Internet use among childhood and young adult cancer survivors who smoke: implications for cessation interventions

Abstract

Objective

To identify patterns of Internet use among childhood and young adult cancer survivors who smoke.

Methods

Baseline assessment data were collected from 2005 to 2008 for the Partnership for Health-2 (PFH-2) study, a web-based smoking cessation intervention for childhood and young adult cancer survivors. Participants were surveyed about their Internet access and use. Sociodemographic, clinical, and psychosocial data also were collected.

Results

Internet access and use was widespread among PFH-2 participants. However, older, less-educated, and female survivors reported less frequent Internet use, even when they had access to the Internet at home and/or at work. These associations were significant in multivariable analyses.

Conclusions

Although the digital divide is narrowing, Internet use and engagement remains socially patterned. web-based prevention interventions are a promising method of reaching this geographically dispersed, high-risk population, but certain subgroups—particularly older and lower socioeconomic status survivors—might be missed by this approach.

Introduction

Smoking rates among childhood cancer survivors are non-trivial, with estimates ranging from 17 to 29% for current smoking [1–5]. Although these rates are comparable to those of healthy individuals, smoking presents unique risks to survivors because of the documented late effects of childhood cancer treatment (e.g., increased risk of cardiovascular disease, secondary malignancy, and pulmonary conditions) [6–9].

Fortunately, there is evidence that we can intervene to reduce smoking rates in this population. Partnership for Health (PFH), an intervention that provided peer-delivered telephone counseling, tailored and targeted print materials, and free nicotine replacement therapy to childhood cancer survivors who smoke, produced a doubling of quit rates compared to a self-help control group [10]. Additionally, long-term cessation rates were higher among those who received the intervention [11]. A central question, however, is how to disseminate successful interventions such as PFH, particularly to a population that is as geographically dispersed as childhood cancer survivors. Internet-based interventions have the potential to increase reach while maintaining intervention efficacy [12], but the feasibility of such interventions hinges, at least in part, on whether survivors have access to and routinely use the Internet.

This study explores patterns of Internet use among childhood and young adult cancer survivors who smoke and, in doing so, assesses which survivors might be reached via Internet-based cessation interventions and which might need to be reached via other channels.

Methods

Study participants

Data for this study come from the Partnership for Health-2 (PFH-2) baseline survey, which was administered from 2005 to 2008. The PFH-2 intervention tests whether web-based and tailored print material formats of the PFH peer-delivered program produce similar rates of cessation as the labor-intensive PFH peer-delivered telephone counseling intervention. Participants were recruited from five U.S. and Canadian cancer centers: St. Jude Children’s Research Hospital, Memorial Sloan Kettering Cancer Center, Princess Margaret Hospital, The Hospital for Sick Children, and Dana-Farber Cancer Institute/ Partners HealthCare. Institutional review board approval was obtained at each recruitment site. Eligibility criteria included having a cancer diagnosis before age 35, not currently being treated for cancer and being out of treatment for at least 2 years, being mentally able to provide informed consent, being reachable by telephone, being able to speak English, and being a current smoker (defined as having smoked at least one puff in the last 30 days). Each recruitment site performed a preliminary screen for eligibility, and an introductory letter was mailed to potentially eligible survivors. The letter included information on the study and how to opt out of future contact. The study survey team then contacted consenting survivors, verifying eligibility and administering the baseline survey. In addition, study information was available on childhood and young adult survivorship websites including http://www.cancer.org, http://www.planetcancer.org, and http://www.ulmanfund.org; interested survivors contacted the study team and provided verbal consent.

Introductory letters were mailed to 4,345 cancer survivors, and an additional 54 people contacted the study team after noticing website postings. A total of 51 survivors were not contacted further, yielding 4,348 potential study participants. Of the 4,348 who were contacted by the study survey team, 18% (n = 773) were alive and eligible; ineligibility was largely due to smoking status. A total of 48% (n = 374) of eligible survivors were enrolled in the study. Additional information on sampling procedures and study design is provided elsewhere [13].

Measures

Several questions assessed Internet access and use. Items were adapted from the National Cancer Institute’s Health Information National Trends Survey (HINTS), which routinely collects nationally representative data about Americans’ understanding and use of cancer-related information [14, 15]. Participants were asked if they owned a computer at home. Internet access was assessed with the question: “Do you have access to the Internet at home, at work, both, or neither?” Participants also reported how frequently they used a computer, checked their e-mail, and used the Internet. Response options included “daily,” “two to six days a week,” “weekly,” “monthly,” “less than monthly,” “rarely,” and “never.” For descriptive and bivariate analyses, categories were collapsed into daily, weekly, and monthly or less frequently. For multivariable analyses, Internet use was dichotomized (daily versus non-daily use).

The PFH-2 baseline assessment also collected socio-demographic data (age, education, gender, race/ethnicity, employment, marital status) and clinical data (childhood cancer diagnosis, time since diagnosis, cancer treatment received, physician office visits, self-reported health status). Participants were asked about their risk perceptions, or how likely it was that they would experience “any serious health problem,” “a diagnosis of cancer,” “heart problems,” or “lung problems” in the future on a scale of 1–7. These four items were summed to create a continuous variable (range = 1–28, with higher values indicating greater risk perception). Psychosocial factors (cancer-related worry, cancer-related distress, depressive thoughts) were captured using items described previously [13].

Statistical analysis

Means and standard deviations for continuous variables and frequencies for categorical variables were obtained for all key variables. Distributional assumptions were tested for skewness, and measures of outliers were assessed. Chi-square statistics and analysis of variance (ANOVA) were used to assess individual differences in sociodemographic, clinical, and psychosocial variables between the three levels of Internet use (see Table 1). A parsimonious multivariable logistic regression model predicting daily Internet use was created to calculate odds ratios (OR) and 95% confidence intervals (CI). We began by building a model that used all variables that were significant at the p = 0.10 level in bivariate models. We then performed backward stepwise elimination of non-significant predictors, and considered the effect of elimination on the coefficients of remaining predictors in the model. Interaction and model assessments were conducted on the final model. All analyses were conducted in SAS Version 9.3.

Table 1.

Internet use among PFH-2 childhood and young adult cancer survivors who smoke and have Internet access^a, 2005–2008 (n = 308)^b

	Total n (%)	Frequency of Internet use
		Daily n (%)	Weekly n (%)	≤Monthly n (%)
Overall	308 (100.0)	204 (66.2)	77 (25.0)	27 (8.8)
Sociodemographic factors
Age*
25 or younger	81 (26.3)	58 (71.6)	17 (21.0)	6 (7.4)
26–30	59 (19.2)	44 (74.6)	11 (18.6)	4 (6.8)
31–35	68 (22.1)	44 (64.7)	18 (26.5)	6 (8.8)
36–40	53 (17.2)	36 (67.9)	13 (24.5)	4 (7.6)
41 or older	47 (15.3)	22 (46.8)	18 (38.3)	7 (14.9)
Education*
Less than high school	19 (6.2)	9 (47.4)	2 (10.5)	8 (42.1)
High school	79 (25.7)	44 (55.7)	25 (31.7)	10 (12.7)
Some college or vocational school	106 (34.4)	72 (67.9)	27 (25.5)	7 (6.6)
College and above	104 (33.8)	79 (76.0)	23 (22.1)	2 (1.9)
Gender*
Male	160 (52.0)	115 (71.9)	29 (18.1)	16 (10.0)
Female	148 (48.1)	89 (60.1)	48 (32.4)	11 (7.4)
Race/ethnicity*
White	267 (86.7)	181 (67.8)	67 (25.1)	19 (7.1)
Non-white	41 (13.3)	23 (56.1)	10 (24.4)	8 (19.5)
Employed during past year
Yes	257 (83.4)	172 (66.9)	64 (24.9)	21 (8.2)
No	51 (16.6)	32 (62.8)	13 (25.5)	6 (11.8)
Married or partnered
Yes	147 (47.9)	90 (61.2)	43 (29.3)	14 (9.5)
No	160 (52.1)	114 (71.3)	33 (20.6)	13 (8.1)
Clinical factors
Childhood cancer diagnosis
Leukemia	71 (23.1)	46 (64.8)	17 (23.9)	8 (11.3)
Hodgkin’s disease	59 (19.2)	37 (62.7)	15 (25.4)	7 (11.9)
CNS malignancy	24 (7.8)	19 (79.2)	3 (12.5)	2 (8.3)
Non-Hodgkin’s lymphoma	21 (6.8)	14 (66.7)	6 (28.6)	1 (4.8)
Bone cancer	24 (7.8)	17 (70.8)	5 (20.8)	2 (8.3)
Other	109 (35.4)	71 (65.1)	31 (28.4)	7 (6.4)
Time since diagnosis (in years), M (SD)*	19.2 (9.7)	18.3 (9.0)	19.8 (11.3)	23.6 (9.2)
Cancer treatment received: surgery
Yes	214 (70.9)	142 (66.4)	51 (23.8)	21 (9.8)
No	88 (29.1)	59 (67.1)	23 (26.1)	6 (6.8)
Cancer treatment received: radiation
Yes	182 (59.7)	117 (64.3)	46 (25.3)	19 (10.4)
No	123 (40.3)	84 (68.3)	31 (25.2)	8 (6.5)
Cancer treatment received: chemotherapy
Yes	232 (76.1)	152 (65.5)	60 (25.9)	20 (8.6)
No	73 (23.9)	51 (69.9)	16 (21.9)	6 (8.2)
Primary care physician office visit in past year
Yes	207 (67.2)	134 (64.7)	55 (26.6)	18 (8.7)
No	101 (32.8)	70 (69.3)	22 (21.8)	9 (8.9)
Oncologist office visit in past year
Yes	122 (39.6)	80 (65.6)	33 (27.1)	9 (7.4)
No	186 (60.4)	124 (66.7)	44 (23.7)	18 (9.7)
Primary care physician or oncologist office visit in past year
Yes	239 (77.6)	158 (66.1)	62 (25.9)	19 (8.0)
No	69 (22.4)	46 (66.7)	15 (21.7)	8 (11.6)
Self-reported health status*
Excellent or very good	106 (34.4)	79 (74.5)	21 (19.8)	6 (5.7)
Good	126 (40.9)	80 (63.5)	30 (23.8)	16 (12.7)
Fair or poor	76 (24.7)	45 (59.2)	26 (34.2)	5 (6.6)
Risk perceptions (range = 1–28), M (SD)	15.4 (3.4)	15.3 (3.2)	15.5 (3.6)	15.1 (3.5)
Psychosocial factors
Cancer-related worry
Yes	43 (14.0)	29 (67.4)	9 (20.9)	5 (11.6)
No	265 (86.0)	175 (66.0)	68 (25.7)	22 (8.3)
Cancer-related distress (range = 0–21), M (SD)	4.3 (4.9)	4.3 (4.7)	3.6 (4.6)	5.8 (6.4)
Depressive thoughts
Yes	172 (56.0)	110 (64.0)	46 (26.7)	16 (9.3)
No	135 (44.0)	93 (68.9)	31 (23.0)	11 (8.2)
Computer ownership*
Yes	283 (91.9)	193 (68.2)	69 (24.4)	21 (7.4)
No	25 (8.1)	11 (44.0)	8 (32.0)	6 (24.0)

^aRefers to those who have Internet access at home and/or at work

^bAcross variables, sample sizes may differ slightly due to missing data

^*Indicates significant difference between groups at the p = 0.10 level or less

Results

Overall, PFH-2 participants reported substantial Internet access and use. Eighty percent owned a computer at home, and 79% reported using a computer on a daily or weekly basis. Over half of survivors reported going online (56%) and checking e-mail (51%) every day, and more than three-quarters reported going online (77%) and checking e-mail (77%) at least weekly. In addition, a majority (83%) had Internet access at home and/or at work (data not shown).

Of those who had access at one or both locations (n = 308), 91% reported going online at least weekly (Table 1). In contrast, of participants without access at either location (n = 64), most (81%) reported rarely or never going online (data not shown). Therefore, we restricted subsequent Internet use analyses to those who had access at home and/or at work.

Daily Internet users tended to be younger, better educated, and male (Tables 1, 2). In bivariate analyses, several clinical and technological characteristics were associated with daily Internet use (e.g., being closer to diagnosis, reporting better health, owning a computer); however, these associations were attenuated in multivariable analyses. Neither risk perceptions nor psychosocial factors were associated with the frequency of Internet use.

Table 2.

Multivariable model predicting daily Internet use among PFH-2 childhood and young adult cancer survivors who smoke and have Internet access^a, 2005–2008 (n = 308)

Variableb	OR (95% CI)c	p value
Age (continuous)	0.94 (0.91–0.97)	<0.001
Education		0.005
Less than high school	0.22 (0.07–0.64)
High school	0.33 (0.16–0.66)
Some college or vocational school	0.57 (0.30–1.11)
College and above	1.00
Gender		0.02
Male	1.88 (1.12–3.14)
Female	1.00
Computer ownership		0.07
Yes	2.28 (0.94–5.52)
No	1.00

^aRefers to those who have Internet access at home and/or at work

^bFactors found to be significant at the 0.10 level in bivariate analyses were included in multivariable modeling: age, education, gender, race/ethnicity, time since diagnosis, self-reported health status, and computer ownership. Using these variables, a parsimonious model was created. The model adjusted for recruitment site

^cDerived using binary logistic regression where the outcome variable was coded as 0 non-daily Internet use and 1 daily Internet use

Discussion

Although overall levels of Internet access and use were high among PFH-2 participants, certain subgroups reported less engagement. Older, less-educated, and female survivors reported significantly less frequent Internet use than their younger, better-educated, and male counterparts. These characteristics also were associated with having no Internet access at home and/or at work (analyses not shown here). Differences in Internet use were particularly strong in the case of age and education—a pattern that has been documented in the general population [16, 17], including among smokers [18, 19]. Studies using nationally representative HINTS data have found that older and less-educated smokers are more likely to be non-Internet users, as are those with lower incomes and those who are unemployed [18, 19].

gf The current study’s results are consistent with recent research showing that the digital divide exists on multiple levels. As the Internet began to diffuse across society, researchers became concerned about its unequal distribution among population subgroups. The digital divide, therefore, initially referred to differences in access to the Internet [20]. Yet in recent years, as Internet access has become more widespread, researchers have shifted their focus to differences in Internet use and engagement—and have described these differences as the “second-level digital divide” [20, 21]. Studies have shown that although gaps in physical access to the Internet have narrowed, important differences in Internet use and engagement remain across age and socioeconomic status (SES) groups [17, 22, 23]. Ultimately, then, the current study’s findings not only contribute to the growing body of research on the second-level digital divide, but they also have important implications for web-based prevention interventions. Specifically, for a web-based intervention to be successful, subjects need to actively participate and engage in the intervention; survivors who report less frequent Internet use might be less likely to engage in an Internet-based intervention, even if physical access is not a barrier. This implication is particularly relevant to childhood and young adult cancer survivors, who may be hard to reach through location-based cessation interventions. Not only are these survivors geographically dispersed, but also many must transition from a pediatric oncology setting to an adult health care system, which may have less cancer survivorship expertise. Younger survivors have grown up in an era of new media and communication technology, and thus there may be greater exposure opportunities for web-based interventions; however, these opportunities might not translate into similar levels of intervention engagement across survivor subgroups.

Several study limitations should be noted. First, PFH-2 may underestimate current Internet use, as data were collected between 2005 and 2008. That said, although overall use may be higher today, we might expect similar patterns of differential use, given evidence of persistent differences in Internet use and engagement across population subgroups [17, 21–23]. Second, the survey only asked about Internet access and use via computers. Future research should consider the role that smartphones and tablets may play in web-based prevention interventions—given the popularity of these new technologies, as well as concerns about potential inequalities in their access and use [24, 25]. Third, data on other SES indicators such as income were not collected; research has shown that income is an important predictor of Internet use [26]. Fourth, the sample was predominantly White and well-educated. This limits generalizability to other racial/ethnic groups with lower education levels; however, the distribution reflects the populations of the participating cancer centers. Fifth, although the study had a response rate of 48%, we were unable to collect non-responder data due to HIPAA restrictions. Thus, generalizability to the larger population of childhood and young adult cancer survivors might be constrained.

Internet-based interventions have the potential to reach childhood and young adult cancer survivors, but given differences in Internet use and engagement across population subgroups, some survivors might be reached more effectively via other strategies. For example, an ongoing study is examining whether a telephone-based quitline intervention will be successful in reaching geographically dispersed survivors [27]; if effective, this approach might be particularly appropriate for older and less-educated survivors. Intervention approach should be informed by patterns of media use and other considerations, including whether treatment effects might vary by communication modality and whether survivors might differentially respond to traditional versus new media modalities (e.g., telephone- versus Internet-based approaches). Moreover, ascertaining how to intervene with this high-risk population has implications beyond tobacco cessation, as there is a growing need for interventions that target health-promoting behaviors (e.g., healthy diet, physical activity) among survivors [28].