Cigarette smoking poses a serious risk to health. Despite public health campaigns, increased taxes on cigarettes and cautionary labelling, cigarette smoking remains the leading cause of preventable death in Canada (Public Health Agency of Canada, 2016). A report by the Centers for Disease Control and Prevention (CDC) states that 480,000 deaths per year in the USA are caused by cigarettes (CDC, 2017). The report further states that, on average, smokers typically die an entire decade earlier than non-smokers; however, non-smokers do not fully escape the health effects of cigarettes. Indeed, 41,000 of these deaths are caused by second hand smoke (CDC, 2017). Thus, as the effects of cigarette smoking are widespread, posing a serious risk to the health of both smokers and non-smokers alike, these effects must be addressed.
Difficulty in Quitting
The majority of smokers want to quit, with approximately one in every two smokers reporting a previous quit attempt in the past year or the intention to quit in the next 6 months (CDC, 2017; Finney Rutten et al., Reference Finney Rutten, Blake, Agunwamba, Grana, Wilson, Ebbert, Okamoto and Leischow2015). Despite the high motivation to quit, few attempts are successful (Beard, West, Michie, and Brown, Reference Beard, West, Michie and Brown2016). In a report investigating short-term quit success (smokers who had at least one quit attempt within the last year), only 13% of respondents remained abstinent by the time they were surveyed (Reid, Hammond, Rynard, Madill, and Burkhalter, Reference Reid, Hammond, Rynard, Madill and Burkhalter2017). On average, a smoker will attempt to quit 30 times before they are successful (Chaiton et al., Reference Chaiton, Diemert, Cohen, Bondy, Selby, Philipneri and Schwartz2016). Further, one of every ten smokers report an aborted attempt (Borland, Partos, Yong, Cummings, and Hyland, Reference Borland, Partos, Yong, Cummings and Hyland2012). Thus, despite a desire to quit smoking, most smokers who plan to quit are unsuccessful in this endeavour. This indicates a need for greater support and more efficacious smoking cessation aids.
Smoking Cessation Aids
There are a wide variety of smoking cessation aids available to help individuals in their cessation attempts. However, they are underutilized, with less than a quarter of individuals who have had a quit attempt in the past year employing a nicotine replacement therapy (NRT), a pharmacological therapy such as bupropion hydrochloride (i.e., Wellbutrin or Zyban), or behavioural therapy (Cokkinides, Ward, Jemal and Thun, Reference Cokkinides, Ward, Jemal and Thun2005). Further, it has been demonstrated that although NRTs do offer benefits in a smoking cessation attempt, their effects are modest. For example, smoking cessation rates without the use of NRTs are 3–5% and may only be increased by 2–3% with the use of NRTs (Hartmann-Boyce, Chepkin, Ye, Bullen, and Lancaster, Reference Hartmann-Boyce, Chepkin, Ye, Bullen and Lancaster2018). Indeed, despite the efficacy of NRTs in increasing the likelihood of a successful quit attempt, there remains a need for a smoking cessation aid with more robust effects which encourages more widespread use during quit attempts.
The ubiquity of mobile phones presents an opportunity to provide accessible smoking cessation aids to the population at large. Cessation programs delivered through mobile phones have shown promise (Whittaker, McRobbie, Bullen, Rodgers, and Gu, Reference Whittaker, McRobbie, Bullen, Rodgers and Gu2016). For example, brief SMS text messages have proven to be effective in increasing the likelihood of a successful quit attempt (Spohr et al., Reference Spohr, Nandy, Gandhiraj, Vemulapalli, Anne and Walters2015). Yet, smoking cessation applications (apps) are used considerably more than both smoking quitlines and text messaging programs in the USA; as of March 2014, there were 3.2 million smoking cessation app downloads compared with approximately 1 million enrolments to quitlines and 140,000 subscriptions to text-message programs (Bricker et al., Reference Bricker, Mull, Kientz, Vilardaga, Mercer, Akioka and Heffner2014). Simultaneously, these smartphone apps offer more advanced tools to the user (Hoeppner et al., Reference Hoeppner, Hoeppner, Seaboyer, Schick, Wu, Bergman and Kelly2016). Smartphone apps are useful tools in smoking cessation as they are easily accessible, visually engaging and simplify progress tracking during a quit attempt (Bricker et al., Reference Bricker, Mull, Kientz, Vilardaga, Mercer, Akioka and Heffner2014).
Despite the promise of smartphone apps as a smoking cessation tool, problems exist with the currently available apps designed for smoking cessation. Most apps are not proactive; they require the user to interact with it, often by entering smoking events into the app. The passive nature of these smoking cessation apps leads to inactivity by the user. In a recent study, less than one-quarter of individuals entered their smoking data each week (Iacoviello et al., Reference Iacoviello, Steinerman, Klein, Silver, Berger, Luo and Schork2017); this inactivity presents an issue for the validity of tracking measures included in the apps. However, it has been shown that apps which proactively send messages to the user are more popular and may be more helpful in a quit attempt (Hoeppner et al., Reference Hoeppner, Hoeppner, Seaboyer, Schick, Wu, Bergman and Kelly2016). As such, it seems an app which automatically records the number of cigarettes smoked and sends messages to the user would be more useful in quit attempts than an app that depends on the user to interact with it.
What is SmokeBeat?
SmokeBeat is a smoking cessation app developed to address the shortcomings of currently available smoking cessation apps. SmokeBeat is used in conjunction with a smartwatch and smartphone, giving it the functionality to recognize and independently record smoking events without the user directly entering them. When the app detects smoking behaviour, it immediately prompts the individual on their mobile device enquiring whether the individual has begun smoking. Furthermore, the app learns and becomes more accurate the more the smoker uses it, with the ability to measure several aspects of smoking behaviour such as smoking duration and number of puffs taken. It can also learn individual smoking patterns to try to predict when smoking episodes are likely to occur for an individual. It can then generate alerts for the smoker to potentially stop a smoking episode before it happens.
Current Study
If the SmokeBeat app can accurately detect and record smoking events while proactively engaging the user, it could overcome the limitations of currently available smoking cessation apps, providing an effective aid in smoking cessation. Therefore, the purpose of this feasibility study was to determine SmokeBeat's sensitivity, its ability to identify true smoking events, and specificity, its ability to discriminate between smoking and other hand motions. As well, we investigated the participants’ preferences regarding apps as smoking cessation aids. It was hypothesized that (a) the detection rate and accuracy of SmokeBeat would increase significantly between the first and second sessions, and (b) participants would prefer using the SmokeBeat app compared with a manual entry app.
Method
Participants
Participants (N = 20, 75% male) were recruited from the Halifax Regional Municipality in Nova Scotia via advertisements in online and community bulletin boards. All participants were required to be daily, dependent smokers, determined by scoring 3 or higher on the Fagerström Test for Nicotine Dependence (FTND; Heatherton, Kozlowski, Frecker, and Fagerström, Reference Heatherton, Kozlowski, Frecker and Fagerström1991). Please see Table 1 for demographic information. During a pre-screening telephone interview, it was confirmed that participants had been smokers for at least the past year, were medically healthy, had no plans of quitting smoking and were not currently using any nicotine replacement therapies. Individuals were excluded from participation if they had a current diagnosis of a serious medical illness, were currently engaged in a quit attempt or planned to quit in the next 30 days. All participants provided written consent to participate. Participants were compensated at a rate of $10 CAD per hour spent in the laboratory, and $10 CAD per day that they used the app outside of the laboratory. The study received ethical approval from the Dalhousie Research Ethics Board.
Demographics information for participants
Note: The final three variables are taken from the Smart Technology Preference Scale. Items were rated from 1 (strongly disagree) to 5 (strongly agree) during the initial session.
Materials
Cigarettes
Participants self-administered their preferred brand of cigarette during each study session.
Demographics and Smoking Patterns Questionnaire
A demographic and smoking history questionnaire was used to collect demographic (e.g., age, sex, ethnicity, marital status, education, occupation) and smoking history (e.g., age of first cigarette) information.
Smart Technologies Preference Scale
The Smart Technologies Preference Scale (STPS) is a nine-item author-compiled questionnaire, developed for this study to examine an individual's level of experience with technology as a smoking cessation aid, as well as their preference for the different types of cessation aids (i.e., interactive and manual).
SmokeBeat App
SmokeBeat (Somatix Inc, 2016) is a newly developed Smartphone app which is used in conjunction with a smartwatch. The smartwatch is used to actively perceive the motions associated with smoking and send timely notifications and messages of motivation to the user through the smartphone. When a potential smoking episode is detected, users are asked to confirm whether they are smoking or not. If the user selects yes, this is noted as a correct smoking detection; if the user selects no, this is noted as a false positive. If the user has to manually input a smoking episode (as it was not detected by the app), this is noted as a manual entry. Connection between the smartwatch and smartphone is also monitored by the app – if there is no signal coming from the smartwatch, this is recorded as no communication. Manual entries during a time of no communication between the devices are noted separately from manual entries with communication between the devices.
Smoke Free App
Smoke Free is a smoking cessation app used on a smartphone, in which participants manually enter their smoking instances. Smoke Free was used as a comparison for the likability and usability of SmokeBeat versus a manual entry app.
Smartwatches
Polar M600 (Polar Electro Canada, Inc., Lachine, QC, Canada) and LG Watch R W110 (LG Electronics, Yeoui-daero, Yeongdeungpo-gu, Seoul, Korea) smartwatches were used as the platform for SmokeBeat to detect smoking behaviour.
Design
The current study used a within-subject design. SmokeBeat's sensitivity was assessed by measuring the app's cigarette detection rate, both at the initial study session and at a follow-up session 2 weeks later. These two timepoints were used as the app's sensitivity was purported to increase with use. SmokeBeat's specificity was assessed by examining the number of false positives throughout the study.
Procedure
Participants attended two 1 h study sessions, spaced 2 weeks apart. When participants arrived at the laboratory for their first study session, they were provided with a smartwatch to be used for the following 2 weeks. If participants did not have a compatible Android phone, they were provided one to use for the duration of the study. Participants were then instructed how to use both the SmokeBeat and Smoke Free apps. Participants were expected to wear the smartwatch and respond to prompts from the SmokeBeat app, asking whether the detected smoking instance was correct. As well, participants were instructed to manually enter each smoking instance in the Smoke Free app.
Following these explanations, participants were randomly assigned to either a structured or unstructured initial session. This was done to determine whether the app would learn better with a more structured approach when compared with a more random approach; in the end, there was no difference between the two methods. In the structured session, participants completed the demographics and smoking questionnaire, the STPS, and smoked two cigarettes of their preferred brand, in that order. In the unstructured session, participants completed questionnaires and smoked the two cigarettes in the whichever order they chose, with a single caveat: they could not smoke and fill out questionnaires concurrently. Eleven participants completed an unstructured first session, and all participants completed an unstructured second session.
For the 2 weeks following the participants’ initial study session, participants were instructed to wear the smartwatch and use both the SmokeBeat and Smoke Free apps. After these 2 weeks, participants returned for their second study session. In this second session, participants completed the STPS and smoked two cigarettes of their preferred brand in the order of their choice. At this point, participants returned the smartwatches and phones, received their compensation, and were debriefed.
Results
Participant App Preferences
Participants’ preferences regarding the use of smart phone apps as smoking cessation aids were analysed using the STPS. When responding to whether apps are effective at helping with quitting smoking, a Wilcoxon signed-ranks test indicated that responses at the second session were significantly higher than at the first session (Z = −2.00, P = 0.046). Also, when responding to whether they found apps as a cessation aid effective, a Wilcoxon signed-ranks test indicated that responses at the second session were significantly higher than at the first session (Z = −2.01, P = 0.044). This indicates that more participants thought that apps are effective as cessation aids after using the app themselves.
Wilcoxon signed-ranks tests were also conducted to compare participants’ thoughts regarding interactive apps compared with non-interactive (manual input) apps, both prior to using the app and after (first and second sessions). The test was significant at both session 1 (Z = −3.13, P = 0.002) and session 2 (Z = −2.65, P = 0.008). This indicates that participants preferred the idea of interactive apps over passive apps, both before using the app and after.
Connection Between Smartwatch and Smartphone
The wireless connection between the smartwatches and smartphones used in the study was an issue throughout the study. This was out of the control of the app developer and the study investigators and may have been due to the smartwatch operating system. The operating system went through many changes and updates throughout the study and, as of writing, the operating system has gone through another revamp. As such, connection issues had to be considered when analysing the app.
Main Analyses
During the 2-week portion of the study when participants went about their daily lives, the sensitivity of the SmokeBeat app varied between users. The intermittent periods of connectivity between the devices led to many manual entries. As such, we focused our analysis for the 2-week period on the times when there was communication between the smartwatch and smartphone. When comparing the cigarette detection rate across all participants during this period of time, the true-positive rate was 64.9% (670 true positives, 363 false negatives).
As the sensitivity of the app was supposed to increase with usage, SmokeBeat's sensitivity in cigarette detection was investigated at sessions 1 and 2. The true-positive rate was 22.5% during session 1, and 41.7% during session 2. When analysing only the instances when there was a connection between the smartwatch and smartphone, the true positive rate for the first session was 34.6%; for session two, the true-positive rate was 100%. This shows that there was a significant increase in cigarette detection rates from session 1 to 2 when there was a connection between the devices (P <0.001, Fisher's exact test). This information is summarized in Table 2. A Mann–Whitney U test indicated that there was no significant difference in cigarette detection between the structured and unstructured sessions (U = 43.00, P = 0.603).
Connection and detection rates for each session
Note: ‘Connection rate’ refers to the percentage of smoking events during which there was a stable connection between the smartphone and smartwatch. The ‘true-positive rate’ is the detection rate of a cigarette being smoked, regardless of whether there was a connection between the smartphone and smartwatch. The ‘true-positive rate when connected’ refers to whether the act of smoking a cigarette was detected when there was a connection between the smartphone and smartwatch.
With regards to specificity, discrimination between hand movements related to smoking and other hand movements, the app appeared to perform well. Out of a total of 670 cigarettes detected by the app (when the smartwatch and smartphone were connected), there were 48 false positives (7.1%). One participant had an irregular pattern regarding false positives: each time they selected ‘no’ to indicate a false positive, it was immediately followed by a manually entered smoking event. This suggest that this participant may have systematically incorrectly selected ‘no’ when asked if they were smoking. When this participant's false positives were removed from the analysis, the false detection rate went down to 5.0%.
Discussion
The aim of this study was to determine SmokeBeat's sensitivity and specificity with regards to cigarette detection, as well as to investigate participants’ preferences related to smoking cessation apps. It was hypothesized that the detection rate and accuracy of SmokeBeat would increase significantly between the first and second sessions. When accounting for the connectivity issues between the smartphone and smartwatch, this hypothesis was confirmed. Our second hypothesis stated that participants would prefer using the proactive SmokeBeat app when compared with a more passive app. This was also confirmed, as participants rated interactive smoking cessation apps as more efficacious and preferred them to non-interactive apps; however, there was a decrease from in the average score from session 1 to 2. One possible explanation for this is that, prior to usage, participants may have had more optimism regarding the technology in general; another possible explanation for this may be that the connection issues may have dampened the participants’ enthusiasm for the technology.
Our results regarding SmokeBeat's detection accuracy and validity were similar to those found by another study. The author found the detection rate of the app to be between 80% and 90% by the end of the study (Dar, Reference Dar2017). This substantiates the claim of the app developer that the detection rates would be around 90% by the end of our study. The study by Dar (Reference Dar2017) also found similarly low false detection rates, around 3% in total. Although, false detections initially appeared to be greater than in previous studies, once these data were accounted for, false detection rates were comparable to previous studies.
Participants in our study had favourable views of smartphone apps as smoking cessation aids, particularly after using the apps we provided them. Specifically, participants preferred using the interactive app when compared with the passive app. These findings are aligned with the findings of Hoeppner et al. (Reference Hoeppner, Hoeppner, Seaboyer, Schick, Wu, Bergman and Kelly2016). Hoeppner et al. found that participants preferred to use apps that were less like tools and more like coaches. The authors go on to say that the more passive smoking cessation apps are likely only helpful to smokers who are highly motivated to quit (Hoeppner et al. Reference Hoeppner, Hoeppner, Seaboyer, Schick, Wu, Bergman and Kelly2016). Interactive apps can assist both smokers who are highly motivated and those who are more hesitant or indecisive.
Proactive smoking cessation apps may become an important aid going forward, as technology becomes more ingrained in everyday life. Individuals may prefer the anonymity that they are afforded when downloading an app on their smartphone as opposed to using other means for smoking cessation. Other studies have shown that some smoking cessation aids are associated with lower odds of abstinence, likely due to inappropriate usage and low adherence in the real world (Balmford, Borland, Hammond, and Cummings, Reference Balmford, Borland, Hammond and Cummings2011; Kotz, Brown, and West, Reference Kotz, Brown and West2014); simply wearing a smartwatch and carrying a phone may prove to be more accessible than following different therapies. Regardless of the reasons for using an app instead of other cessation aids, having an accessible and interactive option for smoking cessation should have a positive impact on public health, given that smoking cessation aids significantly increase success rates of a quit attempt (Gross et al., Reference Gross, Brose, Schumann, Ulbricht, Meyer, Völzke, Rumpf and John2008).
This study was not without limitations. First and foremost, there were connection issues between the smartwatches and smartphones. This led to data not being collected by the smartphones, and the cigarettes would then have to be entered manually. Second, many users experienced issues regarding battery drain on the smartwatches; if the battery on the smartwatch was dead, no data could be collected. Third, there may have been issues with compliance. Participants had to ensure that the smartphone and smartwatch were charged, and that the smartwatch was on their dominant smoking hand. When analysing the data, it was noted that some participants had long periods during the day when the smartwatch was not being worn, which may have influenced the results.
Further, there were limitations due to the nature of the participant sample. There were only 20 participants in this feasibility study, with the majority reporting that they were frequent users of similar technology or at least comfortable with the technology. This may not be representative of all smokers; infrequent users of such technology may have greater difficulty using this app as a smoking cessation tool. Another limitation of the study was that participants had to wear the smartwatch on their smoking hand (i.e., dominant hand). This may have had an impact on the participants’ behaviour, as they may not typically wear a watch in general, particularly on their dominant hand.
As SmokeBeat has been shown to have good sensitivity and specificity in cigarette detection after a period of behaviour learning, future endeavours should focus on recruiting participants who are motivated to quit smoking. Future studies should examine the efficacy of the app as a smoking cessation aid, as this was not an area that we explored in the current study. The efficacy of SmokeBeat as a smoking cessation aid could be compared with non-interactive apps and other traditional smoking cessation methods. In conclusion, when the app was used correctly and there were no connection issues between the smartphone and smartwatch, SmokeBeat performed as the developers claimed.
Financial support
This work was supported by the Lung Association of Nova Scotia Legacy Grant.
Conflict of interest
There were no financial conflicts of interest. However, Somatix (SmokeBeat's developers) provided three smartwatches to be used throughout this study.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
