We use a lab-in-the-field experiment to examine the effectiveness of accountability systems that rely on patient reporting in Kenyan health clinics. We recruit patients and health care providers from public and private health clinics to play a series of modified Trust Games. In the game, patients can send money to providers, who are then able to reciprocate. Patients can then file complaints if they are unhappy with the provider's level of reciprocity. We examine patient and provider behavior in a system where complaints lead to non-monetary consequences in the form of disclosing the complaints to professional peers, a system where complaints lead to monetary penalties, and a system where there are no direct consequences on providers, such as standard complaint boxes (our "control"). We focus on provider reciprocity and patient reporting (or complaining) as our primary behavioral measures in the game. Combining the experimental variation in provider consequences with non-experimental variation in provider and client characteristics such as sector of work, and the existence of personal relationships between clients and providers, we find that: 1) disclosing patients' complaints to providers' professional peers increases providers' pro-social behavior toward patients as much as imposing monetary penalties based on patients' complaints; 2) when complaints lead to tangible consequences (either monetary or non-monetary) for providers, patients are less willing to file such complaints, mainly due to the existence of personal relationships with providers. Overall, our findings support the implementation of citizen reporting systems that leverage peer pressure and reputational concerns.