Posted by Alumni from MIT
October 19, 2023
We analyzed the free text of 150,000 reviews written by U.S. nurses from the beginning of the pandemic through June 2023. For each review, we analyzed whether it mentioned one of 200 topics and assessed whether the nurse spoke about that topic positively or negatively. The topics were derived from research on various elements of employees' experiences, including culture, compensation and benefits, work schedules, and perceived organizational support. We clustered related topics together into two dozen broader themes. We used the topics as features in an XGBoost model that predicted each nurse's overall rating of their current or former employer on a 5-point scale. We then calculated SHAP values (which assign an importance value to each feature in a model) to directly compare the relative importance of different topics in predicting job satisfaction (measured by the overall rating each nurse gave their employer). We also used textual analysis to compare how favorably travel nurses... learn more