Background: Effective strategies to improve diagnostic stewardship around C. difficile infection (CDI) remain elusive. Electronic medical record-based solutions, such as ‘hard’ and ‘soft’ stops, have been associated with reductions in testing, but may not be sustainable due to alert fatigue. Additionally, data on the potential for undertesting, missed diagnoses, and the implications regarding patient harm or clusters of transmission are limited. In this study, we assessed the efficacy of a behavioral approach to diagnostic stewardship, while monitoring for unintended consequences. Methods: This quality improvement study was conducted January 2018–May 2019; baseline period: January–April 2018, implementation period: May–December 2018, sustainment period: January 2019–May 2019. First, we conducted an internal analysis and identified 3 barriers to appropriate testing: clinician’s perceived risk of CDI, inconsistent definition of diarrhea, and lack of involvement of nurses in diagnostic stewardship. A multidisciplinary team to address these barriers was then convened. The team utilized the Bristol stool scale to improve the reliability of diarrhea description, and created a guideline-concordant testing algorithm with clinicians and nurses. The primary outcome was the number of tests ordered. The secondary outcomes were the proportion of inappropriate tests and the proportion of delayed tests. Delayed tests were defined as CDI-compatible diarrhea based on the algorithm where the test was sent >24 hours after symptom onset. Results: During the baseline period, we detected no significant change in number of tests ordered month to month, with 194.2 tests ordered per month on average. During the postimplementation period, the number of tests ordered decreased by ~4.5 each month between January 2018 and May 2019 (P < .0001). The proportion of inappropriate tests steadily decreased from 54% to 30% across the 3 study periods, and the number of delayed testing changed from 11% to 1% then increased to 20% in the sustainment period. There were no cases of toxic megacolon associated with delayed testing. Conclusions: The decision to test for CDI is complex. Interventions that address this issue as a simple ‘right’ and ‘wrong’ fail to address the root cause of CDI overdiagnosis, and they have no embedded mechanism to detect unintended consequences. Our study demonstrates that by taking a behavioral approach and addressing clinicians’ safety concerns, we were able to sustain a significant reduction in testing. We could not determine the significance of the increase in delayed testing given the low numbers; however, further studies are needed to evaluate the safety of CDI reduction strategies through diagnostic stewardship only.