Responses to Tobacco Smoking-Related Health Messages in Young People With Recent-Onset Schizophrenia

Abstract

Virtually no research has examined the responses of youth with recent-onset psychosis (ROP) to smoking-related health warnings. We examined predictors of response and tested hypotheses that participants with ROP would (a) assess warnings as less effective than a healthy comparison (HC) group, and (b) assess video warnings as more effective than pictures. ROP participants (n = 69) had <2 years of prior antipsychotic treatment; the HC group (n = 79) had no major mental illness. Participants viewed 10 pictorial warnings, 8 videos depicting similar messages, and were interviewed regarding tobacco use, health literacy, and smoking knowledge. We assessed response at baseline and at 4-week follow-up. ROP participants were more likely than HC to smoke tobacco (49.3% vs 10.1%) and had lower levels of health literacy and smoking-related knowledge. Cannabis was used by 46.4% of ROP participants. Effectiveness ratings were high for both picture and video warnings with no differences between media. ROP participants compared to HC and nonsmokers compared to smokers were more likely to perceive warnings as effective. Effectiveness was associated with negative affect and greater emotional arousal. We assessed 33 smokers at follow-up; 5 (15%) identified as nonsmokers, 15 (45%) made a quit attempt, and 16 (49%) reported that the warnings influenced their smoking. Results indicate that young people with psychotic disorders respond favorably to health warnings. Effective messages depict health consequences clearly, elicit negative emotions, and may impact smoking behavior. Future research is needed to understand the effects of mode of presentation and message comprehension on smoking behavior.

Introduction

Tobacco smoking remains the leading preventable cause of mortality in the United States.¹ Smoking rates have declined in the United States and internationally^2,3 partially due to public health interventions that restrict tobacco-related marketing and improve education about smoking-related health consequences.⁴ However, smoking rates remain high in persons with mental illness, and schizophrenia in particular. Smoking prevalence among persons with psychosis range from 44% to 77%,^5–10 with over 50% of 404 participants with first episode psychosis reporting current smoking in a recent treatment study.¹¹ People with severe mental illness (SMI) are also more likely to be heavy smokers (>1 ppd),¹² less likely to quit,¹³ and die up to 25 years earlier than the general population, primarily due to smoking-related cardiac illness.¹⁴ Almost one-third of current smokers in the United States have a mental illness¹⁵ and smoking accounts for over half the illness-related deaths in persons with schizophrenia.¹⁶

Public health initiatives such as the WHO Framework Convention on Tobacco Control¹⁷ and legislation such as the Family Smoking Prevention and Tobacco Control Act in the United States (FSPTCA; 2009) have contributed to declining smoking rates by promoting economic and educational disincentives to use tobacco, targeting high-risk groups for education (including adolescents, young adults, and persons with mental illness¹⁸), and encouraging evidenced-based cessation treatments. These initiatives guide tobacco-related health message dissemination, including the placement of graphic warning labels (GWLs) on tobacco products. Disseminating vivid depictions of smoking consequences has been associated with population-based changes in smoking attitudes. Less change is noted in countries that have not yet implemented GWLs.¹⁹ Prior research suggests that effective smoking-related health warnings are large, prominently displayed, and combine pictorial images with words.²⁰ Images illustrating negative health effects of smoking are more effective than other image types, and pictures evoking strong emotions are more effective than less evocative images.²⁰ US implementation of GWLs has been delayed by litigation from the tobacco industry.²¹ Health warnings target both nonsmokers to prevent smoking initiation as well as smokers to encourage cessation. Educational initiatives frequently target adolescents, the critical age period for initiating smoking.²² Adolescents’ and young adults’ responses to smoking-related health warnings suggest that they notice and retain information about health warnings.^23,24 Young smokers, however, may perceive messages as less salient and less likely to influence smoking behavior²⁵ than older smokers.²⁶ Youth are also more likely to recall and respond to health messages with personal testimonials, a novel narrative, and high message sensation value (MSV).^27,28

Video format health messages appear on television and also on unregulated internet sites such as YouTube that are popular among youth. A YouTube content analysis found that more frequently viewed and discussed antismoking videos had higher MSV; however, humorous messages with high MSV are less popular and subvert antismoking messages by reducing credibility.²⁹ We are unaware of published research comparing the effectiveness of video communications to static pictorial messages. Examining the impact of new media on health messaging is warranted, because both tobacco promotion and antismoking messages now widely appear in this medium.

Despite high smoking rates among young persons with psychotic disorders, the temporal association between age of smoking onset and of psychotic symptoms, and data suggesting that changes in tobacco use may be associated with prodomal and psychotic symptoms,⁹ there are virtually no data assessing the impact of health-related warnings in this vulnerable group. This gap stands in stark contrast to evidence that patients with psychotic disorders have been the target of tobacco industry marketing efforts to promote smoking and normalize the behavior among mental health clinicians.³⁰ To fill this gap, we investigated perceptions of effectiveness and affective responses to smoking-related health warnings in young people with recent-onset psychosis (ROP) and in a comparison group of persons without mental illness. Message formats included pictorial warnings and brief video-clips. We assessed smoking-related knowledge and examined the effects of message viewing on subsequent knowledge and behavior. We hypothesized that (a) people with ROP would perceive warnings as less effective than a healthy comparison (HC) group; and that (b) video warnings would be perceived as more effective and processed more intensely than picture/text warnings. We also examined the demographic, clinical, and behavioral predictors of responses to the health messages.

Methods

Design

We assessed ROP and HC participants at baseline, presented the health warnings within a structured protocol, and assessed responses to the warnings immediately and 4 weeks later.

Setting and Participants

The investigation was conducted at an academic medical center serving a socioeconomically diverse urban and suburban catchment area. Eligible participants were tobacco smokers and nonsmokers ages 15 to 40 with no cognitive impairments hindering informed consent or study completion. Written informed consent was obtained to participate (or parental consent and written assent for participants <18) for this IRB-approved investigation. Data were collected between January, 2013 and March, 2015.

Participants were recruited from specialized early-phase psychosis programs. Eligible participants had a diagnosis of schizophrenia, schizophreniform disorder, schizoaffective disorder, or psychotic disorder NOS, and less than 2 years of prior antipsychotic treatment. Exclusion criteria included medical impairments, substance-induced psychotic disorder, psychotic symptoms preventing study completion, or acute risk of suicidal/homicidal behavior. Initial diagnostic eligibility was confirmed by the Structured Clinical Interview for Axis I DSM-IV Disorders (SCID).³¹ These data were later reviewed in a consensus conference³² for final diagnostic assignment.

HC participants were recruited from the local community with promotional flyers. Exclusion criteria included taking psychotropic medications or meeting criteria for any current Axis I disorder except simple phobias on a SCID-NP interview.³³

Procedures

After assessment of clinical status, tobacco use history, smoking characteristics, smoking knowledge and health literacy, participants viewed 18 health warnings in random order for a minimum of 15 seconds each. Videos and text/picture warnings were interspersed and sequencing was randomized to control for order effects. After viewing each warning, participants rated its effectiveness and their emotional response to the warning. After viewing all stimuli, participants were asked to formulate as many questions as possible about smoking. Repeat assessment of smoking-related knowledge, attitudes, and behavior occurred 4 weeks later.

Health Warnings

Nine print warnings from the proposed US cigarette labels³⁴ were supplemented with 10 GWL images from other countries.^35–38 We selected messages using younger models that depicted relevant messages for young people (eg, smoking affects physical appearance or sexual functioning). All warnings were formatted to conform to the text, typography and layout of the US warnings.

These 19 preliminary images were presented at two 90-minute focus groups attended by 3 men and 3 women with schizophrenia (4 smokers and 2 nonsmokers, mean age = 29.62, SD = 6.12). Group content analysis suggested that effective warnings depicted actual individuals (as opposed to cartoon drawings or abstract/nonhuman images) and elicited strong negative emotions (eg, sadness, empathy, or disgust). Participants had difficulty grasping some of the relationships between the health warning text and associated pictures. Effective messages clearly explained the health consequence depicted in the picture. Ten warnings were selected for subsequent study based on focus group efficacy ratings.

Eight videos with equivalent health-related messages were developed. Educational messages paralleled the text/picture warnings, lasted 2 minutes or less, deployed a testimonial style demonstrated to be preferable in prior research with young people,^39,40 included young actors from diverse ethnic/racial backgrounds, and ranged in emotional tone from light and humorous to dark and fear-inducing. The final set of warnings is shown in figure 1.

Fig. 1.

Final set of tobacco-related health warnings.^34–38

Clinical Assessments

Current psychiatric functioning, positive symptoms, and current depressed mood were assessed with the Brief Psychiatric Symptom Rating Scale- A (BPRS-A)⁴¹ and negative symptoms with the Hillside Clinical Trials version of the Schedule for the Assessment of Negative Symptoms (SANS).⁴² Assessments were performed by trained Doctoral- or Master’s-level psychometricians.

Smoking and Related Characteristics

A Smoking Behavior and Engagement Battery was developed including (a) items assessing quantity/frequency of current smoking behavior and quit attempts. Current smoking was defined as any cigarette use in the past 30 days and >100 cigarettes smoked lifetime. This battery also included (b) a Likert-scale item assessing recent health warning exposure, (c) The Fagerstrom Test of Nicotine Dependence (FTND)⁴³ to assess smoking behavior in current smokers; and (d) selected items from the Population Assessment of Tobacco and Health survey to obtain current and prior tobacco product use.⁴⁴ Tobacco-related marketing exposure was assessed using a 17-item checklist used in tobacco research in youth.⁴⁵

Smoking and Health Knowledge

Participants completed an 18-item smoking knowledge questionnaire developed for this investigation. This questionnaire used simple dichotomous questions based broadly on content areas used in prior research among participants with psychotic disorders.⁴⁶ Preliminary psychometric examination indicated good internal consistency (α = 0.78) suggesting that we were tapping an underlying “smoking knowledge” construct. Participants also completed the short-form of the Test of Functional Health Literacy (S-TOFHLA)⁴⁷ to assess participants’ abilities to process and apply health-related information.

Perceptions of and Responses to Health Warnings

Participants rated warning effectiveness using 5-point Likert-scale items adapted from a study of young smokers’ reactions to warning labels.⁴⁸ Items tapped each message’s effectiveness at (a) preventing smoking initiation; (b) encouraging quit attempts; and (c) increasing concern about the impact of smoking on health. Emotional response was rated on 9-point visual analog scales using the Valence (degree of positive/negative affect) and Arousal items of the Self-Assessment Manikin (SAM).⁴⁹ After rating all stimuli, participants were asked to formulate questions about smoking, guided by a structured question eliciting protocol⁵⁰ used in interventions to activate patients in psychiatric⁵¹ and primary care medical treatment.⁵² Participants were encouraged to produce as many questions as possible in response to the following eliciting statement or question focus: “questions to help you make decisions about smoking cigarettes.”⁵⁰

Data Analysis

We used t tests for independent means to evaluate differences between continuous variables and χ² analyses to test differences among categorical variables. Univariate analyses were conducted to compare variables according to study group (ROP vs HC).

Each patient generated measures of perceived effectiveness, affect, and arousal for each of the 18 stimuli; item analysis procedures identified good-to excellent internal consistency within the 3 effectiveness questions for both picture/text warnings (mean Chronbach’s α = 0.88, range = 0.82–0.94) and videos (Mean α = 0.89, range = 0.82–0.94). A total effectiveness score (score range 3 to 15) was computed for each health warning, and 2 summative effectiveness scores were calculated according to message modality (picture/texts and videos). A matrix of bivariate Pearson r analyses examined patterns of association between perceived effectiveness and affective dimensions (valence and arousal).

Exploratory t tests compared ratings for each stimulus according to study group followed by a 2-way analysis of variance (ANOVA) to evaluate interactions between study group (ROP vs HC) and smoking status on perceived warning effectiveness.

A multiple linear regression analysis was conducted to identify significant predictors of perceived warning effectiveness from among demographic, clinical, and smoking-related dependent variables. Backward elimination procedures were used until the model only contained variables with P values < .05. Significance tests were generally set at α = 0.05 except for exploratory analyses of the 18 warning stimuli, where a Bonferroni multiple comparison correction was applied. Analyses were completed using the Statistical Package for the Social Sciences (SPSS) V. 22.

Results

Overall, 158 individuals consented to participate (77 with ROP and 81 HC) and 148 (69 patients with ROP and 79 HC) viewed all the health warnings and completed initial assessments. ROP participants met DSM-IV criteria for Schizophrenia-undifferentiated type (n = 34), paranoid-type (n = 19), Schizoaffective Disorder (n = 4) Psychotic Disorder NOS (n = 2), or Schizophreniform disorder (n = 10). Follow-up assessments were obtained from 117 participants (79% of those who viewed the health warnings; 53 with ROP and 64 HC).

Participants with ROP were approximately 5 years younger (t = 5.93, df = 146, P < .001), more likely to be male (χ² = 26.24, df = 1, P < .001), and had lower levels of education (χ² = 55.70, df = 4, P < .001; table 1). Both groups were racially diverse; Hispanic ethnicity was more common among the ROP vs HC group (χ² = 4.16, df = 1, P = .04).

Table 1.

Demographic and Clinical Characteristics

Characteristic	Recent-Onset Psychosis (n = 69)	Healthy Comparison Group (n = 79)	Statistics
Demographic variables
Male (%)	55 (79.7%)	30 (38%)	χ2 = 26.24, df = 1, P < .001
Mean age (SD)	22.67±5.2	27.97±5.6	t = 5.93, df = 146, P < .001
Ethnicity (% of Hispanic origin)	13 (18.8%)	8 (7.6%)	χ2 = 4.16, df = 1, P = .04
Race (%)
Caucasian	25 (26.2%)	38 (48.1%)	χ2 = 3.66, df = 3, P = .30
African American	22 (31.9%)	22 (27.8%)
Asian	15 (21.7%)	16 (20.3%)
Mixed race/other	7 (10.1%)	3 (3.8%)
Level of education (%, n = 64)
≤High school graduate	22 (31.9%)	0 (0%)	χ2 = 55.70, df = 4, P < .001
High school graduate	14 (40.63%)	7 (10.9%)
Some college	28 (40.6%)	16 (25%)
College graduate	3 (4.3%)	22 (34.4%)
Graduate degree	2 (2.9%)	19 (29.7%)
Smoking variables
Current smokers	34 (49.3%)	8 (10.1%)	χ2 = 27.77, df = 1, P < .001
Fagerstrom total score (in smokers)	2.36±2.16	2.37±1.77	t = 0.15, df = 39, P = .99
Smoking knowledge (M, SD, n = 63)	84.57±17.95	91.71±7.93	t = 3.17, df = 140, P = .002
Health literacy (S-TOFHLA) (M, SD)	30.44±6.9	35.24±1.1	t = 6.13, df = 146, P < .001
Number of smoking-related questions	12.12±6.25	17.36±11.35	t = 3.41, df = 145, P < .001

Note: S-TOFHLA, Test of Functional Health Literacy.

Table 2.

Clinical Profile of Participants With Recent-Onset Psychosis (ROP, n = 69)

Characteristic	M ± SD	Nonsmokers (n = 35)	Smokers (n = 34)	Statistics
Positive symptoms
BPRS (total score)	30.35±8.56	27.41±6.85	33.29±9.17	t = 2.98, df = 66, P = .004
BPRS-hallucinatory behavior	2.39±1.58	2.17±1.38	2.62±1.76	t = 1.18, df = 67, P = .25
BPRS-conceptual disorganization	1.51±0.98	1.40±0.91	1.62±1.04	t = 0.92, df = 67, P = .36
BRPS-unusual thought content	2.72±1.61	2.43±1.29	3.03±1.85	t = 1.57, df = 67, P = .12
BPRS-depression	1.84±1.13	1.54±0.82	2.14±1.33	t = 2,28, df = 67, P = .03
Negative symptoms
SANS-affective flattening	1.87±0.79	1.85±0.74	1.88±0.86	t = 0,13, df = 65, P = .90
SANS-alogia	1.46±0.76	1.32±0.53	1.62±0.92	t = 1,61, df = 66, P = .11
SANS-avolition	1.88±0.76	1.91±0.79	1.85±0.76	t = 0.33, df = 65, P = .74
SANS-asociality-anhedonia	1.74±0.80	1.60±0.69	1.88±0.88	t = 1.48, df = 67, P = .14
Comorbid substance abuse disorder	32 (46.4%)	9 (25.7%)	23 (67.6%)	χ2 = 12.19, df = 1, P < .001
Comorbid alcohol abuse disorder	17 (24.6%)	8 (22.9%)	9 (26.5%)	χ2 = 0.12, df = 1, P = .73
Medication treatment
Risperidone	51 (73.9%)
Aripiprazole	10 (14.5%)
Olanzapine	2 (2.9%)
Clozapine	2 (2.9%)
Haloperidol	2 (2.9%)
Quetiapine	1 (1.5%)
Quetiapine+Aripiprazole	1 (1.5%)

Note: BRPS, Brief Psychiatric Symptom Rating Scale.

Antipsychotic treatment among ROP participants by frequency were: risperidone (n = 51 or 74%), aripiprazole (n = 10 or 14.1%), haloperidol (n = 2 or 2.9%), olanzapine (n = 2 or 2.9%), clozapine (n = 2 or 2.9%), quetiapine (n = 1 or 1.5%) and combination aripiprazole and quetiapine (n = 1 or 1.5%). ROP participants had relatively low levels of positive and negative symptoms at baseline and mean BPRS total scores were in the mildly ill range.⁵³ Thirty-two ROP participants (46.4% of ROP participants) had comorbid substance abuse disorders. All had a cannabis use disorder; additional diagnoses included cocaine abuse (n = 2), opioid abuse (n = 1), and polysubstance dependence (n = 1).

Nicotine dependence among smokers was in the low range in both groups. Just under half of the ROP group reported current tobacco smoking (n = 34 or 49.3%) compared to 10% of the HC group (χ² = 27.77, df = 1, P < .001). Health literacy levels were high overall (with 138/148 or 93.5% of all participants in the “adequate” range of functioning); however, participants with ROP had lower mean health literacy scores (t = 6.13, df = 146, P < .001) and lower levels of smoking knowledge (t = 3.17, df = 140, P = .002). Smokers with ROP had higher mean total BPRS scores (P = .004), higher BPRS Depression Scores (P = .03) and were more likely to have comorbid substance abuse (P < .001).

Exploratory examination of knowledge item responses suggested that a greater proportion of ROP participants believed that smoking “does not increase the chances of young people having a stroke” (P = .001), “only affects the brain if you smoke for 10 years or longer” (P = .02), and that it took “years of smoking before you have trouble quitting” (P = .04). ROP participants also were more likely to endorse that smoking “did not increase risks of emphysema or lung disease” (P = .003), heart disease (P = .005), cancer (P = .04), or miscarriage (P = .01). Finally, ROP participants were also more likely to perceive that smoking “did not reduce the amount of oxygen the baby gets” in pregnant women (P = .04) or believe that arsenic was a cigarette ingredient (P = .04).

Perceptions of Health Warning Effectiveness

Over 90% of participants perceived the health warnings as effective. Only 12 subjects (8.1%) rated pictures as less than moderately effective, and only 14 participants (9.5%) rated videos as less than moderately effective. We observed high mean effectiveness ratings for both the set of 10 pictures (12.28±2.04) and the 8 videos (M = 12.20±2.19) that were not statistically different (t = 0.68, df = .146, P = .50).

Perceived Effectiveness and Affective Response

The association between perceived effectiveness and affective response (positive/negative valence and arousal) was assessed for each stimulus (table 3). Moderate-to-high associations were observed (mean r = 0.42±0.11) between effectiveness ratings and valence, with greater effectiveness associated with more negative or sad affective response. The strength of association was weaker in response to health messages that used humor to deliver the health warning. Significant but more modest associations were observed between perceived effectiveness and arousal level (mean r = 0.28±0.09), with more effective messages evoking greater levels of arousal/stimulation.

Table 3.

Associations Between Perceived Effectiveness and Affective Response (Valence and Arousal) in Total Sample (n = 148)

Stimulus/ Message	Association Between Perceived Effectiveness
	Valence (r)	Arousal (r)
Picture A: Smoking isn’t pretty.	0.54***	−0.36***
Picture B: Cigarettes cause heart disease.	0.41***	−0.33***
Picture C: Smoking causes fatal lung disease.	0.47***	−0.21**
Picture D: Smoking contains benzene, nitrosamines, formaldehyde, and hydrogen cyanide.	0.50***	−0.32***
Picture E: Smoking is a major cause of strokes.	0.49***	−0.39***
Picture F: Secondhand smoke can cause stillborn births, low birth weight, and other pregnancy problems.	0.30***	−0.12
Picture G: Smoking causes impotency.	0.45***	−0.35***
Picture H: Cigarettes are addictive.	0.42***	−0.39***
Picture I: Smoking causes premature births.	0.42***	−0.01
Picture J: Cigarettes are addictive.	0.57***	−0.29***
Video A: Smoking is addictive.	0.43***	−0.28***
Video B: Smoking causes lung cancer.	0.55 ***	−0.29***
Video C: Smoking causes heart attack.	0.46***	−0.34***
Video D: Cigarettes contain harmful chemicals.	0.40***	−0.25**
Video E: Smoking causes impotence.	0.25**	−0.25**
Video F: Smoking harms physical appearance.	0.19*	−0.31***
Video G: Smoking harms the developing fetus.	0.53***	−0.27***
Video H: Secondhand smoke causes disease.	0.52***	−0.32***
Mean	0.43±0.11	−0.28±0.09

Note: *P < .05; **P < .01; ***P < .001.

Perceived Effectiveness, ROP, and Smoking Status

Given the similar effectiveness and arousal ratings for picture text warnings and videos, these 2 scores were combined for subsequent analyses. Two-way ANOVA identified significant main effects for smoking status (F = 8.60, df = 3,143, P = .004) and illness group (F = 6.37, df = 3,143, P = .01) on perceived effectiveness. Post hoc examination of scores revealed that ROP participants perceived warnings as “more effective” (M = 12.47, SD = 2.12, range = 5 to 15) than healthy controls (M = 12.04, SD = 1.96), and nonsmokers in both study groups (M = 12.44, SD = 1.90) perceived warnings as more effective than smokers (M = 11.74, SD = 2.30). Multivariate regression analysis (table 4) identified only smoking status and illness group as significant predictors of perceived effectiveness from among the set of potential variables including demographics (age, sex, ethnicity) smoking-related knowledge scores, and health literacy scores (F value for the final model = 4.67, df = 2,147, P = .01).

Table 4.

Variables Significantly Associated With Perceived Effectiveness to the Health Warnings

Independent Variable	B	SE	β	t Value	P
Group (ROP vs HC)	−1.72	0.73	−0.21	−2.37	.019
Smoking status	−2.22	0.80	−0.25	−2.76	.006

Note: HC, healthy comparison; ROP, recent-onset psychosis.

Response to Health Warnings at Follow-up

One month after viewing the warnings, HC compared with ROP participants were more likely to have discussed the health warnings (χ² = 8.24, df = 1, P = .004) and were able to generate more smoking-related questions (t = 3.41, df = 145, P < .001). Participants in both groups reported significantly greater attention to health warnings at the follow-up assessment when compared to their recollections in the month prior to viewing the study warnings (F = 12.61, df = 1,1,111, P < .001). Between-groups analysis indicated that change in level of attention was greater in ROP than HC participants, but this finding failed to attain statistical significance (F = 3.68, df = 1,1,111, P = .06).

Of the participants returning for follow-up, 5 (15%) of the 33 smokers identified as “nonsmokers” at follow-up (4/27 with ROP and 1/6 HC), and an additional 10 ROP participants reported making 1 or more quit attempts. Five participants had tried nicotine replacement, and 1 had joined a quit program. None initiated smoking between baseline and follow-up. Just under half of the follow-up sample of smokers (16/33 or 48.5% of baseline smokers and 10/15 or 66.7% of the quit attempters) reported that their decisions about smoking were at least somewhat influenced by the health warnings. Among smokers, quit attempters perceived the health warnings as eliciting more negative affect than nonattempters (t = 2.46, df = 30, P = .02).

Discussion

To our knowledge this is the first report documenting reactions to health warnings in persons with schizophrenia, the first to compare the reactions of persons with schizophrenia to healthy controls, and the first report comparing responses to picture/text warnings and brief video messages. ROP participants were more likely to be current tobacco smokers and had limited knowledge of smoking-related health risks. Overall, both pictorial and video health warnings were rated as highly effective. Controlling for demographic and pre-assessment knowledge and literacy differences, ROP participants perceived health warnings as more effective than healthy controls. The ranking of perceived warning effectiveness was: nonsmokers with psychotic disorders (highest effectiveness), followed by healthy nonsmokers, smokers with psychotic disorders, and healthy smokers (least effectiveness). Results suggest that health messages containing clear messages delivered with pictures or video are highly relevant in this population.

Perceived warning effectiveness was associated with higher negative affect and emotional arousal, particularly for serious messages depicting life-threatening health consequences. These results are broadly consistent with tobacco message findings with the general youth population^24,28 and suggest that pictorial health warnings are not contraindicated in people with psychosis.

Contrary to our hypothesis, participants responded equally to video and picture/text warnings. This supports the use of both pictorial and video media messages to motivate and educate young people with psychotic disorders, as well as disseminating clearly presented, loss-framed and arousing pictures on cigarette packages. Of note, we excluded messages depicting cartoons, x-rays and other abstract concepts that were not well understood in the initial ratings by ROP participants.

During the month following structured health warning exposure, participants responded to them cognitively and behaviorally. Remarkably, 45.4% of the participants reported they tried to quit and 15% became abstinent. Prior research testing the impact of health warnings on subsequent behavior has shown variable and modest results.⁵⁴ Our findings are broadly consistent if not slightly higher than studies examining relations between health warning exposure and subsequent quit attempts.^55,56 Our results contrast with the only study of attitudes about smoking-related health warnings in persons with psychotic disorders, in which qualitative analysis of interviews suggested that health warnings may not only fail to motivate smokers with psychosis to quit, but in fact may cue viewers to the perceived benefits of tobacco use.⁴⁶

This exploratory investigation has limitations. First, our HC group was a convenience sample recruited by advertisements rather than a matched control group. The HC sample was older, and had a different sex and educational distribution. Analyses of interactions between illness group and smoking status on warning effectiveness may have been subject to Type II error related to the low prevalence of smoking in the HC group. Second, we recruited participants from a single site in an area with high tobacco taxation rates and widely disseminated, multimedia, and graphic antismoking campaigns. Observed warning response may have been influenced by exposure to prior messages. By design, our protocol facilitated processing of multiple health messages to obtain participants’ responses to them, unlikely to occur in real-world settings. Thus the 1-month follow-up behavioral responses are likely magnified. Further, without a control condition we cannot confirm that the impact reported at follow-up was not due to study assessments, unmeasured factors, or chance. Fourth, 31 subjects were lost to follow-up and this may have skewed the data on subsequent smoking behavior. Finally, we limited our area of inquiry to perceived effectiveness and affective response to health warnings as these are previously identified predictors of health warning impact on future smoking behavior.⁵⁷

Our results nonetheless identify important areas for future study. Our ROP participants who did not smoke responded positively to the warnings; suggesting exploring if graphic warnings could prevent smoking initiation in this highly vulnerable population. Our relatively low-intensity intervention (processing a series of warnings on a computer monitor) impacted short-term smoking behavior in ROP smokers. This challenges conceptions that attitudes and behavior in this population are unmalleable. Overall, our results show that public health efforts using graphic warnings, presented in a distraction-free environment, can reach young people with psychotic disorders. This group may need more education about smoking-related health consequences, and help translating messages into concrete long-term behavior change. Future research could compare responses to graphically rich messages and existing text-based warnings, and describe how messages are processed in real-world environments—eg, when they are disseminated on cigarette packaging through media campaigns and through strategic placement in locations where cessation support and education might be obtained (eg, in mental health clinic environments, waiting rooms, offices, etc.).

Future research might also examine interrelations among smoking knowledge, warning comprehension, and other social-cognitive variables known to mediate health-related decision making in general⁵⁸ and smoking behavior in persons with SMI.^59–61 Causal pathways between viewing health warnings and subsequent behavior are inadequately understood in general populations⁵⁵ and have yet to be studied in persons with SMI. Finally, our findings raise the possibility that pictorial and video health warnings could be incorporated into motivational interventions and cessation decision support tools targeting ROP smokers. The environmental context in which these interventions are implemented would need to assess potential barriers to implementation related to cultural norms normalizing tobacco use within mental health treatment environments.⁶²

In summary, young people with schizophrenia responded positively to graphic health warnings. Exposure to and processing of these messages led to cognitive and behavioral cessation activity. Given the ongoing high rates of smoking initiation and persistence leading to early morbidity and mortality in this group, future research is warranted to explore the role of these messages in addressing tobacco use disorder in schizophrenia.

Funding

This work was supported by an administrative supplement to National Institutes of Mental Health (grant P50 MH080173 [PI: A.K.M.]) provided under the Research Relevant to the Family Smoking Prevention and Tobacco Control Act.