A Feasibility Study of Home-Based Contingency Management with Adolescent Smokers of Rural Appalachia

Abstract

Cigarette smoking among adolescents remains a significant public health concern. This problem is compounded in regions such as rural Appalachia where rates of smoking are consistently higher than national averages and access to treatments is limited. The current research evaluated a home-based contingency management program completed over the Internet with adolescent smokers recruited from rural Appalachia. Participants (N = 62) submitted three video recordings per day showing their breath carbon monoxide (CO) levels using a handheld CO monitor. Participants were assigned to either an active treatment condition (AT: n = 31) in which reductions in breath CO were reinforced or a control treatment condition (CT: n = 31) in which providing timely video recordings were reinforced with no requirement to reduce breath CO. Results revealed that participants in the AT condition reduced their breath CO levels significantly more so during treatment than participants in the CT condition. Within-group comparisons revealed that participants in both conditions significantly reduced their breath CO, self-reported smoking, and nicotine dependence ratings during treatment. However, only participants in the AT condition significantly reduced urinary cotinine levels during treatment, and only participants in this condition maintained all reductions until six-week post treatment. Participants in the CT condition only maintained self-reported smoking reductions until post-treatment assessments. These results support the feasibility and initial efficacy of this incentive-based approach to smoking cessation with adolescent smokers living in rural locations.

Appalachia is a large region of the Eastern United States, including all of West Virginia and parts of 12 other states, where rates of smoking for adolescents and adults are consistently higher than national averages (Behavioral Risk Factor Surveillance System [BRFSS], 2011). Several factors may contribute to these elevated rates of smoking, including a lower average socioeconomic status, lower levels of education, and in many areas of Appalachia, and a reliance on tobacco production as a means of economic support (e.g., Meyer et al., 2008). Youths who grow up in tobacco-growing regions start smoking at younger ages (Noland et al., 1990), which also is a pattern observed among adolescents in Appalachia (Meyer, et al., 2008). Not surprisingly, cancer incidence rates in this region are higher than the U.S. national average, and the prevalence of cardiovascular disease (which cigarette smoking contributes to) exceeds that of much of the developed world (Wingo et al., 2008).

Research to evaluate intensive behavioral smoking cessation interventions specifically oriented to adolescents who live in rural areas is limited (but see Horn et al., 2005; Woodruff et al., 2001, for exceptions). Such research could lead to significant public health benefits in a population that has been largely neglected in this regard; however, the implementation of such programs in regions like Appalachia involves barriers that require special consideration. For example, much of Appalachia is geographically isolated, with 42% of the population being defined as rural compared to 20% of the population nationally (Appalachian Regional Commission, 2001). Often, such remoteness leads to difficulties in transportation to and from treatment centers, especially for younger smokers who are highly dependent on others for transport. Consistent with this notion, lack of transportation for younger smokers in Appalachia has been cited as the primary barrier to smoking cessation treatment programs (Kruger, et al., 2012).

The purpose of this research was to implement and evaluate a web-based contingency management (CM) program that could be completed from home with adolescent smokers recruited from rural Appalachian Kentucky, Ohio, and West Virginia. Typically, these abstinence reinforcement therapies provide tangible reinforcers (usually cash or vouchers exchangeable for purchases) of increasing value for frequently verified target behaviors (e.g., abstinence from smoking), with the verifications being made in person to ensure validity of the samples. The efficacy of CM programs for drug use treatment and smoking cessation has been well established (see Petry, 2000). For example, through meta-analytic review CM has been shown to have larger effects for reducing substance use than psychosocial treatments such as relapse prevention and cognitive behavioral therapy (Dutra et al., 2008). However, other analyses have shown that these effects from CM decline substantially following treatment and that CM may lead to larger effects for treating opiate or cocaine use compared to tobacco (Prendergast et al., 2006).

In addition, previous research suggests that CM-based treatments can be implemented in an adolescent population and have positive effects on teen smoking cessation. For example, Krishnan-Sarin et al. (2013) report better seven day abstinence outcomes for high school students in CM alone and CM plus cognitive behavioral therapy (CBT) compared to CBT alone. Moreover, adolescents in CM alone had the lowest rates of cigarette smoking at follow-up. Similarly, Morean et al. (2015) found that adolescents who self-reported higher behavioral impulsivity and limitations in self-regulation had better end-of-treatment outcomes in either a CM alone or CM and CBT compared to CBT alone.

The web-based CM program evaluated here has been reported previously in various forms with non-rural adult and adolescent smokers (Dallery, Glenn & Raiff, 2007; Reynolds et al., 2008) and with rural adult smokers who had home Internet service (Stoops et al., 2009). The program requires participants to use a software platform called Mōtiv8 to make video recordings of themselves providing breath samples to verify smoking status. The Mōtiv8 system provides immediate feedback to participants concerning progress in the program (in the form of a graph) and updates on program earnings. Study personnel can access participant videos through Mōtiv8 to verify the validity of submitted breath-sample videos and to confirm program earnings. Importantly, this system makes it possible for rural smokers with limited transportation (e.g., adolescent smokers) to participate in an intensive CM program from home and at distances from a research/clinical facility that would not be possible with more traditional in-person sample verifications. To our knowledge, this is the first published research to evaluate a CM program with rural adolescent smokers.

Eligible adolescent smokers (N = 62 completing participants) were randomized into one of two research arms: an active treatment (AT) arm in which they received spending vouchers for providing timely breath samples indicating abstinence from smoking (CO ≤ 4 ppm) or a control treatment (CT) arm in which vouchers were received for providing timely breath samples but with no requirement for CO verified abstinence. Both voucher-reinforcement intervention programs lasted six weeks; however, smoking-status data were collected at six weeks post treatment (12 weeks following enrollment) to determine if any observed program effects were still present at this post-treatment time point. The purpose of this study was to determine the feasibility of implementing a CM program with rural adolescent smokers and to collect initial efficacy data for this program with this population.

Method

Participants

Ninety non-treatment-seeking adolescents were enrolled in the study. However, 14 dropped out of the study (n = 6 dropped out before randomization, n = 5 from AT, and n = 3 from CT). Also, a total of 14 other participants were disqualified because they were identified as non-smokers (n = 3: 2 from AT, 1 from CT), we could not establish Internet service at their homes (n = 4: 3 from AT, 1 from CT), they had broken or had stolen loaner equipment (n = 6: 3 from AT, 1 from CT, 2 before randomization), and applicant falsified consent (before randomization). Therefore, 62 participants (31 female) completed the study. All participants were recruited from local high schools, advertisement posters, and word-of-mouth referrals. To qualify, participants were required to have an afternoon breath CO level of ≥9ppm (Bedfont Instruments, Micro III, UK) and/or a urinary quantitative cotinine value ≥100ng/ml. These conditions were met by all 62 participants. Cotinine content was determined using a homogenous enzyme immunoassay at J2 Laboratories in Tucson, AZ. Potential participants were prescreened by phone, and those who met eligibility requirements were enrolled in the 6 week programs.

General procedure

Before beginning a pre-treatment laboratory session (see below for further description), an IRB approved consent was reviewed and signed by participants’ parents/guardians and assent forms were reviewed and signed by adolescent participants, unless they were 18 years or older. Consent and assent were obtained either at our laboratory or in our mobile laboratory (see description below).

A dynamic-balanced-randomization procedure (Signorini et al., 1993) was used to assign participants to conditions at the end of the CM baseline phase (see below). Randomization was discontinued if imbalances (20% or greater across conditions) were detected across conditions in a) the distribution of males and females, b) the distribution of “heavy smokers” (i.e., ≥15 daily cigarettes) or “light smokers” (i.e., ≤4 daily cigarettes) c) or the distribution of participants not submitting breath samples during the baseline phase (i.e., ≤15 of the 18 possible samples). Any of these variables could have impacted treatment outcomes if they were disproportionately distributed across the AT and CT conditions.

As described above, the web-based monitoring system involved an Internet website called Mōtiv8 to verify breath CO measurements. Participants were allowed to use their own home computers and Internet services, or they were loaned a laptop (Dell Latitude 2110 with Intel Atom processor) with Verizon Wireless or satellite (StarBand^®) access to the Internet. Satellite was only used when participants did not have sufficient cell signal at their homes to support Verizon Wireless access to the Internet. If satellite was used, the dish was temporarily installed at participants’ homes by study personnel. An outcome of interest for this research was the frequency with which these loaner systems were required for teens to participate in the trial versus teens who already had adequate Internet service at home for the CM program. All participants were loaned a Web camera (Logitech) and a breath CO monitor (piCO Smokerlyzers; Bedfont Inc., Medford, NJ). Research personnel set up the CO sampling equipment and practiced the sampling procedure with participants. Participants were given a username and password to log into Mōtiv8 website and instructed when to begin the program.

During the CM programs, three breath samples were required per day—each separated by at least 5 hours but not more than 8 hours. CO samples were video recorded and securely uploaded to the Mōtiv8 server. Participants were instructed to use the web-camera to show: a) the CO monitor set to zero, b) themselves taking a deep breath and holding it for 15s, c) fully exhaling into the mouthpiece, and d) the final CO reading (see Dallery et al., 2007 and Stoops et al., 2009 for further discussion). Participants then entered their CO reading directly into the Mōtiv8 site and submitted it. All videos and entered CO values were reviewed by study personnel on a daily basis for completeness and accuracy. Video samples not meeting all of the above criteria were considered invalid and ineligible for vouchers.

Participants in both conditions were allowed two “freebies” to count for late or missed breath samples. However, freebies could not be used by participants to replace invalid samples or samples with non-criterion CO levels. When applicable, one freebie could be used only during the abstinence phase, but the other could be used (if used at all) during any of the other program phases.

Participants received their earnings in the form of vouchers, which could be redeemed for purchases at any time. A designated staff member placed on-line orders for participants to pre-approved stores (e.g., Walmart, Best Buy, Macy’s) after requested items were individually reviewed for appropriateness (e.g., no tobacco, weapons, or sexually explicit materials). Items were shipped directly to participants, and the dollar amount plus shipping was deducted from the participant’s total earnings.

Program Phases and Contingencies for the AT Condition

The program phases and contingencies reported here were based on past research using this web-based program with different populations (e.g., Dallery, Glenn & Raiff, 2007; Reynolds et al., 2008; Stoops et al., 2009).

Baseline phase

The baseline phase consisted of the first seven days of the program. The purpose of the baseline phase was to help participants become acclimated to submitting CO breath samples and also to determine their baseline average CO level. During baseline, participants were required to provide three timely breath samples per day with no criteria for CO level. Participants received $6 each day for providing all three samples, but they receive nothing for days with missed or late samples. The participant could potentially earn $42 for this phase.

Shaping phase

Shaping was the second program phase, which lasted the next four days. The purpose of the shaping phase was to allow participants to receive reinforcements for gradual decreases in CO level in preparation for the abstinence phase, where contingencies were only given if CO level were at or below 4 ppm. During the shaping phase, participants were expected to reduce CO levels so that their readings would be ≤ 4 ppm by the last sample of this phase. Before this phase began, study personnel calculated the average breath CO level for each participant over the first six days of baseline to determine the rate of CO reduction required over the twelve breath samples of the shaping phase to reach 4 ppm by the last sample. If CO levels needed to be reduced by more than 3 ppm per day (i.e., by more than 1 ppm per sample), the larger reductions were scheduled for the earlier samples of that day. This was done because CO accumulates in the lungs over the course of a day, thus making it more difficult to reduce breath CO for samples submitted later in the day. Just prior to starting this shaping phase, the participant was given a full schedule of specific “goal-CO” values by e-mail and phone for each of the twelve breath samples of this phase. The seventh day of the baseline phase was used to calculate CO reductions and to provide that information to participants. Participants received $3 for each breath sample at or below their goal-CO value, for a total potential earnings of $36 for this phase.

Abstinence phase

The abstinence phase lasted 21 days. The purpose of the abstinence phase was to promote continued abstinence. The goal for this period was for participants to provide three timely breath samples per day with CO levels meeting the criterion for abstinence (i.e., CO ≤4 ppm). This phase involved an ascending pay schedule with a reset component. The participant received $3 for his/her first breath sample meeting criterion for abstinence, and this amount was increased in increments of $0.25 for each consecutive criterion sample. Additionally, there was a $5 bonus for every five consecutive criterion samples. However, if a participant provided a non-criterion sample (i.e., samples ≥4 ppm), he/she did not get paid for that sample and the next criterion sample was reset to a payment of $3, the original payment amount. The ascending pay schedule resumed as before from $3. After three consecutive criterion samples, the payment amount was returned to the highest established value before the reset occurred. This ascending pay schedule was included as part of the abstinence phase because of the difficulty of maintaining initial abstinence from smoking, with program incentives increasing during the timeframe when participants are more likely to relapse. With the ascending pay schedule and bonuses for consecutive criterion samples, participants could potentially earn $737.25 for this phase.

Thinning phase

After the abstinence phase, the fourth phase was thinning, which lasted five days. The purpose of the thinning phase was to allow participants to gradually transition from incentives for abstinence. For the thinning phase, participants provided three timely samples per day with each having a CO recording ≤4 ppm. Participants received $6 each day three criterion samples were received, for a total potential earning of $30 for this phase.

Return-to-baseline phase

The fifth and final phase was a second baseline period, or return-to-baseline, which lasted five days. The purpose of this program phase was to remove incentives for abstinence and to evaluate whether participants immediately returned to smoking after the incentives for abstinence were removed. Participants were paid $6 per day for providing three timely samples each day, with no specified criterion CO level. This phase was included for the purpose of determining consistency of program effects in the days immediately following removal of program CO contingencies. Participants could potentially earn $30 for this phase.

Program Phases and Contingencies for the CT Condition

The five phases of the CT condition were identical to the phases of the AT condition. Also, the payment schedules and the expected timing of samples were the same as they were for AT. However, there was no set criterion for CO level at any program phase. Participants earned vouchers for providing valid and timely breath samples at any CO level.

Assessments

Pre-treatment assessments

Within two weeks of beginning CM, participants completed a ~2.5 hour laboratory session in our research office or our mobile behavioral research laboratory. The Mobile Research Lab (MRL) includes three computer stations, a bathroom, a private interview room, and a biological samples freezer. Use of the MRL allowed for laboratory data collection across a variety of locations, especially when the research office was too far for participants to travel. Research staff were blind to treatment condition, as randomization had not yet occurred. Participants completed the following self-report measures: 1) demographic questionnaire specifically designed for this study, 2) assessment of IQ (Kaufman and Kaufman, 2004), 3) Time Line Follow-Back (TLFB) procedure for daily smoking frequency over the preceding two weeks (see Lewis-Esquerre et al., 2005), 4) Stage of Change Ladder (SCL; Rustin & Tate, 1993) to measure changes in the transtheoretical model of quitting smoking, and 4) Modified Fagerström Test for Nicotine Dependence (mFTND; Rojas et al., 1998). Participants also submitted breath and urine samples to verify smoking status, were asked about smokeless tobacco use over the previous 14 days and completed laboratory behavioral measures of impulsivity (see Harris et al., 2014, for description of these measures); however, impulsivity results are not reported here. Participants were compensated between $40 and $50 depending on behavioral-task performance.

During-treatment assessments

During participation in the CM programs, data were collected the day before starting the program and at the approximate midpoints of the baseline (days 3–4 during baseline), abstinence (days 10–12 during abstinence) and return-to-baseline (days 2–3 of return-to-baseline) phases. These data included urine samples for cotinine analyses, the TLFB, SCL, and the mFTND. Participants were compensated $10 per during-treatment assessment, for a total of $40.

Post-treatment and follow-up assessments

Within two weeks of finishing a program, participants completed a post-treatment laboratory session. This session included the same measures as the pre-treatment assessment. Participants were compensated between $40 and $50. Also, for the six weeks following the end of treatment, there were weekly assessments that included measures of smoking status and urges to smoke. However, only the data collected at the final time point (six weeks post-treatment) are presented here, as data from this time point provides the most temporally-extended estimate of continued program effects post-treatment. Participants earned $10 per follow-up assessment, for a total of $60.

Statistical analysis

Demographics were compared using chi-square, Fisher’s exact test and t-test, as appropriate. Because multiple breath CO levels were collected during each CM program phase, the mean CO level at each phase was calculated and utilized as the primary outcome measure of abstinence from smoking. Group difference effect sizes (d) are reported for program breath CO levels. The percentage of CO measurements collected during each phase of the CM programs also was calculated for every participant as a secondary outcome of overall program adherence. Overall program adherence was tested using Wilcoxon rank sum test because these data were bimodal. The change in mean CO level for each program phase was modeled with a mixed model controlling for the five repeated measures of CO and adjusting for overall adherence (i.e., percentage of program breath samples submitted) and gender. Because the phases were temporal, an autoregressive correlation structure was utilized to model the intraclass correlation. All reported pairwise p-values were adjusted using the Holm-Sidak method for between and within condition comparisons independently. Since the first 5 measures of CO in the model are means and the 6 week is a single measure, the 6 week follow-up data was not included in the CO model.

Secondary analyses were performed to explore the possibility that any observed AT and CT group differences in program breath CO levels may be attributable to group differences in program adherence. We conducted sensitivity analyses in which CO values were generated and imputed for missing values based on available CO data from the CT condition. By basing estimates and imputed data on CO values from the CT condition (where CO values are likely to be higher than in the AT condition) we would be making a conservative estimate of what missing CO values may have been in the AT condition.

Daily cigarette usage and cotinine were modeled similarly at the four phases in which they were measured: baseline, adherent, return to baseline and 6 week follow-up. Smokeless tobacco use was recorded as a binary variable and was modeled with an analogous model to control for the dichotomy and with participant sex removed since only males answered positively. For the SCL and mFTND measures, change from the pre-treatment value was calculated for the baseline, abstinence, return-to-baseline program phases and 6 week follow-up. These calculated changes were then analyzed with mixed models similar to the other outcome measures. Changes from baseline to the both the return to baseline phase and the 6 week follow-up measure were estimated from the model. All statistical analyses were performed using SAS 9.4. Alpha was set at .05 for all statistical tests

Results

Participants

Table 1 shows self-reported demographic, substance use, and smoking background data for participants assigned to AT and CT conditions and participants who dropped out of the CM programs. There were no statistically significant differences between these groups on any of these variables.

Table 1.

Participant Demographics and Baseline Information (N = 76)

	AT (n=31)	CT (n=31)	Dropout (n=14)
Age (years; M [SD])	16.58 (1.54)	16.71 (1.32)	17.57 (.94)
Sex (% female)	48.4	51.6	50
Race (n; black:white:other)	1/28/2	2/28/1	0/14/0
KBIT 2 (IQ Standard Score; M [SD])	91.60 (8.32)	92.32 (12.35)	88.36 (10.50)
Cigarettes (number per day; M [SD])1	14.25 (8.68)	12.13 (8.67)	15.41 (11.68)
Nicotine Dependence (M [SD])	5.45 (2.28)	4.68 (3.93)	5.14 (1.91)
Carbon monoxide (ppm; M [SD])	11.16 (6.58)	10.52 (6.87)	11.79 (6.53)
Cotinine (ng/ml; M [SD])	1335.46 (931.65)	1206.23 (697.35)	1476.27 (555.30)
At least one parent smokes (% reporting yes)	83.9	90.3	92.9
How many friends smoke (M [SD])2	3.90 (.91)	3.42 (1.15)	3.71 (.91)
Closest/best friend smokes (% reporting yes)	87.1	83.9	71.4
Contemplation Ladder (Scale 0 to 10; M [SD])	7.63 (2.11)	7.03 (2.33)	6.71 (1.94)
Marijuana (M [SD])3	2.35 (1.78)	2.13 (1.91)	1.50 (1.70)
Alcohol (M [SD])3	2.06 (.77)	1.61 (1.14)	1.36 (1.15)
Smokeless Tobacco (M [SD]) 4	0.85 (2.78)	0.79 (2.80)	1.86 (4.38)

¹Cigarettes per day were calculated using a timeline follow back calendar to determine cigarettes smoked each day during the past 14 days.

²Friends who smoke was assessed using the following question: “How many of your friends smoke cigarettes/black & milds?”: 1 = none, 2 = some, 3 = half, 4 = most, 5 = all.

³Drug use was assessed with the following question: “Thinking about the past six months, how often have you used the following substances?”: 0 = never tried, 1= tried it, 2 = 1–2 times/month, 3 = once a week, 4 = 2–4 times/week, 5 = 5 or more times a week.

⁴Smokeless tobacco use was assessed with the following question: “Do you use any form of smokeless tobacco?”

Note. There were no significant group differences at p < .05.

Program efficacy

There was an overall difference in mean breath CO levels across the AT and CT conditions (p=0.002). As shown in Figure 1, the mean program CO levels were significantly different across conditions for all program phases except baseline. The adjusted mean CO value during the abstinence phase was 4.65 ppm for the AT group and 9.49 ppm for the CT group (p=0.0002, d = 1.252). While the difference in adjusted mean CO value was not as dramatic for other post-baseline phases, the differences were still highly statistically significant (ps<0.03, ds > .730). Participants in the AT condition reduced their average CO from baseline to return-to-baseline phase by 4.4 ppm (p<0.0001), and participants in the CT condition reduced CO levels by 1.7 ppm, which approached statistical significance (p=0.06). The difference in change between the conditions showed a trend, but was not statistically significant (p=0.09).

Figure 1.

Mean (±SEM, denoted by error bars) breath CO recordings for participants in the AT and CT study conditions at each treatment program phase.

Note. + <0.01, * <0.05 different from control; # <0.01, @ <0.05 different from baseline.

There was no statistically significant between-group difference on the TLFB calendar; however, within-group comparisons revealed significant effects (Figure 2). By the end of treatment, the average number of cigarettes smoked each day had dropped in the AT group from 11.3 cigarettes to 6.0 cigarettes (p=0.0003), and the CT group dropped from 11.8 to 8.1 cigarettes per day (p=0.01). While both groups reduced their self-reported smoking, the average change from baseline to return-to-baseline phases between groups was not statistically significant. The decrease in the average daily cigarettes was sustained at the 6 week follow-up and remained statistically significant within each group, but not between groups. The AT group was still smoking 4.0 cigarettes less (p=0.01) while the CT group was smoking 3.3 cigarettes less daily (p=0.02).

Figure 2.

Mean (±SEM, denoted by error bars) number of cigarettes smoked per day based on TLFB calendar reporting by participants in the AT and CT study conditions at three treatment program phases and at six weeks post treatment.

Note. + <0.01, * <0.05 different from control; # <0.01, @ <0.05 different from baseline.

There was no statistically significant difference in cotinine values between groups. However, cotinine values dropped dramatically for both treatment conditions between the baseline phase and the abstinence phase, with an increase during the return-to-baseline phase (Figure 3). The AT condition demonstrated a significant decrease from baseline to return-to-baseline (p=0.04) and it continued at the 6 week follow-up (p=0.01).

Figure 3.

Mean (±SEM, denoted by error bars) urinary cotinine values for participants in the AT and CT study conditions at three treatment program phases and at six weeks post treatment.

Note. + <0.01, * <0.05 different from control; # <0.01, @ <0.05 different from baseline.

Overall, only 8 males responded positively to using smokeless tobacco in the previous 14 days during any point in the study. No differences were seen between conditions during any phase. Similarly, no change in SCL ratings from the pre-treatment evaluation over the course of the study was very small. No significant changes or differences were observed.

Finally, no significant differences were found between the two conditions in mFTND scores; however, both conditions experienced decreases, but only the AT group reached significance. Participants in the AT condition decreased an average of 1.7 points (p=0.001); and in the CT condition participants dropped an average of 1.0 point (p=0.06). At the 6 week follow-up the AT group maintained the significant decrease in scores by 1.5 points (p=0.007).

Percentage of breath samples obtained through Mōtiv8

For the entire sample, the median percentage of submitted program breath samples through Mōtiv8 was 46.03%. There was a significant difference in percentage of samples submitted between AT and CT conditions. For participants in the AT condition, the median percentage of submissions was 38.10%, while participants in the CT condition submitted 69.05% (p=0.011). Furthermore, there was a significant difference in vouchers earned. On average, participants in the AT condition earned $147.90 and participants in the CT condition earned $345.26 (p=0.003). Overall, females tended to be more adherent than males (61.11% vs 38.10%, p=0.083). Females in the CT condition were significantly more adherent (81.00%; p=0.01) than males or females in the AT condition and males in the CT condition.

Because the females in the CT condition submitted 81% of their program breath samples, these data were used for sensitivity analyses to estimate missing data for all participants. The results from these sensitivity analyses revealed that outcomes were qualitatively the same as the results reported above without imputed values for missing CO data—with the exception of the return to baseline phase. After imputing missing CO values, the group differences in program breath CO levels only approached statistical significance (p = .06). These secondary analyses suggest that the overall observed group differences in program breath CO levels are not likely attributable to group differences in program adherence.

Loaner equipment for Internet access

Due to an early error in record keeping, we have data on use of loaner equipment only for the last 42 participants (67.7% of the total sample). Of these participants, 90.48% borrowed a laptop and 6.25% required the satellite for Internet service.

Discussion

The primary goal of this research was to evaluate efficacy of a home-based contingency management program for smoking cessation with adolescent smokers recruited from rural Appalachia. In comparing outcomes for the AT and CT conditions, there were minimal between subject effects. However, participants in the AT condition reduced their program breath CO levels significantly more so than participants in the CT condition, which is consistent with earlier findings in adult smokers (e.g., Dallery, Glenn & Raiff, 2007; Stoops et al., 2009).

There were no other statistically significant group differences in smoking related variables. However, there were significant within-group effects for the AT condition. For example, both groups reported significant decreases in self-reported smoking and mFTND scores, but only participants in the AT condition showed a significant decrease in cotinine levels. Furthermore, at six weeks post-treatment participants in the AT condition still had significantly reduced levels of breath CO, cotinine, self-reported smoking, and mFTND ratings when compared to baseline assessments. Participants in the CT condition only had significantly lower rates of self-reported smoking and breath CO. Taken collectively, these findings support the conclusion that the AT program was effective in reducing smoking in this population of adolescent smokers.

A unique feature of this study is that it was conducted in a rural region of Appalachia where broadband service is limited. Because of the study location, loaner equipment was made available to participants on an as-needed basis so that they could access the internet from home to submit video breath samples. The vast majority of participants for which we tracked use of loaner equipment made use of these systems (> 95%) and could not have participated in the trial otherwise. In terms of program feasibility, this finding illustrates the importance of providing at least some internet-access support to adolescent smokers in this region who are attempting to use web-based programs from home. Notably, most of the participants who required loaner equipment were able to use Verizon Wireless to access the internet (90%), and only a handful of participants made use of StarBand® satellite service.

Overall, program adherence in terms of timely breath samples submitted through Mōtiv8 was low, with an overall median of just under 50% of potential samples received. Participants in the AT condition submitted significantly fewer breath samples than participants in the CT condition. This finding across conditions is not surprising given that participants in the AT condition were required to both provide timely samples and reduce breath CO to earn vouchers whereas participants in the CT condition were only required to provide timely samples. However, results also revealed that this group difference was largely accounted for by the female participants in the CT condition, who on average submitted more than 80% of their samples. Male participants in the CT condition did not differ significantly in percentage of samples submitted from male or female participants in the AT condition. It is not clear why females in the CT condition provided more breath samples than any other participant groups. Perhaps females were more motivated by program incentives and were more tolerant of the program demands and schedule. However, females in the AT condition submitted far fewer breath samples, suggesting that the requirement to reduce breath CO in the AT condition may have neutralized possible sex differences in program adherence.

Program adherence for this study was substantially lower than reported in a similar study with adult rural smokers (Stoops et al., 2009). In that study, more than 65% of samples were obtained from both active and control groups. However, participants in this study had to submit three breath samples per day, whereas the earlier research required only two samples per day. On average, adolescent smokers are lighter smokers who smoke more sporadically over the course of the day compared to adult smokers. As a result, in planning this program it was anticipated that more frequent sampling would be required for adolescent smokers to offset the greater likelihood of false negative CO readings. However, future research using this type of program with younger smokers will have to weigh the benefits of more frequent sampling against the possible impact of this approach on program adherence. Unfortunately, the current study does not provide data to clarify whether the low adherence rates observed here were due to frequent CO sampling or the fact that these were adolescent smokers. Notably, secondary sensitivity analyses suggested that observed group differences in program CO levels were not likely caused by group differences in program adherence.

There were limitations to this research that should be acknowledged. First, the sample was relatively small. As a result, there was limited statistical power to detect possible group differences, e.g., TLFB calendar (Figure 2) and mFTND ratings (Figure 4). With a larger sample, these observed group differences may have become statistically significant. Another limitation involved the sample being almost entirely white; however, given that more than 90% of the population in this region is white (Appalachian Regional Commission, 2014) this sample is roughly representative of race for rural Appalachia in these states. Also, cotinine levels remained high during the treatment programs, which may indicate other forms of nicotine intake (e.g., nicotine replacement or smokeless tobacco) that did not elevate breath CO. This could be a limitation of using only breath CO to determine nicotine intake during the CM program. Finally, program adherence and resulting group differences in program compensation could be considered limitations of this research.

Figure 4.

Mean (±SEM, denoted by error bars) mFTND change scores based on ratings by participants in the AT and CT study conditions at three treatment program phases and at six weeks post treatment.

Note. + <0.01, * <0.05 different from control; # <0.01, @ <0.05 different from baseline.

This research implemented and evaluated a home-based smoking cessation intervention with adolescent smokers recruited from a rural region of Appalachia that has limited broadband service and where rates of smoking are consistently higher than national averages. Participants assigned to a condition in which reductions in breath CO were reinforced showed greater overall decreases in breath CO compared to control participants, and there were significant within-group effects six-weeks after treatment for those receiving CO-reduction reinforcement. Also, study loaner equipment that could be used to access the Internet and complete this incentive-based program was highly utilized by participants, suggesting loaner equipment may be an important consideration in efforts to increase feasibility of such programs in this region. However, future research should take steps to improve program adherence, either by decreasing participant burden (e.g., fewer smoking status verifications per day) or by providing more coaching support to participants who miss scheduled samples. It is possible that improving program adherence will correspondingly improve smoking-related outcomes from this program.