The use of nursing assistants has increased across health systems in the past 20 years, to alleviate licensed nurses' workload and to meet rising health care demands at lower costs. Evidence suggests that, when used as a substitute for licensed nurses, assistants are associated with poorer patient and nurse outcomes. Our multimethods study evaluated the impact of a policy to add nursing assistants to existing nurse staffing in Western Australia's public hospitals, on a range of outcomes. In this article, we draw the metainferences from previously published quantitative data and unpublished qualitative interview data. A longitudinal analysis of patient records found significantly higher rates adverse patient outcomes on wards that introduced nursing assistants compared with wards that did not. These findings are explained with ward-level data that show nursing assistants were added to wards with preexisting workload and staffing problems and that those problems persisted despite the additional resources. There were also problems integrating assistants into the nursing team, due to ad hoc role assignments and variability in assistants' knowledge and skills. The disconnect between policy intention and outcomes reflects a top-down approach to role implementation where assistants were presented as a solution to nurses' workload problems, without an understanding of the causes of those problems. We conclude that policy makers and managers must better understand individual care environments to ensure any new roles are properly tailored to patient and staff needs. Further, standardized training and accreditation for nursing assistant roles would reduce the supervisory burden on licensed nurses.