€¦ · web viewidentification of three different types of smokers who are not motivated to...

Subtypes of Unmotivated Smokers 1

Identification of Three Different Types of Smokers Who Are Not Motivated to Quit:

Results from a Latent Class Analysis


Abstract

Objective: The majority of smokers are not motivated to quit within 30 days. We examined

whether these smokers are a homogeneous group, hypothesizing that subtypes of unmotivated

smokers could be identified. Methods: 500 smokers not ready to quit within 30 days

completed an online survey assessing variables known to be associated with quitting.

Results: Latent Class Analysis revealed three unmotivated smoker subtypes. ‘Health-

Concerned Smokers,’ (HCS; n=166) had a significantly greater proportion of previous

smoking-related illness and high risk perceptions. ‘Smokers with Psychosocial Barriers’

(SPB, n=192) had a significantly greater proportion of younger smokers, partners who

smoked, other household smokers, and children. ‘Unconvinced Smokers’ (UCS, n=142) had

the lowest proportion of those who: were motivated/confident to quit, had smoking-related

illnesses, and perceived the risks of smoking/benefits of quitting. UCS had the highest

proportion with optimistic bias, and no prior quit attempts. A greater proportion of HCS had

high motivation to quit vs. SPB and UCS. In model validation, 60.6% of UCS said they

‘never plan to quit’ vs. 31.8% of SPB and 22.3% of HCS; SPB and HCS had lower odds of

never planning to quit vs. UCS. Of those who plan on quitting at some point, 75.2% of HCS

and 62.6% of SPB plan on quitting within 1 year, vs. 46.4% of UCS; the cumulative odds of

planning to quit later were higher among UCS. Conclusions: Smokers who are not

motivated to quit are not a homogeneous group; tailored intervention approaches and targeted

messages might be needed to motivate quitting.

Keywords: Smoking; Motivation; Latent Class Analysis


Introduction

Although the prevalence of smoking in the United States has declined to 15%

(Centers for Disease Control and Prevention, 2015), and 70% of smokers want to stop

smoking at some point (Centers for Disease Control and Prevention, 2011), two questions

remain: 1) how can motivation to quit be accelerated before smoking-related illness occurs

(or worsens) among the 70% of smokers who want to quit at some point, and 2) how can the

remaining 30% who never see themselves quitting smoking be encouraged to try to quit

smoking? Very little is known about the characteristics of smokers who are not ready to quit.

Currently available evidenced-based treatments are specifically designed for smokers

who are ready to quit within 30 days, which constitute only 12% of smokers (Office for

National Statistics, 2009). Similarly, the majority of research trials recruit only smokers who

are interested in quitting within 30 days, excluding all others. It cannot be assumed that

interventions that are effective for smokers who are motivated to quit will be effective for

those who are not motivated to quit, because there are important differences in demographic,

psychosocial and smoking behavior characteristics between these two groups (Bartlett,

Borrelli, Armitage, Tooley, & Wearden, 2015; Borrelli, Bartlett, Tooley, Armitage, &

Wearden, 2015; Christiansen, Reeder, TerBeek, Fiore, & Baker, 2015). Therefore, while we

have a good understanding of what will help smokers who want to quit smoking do so, it is

unclear what interventions might be most appropriate to utilize with smokers who are not

motivated to quit.

It is important to characterize smokers who are not motivated to quit by examining

within-group differences so that interventions can be designed to be more effective and

efficient. The first step in this process is to determine whether smokers who are not motivated

to quit are a homogeneous group (necessitating a singular intervention approach) or whether

they are comprised of distinct subgroups (which may warrant differential intervention


approaches and messages). In general, smoking cessation studies have shown that tailored

smoking cessation interventions outperform non-tailored interventions (e.g., Armitage &

Arden, 2008; Westmaas, Bontemps-Jones, Hendrinks, Kim, & Abroms, 2017; Head, Noar,

Iannarino, & Grant, 2013), with less support for tailoring using only stage of change (Cahill,

Lancaster, & Green, 2010).

Previous studies have found evidence for subtypes of smokers who are not motivated

to quit, but include only smokers who were participating in smoking cessation trials and a set

of variables limited to the Transtheoretical Model (Norman, Velicer, Fava, & Prochaska,

2000; Schorr et al., 2008). For example, Schorr, et al. (2008) identified four subtypes of

smokers in the precontemplation stage (do not plan on quitting in the next 6 months) based on

two variables from the Transtheoretical model (pros and cons of non-smoking and self-

efficacy to quit smoking). The four clusters were: 1) Progressive smokers (balanced on pros

and cons of non-smoking but indicated higher overall pros and cons and self-efficacy than

other clusters, 2) Immotive smokers (low self-efficacy, high cons of non-smoking, and

moderate pros of non-smoking compared to other clusters), 3) Disengaged pessimistic

smokers (low pros and cons of nonsmoking and low self-efficacy), and 4) Disengaged

optimistic smokers (low pros and cons of non-smoking and high self-efficacy). These

findings validated the results of a study by Norman et al. (2000). However, the Schorr et al.

(2008) and the Norman et al. (2000) trials included only smokers who were participating in

smoking cessation trials. In fact, all prior studies on subtyping smokers who are not

motivated to quit have used a restricted set of variables guided solely by the Transtheoretical

Model and limited to the pros and cons of smoking (Bommele et al., 2015), temptations to

smoke (Anatchkova, Velicer, & Prochaska, 2006; Norman, et al., 2000), or self-efficacy

(Dijkstra & De Vries, 2000; Schorr, et al., 2008). Thus, other important determinants of

smoking could have been overlooked (Armitage, 2009, 2015; Borrelli, 2010). Other studies


on subtyping smokers conduct analyses on data that are more than 15 years old (Balmford,

Borland, & Burney, 2008), and smokers who are unmotivated to quit today may be

phenotypically different from those in past studies (e.g., following bans of smoking in public

places).

The present paper uses Latent Class Analysis (LCA) to examine whether distinct

subgroups of smokers who are not motivated to quit can be derived empirically, exploring

differential profiles based on a variety of demographic (Caine, Smith, Beasley, & Brown,

2012; Monden, de Graaf, & Kraaykamp, 2003; Schuck, Otten, Kleinjan, Bricker, & Engels,

2014), psychosocial (Borrelli, Hayes, Dunsiger, & Fava, 2010; Borrelli & Mermelstein, 1994;

Ho, Alnashri, Rohde, Murphy, & Doyle, 2015; Savoy et al., 2014; Schuck, et al., 2014;

Stockings et al., 2013) and smoking behavior (Broms, Silventoinen, Lahelma, Koskenvuo, &

Kaprio, 2004; Tzelepis et al., 2013; Zhou et al., 2009) variables that have been shown to be

associated with smoking. We defined smokers as “unmotivated to quit” if they were not

ready to quit within 30 days, in order to be consistent with systematic reviews (Asfar, Ebbert,

Klesges, & Relyea, 2011), nationwide surveys, and other studies on smokers who are not

motivated to quit (Carpenter, Alberg, Gray, & Saladin, 2010; Carpenter, Hughes, Solomon, &

Callas, 2004). We also chose this definition because the majority of clinical trials exclude

smokers who are not motivated to quit within 30 days, so there is a dearth of information on

this group of smokers.

We hypothesized that there are latent variables that define different subtypes of

smokers who are not ready to quit within 30 days; that smokers within subtypes are similar

with respect to their attitudes and behaviors regarding smoking; and that smokers between

subtypes are different on these variables. We also conducted model validation by examining

differences on variables extraneous to the LCA analysis. Specifically, we examined

differences between the subtypes in their estimated timeline for quitting smoking, as well as


differences in whether or not they ever see themselves quitting. Exploring the differential

profiles of subtypes of smokers could help to tailor intervention programs and deliver more

effective interventions to smokers who are not motivated to quit.

Materials and Methods

Study Participants

The sample was recruited as part of a larger cross-sectional study on smokers’

motivation to quit. In order to be eligible for the study, participants had to be 18 years or

older and current regular smokers (defined as smoking ≥3 tobacco cigarettes per day and at

least 100 cigarettes in their lifetime). For the current study, we analyzed participants (n =500)

who reported that they did not plan to quit smoking within 30 days (n = 250 from the UK, n =

250 from the US).

Procedure

Participants were recruited by Toluna, Inc to complete an online survey. Participants

first completed an online screener to determine study eligibility and if they were eligible, they

viewed an online consent form. Those who agreed with the terms of the consent form

indicated their consent by clicking an “agree” button, after which they were directed to the

survey. Of those who completed the screener and were eligible to participate, 42 did not

complete the questionnaire. Of those who completed the questionnaire, each respondent was

checked and verified as unique. Participants were removed from the data if they selected the

same response option uniformly (n =8), completed the survey in less than half the median

completion time (n =24), or responded randomly (n = 0). These quality criteria represent best

practice in the industry (World Association for Social, Opinion and Market Research & the

Global Research Business Network, 2015) and could indicate poor quality responses

(Malhotra, 2008). We compared non-completers (n=42 who completed the screener but not

the questionnaire) with those who completed the questionnaire and found no significant


differences in age (t (540) = .91, p=.36), cigarettes smoked per day (t (540) = -.36, p=.72) or

gender χ2 (1) = 1.43, p = .15).

Participants were given ‘panel points’ for completion, which can be exchanged for

vouchers, cash, or a lottery on the Toluna, Inc. website. Toluna, Inc abides by international

standards of data security and protection (e.g. ESOMAR and ISO27001) and only

anonymized data were transferred to the research team. Ethical approval for the study was

obtained from both the institutions involved in the study (The Miriam Hospital and The

University of Manchester).

Measures

Demographics: Gender, age (<40 years vs. <40 years), race/ethnicity (minority vs. other),

marital status (partnered vs. not), employment (full- or part-time) and number of children in

the home (>1 vs. none) were assessed.

Smoking Behavior. The number of cigarettes smoked per day was dichotomized into ‘>10

cigarettes per day’ vs. ‘< 10 cigarettes per day1.’ This cut-off has been commonly used to

discriminate light vs. heavy smokers (Ahluwalia et al., 2006; Salgado-Garcia, Cooper, &

Taylor, 2013). Nicotine dependence (Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991)

was assessed by whether or not participants smoked within 30 minutes of waking (yes/no)

because this single item is highly correlated with the full Fagerstrom Test for Nicotine

Dependence (Baker et al., 2007). Participants indicated whether they had ever tried to quit

smoking for 24 hours because they were trying to quit smoking (yes/no), whether they had a

partner or spouse who smoked (yes/no), and the number of smokers in the home besides

themselves (0 vs. 1 or more). Smoking-related health status was assessed by two items

assessing current or past illness caused or made worse by smoking (yes/no).

Psychosocial variables: Although participants were screened into the study because they had

‘no plans to quit within 30 days’, we wanted to assess the level of motivation to quit smoking


among our sample even though they are not behaviorally ready to quit within 30 days.

Motivation to quit was assessed with the question ‘Right now, how much do you want to quit

smoking?’ measured on a continuous 10-point scale ranging from “1” ‘do not want to quit’ to

“10” ‘very much want to quit.’ Confidence to quit was assessed with the question ‘Right

now, how confident are you that you can stop smoking?’ measured on a continuous 10-point

scale ranging from “1” ‘not at all confident’ to “10” ‘very confident to quit’ smoking.”

Motivation and confidence scales were each dichotomized as >5 vs. 5 or less. Because there

was no empirical guidance from the literature, these scales were dichotomized, a priori, at the

mid-point. Depressed mood was measured by the 10-item Center for Epidemiologic Studies

Depression Scale (CESD10). Analyses used the validated cut off of >10 to indicate

depression (Andresen, Malmgren, Carter, & Patrick, 1994).

Risk Perception of Smoking. Perceived vulnerability (degree to which the smoker feels

vulnerable to the health effects of smoking) was measured with three items assessing

perceived risk of heart disease, lung cancer, and other lung disease. Each of the three diseases

were assessed on a 7-point scale ranging from “1” ‘no chance’ to “7” ‘certain to happen’

dichotomized as having an average of >4 across the three items vs. an average of 4 or less.

(Borrelli, Hayes, et al., 2010). Precaution effectiveness (degree to which the smoker believes

that quitting will have health benefits) was measured with three items assessing the

anticipated benefits of quitting smoking on heart disease, lung cancer and other lung disease.

Each of the three diseases were assessed on a 5-point scale ranging from “1” ‘no decrease in

risk’ to “5” ‘complete elimination of risk’ dichotomized as having an average of >3 across

the three items vs. an average of 3 or less (Borrelli, Hayes, et al., 2010). Optimistic bias

(smoker’s belief that they are at less risk than other smokers) was assessed by three items

regarding whether their chances of having heart disease, lung cancer or other lung disease

was lower or higher than other smokers (Borrelli, Hayes, et al., 2010). Optimistic bias was


also assessed with the question ‘if you were to develop an illness from your smoking, do you

think it would be: “1” ‘less serious than the illnesses typically developed by smokers’, “2”

‘more serious’, or “3” ‘about the same as what happens to other smokers’ and dichotomized

as ‘as or more serious than other smokers’ vs. ‘less serious than other smokers’ (Borrelli,

Hayes, et al., 2010).

External validation variables: We did not include the following variables in the LCA analysis

so that we could assess whether the LCA subgroups had patterns that would provide some

support for class membership. Participants were asked whether or not they ever saw

themselves quitting (yes/no), and, of those who said ‘yes,’ they indicated their how soon they

planned to quit: ‘1- 6 months’, ‘7-11 months’, ‘1-5 years’, ‘6 or more years.’

Statistical Analysis

Latent class analysis (LCA) was used to explore the presence of distinct subgroups of

unmotivated smokers and characterize the classes. LCA identifies mutually exclusive discrete

latent variables derived from two or more observed indicators. LCA utilizes categorical data

in a mixture model discriminated by latent variables. LCA was performed in R (Version

3.1.0) using the poLCA package. The approach employed by poLCA uses the Expectation

Maximization (EM) and Newton-Raphson algorithms to estimate model parameters for

polytomous variables.

The LCA method requires specification of the number of classes prior to performing

analysis. Given that this was an exploratory analysis, we did not have a priori knowledge

of the number of latent classes and the corresponding ideal sample size. While there is no

formal approach to calculating sample sizes for LCA, Dziak and colleagues (2014) performed

simulations to identify sample sizes for detecting the appropriate number of classes. Using

these guidelines, for the range of classes considered (2 to 6 classes) with a moderate effect

size, a sample size of at least 450 participants is estimated to be appropriate. This further


aligns with sample sizes in simulated and LCA studies using primary data collection (e.g.

Rosa, Aloise-Young, & Henry; 2014 and Bommele, Kleinjan, Schoenmakers, Burk, van den

Eijnden, & van de Mheen; 2015).

Various criteria to identify the number of classes in LCA models have been

recommended, such as the Bayesian Information Criterion (BIC; Schwartz, 1978) and

Akaike’s Information Criterion (AIC; Akaike, 1987). Studies have shown that the BIC is

strongest indicator of number of classes (Collins, Fidler, Wugalter, & Long, 1993; Hagenaars

& McCutcheon, 2002; Kass & Raftery, 1995; Keribin, 1998; Nylund, Asparouhov, &

Muthen, 2007), thus the BIC was the primary criterion for deciding on the number of classes

in the current study, with AIC considered secondarily (Nylund, et al., 2007). The BIC was

examined across 1000 iterations of LCA model-fitting for each of k=2 to k=7 number of

classes, to ensure a global minimum was found. A smaller BIC suggests a better fit (Kass &

Raftery, 1995). The AIC was considered in the same manner. Variables that were

hypothesized to be important for discriminating subtypes of smokers were identified a priori.

Model covariates were ethnicity, gender, country and age.

After fitting 1,000 iterations of k=2 to k=7 classes, the average of each model fit

criteria was taken across the iterations and the number of classes was chosen primarily based

on the smallest mean BIC across iterations. The final model was chosen within the

appropriate number of classes as the model with the smallest BIC in the 1,000 iterations. The

variables used in identifying the classes were then utilized to characterize each of the classes,

conditional on the covariates. Specifically, the class-conditional response probabilities for

each variable among the individuals in each class were calculated. These could then be used

to draw conclusions on the underlying classes of smokers.

As a validation of the LCA classification, we compared the classes on the two

external measures which were excluded from model fitting (whether or not they ever saw


themselves quitting and when they planned to quit). We tested the association between the

LCA classes and whether or not an individual ever planned to quit smoking using logistic

regression. We also tested the association between the classes and how soon an individual

planned to quit smoking (among those who did not say they ‘never’ plan on quitting

smoking). We used ordinal regression with cumulative probabilities (non-parallel cumulative

logit) in order to preserve the ordering between unequally spaced ‘time to quit’ categories.

Results

Class Membership

The BIC reached its minimum, on average, in the 3-class model (Table 1), and was

therefore chosen as the appropriate number of classes. The four class model had the second

smallest average BIC; however, the entropies were similar so the 3-class model was retained.

Demographic characteristics of the sample (Table 2, 1 st column)

Participants were 50.3 years old on average (SD=14.1), 42.8% male, 50% residing in

the US, 90.0% were White, 72.6% had at least a high-school education, 49.0% were in

employed and 59.8% were partnered. Participants smoked an average of 17.3 cigarettes/day

(SD = 12.6) and 75.0% smoked their first cigarette within 30 minutes of waking.

Observed predictor characteristics of the model classes

Table 3 presents the estimated class-conditional response probabilities and tests for

proportion between classes. Based on the pattern of item endorsement, the classes were

labelled as follows: Class 1 (n=142): Unconvinced Smokers (UCS); Class 2 (n=192):

Smokers with Psychosocial Barriers (SPB); Class 3 (n=166) Health Concerned Smokers

(HCS).

Compared with each of the other two classes, HCS had a significantly greater

proportion of people with past smoking-related illnesses, high perceived risk for developing


smoking-related illnesses, and, although the entire sample was not behaviorally ready to quit

within 30 days, HCS had higher motivation to quit smoking. Compared with UCS, HCS had

a significantly greater proportion of those who reported current smoking-related illness and a

significantly lower proportion of optimistically biased smokers. HCS and SPB were similar

in that they had twice the proportion of smokers who tried to quit in the past year and twice

the proportion of depressed smokers vs. UCS.

SPB had a high proportion of partners who smoked (89.4%), universally reported that

there was another smoker in the home besides themselves, and were more likely to have

children in the home vs the other two classes. There were several variables on which the

proportion of SPB fell in between HCS and UCS (past smoking related illness, perceived

vulnerability, optimistic bias, and motivation to quit) resulting in a profile that was more

favorable towards quitting than UCS and less favorable towards quitting than HCS.

Approximately half of SPB (51.5%) and HCS (48.6%) reported CESD10 scores that were

consistent with depression, vs. 20.8% in UCS.

Compared with each of the other two classes, UCS had a significantly higher

proportion of smokers with lower motivation and confidence to quit smoking, as well as

smokers who never made a quit attempt. UCS also had a significantly lower proportion of

smokers who: tried to quit in the past year, reported past or current smoking-related illness,

had children in the home, believed they were at risk for smoking-related illnesses and

believed that the risks of smoking could be attenuated by quitting. UCS also had significantly

greater proportion who believed that they were at less risk of developing a smoking-related

illness than the typical smoker vs. SPB and HCS. UCS had the lowest proportion of those

who were dependent upon nicotine and the lowest proportion of those who were depressed.

UCS had a greater proportion of men, smokers who were over 40 years of age, and those who

were unemployed (Table 2).


Class Validation Analyses

As a validation of the classification, we compared the three classes on the two

external measures that were not included in the LCA model. We hypothesized that the odds

of smokers who ‘never see themselves quitting’ would be highest among UCS, followed by

SPB and HCS smokers, respectively. 184 participants reported that they ‘never see

themselves quitting:’ 60.6% of UCS never saw themselves quitting compared with 31.8% of

smokers in SPB and 22.3% of smokers in HCS. We fit a logistic regression model with UCS

as the reference class and never seeing oneself quit as the outcome. As hypothesized, the

odds of SPB (OR=0.30, Z=-5.16, p=<0.001) and HCS (OR=0.19, Z=-6.62, p=<0.001) never

seeing themselves quitting smoking were significantly lower than UCS. HCS had lower odds

of never seeing themselves quit smoking than SPB (OR=0.62, Z=-2.00, p=0.05).

We further hypothesized that, among smokers who see themselves quitting smoking

at some point in the future (HCS n=129; SPB n=131; UCS n=56), UCS would have greater

odds of planning to quit at a later time point than HCS and SPB. We fit a cumulative logits

model to compare the cumulative change in odds between the four categories of time

planning to quit, with UCS as the reference class. As hypothesized, the estimated odds of

planning to quit at later time periods versus sooner time periods were lower among SPB and

HCS versus UCS. The odds of an individual planning on quitting smoking in 6 or more years

versus 5 years or less was 0.47 times lower for SPB (Z=-1.30, p=0.19) and 0.48 times lower

for HCS (Z=-1.27, p=0.20) versus those in the UCS class. The odds of planning to quit in 1

year or more versus less than 1 year was significantly lower in both SPB (OR=.52, Z=-2.04,

p=0.04) and HCS (OR=.28, Z=-3.72, p<0.001) versus UCS. The odds of planning to quit in a

time period > 6 months versus within the next 6 months was significantly lower for HCS

(OR=0.51, Z=-2.05, p=0.04) compared to UCS. Comparing HCS and SPB, the odds of

planning to quit at 1 year or longer versus less than 1 year was 0.55 times lower for HCS as


compared to SPB (Z=-2.18, p=0.03). Other model comparisons between HCS and SPB were

not statistically significant.

Discussion

The present paper uses LCA to examine whether distinct subgroups of smokers who

are not motivated to quit can be empirically derived. We explored differential profiles based

on a variety of psychosocial and smoking behavior variables previously shown to be

associated with quitting smoking. Prior research on subtyping smokers has examined only a

limited range of variables associated with the Transtheoretical Model (Anatchkova, et al.,

2006; Bommele, et al., 2015; Dijkstra & De Vries, 2000; Norman, et al., 2000), smokers

participating in intervention studies (Norman, et al., 2000; Schorr, et al., 2008; Smit, Hoving,

& de Vries, 2010), and college students (Rosa, Aloise-Young, & Henry, 2014). These studies

found four (Anatchkova, et al., 2006; Schorr, et al., 2008; Norman, et al., 2000), five

(Dijkstra & De Vries, 2000), or six (Bommele, et al., 2015) clusters of smokers, but only a

few variables from one theoretical model were examined as discriminating factors. We

hypothesized that because smoking behavior is so complex (particularly among those who are

not ready to quit), a broader constellation of variables would be needed to adequately capture

the phenomena of interest.

We found evidence to support three distinct subtypes of unmotivated smokers, with

each class having defining features that were significantly different from each of the other

two classes. HCS had a profile that was most favorable towards quitting smoking (previous

and current smoking-related illness, smoking-related risk perceptions, and motivation to

quit), whereas UCS had a profile that was least favorable towards quitting smoking (low

motivation and confidence to quit, low perception of smoking-related risks, optimistically

biased, few anticipated benefits of quitting, few quit attempts). HCS had twice as many


smokers with lifetime and past year quit attempts vs. UCS. These patterns were validated by

our findings that UCS were more likely to never see themselves quitting vs. HCS, and that, of

those who plan on quitting at some point, UCS planned to quit at later time points than HCS.

These patterns indicate that HCS may be the ‘low hanging fruit’ and perhaps

responsive to proactive, low cost interventions. Because a high proportion of HCS have

medical comorbidities due to smoking, identification of these smokers through electronic

health records and inviting them to use a smoking cessation service or quit line (Fu et al.,

2014; Haas et al., 2015) may be particularly effective for HCS. Other proactive options such

as integrating smoking cessation through existing medical infrastructures such as home health

care (Borrelli et al., 2005), dental settings (Omana-Cepeda, Jane-Salas, Estrugo-Devesa,

Chimenos-Kustner, & Lopez-Lopez, 2016), hospitalization (Rigotti et al., 2014) or primary

care (Naughton et al., 2014) might also be effective. Presenting biomarker feedback on the

effects of smoking on their health could be ‘preaching to the choir,’ because our analyses

showed that a substantial proportion of HCS were already convinced of the health effects of

smoking. Given that half of HCS had CESD scores consistent with depression, it would be

important to screen for mood changes before and after quit attempts, as well as to provide

mood management strategies prophylactically.

UCS might not achieve benefit from interventions focused on building risk perception

regarding the health effects of smoking, given their low risk perception, high optimistic bias,

and lack of experience with smoking-related illness. However, it might be possible to

‘trigger’ quit attempts by offering external cues to quit (Nicotine Replacement Therapy,

Quitlines) and ‘practice quit attempts’ rather than waiting for internal cues, such as sufficient

motivation (Carpenter et al., 2011; Jardin et al., 2014). Smoking reduction may be more

attractive to UCS than strict cessation (Shiffman et al., 2007), and may help build confidence

to quit and increase perceived anticipated benefits of quitting, both of which were low among


UCS. However, the effect of reduction on future cessation rates is unclear (Hughes &

Carpenter, 2006; Schauer, Malarcher, & Babb, 2015). Another possibility is to design

interventions that target factors that hinder smoking cessation, such as employment,

education and housing (Haas, et al., 2015). This may be a particularly effective approach for

UCS, given that more than half were unemployed. Finally, because UCS do not have

experience with smoking-related illness, they are less likely to come into contact with the

medical system. Therefore, it may be important to leverage mobile health approaches for

motivating quitting with this population. Smokers unmotivated to quit have a high prevalence

of mobile (91.6%) and smartphone ownership (71%), as well as frequent use of Facebook,

texting, and visiting health-related websites ( Borrelli, et al., 2015).

Overall, SPB had a more favorable profile towards quitting than UCS and less

favorable profile than HCS. SPB was defined by having a high proportion of smokers in the

home, partners who smoked, and children. SPB were also significantly different from the

other two classes on motivation to quit, perceived vulnerability, optimistic bias and smoking

related illness. SPB also had the highest percentage of smokers with lifetime and past year

quit attempts. Biomarker feedback regarding how smoking is affecting themselves and others

in the home may be effective for SPB, given that a very high proportion had families and that

they fell in the middle of the other two groups in terms of perceived risk and optimistic bias.

Previous cessation induction studies have shown that a combination of biomarker feedback

and motivational counseling is effective for smoking cessation (Borrelli, McQuaid, Novak,

Hammond, & Becker, 2010; Borrelli et al., 2016; Emmons et al., 2000). Because of the high

rate of unsuccessful quit attempts, this group may benefit from longer term interventions to

promote maintenance of abstinence. In addition, validation analyses also supported findings

that SPB had a more favorable profile towards quitting than UCS and less favorable profile

than HCS. UCS was more likely to endorse that they ‘never see themselves quitting’ versus


SPB, but SPB was more likely to endorse this vs. HCS. Similarly, SPB was more likely to

want to quit sooner verses UCS, but less likely to want to quit sooner than HCS.

The strengths of our study include the large sample and assessment of a wide range of

psychosocial and smoking behavior variables. Because our sample was comprised of

predominately white smokers, the results may not be generalizable to ethnic minority

smokers. Also, our sample may have over-represented females and older, more dependent

smokers. The data were collected online so the results may not be generalizable to those

without internet access. However, 82% of adults in the UK and 84% of adults in the US use

the internet (Pew Research Centre, 2015; Office for National Statistics, 2016). The rate of

adoption amongst older adults, those with lower education attainment, and black and minority

ethnic groups has accelerated in the last 15 years and although a gap persists, it has narrowed

substantially (Pew Research Centre, 2015). Although steps were taken to minimize unreliable

data reporting, as with any survey, there is no guarantee of validity of responses.

Validation analyses should be viewed as preliminary, due to the cross-sectional design

and small sample size. We considered dividing the dataset into an exploratory and

confirmatory datasets, but given that our sample is already stratified between two populations

(US and UK), this would result in subsamples with sizes less than that recommended for

performing LCA analyses. Subsamples of the overall sample of 500 further resulted in

sparsity of the predictor used in the identification of groups, so we proceeded to validate on

variables external to the LCA analyses, as has been done in prior research. In addition, while

LCA considers all possible patterns of observed variables, the data quickly become sparse as

the number of included variables increases. The exclusion of relevant, unknown variables

may have affected the model. Another limitation of our study is that, while the majority of

our variables were either naturally dichotomous or were continuous scales with validated cut-

points, some of our continuous variables were transformed into binary variables based on the


median or a standard clinical threshold (e.g., the number of cigarettes smoked per day was

dichotomized into >10 cigarettes per day vs. <10 cigarettes per day because this has been

commonly used to discriminate light vs. heavy smokers (Ahluwalia et al., 2006; Salgado-

Garcia et al., 2013). Although software exists to analyze latent groups with solely

dichotomous (LCA) or solely continuous variables (latent profile analysis), at the present

time, software that uses a mixture of dichotomous and continuous variables does not exist for

exploratory, interpretable latent group analysis.

Conclusions

Because there is a large population of smokers who are not motivated to quit

smoking, it is important to determine if this population is homogeneous in which ‘one size

fits all’ approach to intervention would be effective to motivate quitting. The results of the

current study suggest that smokers who are unmotivated to quit are a heterogeneous

population, characterized by different profiles and features. This has implications for

interventions that could be tailored for these different subtypes of smokers, resulting greater

efficacy and potential cost-effectiveness. With future research, a screening tool could be

developed with the goal of administration at the point of care to identify the type of

unmotivated smoker and assist with how to approach the smoker and plan treatment. Even

being about to discriminate the Unconvinced Smokers (UCS), from those who would be

amenable to treatment (HCS and SPB) would be a step forward. Future directions include

testing the observed classes in other samples of vulnerable smokers (e.g., smokers with

severe mental illness, comorbid substance abuse) and testing the model with a longitudinal

design.


REFERENCES

Ahluwalia, J. S., Okuyemi, K., Nollen, N., Choi, W. S., Kaur, H., Pulvers, K., & Mayo, M. S.

(2006). The effects of nicotine gum and counseling among African American light

smokers: a 2 x 2 factorial design. Addiction, 101(6), 883-891. doi: 10.1111/j.1360-

0443.2006.01461.x

Akaike, H. (1987). Factor analysis and AIC. Psychometrika, 52(3), 317-332.

Anatchkova, M. D., Velicer, W. F., & Prochaska, J. O. (2006). Replication of subtypes for

smoking cessation within the precontemplation stage of change. Addict Behav, 31(7),

1101-1115. doi: 10.1016/j.addbeh.2005.08.002

Andresen, E. M., Malmgren, J. A., Carter, W. B., & Patrick, D. L. (1994). Screening for

depression in well older adults: evaluation of a short form of the CES-D (Center for

Epidemiologic Studies Depression Scale). Am J Prev Med, 10(2), 77-84.

Armitage, C. J. (2009). Is there utility in the transtheoretical model? Br J Health Psychol,

14(Pt 2), 195-210. doi: 10.1348/135910708X368991

Armitage, C. J. (2015). Changing behaviour, slow and fast: commentary on Peters, de Bruin

and Crutzen. Health Psychol Rev, 9(1), 30-33. doi: 10.1080/17437199.2014.912537

Armitage, C. J., & Arden, M. A. (2008). How useful are the stages of change for targeting

interventions? Randomized test of a brief intervention to reduce smoking. Health

Psychol, 27(6), 789-798. doi: 10.1037/0278-6133.27.6.789

Asfar, T., Ebbert, J. O., Klesges, R. C., & Relyea, G. E. (2011). Do smoking reduction

interventions promote cessation in smokers not ready to quit? Addict Behav, 36(7),

764-768. doi: 10.1016/j.addbeh.2011.02.003

Baker, T. B., Piper, M. E., McCarthy, D. E., Bolt, D. M., Smith, S. S., Kim, S. Y., . . .

Transdisciplinary Tobacco Use Research Center (2007). Time to first cigarette in the

morning as an index of ability to quit smoking: implications for nicotine dependence.

Nicotine Tob Res, 9 Suppl 4, S555-570. doi: 10.1080/14622200701673480

Balmford, J., Borland, R., & Burney, S. (2008). Exploring discontinuity in prediction of

smoking cessation within the precontemplation stage of change. Int J Behav Med,

15(2), 133-140. doi: 10.1080/10705500801929759

Bartlett, Y. K., Borrelli, B., Armitage, C. J., Tooley, E., & Wearden, A. (2015). Approaches

used to quit smoking: Differences between smokers who are motivated to quitand

those who are not motivated to quit. Paper presented at the The British Psychological

Society Division of Health Psychology Annual Conference, London, England.


Bommele, J., Kleinjan, M., Schoenmakers, T. M., Burk, W. J., van den Eijnden, R., & van de

Mheen, D. (2015). Identifying Subgroups among Hardcore Smokers: a Latent Profile

Approach. PLoS One, 10(7), e0133570. doi: 10.1371/journal.pone.0133570

Borrelli, B. (2010). Smoking Cessation: Next Steps for Special Populations Research and

Innovative Treatments. Journal of Consulting and Clinical Psychology, 78(1), 1-12.

doi: 10.1037/a0018327

Borrelli, B., Bartlett, Y. K., Tooley, E., Armitage, C. J., & Wearden, A. (2015). Prevalence

and Frequency of mHealth and eHealth Use Among US and UK Smokers and

Differences by Motivation to Quit. J Med Internet Res, 17(7), e164 PMC4526978.

doi: 10.2196/jmir.4420

Borrelli, B., Hayes, R. B., Dunsiger, S., & Fava, J. L. (2010). Risk perception and smoking

behavior in medically ill smokers: a prospective study. Addiction, 105(6), 1100-1108.

doi: 10.1111/j.1360-0443.2010.02900.x

Borrelli, B., McQuaid, E. L., Novak, S. P., Hammond, S. K., & Becker, B. (2010).

Motivating Latino caregivers of children with asthma to quit smoking: a randomized

trial. Journal of Consulting and Clinical Psychology, 78(1), 34-43. doi:

10.1037/a0016932

Borrelli, B., McQuaid, E. L., Tooley, E. M., Busch, A. M., Hammond, S. K., Becker, B., &

Dunsiger, S. (2016). Motivating parents of kids with asthma to quit smoking: the

effect of the teachable moment and increasing intervention intensity using a

longitudinal randomized trial design. Addiction, 111. doi: 10.1111/add.13389

Borrelli, B., & Mermelstein, R. (1994). Goal setting and behavior change in a smoking

cessation program. Cognitive Therapy and Research, 18(1), 69-83.

Borrelli, B., Novak, S., Hecht, J., Emmons, K., Papandonatos, G., & Abrams, D. (2005).

Home health care nurses as a new channel for smoking cessation treatment: outcomes

from project CARES (Community-nurse Assisted Research and Education on

Smoking). Prev Med, 41(5-6), 815-821. doi: S0091-7435(05)00120-9

Broms, U., Silventoinen, K., Lahelma, E., Koskenvuo, M., & Kaprio, J. (2004). Smoking

cessation by socioeconomic status and marital status: the contribution of smoking

behavior and family background. Nicotine Tob Res, 6(3), 447-455. doi:

10.1080/14622200410001696637


Caine, V. A., Smith, M., Beasley, Y., & Brown, H. L. (2012). The impact of prenatal

education on behavioral changes toward breast feeding and smoking cessation in a

healthy start population. J Natl Med Assoc, 104(5-6), 258-264.

Carpenter, M. J., Alberg, A. J., Gray, K. M., & Saladin, M. E. (2010). Motivating the

unmotivated for health behavior change: a randomized trial of cessation induction for

smokers. Clin Trials, 7(2), 157-166. doi: 10.1177/1740774510361533

Carpenter, M. J., Hughes, J. R., Gray, K. M., Wahlquist, A. E., Saladin, M. E., & Alberg, A.

J. (2011). Nicotine therapy sampling to induce quit attempts among smokers

unmotivated to quit: a randomized clinical trial. Arch Intern Med, 171(21), 1901-

1907. doi: 10.1001/archinternmed.2011.492

Carpenter, M. J., Hughes, J. R., Solomon, L. J., & Callas, P. W. (2004). Both smoking

reduction with nicotine replacement therapy and motivational advice increase future

cessation among smokers unmotivated to quit. J Consult Clin Psychol, 72(3), 371-

381. doi: 10.1037/0022-006X.72.3.371

Centers for Disease Control and Prevention. (2011). Quitting smoking among adults - United

States, 2001-2010. Morbidity and Mortality Weekly Report, 60(44), 1513-1522.

Centers for Disease Control and Prevention. (2015). Early release of selected estimates based

on data from the national health interview survey, January-March 2015.

Christiansen, B. A., Reeder, K. M., TerBeek, E. G., Fiore, M. C., & Baker, T. B. (2015).

Motivating Low Socioeconomic Status Smokers to Accept Evidence-Based Smoking

Cessation Treatment: A Brief Intervention for the Community Agency Setting.

Nicotine Tob Res, 17(8), 1002-1011. doi: 10.1093/ntr/ntu345

Collins, L. M., Fidler, P. L., Wugalter, S. E., & Long, J. D. (1993). Goodness-of-Fit Testing

for Latent Class Models. Multivariate Behav Res, 28(3), 375-389. doi:

10.1207/s15327906mbr2803_4

Dijkstra, A., & De Vries, H. (2000). Clusters of precontemplating smokers defined by the

perception of the pros, cons, and self-efficacy. Addict Behav, 25(3), 373-385.

Dziak, J. J., Lanza, S. T., & Tan, X. (2014). Effect size, statistical power, and sample size

requirements for the bootstrap likelihood ratio test in latent class analysis. Structural

equation modeling: a multidisciplinary journal, 21(4), 534-552.


Emmons, K. M., Sorensen, G., Klar, N., Digianni, L., Barclay, G., Schmidt, K., & Hammond,

S. K. (2000). Healthy baby second-hand smoke study: project brief. Tob Control, 9

Suppl 3, III58-60.

Fu, S. S., van Ryn, M., Sherman, S. E., Burgess, D. J., Noorbaloochi, S., Clothier, B., . . .

Joseph, A. M. (2014). Proactive tobacco treatment and population-level cessation: a

pragmatic randomized clinical trial. JAMA Intern Med, 174(5), 671-677.

Haas, J. S., Linder, J. A., Park, E. R., Gonzalez, I., Rigotti, N. A., Klinger, E. V., . . .

Williams, D. R. (2015). Proactive tobacco cessation outreach to smokers of low

socioeconomic status: a randomized clinical trial. JAMA Intern Med, 175(2), 218-226.

Hagenaars, J. A., & McCutcheon, A. L. (2002). Applied Latent Class Analysis: Cambridge

University Press.

Heatherton, T. F., Kozlowski, L. T., Frecker, R. C., & Fagerstrom, K. O. (1991). The

Fagerstrom Test for Nicotine Dependence: a revision of the Fagerstrom Tolerance

Questionnaire. Br J Addict, 86(9), 1119-1127.

Ho, S. Y., Alnashri, N., Rohde, D., Murphy, P., & Doyle, F. (2015). Systematic review and

meta-analysis of the impact of depression on subsequent smoking cessation in patients

with chronic respiratory conditions. Gen Hosp Psychiatry, 37(5), 399-407. doi:

10.1016/j.genhosppsych.2015.05.002

Hughes, J. R., & Carpenter, M. J. (2006). Does smoking reduction increase future cessation

and decrease disease risk? A qualitative review. Nicotine Tob Res, 8(6), 739-749. doi:

10.1080/14622200600789726

Jardin, B. F., Cropsey, K. L., Wahlquist, A. E., Gray, K. M., Silvestri, G. A., Cummings, K.

M., & Carpenter, M. J. (2014). Evaluating the effect of access to free medication to

quit smoking: a clinical trial testing the role of motivation. Nicotine Tob Res, 16(7),

992-999. doi: 10.1093/ntr/ntu025

Kass, R. E., & Raftery, A. E. (1995). Bayes Factors. Journal of the American Statistical

Association, 90(No. 430), 773-795. doi: 10.2307/2291091

Keribin, C. (1998). Consistent estimate of the order of mixture models. Comptes Rendues de

l'Academie des Sciences, Serie l-Mathematiques, 326, 243-248.

Malhotra, N. (2008). Completion time and response order effects in web surveys. Public

Opin Q, 72 (5), 914-934.


Monden, C. W., de Graaf, N. D., & Kraaykamp, G. (2003). How important are parents and

partners for smoking cessation in adulthood? An event history analysis. Prev Med,

36(2), 197-203.

Naughton, F., Jamison, J., Boase, S., Sloan, M., Gilbert, H., Prevost, A. T., . . . Sutton, S.

(2014). Randomized controlled trial to assess the short-term effectiveness of tailored

web- and text-based facilitation of smoking cessation in primary care (iQuit in

Practice). Addiction, 109(7), 1184-1193. doi: 10.1111/add.12556

Norman, G. J., Velicer, W. F., Fava, J. L., & Prochaska, J. O. (2000). Cluster subtypes within

stage of change in a representative sample of smokers. Addict Behav, 25(2), 183-204.

Nylund, K. L., Asparouhov, T., & Muthen, B. O. (2007). Deciding on the number of classes

in latent class analysis and growth mixture modeling: a Monte Carlo simulation study.

Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535-569.

Office for National Statistics. (2009). Smoking related behaviour and attitudes, 2008/09.

http://webarchive.nationalarchives.gov.uk/20160105160709/http://www.ons.gov.uk/

ons/rel/lifestyles/smoking-related-behaviour-and-attitudes/2008-09/index.html

Retrieved from

http://webarchive.nationalarchives.gov.uk/20160105160709/http://www.ons.gov.uk/

ons/rel/lifestyles/smoking-related-behaviour-and-attitudes/2008-09/index.html

Office for National Statistics. (2016). Internet access -households and individuals: 201.

Retrieved from:

https://www.ons.gov.uk/peoplepopulationandcommunity/householdcharacteristics/

homeinternetandsocialmediausage/bulletins/internetaccesshouseholdsandindividuals/

2016

Omana-Cepeda, C., Jane-Salas, E., Estrugo-Devesa, A., Chimenos-Kustner, E., & Lopez-

Lopez, J. (2016). Effectiveness of dentist's intervention in smoking cessation: A

review. J Clin Exp Dent, 8(1), e78-83. doi: 10.4317/jced.52693

Perrin, A., & Duggan, M. (2015). Pew Research Centre: Americans’ Internet Access 2000-

2015. Retrieved from: http://www.pewinternet.org/2015/06/26/americans-internet-

access-2000-2015/

Rigotti, N. A., Regan, S., Levy, D. E., Japuntich, S., Chang, Y., Park, E. R., . . . Singer, D. E.

(2014). Sustained care intervention and postdischarge smoking cessation among

hospitalized adults: a randomized clinical trial. JAMA, 312(7), 719-728. doi:

10.1001/jama.2014.9237

http://www.pewinternet.org/2015/06/26/americans-internet-access-2000-2015/

http://www.pewinternet.org/2015/06/26/americans-internet-access-2000-2015/


Rosa, J. R., Aloise-Young, P. A., & Henry, K. L. (2014). Using motives for smoking to

distinguish between different college student smoker typologies. Psychol Addict

Behav, 28(4), 1297-1304.

Salgado-Garcia, F. I., Cooper, T. V., & Taylor, T. (2013). Craving effect of smoking cues in

smoking and antismoking stimuli in light smokers. Addict Behav, 38(10), 2492-2499.

doi: 10.1016/j.addbeh.2013.04.008

Savoy, E., Reitzel, L. R., Scheuermann, T. S., Agarwal, M., Mathur, C., Choi, W. S., &

Ahluwalia, J. S. (2014). Risk perception and intention to quit among a tri-ethnic

sample of nondaily, light daily, and moderate/heavy daily smokers. Addict Behav,

39(10), 1398-1403. doi: 10.1016/j.addbeh.2014.05.002

Schauer, G. L., Malarcher, A. M., & Babb, S. D. (2015). Gradual reduction of cigarette

consumption as a cessation strategy: prevalence, correlates, and relationship with

quitting. Nicotine Tob Res, 17(5), 530-538. doi: 10.1093/ntr/ntu172

Schorr, G., Ulbricht, S., Schmidt, C. O., Baumeister, S. E., Ruge, J., Schumann, A., . . .

Meyer, C. (2008). Does precontemplation represent a homogeneous stage category? A

latent class analysis on German smokers. J Consult Clin Psychol, 76(5), 840-851. doi:

10.1037/a0013037

Schuck, K., Otten, R., Kleinjan, M., Bricker, J. B., & Engels, R. C. (2014). Predictors of

cessation treatment outcome and treatment moderators among smoking parents

receiving quitline counselling or self-help material. Prev Med, 69, 126-131. doi:

10.1016/j.ypmed.2014.09.014

Schwartz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6, 461-

464.

Shiffman, S., Hughes, J. R., Ferguson, S. G., Pillitteri, J. L., Gitchell, J. G., & Burton, S. L.

(2007). Smokers' interest in using nicotine replacement to aid smoking reduction.

Nicotine Tob Res, 9(11), 1177-1182. doi: 10.1080/14622200701648441

Smit, E. S., Hoving, C., & de Vries, H. (2010). Does a typical contemplator exist? Three

clusters of smokers in contemplation. Health Educ Res, 25(1), 61-73. doi:

10.1093/her/cyp059

Stockings, E., Bowman, J., McElwaine, K., Baker, A., Terry, M., Clancy, R., . . . Wiggers, J.

(2013). Readiness to quit smoking and quit attempts among Australian mental health

inpatients. Nicotine Tob Res, 15(5), 942-949. doi: 10.1093/ntr/nts206


Tzelepis, F., Paul, C. L., Walsh, R. A., Wiggers, J., Duncan, S. L., & Knight, J. (2013).

Predictors of abstinence among smokers recruited actively to quitline support.

Addiction, 108(1), 181-185. doi: 10.1111/j.1360-0443.2012.03998.x

World Association for Social, Opinion and Market Research and the Global Research

Business Network (2015) ESOMAR/ GRBN Guideline for online sample quality.

Retrieved from: https://www.esomar.org/knowledge-and-standards/codes-and-

guidelines/guideline-on-online-sample-quality.php

Zhou, X., Nonnemaker, J., Sherrill, B., Gilsenan, A. W., Coste, F., & West, R. (2009).

Attempts to quit smoking and relapse: factors associated with success or failure from

the ATTEMPT cohort study. Addict Behav, 34(4), 365-373. doi:

10.1016/j.addbeh.2008.11.013

Footnotes

1LCA requires that dichotomous variables be used in analyses. This allows for

interpretable classification, ensures equivalence of measure, and creates qualitatively

indicative categories from discrete scales that have the potential for heteroscedasticity.

Variables were dichotomized primarily based on scientific motivation and secondarily on the

mean of the bounded scale or observed values (see Tables 2 and 3).

https://www.esomar.org/knowledge-and-standards/codes-and-guidelines/guideline-on-online-sample-quality.php

https://www.esomar.org/knowledge-and-standards/codes-and-guidelines/guideline-on-online-sample-quality.php


Table 1. Latent Class Model Fit Statistics (N=500)

Fit Indices 2-Class Model

3-Class Model

4-Class Model

5-Class Model

6-Class Model

7-Class Model

BIC 9395.745 9237.683 9305.711 9872.371 10148.96 10279.32

AIC 9239.805 8993.236 8972.757 9450.91 9638.996 9680.841

Chi Square 691637.6 227162.3 186743.7 140017.4 124055.7 122348.7

Parameters 37 58 79 100 121 142

Log-likelihood

-4582.9 -4438.62 -4407.38 -4625.46 -4698.5 -4698.42

Entropy 9.217 8.932 8.9 9.378 9.533 9.544


Table 2. Covariate and demographic characteristic probabilities and test of proportions.

Total Sample

% (n)

UCS1 Smokers

% (n)

SPB2 Smokers

% (n)

HCS3 Smokers

% (n)

Overall Two-sided Test P-Value (2)

UCS vs. SPB

P-Value4

HCS vs. UCS P-Value4

SPB vs. HCS

P-Value4

% male 42.8 (214) 54.2 (77) 40.1 (77) 036.1 (60) 0.004 (11.145) 0.007a 0.001b 0.255

% residing in US 50.0 (250) 45.8 (65) 50.5 (97) 53.0 (88) 0.441 (1.637) 0.227 0.125 0.358

% age 40 years 75.8 (379) 83.8 (119) 64.1 (123) 82.5 (137) <0.001 (23.477) <0.001c 0.442 <0.001d

% Caucasian/white 90 (450) 90.8 (129) 88.0 (169) 91.6 (152) 0.496 (1.401) 0.260 0.492 0.178

% high school educated or beyond 72.6 (363) 72.5 (103) 68.8 (132) 77.1 (128) 0.209 (3.127) 0.265 0.214 0.049e

% full- or part-time employed 49.0 (245) 42.3 (60) 49.5 (95) 54.2 (90) 0.110 (4.411) 0.115 0.024f 0.215

% married, engaged or living with partner 59.8 (299) 45.8 (65) 87.0 (167) 40.4 (67) <0.001 (96.711) <0.001g 0.200 <0.001h

Notes: 1Unconvinced Smokers, 2Smokers with Psychosocial Barriers, 3Health Concerned Smokers 4 P-values for when testing that the larger of the two proportions is statistically significant

Pearson’s 2 test statistics for p-values significant at the 0.05 level are: a6.00, b9.41, c14.96, d14.36, e2.72, f3.92, g63.41, h83.41.


Table 3. Response probabilities and tests of proportions of LCA predictors.

UCS1 Smokers

% (SD)

SPB2 Smokers

% (SD)

HCS3 Smokers

% (SD)


UCS vs. SPB

P-Value4HCS vs. UCS P-Value4

SPB vs. HCS

P-Value4

Demographics

% employed full or part time 43.1 (0.05) 50.0 (0.041) 52.8 (0.046) 0.219 (3.041) 0.1262 0.056 0.336

% with no children in household 80.4 (0.042) 49.5 (0.039) 63.6 (0.044) <0.0001 (33.468) <0.0001a 0.0008b 0.005c

Smoking behavior

% smoking >10 cigarettes/day 75.3 (0.046) 81.7 (0.033) 82.0 (0.036) 0.257 (2.716) 0.101 0.097 0.5

% smoking within 30 mins of waking 61.9 (0.052) 79.0 (0.034) 81.3 (0.035) 0.0001 (18.013) 0.0005d 0.0001e 0.338

% never quit smoking for 24 hrs 55.3 (0.05) 35.3 (0.039) 28.5 (0.042) <0.0001 (24.67) 0.0002f <0.0001g 0.103

% with partner/spouse smoker 0.0 (0) 89.4 (0.047) 0.0 (0) <0.0001 (419.467) <0.0001h 0.5 <0.0001i

% with 1 or more other smokers in household 11.8 (0.034) 100.0 (0) 15.0 (0.049) <0.0001 (355.452) <0.0001j 0.260 <0.0001k

Continued on next page


UCS1 Smokers

% (SD)

SPB2 Smokers

% (SD)

HCS3 Smokers

% (SD)


UCS vs. SPB


SPB vs. HCS

P-Value4

Health status

% with past illness caused or made worse by smoking 2.5 (0.023) 31.0 (0.047) 50.6 (0.053) <0.0001 (85.734) <0.0001l <0.0001m 0.0001n

% with current illness caused or made worse by smoking 1.5 (0.017) 22.1 (0.042) 29.9 (0.044) <0.0001 (42.618) <0.0001o <0.0001p

0.0590

Motivation and confidence to quit

% motivated to quit 16.4 (0.041) 46.1 (0.046) 68.1 (0.047) <0.0001 (82.589) <0.0001q <0.0001r <0.0001s

% confident to quit 17.9 (0.041) 28.6 (0.04) 25.6 (0.04) 0.075 (5.184) 0.017t 0.068 0.303

Mood

% 10 on CESD short form 20.8 (0.043) 51.5 (0.042) 48.6 (0.045) <0.0001 (36.305) <0.0001u <0.0001v 0.330

Risk perception

% “likely, very likely or certain” to develop disease if continue smoking5 10.0 (0.036) 50.3 (0.041) 68.6 (0.046) <0.0001 (109.921) <0.0001w <0.0001x 0.0003y

% endorsing large decrease or elimination of risk of disease if quit smoking6 49.5 (0.051) 76.8 (0.034) 79.7 (0.039) <0.0001 (40.074) <0.0001z <0.0001aa 0.294


UCS1 Smokers

% (SD)

SPB2 Smokers

% (SD)

HCS3 Smokers

% (SD)


UCS vs. SPB


SPB vs. HCS

P-Value4

% endorsing a “little” or “a lot” lower chance of disease vs. other smokers7 41.2 (0.058) 18.2 (0.033) 5.6 (0.022) <0.0001 (60.596) <0.0001bb <0.0001cc 0.0003dd

% endorsing that, if they were to develop a smoking-related illness, it would be “as or more serious” as other smokers’7 74.1 (0.052) 92.2 (0.021) 98.9 (0.011) <0.0001 (51.668) <0.0001ee <0.0001ff 0.003gg

Notes: Notes: 1Unconvinced Smokers, 2Smokers with Psychosocial Barriers, 3Health Concerned Smokers 4 P-values for when testing that the larger of the two proportions is statistically significant, 5 perceived vulnerability, 6 precaution effectiveness/outcome, expectation, 7 optimistic bias

Pearson’s 2 test statistics for p-values significant at the 0.05 level are: a32.01, b9.85, c6.55, d10.86, e13.41, f12.41, g21.64, h257.74, i281.67, j267.03, k265.80, l41.51, m84.76, n13.43, o28.26, p42.22, q30.99, r80.60, s16.52, t4.54, u31.21, v24.54, w58.11, x105.85, y11.53, z25.57, aa29.70, bb20.34, cc54.57, dd11.96, ee18.98, ff40.47, gg7.40.


Figure 1. Proportion of classes with question responses.

€¦ · web viewidentification of three different types of smokers who are not motivated to...

Documents