Background: Most of the integration-related policies are facing implementation problems. During 2012/07-2014/12, we implemented an trial-based intervention in Qianjing, a rural district in southwestern China, to test the effects of an innovative integrated care delivery model targeted for rural patients (RICDM ) with hypertension. Using the difference-in-differences analysis, we found that the blood pressure did decrease in treatment group compared to the control group however the health-related quality of life did not, which violated our assumptions. To explore whether the provider performance interfered with the outcomes, we did a process evaluation mainly to compare their behaviors during trial implementation. Methods: All the doctors’ behaviors were bimonthly-recorded using diaries, and their attitudes were evaluated using questionnaires adapted from D’Amour’s inter-professional collaboration model. We firstly observed the referral amounts and provider communication frequency on four inter-professional cooperation types (village to town, town to county, county to town, and town to village – the former indicated the initiated party). Then we conducted interviews with the doctors, investigating the following aspects that potentially affected the collaboration: willing to cooperate, actual motivation, mutual trust, personal relations and ability to cooperate. All variables were compared between groups before and after intervention. After then we did a regression analysis examining the relations between the cooperation level and the identities of doctors. The implementation degree was also estimated by multiplying the percentage of reach scope for each step of trial process. Results: 120 doctors were included. There were no significant differences between treatment group and control group either on the referral amounts or the communication frequency (all P >0.05). Most of the cooperation aspects showed a low- and middle-level performance (>90%). The collaboration ability in treatment group was significantly higher than control group (P 0.05). And the doctors from county hospitals showed a lower performance in each of the aspect compared to doctors from town hospitals and village clinics. Associated with the professional identity, the medical staff showed relatively low interests in collaboration compared to primary care professionals. By observing and analyzing the processes of exposure, enrollment, reach and follow-ups, we had found out the implementation degree to be only 61%. For each part, the patient were reached out by 100%, the supplier 72%, and the hospitals 85%. Lessons Learned: The RICDM improved patient health, but did not fully achieve desired effects. We inferred that one possible reason might lied in the providers’ reluctance to cooperate. By furtherly investigating into the provider behaviors and implementing process, we had proved our guess and the actual model effect might therefore be underestimated. However the study did not successfully connect the health-related outcomes with the provider behaviors through statistical analysis, so we could not reach a direct conclusion that the incomplete implementation was the exact reason for imperfect model effects. Further studies need to be proposed and also potential bias from provider side should be properly dealt with in the stage of study design.