A research methodology for the development of risk assessment tools in social work practice.

Abstract

A social work professor uses a series of 3 studies to develop test and revise a prediction model of oral contraceptive (OC) continuation among adolescents 3 months after beginning OC use. She uses discriminant analysis to select those variables which best predict return to the family planning clinic for follow up care and continued OC use. It identifies sexually transmitted disease irregular menstrual periods a Pap smear >1 year before the 1st clinic visit considerable general health problems referral for additional medical tests and long interval since 1st coitus and seeking birth control as the variables of teenagers least likely to return to the clinic. The initial model includes demographic and medical history data plus the 6 variables. It shows that the variables could moderately predict discontinued OC use (66.2% of correctly predicted cases). Next the researcher revises the model by adding psychosocial variables and more variables (score on personal pregnancy risk scale score on costs of pill use scale number of options generated and score on decision to use pill scale). The 8 variable model correctly predicts 78.6% of cases (false negative rate=42.9% and false positive rate=9.4%). This model achieves 75% of correctly predicted cases with another set of data and false negative and false positive rates of 13% and 40%. This final model could predict outcome better than chance alone but it remains susceptible to some error. This risk assessment approach can be used to screen for other behaviors. This methodology minimizes the false negative predictions. This is important because false negatives can lead to stigmatization and an adverse effect. This multifactorial model can help social workers to develop more effective prevention strategies.