Prediction of patients who are dangerous to others

Abstract

PROBLEMThe identification of psychiatric patients who are dangerous to others is an important but often very difficult task. Because actual assaultive behavior among psychiatric patients is a relatively rare occurrence, however, study of this phenomenon and development of predictor measures require large numbers of patients. Recent state-wide automation of demographic and clinical information for all patients within the Missouri Division of Mental Health (DMH) ( 8 , l l~ n ) has provided one basis for such large sample studiec:". 2 , 5 . 6 * g ) . METHODFrom the Missouri DMH clinical data base, 5525 patients were identified whose automated records included demographic, admissions and initial mental status information, in addition to explicit criterion information as to the presence or absence of attempts to harm, or in fact having harmed others, ix., "assaultive behavior". Demographic and admissions information, including such things as sex, race, age, religion, education, marital status, occupation, referral source, previous psychiatric care, type of admission, and hospital involved, was obtained originally on an Admission Face Sheet ( l o ) . Mental status information was recorded originally on a 111-item checklist (11) specifically designed and developed for the DMH automated psychiatric information system. Criterion information, i.e., whether the patient had attempted to harm or in fact harmed others, was recorded originally on an Emergency Room or Admission Checklist ( l o ) , which also was developed for the automated information system. This form includes approximately 100 descriptors that have to do with preadmission behavior rather than behavior observed during the interview, as is the case with the mental status checklist. The principal focus is on symptoms that occur in the month prior t o admission, although for about one-third of the items the interviewer is asked to consider the patient's entire life. Typical items include suicide attempt, arrests, definite drinking problem, extravagance with money, poor personal hygiene, truancy, attempts to harm others, use of barbiturates, and poor job performance.The patient sample consisted of psychiatric inpatients admitted to four Missouri DMH mental health centers. Approximately 60% of the sample was male average age was 43 and the average education was 10 years. Approximately 25% of the sample never had been married, 34% currently were married, 28% were divorced or separated, and 10% were widowed. First admissions accounted for 56% of the sample; 72% were voluntary commitments. Of the sample, l60/0 was diagnostically classified as schizophrenic, 13% as affective disorders, 17% as organic brain syndromes, 3% as mentally deficient, 4% as neurotic, and 5% as personality disorders; 27% of the sample had alcohol or drug abuse diagnoses.The 5525 patients were assigned randomly to a derivation sample (N = 2763) and a validation sample (N = 2762), and a stepwise linear discriminant function (LDF) was used t o derive and cross-validate prediction equations with rega...

A Follow-Up of Patients Found “Unfit to Stand Trial” or “Not Guilty” Because of Insanity*

“…The number and kind (economic) of new offences committed upon release by their sample closely resembled those committed by persons discharged from prison. Other studies of patients released from maximum security (2,4,6,7,13,18) have found varying failure rates. Because these studies were so diverse in methodology, varying failure rates across studies are to be expected.…”

Predicting violence in schizophrenic and non-schizophrenic patients: A prospective study

“…Efforts to develop predictive scales have fared better in studies of psychiatric patients. Hedlund, Sletten, Altman, and Evenson (1973), with an extremely large number (5,525) of inpatients in psychiatric facilities, used discriminant analyses with 100 demographic, admissions, and initial mental status variables to predict whether the patient had attempted to harm or had actually harmed others. Predictive equations were developed from half of the sample and validated on the remaining half.…”

Politique institutionnelle de libération : identification des individus dangereux. Une revue de la littérature

“…Dans une étude auprès de 5 525 patients, Hedlund, Altman et Evenson (1973) ont tenté de prédire à posteriori quels patients avaient été dangereux envers les autres en utilisant un grand nombre d'items informatisés sur l'état mental ainsi que des variables démographiques, concernant l'admission ou liées au diagnostic. Le taux de faux positifs produit par l'équation de régression variait de 60 % pour la prédiction à posteriori du comportement violent à 80 % pour le fait d'avoir réellement blessé quelqu'un.…”