent) divided by the sum of true negatives and false positives. Sensitivity involves the number of true positives (predicted violent applicants who actually committed violence) divided by the sum of the true positives and the false negatives. Predictions less influenced by base rates, especially the low estimated occurrence rates for domestic violence hold a great deal of interest for us. Of course, 100% predictive accuracy lies beyond the reach of psychologists now and in the reasonably foreseeable future. Human behavior eludes complete quantification. A statistical procedure, known as receiver operating characteristic (ROC) analysis accommodates calculating sensitivity-specificity pairs independent of base rates. These pairings establish a threshold for accuracy prediction from which measuring prediction above chance becomes possible. This procedure establishes the threshold for an evaluator (receiver) and serves as the detector (an alarm) in decision making. Evaluators using a minimal Apass or fail rating of psychological fitness may favorably recommend more applicants prone to violence. This results from a lower threshold (sensitivity-specificity ratio). More stringent evaluation ratings, such as multi-tiered suitability ratings, may provide above average or superior recommendation ratings to fewer violence prone applicants, but rates low or rejects otherwise suitable applicants who never subsequently engage in violent behaviors. These constitute critical issues for Assessing Predictions 8 evaluators and future research. CONCLUSION This paper introduces the statistical methodology of ROC analysis into the discussion of pre-employment psychological screening of law enforcement applicants. Recognition of this methodology for future research and current practice appears relevant when considering the base rates of domestic violence among law enforcement officers. Mental health professionals have demonstrated in research on other clinical groups a better than random chance of predicting violence. Sound reasoning to support accuracy of prediction by pre-employment evaluators exists. Adhering to current preemployment guidelines generally accepted as the standard of practice will enhance the accuracy of behavioral predictions about future job performance. Assessing Predictions 9 IACP POLICE PSYCHOLGICAL SERVICES SECTION GUIDELINES FOR PREEMPLOYMENT PSYCHOLGICAL EVALUATION SERVICES (1998) OVERVIEW The following statements are guidelines for professional practice in the area of pre-employment psychological evaluations of candidates for public safety positions. These positions include but are not limited to positions where incumbents have arrest authority or the legal authority to detain and/or confine individuals. These guidelines are presented as a recommended professional policy for public safety agencies and individuals who are charged with the responsibility of conducting defensible preemployment psychological screening programs. DEVELOPMENT 1. Pre-employment psychological assessments should be used as one component of the overall selection process. 2. Before conducting their own clinical assessments of candidates, practitioners should be familiar with the research literature available on psychological testing for public safety positions, as well as the state and federal laws relevant to this area of practice, including the Americans with Disabilities Act (ADA). 3. Only licensed or certified psychologists trained and experienced in psychological test interpretation and law enforcement psychological assessment techniques should conduct psychological screening for public safety agencies. 4. Data on attributes considered most important for effective performance in a particular Assessing Predictions 10 position should be obtained from job analysis, interview, surveys, or other appropriate sources. 5. Efforts should be made to provide agency administrators with information regarding the benefits and limitations of psychological assessment procedures so that realistic goals may be set. 6. Provisions should be made for the security of all testing materials (e.g., test booklets). Provisions should also be made for the security of, access to, and retention of the psychological report and raw data. 7. Prior to the administration of any psychological instruments and psychologist interview, the candidate should sign an informed consent to the conditions of the evaluation. TESTING 8. A test battery including objective, job-related, validated psychological instruments should be administered to the applicant. It is preferable that test results be available to the evaluator before screening interviews are conducted. 9. Written tests selected should be validated for use with public safety candidates. 10. If mail order or computerized tests are employed, the licensed or certified psychologist conducting the follow-up interview should verify and interpret individual results. 11. The pre-employment psychological evaluation must be conducted in accordance with the Americans with Disabilities Act (ADA) guidelines after a conditional offer of employment has been tendered. Although some personality tests, and other methods of inquiry, may be conducted at the pre-offer stage, they do not in and of themselves Assessing Predictions 11 constitute the pre-employment psychological evaluation. INTERVIEW 12. Individual, face-to-face, interviews with candidates should be conducted before a final psychological report is submitted. 13. A semi-structured, job-related interview format should be employed with all candidates. 14. Interviews should be scheduled to allow for sufficient time to cover appropriate background and test results verification. EVALUATION 15. Public safety agency administrators directly involved in making employment decisions should be provided with written reports. These reports should evaluate the suitability of the candidate for the position based upon an analysis of all psychological material including test data and interview results. Reports to the agency should contain a rating and/or recommendation for employment based upon the results of the screening, justification for the recommendation and/or rating, and any reservations that the psychologist might have regarding the validity or reliability of the results. 16. While a clinical assessment of overall psychological suitability may be made, clinical diagnoses or psychiatric labeling of candidates should be avoided when the goal of the assessment is to identify candidates whose psychological traits may adversely affect specific job performance. In all cases, the screening should be focused on an individual candidate’s ability to perform the essential functions of the position under consideration. 17. Specific cut-off scores should be avoided, unless there is clear statistical evidence Assessing Predictions 12 that such scores are valid and have been cross-validated in research studies by the test developer and/or in the agency where they will be used. If cut-off scores are utilized, the report should acknowledge their use and the basis for using the specific cut-off level. Conclusions concerning a candidate’s qualifications should be based on consistencies across data sources rather than on a single source. 18. Clear disclaimers should be made so that reports evaluating current emotional and behavioral traits or suitability for a public safety position will not be deemed valid after a specific period of time. FOLLOW-UP 19. Pre-employment test results should not be used for purposes other than making pre-employment decisions and for monitoring the candidate during the probationary period. Follow-up research may be conducted with agency approval and where individual identities are protected. Pre-employment reports should not be used for positions or agencies not expressly considered by the psychologist at the time of the evaluation, nor should they be used for subsequent disciplinary or forensic matters. 20. Continuing collaborative efforts by the hiring agency and evaluating psychologist should be made to validate final Asuitability@ ratings using behavioral criteria measures. 21. Each agency should maintain adverse impact analyses in order to detect any discriminatory patterns of the psychological screening program. 22. Psychologists should be prepared to defend their procedures, conclusions, and recommendations if a decision based, even in part, on psychological results is challenged. Assessing Predictions 13 Assessing Predictions 14 REFERENCES Americans with Disabilities Act of 1990 (1991). Pub. L. No. 101-336, sec. 104 Stat. 328. Curran, S. F. (1996). Positioning psychological services for the 21st century. In J. T. Reese & R. M. Solomon (Eds.), Organizational issues in law enforcement. (pp. 375382). Washington, DC: Federal Bureau of Investigation. Havassy, V. (1996, October) Psychological preemployment practices survey results of IACP PSS members. Paper presented at the meeting of the Police Psychological Services Section of the International Association of Chiefs of Police, Phoenix, AZ. International Association of Chiefs of Police. (1986, revised 1998). Guidelines for providers of pre-employment psychological evaluation services to law enforcement agencies. (Police Psychological Services Section). Alexandria, VA: Author. Inwald, R. E., Resko, J. A., & Trainer, B. (1997). Assessing violent/aggressive characteristics: current studies with public safety officers/violent offenders. Kew Gardens, NY: Hilson Research. Jones, T. Beidleman, W. B. & Fowler, R. D. (1981). Differentiating violent and nonviolent prison inmates by use of selected MMPI scales. Journal of Clinical Psychology, 37 (3), 673-678. Lidz, C. W., Mulvey, E. P., & Gardner, W. (1993). The accuracy of predictions of violence to others. Journal of the American Medical Association, 269 (8), 1007-1011. Mossman, D. (1994). Assessing predictions of violence: being accurate about accuracy. Journal of Consulting and Clinical Psychology, 62 (4), 783-792. Sandberg, D. A., McNiel, D. E., & Binder, R. L. (1998). Characteristics of psychiatric Assessing Predictions 15 inpatients who stalk, threaten, or harass hospital staff after discharge. American Journal of Psychiatry, 155 (8), 1102-1105. Scrivner, E. (1994). The role of police psychology in controlling excessive force. (Research Report NCJ 146206). Washington, DC: U.S. Department of Justice. Shaw, J. H. (1992). Psychological services section history. Police Chief, May, 54. Solomon, R. M. (1992). Developing professional guidelines. Police Chief, May, 5356.