Method The intervention counties received supervisor and manager training in recruitment, selection and retention skills; toolkits and technical assistance for transfer of learning; and resources such as a Realistic Job Preview DVD. Data collected from all 34 counties included an online survey administered five times in four years to child welfare workers. The survey of 101 Likert-type items asked about worker perceptions, characteristics and attitudes shown from previous research to be related to retention, from which seventeen scales were created, including an Intent to Leave question. The second source of data, the human resource (HR) database, was created to collect basic employment information prospectively from December 2005 through August 2008.
Results Results of the survey were analyzed using multi-level modeling procedures with scale scores. Following the intervention, analysis of pre- post-intervention data from 815 workers found significant differences between the intervention and control groups on 11 outcomes, including self-efficacy, role clarity, organizational commitment, shared mission, supervisor practice support and intent to leave. These results support the assertion that intervention counties will show statistically significant differences from the control counties on relevant survey scales. A Cox regression survival analysis was performed to assess the effectiveness of the intervention, and the results showed that the intervention was statistically significant in stemming turnover. Survival analysis indicated that the intervention group workers were 44% less likely to experience an undesirable exit than workers in the control group.
Implications We can conclude from these analyses that a rigorous research methodology can be used to test the effectiveness of a training intervention in stemming unwanted turnover in child welfare. The findings have implications for such interventions as supervisor training, improved recruitment and selection methods and agency climate changes.