Text this: Comparing the performance of narrow vs. broad search strategies when using machine learning-based software for title/abstract screening.