Delivery of Smoking Cessation Services and Cessation Attempts Across a Public, Safety-Net Primary Care System

Background: Smoking cessation rates are low in safety-net settings, contributing to high smoking-related morbidity and mortality. Understanding factors associated with cessation attempts can inform interventions. Objective: To evaluate factors associated with smoking cessation attempts. Design: Retrospective analysis using electronic health record (EHR) data on individuals with at least three primary care encounters from 2016 to 2019 in the San Francisco Health Network (SFHN), a network of clinics serving publicly insured and uninsured residents in San Francisco. Participants: Patients engaged in primary care in the San Francisco Health Network. Main Measures: The outcome was recent cessation attempt, dened as change in smoking status from “current smoker” at the index visit to “former smoker” at visit 2 or 3. We measured demographics, tobacco-related comorbidities, and cessation treatment characteristics (i.e., counseling and pharmacotherapy). To better characterize subpopulations that may benet from targeted interventions, we described characteristics of smokers with hypertension, depression, diabetes, or HIV. Key Results: Of the 51,554 adults identied across 15 SFHN primary care clinics, 11,622 (22.7%) were current smokers. Approximately 26% of smokers made a recent cessation attempt. Medical assistant (90%) and provider counseling (73%) rates were high, while behavioral assistant counseling rate (17%) was low. All counseling types had lower odds of cessation attempts in multivariable analysis. Smokers with depression (AOR 1.18, 95%CI 1.05-1.33) and ischemic heart disease (AOR 1.36, 95%CI 1.06-1.74) had higher odds of attempts. Among comorbidity groups, cessation attempts ranged from 21-26%, and smokers with HIV received the lowest rates of cessation counseling. Conclusions: Although rates of basic cessation counseling were high, efforts were associated with lower odds of making a cessation attempt. Using intensive interventions to target populations with comorbidities could be opportunities to increase cessation engagement.


Introduction
Despite decades of efforts to reduce tobacco use in the United States (US), cigarette smoking accounts for one in ve deaths annually and remains a leading cause of preventable death. 1 The prevalence of smoking and burden of tobacco-caused diseases further remain disproportionately high in individuals with low socioeconomic status, mental illness, substance use disorders, and communities of color. 2,3 Safety-net health systems that provide care for populations with high smoking rates are therefore uniquely poised to provide smoking cessation services, though many competing priorities and limited resources make providing cessation care challenging.
Our prior work, using a PRECEDE-PROCEED evaluation framework, explored enabling factors and barriers to delivery of cessation services within a subset of safety-net clinics. 7 We found that due to competing priorities, clinic staff trained in offering cessation counseling (e.g., behavioral assistants) were not able to meet the demands of "warm hand offs" for counseling during encounters, requiring rescheduling visits that resulted in low follow up. Lack of both coordination of cessation services and ownership of counseling by clinical workforces also posed barriers to successful delivery of cessation services. 7 Less is known about patient-level factors associated with smoking cessation attempts and receipt of cessation services within safety-net populations who have comorbid chronic physical and mental health conditions. This study expands on prior work by using the electronic health record (EHR) to identify factors associated with smoking cessation attempts among all primary care clinics in the San Francisco Health Network (SFHN), the largest network of primary care clinics, specialty clinics, acute care hospitals, and behavioral health clinics serving publicly insured and uninsured patients in San Francisco. In recent years, the SFHN has undertaken various quality improvement initiatives to improve smoking cessation care across the entire health system. Understanding patient-level characteristics and factors associated with cessation attempts can enable opportunities to develop future interventions within SFHN and other urban safety-net systems. We hypothesized that delivery of cessation services, including any counseling from medical assistants, providers, and behavioral assistants, as well as pharmacotherapy, would be associated with cessation attempts.

Study setting and population
We extracted patient-level EHR data on smoking status and delivery of smoking cessation services from 15 primary care clinics across SFHN (Appendix 1).
Four of the 15 clinics (Clinics 2-6) are academic primary care practices housed within a university-a liated public hospital, with three general primary care clinics and one clinic focusing on care for patients living with HIV. The remaining 11 clinics are community-based clinics dispersed across San Francisco with varying patient demographics. Clinic 1 is located in the downtown area and provides care predominantly to people experiencing homelessness, whereas other clinics serve a higher proportion of Black, Latinx, Asian, and other racial/ethnic communities.
At each clinic visit, medical assistants (i.e., frontline medical staff) screen patients for smoking status as "current smoker," "past smoker," or "never smoker" and input their status into the EHR. To be consistent with prior literature and clinical practice, we refer to "past smokers" as "former smokers" for the remainder of the study. 7 For current smokers, medical assistants can refer to one or more cessation resources including those speci c to the clinic (e.g., on-site smoking cessation groups or counseling by the behavioral assistants) and/or general resources (e.g., the California Smoker's Helpline). 14 Providers and medical assistants during the encounter may also refer to behavioral assistants for counseling. Behavioral assistants are ancillary staff trained to offer smoking cessation coaching, and they offer on-site or telephone counseling to patients. Patients can also receive counseling from their provider and/or initiate smoking cessation pharmacotherapy through a prescription from their provider or a pharmacist during the clinical encounter. This study was approved by the University of California, San Francisco Committee on Human Research (#18-26398).

Study Design
The study utilized the PRECEDE-PROCEED model as its theoretical framework, an implementation science model used for assessing the health needs within a public health program, and for designing, implementing, and evaluating programs to meet these needs. 15 In our prior work, we conducted social, epidemiological, and ecological assessments to identify predisposing and reinforcing factors (PRECEDE) to increase delivery of cessation services. 7 We found that provision of medical services using team-based care models is feasible, and that the EHR could be a useful tool to facilitate delivery of cessation care.
Here, we explore potential Policy, Regulatory, and Organizational (PROCEED) factors in order to facilitate the implementation of system-level interventions to increase delivery of cessation services and evaluation of such interventions.
We extracted data from the EHR on any patient of the 15 clinic sites who had at least one recorded smoking status and at least three unique primary care encounters between May 2016 to May 2019. The index visit was the rst encounter closest to the study's start date. Current smokers reported smoking at the index visit, whereas non-smokers were either former smokers or never smokers. We extracted data from the EHR using i2i Tracks (http://www.i2isys.com/p/i2itracks) software and Structured Query Language (SQL).
We extracted patient level factors including age, gender, primary language, race/ethnicity, health insurance type, and smoking-related comorbidities from the EHR. Health insurance was categorized as either Medicare, Medicaid (named "Medi-Cal" in California), Healthy San Francisco/Healthy Workers (a countybased program providing medical care to low-income adults), private, other coverage, or uninsured. Smoking-related health comorbidities were identi ed using ICD-9 and ICD-10 codes listed as diagnoses within the patient's medical record (Appendix 2).

Data Analysis
We used descriptive statistics to summarize the data, with frequencies and percentages for categorical variables and means with standard deviations (SD) for continuous variables. We compared current smokers and non-smokers based on smoking status at the index visit. We used Pearson's chi-square test for categorical variables and the t-test for continuous variables when comparing demographic characteristics, health insurance type, comorbidities, clinic, and referral characteristics.
We evaluated factors associated with making a recent smoking cessation attempt. Recent smoking cessation attempt was de ned as a transition in smoking status from current smoker at the index to former smoker either at visit 2 or 3. We chose independent variables previously shown to be associated with smoking cessation attempts, including demographics, health insurance type, clinic, visit number, comorbidities, and receipt of cessation services including counseling (medical assistant counseling, provider counseling, behavioral assistant counseling) and pharmacotherapy. 7 We examined the multivariable association between recent cessation attempts and covariates using generalized estimating equations (GEE), accounting for nesting patients within clinics with an exchangeable correlation structure. We further characterized rates of recent cessation attempts as well as the relapse rates by clinic. We de ned a relapse as a change in smoking status from former smoker at visit 2 to current smoker at visit 3.
To better understand disparities in delivery of smoking cessation services and delineate which subpopulations may bene t from more intensive efforts, we described characteristics within patient groups with various comorbidities (i.e., hypertension, diabetes, depression, and/or HIV), including their demographics, health insurance type, clinic, receipts of smoking cessation counseling and pharmacotherapy, and cessation attempts. All analyses were completed using SAS version 9.4.

Rates of Cessation Attempts
Within the cohort of current smokers at the index visit, 26% made a recent smoking cessation attempt at either visit 2 or 3 (Fig. 1). Academic, hospital-based clinics had some of the highest cessation attempt rates. Among those who had made a cessation attempt at visit 2, on average 45% relapsed at visit 3 (range 0-50%) (Appendix 3).

Factors Associated with Recent Smoking Cessation Attempts
All forms of smoking cessation counseling were signi cantly associated with lower odds of making a cessation attempt. Receipt of smoking cessation pharmacotherapy did not reach statistical signi cance. Among patient factors, only older age was associated with lower odds of making a cessation attempt

Demographic and Referral Characteristics among Smokers with Speci c Comorbidities
Among current smokers with available EHR comorbidity data, we strati ed patients based on the presence of comorbidities of hypertension (N = 3,441), diabetes (N = 1,446), depression (N = 2,903), and HIV (N = 1,140) ( Table 3). Current smokers in all comorbidity groups were on average middle aged (though those with HIV were slightly younger) and predominantly male. Asian smokers were more represented among populations with hypertension (16%) and diabetes (22%) and less so among those with depression (10%) or HIV (5%). Latinx smokers were more represented among groups with diabetes (17%), depression (17%), and HIV (18%). Black/African American smokers were well represented in all comorbidity groups, especially among those with hypertension (44%). White smokers were well represented in all groups, especially among those with depression (33%) and HIV (38%).
There were high rates of current smokers across comorbidity groups receiving medical assistant (range 92-96%) and provider counseling (74-80%). There was much lower delivery of behavioral assistant counseling (range 7-24%), especially among current smokers with HIV. Notably, those with HIV had consistently the lowest rates of receiving any type of cessation counseling. All comorbidity groups had similar low percentages of receiving cessation pharmacotherapy (17-19%). Cessation rates across comorbidity groups were also similar (21-26%).

Discussion
Among 51,554 individuals across 15 safety-net primary care clinics, we found 23% were current smokers, and 26% made a recent cessation attempt. Safetynet clinics delivered medical assistant and provider counseling at high rates, though rates were much lower for behavioral assistant counseling and pharmacotherapy. Contrary to our hypothesis, individuals receiving any type of cessation counseling were less likely to make a cessation attempt. This was dissimilar to our prior work, which found higher odds of making a cessation attempt in those who received medical assistant counseling and provider counseling, albeit the prior study was limited to only four clinic sites with a much smaller study sample. 7 Our results highlight how safety-net clinics are able to adequately deliver basic cessation interventions such as provider and medical assistant counseling, and that enhancements in EHR functions allowed demonstration of cessation service delivery. However, delivery of cessation services may not always correlate with cessation attempts, especially in the context of large health systems with diverse subpopulations. For subpopulations with high burden of comorbidities, basic cessation services may be insu cient, highlighting a need for more intensive efforts. 8,16 Such efforts may include counseling from care providers of multiple disciplines, combining referrals from different encounters or providers, and streamlined infrastructure to ensure e cient delivery of cessation resources.
Several opportunities for interventions among subgroups exist. Although members of Black/African American, Latinx, and Asian communities are less likely to ever smoke and are more likely to be lighter smokers than their White counterparts, they also face disproportionately worse smoking-related health outcomes. [17][18][19] And despite higher interest in quitting than White individuals and past-year quit attempts, Black/African American individuals have lower rates of sustained cessation. These racial disparities can be attributed to structures of systemic racism, including barriers to accessing care, lower receipt of cessation counseling and pharmacotherapy, and increased targeted marketing of tobacco products to racial/ethnic minorities, making sustained cessation more challenging among these communities. [20][21][22][23] We found communities of color were well represented across comorbidity groups, with Asian smokers well represented among groups with hypertension and diabetes, Latinx smokers represented among those with diabetes, depression, and HIV, and Black/African American represented across all comorbidity groups, especially hypertension. As efforts in addressing diabetes, hypertension, depression, and HIV have all displayed success in improving health outcomes by using a chronic disease management framework, pairing smoking cessation with other chronic disease management efforts may help address racial/ethnic disparities in smoking outcomes. 20,24,25 Such interventions include telephone or in-person outreach to targeted populations, linking cessation counseling with efforts to improve blood pressure or diabetes care targets, or community engagement practices to inform cessation interventions. Latinx and Non-English speaking patients also had higher odds of recent cessation attempts, highlighting the importance of culturally informed and language concordant cessation counseling and resources. 26 More intensive efforts would align with equity goals to reduce racial disparities across health systems.
Our ndings demonstrate how EHRs can be an effective tool for identifying smokers and delivering basic smoking cessation services within the context of rapid cycle quality improvement work. In the past decade, nancial incentive programs for meaningful use of EHRs have increased tobacco screening, documentation of smoking status, and delivery of cessation services in safety-net settings. 11,27−29 Additionally, the EHR has shown to be effective in rapidly identifying factors associated with cessation attempts and the receipt of referral services, which could be used to drive quality improvement activities to improve health outcomes. 7,13 In the face of many competing priorities, health systems are required to meet minimum criteria to obtain reimbursement and incentives from public insurers.
For example, The Public Hospital Redesign and Incentives in Medi-Cal program (PRIME) requires evidence-based quality improvement goals for clinics, including screening for smoking status and counseling every two years. 30 However best practices guidelines recommend assessments at every clinical encounter to optimize chances of cessation, 31 highlighting how more intensive interventions than those required by public insurers may be needed to improve patient outcomes. Therefore, clinics should acknowledge the need to meet minimum requirements for reimbursement, but also to take measures to strive for best practice recommendations. Health systems can do so by streamlining efforts, including assigning responsibilities to each health team member for providing smoking cessation services or providing guidance on how frequently these services should be provided to avoid redundancy and waste of resources. The EHR is also advantageous in identifying populations that need these intensive interventions, and the PRECEDE-PROCEED model can be used to develop interventions in these contexts. For example, the SFHN implemented an EPIC Enterprise EHR in August 2020. Our evaluation using the PRECEDE-PROCEED is therefore timely in providing critical information to developing system-level approaches to support cessation efforts throughout the network. Already, efforts from this work have led to creation of a tobacco registry that includes a better screening tool for tobacco use embedded within the new EHR and templates to document counseling interventions. The registry can be used to track receipt of cessation services and drive practice changes in delivery of cessation care.
In our study, we found that only a quarter of current smokers made a recent cessation attempt, and of those who made a smoking cessation attempt at visit 2, about half of them relapsed by visit 3. These rates of cessation attempts were lower than the estimated 44% of smoking cessation attempts in the past year made by the general US population. 32 Little is known about the rates of cessation attempts in primary care settings, with few studies estimating roughly 36-39% of patients making a recent cessation attempt and 15-20% of patients maintain cessation at one year. 24,25,33,34 Ours was one of the few studies evaluating cessation attempts and relapse rates at clinic or system levels within a safety-net system. Because most smokers who attempt cessation are likely to relapse with high average lifetime number of quit attempts before sustained cessation, 35,36 program initiatives should pay increasing attention toward sustaining cessation attempts and streamlining interventions to determine which groups require more intensive efforts.
There are several limitations to our study. EHR data relied on patient self-report, and smoking status was not biochemically veri ed, leading to a potential misclassi cation bias. However, our repeated assessments of smoking status over time may have reduced potential misclassi cations. By excluding people with missing smoking status in the analysis, we may have also introduced some bias. 7 Still, our large sample size may have protected against this and allows our data to be generalizable, as the inclusion of a diverse array of patients and clinics may be representative of other safety-net settings. The quality of smoking status data collection could have varied across clinic sites, though all clinics had the same EHR with a structured format for data collection. 28 Finally, for some clinics especially those serving young adult populations, the actual numbers of patients who attempted recent cessation were small, leading to in ated percentages of relapse in these clinics.
In conclusion, the EHR can be used to e ciently understand and identify opportunities for improvement in delivering smoking cessation services, especially in subpopulations that may require more intensive, directed efforts to achieve sustained cessation. [37][38][39] Safety-net providers and clinic leaders could consider using the EHR to enhance the reach and e cacy of smoking cessation services, and target subpopulations with high needs in order to reduce racial and health disparities in safety-net settings.