|Home | About | Journals | Submit | Contact Us | Français|
Among adult emergency department (ED) patients, we sought to examine how estimates of substance use prevalence and the need for interventions can differ, based on the type of screening and assessment strategies employed.
We estimated the prevalence of substance use and the need for interventions using the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) in a secondary analysis of data from two cross-sectional studies using random samples of English- or Spanish-speaking 18–64-year-old ED patients. In addition, the test performance characteristics of three simplified screening strategies consisting of selected questions from the ASSIST (lifetime use, past three-month use, and past three-month frequency of use) to identify patients in need of a possible intervention were compared against using the full ASSIST.
Of 6,432 adult ED patients, the median age was 37 years-old, 56.6% were female, and 61.6% were white. Estimated substance use prevalence among this population differed by how it was measured (lifetime use, past three-month use, past three-month frequency of use, or need for interventions). As compared to using the full ASSIST, the predictive value and accuracy to identify patients in need of any intervention was best for a simplified strategy asking about past three-month substance use. A strategy asking about daily/near-daily use was better in identifying patients needing intensive interventions. However, some patients needing interventions were missed when using these simplified strategies.
Substance use prevalence estimates and identification of ED patients needing interventions differ by screening strategies used. EDs should carefully select strategies to identify patients in need of substance use interventions.
Recent research indicates high prevalences of alcohol, smoking and drug use among United States (U.S.) emergency department (ED) adult patients.1–6 However, estimated ED patient substance-use prevalence and anticipated need for interventions likely are impacted by the screening strategies used to measure them, such as who is screened; where, when, how and by whom screening is conducted; the simplicity or complexity of these strategies; the types of screening instruments used; and the domains these instruments measure. Variability across screening strategies may lead to disparate or contradictory recommendations for addressing substance use among ED patients, which in turn could have consequent deleterious effects on ED and public health priorities chosen for substance use prevention, treatment efforts and the allocation of extramural funding to evaluate interventions.
Although ED-based screening and initiation of consequent interventions are recommended for alcohol and smoking,7–9 this practice is not yet routine.10–18 Comprehensive assessments of substance use and intervention need can be a time-consuming process, which can discourage their use. An enticing way to assess ED patients substance use is to use simplified screening-question strategies involving one or two screening questions that might identify individuals in possible need of an intervention (e.g., “In the past 3 months, did you drink alcohol?”).3,4 This strategy might be applied by ED staff at triage or incorporated into the electronic medical record (EMR), and the results could prompt or obviate the need for an intervention during the clinical care encounter. Before such simplified strategies can be recommended, their yield and accuracy against relatively more comprehensive screening and assessment strategies need to be evaluated.
Our primary aim was to examine how estimates about the prevalence of substance use among adult ED patients differ when these estimates are based on the following: any lifetime, any past three-month, or past three-month frequency of substance use; or according to the need for any, a brief, or an intensive intervention per the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST).19 Our secondary goal was to compare the ability of three simplified screening-question strategies to identify adult ED patients in need of any intervention vs. no intervention, and in need of an intensive intervention vs. no intensive intervention (i.e., no intervention needed or only a brief intervention [BI] needed), to the full ASSIST as the “gold standard.” The three simplified strategies constituted screening based on any lifetime use, any past three-month use, or past three-month frequency of use.
This investigation was a secondary analysis of two concurrent studies on substance use at two EDs affiliated with a medical school in the same hospital system and city from July 2010 to December 2012. The data were based on two cross-sectional studies that involved surveying random samples of adult ED patients. The hospital institutional review board approved the study.
For both studies, bilingual (English- and Spanish-speaking) research assistants (RAs) randomly selected ED patients for possible study inclusion, reviewed their EMR for exclusion criteria, and confirmed study eligibility through a brief interview. A random sample of patients present in the ED during study collection periods was approached and evaluated for study inclusion through random selection to their patient care rooms. If the ED EMR indicated that a patient potentially was study eligible, a RA would confirm study eligibility through a brief interview. Data collection for the study was performed from 8 AM to midnight seven days/week when bilingual (English- and Spanish-speaking) RAs were available to conduct the study.
Patients were study eligible if they were 18–64 years-old; English- or Spanish-speaking; not critically ill or injured; not prison inmates, under arrest, nor undergoing home confinement; not presenting for an acute psychiatric illness; not requesting treatment for substance use; not intoxicated; and did not have a physical or mental impairment that prevented them from providing consent or participating in the study. The study population aimed to reflect the general adult ED population that would be included in a screening, brief intervention, and referral to treatment (SBIRT) program, i.e., excluding those presenting for evaluation of their substance use, those acutely intoxicated, and those undergoing a formal substance use or psychiatric evaluation.
Participants completed the ASSIST, which we adapted for these studies. (See supplemental material for an English-language copy of the study instrument.)19 The cognitive-based assessments, pilot testing and evaluations of the adapted ASSIST have been described in detail previously.20 In brief, this adaption involved preparing it for audio computer self-administered interviewing (ACASI) to increase veracity of responses of sensitive or stigmatizing information (i.e., substance use/misuse)21–25 and improve the flow of the instrument; clarifying questions, responses, and instructions; distinguishing misuse from the use of prescription drugs; and adding or expanding drug categories (e.g., barbiturates, benzodiazepines, prescription opioid analgesics). Cronbach’s α ranged from 0.86 to 0.95 for the drug categories assessed in our adapted ASSIST.
As shown in Figure 1, participants first were asked in the ASSIST about any lifetime substance use by substance category, and if they indicated lifetime use of a given substance, they were asked about any past three-month use and frequency of use during the past three months (ASSIST Questions 1 and 2). Following these initial questions, the ASSIST proceeded through an evaluation of substance-use severity. Substance-specific scores are calculated for those who have used a given substance; if an individual has not used that substance, no score is calculated. Per World Health Organization (WHO) recommendations, an ASSIST score of ≥4 points for smoking or any drug category or a score of ≥11 points for alcohol suggests a need for BI, and a score of ≥27 points for any substance suggests a need for a more intensive intervention.19 Those who report use in the past three months are assessed for the need of a BI or an intensive intervention, while those who report ever using a given substance (lifetime use) but deny past 3-month use are assessed for the need for a BI only.
In addition to the ASSIST, we queried participants about the specific drugs that they had used within the past three months and whether or not these drugs had been injected or prescribed. (See supplemental material [Appendix A].) The reading level of study questionnaires in English was at a Flesch-Kincaid grade level of 6.6 (Microsoft Word; Microsoft Corp., Redmond, WA) and in Spanish was at a Huerta Reading Ease score of 80, both indicating an easy reading level.26 Participants completed the questionnaires in approximately 10–15 minutes.
The RAs received 40 hours of training on the study protocol, including mock interviews with the study investigators and pilot testing of the study protocol prior to collecting data for the study. The RAs met with the study investigators throughout the study to discuss procedural issues arising from the conduct of the study. To ensure fidelity to the study protocol, study investigators directly observed RAs during participant encounters. Deviations from the study protocol were addressed and suggestions for improvement were provided.
Analysis of the study and presentation of study findings followed current Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations for cross-sectional studies (www.strobe-statement.org). We summarized study eligibility assessments and enrollment using current recommendations,27 participant demographic characteristics, and responses to the ASSIST. ASSIST scores were calculated for each participant. The need for any, a brief, or more intensive intervention was calculated according to WHO recommendations for all participants, and as stratified by those reporting no past three-month or any past three-month use of a given substance.19
For the primary aim, substance use prevalence was estimated for each substance use category (e.g., smoking, alcohol, benzodiazepines) as stratified by responses to Questions 1 and 2 of the ASSIST (any lifetime use, any past three-month use, and past three-month frequency of use), and by the need for interventions using ASSIST scores. Substance use was ranked in order of decreasing magnitude of prevalence or frequency, respectively, across all substance categories according to these strata. We estimated accompanying 95% confidence intervals (CIs) for the rankings. The ranks were provided to assist in distinguishing differences among proportions within each category.
For the secondary aim, we calculated the test performance characteristics for the ability of the three simplified screening question strategies to identify individuals in need of any intervention vs. no intervention, and in need of an intensive intervention vs. no intensive intervention (i.e., no intervention needed or only a BI needed), as compared to the full ASSIST as the “gold standard.” The three simplified strategies constituted screening based on any lifetime use, any past three-month use, or past three-month frequency of use (per ASSIST Questions 1 and 2). Sensitivity, specificity, negative and positive predictive values, and accuracy with corresponding 95% CIs were estimated. We performed all analyses using STATA 13 (Stata Corporation, College Station, TX).
Participant eligibility assessment and enrollment results are depicted in Figure 2. As shown, of the 9,813 randomly selected 18- to 64-year-old English- or Spanish-speaking ED patients, 78% (7,643) were assessed in person for study eligibility. Of these, 97% (7,409) were eligible to complete the ASSIST, and 87% (6,432) completed it and comprised the final study sample. The majority of participants were female, white/non-Hispanic, most had 12 or fewer years of formal education, had private healthcare insurance, were married or part of an unmarried couple, were employed, not homeless, and received their medical care from a private clinic/practice (Table 1).
In regards to specific substances used among these ED patients, the highest past three-month use prevalences for illicit drugs were cocaine (4.4%), crack (2.6%), and ecstasy/3,4-methylenedioxy-N-methylamphetamine (MDMA) (1.0%) (Supplemental Table 1). Although among all participants the highest past three-month use prevalences for prescription opioids were acetaminophen and hydrocodone (4.0%), acetaminophen and oxycodone (3.6%), and oxycodone (1.1%), only 2.2% of all participants stated that had been prescribed acetaminophen and hydrocodone, 1.9% acetaminophen and oxycodone, and 0.7% oxycodone within the past three months. Lifetime prevalence of injection-drug use was 4.4% and was 1.7% for past three-month use, predominately heroin (0.93%) and cocaine (0.56%).
When estimating the prevalence of substance use based on lifetime use or past 3-month use, the rank order was the same for the first three substances (alcohol, smoking, and then marijuana) (Table 2). When estimating substance use prevalence according to daily/near-daily use, smoking had the highest rank, followed by marijuana, alcohol, and then prescription opioids. There were subtle differences in the ranks for the remaining drugs across these three ways of estimating the prevalence of substance use.
When estimating substance use prevalence based on the need for interventions (Table 3a and and3b),3b), the need for any intervention or at least a BI among these patients was greatest for smoking, marijuana, and then alcohol; however, the need for an intensive intervention was greatest for smoking, alcohol, and then marijuana. The need for any, a brief, or an intensive intervention differed slightly for the remaining substances.
The relative rank order of the need for a BI also differed when ASSIST score results were stratified by lifetime only vs. past three-month use. Among those who reported lifetime-only use, the need for a BI was greatest for opioids, methadone or buprenorphine, prescription opioid analgesics, barbiturates and then cocaine or crack. However, among those who reported past three-month substance use, the need for a BI was greatest for smoking, marijuana, amphetamines, prescription opioid analgesics, and then methamphetamines. The need for an intensive intervention based on ASSIST scores among those who reported past three-month substance use was greatest for opioids, gamma hydroxybutyrate (GHB), methadone or buprenorphine, cocaine or crack, and then barbiturates. The relative need for an intervention differed for specific substances when comparing past or no past three-month use of these substances. For example, the need for a BI for illicit opioids was greater among those who reported no past three-month use of this substance; whereas for smoking, the need for a BI was greater for those who reported smoking in the past three months.
The performance of using simplified screening question strategies (any lifetime use, any past three-month use, or daily/near-daily use) as compared to the full ASSIST as the “gold standard” to identify need for WHO-recommended substance use interventions is shown in Table 4a–4d. Across all substances, sensitivity decreased while specificity increased when using querying about lifetime use vs. past three-month vs. daily/near-daily use. Also across substances, querying about past three-month use generally performed better in regards to predictive value and accuracy for any intervention need, but querying about daily/near-daily use was generally better for identifying intensive intervention need.
The test performance characteristics of these three simplified screening strategies varied by substance category. For any smoking intervention, predictive values and accuracy of querying about past three-month use was higher than screening for lifetime alone or frequency of use (daily/near-daily use). However, frequency of use performed better for identifying need for smoking intensive interventions, although the positive predictive value was low. For any or alcohol intensive interventions, querying about frequency of use performed better than the other screening strategies. For any marijuana interventions, querying about past three-month use performed better, but querying about frequency of use was better for intensive interventions. For all other drugs, querying about past three-month use was better than other queries for any intervention need, yet querying about frequency of use was better for intensive intervention need.
Our investigation demonstrates how estimated substance use prevalence among ED patients can differ depending on how it is derived. For example, prescription opioid analgesics is only sixth when estimating its prevalence based on lifetime use, but rises to fourth in prevalence when considering past three-month use, daily/near-daily use, or the need for a BI based on past three-month use. Likewise, smoking becomes the number one substance use problem when considering the need for interventions, rather than measuring lifetime or past three-month use prevalence. Subtle small differences in estimates such as these can have large public health implications on which substances are prioritized for interventions in United States EDs.
This investigation adds a unique perspective to the small number of large sample studies estimating the extent of adult ED patient substance use. The results from these prior studies also illustrate how differences in who is screened, how they are screened, and how prevalence is estimated can change perspectives on the extent of substance use in this setting. Wu, et al. conducted a secondary analysis of past year prevalence of alcohol and drug use disorders using ACASI-collected data from persons ≥18 years-old in the U.S. who participated in the 2007–2009 National Survey on Drug Use and Health and who had visited an ED at least once within the past year.2 Approximately 66% of those who had visited an ED drank alcohol, 12% had used at least one of nine different drugs within the past year, 10% met criteria for alcohol abuse or dependence and 4% for drug abuse or dependence per Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria. Wu, et al. observed that the prevalence of substance use disorders was greatest for heroin, sedatives, cocaine, and then opioids. As noted, Wu, et al. reported a lower prevalence of drug use and markedly different order of drugs of concern than for our study. These differences likely are due to the study methodology, drug categories used, their inability to measure recency of substance use, and use of DSM-IV criteria for abuse and dependence.
Four recent U.S. ED-based studies screened patients concurrently with their ED visit, as we did in our investigation. Two employed a brief screening strategy administered aloud by nurses at ED triage. From 2009 to 2012 among ≥18-year-old ED patients in Macon, Georgia, Johnson, et al. evaluated three one-question screening tools for past year tobacco product use, alcohol use (≥4 drinks in a day for women, ≥5 drinks in a day for men), and drug use (“pot [marijuana], use of another street drug, or use of a prescription painkiller, stimulant or sedative for a non-medical reason”), which had been adapted from primary care settings.3 Approximately 22% screened positive for at-risk alcohol or drug use. Hankin, et al. from 2009 to 2010 employed a screening approach similar to that by Johnson, et al. among a predominately Black, non-Hispanic population at the Grady Hospital ED in Atlanta, Georgia.4 Patients with a positive screen were administered the ASSIST through an interview with a health educator. Of those screened, 27.6% had a positive one-question screen for alcohol or drug use within the past year; marijuana and cocaine were the most commonly used drugs. The lower yields of screening for these studies as compared to our study might be due to the type of screening instruments used and mode of screening (nurse-administered aloud at ED triage). Employing three different screening instruments at EDs in six states among ≥18-year-old English-speaking patients, Sanjuan, et al.5 and Konstantopoulos, et al.6 found that 39% reported daily tobacco use, 45% risky alcohol use, 22% any past 30-day drug use, and 17% moderate to severe drug problems. The most frequently reported drugs used were marijuana, cocaine, “street opioids” and prescription opioids. Although Spanish-speaking ED patients were excluded and different instruments were used, the prevalence of use from these two analyses is relatively similar to ours.
In their practical application, our study findings also indicate how screening strategies can affect adult ED patients identified (or missed) as possibly needing an intervention. For example, a simplified screening question strategy that queries about past three-month use to identify adult ED patients needing any intervention would perform well for smoking, marijuana, and other drugs, but would perform poorly for alcohol. Likewise, querying about daily/near-daily use to identify need for an intensive intervention would perform well for marijuana, other drugs, and alcohol, but less well for smoking.
As the ASSIST assessments that stratify by lifetime or past three-month use remind us, remarkably high proportions of ED patients who deny current or recent substance use still qualify for a BI per WHO recommendations. This finding indicates that asking about lifetime use and not just past three-month or frequency of use remains an important facet in identifying those who might benefit from substance use interventions. This stratification also demonstrates that the relative need for an intervention can vary substantially when considering past three-month use, as the example of illicit opioids and smoking illustrate. One implication for ED clinical practice is to recognize that a need for an intervention is based on more than current substance use. Querying about lifetime and current use and frequency of use can help to identify those who need assistance. The challenge for the future is to learn how to screen efficiently yet effectively in EDs to identify those who might benefit from assistance because of their substance use. Unfortunately, we do not yet have a single perfect instrument to screen adult ED patients quickly and easily for substance abuse in a manner that identifies all of those in need of an intervention, without overburdening providers and patients, and minimizes false positive and negative results.
One caveat is that effective interventions for adult ED patients to help them reduce or eliminate their substance use continue to elude us. Although BIs have been recommended28 for alcohol among adolescent and adult ED patients, studies evaluating them have had mixed results.29–42 Three different recently published studies also observed the following: (1) no short-term (three-month follow-up) benefit from a BI in reducing drug use or increasing uptake of drug treatment services;43 (2) no differences in past 30-day drug use abstinence six months post-enrollment44; and no advantage to a BI with telephone boosters as compared with screening, assessment and referral to treatment or minimal screening only in terms of drug use at three, six, and 12 months post-enrollment.45 As such, although instruments such as the ASSIST might indicate the need for an intervention, the optimal type of intervention needed is not yet known.
This study had several limitations. Although care was taken to reach a representative sample of adult patients at these two EDs not presenting for a substance use or psychiatric evaluation, those who were excluded might have different substance use profiles and need for interventions. An advantage of this study is that random sampling of patients was used and Spanish-speaking patients were included, which should increase the internal and external validity of the study findings. Of course, patients at EDs in other locales might have a different spectrum of substance use. Although we provide estimates of predictive value and accuracy for the simplified screening strategies, such estimates are based on the prevalence at these EDs, and likely would be different at other EDs. Nevertheless, these parameters provide an illustration of the application of the screening strategies. Even though confidentiality and fidelity of responses likely was enhanced by the use of the ACASI approach,21–25 we cannot be assured that all participants answered the questions truthfully. As would be expected, other screening strategies and substance-use evaluation instruments other than the ASSIST might yield different results. In addition, because we only used questions taken from the ASSIST and not any other instrument, we cannot determine how the three simplified screening strategies would compare to other screening instruments. Comparisons to different instruments might have led to different conclusions.
The results of this investigation indicate that estimated substance use prevalence among adult ED patients differs according to how it is measured. In addition, the yield from simplified screening question strategies can vary by substance, and risk missing patients who might benefit from an intervention. Public policy makers and EDs contemplating substance use screening programs should be cognizant that the priorities for what substances need intervention and the types of interventions needed are highly dependent on how adult ED patients are assessed for substance use.
Section Editor: Chadd K. Kraus, DO, DrPH, MPH
Full text available through open access at http://escholarship.org/uc/uciem_westjem
Conflicts of Interest: By the WestJEM article submission agreement, all authors are required to disclose all affiliations, funding sources and financial or management relationships that could be perceived as potential sources of bias. The authors disclosed none.