Vertical integration is one possible way to improve the performance of a health care system; however, its effects are inconsistent, and there is a lack of evidence from undeveloped nations. This study aims to systematically review the evidence regarding effects of vertical integration on health care system in China. We searched PubMed, Embase, Cochrane Library, Web of Science, ProQuest Health & Medicine Collection, CNKI and Wanfang databases from April 2009 (initiation of new health care reform) to May 2021 for randomized controlled trials (RCTs), controlled before and after (CBA) trials, cohort studies, and interrupted time series (ITS) trials. Vertical integration in the included studies must involve both primary health institutions and secondary or tertiary hospitals. After screening 3109 records, we ultimately analysed 47 studies, including 27 CBA trials, 18 RCTs and 2 ITS trials. The narrative synthesis show that all but three studies indicated that vertical integration improved efficiency (utilization and cost of health services), quality of public health services and medical services, health provider-centred outcomes (knowledge and skill) and patient-centred outcomes (patients’ clinical outcomes, behaviour and satisfaction). Although the heterogeneity of vertical integration interventions across different studies, the meta-analysis reveal that it lowered diastolic blood pressure (MD -8.41, 95% CI -15.18– -1.65) and systolic blood pressure (MD-5.83, 95% CI -9.25– -2.40) among hypertension patients; and it lowered HbA1c levels (MD -1.95, 95% CI -2.69– -1.21), fasting blood glucose levels (MD -1.02, 95% CI -1.53– -0.50) and 2-hour postprandial blood glucose levels (MD -1.78, 95% CI -2.67– -0.89). The treatment compliance behaviour was improved for hypertension participants (RR 1.08, 95% CI 1.04–1.13) and for diabetes patients (RR 1.32, 95% CI 1.08–1.61). Vertical integration in China was prone to improve the efficiency, quality of care, health provider-centred outcomes and patient-centred outcomes, but high quality of original studies are highly needed.