|Home | About | Journals | Submit | Contact Us | Français|
This report describes the development and measurement characteristics of a new measure of smoking knowledge, attitudes, and practices (S-KAP) among treatment providers. Data are based on survey responses from 336 paid staff working in one of three drug abuse treatment or HIV care settings. Exploratory factor analysis, used to examine the factor structure, pointed towards five underlying factors: a single “knowledge” factor, three “attitude” factors (‘treatment barriers,’ ‘counselor self-factor. The Knowledge scale had a standardized Cronbach’s alpha coefficient of .85. The coefficients for Barriers, Self-Efficacy, and Attitudes were .81, .72, and .74, respectively. The Practice scale had a standardized Cronbach’s alpha coefficient of .91. These results indicate that the proposed scales have reasonably good psychometric characteristics and will allow researchers to quantify staff knowledge, attitudes, and practices regarding smoking cessation treatments and issues.
Persons with either a mental or substance abuse disorder are twice as likely to smoke as those without, and 44% of all cigarettes smoked in the U.S. are smoked by this population (Lasser et al., 2000). Persons involved in substance abuse smoke at higher rates than the general population (Sobell, Sobell, & Agrawal, 2002). Specifically, smoking rates above 70% are reported for persons with substance abuse involvement (Fiore et al., 2008; Richter, Ahluwalia, Mosier, Nazir, & Ahluwalia, 2002), and smoking rates over 85% have been reported among persons enrolled in methadone maintenance treatment (Richter, Choi, McCool, Harris, & Ahluwalia, 2004). Persons with substance abuse problems are more heavily addicted to nicotine than non-substance abusers who smoke (Hughes, 2002). Alcohol dependent individuals die from smoking-related causes more frequently than from alcohol-related causes (Hurt et al., 1996). Severely dependent substance abusers are also less successful in their attempts to quit smoking than non-substance abusers (Joseph, Nichol, & Anderson, 1993), although many persons in drug treatment are interested in quitting smoking (Saxon, McGuffin, &Walker, 1997).
Drug treatment clinics may be an optimal venue for delivering smoking cessation interventions. Persons entering drug treatment often have reached a teachable moment when they are motivated to change addictive behaviors (Williams, Brown, Patton, Crawford, & Touquet, 2005). Blending smoking cessation into existing drug treatment programs may reduce attendance problems that occur when patients are referred to external smoking cessation clinics (Sherman, Yano, Lanto, Wang, & Lee, 2001). Finally, the integration of nicotine dependence services allows drug treatment programs to provide a consistent message targeting all addictive substances and focus on positive lifestyle change (Sussman, 2002).
Despite the benefits of smoking cessation and national guidelines that direct healthcare professionals to address smoking in patients (Fiore et al., 2008), smoking is often overlooked in drug programs. Resistance by drug treatment staff to treating nicotine dependence is documented (Hahn, Warnick, & Plemmons, 1999), rooted partly in traditional wisdom which suggests that those in treatment should avoid major life changes (including smoking cessation) early in recovery, and that stopping smoking may jeopardize drug/alcohol recovery (Burling, Ramsey, Seidner, & Kondo, 1997). However, some research has indicated that nicotine dependence treatment does not jeopardize recovery, and may help maintain it (Bobo, McIlvain, Lando, Walker, & Leed-Kelly, 1998). Prochaska, Delucchi, and Hall (2004) found that smoking cessation efforts concurrent with addiction treatment were associated with a 25% increase in long-term abstinence from other drugs.
In a review of the literature concerning the range of barriers to provision of nicotine dependence services in addictions treatment settings (Guydish, Passalacqua, Tajima, & Manser, 2007), three barriers were prominent. Lack of knowledge and training was reported as a barrier to smoking cessation treatment in a variety of settings, including Veterans Affairs (VA) programs (Willenbring et al., 2004) and methadone clinics (McCool, Richter, & Choi, 2005), and by both clinic leaders (McCool et al., 2005) and staff (Gill & Bennett, 2000) in the U.S. and Australia (Walsh, Bowman, Tzelepis, & Lecathelinais, 2005). This widely reported barrier is of interest because studies have found that an increase in staff knowledge or training resulted in a higher provision of smoking cessation treatment (Bobo, Slade, & Hoffman, 1995) and changes in staff attitudes toward smoking cessation treatment (Gill & Bennett, 2000). Second, staff beliefs that smoking cessation is stressful and may represent a risk to sobriety for another addiction are common in the treatment field. In Knapp, Rosheim, Meister, and Kotte (1993) for example, 77% of counselors recommended that the most appropriate time to address client smoking was after one year of sobriety. The perception that smoking cessation is a risk to sobriety, while appearing frequently in the review, was sometimes endorsed by only a small proportion of staff. The third frequently occurring barrier was that program staff members are themselves smokers, and staff smoking prevalence rates in the papers reviewed ranged from 14% to 40% (Guydish et al., 2007).
These barriers to providing smoking cessation intervention have been conceptualized as organizational barriers, and at least one intervention has been developed to specifically address such barriers (Hoffman et al., 2004). However, measuring staff knowledge, attitudes, and practices is challenging, and most research to date has relied on comparisons of individual staff survey items to assess changes in these constructs. In this study we used staff survey data from multiple clinics to develop measurement scales. Because scales offer more robust and stable measures of underlying constructs than single item measures, these scales may be useful to other investigators concerned with changing the organizational climate related to smoking intervention. This paper reports on the development of a smoking knowledge, attitudes, and practices (S-KAP) instrument, which includes scaled measures of these constructs and may be useful to treatment program administrators, and to others interested in changing organizational philosophy and practices related to smoking cessation.
Data for scale development and analysis were drawn from two workforce survey studies related to nicotine dependence. The first study was designed to investigate organizational changes that may occur when smoking cessation treatments are tested, through clinical trials, in drug abuse treatment programs and HIV-care clinics. This study included two intervention clinics where the nicotine dependence treatments were tested, and three comparison clinics where no treatments were tested (Chun, Guydish, & Delucchi, in press). Three community based drug abuse treatment clinics were later added to the study, to enable comparison of knowledge, attitudes and practices across a range of clinical settings, including VA drug treatment clinics (where there are national mandates for smoking assessment and treatment), hospital -based HIV and drug treatment settings, and community-based drug treatment settings (Tajima, Guydish, Delucchi, Passalacqua, & Moore, in press). The second study investigated the effectiveness of a manualized organizational change intervention in three residential treatment programs that were designed to support programs in integrating nicotine dependence treatment (Hoffman et al., 2004; Williams et al., 2005; Ziedonis et al., 2007).
Survey items were drawn from prior studies of healthcare professionals (Borrelli et al., 2001; Goldstein, DePue, & Monroe, 1998; Velasquez et al., 2000), the National Cancer Institute’s “four A’s” approach (Glynn & Manley, 1989), and the Fagerstrom Test for Nicotine Dependence (Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991). Knowledge items were selected from the CDC Adult Tobacco Survey (Centers for Disease Control and Prevention, n.d.) and the California Adult Tobacco Survey (California Department of Health Services, 2004). Survey items were selected to describe respondent demographics, staff smoking status and interest in quitting, knowledge about the risks of smoking, attitudes toward treating nicotine dependence in the context of drug abuse treatment, practices used to address smoking among clinic clients, and barriers to providing such services. Survey drafts were reviewed and modified with input from the study team and from smoking researchers. Scale development was not the original aim of this work, so other standard test construction approaches were not used in development of the survey. Once data had been collected, we explored whether scales emerged from the items selected.
The survey was administered to staff in 11 treatment programs. Eight of the programs were located in California while the remaining programs were located in Oregon, Ohio, and Massachusetts. Two programs were hospital-based HIV care clinics while the remaining programs were drug abuse treatment clinics. Among the drug abuse treatment clinics, three were VA outpatient clinics, three were residential treatment programs, two were community-based outpatient drug abuse treatment programs, and one was a hospital-based methadone program.
In each program, all paid employees who worked at least 20% time, including both administrative and clinical staff, were eligible for participation.
In each clinic, one staff member was designated as a liaison to the study team. The research liaison provided the study team with a list of names of all program staff meeting study eligibility criteria, and arranged a staff meeting when the study could be introduced and those eligible could be invited to participate. The study team assigned a research identification number to each staff person, and prepared survey packets in advance containing informed consent documents, the survey, and a pre-addressed and stamped return envelope.
Staff meetings were held at lunch time to minimize disruption of clinic activities, and lunch was provided to encourage attendance. A member of the study team (8 clinics) or the research liaison (3 clinics) explained the nature of the study and distributed the survey packets. Survey packets had a removable label on the outside that contained the name of the staff member, and all other documents were identified by the research identification number assigned to that staff member. When the packet was distributed to the staff member, the name label was removed. Some clinics required signed consent forms as a condition of participation while other clinics allowed the use of a participant information sheet detailing study procedures, but not requiring participant signature. The study team member or research liaison was available throughout the meeting to answer questions.
Staff who wished to participate in the study had the option of completing the survey immediately and handing it to the researcher, or completing it later and returning it in the enclosed envelope. Staff who did not wish to participate could tell the researcher directly, or return a blank survey and indicating “decline to participate” on the front page. Staff members absent from the meeting were recruited by phone, using their work number, and received a survey packet by mail. The survey was brief, self-administered, and confidential.As an incentive, staff members received a $25 gift card redeemable at local vendors. All study procedures were approved by the Institutional Review Board of the University of California, San Francisco.
Summary statistics were used to characterize both the participants and their responses. Exploratory factor analysis using principal factor analysis with varimax rotation was employed to extract the underlying factor structure of the items, including the Minimal Average Partial-correlation approach (MAP) (Zwick & Velicer, 1986) to aid in determining the optimal number of factors to extract. Communalities were estimated by using the squared multiple correlation of each item with all other items, and computations were accomplished via SAS Proc Factor.
Scale properties were indexed by Cronbach’s alpha coefficient with 95% confidence intervals using Kistner and Muller’s F approximation (Kistner & Muller, 2004). Correlations of items with total scale scores were checked.
Across all participating agencies there were 384 eligible staff members, and 363 (95%) responded to the survey. Some respondents (n=27) indicated on the survey that they declined to participate, giving 336 (88%) completed surveys.
Participant characteristics are summarized in Table 1. Two thirds were women and over half were White (58%), while 25% were African-American and 11.7% were Hispanic. Over a third (35.2%) had some college training but less than a bachelor degree, 18.8% held bachelor degrees, and 34% held postgraduate degrees. Fewer than half (42.3%) reported that they were in recovery from addiction, 37.9% held a professional license or certification, 71.1% had direct patient care responsibility, and 26% were current smokers.
The survey contained 50 items concerning knowledge, attitudes, and practices related to smoking. All items were analyzed, but preliminary results indicated that four items did not fit well on any of the scales, in that they did not correlate with the scale totals, and either did not contribute towards or detract from the alpha coefficients. These items, listed below, were set aside and not used in further analysis:
As the scoring for the remaining 46 items varied in format (see Table 4), we recoded four items with a three-point response from the common 1, 2, 3 scoring to 1, 3, 5, allowing all items to be analyzed using a common scale ranging from 1 to 5. The eigen structure of the data had eight eigenvalues greater than 1.0 with the last one at 1.04, a small number of items loading on the sixth, seventh and eighth factors, and some items split between factors. The MAP analysis pointed towards six factors but the sixth one was formed by only three items, two of which split between factors. Examination of a five factor solution pointed towards scales with apparent substantive meaning. For the sake of parsimony we generated those five scales by averaging item scores. The rotated factor pattern is shown in Table 2.
Review of the item content suggested labels for the scales. As expected, one related to knowledge about the effects of smoking (Knowledge, Factor 2) and one to smoking cessation practices (Practices, Factor 1). The remaining three scales were related to perceived barriers to smoking cessation (Barriers, Factor 3), their perceived self-efficacy to implement smoking cessation treatment (Efficacy, Factor 5) and their beliefs and attitudes towards cessation (Beliefs, Factor 4). Two items which loaded primarily on the first factor (Practice) also loaded noticeably on the fifth (Efficacy). These items are: “I know where to refer patients for help with smoking cessation” and “I have the required skills to help my patients quit smoking.” These two items were moved to the Efficacy scale for the sake of face validity and virtually no decrements in psychometric characteristics were found. Scale statistics and inter-correlations are shown in Table 3 and the final scales with response options are provided in Table 4.
Alpha coefficients for the five scales (top line in Table 3) ranged from .72 to .91 with all items contributing positively to their scale. Confidence intervals for the coefficients, shown in parentheses, were reasonably narrow. The correlations indicate a fair amount of independence among the scales except for the Practice scale correlating with Self-Efficacy (r=.56) and Beliefs (r=.52), and Self-Efficacy also correlating with Beliefs (r=.50).
The aim of this study was to apply factor analytic techniques to staff survey questions frequently used in the study of smoking in drug abuse treatment settings, and to identify scales that may measure relevant constructs more robustly than individual survey items. The results identified five scales having reasonably good basic psychometric properties. These scales reflect knowledge of smoking risks, attitudes toward treating nicotine dependence in the context of drug abuse treatment, counselor self-efficacy in providing such services, barriers to providing such services, and practices used to address smoking among clinic clients. These scales may be useful, for example, for assessing whether staff training regarding nicotine dependence results in increased smoking-related knowledge, more favorable attitudes towards treating smoking in drug abuse treatment settings, or increased self-efficacy when addressing smoking with clients. Interventions designed to change clinical practice, like the Addressing Tobacco Through Organizational Change Intervention (Ziedonis et al., 2007), or large scale policy interventions designed to increase smoking cessation intervention in statewide treatment systems, like those in New Jersey (Williams et al., 2005) or New York (Tobacco-Free Services Title 14, 2008), could use the practice scale to assess whether counselors are delivering more smoking-related services to clients over time.
Limitations to the current study include the process of scale development used and the need for large and diverse samples in factor analysis. The process of scale development is usually an iterative process, beginning with the collection of a set of items selected with the goal of forming those items into a scale as the organizing force. The items used in this analysis were not preselected with scale development in mind but, as in a number of staff survey studies related to smoking, used individual items to measure knowledge, attitudes, and practices. Because of this, there was variation in response choices across items. Even with this variation, however, the items formed scales with face validity and internal consistency, offering a strong first step towards the development of more refined items and scales. Future work focused on these scales should first consider confirming these findings, preferably in a larger sample to confirm that these scales replicate. It may also be appropriate to use some form of control group to examine change in these measures over time. Finally, the psychometric properties may be improved by refining the responses so that they use a common format.
The four “orphaned” items, those not fitting in with the others, are interesting to consider. The two factual questions, one related to light cigarettes and one to the benefits of quitting for someone with 20 years of smoking, apparently address facts not well-known among clinical providers. Compared to other bits of knowledge about smoking cessation, these are more recent findings, but point to the need to assure that providers have the most up-to-date information available. Answers to the question as to whether nicotine dependence treatment should be offered to clients who smoke may reflect a complex issue regarding the various factors that go into treatment selection as well as the issues stemming from traditions as discussed in the introduction of this article. The final question has a fairly complex format which may have confused some respondents, and the clinicians’ answers may reflect the same range of treatment selection factors influencing the answers given to the third of these four questions.
Persons enrolled in drug abuse treatment smoke at a rate 3-4 times that of the general population and. because of this, bear a disproportionate burden of illness related to smoking. Individual treatment programs have initiated efforts to better address smoking among both clients and staff (Bernstein & Stoduto, 1999; Hurt, Croghan, Offord, Eberman, & Morse, 1995; McDonald, Roberts, & Descheemaeker, 2000), the VA system has made broad and enduring efforts to better address smoking across all specialties of medical/psychiatric/addiction care (Isaacs, Schroeder, & Simon, 2004), and at least three states have launched policy initiatives to better address smoking among persons in publicly-funded drug abuse treatment systems (Williams et al., 2005; Tobacco-Free Services Title 14 NYCRR, 2008; Toussaint, VanDeMark, Silverstein, & Stone, in press). As drug treatment programs and systems implement innovative smoking cessation interventions and policy approaches to better address nicotine dependence, there is an increasing need for improved strategies to measure outcomes of these interventions and approaches. The measurement scales reported here offer tools to measure program change in response to these interventions.
This work was supported by grant R01 DA020705, U10 DA015815, and P50DA09253 from the National Institute on Drug Abuse.
Kevin L. Delucchi, Ph.D., Professor of Biostatistics in Psychiatry at the University of California, San Francisco. His primary field of research is the application of statistical methods to studies of drug and alcohol abuse.
Barbara Tajima, Ed.M., Senior Public Administrative Analyst at the Institute for Health Policy Studies at University of California, San Francisco. Her areas of interest are addressing nicotine dependence in substance abuse treatment programs, treatment effectiveness and the adoption of new treatments into practice.
Joseph Guydish, Ph.D., MPH, Professor of Medicine at the University of California, San Francisco. His research is in the area of access, delivery, and organization of substance abuse treatment services, treatment effectiveness, and adoption of new treatments into practice settings. He is currently testing strategies designed to support drug abuse treatment programs in better addressing nicotine dependence.