ABSTRACT Risky alcohol use is a serious public health problem contributing to significant mortality, morbidity and healthcare costs each year. Although it is common among primary care patients – around 20% of adult primary care patients in the U.S. report unhealthy alcohol use – it is rarely addressed in primary care. Risky use is even more prevalent among those with chronic medical conditions such as hypertension, diabetes, and depression, and can affect disease progression and severity, self-management and treatment outcomes. Little is known about the effectiveness of brief intervention (BI) on drinking and health outcomes when delivered as part of mainstream primary care. This population-based, secondary analysis study examines the impact of an ongoing system-wide alcohol screening, brief intervention and referral to treatment (SBIRT) initiative in a large health care delivery system, Kaiser Permanente Northern California, which in 2013 incorporated SBIRT into its adult primary care workflow. Using electronic health record (EHR) data from 1/1/2014 to 06/30/2019 for the approximately ~479,070 members identified as unhealthy drinkers, about 191,628 of whom will have had return primary care visits and thus follow-up data, we will examine the effects of receiving a BI on the drinking outcomes, health outcomes, and costs and health services utilization among those screening positive for risky drinking compared to those not receiving a BI, in order to understand how, for whom and under what circumstances BIs work for risky drinkers. Our study addresses many of the gaps in the literature, including generalizability to real-world healthcare settings where screening and BI occurs as part of routine healthcare delivery and data is collected as part of the care delivery process by clinicians, going beyond the populations studied in previous RCTs of alcohol screening and brief intervention. We will include all adult primary care patients identified as risky drinkers, with all levels of drinking severity, many of who have chronic medical and psychiatric problems. Our large sample’s ethnic and racial diversity will allow us to examine differences in BI effectiveness by age, gender, and ethnicity, which will provide critical information for addressing disparities in the detection and early intervention of alcohol problems in vulnerable populations. The availability of demographic, clinical comorbidity and services use data in the EHR provides us the opportunity to examine the dynamics of BI and its impact on health and services more holistically. We use an innovative model to conceptualize the relationships between BIs and outcomes, and novel statistical methods to make causal inference from this observational study using marginal structural models, and we address selection bias using inverse probability weighting. Our research hypotheses of reduced excessive alcohol use, improvement in health outcomes and decreases in unnecessary healthcare costs and utiliz...