Comparison of Methods for Alcohol and Drug Screening in Primary Care Clinics

This quality improvement study compares commonly used screening methods for alcohol and drug use among adult patients at primary care clinics to examine their association with implementation outcomes and to evaluate the best approach for implementing electronic health record–integrated screening.


Introduction
Alcohol and drug use are among the top 10 causes of preventable death in the US. 1 More than 72 500 deaths in 2017 were associated with alcohol use, 2 and drug overdose deaths now exceed 81 000 deaths per year. 3 The US Preventive Services Task Force recommends screening for both alcohol and drug use among adult patients during primary care visits. 4,5 While the drug screening recommendation reflects a recent change, alcohol screening has been recommended for more than 2 decades and ranks as the third-highest prevention priority for adults in the US. 6,7 Despite the substantial health burden of alcohol and drug use, screening and interventions to address these issues are rarely incorporated into routine medical care. [8][9][10][11][12] Barriers to screening for alcohol use include systems-level problems, such as time and workflow constraints, and underlying issues of stigma and lack of clinician knowledge. [13][14][15][16][17] These barriers are even more pronounced for drug screening because of the variety of substances (ranging from cannabis to heroin and the nonmedical use of prescribed medications), the illegality of some drugs, the greater stigmatization among patients, and the knowledge deficits among clinicians. [18][19][20] Electronic health records (EHRs), which can facilitate systematic screening, guide clinician actions, and record results in structured data fields, have been underused for substance use. 21,22 Common data elements for alcohol and drug information have been defined and recommended for integration into EHRs 23 ; however, in most systems, this information is still gathered in social history fields that do not include validated screening questionnaires and are inconsistently used.
To inform strategies for implementing substance use screening and interventions in primary care settings, the Clinical Trials Network of the National Institute on Drug Abuse conducted a study of EHR-integrated screening in 2 large urban academic health care systems. The primary goal was to facilitate the implementation of screening approaches that were feasible, had good potential for sustainability, and would optimize the screening rate and capture of screening data in the EHR. To evaluate how best to implement EHR-integrated screening, the present study compared screening methods to examine their association with implementation outcomes during the first year of screening at 6 participating primary care clinics.

This study was approved by the institutional review boards of the New York University Grossman
School of Medicine, the Icahn School of Medicine at Mount Sinai, and Partners Healthcare System. A waiver of informed consent was granted because the screening was conducted as part of routine clinical care, the study posed minimal risk to participants, and it would not have been feasible to obtain individual consent from all clinic patients. The study followed the Standards for Quality Improvement Reporting Excellence (SQUIRE) reporting guideline for health care quality improvement studies. 24

Study Design and Setting
In this quality improvement study assessing implementation feasibility, clinics were required to use screening tools included in the National Institute on Drug Abuse Common Data Elements. 23 All clinics received support to compensate a clinical champion (10% full-time equivalent) and expert consultation from the research team to assist them in choosing a screening approach that clinic leaders believed would be the most feasible and effective in their setting. The clinics used existing EHR systems and were not provided with additional clinical staff or resources to perform screening.
Although the study was not randomized, clinics varied in their screening approaches and implementation strategies. We sought to characterize and evaluate the screening implementation outcomes achieved using these diverse tactics.
The study sites were 2 urban academic health care systems; site A was in New York City, and site B was in Boston, Massachusetts. At site A, which included 2 primary care clinics (A1 and A2), screening was advocated by a site principal investigator (R.N.R.), who is a clinical leader in the treatment of substance use disorders. At site B, which included 4 primary care clinics (B1 through B4), screening was part of a systemwide initiative to integrate substance use care into general medical settings. Clinics were general internal medicine practices serving adult patients, and none of the clinics were systematically screening patients for alcohol or drug use before the study began.
One clinic from each site was the primary teaching practice for the internal medicine residency program, which was located near the hospital. The other clinics were in community settings, and 1 clinic (A2) was a faculty practice. All clinics used Epic EHR software (Epic Systems Corp).
Before the initiation of screening, barriers were assessed through focus groups and interviews with stakeholders, 25 and the EHR tools developed to support screening were tailored through multiple rounds of usability testing. Screening was initiated at 1 clinic within each health care system in January 2017 (clinic A1) and July 2017 (clinic B1). Screening was then implemented at the remaining clinics (A2, B2, B3, and B4) between February and October 2018. Implementation outcomes were collected for 12 months after initiation of screening at each clinic, and data collection ended in October 2019. The primary outcome was the screening rate for alcohol and drug use. Secondary outcomes were the prevalence of unhealthy substance use detected via screening and clinician adoption of a brief counseling script.

Screening Program Elements
Clinics used screening program elements that had been reported to increase the adoption of screening and interventions by health care professionals. [26][27][28][29] These elements 23,30-37 are summarized in Table 1 and included the training of clinic staff and the use of validated screening tools (the single-item screening questions for alcohol and drugs, the 3-item Alcohol Use Disorders Identification Test-Consumption items [AUDIT-C] and the 10-item Drug Abuse Screening Test [DAST-10]), a clinical reminder in the EHR indicating that a patient was due for screening (based on age, type of visit, and no receipt of screening within the past 12 months), a brief EHR-integrated counseling script suggested for use with patients with moderate-to high-risk alcohol or drug use, and clinical champions (ie, clinicians who advocate for change, motivate others, and use their expertise to facilitate the adoption of an intervention). The counseling script was created for the study and included the 4 major components of a brief negotiated interview: raising the subject, providing feedback, enhancing motivation, and negotiating a plan. 36

Screening Approach and Implementation Strategies
Participating health care systems and clinics selected the screening approach that they deemed most feasible given their resources, clinical workflows, and patient populations. There was variation in the type of visit targeted for screening (annual examination only vs any visit), with clinics A1 and A2 choosing to conduct screenings at any visit and clinics B1 through B4 choosing to conduct screenings at annual examinations only ( Table 2). The mode of screening administration (staff-administered vs self-administered by the patient) also varied, with 5 clinics choosing self-administration and only 1 clinic (A1) choosing staff administration (Table 2).
Clinics varied in the extent of practice facilitation and usability testing of the EHR tools developed to support screening. Practice facilitation is an implementation strategy in which trained facilitators work with clinic leaders and staff to implement evidence-based practices. [39][40][41] All clinics received practice facilitation from study investigators (T.A.K., E.F., B.I., and M.H.) for the first 12 months of the screening program, and the extent of facilitation was classified as robust vs standard based on training, days spent on site by the facilitator, and frequency of meetings with clinical champions. Usability testing involves observing human-computer interaction as representative end users use a prototype and provide feedback on system design, interface, information content, and mode of delivery. It is an iterative process in which each round of usability testing is followed by All patients received single-item screening questions for alcohol and drug use; responses >0 were considered positive results [30][31][32] Patient with positive alcohol screening results received the AUDIT-C 33 ; patients with positive drug screening results received the DAST-10, 34,35 which provides a single summary score and does not specify drug classes used Established cutoffs were used to categorize results as representing moderate-or high-risk use a

Clinical reminders
Best practice alert appears in the EHR, indicating that a patient is due for screening (based on age, visit type, and no screening within the past 12 mo) Counseling script EHR-integrated counseling script created for the study provided guidance for conducting and documenting a brief intervention to address substance use Training of clinic staff recommended using the script for patients with moderate-to high-risk alcohol or drug use Accessed through a dot phrase in Epic EHR software: 1 keystroke to start, with fillable fields to document patient responses (2 fields at site A and up to 7 fields at site B) Designed to be delivered in approximately 5 min Guided clinicians through the 4 major components of a brief negotiated interview: raising the subject, providing feedback, enhancing motivation, and negotiating a plan 36 For patients with high-risk use, the script suggested placing a referral order for an appointment with clinic social workers or a peer navigator

Screening Implementation Outcomes
Implementation outcome data were extracted from the EHR for 12 months after initiation of the screening program at each clinic. Data collected comprised (1)

Statistical Analysis
Clinics sent extracted EHR data that captured screening information to the study's data and statistical coordinating center (The Emmes Company, Rockville, Maryland) throughout the first 12 months; these data were used to generate summary reports of the implementation outcome measures for each clinic. A separate data extraction was performed at the end of the study period to collect sociodemographic information from the EHR, which was used to describe the patient populations at each clinic. Descriptive statistics were used to characterize the rates and frequencies of predefined implementation outcomes.

Results
Patients of the 6 clinics had a mean (SD) age ranging from 48 The prevalence of moderate-to high-risk alcohol use detected by screening was lowest at clinic A1 (1.6%), which used a staff-administered screening approach. At clinics that used a selfadministered approach, the prevalence of moderate-to high-risk alcohol use was higher, ranging from 14.7% at clinic A2 to 36.6% at clinic B1. The prevalence of moderate-to high-risk drug use detected by screening was low at all clinics, ranging from 0.3% at clinic A2 to 1.0% at clinics B1 and B2.
The counseling script was used infrequently, with rates ranging from 0.1% at clinic B3 to 12.5% at clinic A1 (Table 4). Clinics A1 and A2, which used implementation strategies comprising robust practice facilitation and more EHR usability testing, had the highest rates of counseling (12.5% and 1.4%, respectively) across the 6 clinics.

Discussion
In this quality improvement study of the feasibility and implementation of EHR-integrated screening, 71.8% of eligible patients received screening for substance use over the course of 12 months, via the use of validated questionnaires during routine primary care visits. This screening represents a substantial change in practice for the participating clinics, none of which were systematically screening patients for alcohol or drug use before our study intervention began. The successful implementation of screening aligns these clinics with the current US Preventive Services Task Force guidelines for alcohol and drug screening among adult patients in primary care settings. 4,5 Notably, the screening program was implemented using existing clinic staff and EHRs, although the study provided support for practice facilitation and usability testing as well as modest funding for clinical champions. Screening rates were higher than those observed in a number of previous a Clinics A1 and B3 were the primary outpatient clinical training sites for residents in general medicine at their respective sites.
b Characteristics of adult patients who had primary care visits during the study period. Demographic data were extracted from the electronic health record separately from implementation outcome data; therefore, there are small discrepancies in the number of patients included in Table 3 (N = 93 023) and the number included in the implementation outcomes data shown in Table 4 (N = 93 114).
c At clinic A2, 3 patients reported their sex as other. At clinic B2, 1 patient reported their sex as other.
d Other races included American Indian/Alaska Native, Native Hawaiian/other Pacific Islander, other, multiracial, unavailable or declined, and missing. Hispanic ethnicity was an optional field in the EHRs and was missing for most patients; therefore, it was not included as an additional variable in the study. e Other insurance included coverage from the US Department of Veterans Affairs, workers' compensation, professional associations (eg, law enforcement), county jails, screening implementation studies. 26,29,44 At some clinics, screening prevalence was consistent with results from the Veterans Health Administration, which has prioritized alcohol screening during primary care visits for years. 45 However, we did observe differences in screening rates that appeared to be associated with the screening approach and implementation strategies used in the participating clinics.
Screening rates were highest at site A clinics, which adopted the approach of screening patients during any primary care visit. The screening rates at site A (89.6%-94.7%) were slightly higher than those reported in a recent pragmatic clinical trial conducted at primary care practices in Washington state. 46 The site B clinics, which specifically targeted annual examinations for screening, had lower and more variable screening rates (24.2%-72.0%). Although it is typical for primary care practices to conduct screening only during a dedicated annual physical or preventive care visit, patients may miss appointments; therefore, offering screening during any type of visit provides more opportunities to detect substance use. Notably, at all sites, the performance of screening was suggested only once per year because screening at every visit may produce patient and clinician fatigue and decrease the accuracy of results. 47 Perhaps the most notable finding was the substantially lower detection of unhealthy alcohol use at the 1 clinic (A1) that used a staff-administered screening approach in comparison with clinics that used a self-administered approach. Based on survey data, the anticipated prevalence of unhealthy alcohol use in the general population is approximately 30%. 48,49 In medical settings, the rates of positive screening results are typically lower, which may be associated with patients' reluctance to disclose substance use. 25,[50][51][52] Two study clinics achieved rates of positive screening results for moderate-to high-risk alcohol use that exceeded general population estimates, and all 5 clinics using a self-administered approach detected a greater than 14% prevalence of moderate-to high-risk alcohol use (vs <2% for the clinic that used staff-administered screening). Our findings are consistent with those of previous studies using self-administered screening, which typically produces more accurate reporting among patients with stigmatized conditions. [53][54][55] Our findings could also reflect problems with the quality of screening when administered by staff, who may change the wording of validated screening questions in an effort to hasten the process or reduce perceived patient discomfort. 56,57 Screening rates were similar for alcohol and drug use, likely reflecting the fact that most clinics administered the alcohol and drug screenings simultaneously. However, the detection of unhealthy drug use via screening was low and did not appear to vary based on the screening approach or implementation strategy used. Although population rates of illicit drug use are lower than those of alcohol use, the prevalence of past-year drug use may be as high as 21% when cannabis is included. 58 The low screening-detected prevalence of drug use likely reflects patients' discomfort with disclosing a behavior that is illegal and stigmatized and one that they may believe could negatively impact their medical care. The drug screening tools used in our study did not distinguish between the use of cannabis and other drugs. Given the changing legal status and social acceptability of cannabis, it is possible that using a screening instrument that contains separate questions about cannabis and other illicit drugs, such as the Tobacco, Alcohol, Prescription Medications, and Other Substance Use (TAPS) tool, 59 would produce higher reporting of drug use. 60,61 Regardless of the screening approach, increasing patients' comfort with disclosing drug use may also require improving clinicians' attitudes. 17,19,25 Clinics differed in their implementation strategies, with site A clinics having more robust practice facilitation and usability testing than site B clinics, which may have been associated with the higher rates of adoption of screening and counseling. Practice facilitation can be a beneficial implementation strategy, although results are not uniform across studies. [39][40][41] Usability testing is helpful for implementing any practice change that involves modifications to the EHR. 43,62,63 In our study, it is not possible to differentiate the relative contributions of practice facilitation and usability testing, although the clinic that performed the most usability testing (A1) used the counseling script at a slightly higher rate. Further studies are needed to examine the independent, and possibly synergistic, associations between these implementation strategies.
Clinician use of the counseling script among patients with moderate-to high-risk alcohol or drug use was low at all clinics. This low adoption may reflect the fact that counseling was recommended but not required and that delivering the script was a complex and relatively time-consuming task (requiring Ն5 minutes). 64,65 Focus groups conducted in an earlier stage of this study voiced concern that substance use counseling would be too time consuming to include during regular primary care visits. 25 Further barriers to adoption could have been clinician discomfort and lack of knowledge about alcohol and drug use, which may have made them reluctant to engage patients in conversation about these behaviors. 14,66

Limitations
This study has limitations. As an implementation feasibility study, it sought to adapt to the existing conditions and resources of the participating clinics. Clinics were not randomized, so it is not possible to conclude with confidence that the differences we observed are associated with the screening approaches or implementation strategies used rather than other practice characteristics. In addition, clinics had overlapping combinations of implementation strategies, so we are unable to pinpoint whether robust practice facilitation or usability testing has more substantial implications for differential screening outcomes. However, the extent of the differences, particularly in screening rates and the prevalence of unhealthy alcohol use detected via screening, suggests that the screening approach selected by clinics was associated with the outcomes. We had limited ability to measure clinician counseling, and it is possible that clinicians discussed screening results with patients without using the counseling script. Conducting medical record reviews or using natural language processing to analyze clinical notes may have captured this information, but these methods were outside the scope of the study. We were also unable to measure referrals for substance use treatment among patients who received positive screening results for high-risk substance use.
Although study clinics varied in size, location, and patient population and included both faculty and resident clinicians, they were all within urban academic health care systems; thus, they are not representative of all primary care practices.

Conclusions
By examining the outcomes of common approaches to screening for substance use in primary care settings, the findings of this quality improvement study can guide clinics and health care systems that are seeking to implement screening for alcohol and drug use. Interest in screening will likely increase, motivated by US Preventive Services Task Force recommendations, 4,5 the Healthcare Effectiveness Data and Information Set (HEDIS) measure for alcohol screening and brief interventions, 67 and increases in substance use associated with the COVID-19 pandemic. 68 The high screening rate achieved in our study clinics supports the feasibility of EHR-integrated screening for substance use as part of routine primary care. This study also suggests best practices, including the use of selfadministered screening tools and the performance of screening at any type of visit. The implementation strategies of robust practice facilitation and usability testing, although more resource intensive, were also associated with greater adoption of screening and counseling.
Many previous studies have reported the challenges of implementing substance use screening and interventions in primary care settings. 26,29,44,46,69,70 The health care systems participating in the present study have maintained screening and are now adopting our EHR-integrated screening tools systemwide, which highlights the success of their implementation. However, more research is needed regarding beneficial interventions to address moderate-risk drug use during primary care visits, strategies for motivating clinicians to engage patients in discussions of substance use, and the resources required to do so.