Determinants of Successful Mobile Smoking Cessation Interventions: Cross-Sectional Study.

Determinants of Successful Mobile Smoking Cessation Interventions: Cross-Sectional Study

Abstract

Background

In recent years, digital interventions for smoking cessation have seen increasing use, but program completion rates vary widely. Therefore, understanding the factors associated with program completion in cessation programs is critical for evaluating the feasibility and acceptability of the digital intervention program.

Objective

This study aims to identify the predictors (sex, age group, and health) of program completion in a cognitive behavioral therapy–based mobile smoking cessation program.

Methods

Participants enrolled in Adiquit, a mobile app–based smoking cessation program, were classified as completers if they had entered the Withdrawal-Relapse phase and had reported no lapse for 15 consecutive days or had finished all program sessions. User data for this study covered the period from launch through July 2024. Users who did not meet the completion criteria by the end of the study were classified as noncompleters. Predictors were assessed during the Entry and Equilibrium phases, including demographic characteristics, psychosocial factors, and early engagement metrics. Logistic regression was used to identify factors associated with completion.

Results

The three strongest predictors of program completion, as identified by the multivariate model, are higher early engagement (session attendance and screen time), stronger perceived social support, and positive prequitting feelings.

Conclusions

The results of this study will help clarify the variables that lead to successful completion of such a digital intervention program, as well as their importance by sex, age group, health status, and employment status.

Introduction

Tobacco consumption remains a major global health hazard [1,2]. The smoke in a cigarette contains more than 7000 harmful compounds, causing respiratory, cardiovascular, and cancer diseases [3]. According to the World Consumption of Cigarettes 1880-2020 report, around 5.2 trillion cigarettes were consumed worldwide in 2020, with China and Europe accounting for the highest consumption [4]. Despite global prevalence programs, reports from the World Health Organization showed only modest declines: 24.4% in 2015, 22.3% in 2020, and a projected 20.4% in 2025 [5]. This has led to a rising concern regarding the inclusion and promotion of various measures for smoking cessation, as well as a need to find an underlying cause to help in decreasing this rate. Mobile-based digital intervention programs have gained prominence due to affordability, accessibility, and independence from direct involvement of health care professionals [6,7]. Rooted in cognitive behavioral therapy (CBT) [8], these apps prepare users for quitting and support abstinence by including gradual behavioral changes [9]. A few examples of such apps include Tobbstop (Universitat Rovira i Virgili) [10], iCanQuit (Fred Hutchinson Cancer Center) [11], QuitGuide (National Cancer Institute) [12], QuitAction (despDev) [13], and SmokeFreeBuddy (23 Ltd) [7]. According to a meta-analysis [14], which included 19 controlled trials, intervention therapy was found to be more effective for smoking cessation than self-help guidelines and no intervention program. While the success rate of quitting without any support is only 5% to 7%, incorporating an app‑based intervention program may increase this rate up to 20% [15].

The process of smoking cessation is complex, and the success of an intervention program depends on multiple interrelated factors, including sociodemographic characteristics, smoking history, underlying health conditions, and financial or employment status [16,17]. Age often shapes these dynamics: older smokers show higher levels of dependence and comorbidities but lower engagement with digital interventions [18]. Sex differences also emerge: women achieve higher cessation rates in community settings [19], whereas men respond better to gamification and engagement metrics [20,21]. Women often use apps for emotional regulation, whereas men use them for instrumental purposes [22]. Self-efficacy strongly predicts success [23], whereas nicotine dependence, hypertension, and cardiovascular disease are associated with quitting but not with completion of the program [24]. Employment status is also important, as occupational stress and workplace policies affect smoking behavior [25,26]. The literature shows inconsistent correlations between these factors and program completion [16,17,27]. The selection of sex, age group, health, and employment status as stratification variables is grounded in three considerations: (1) prior evidence consistently identifies these as moderators of cessation outcomes in both clinical and digital contexts [6,19,26]; (2) each factor is linked to distinct psychological, behavioral, and social mechanisms relevant to quitting—for example, chronic illness may heighten health motivation, while economic status can influence daily routine and stress; and (3) these variables were available for most participants, enabling adequately powered stratified analyses.

The design of digital intervention programs also plays a major role in shaping effectiveness [21]. Zhang et al [21] identified 4 critical factors: external support, personalization, active tracking of user progression, and effective management of withdrawal symptoms [21]. Yet attrition remains high due to poor engagement [28]. Apps that use gamification, rewards, feedback, and frequent interventions reduce dropout rates [6,28]. The digital intervention app Adiquit, used in this study, was designed using established behavioral principles and is therefore expected to be effective in supporting smoking cessation. This study addresses three prespecified research questions: RQ1: Which variables measured within a mobile-based smoking cessation app (Adiquit) are significantly associated with program completion? RQ2: Do the identified predictors of program completion differ across clinically relevant subgroups defined by sex, age group, health status, and employment status? RQ3: Do app-derived behavioral and engagement variables (eg, number of sessions completed, screen time, and social support) demonstrate stronger associations with program completion? Verified abstinence was not included within the scope of this study, and program completion was the primary outcome. The findings were contextualized against a targeted review of comparable app-based cessation studies, with the goal of assessing the generalizability of identified predictors across digital intervention platforms.

Methods

Details of the Cessation Intervention Program and App Description

The overall therapy process in the Adiquit app is based on the principles of CBT [29] and is structured as an intensive, multiphase process aimed at guiding people through their smoking cessation journey (Figure 1). The app is designed in accordance with evidence-based principles for digital intervention design, including external support, personalization, active engagement tracking, and withdrawal symptom management. The program starts with the Envisage and Explore (EE) phase, which lasts about 10 sessions and focuses on psychoeducation, the rationale for CBT, motivation evaluation, and personalization of the quit date, with baseline data collected but no social support components. This is followed by the Envisage Quitting (EQ) phase, a critical single session in which the user makes the final decision to quit and identifies their social support network. The Follow-up (FU) phase lasts 21 sessions and focuses on cognitive restructuring, recognizing high-risk situations, increasing self-efficacy, behavioral activation, craving management, and lapse-prevention strategies, which are reinforced through gamified daily check-ins and in-app reminders to engage the support network. Finally, the Withdrawal-Relapse (WR) phase, which spans more than 70 sessions, focuses on abstinence maintenance, relapse prevention, progress tracking, and reward milestones, with features for sharing milestones and keeping participants motivated. As with other CBT-based digital cessation tools [30], its theoretical basis suggests potential utility for smoking cessation; however, no prior independent efficacy studies, especially for Adiquit, are available.

User Information and Data Collection

The app offers a free version that provides access for the first few sessions, and users who opt for a paid subscription (€19.99 per month, where €1=US $1.08, as of July 31, 2024) can access subsequent sessions. Only users who subscribed to the paid tier were included (N=1311 at baseline), as free-tier users were restricted to introductory sessions and lacked exposure to the core intervention components required to assess the study’s primary predictors. According to the available information as of July 2024, 1311 users have opted for the paid service and are considered to be in the initial stage. However, after an initial screening based on the availability of basic user information and on inclusion criteria for the region (the Czech Republic) and age (14‐70 years), only 692 users were included in this study. Ethical approval for the study design and data collection was obtained from the Committee for Ethics in Research of the Scientific Council of the Czech Technical University in Prague. Informed consent for the use of anonymized data for research purposes was obtained electronically from all users during app registration. A detailed description of the approval and consent procedures, privacy safeguards, and compensation is provided in the Ethical Considerations subsection. Basic information, including age, sex, current health conditions, and previous smoking history, was collected at the initial sessions. User responses to several questionnaires (Multimedia Appendix 1) assessing their feelings and perceptions toward the therapy were collected during the therapy, including the preparatory phase. Information on a user’s therapy progress, including sessions and total time invested, was collected from the user log. Individuals who did not complete the programs served as the primary comparator group in this study, but missing outcomes were not included.

Measures

The information collected from the digital intervention app, including the user’s basic and sociodemographic information, health status, and smoking history, was processed based on whether each variable was binary, categorical, or ordinal. The intervention was delivered via a mobile app and was divided into 4 phases over 4 months, during which the user received a 3- to 10-minute therapeutic dialogue every day covering psychoeducation, CBT-based coping strategies, and withdrawal management. Sessions were asynchronous, with automated reminders sent daily. Personalization was limited to pacing and feedback based on user input. Behavioral information, including the presence of external or social support, the user’s motivation, and feelings toward the therapy, was collected during the initial phases and, in subsequent phases, through the user’s responses to questionnaires. The user’s log data were used to track their involvement in the cessation program by measuring the number of completed sessions and the time spent in the app (screen time). However, only the log information from the first 10 days was considered, as it has been reported that the user disengagement from any intervention program is highest during the first 10 days [31] and is considered an important predictor of program completion. Table 1 presents the list of parameters or measures collected from the app, categorized into 2 groups: digital intervention–based parameters and nondigital intervention–based parameters, including biological, sociodemographic, and information related to previous smoking history and health.

All predictors included in this analysis were measured prior to the onset of the WR phase, which defines the outcome window. Specifically, predictors were derived from the following app phases: (1) sociodemographic, health, and smoking history variables were collected at app registration (EE phase); (2) behavioral questionnaires assessing social support, prequitting feelings, self-efficacy, and motivation were administered during the EQ phase; and (3) log-based engagement metrics were derived from the first 10 days of app usage, which spans the EE and EQ phases. No predictors were measured concurrently with or following the WR phase, precluding reverse causation bias.

Table 1. List of data and measures obtained from the app-based digital intervention.

Group and parameter subgroup Description Information extracted

Nondigital intervention–based

 Sociodemographic Includes the basic information of all users Age, sex, educational status, and town size

 User smoking information Includes the parameters based on the previous smoking history and the type of smoking product used Age for smoking initiation, cigarettes per day, last nonsmoking period duration, and number of previous quitting attempts

 User health information Includes the parameters with the user’s previous health status and medication history Health status (healthy or presence of any chronic disease) and medication history (currently on long-term medication or not)

Digital intervention–based

 Behavioral information Includes all information depicting the user’s motivation and intention toward therapy, either at the initial stage or during the therapy progression Presence of social support, prequitting feelings, and questionnaires to measure self-efficacy, motivation, and user response to lapse and high-risk situations

 Log-based information Includes the measures derived from the log information The number of sessions and the screen time during the first 10 days

Outcome Definition

The primary outcome in this study is “program completion,” operationally defined as (1) completion of the WR phase or (2) active engagement in the WR phase with no self-reported lapse during the preceding 15 days. The outcome criterion was selected for the following reasons: (1) the 15-day lapse-free window aligns with 2-week point prevalence abstinence criteria widely used in cessation trials [32]; (2) entry into the WR phase requires successful completion of preparatory CBT modules, including cognitive restructuring, trigger identification, and quitting date commitment: all of which are theoretically necessary precursors to sustained abstinence [11]; and (3) data from Baumel et al [33] indicate that sustained engagement beyond the first 10 days of digital health intervention is strongly predictive of long-term behavior change outcomes. Here, program completion refers to completing all therapy sessions, whereas verified smoking cessation requires biochemical validation, which was not performed in this study. The proposed outcome in this study reflects persistent behavioral engagement and therapy adherence, which are considered proxies for program completion in digital intervention research [20,34].

Feasibility and Acceptability

Feasibility was defined as the extent to which users could progress through the program stages, as measured by attrition and completion rates. Acceptability was defined as the extent to which users engaged with the program, as inferred from early session completion (the first 10 days), reported social support, and overall program completion. No direct satisfaction surveys were conducted; therefore, acceptability was assessed indirectly through these engagement indicators.

Statistical Analysis

Descriptive Statistics

After preprocessing, the information collected from the cessation app is split into 2 groups: users who have not finished and users who have finished therapy. Descriptive statistics for the 2 groups were used to understand their distributions. Based on the nature of the collected information (continuous, binary, or ordinal), the choice of the statistical test (t test, χ2 test, or Spearman rank correlation test) was made. Prior to analysis, the completeness of each variable was assessed. Variables with less than 70% response availability were excluded from further analyses (n=4 questionnaire items). For the remaining variables, missing data were handled using complete-case analysis, justified by the observation that missingness was largely attributable to nonprogression through therapy phases (ie, not due to data collection failure), consistent with a missing-at-random assumption.

Univariate Logistic Regression

A univariate logistic regression (LR) model was used to assess the individual effect of each predictor on program completion, with each predictor as the independent variable and program completion status as the dependent variable. The results of the univariate analysis are presented in terms of the regression coefficient, SE, z-statistic, P-value, and odds ratio (OR, 95% CI). A predictor with a P<.05 in the univariate analysis was considered significant.

Multivariate LR

A multivariate analysis was also performed using all possible combinations of 5 variables from the group of predictors with P<.05 in the univariate analysis. Backward stepwise elimination, guided by the Bayesian information criterion, was used to derive the most stable model. Model stability was assessed via 200 bootstrap iterations, with the concordance C-statistic and calibration slope reported. Subsequently, the performance of the stratified multivariate model was assessed within subgroups defined by sex (male and female), age group (younger and older), health status (healthy and diseased), and employment status (employed and unemployed) to examine whether the predictors had group-specific effects on program completion.

Comparison With Previous Literature

A systematic search was conducted in PubMed and Scopus using the keywords: “smoking cessation” OR “Tobacco Cessation” AND (“mobile app” OR “smartphone” OR “digital intervention”) AND (“mobile app” OR “smartphone” OR “digital intervention”) AND (“predictors’’ OR “factors’’), limited to publications from 2020 to 2024. Among the documents obtained, 6 met the inclusion criteria, namely the use of multivariate analysis, app-derived variables, and cessation or program completion outcomes.

Ethical Considerations

This study followed the Declaration of Helsinki and Czech/European Union (EU) regulations on human research and data protection. Approval was obtained from the Ethics Committee of the Czech Technical University in Prague (approval number: 0000-13/25/51902/EKČVUT, June 2, 2025). Informed consent was obtained electronically during registration in the Adiquit app, with participants informed about the data categories, the voluntary nature of participation, and their withdrawal rights. The app is registered in the Czech Republic, where the legal age of majority is 18; however, adolescent smokers (aged 14‐17 years) were included with parental/guardian consent, consistent with the Czech Research Ethics Board’s guidelines for minor participation in noninterventional digital health research. Only anonymized records of consenting users were analyzed, with identifiers removed and data stored securely on access-controlled servers. Data processing complied with General Data Protection Regulation and Czech law, with no transfers outside the EU [35]. Participants’ rights were respected, and no financial incentives or compensation were provided for participation in this study. The analyzed data were generated as part of users’ voluntary use of the Adiquit app, which they accessed independently via a paid subscription.

Results

Descriptive Statistics of the User Data

The descriptive statistics for the data collected from the cessation app for the 2 population groups are presented in Table 2. A total of 692 (333 male and 359 female) individuals were included in this study. Neither male nor female individuals showed a significant difference between the 2 groups (Figure 2). Of all users, only 202 were older than 40 years. The users were divided into 3 categories based on the number of cigarettes smoked per day: light, medium, and heavy smokers. The P-value from the χ2 test was .06, indicating that the difference between the 2 groups is not statistically significant. Other factors, including the duration of the last nonsmoking period, educational background, health status, town size, and employment status, were also compared between the 2 population groups, but no statistically significant differences were observed. However, when compared based on the availability of social support, a significant difference was observed in the χ2 test (P<.001). According to the collected data, 312 users received external support during therapy, whereas 380 did not. Significant differences in the log information between the 2 population groups were observed in each case. The group of users who completed therapy usually had a higher mean number of sessions completed and greater screen time in the first 10 days, with P<.001.

Table 2.Descriptive statistics of user information collected from the app.

Parameters Not completed (N=361), n (%) Completed (N=331), n (%) P valuea

Age group         .60

 Younger (<40 years), (n=490) 252 (69.80) 238 (71.90)

 Older (>40 years), (n=202) 109 (30.19) 93 (28.09)

Sex     .52

 Male (n=333) 169 (46.8) 164 (49.55)

 Female (n=359) 192 (53.18) 167 (50.45)  

Smoking category     .06

 Light smoker (n=169) 82 (22.71) 87 (26.28)

 Medium smoker (n=158) 76 (21.05) 82 (24.77)  

 Heavy smoker (n=365) 203 (56.23) 162 (48.94)  

Smoking since age (years), mean (SD) 17.5 (5.02) 17.68 (4.73) .68

Last nonsmoking period durationb .67

 <1 week (n=259) 135 (37.39) 124 (37.47)

 1 week-1 month (n=158) 85 (23.54) 73 (22.05)  

 1‐3 months (n=115) 59 (16.34) 56 (16.91)  

 3‐6 months (n=47) 27 (7.47) 20 (6.04)  

 6‐12 months (n=48) 28 (7.75) 20 (6.04)  

 >1 year (n=65) 27 (7.47) 38 (11.48)  

Educational background          .53

 High school (n=436) 223 (61.77) 213 (64.35)

 University and higher (n=256) 138 (38.223) 118 (35.64)  

Town size      .78

 Urban (n=174) 93 (25.76) 81 (24.47)

 Town (n=253) 126 (34.90) 127 (38.36)  

 City (n=265) 142 (39.33) 123 (37.16)  

Social support        <.001

 Having social support (n=312) 109 (30.19) 203 (61.32)

 No social support (n=380) 252 (69.80) 128 (38.67)  

Health status          .47

 Healthy (n=464) 247 (68.42) 217 (65.55)

 Diseased (n=228) 114 (31.57) 114 (34.44)

Employment status     .23

 Unemployed (n=149) 71 (19.66) 78 (23.56)

 Employed (n=543) 290 (80.33) 253 (76.43)  

Adherence value

 Number of sessions in first 10 days, mean (SD) 8.73 (4.38) 12.53 (4.75) <.001

 Screen time in first 10 days (min), mean (SD) 38.35 (26.67) 55.79 (27.21) <.001

 Number of sessions during WTc, mean (SD) 16.36 (15.92) 26.35 (16.46) <.001

 Screen time during WT (min), mean (SD) 59.86 (52.22) 89.27 (47.13) <.001

aP values for all parameters except those under “Adherence value” were obtained using the χ2 test, and P values for variables under “Adherence value” were obtained using the independent t test.

bLight smoker: <10 cigarettes/day; medium smoker: 10‐20 cigarettes/day; heavy smoker: >20 cigarettes/day. Urban: <4000 inhabitants; town: ≥4000 inhabitants, but less than that of a city; city: >20,000 inhabitants.

cWT: whole therapy.

Predictors of Program Completion

Participation Flow and Attrition

Feasibility was demonstrated by the fact that 662 of 692 registered users progressed to the first engagement stage (EE) and that 187 completed all sessions. However, 331 users, including those who completed all sessions or who were in the WR phase for 15 days without any lag, were considered program completers. Attrition was highest at the EQ stage (173/692, 25%), indicating a critical point where additional support may be needed (Table 3). Acceptability was inferred from engagement metrics: 61.32% (203/331) of users who reported social support completed the program, and higher completion rates were observed among those with greater early session activity. These findings suggest that the intervention was feasible to deliver and acceptable to a subset of users, though direct measures of satisfaction were not collected.

Table 3.User attrition rates at different stages of the therapy phase (N=692).

Therapy stage Users remaininga, n Dropped out before this stageb, n (%)

Registered 692 —c

Reached EEd 662 30 (4.33)

Reached EQe 489 173 (25)

Reached FUf 430 59 (8.52)

Reached WRg 426 4 (0.57)

Finished 187 239 (34.53)

aUsers who reached at least that stage.

bUsers whose last recorded activity was in the previous stage (ie, they did not make it this far).

cNot applicable.

dEE: Envisage and Explore.

eEQ: Envisage Quitting.

fFU: Follow-Up.

gWR: Withdrawal Relapse.

Univariate Analysis

Univariate LR was used to assess the influence of individual predictors on program completion, and the results are presented in Table 4. Based on the results obtained, social support during therapy and prequitting feelings were among the most influential predictors, with P<.001. The participants’ prequitting feelings were assessed during the preparatory phase by asking whether they were afraid of failure during the current cessation attempt. In addition to these parameters, the reason for quitting (eg, job-related) was also found to influence program completion (P=.04), with a negative correlation. The log data (number of sessions and screen time) collected over the first 10 days were significant (P<.001) and showed a positive correlation with program completion.

Table 4.Influence of parameters on program completion.

Parameters and subgroup Regression coefficient SE z-statistic P value<|z| Odds ratio (95% CI)

Biological components

 Sociodemographic parameters

  Age –0.215 0.186 –1.153 .25 0.807 (–0.580 to 0.150)

  Age group (younger or older) –0.159 0.141 –1.125 .26 0.853 (–0.435 to 0.118)

  Sex –0.030 0.110 –0.274 .78 0.970 (–0.245 to 0.185)

  Educational background –0.157 0.125 –1.249 .21 0.855 (–0.402 to 0.089)

  Town size –0.038 0.034 –1.121 .26 0.963 (–0.103 to 0.028)

 Employment

  Full-time employee –0.144 0.092 –1.56 .12 0.866 (–0.324 to 0.037)

  Full-time student –0.204 0.171 –1.19 .23 0.816 (–0.539 to 0.132)

  Part-time student 0.337 0.338 0.995 .32 1.400 (–0.326 to 0.999)

  Unemployed –0.375 0.392 –0.957 .34 0.688 (–1.142 to 0.393)

  Sick leave –0.375 0.392 –0.957 .34 0.688 (–1.142 to 0.393)

  Part-time job 0.201 0.449 0.446 .66 1.222 (–0.680 to 1.082)

  Self-employed –0.123 0.286 –0.428 .67 0.885 (–0.684 to 0.438)

  Parental leave –0.182 0.428 –0.426 .67 0.833 (–1.022 to 0.657)

  Pensioner –0.182 0.428 –0.426 .67 0.833 (–1.022 to 0.657)

 Employment status

  Employed/unemployed –0.137 0.086 –1.587 .11 0.872 (–0.305 to 0.032)

 Smoking history

  Smoking since age regularly –0.230 0.253 –0.911 .36 0.794 (–0.726 to 0.032)

  Cigarettes per day –0.311 0.189 –1.641 .10 0.733 (–0.681 to 0.060)

  Smoking category –0.052 0.031 –1.657 .09 0.949 (–0.113 to 0.010)

  Last quitting attempts count 0.016 0.189 0.085 .93 1.016 (–0.355 to 0.387)

  Last nonsmoking period duration –0.015 0.025 –0.583 .56 0.985 (–0.065 to 0.035)

 Nicotine product last used

  PUPWa (cigarette) –0.127 0.103 –1.23 .22 0.881 (–0.328 to 0.075)

  Heated tobacco –0.134 0.518 –0.258 .79 0.875 (–1.148 to 0.881)

 Health and medication status

  Healthy –0.130 0.093 –1.392 .16 0.879 (–0.312 to 0.053)

  Cardiovascular disease 0.693 0.612 1.132 .26 2.000 (–0.507 to 1.893)

  Respiratory disease –0.170 0.261 –0.65 .52 0.844 (–0.682 to 0.342)

  Cancer 0.134 0.366 0.365 .72 1.143 (–0.584 to 0.851)

  Mental disorders –0.059 0.243 –0.243 .81 0.943 (–0.534 to 0.417)

  History of COVID-19 –0.123 0.128 –0.958 .34 0.885 (–0.374 to 0.128)

  Taking regular medication –0.108 0.093 –1.161 .25 0.898 (–0.291 to 0.074)

App-based intervention components

 Social support

  Having social support 0.622 0.119 5.237 <.001 1.862 (0.389 to 0.855)

 Initial feelings about the therapy

  Prequitting feelings 0.263 0.075 3.492 <.001 1.301 (0.115 to 0.411)

  Importance of quitting –0.038 0.021 –1.783 .08 0.963 (–0.079 to 0.004)

  Commitment to the quitting plan 0.001 0.026 0.026 .98 1.001 (–0.051 to 0.052)

 Last withdrawal method

  Without any help –0.099 0.097 –1.021 .31 0.905 (–0.290 to 0.091)

  Nicotine substitution –0.127 0.152 –0.836 .40 0.880 (–0.426 to 0.171)

  Medication 0.288 0.540 0.533 .59 1.333 (–0.771 to 1.346)

  Mobile app 0.085 0.238 0.356 .72 1.088 (–0.381 to 0.550)

 Reason for quitting

  Job –0.799 0.401 –1.989 .04 0.450 (–1.585 to –0.012)

  Health –0.074 0.078 –0.941 .35 0.929 (–0.228 to 0.080)

  Social pressure –0.052 0.086 –0.601 .55 0.950 (–0.220 to 0.117)

  Addiction free 0.042 0.109 0.381 .70 1.042 (–0.172 to 0.255)

  Finance –0.024 0.098 –0.244 .81 0.976 (–0.215 to 0.168)

  Family 0.037 0.192 0.192 .85 1.038 (–0.340 to 0.414)

 Log-based information

  Number of sessions (first 10 days) 0.674 0.187 3.597 <.001 1.962 (0.307 to 1.041)

  Screen time (first 10 days), (min) 0.806 0.244 3.298 .001 2.239 (0.327 to 1.285)

aPUPW: product used in past week.

Multivariate Analysis

Table 5 presents the results of the multivariate LR model based on 692 participants, demonstrating moderate explanatory power (pseudo-R²=0.193) and good discriminative ability (area under the receiver operating characteristic curve [AUROC]=0.772, SD 0.057), though calibration was imperfect (Hosmer-Lemeshow χ2 statistic [HL χ²8]=20.52; P=.009). Prequitting feelings (OR 2.447, 95% CI 1.738-3.443), early engagement through the number of sessions completed in the first 10 days (OR 1.158, 95% CI 1.107-1.211), and social support (OR 2.274, 95% CI 1.606-3.218) emerged as strong, highly significant predictors (P<.001), consistently retained in bootstrapping. In contrast, screen time during the first 10 days showed no significant effect (OR 1.000; P=.10). Collectively, these findings underscore the importance of motivation, early behavioral engagement, and supportive networks as reliable drivers of program success, while passive screen exposure was not a meaningful determinant.

Table 5.Result of the multivariate models chosen after using the above method.

Predictors Odds ratio (95% CI) P value Bootstrap inclusion (%)

Prequitting feelings 2.447 (1.738-3.443) <.001a 100

Sessions (first 10 days) 1.158 (1.107-1.211) <.001a 100

Social support 2.274 (1.606-3.218) <.001a 99

Screen time (first 10 days), min 1.000 (1.000-1.000) .10b 36

aStatistically significant.

bNot statistically significant.

Effects of Sex, Age Group, Health, and Employment Status

Table 6 represents the influence of the predictors, stratified by sex, age group, health status, and employment status. The stratified multivariate analysis across 4 demographic dimensions reveals a remarkably consistent pattern: sessions completed within the first 10 days and social support are the most cross-cutting predictors of therapy completion, with significance found in the majority of subgroups regardless of sex, age, or health status. The stratified results indicate that early involvement and psychological preparation are strong, population-level drivers of program completion, whereas demographic factors act as contextual moderators rather than independent predictors.

Table 6.Stratified multivariate logistic regression results based on subgroup.

Stratum Social support Prequitting feelings Sessions (first 10 days)

ORa (95% CI) P value OR (95% CI) P value OR (95% CI) P value

Sex

 Male (n=333) 3.39 (2.00‐5.77) <.001 2.48 (1.52‐4.06) <.001 1.20 (1.13‐1.28) <.001

 Female (n=359) 1.74 (1.09‐2.79) .02 2.51 (1.56‐4.03) <.001 1.16 (1.10‐1.23) <.001

Age group

 Younger (<40 y, n=490) 2.95 (1.94‐4.48) <.001 2.69 (1.78‐4.08) <.001 1.19 (1.13‐1.25) <.001

 Older (>40 y, n=202) 1.30 (0.68‐2.48) .42 2.24 (1.23‐4.08) .008 1.18 (1.10‐1.26) <.001

Health status

 Healthy (n=464) 3.08 (1.99‐4.76) <.001 2.10 (1.42‐3.12) <.001 1.18 (1.12‐1.24) <.001

 Diseased (n=228) 1.33 (0.72‐2.46) .36 4.80 (2.34‐9.84) <.001 1.18 (1.10‐1.27) <.001

Employment status

  Employed (n=543) 2.35 (1.58‐3.50) <.001 2.34 (1.58‐3.47) <.001 1.21 (1.15‐1.27) <.001

  Unemployed (n=149) 2.32 (1.08‐4.95) .03 3.45 (1.73‐6.88) <.001 1.09 (1.01‐1.18) .03

aOR: odds ratio.

Comparison With Previous Literature

Table 7 presents a targeted (nonsystematic) review of studies using app-based interventions for smoking cessation, published between 2020 and 2024, that share comparable design elements (digital interventions, LR or similar analyses, and predictors of app engagement). The comparison reveals consistent evidence that psychological and behavioral variables, as well as sociodemographic characteristics, have a major influence on app-based smoking cessation interventions. Previous research has shown that self-efficacy, motivation, self-affirmation, education, and usability are important predictors of quitting success or program completion. The new study using Adiquit builds on previous findings by including log-based variables, such as screen time and session frequency, as well as contextual factors, such as social support and reasons for quitting (eg, job-related). These findings imply that, in addition to demographics and smoking history, dynamic app engagement and psychological preparation are important factors in influencing cessation outcomes.

Table 7.Comparison of the findings of this study with previously reported literature.

Study Study population (N) App-based intervention used Subgroup of variables used Main findings

Rajani et al [36], 2020 116 Kwit and QuitGenius Sociodemographic, smoking history, self-efficacy, and motivation An increase in self-efficacy and motivation from the start to the mid-therapy is positively associated with quitting. No change in the age.

Seaman et al [37], 2021 647 Smoke Free app Sociodemographic, smoking history, user feelings, and self-affirmation Users with a greater tendency to continuous self-affirmation are more likely to quit.

Marler et al [34], 2021 391 Pivot program Sociodemographic, education, income, and baseline intention to quit Self-efficacy, positive baseline attitude, and higher education contribute to maintaining long-term abstinence.

Borrelli et al [38], 2024 66 Behavioral activation app Sociodemographic, smoking history, and app satisfaction and engagement Baseline motivation is an important variable for quitting smoking.

Chen et al [39], 2024 131 quitSTART Sociodemographic and app usability Higher app usability positively influences the program completion.

This study 692 Adiquit Sociodemographic, smoking, and health history, and app-based variables, including psychological, behavioral, and log-based variables Social support, prequitting feelings, and log parameters (screen time and number of sessions) in the first 10 days, and reason for quitting as job significantly influence the program completion.

Discussion

Principal Findings

One of the significant findings from this study’s descriptive statistics is the role of social support in program completion. Social support provides users with a sense of being cared for, loved, and valued, and it also provides a sense of belonging to a community or network [40]. Literature suggests that for humans, social support acts as a stress buffer and has been found effective in improving the success rate of many cessation therapies focused on quitting alcohol, tobacco, opioids, etc [41,42]. As shown in Table 2, 61.32% (203/331) of users with social support completed the Cessation Program, compared with 38.67% (128/331) in the group without social support. Similar findings were observed in the univariate analysis, where social support emerged as a significant predictor (P<.001) and positively influenced program completion. A user with social support had 86% (OR 1.862, 95% CI 0.389-0.855) higher odds of completing therapy than a user without social support. The second important finding of the study is the role of a positive perception of therapy completion in the initial days. Information regarding the user’s attitude toward the therapy was collected during the preparatory phase through a prequitting feelings questionnaire (eg, “Are you afraid that you may fail in this therapy?”). The response was collected as either “yes, I am afraid” or “no, I will not fail this time.” Hence, a higher value in the prequitting feelings reflects a positive attitude toward the therapy. In many previous studies, a positive perception of quitting or a positive attitude toward program completion has been associated with better therapy outcomes [43,44]. The results of this study also support this finding and suggest that a positive attitude toward the therapy at the initial stage is associated with a higher likelihood of completing the therapy than a negative attitude. The log information is considered a measure of user engagement in the therapy. According to descriptive statistics and univariate analysis, log-transformed measures (number of sessions completed and total time spent in the app) were positively associated with program completion. The univariate analysis also revealed that users whose primary quitting reason was job showed a lower likelihood of completing the therapy. Gill et al [45] discussed the impact of several reasons for quitting on program completion and investigated the role of intrinsic and extrinsic motivators as reasons for quitting in program completion [45]. They reported that extrinsic factors, such as finances, job satisfaction, and physical appearance, are negatively correlated with program completion. On the contrary, intrinsic factors such as health, well-being, and reduced anxiety increase the likelihood of abstaining and are expected to positively influence cessation outcomes. This explains the negative correlation between the quitting reason (job) and the quitting success, as reported in this study. The study also proposes a stratified multivariate model with bootstrapping and cross-validation, which suggests that social support, prequitting feelings, and number of sessions completed in the first 10 days are important predictors of program completion. Table 6 shows the significance of the obtained predictors in the multivariate analysis across specific subgroups (age, sex, health, and employment status). The stratified multivariate analysis across 4 demographic dimensions reveals a remarkably consistent pattern: sessions completed within the first 10 days and social support are the most cross-cutting predictors of therapy completion, with significance found in the majority of subgroups regardless of sex, age, health, and employment status. However, when the analysis was performed across different groups, some inconsistencies in the findings were observed in the diseased population and the older age group (Table 6), where social support was not significant (P=.42). Prequitting feelings are also broadly significant, though with slightly greater fluctuation in smaller strata such as diseased groups, an expected result of lower statistical power rather than a true moderating influence. When age group is considered as a predictor within strata, it contributes inconsistently and rarely approaches significance, supporting the core model’s finding that it has limited independent predictive power. Notably, the older age subgroups have the weakest and least consistent results overall, most likely due to smaller sample sizes and the added complexity of overlapping health and socioeconomic stresses, which may dilute the influence of engagement-based variables. The effect sizes in the healthy stratum, which comprised the majority of participants, closely matched those of the original model, providing confidence in the generalizability of the fundamental findings within these groups. This study has advantages in terms of population size and subgroups of variables considered. Most of the findings presented in this study align with the literature (Table 7), validating and strengthening the efficacy of the intervention method used. However, due to differences in population sizes, study settings, and design methodologies, a direct comparison of the results may be imprecise. The feasibility of the intervention was supported by progression through multiple stages and completion by 27.02% (187/692) of registered users. Acceptability was indirectly evidenced by early engagement and social support, which were associated with higher completion. However, the absence of direct user feedback limits conclusions about perceived acceptability.

Limitations, Scope for Further Research, and Conclusions

This study focuses on identifying the influential predictors in a digital intervention therapy that helps users quit smoking. Data from 692 users were used to examine the effects of several variables, including sociodemographic, smoking, health history, employment status, and app-derived factors, on program completion. The results suggest that important predictors of program completion in a digital intervention for smoking cessation include social support, prequitting feelings, and the number of sessions completed in the first 10 days. However, these factors may not have a uniform dominant effect when population subgroups are considered by sex, age, health, and employment status. Hence, a population-specific approach is recommended while identifying the influential parameters of program completion in a digital intervention. The study was conducted exclusively among paid subscribers of the Adiquit app, representing a self-selected subgroup of motivated individuals. This limits external validity to the broader population of smokers, particularly those with lower digital health literacy or socioeconomic resources. Future research should examine predictors of cessation engagement across both free and paid user tiers. The retrospective, observational design of this study precludes causal inference. Hence, all identified associations should be interpreted as correlational, and future experimental or longitudinal studies are required to establish causal pathways between the identified predictors and program completion. One major limitation of the study is the lack of clinical validation of users’ quitting. No clinical validation was performed using cotinine levels or exhaled carbon monoxide. Hence, this study is more focused on users’ program completion rather than actual smoking cessation. The presence of potential confounding effects among predictors and population groups was not addressed and should be explored in the future. Feasibility and acceptability were assessed indirectly through attrition and engagement metrics. No direct measures of user satisfaction or qualitative feedback were collected, which restricts the interpretation of acceptability. Future studies should incorporate validated acceptability instruments and user-reported outcomes. In summary, this study highlights the novel role of early engagement metrics and social support in predicting program completion in a digital, CBT-based cessation program delivered via the Adiquit app. Future research should validate these predictors across diverse populations and integrate them into adaptive intervention designs.

Supplementary material

References

1.WHO global report on trends in prevalence of tobacco use 2000–2030 World Health Organization2024 URLhttps://iris.who.int/server/api/core/bitstreams/1903bfca-6c2f-470c-bd52-237edf4828ca/contentAccessed22-09-2026

2.Chen Q Zhang C Zhong F Huang Y Zeng Y Zhang S Global burden of disease related to tobacco products and trends projected: 1990-2021 Addict Behav Oct2025 169 108391 doi10.1016/j.addbeh.2025.108391 Medline40414139

3.Wei H Rui J You M et alMaterials for the selective removal of toxic compounds in cigarette smoke: a review Sep Purif Technol Feb192025 354 128908 doi10.1016/j.seppur.2024.128908

4.Buczkowski K Dachtera-Frąckiewicz M Luszkiewicz D Klucz K Sawicka-Powierza J Marcinowicz L Reasons for and scenarios associated with failure to cease smoking: results from a qualitative study among Polish smokers who had unsuccessfully attempted to quit Patient Prefer Adherence 2021 15 2071 2084 doi10.2147/PPA.S320798 Medline34556977 PMC8453429

5.WHO global report on trends in prevalence of tobacco use 2000-2025, fourth edition World Health Organization2021 URLhttps://iris.who.int/server/api/core/bitstreams/94b19096-c193-49c4-bcbb-653aa34ee29a/contentAccessed22-09-2026

6.Hartmann-Boyce J Livingstone-Banks J Ordóñez-Mena JM et alBehavioural interventions for smoking cessation: an overview and network meta-analysis Cochrane Database Syst Rev 0142021 1 1 CD013229doi10.1002/14651858.CD013229.pub2 Medline33411338 PMC11354481

7.Schwaninger P Berli C Scholz U Lüscher J Effectiveness of a dyadic buddy app for smoking cessation: randomized controlled trial J Med Internet Res Sep92021 23 9 e27162doi10.2196/27162 Medline34499045 PMC8461528

8.Brendryen H Kraft P Schaalma H Looking inside the black box: using intervention mapping to describe the development of the automated smoking cessation intervention “happy ending” J Smok Cessat Jun12010 5 1 29 56 doi10.1375/jsc.5.1.29

9.Nian T Guo K Liu W et alNon-pharmacological interventions for smoking cessation: analysis of systematic reviews and meta-analyses BMC Med Sep292023 21 1 378doi10.1186/s12916-023-03087-z Medline37775745 PMC10542700

10.Pallejà-Millán M Rey-Reñones C Barrera Uriarte ML et alEvaluation of the Tobbstop mobile app for smoking cessation: cluster randomized controlled clinical trial JMIR mHealth uHealth Jun262020 8 6 e15951doi10.2196/15951 Medline32589153 PMC7381259

11.Bricker JB Watson NL Mull KE Sullivan BM Heffner JL Efficacy of smartphone applications for smoking cessation: a randomized clinical trial JAMA Intern Med Nov12020 180 11 1472 1480 doi10.1001/jamainternmed.2020.4055 Medline32955554 PMC7506605

12.Santiago-Torres M Mull KE Sullivan BM Kendzor DE Bricker JB Efficacy and utilization of smartphone applications for smoking cessation among low-income adults: secondary analysis of the iCanQuit randomized trial Drug Alcohol Depend Feb12022 231 109258 doi10.1016/j.drugalcdep.2021.109258 Medline35026491 PMC8810613

13.Xie JH Qiu YF Zhu L et alEvaluation of the smoking cessation effects of QuitAction, a smartphone WeChat platform Tob Induc Dis 2023 21 49doi10.18332/tid/161257 Medline37057059 PMC10088363

14.Alazbah A Fakeeh K Rabie O Exploring human crowd patterns and categorization in video footage for enhanced security and surveillance using computer vision and machine learning arXiv Preprint posted online onAug262023 doi10.48550/arXiv.2308.13910

15.Li S Qu Z Li Y Ma X Efficacy of e-health interventions for smoking cessation management in smokers: a systematic review and meta-analysis EClinicalMedicine 2024 68 102412doi10.1016/j.eclinm.2023.102412 Medline38273889 PMC10809126

16.Madureira J Camelo A Silva AI et alThe importance of socioeconomic position in smoking, cessation and environmental tobacco smoke exposure during pregnancy Sci Rep Sep242020 10 1 15584doi10.1038/s41598-020-72298-8 Medline32973155 PMC7515871

17.Lightfoot K Panagiotaki G Nobes G Effectiveness of psychological interventions for smoking cessation in adults with mental health problems: a systematic review Br J Health Psychol Sep2020 25 3 615 638 doi10.1111/bjhp.12431 Medline32678937

18.Kwon DM Santiago-Torres M Mull KE Sullivan BM Bricker JB Older adults who smoke: do they engage with and benefit from web-based smoking cessation interventions? Prev Med Aug2022 161 107118 doi10.1016/j.ypmed.2022.107118 Medline35718119 PMC12955718

19.Gram IT Antypas K Wangberg SC Løchen ML Larbi D Factors associated with predictors of smoking cessation from a Norwegian internet-based smoking cessation intervention study Tob Prev Cessat 2022 8 38doi10.18332/tpc/155287 Medline36382026 PMC9620393

20.Rajani NB Bustamante L Weth D Romo L Mastellos N Filippidis FT Engagement with gamification elements in a smoking cessation app and short-term smoking abstinence: quantitative assessment JMIR Serious Games Feb12023 11 e39975doi10.2196/39975 Medline36724003 PMC9932870

21.Zhang M Wolters M O’Connor S Wang Y Doi L Smokers’ user experience of smoking cessation apps: a systematic review Int J Med Inform 072023 175 105069 doi10.1016/j.ijmedinf.2023.105069 Medline37084673

22.Gao Y Yang S Jiang N Zhou W Chu Q Liang Y Exploring gender equality in digital health: a qualitative study of community-dwelling older Chinese women Digit Health 2026 12 20552076261440176doi10.1177/20552076261440176 Medline41938727 PMC13049349

23.Gallus S Cresci C Rigamonti V et alSelf-efficacy in predicting smoking cessation: a prospective study in Italy Tob Prev Cessat 2023 9 April 15doi10.18332/tpc/162942 Medline37125003 PMC10141785

24.Duncan MS Freiberg MS Greevy RA JrKundu S Vasan RS Tindle HA Association of smoking cessation with subsequent risk of cardiovascular disease JAMA Aug202019 322 7 642 650 doi10.1001/jama.2019.10298 Medline31429895 PMC6704757

25.Sorensen G Allen JD Adamkiewicz G Yang M Tamers SL Stoddard AM Intention to quit smoking and concerns about household environmental risks: findings from the Health in Common Study in low-income housing Cancer Causes Control Apr2013 24 4 805 811 doi10.1007/s10552-013-0149-5 Medline23334887 PMC3602321

26.Troelstra SA Boot CRL Harting J Geuskens GA Kunst AE van der Beek AJ Associations of sustained smoking and smoking cessation with work-related outcomes: a longitudinal analysis Int Arch Occup Environ Health Apr2021 94 3 529 537 doi10.1007/s00420-020-01598-3 Medline33170345 PMC8032580

27.Valiente R Escobar F Urtasun M Franco M Shortt NK Sureda X Tobacco retail environment and smoking: a systematic review of geographic exposure measures and implications for future studies Nicotine Tob Res Aug42021 23 8 1263 1273 doi10.1093/ntr/ntaa223 Medline33155040

28.Nwosu A Boardman S Husain MM Doraiswamy PM Digital therapeutics for mental health: is attrition the Achilles heel? Front Psychiatry 2022 13 900615doi10.3389/fpsyt.2022.900615 Medline35982936 PMC9380224

29.Jha D Rauniyar A Srivastava A Hagos DH Tomar NK Sharma V et alEnsuring trustworthy medical artificial intelligence through ethical and philosophical principles arXiv Preprint posted online onSep212023 URLhttps://arxiv.org/html/2304.11530v4Accessed22-09-2026

30.Tudor-Sfetea C Rabee R Najim M et alEvaluation of two mobile health apps in the context of smoking cessation: qualitative study of cognitive behavioral therapy (CBT) versus non-CBT-based digital solutions JMIR mHealth uHealth Apr182018 6 4 e98doi10.2196/mhealth.9405 Medline29669708 PMC5932330

31.Baumel A Muench F Edan S Kane JM Objective user engagement with mental health apps: systematic search and panel-based usage analysis J Med Internet Res Sep252019 21 9 e14567doi10.2196/14567 Medline31573916 PMC6785720

32.Hughes JR Carpenter MJ Naud S Do point prevalence and prolonged abstinence measures produce similar results in smoking cessation studies? A systematic review Nicotine Tob Res 072010 12 7 756 762 doi10.1093/ntr/ntq078 Medline20504946 PMC2893294

33.Baumel A Edan S Kane JM Is there a trial bias impacting user engagement with unguided e-mental health interventions? A systematic comparison of published reports and real-world usage of the same programs Transl Behav Med Nov252019 9 6 1020 1033 doi10.1093/tbm/ibz147 Medline31689344

34.Marler JD Fujii CA Galanko JA Balbierz DJ Utley DS Durability of abstinence after completing a comprehensive digital smoking cessation program incorporating a mobile app, breath sensor, and coaching: cohort study J Med Internet Res Feb152021 23 2 e25578doi10.2196/25578 Medline33482628 PMC7920755

35.Proposal for a regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) and amending certain Union legislative acts European Commission2021 URLhttps://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex:52021PC0206Accessed22-09-2026

36.Rajani NB Mastellos N Filippidis FT Self-efficacy and motivation to quit of smokers seeking to quit: quantitative assessment of smoking cessation mobile apps JMIR mHealth uHealth Apr302021 9 4 e25030doi10.2196/25030 Medline33929336 PMC8122290

37.Seaman EL Robinson CD Crane D et alAssociation of spontaneous and induced self-affirmation with smoking cessation in users of a mobile app: randomized controlled trial J Med Internet Res Mar52021 23 3 e18433doi10.2196/18433 Medline33666561 PMC7980123

38.Borrelli B Bartlett YK Fulford D Frasco G Armitage CJ Wearden AJ Behavioral activation mobile app to motivate smokers to quit: feasibility and pilot randomized controlled trial JMIR Form Res Apr42024 8 1 e54912doi10.2196/54912 Medline38573739 PMC11027050

39.Chen Z Siegel LN Prutzman YM Wiseman KP Characterizing perceived usability and its correlation with smoking cessation: an analysis of user assessments of the smoking cessation app quitSTART Internet Interv Mar2024 35 100714doi10.1016/j.invent.2024.100714 Medline38313141 PMC10835281

40.Li C Li H Suomi R Antecedents and consequences of the perceived usefulness of smoking cessation online health communities Internet Res Dec192022 32 7 56 86 doi10.1108/INTR-04-2020-0220

41.Hagman BT Falk D Litten R Koob GF Defining recovery from alcohol use disorder: development of an NIAAA research definition Am J Psychiatry Nov12022 179 11 807 813 doi10.1176/appi.ajp.21090963 Medline35410494

42.Witkiewitz K Wilson AD Pearson MR et alProfiles of recovery from alcohol use disorder at three years following treatment: can the definition of recovery be extended to include high functioning heavy drinkers? Addiction 012019 114 1 69 80 doi10.1111/add.14403 Medline30063267 PMC6289769

43.Li M Koide K Tanaka M Kiya M Okamoto R Factors associated with nursing interventions for smoking cessation: a narrative review Nurs Rep Feb12021 11 1 64 74 doi10.3390/nursrep11010007 Medline34968313 PMC8608102

44.Heris CL Chamberlain C Gubhaju L Thomas DP Eades SJ Factors influencing smoking among Indigenous adolescents aged 10-24 years living in Australia, New Zealand, Canada, and the United States: a systematic review Nicotine Tob Res Oct292020 22 11 1946 1956 doi10.1093/ntr/ntz219 Medline31773158

45.Gill KK van der Moolen S Bilal S Phenomenological insight into the motivation to quit smoking J Subst Abuse Treat Dec2021 131 108583 doi10.1016/j.jsat.2021.108583 Medline34535321