Introduction: Smoking after a cancer diagnosis represents a modifiable health risk. It is recommended that oncology clinicians address tobacco use among their patients using the 5As brief model: Asking about use, Advising users to quit, Assessing willingness to quit, Assisting in quit attempts (counseling and medication), and Arranging follow-up. However, cross-sectional studies have found limited adoption of 5As (especially Assist and Arrange) in oncology settings. Further investigation is needed to understand changes in, and factors associated with, 5As delivery over time. Methods: Patients recently diagnosed with cancer and reporting current smoking (N = 303) enrolled in a smoking cessation clinical trial and completed three longitudinal surveys; at pre-intervention baseline and 3- and 6-month follow-up post-enrollment. Patient-level correlates of 5As receipt at baseline, 3 months, and 6 months were identified using multilevel regression models. Results: At baseline, patient-reported rates of 5As receipt from oncology clinicians ranged from 85.17% (Ask) to 32.24% (Arrange). Delivery declined from baseline to 6-month follow-up for all 5As, with the largest declines observed for Ask, Advise, Assess, and Assist-Counseling. Diagnosis of a smoking-related cancer was associated with greater odds of 5As receipt at baseline but lower odds at 6-month follow-up. At each time point, female gender, religiosity, advanced disease, cancer-related stigma, and smoking abstinence were associated with lower odds of 5As receipt, while reporting a recent quit attempt prior to enrollment was associated with higher odds of 5As receipt. Conclusion: Oncology clinicians’ 5As delivery declined over time. Clinician delivery of the 5As varied based on patients’ sociodemographics, clinical and smoking characteristics, and psychosocial factors.

© 2023 The Author(s). Published by S. Karger AG, Basel

Introduction

Approximately 10–30% of patients with cancer continue to smoke cigarettes post-diagnosis [1, 2]. Continued smoking after a cancer diagnosis is linked to worse quality of life, treatment effectiveness, treatment outcomes, risk of recurrence, risk of developing a second primary cancer, and survival, whereas quitting after diagnosis has the potential to improve these outcomes [3–6].

To address patients’ tobacco use, clinical practice guidelines recommend that healthcare clinicians deliver 5As at every visit (Ask about smoking, Advise cessation, Assess readiness to quit, Assist with motivation and/or cessation, and Arrange follow-up) [7]. Clinical practice guidelines from the National Comprehensive Cancer Network (NCCN) underscore the importance of oncology clinician-delivered smoking cessation interventions, and the American Society of Clinical Oncology (ASCO) includes documentation and counseling of all patients who smoke as a core quality indicator as part of their Quality Oncology Practice Initiative [8, 9]. While there has been some progress as evidenced by increased rates of smoking cessation counseling and referrals from 2015 to 2019, overall rates of treatment and referral remain low (<50%), indicating much room for improvement [10]. To date, the vast majority of studies investigating rates of 5As delivery in oncology settings are cross-sectional and have explored clinician-level (rather than patient level) reporting of 5As delivery [11], limiting understanding of potential inequities in 5A receipt and the degree to which patients who continue to smoke, or have recently quit, receive appropriate follow-up and ongoing cessation support. When queried about barriers to 5As delivery, oncology clinicians commonly cite lack of time and referral pathways [11, 12]. A collaborative care approach (in which patients are referred to a tobacco treatment program embedded in the cancer center and treatment is provided in consultation with oncology care providers) may enhance 5As delivery by minimizing burden on oncology clinicians while still involving them in ongoing smoking cessation treatment [13].

The present study uses data collected from a randomized trial comparing effects of combined pharmacotherapy and intensive smoking cessation counseling to standard tobacco treatment for patients with newly diagnosed cancers in the context of a collaborative care model involving oncologists who provided usual care for patients [14]. Through exploratory secondary analyses of data collected over patients’ 6 months of trial participation, the present study aims to (1) characterize oncology clinicians’ rates and patterns of 5As delivery for recently diagnosed cancer patients during the 6 months following trial enrollment (as reported by patients) and (2) identify patient-level sociodemographic, disease, smoking, and psychosocial factors associated with variation in 5As receipt. The results of this study allow us to explore patient receipt of the 5As in an oncology setting and to identify subpopulations of patients potentially requiring greater attention.

Patients and MethodsParticipants and Setting

Data were collected as part of a randomized unblinded clinical trial (NCT01871506). Study design, attrition, power, and cessation outcomes are reported in greater detail elsewhere [14, 15]. Recruitment was standardized across the 2 study sites, Massachusetts General Hospital/Dana Farber Cancer Institute (MGH/DFCI) and Memorial Sloan Kettering Cancer Center (MSKCC). All new patients attending oncology appointments completed a brief screener, and patients were eligible if they self-reported smoking a cigarette, even a puff, within the last 30 days. Beyond self-reported current smoking, eligibility criteria were designed to be as inclusive as possible. Patients had to be at least 18 years old, English- or Spanish-speaking, with a recently diagnosed cancer within 3 months of diagnosis or 4 office visits. Patients were only excluded if they did not plan to receive cancer care at either study site, were psychiatrically or medically unstable, or otherwise unable to provide informed consent as determined by study investigators. Potentially eligible patients were either approached in-clinic or sent an opt-out letter inviting participation (see Fig. 1: CONSORT diagram). This study protocol was reviewed and approved by the MGH/Partners Health System Institutional Review Board (#2013P001036/PHS) and the Memorial Sloan Kettering Cancer Center Institutional Review Board (#14–079) and was conducted in accordance with the Declaration of Helsinki. All subjects provided written informed consent prior to participation.

Fig. 1.

Flowchart of enrollment and intervention to test the effectiveness of two models of tobacco treatment integrated into cancer are at diagnosis. Trial, November 2013 to July 2017. aPatients could give multiple reasons for refusal. The research assistant categorized the reasons of refusal that patients offered according to options available on the study screening tool; those reasons that did not fit into one of these predefined categories were discussed with the team to determine fit with existing categories or establishment of new categories. bMultiple reasons for ineligibility could have been indicated on the screener. As such, the number of reasons exceed the number of patients ineligible. cThose who were never randomized were those who signed a consent form but did not complete a counseling session and were thus not randomized to a treatment group. Reasons included participants who were not able to be reached by the study counselor, participants who withdrew citing other cancer care demands, and participants who became ineligible over time. dA computer was used to randomize participants. Blocks of 6 were used for treatment assignment and were stratified by cancer center clinic and study site. eThose who did not complete follow-up survey include those who were not able to be reached at all to complete the follow-up assessment. fFollow-up survey completion rate = completed/completed+ refused+withdrew+did not complete follow-up survey. Participants who were deceased or medically ineligible (e.g., in inpatient hospice or psychiatrically impaired) at follow-up were not included in the final outcome analyses (n = 5 intensive treatment and n = 15 standard treatment). Thus, for the intensive treatment, the denominator = 148, and for the standard treatment, the denominator = 135.

The study compared intensive tobacco treatment (4 counseling sessions over 1 month, biweekly counseling for 2 months, and 3 monthly booster sessions over 6 months plus choice of free FDA-approved cessation medication) with standard tobacco treatment (4 counseling sessions over 1 month and medication advice). Oncology clinicians were informed of the study and could refer patients but were not required to provide smoking cessation treatment as part of the trial, although delivery of the 5As was encouraged as part of usual care. Both treatments were patient-level interventions using a collaborative care model, whereby oncology clinicians were sent an email notification when a patient started tobacco treatment, encouraging clinicians to provide cessation advice and encouragement. Tobacco treatment specialists delivering the intervention documented all sessions in patients’ electronic health record (EHR), including details regarding tobacco treatment plan, goals, and progress.

Measures

Participants completed surveys at baseline prior to randomization and at 3- and 6-month post-enrollment follow-up. Baseline surveys were collected in person or by telephone, and follow-up surveys were collected in person or by telephone, mail, or email based on patient preference. Participants self-reported sociodemographics, smoking characteristics, and psychosocial factors relevant to tobacco use and clinical care, while clinical characteristics were abstracted from participants’ medical records.

Sociodemographics assessed at baseline were age, gender, race, ethnicity, education, marital status, and employment. Clinical characteristics assessed at baseline were cancer stage and diagnosis. Based on classification of smoking-related cancers in the 2014 Surgeon General Report [16], we categorized cancer diagnosis into smoking related (e.g., lung, esophageal, head and neck, bladder, kidney, liver, pancreatic, colorectal, anal, small intestinal, gastric, or cervical) and not smoking related (e.g., prostate, testicular, penile, breast, lymphoma, melanoma, or non-cervical gynecological cancer).

Smoking characteristics assessed at baseline included time to first cigarette (a proxy for nicotine dependence), quit attempt history, presence of additional smokers in the household, and smoking-related comorbid disease (including asthma, emphysema/COPD, hypertension, heart attack, stroke). Quit status was determined by patient self-reported continuous abstinence from cigarette smoking at 3 and 6 months, which was biochemically verified (salivary cotinine or expired carbon monoxide) [14].

Psychosocial factors included diagnosis of a serious mental illness (e.g., major depressive disorder, bipolar disorder, schizophrenia), religiosity (assessed at baseline), anxiety and depression symptoms (assessed at all 3 time points), and cancer-related stigma (assessed at baseline and 3 months). Serious mental illness was a binary variable determined by a combination of EHR and patient self-report. Religiosity was measured with a single Likert-style question about the extent to which the respondent considered themselves a religious or spiritual person, ranging from 1 (not religious or spiritual) to 4 (very religious or spiritual). Although use of a single item measure of religiosity/spirituality provides limited understanding, single item measures of religious importance are widely used in contexts where more extensive measurements are not feasible and demonstrate adequate single-item construct validity [17, 18].

Anxiety was measured using the GAD-7 (range 0–21), the sum of 7 items encompassing generalized anxiety disorder symptoms; higher scores indicate greater anxiety [19]. Depression was measured using the PHQ-9 (range 0–27), the sum of 9 items representing symptoms of major depression; higher scores indicate more depressive symptoms [20]. Patient-perceived cancer-related stigma was measured using the sum of 5 items on 4-point Likert scale assessing extent to which patients endorsed feeling internalized shame and blame related to their cancer (range 5–20) [21].

Participant-reported 5As receipt was measured at baseline, 3- and 6-month follow-up in a manner consistent with previous research [22] by asking patients the following question: did your oncology care provider(s) (e.g., doctor, nurse) do any of the following? (1) Ask about your current tobacco use; (2) Advise you to quit; (3) Assess your readiness to quit; (4) Assist you in quitting smoking; and (5) Arrange follow-up. Assist was further separated into (4a) Assist-Talk (talk to you about quitting smoking); (4b) Assist-Counseling (recommend cessation counseling); and (4c) Assist-Medication (recommend nicotine replacement therapy or other cessation pharmacotherapy). Responses to each question were binary (yes/no). The time frame assessed at baseline was “during your last visit,” while the time frame assessed at 3 and 6 months was “in the past 3 months.”

Statistical Analyses

Analyses were conducted using R version 3.5.1. Table 1 presents descriptive analyses including frequencies, means, and standard deviations for sociodemographic, clinical, and psychosocial variables at baseline. Handling of missing data and attrition are discussed in detail elsewhere [14]. Briefly, scales that were 80% complete were completed with the mean score from available items, and remaining missing data were imputed using a 2-step Markov chain Monte Carlo procedure. At baseline, rates of missing data were relatively low (<5%) for nearly all items, and approximately 30% of the sample was lost to follow-up at 3 and 6 months (Fig. 1). Previous analyses [14] indicated that younger age and more advanced cancer were associated with loss to follow-up at both 3 and 6 months; thus, these factors were included and retained as predictors in all models.

Table 1.

Patient characteristics at baseline (N = 303)

VariableM (SD)/n (%)RangeAge, years58.34 (9.71)21–86Gender Male133 (43.89) Female170 (56.11)Race American Indian/Alaska Native3 (0.99) Asian2 (0.66) Black/African American31 (10.23) White265 (87.46) Other2 (0.66)Ethnicity* Not Hispanic/Latino282 (93.07) Hispanic/Latino11 (3.70)Marital/relationship statusc Not married/partnered131 (43.23) Married/partnered164 (54.13)Employment statusc Not full-time employed196 (64.86) Full-time employed100 (33.00)Educationc High school or less93 (30.69) Some college or more202 (66.67) Cigarettes per day14.08 (9.89)1–70Time to first cigarettec Greater than 30 min after waking83 (27.39) Less than 30 min after waking214 (70.60)Recent quit attemptc Greater than 6 months ago119 (39.27) Less than 6 months ago81 (26.70) Importance of quitting9.28 (1.60)0–10Household smokersc None123 (40.59) One or more93 (30.69) Perceived cancer-related stigma10.43 (3.90)5–20Comorbid smoking-related disease None155 (51.16) Comorbid smoking-related disease148 (48.84)Smoking-related cancera Not smoking-related cancer122 (40.25) Smoking-related cancer181 (59.74)Cancer type Thoracic93 (30.69) Breast77 (25.41) Genitourinary51 (16.83) Gastrointestinal29 (9.57) Head and neck31 (10.23) Lymphoma9 (2.97) Gynecological7 (2.31) Melanoma6 (1.98)Cancer stage 017 (6.16) I86 (31.16) II67 (24.28) III53 (19.20) IV53 (19.20) Nonsolid indolent3 (0.99) Nonsolid advanced5 (1.65) NA/unknown19 (6.27)Stage of diagnosisc,b Not advanced182 (60.07) Advanced111 (36.63)5As Ask247 (81.5) Advise228 (75.2) Assess220 (72.6) Assist (talk)139 (45.9) Assist (counseling)166 (54.8) Assist (medication)131 (43.2) Arrange93 (30.7) Anxiety (GAD-7)9.5 (5.9)0–21 Depression (PHQ-9)8.6 (5.9)0–27 Religiosity2.6 (0.9)1–4

Overall rates of oncology clinician 5As delivery at each follow-up were determined by dividing the number of patient participants who reported receiving each of the 5As by the total number of participants providing data for that item for 3-month (217 respondents) and 6-month (221 respondents) follow-up (reported graphically in Fig. 2). While baseline analyses represented pre-intervention 5As as raw percentages [23], we use valid percentages (percent endorsing each A out of the total number providing follow-up data at 3 and 6 months rather than total baseline sample) so as not to under-represent rates of 5As at 3 and 6 months due to loss to follow-up.

Fig. 2.

Percentage of sample reporting 5As at each time point. Note: percentages represent proportion of the sample providing data at each time point so as not to under-represent rates of 5As at 3 and 6 months due to loss to follow-up.

To assess the associations between predictors and each of the 5As across the 3 time points, 7 multilevel linear models with time points nested within participants were estimated using a logit link in the lme4 package [24]. The objective was to create a model for each of the 5As over time, selecting predictor variables for each model in an exploratory manner based on fit and parsimony. Models were built systematically; we first compared the fit of models with a fixed versus random time parameter. Random effects of time did not improve the fit of models; thus, time was modeled only as fixed. Next, we added both between- and within-person fixed effects including study arm, study site, age, cancer stage, and quit status, as well as interactions between each variable and time. We tested interactions with time to determine whether any factors influenced the rate of change in 5As over time. To select variables to include in initial models, Pearson correlations between sociodemographic, clinical, smoking, and psychosocial variables and each of the 5As were examined; those with α levels below 0.10 were included (see Table 2 for a comprehensive list of sociodemographic, clinical, smoking, and psychosocial variables included in initial models and later removed). Level 1 variables (assessed across multiple time points) included in initial models included time, quit status, generalized anxiety, depressive symptoms, and stigma; level 2 variables (assessed only at baseline) included site, arm, age, stage, marital status, time to first cigarette, household smokers, smoking-related cancer, serious mental illness, smoking-related disease, recent quit attempt, and religiosity.

Table 2.

Variables tested in initial multilevel logistic regression models for each of the 5As

1. AskTime, site, armSociodemographic: age, race/ethnicityaClinical: stage, smoking-related cancerSmoking: quit status, smoking-related diseasea, recent quit attemptPsychosocial: serious mental illnessa, generalized anxietya, depressive symptomsa, stigma, religiosity2. AdviseTime, site, armSociodemographic: age, race/ethnicitya, marital statusaClinical: stage, smoking-related cancerSmoking: quit status, smoking-related diseasea, recent quit attempt, time to first cigaretteaPsychosocial: generalized anxietya, stigmaa3. AssessTime, site, armSociodemographic: age, genderaClinical: stageSmoking: quit status, smoking-related diseasea, recent quit attempt, household smokerPsychosocial: generalized anxietya, depressive symptomsa, stigmaa, religiositya4. Assist-TalkTime, site, armSociodemographic: ageClinical: stageSmoking: quit status, recent quit attemptPsychosocial: serious mental illnessa, generalized anxietya, depressive symptomsa5. Assist-CounselingTime, site, armSociodemographic: age, race/ethnicityaClinical: stageSmoking: quit status, smoking-related diseasea, recent quit attempt, household smokeraPsychosocial: generalized anxietya, depressive symptomsa, stigma, religiosity6. Assist-MedicationTime, site, armSociodemographic: age, genderaClinical: stage, smoking-related canceraSmoking: quit status, smoking-related diseasea, recent quit attempta, time to first cigarettea, household smokeraPsychosocial: generalized anxietya7. ArrangeTime, site, armSociodemographic: age, race/ethnicitya, genderClinical: stageSmoking: quit status, smoking-related diseasea, recent quit attemptPsychosocial: generalized anxietya, depressive symptomsa, stigmaa, religiosity

Final models for each of the 7 5A variables (presented in Table 3) were then determined based on comparison of fit statistics (i.e., AIC, BIC, model convergence, residuals) of a series of nested models. Through this process of nested model comparison, several variables and interactions were dropped from each model for nonsignificant associations, model fit, and model parsimony. Variables with α ≥ 0.10 were retained in the model if their removal significantly worsened model fit (AIC, BIC), if they were related to study design (e.g., treatment arm, study site), or if they were predictive of attrition (e.g., age, quit status, and cancer stage). All model assumptions and outliers were examined, and models were estimated with and without outliers to test the robustness of effects and consistency of results. Variables included in final models for each of the 5As are included in Table 3. All significant interactions included in the final models were decomposed to obtain simple slopes. As the primary outcomes are binary, simple slopes were transformed from log odds to probabilities for ease of interpretation prior to graphing. Probability graphs for all significant interactions are presented in Figure 3.

Table 3.

Variables associated with the 5As across study time points in multi-level logistic regression model

OR [95% CI]BlogSEzp value1. AskTime0.58 [0.30, 1.10]−0.550.33−1.660.10Site1.54 [0.52, 4.52]0.430.550.790.43Arm0.64 [0.23, 1.81]−0.440.53−0.840.40Age0.97 [0.91, 1.03]−0.030.03−1.030.30Stage0.60 [0.27, 1.34]−0.510.41−1.240.22Quit status0.26 [0.12, 0.64]1.340.46−2.930.003**Smoking-related cancer4.26 [1.37, 13.23]1.450.582.510.01*Stigma0.90 [0.81, 0.99]−0.110.05−2.090.04*Recent quit attempt1.72 [0.77, 3.86]0.550.411.320.19Time × arm2.38 [1.17, 4.86]0.870.362.390.02*Time × age1.07 [1.03, 1.13]0.070.023.080.002**Time × smoking-related cancer0.36 [0.17, 0.77]−1.030.39−2.640.008**2. AdviseTime0.31 [0.09, 1.05]−1.180.63−1.880.06Site0.99 [0.44, 2.19]−0.010.41−0.040.97Arm1.08 [0.61, 1.92]0.080.290.260.80Age0.94 [0.90, 0.99]−0.060.03−2.270.02*Stage0.43 [0.23, 0.79]−0.850.31−2.720.007**Quit status1.25 [0.20, 7.73]0.220.930.240.81Smoking-related cancer2.82 [1.20, 6.61]1.040.442.380.02*Recent quit attempt1.53 [0.84, 2.79]0.420.311.390.17Time × age1.07 [1.03, 1.12]0.070.023.49<0.001***Time × quit status0.26 [0.07, 0.90]−1.350.64−2.120.03*Time × smoking-related cancer0.31 [0.16, 0.61]−1.170.34−3.41<0.001**3. AssessTime0.09 [0.02, 0.44]−2.430.82−2.970.003**Site0.99 [0.41, 2.41]−0.010.45−0.020.98Arm1.32 [0.70, 2.49]0.280.320.850.39Age1.01 [0.97, 1.04]0.010.020.490.62Stage0.9 [0.48, 1.72]−0.100.33−0.310.76Quit status1.48 [0.17, 12.82]0.391.100.360.72Household smoker0.39 [0.16, 0.94]−0.950.45−2.100.04*Recent quit attempt2.21 [1.14, 4.30]0.790.342.340.02*Time × quit status0.20 [0.04, 0.97]−1.630.82−2.000.046*Time × household smoker2.06 [1.02, 4.16]0.720.362.010.044*4. Assist-TalkTime0.17 [0.05, 0.61]−1.750.64−2.730.006**Site0.60 [0.31, 1.16]−0.520.34−1.520.13Arm1.40 [0.87, 2.26]0.340.241.380.17Age1.02 [0.99, 1.05]0.020.011.450.15Stage0.81 [0.49, 1.33]−0.210.25−0.840.40Quit status3.37 [0.55, 20.65]1.210.931.310.19Recent quit attempt1.79 [1.10, 2.91]0.580.252.330.02*Time × quit status0.19 [0.05, 0.67]−1.680.65−2.560.01*5. Assist-CounselingTime0.37 [0.23, 0.60]−0.990.24−4.08<0.001***Site0.82 [0.36, 1.86]−0.190.42−0.470.64Arm0.76 [0.36, 1.60]−0.280.38−0.730.47Age1.04 [1.00, 1.07]0.030.022.060.04*Stage1.16 [0.64, 2.11]0.150.300.490.62Quit status0.23 [0.11, 0.54]−1.43