Real-World Impact of Quitline Interventions for Provider-Referred Smokers Glory Song, MPH, Anna S. Landau, MPH, Timothy J. Gorin, MPH, Lois Keithly, PhD Background: The healthcare provider–referral quitline model has potential to help identify and connect more smokers to effective cessation services as compared to the self-referral model alone. However, research is limited as to whether provider-referred smokers, who may have more barriers to quitting, can have similar rates of quit success using traditional quitline interventions as selfreferred smokers. Purpose: To (1) determine how provider-referred smokers may differ from self-referred smokers in their demographics, service utilization, and quit rates and (2) quantify the impact of traditional quitline services on provider-referred smokers’ ability to quit. Methods: Data were collected for 2,737 provider-referred and 530 self-referred Massachusetts quitline clients between November 2007 and February 2012. Analysis was performed in 2012. Wald chi-square tests and two-sample t-tests were used to identify differences between the two referral populations. A multivariable logistic regression model was used for each referral population, and smoker quit status at follow-up was the primary outcome. Tests and models were weighted using inverse probability of treatment weights propensity score weighting method. Results: Compared with self-referred smokers, provider-referred smokers were more likely to be non-white, less educated, and have public insurance. They were less ready to quit and had lower service utilization and quit rates. In both referral populations, clients who used services had greater odds of quitting than those who did not. Conclusions: Expanding the provider-referral model may require quitlines to address the various risk factors associated with this population. Providers serve critical roles in preparing patients for quitline participation prior to referral. (Am J Prev Med 2014;47(4):392–402) Published by Elsevier Inc. on behalf of American Journal of Preventive Medicine
Introduction
T
he clinical and real-world effectiveness of telephone-based tobacco quitlines is well established.1–4 Smokers who use quitlines are significantly more likely to make quit attempts and quit than smokers who do not use quitlines.5 However, much of the previous research examined traditional quitline service models, in which the smoker initiates the first call to a quitline. Recently, interest has increased nationally in the expansion of a healthcare provider–referral From the Tobacco Cessation and Prevention Program (Song, Landau, Gorin) and the Massachusetts Department of Public Health (Keithly), Boston, Massachusetts Address correspondence to: Glory Song, MPH, Tobacco Cessation and Prevention Program, Massachusetts Department of Public Health, 250 Washington Street, Boston MA 02108. E-mail: [email protected]. 0749-3797/$36.00 http://dx.doi.org/10.1016/j.amepre.2014.05.043
392 Am J Prev Med 2014;47(4):392–402
model for quitlines, in which patient smokers are referred by their provider, either via electronic health record or fax, to a public quitline. It is proposed that this second quitline model, operating in complementary fashion alongside the first, could help identify and connect many more smokers to quitline services, thereby making a greater impact on the reduction of smoking prevalence compared to the self-referred model alone.6 Despite the appeal of the provider-referral system as a complementary, farther-reaching service model than the traditional self-referral model,7–9 research on the realworld impact of traditional quitline interventions (telephone counseling, nicotine replacement therapy [NRT], and self-help materials) on provider-referred smokers’ ability to quit is limited. Prior studies have found that patients referred to quitline services do quit at greater rates than those who received only standard in-practice care.10,11 However, few studies have attempted to
Published by Elsevier Inc. on behalf of American Journal of Preventive Medicine
Song et al / Am J Prev Med 2014;47(4):392–402
quantify the impact of individual quitline interventions on provider-referred smokers. It also remains an empirical question whether quitlines can expect providerreferred smokers to have similar rates of participation and quit success as traditional (self-referred) callers. Studies have found that actively recruited smokers, such as those referred by healthcare providers, differ significantly from smokers who enroll on their own on several sociodemographic and smoking-related characteristics.12,13 Specifically, compared with non-referred smokers, healthcare provider–referred patients were found to have more comorbidities, less motivation to quit, less education, and less health insurance coverage.14 Separately, other studies have found that healthcare providers’ implementation of tobacco treatment models varies significantly and that patient enrollment rates and quit outcomes for a quitline can be affected by the degree of intervention and preparation received by the patient prior to referral.15 These potential differences highlight the importance of determining the extent to which quitlines can improve the cessation outcomes of provider-referred smokers. This paper contributes to the growing body of literature on provider referrals to quitlines by examining differences in demographics, service utilization, and quit outcomes between provider-referred and self-referred clients, and by measuring the real-world impact of cessation services for provider-referred clients. The Massachusetts Smokers’ Helpline, with more than 90% of its annual client volume derived from provider referrals (both electronic and fax), offers a unique opportunity to examine the potential benefits and limitations of this referral model.
Methods Helpline Protocol Overview The Massachusetts Smokers’ Helpline offers evidence-based proactive counseling, NRT, and self-help materials. Self-referred clients enroll in services via calling the national phone number, 1-800-QUITNOW, and provider-referred clients are referred to the quitline by their provider electronically or via fax. Both groups are eligible to receive proactive counseling and self-help materials. Provider-referred clients are eligible to be screened for a 2-week supply of nicotine replacement patches. Proactive counseling consists of five individualized telephone sessions with a Helpline quit coach, although clients may receive additional sessions if they experience relapse during the course of the program.
Data Collection and Sampling A retrospective analysis was performed in 2012 of Massachusetts provider-referred and self-referred clients who entered the quitline between November 2007 and June 2011. For this analysis, individuals who called the quitline as part of any free patch October 2014
393
promotions were excluded (n¼30,276), as these individuals largely did not participate in counseling and are not representative of the average quitline caller. Figure 1 shows the exclusion criteria used to select the final client cohort for analysis. The final complete case cohort used for analysis consisted of 2,737 provider-referred clients and 530 self-referred clients. Client data for demographic, smoking behavior, source of referral, services requested, and details of counseling sessions were collected during the course of service delivery. Client data for services used, quit outcomes, and use of additional quit aids were collected via telephone-based evaluation administered 6–8 months following the time of enrollment (May 2008 to February 2012), conducted on a rolling basis over each year. For this follow-up evaluation, 84% of provider-referred clients were attempted for contact (16% either had incomplete contact information or opted out) and, owing to budget constraints, only 32% of self-referred clients were randomly sampled for attempted contact. The response rate among those sampled for follow-up was 50% for provider-referred clients and 47% for self-referred clients.
Measures The main service use variables were number of counseling sessions used (zero to five or more); use of self-help materials (yes or no); and for the provider-referred population only, use of any of the 2week supply of NRT offered through the quitline (yes or no). Additionally, a four-tiered level of service variable was created to measure the effect of combination service use (counseling with NRT, NRT only, counseling only, or neither). The primary outcome was quit status (quit or not quit) measured by client self-report to having been “currently quit” at the time of the 6–8-month follow-up. Clients were asked, Do you currently use tobacco every day, some days, or not at all? Those who responded not at all were considered “currently quit.” Demographic variables (gender, age group, race, education, and insurance status) and smoking behavior variables (when client plans to quit smoking, time to first cigarette, and number of cigarettes used per day) were included for analysis. Additionally, the self-reported use of additional medications, including NRT (patch, gum, lozenge, inhaler, or nasal spray), varenicline (Chantix) or Buproprion SR, and Zyban or Wellbutrin SR (WZB), was included for analysis. Finally, for provider-referred clients, a variable that categorized the type of institution clients were referred from (hospital, provider practice, community health center, outpatient clinic, and human services/others) was included to examine its impact on client quit outcome.
Data Analysis Pearson chi-square tests for categorical variables and t tests for continuous variables were used to identify significant differences between clients with and without follow-up in each referral population (Table 1). Wald chi-square tests for categorical variables and t-tests for continuous variables were used to identify significant differences between provider-referred and self-referred clients across demographic, smoking behavior, service utilization rates, and quit outcomes (Table 2). Only those with complete follow-up data were included for analysis (n¼3,267).
394
Song et al / Am J Prev Med 2014;47(4):392–402 Massachusetts Helpline clients who enrolled between November 2007 and June 2011 n=11,582
Clients with missing data for any covariate were excluded n=1,228 (10.6%)
Clients with follow-up data
Clients without followup data
n=3,267 (31.6%)
n=7,087 (68.4%)
Provider-referred clients with complete follow-up data
Self-referred clients with complete followup data
n=2,737
n=530
Used for baseline comparison (Table 1)
Used for data analysis (Table 2 and Table 3)
Figure 1. Exclusion criteria for final data set selection. Note: This analysis excluded clients who came through the quitline between November 2007 and June 2011 as part of a free patch promotion (n¼30,276). These individuals largely did not participate in counseling and are not representative of the average quitline caller.
To measure the odds of quitting associated with the use of quitline services and determine key predictors of quit success, two separate multivariate logistic regression models were used, one for each referral population (Table 3). In each model, the service use variables predicted client quit status, after adjusting for other covariates. Only the provider-referred model included the use of a quitline-supplied NRT variable and referral institution type variable as covariates. To test whether variation existed among clients referred from different institutions within each institution type, a mixed-effects model was constructed for the providerreferred population that treated each unique referral site as a cluster. The two referral populations were not analyzed together because of the many differences that existed between them. Unlike selfreferred clients, provider-referred clients were eligible for a 2-week supply of nicotine patches from the quitline, may have received a physician intervention prior to referral, and may have had higher prevalence of physical and mental health comorbidities. Altogether, these and other factors may differentially influence the two populations’ quit outcomes. The sampling and evaluation completion rates determined the proportion of clients in the provider-referred and self-referred populations with complete follow-up evaluation data (41% and 14%, respectively). To account for any selection bias resulting from differences between those with and without follow-up, and to improve the overall representativeness of the sample population, an inverse probability of treatment weights (IPTW) propensity score method was applied. Logistic regression was used to estimate the propensity scores for client follow-up status,
predicted from nine covariates including demographics, smoking characteristics, and referral population that were collected for everyone. The propensity score weight was then calculated as the inverse of the propensity score for each client (IPTW).16 This weight was applied to the Wald chi-square test comparisons between the two referral populations and to the two logistic regression models. Data cleaning and manipulation were performed in Microsoft ACCESS (Redmond, WA) and SAS, version 9 (SAS Institute Inc., Cary NC). All statistical testing, including hypothesis tests and propensity score regression analyses, was performed in SAS.
Results Comparison of Client Characteristics by FollowUp Status In both referral groups, compared to those followed up with, those not followed up with were more likely to be younger than age 35 years and less likely to be older than age 55 years (Table 1). The population without follow-up was also more likely to have a high school education or less. In both referral groups, compared to those with follow-up, the population without follow-up was more likely to have public insurance or no insurance. Self-referred clients without follow-up on average smoked one additional cigarette per day compared to those with follow-up (Table 1). www.ajpmonline.org
395
Song et al / Am J Prev Med 2014;47(4):392–402
Table 1. Comparison of characteristics between clients with and without follow-up, n (%) Provider-referred (n¼6,682) No follow-up (n¼3,945)
Variables
Follow-up (n¼2,737)
Self-referred (n¼3,672) p-valuea
No follow-up (n¼3,142)
Follow-up (n¼530)
p-valuea
DEMOGRAPHIC VARIABLESb Gender
0.1576
0.7710
Male
1,661 (42.1)
1,200 (43.8)
1,266 (40.3)
210 (39.6)
Female
2,284 (57.9)
1,537 (56.2)
1,876 (59.7)
320 (60.4)
o0.0001
Age group (years)
o0.0001
18–34
901 (23.1)
415 (15.3)
842 (27.1)
100 (19.0)
35–54
2,064 (52.3)
1,441 (52.6)
1,547 (49.2)
257 (48.5)
936 (24.0)
851 (31.4)
722 (23.2)
168 (32.0)
Z55 Race
0.0745
White
2,966 (76.0)
2,094 (77.4)
Black
440 (11.3)
Other
495 (12.7)
0.6966 2,620 (84.2)
441 (84.0)
319 (11.8)
204 (6.6)
39 (7.4)
294 (10.9)
287 (9.2)
45 (8.6)
o0.0001
Education
0.0095
High school or less
2,398 (61.5)
1,523 (56.3)
1,648 (53.0)
246 (46.9)
College or beyond
1,503 (38.5)
1,184 (43.7)
1,463 (47.0)
279 (53.1)
o0.0001
Insurance None
280 (7.1)
148 (5.4)
Public
2,400 (60.9) 1,265 (32.1)
Private SMOKING BEHAVIOR
o0.0001 449 (14.3)
41 (7.7)
1,553 (56.7)
1,622 (51.6)
265 (50)
1,036 (37.8)
1,071 (34.1)
224 (42.3)
b
Plan to quit
0.9326
0.3205
Currently quit
499 (12.6)
346 (12.6)
266 (8.6)
57 (10.9)
Within 1 week
787 (20)
530 (19.4)
974 (31.3)
164 (31.2)
Within 1 month
1,572 (39.9)
1,088 (39.8)
1,060 (34.1)
166 (31.6)
After 1 month or don’t know
1,074 (27.5)
759 (28.2)
811 (26.1)
138 (26.3)
Time to first cigarette (minutes)
0.4014
0.3660
r30
3,185 (81.7)
2,188 (80.9)
2,609 (83.9)
432 (82.3)
430
716 (18.4)
519 (19.2)
502 (16.1)
93 (17.7)
17.9 (10)
17.5 (9.5)
Cigarettes per day (M [SD])
0.0980
20.9 (11)
19.7 (10.8)
0.0370
Note: Boldface indicates statistical significance. a Pearson χ2 tests were used for comparison of categorical variables and two-sample t tests were used for continuous variables. b This information was collected at time of intake.
Comparison of Client Characteristics by Referral Population Compared to self-referred clients, provider-referred clients were more likely to be male, aged 35–54 years, October 2014
of black or other race, have a high school education or less, and have public insurance. Self-referred clients were more likely to be aged 18–34 years and have either no insurance or private insurance (Table 2). Self-referred
396
Song et al / Am J Prev Med 2014;47(4):392–402 a
Table 2. Weighted comparison of characteristics, service use, and quit outcomes between provider-referred and selfreferred clients with follow-up Variables
Provider-referred (n¼2,737)
Self-referred (n¼530)
p-valueb
DEMOGRAPHIC VARIABLESc Gender
0.0234
Male
1,200 (43.8)
210 (39.6)
Female
1,537 (56.2)
320 (60.4)
Age group (years)
0.0036
18–34
415 (15.3)
100 (19.0)
35–54
1,441 (52.6)
257 (48.5)
851 (31.4)
168 (32.0)
Z55 Race
0.0006
White
2,094 (77.4)
441 (84.0)
Black
319 (11.8)
39 (7.4)
Other
294 (10.9)
45 (8.6)
Education
0.0005
High school or less
1,523 (56.3)
246 (46.9)
College or beyond
1,184 (43.7)
279 (53.1)
None
148 (5.4)
41 (7.7)
Public
1,553 (56.7)
265 (50)
1,036 (37.8)
224 (42.3)
Insurance
0.0025
Private c
SMOKING BEHAVIOR
o0.0001
Plan to quit Currently quit
346 (12.6)
57 (10.8)
Within 1 week
530 (19.4)
165 (31.1)
1,088 (39.8)
168 (31.7)
759 (28.2)
138 (26.3)
Within 1 month After 1 month or don’t know Time to first cigarette (minutes)
0.3340
r30
2,188 (80.9)
432 (82.3)
430
519 (19.2)
93 (17.7)
Cigarettes per day (M [SD])
17.5 (9.5)
19.7 (10.8)
o0.0001
USE OF QUITLINE SERVICES o0.0001
Self-help materiald Any
1,472 (53.8)
338 (63.8)
None
1,265 (46.2)
192 (36.2) o0.0001
Counseling sessions Z5 4 3
125 (4.6) 91 (3.3) 108 (4)
51 (9.6) 29 (5.5) 31 (5.8) (continued on next page)
www.ajpmonline.org
397
Song et al / Am J Prev Med 2014;47(4):392–402 a
Table 2. Weighted comparison of characteristics, service use, and quit outcomes between provider-referred and selfreferred clients with follow-up (continued) Variables 2
Provider-referred (n¼2,737)
Self-referred (n¼530)
136 (5)
29 (5.5)
1
406 (14.8)
105 (19.8)
0
1,871 (68.4)
285 (53.8)
Two-weeks of NRT patch
d,e
N/A
Any
1,067 (39)
N/A
None
1,670 (61)
N/A
Levels of services used
N/A
Any counseling and NRT patch
341 (12.5)
N/A
Any NRT patch only
726 (26.5)
N/A
Any counseling only
525 (19.2)
N/A
1,145 (41.8)
N/A
No counseling or NRT patch PROVIDER REFERRAL INSTITUTION TYPE
c
N/A
Provider practice
492 (18)
N/A
Community health center
287 (10.5)
N/A
Outpatient clinic
462 (16.9)
N/A
Human services/others
222 (8.1)
N/A
1,274 (46.6)
N/A
Hospitals ADDITIONAL QUIT AIDS USE
p-valueb
d
Additional nicotine replacementf Any None
0.0003 813 (29.7)
195 (36.8)
1,924 (70.3)
335 (63.2) o0.0001
Chantix Any None
361 (13.2)
138 (26)
2,376 (86.8)
392 (74)
Wellbutrin, Zyban, or Buproprion Any None
0.0504 235 (8.6)
33 (6.2)
2,502 (91.4)
497 (93.8)
d
QUIT OUTCOME Quit status Quit Not quit
0.0004 549 (20.1)
139 (26.2)
2,188 (79.9)
391 (73.8)
Note: Boldface indicates statistical significance. a Hypothesis test for each variable was weighted using the inverse probability of treatment weights propensity score weighting method. b Wald χ2 test were used for comparison of categorical variables and two-sample t-tests were used for continuous variables. c This information was collected at time of intake. d This information was collected at the time of the 6–8-month follow-up evaluation. e Some provider-referred clients did not receive patches owing to medical ineligibility or voluntary opt-out. Among 2,737 provider-referred clients, 50% (n¼1,342) self-reported to receiving patches from the quitline. f Additional nicotine replacement includes use of gum, inhaler, lozenge, and/or patches (for provider-referred clients, this indicates patches used outside of the free 2-week supply that may have been provided by the quitline). NRT, nicotine replacement therapy patches provided by the quitline.
October 2014
398
Song et al / Am J Prev Med 2014;47(4):392–402
Table 3. Odds of quitting associated with service use and key predictors of smoking abstinence,a AOR (95% CI) Provider-referred (n¼2,737)
Self-referred (n¼530)
p-value
Predictorsb
p-value
DEMOGRAPHIC VARIABLES Age group (years) Z55
1.65 (1.28, 2.13)
o0.0001
1.72 (0.82, 3.58)
0.1496
18–34
1.45 (1.16, 1.80)
0.0010
1.09 (0.60, 1.99)
0.7698
35–54
ref
ref
ref
ref
USE OF QUITLINE SERVICES Self-help material Any None
1.25 (1.06, 1.46)
0.0062
ref
1.62 (1.01, 2.61)
0.0458
ref
ref
o0.0001
Counseling sessions Z5
3.63 (2.68, 4.92)
o0.0001
4.46 (2.23, 8.92)
4
2.63 (1.84, 3.77)
o0.0001
4.15 (1.60, 10.74)
0.0034
3
2.08 (1.48, 2.93)
o0.0001
0.83 (0.34, 2.03)
0.6815
2
1.29 (0.92, 1.82)
0.1403
1.63 (0.69, 3.88)
0.2671
1
1.01 (0.80, 1.28)
0.9202
0.62 (0.33, 1.16)
0.1312
0
ref
ref
ref
N/A
N/A
N/A
N/A
Two weeks of NRT patch Any None
1.99 (1.60, 2.48)
o0.0001
ref
Levels of services used Any counseling and NRT patch
3.45 (2.60, 4.58)
o0.001
N/A
N/A
Any NRT patch only
2.19 (1.72, 2.80)
o0.001
N/A
N/A
Any counseling only
1.73 (1.40, 2.14)
o0.001
N/A
N/A
N/A
N/A
No counseling or NRT patch
ref
PROVIDER REFERRAL TYPE Provider practice
1.40 (1.13, 1.73)
0.0023
N/A
N/A
Community health center
1.14 (0.86, 1.50)
0.3684
N/A
N/A
Outpatient clinic
1.09 (0.87, 1.36)
0.4439
N/A
N/A
Human services/others
0.78 (0.56, 1.08)
0.1413
N/A
N/A N/A
Hospitals
ref
ref
N/A
Currently quit
6.84 (5.26, 8.90)
o0.0001
3.82 (1.83, 8.00)
0.0004
Within 1 week
1.76 (1.38, 2.23)
o0.0001
0.72 (0.39, 1.32)
0.2860
SMOKING BEHAVIOR Plan to quit
(continued on next page)
www.ajpmonline.org
399
Song et al / Am J Prev Med 2014;47(4):392–402 a
Table 3. Odds of quitting associated with service use and key predictors of smoking abstinence, AOR (95% CI) (continued) Provider-referred (n¼2,737) p-value
Predictorsb Within 1 month After 1 month or don’t know
Self-referred (n¼530)
1.26 (1.02, 1.56)
0.0309
p-value 1.18 (0.66, 2.09)
ref
ref
ref
430
1.50 (1.24, 1.82)
o0.0001
2.03 (1.17, 3.52)
r30
ref
ref
ref
0.5802 ref
Time to first cigarette (minutes) 0.0117 ref
ADDITIONAL QUIT AIDS USE Additional nicotine replacementc Yes
1.48 (1.16, 1.90)
No
ref
0.0016
2.60 (1.52, 4.46)
0.0005
ref
Chantix Yes
2.21 (1.63, 3.00)
No
ref
o0.0001
5.93 (3.22, 10.92)
o0.0001
ref
Wellbutrin, Zyban, or Buproprion Yes
2.85 (1.84, 4.42)
No
ref
o0.0001
2.53 (1.14, 5.62)
0.0228
ref
Note: Boldface indicates statistical significance. a Both logistic regression models predicted quit status (quit or not quit) at the 6–8-month follow-up and were weighted using the inverse probability of treatment weights propensity score method. b Other covariates included in both models are race, gender, education level, and insurance type. These covariates were not significant at α¼0.05 for either referral model and therefore are not presented in the table. c Additional nicotine replacement includes use of gum, inhaler, lozenge, and/or patches (for provider-referred clients, this indicates patches used outside of the free 2-week supply that may have been provided by the quitline). NRT, nicotine replacement therapy patches provided by the quitline.
clients were also more likely to be ready to quit within a week, whereas provider-referred clients were more likely to be ready to quit within a month. There was no significant difference in time to first cigarette, although self-referred clients on average smoked two more cigarettes per day than provider-referred clients (Table 2).
Service Use and Quit Outcomes by Referral Population Thirty-nine percent of provider-referred clients self-reported using the 2-week quitline-supplied NRT (Table 2). Compared to self-referred clients, provider-referred clients were less likely to use self-help materials (54% vs 64%) and any counseling sessions (32% vs 46%). They were also less likely to use additional NRT (30% vs 37%) and Chantix (13% vs 26%). Twenty percent of provider-referred clients quit compared to 26% of self-referred clients who quit. October 2014
Odds of Quitting Associated with Service Use Provider-referred clients who read the self-help materials had 1.2 times the odds of quitting compared with those who did not (Table 3). The effect of counseling approximated a dose–response relationship, where increased number of sessions was positively associated with higher odds of quitting. Clients who used any of the 2-week quitline-supplied NRTs had twice the odds of quitting than those who did not. Provider-referred clients who used only counseling and no quitline-supplied NRT experienced 1.7 times the odds of quitting, and those who only used quitline-supplied NRT and no counseling experienced 2.2 times the odds of quitting compared to clients who used neither. However, clients who used counseling together with NRT experienced the greatest success, having 3.4 times the odds of quitting compared to those who used neither service (Table 3). Self-referred clients who used self-help materials had 1.6 times the odds of quitting compared to non-users. Compared to provider-referred clients who, on average,
400
Song et al / Am J Prev Med 2014;47(4):392–402
saw significantly improved outcomes after three counseling sessions, self-referred clients saw improved outcomes only after four sessions (Table 3).
Predictors of Abstinence Among ProviderReferred Clients Among provider-referred clients, those aged 35–64 years had the lowest odds of quitting. By comparison, those aged 18–34 years and 65 years and older had 1.5 and 1.6 times the odds of quitting, respectively (Table 3). Readiness to quit was also an important predictor of quit outcome. Clients ready to quit within a week and within a month had 1.8 times and 1.3 times, respectively, the odds of quitting compared to those who planned to quit after one month or “didn’t know.” Those “currently quit” at time of enrollment had 6.8 times the odds of quitting. However, more than 50% of these clients had relapsed by the time of the 6–8 month follow-up (data not shown). Tobacco dependence was another important predictor; those who were able to wait more than 30 minutes before smoking their first cigarette had 1.5 times the odds of quitting compared to those who first smoked within 30 minutes after waking (Table 3). Those who used additional quit aids and medication experienced greater success, with clients who used nicotine replacement aids having 1.5 times the odds of quitting than non-users, those who used Chantix having 2.2 times the odds of quitting than non-users, and those who used Wellbutrin, Zyban, or Buproprion having 2.8 times the odds of quitting than non-users (Table 3). Finally, private practice–referred clients had 1.4 times the odds of quitting compared to hospital-referred clients. No other difference between comparison pairs was significant. Results from the mixed-effects model showed that the different referral institutions clustered within an institution type explained little of the remaining variation in residuals after accounting for the other independent variables in the model (variance component¼0.007, p¼0.372).
Predictors of Abstinence Among Self-Referred Clients Self-referred clients who could wait more than 30 minutes before smoking their first cigarette had twice the odds of quitting compared to those who smoked within 30 minutes after waking (Table 3). Self-referred clients who used nicotine replacement aids had 2.6 times the odds of quitting than non-users; those who used Chantix had 5.9 times the odds of quitting than non-users; and those who used Wellbutrin, Zyban, or Buproprion had 2.5 times the odds of quitting than non-users (Table 3).
Discussion This study found that provider-referred smokers who used quitline services had greater odds of quitting than those who did not. Those who used a combination of counseling and NRT had the greatest odds of quitting. Overall, provider-referred clients were more likely to be non-white, have less education, and have public insurance and were less ready to quit than self-referred clients. Provider-referred clients, having recently visited a provider, may also have been more likely to have physical or mental health comorbidities. All of these characteristics are associated with poorer cessation outcomes,17–20 and this was observed in lower rates of service utilization and quit success among provider-referred clients. These results generate a number of questions about quitline and provider roles under a provider-referral model. First, how might a quitline address the socioeconomic, motivational, and comorbid health concerns more often associated with provider-referred patients? Findings from the literature indicate that lower-SES smokers may need to receive more treatment content and have access to more intensive pharmacotherapy.21,22 Although offering medication or more services may not be feasible for public quitlines, any quitline could incorporate scripts that inform clients of available cessation coverage by insurance plans and encourage clients to utilize their benefits. To address readiness to quit among providerreferred clients, quitlines could aim to minimize the wait time between provider referral and actual service provision to the client. Increased wait time before receipt of service can increase the room for ambivalence for any smoker, but may be especially detrimental for provider-referred smokers who are less ready to make a quit attempt.23 Finally, to address disease comorbidities that are often associated with provider-referred smokers, several quitlines have piloted the integration of disease and risk behavior management into their quitline counseling.24,25 However, more research is needed regarding the effectiveness and sustainability of these programs. Second, what is a provider’s role in a referral model, both in assessing patient readiness to quit and in preparing patients for quitline services? In reality, not all patients have received an evidence-based intervention or are ready to quit at the time of enrollment in quitline services. When patients who are not ready to quit are referred, many are lost in the callback process and do not engage with the quitline at all. The Massachusetts reach rate for referred patients using three callback attempts is approximately 40%, and 50% using five callback attempts (data not shown). Similar rates have been observed by other quitlines.11,14,26 The difficulty reaching provider-referred clients highlights the critical need for effective and frequent www.ajpmonline.org
Song et al / Am J Prev Med 2014;47(4):392–402
provider training, outreach, feedback reporting, clinical champions on site, and systems support to ensure that providers can better assess and equip patients for quitline participation to improve the quality of referrals.27
Limitations The low proportion of clients in both populations with complete follow-up data raises concerns about potential selection bias that may affect the representativeness of the study sample to the overall population. A propensity score weighting method was used to minimize this effect. Results from the weighted regression models and unweighted regression models (not shown) were very similar. Any difference between the study sample and those lost to follow-up likely did not affect the study results and should not account for the associations that were observed. Therefore, the observed associations likely reflect the true associations in the population. The lower response rate among self-referred compared to provider-referred (47% vs 50%, p¼0.0174) clients may potentially have yielded an overestimation of the quit rate for self-referred clients, as those with poorer outcomes are more difficult to contact for follow-up.28 However, after adjustment with propensity score weights to minimize non-response bias, the difference in quit rates between the two groups remained significant. Another potential limitation is that quitline services were not allocated in a randomized setting. Inclusion of other covariates such as physical and mental health status may further help correct for any bias in the estimate of service effects. However, the models did include many key covariates, and strong c statistics and adjusted R2 values for both models (c¼0.75, adjusted R2¼0.24 for provider-referred clients; c¼0.81, adjusted R2¼0.39 for self-referred clients) suggest reasonable goodness of fit.
Conclusions Overall, provider-referred clients who utilized quitline services had higher odds of quitting compared to those who did not, and one in five provider-referred clients who used the quitline quit smoking. However, this quitline model also has limitations marked by lower enrollment, lower utilization, and poorer outcomes compared to self-referred smokers who are offered similar services. Changes in quitline infrastructure and protocols, including academic detailing and reduced wait times, can be made to improve enrollment, utilization, and outcomes for this special population. This study was supported in part by a cooperative agreement from CDC’s American Reinvestment and Recovery Act (ARRA) program and the Communities Putting Prevention October 2014
401
to Work (CPPW) program. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the USDHHS or CDC. The authors would like to thank Dr. Wenjun Li for providing help with data analysis, Mark Paskowsky for providing comments on earlier drafts of this paper, and the staff at both the Massachusetts Smokers’ Helpline and the Center for Tobacco Treatment Research and Training at University of Massachusetts Medical School for the quality of their work. No financial disclosures were reported by the authors of this paper.
References 1. Lichtenstein E, Glasgow RE, Lando HA, Ossip-Klein DJ, Boles SM. Telephone counseling for smoking cessation: rationales and metaanalytic review of evidence. Health Educ Res 1996;11(2):243–57. 2. Hopkins DP, Briss PA, Ricard CJ, et al. Reviews of evidence regarding interventions to reduce tobacco use and exposure to environmental tobacco smoke. Am J Prev Med 2001;20(2S):S16–S66. 3. Stead LF, Perera R, Lancaster T. Telephone counselling for smoking cessation. Cochrane Database Syst Rev 2006(3):CD002850. 4. Stead LF, Perera R, Lancaster T. A systematic review of interventions for smokers who contact quitlines. Tob Control 2007;16(1S):Si3–Si8. 5. Fiore M, Jaen CR, Baker TB. Treating tobacco use and dependence: 2008 update. Rockville MD: USDHHS, 2008. 6. Warner DD, Land TG, Rodgers AB, Keithly L. Integrating tobacco cessation quitlines into health care: Massachusetts, 2002–2011. Prev Chronic Dis [serial online] 2012;9:E133. www.ncbi.nlm.nih.gov/pmc/articles/PMC3468303/. 7. Rothemich SF, Woolf SH, Johnson RE, et al. Promoting primary care smoking-cessation support with quitlines: the QuitLink randomized controlled trial. Am J Prev Med 2010;38(4):367–74. 8. Zhu SH, Lee M, Zhuang YL, Gamst A, Wolfson T. Interventions to increase smoking cessation at the population level: how much progress has been made in the last two decades? Tob Control 2012;21(2): 110–8. 9. Greenwood DA, Parise CA, MacAller TA, et al. Utilizing clinical support staff and electronic health records to increase tobacco use documentation and referrals to a state quitline. J Vasc Nurs 2012;30(4):107–11. 10. Borland R, Balmford J, Bishop N, et al. In-practice management versus quitline referral for enhancing smoking cessation in general practice: a cluster randomized trial. Fam Pract 2008;25(5):382–9. 11. Mahabee-Gittens EM, Gordon JS, Krugh ME, Henry B, Leonard AC. A smoking cessation intervention plus proactive quitline referral in the pediatric emergency department: a pilot study. Nicotine Tob Res 2008;10(12):1745–51. 12. Fortmann SP, Killen JD. Who shall quit? Comparison of volunteer and population-based recruitment in two minimal-contact smoking cessation studies. Am J Epidemiol 1994;140(1):39–51. 13. McBride CM, Curry SJ, Grothaus LC, et al. Use of self-help materials and smoking cessation among proactively recruited and volunteer intervention participants. Am J Health Promot 1998;12(5):321–4. 14. Willett JG, Hood NE, Burns EK, et al. Clinical faxed referrals to a tobacco quitline: reach, enrollment, and participant characteristics. Am J Prev Med 2009;36(4):337–40. 15. Gordon JS, Andrews JA, Crews KM, Payne TJ, Severson HH. The 5A’s versus 3A’s plus proactive quitline referral in private practice dental offices: preliminary results. Tob Control 2007;16(4):285–8. 16. Leslie S, Thiebaud P. Using propensity scores to adjust for treatment selection bias. Proceedings of the SAS Global Forum, Statistics and
402
17. 18.
19.
20. 21.
22.
Song et al / Am J Prev Med 2014;47(4):392–402
Data analysis. Cary NC: SAS Institute Inc, 2007. www2.sas.com/ proceedings/forum2007/184-2007.pdf. Glassman AH, Helzer JE, Covey LS, et al. Smoking, smoking cessation, and major depression. JAMA 1990;264(12):1546–9. Nides MA, Rakos RF, Gonzales D, et al. Predictors of initial smoking cessation and relapse through the first 2 years of the Lung Health Study. J Consult Clin Psychol 1995;63(1):60–9. DiClemente CC, Prochaska JO, Fairhurst SK, Velicer WF, Velasquez MM, Rossi JS. The process of smoking cessation: an analysis of precontemplation, contemplation, and preparation stages of change. J Consult Clin Psychol 1991;59(2):295–304. Coleman T. Motivation, physical activity and smoking cessation. Patient Educ Couns 2010;79(2):141–2. Sheffer CE, Stitzer M, Landes R, Brackman SL, Munn T, Moore P. Socioeconomic disparities in community-based treatment of tobacco dependence. Am J Public Health 2012;102(3):e8–e16. Piper ME, Cook JW, Schlam TR, et al. Gender, race, and education differences in abstinence rates among participants in two randomized smoking cessation trials. Nicotine Tob Res 2010;12(6):647–57.
23. Baca CT, Yahne CE. Smoking cessation during substance abuse treatment: what you need to know. J Subst Abuse Treat 2009;36(2): 205–19. 24. Celestino P, Krupski L. NYS smokers’ quitline implementing a brief alcohol use intervention to boost quit success. North American Quitline Consortium Conference; 2012 Aug 13–14; Kansas City MO. 25. Bush T, Beebe L, Levine M, Hsu C. Smoking, weight and outcomes in quitline research. North American Quitline Consortium Conference; 2012 Aug 13–14; Kansas City MO. 26. Shelley D, Cantrell J. The effect of linking community health centers to a state-level smoker’s quitline on rates of cessation assistance. BMC. Health Serv Res 2010;10:25. 27. Sheffer MA, Baker TB, Fraser DL, Adsit RT, McAfee TA, Fiore MC. Fax referrals, academic detailing, and tobacco quitline use: a randomized trial. Am J Prev Med 2012;42(1):21–8. 28. An L, Betzner A, Luxenberg ML, Rainey J, Capesius T, Subialka E. Measuring quit rates. Phoenix AZ: Quality Improvement Initiative, North American Quitline Consortium, 2009.
www.ajpmonline.org
Introduction
T
he clinical and real-world effectiveness of telephone-based tobacco quitlines is well established.1–4 Smokers who use quitlines are significantly more likely to make quit attempts and quit than smokers who do not use quitlines.5 However, much of the previous research examined traditional quitline service models, in which the smoker initiates the first call to a quitline. Recently, interest has increased nationally in the expansion of a healthcare provider–referral From the Tobacco Cessation and Prevention Program (Song, Landau, Gorin) and the Massachusetts Department of Public Health (Keithly), Boston, Massachusetts Address correspondence to: Glory Song, MPH, Tobacco Cessation and Prevention Program, Massachusetts Department of Public Health, 250 Washington Street, Boston MA 02108. E-mail: [email protected]. 0749-3797/$36.00 http://dx.doi.org/10.1016/j.amepre.2014.05.043
392 Am J Prev Med 2014;47(4):392–402
model for quitlines, in which patient smokers are referred by their provider, either via electronic health record or fax, to a public quitline. It is proposed that this second quitline model, operating in complementary fashion alongside the first, could help identify and connect many more smokers to quitline services, thereby making a greater impact on the reduction of smoking prevalence compared to the self-referred model alone.6 Despite the appeal of the provider-referral system as a complementary, farther-reaching service model than the traditional self-referral model,7–9 research on the realworld impact of traditional quitline interventions (telephone counseling, nicotine replacement therapy [NRT], and self-help materials) on provider-referred smokers’ ability to quit is limited. Prior studies have found that patients referred to quitline services do quit at greater rates than those who received only standard in-practice care.10,11 However, few studies have attempted to
Published by Elsevier Inc. on behalf of American Journal of Preventive Medicine
Song et al / Am J Prev Med 2014;47(4):392–402
quantify the impact of individual quitline interventions on provider-referred smokers. It also remains an empirical question whether quitlines can expect providerreferred smokers to have similar rates of participation and quit success as traditional (self-referred) callers. Studies have found that actively recruited smokers, such as those referred by healthcare providers, differ significantly from smokers who enroll on their own on several sociodemographic and smoking-related characteristics.12,13 Specifically, compared with non-referred smokers, healthcare provider–referred patients were found to have more comorbidities, less motivation to quit, less education, and less health insurance coverage.14 Separately, other studies have found that healthcare providers’ implementation of tobacco treatment models varies significantly and that patient enrollment rates and quit outcomes for a quitline can be affected by the degree of intervention and preparation received by the patient prior to referral.15 These potential differences highlight the importance of determining the extent to which quitlines can improve the cessation outcomes of provider-referred smokers. This paper contributes to the growing body of literature on provider referrals to quitlines by examining differences in demographics, service utilization, and quit outcomes between provider-referred and self-referred clients, and by measuring the real-world impact of cessation services for provider-referred clients. The Massachusetts Smokers’ Helpline, with more than 90% of its annual client volume derived from provider referrals (both electronic and fax), offers a unique opportunity to examine the potential benefits and limitations of this referral model.
Methods Helpline Protocol Overview The Massachusetts Smokers’ Helpline offers evidence-based proactive counseling, NRT, and self-help materials. Self-referred clients enroll in services via calling the national phone number, 1-800-QUITNOW, and provider-referred clients are referred to the quitline by their provider electronically or via fax. Both groups are eligible to receive proactive counseling and self-help materials. Provider-referred clients are eligible to be screened for a 2-week supply of nicotine replacement patches. Proactive counseling consists of five individualized telephone sessions with a Helpline quit coach, although clients may receive additional sessions if they experience relapse during the course of the program.
Data Collection and Sampling A retrospective analysis was performed in 2012 of Massachusetts provider-referred and self-referred clients who entered the quitline between November 2007 and June 2011. For this analysis, individuals who called the quitline as part of any free patch October 2014
393
promotions were excluded (n¼30,276), as these individuals largely did not participate in counseling and are not representative of the average quitline caller. Figure 1 shows the exclusion criteria used to select the final client cohort for analysis. The final complete case cohort used for analysis consisted of 2,737 provider-referred clients and 530 self-referred clients. Client data for demographic, smoking behavior, source of referral, services requested, and details of counseling sessions were collected during the course of service delivery. Client data for services used, quit outcomes, and use of additional quit aids were collected via telephone-based evaluation administered 6–8 months following the time of enrollment (May 2008 to February 2012), conducted on a rolling basis over each year. For this follow-up evaluation, 84% of provider-referred clients were attempted for contact (16% either had incomplete contact information or opted out) and, owing to budget constraints, only 32% of self-referred clients were randomly sampled for attempted contact. The response rate among those sampled for follow-up was 50% for provider-referred clients and 47% for self-referred clients.
Measures The main service use variables were number of counseling sessions used (zero to five or more); use of self-help materials (yes or no); and for the provider-referred population only, use of any of the 2week supply of NRT offered through the quitline (yes or no). Additionally, a four-tiered level of service variable was created to measure the effect of combination service use (counseling with NRT, NRT only, counseling only, or neither). The primary outcome was quit status (quit or not quit) measured by client self-report to having been “currently quit” at the time of the 6–8-month follow-up. Clients were asked, Do you currently use tobacco every day, some days, or not at all? Those who responded not at all were considered “currently quit.” Demographic variables (gender, age group, race, education, and insurance status) and smoking behavior variables (when client plans to quit smoking, time to first cigarette, and number of cigarettes used per day) were included for analysis. Additionally, the self-reported use of additional medications, including NRT (patch, gum, lozenge, inhaler, or nasal spray), varenicline (Chantix) or Buproprion SR, and Zyban or Wellbutrin SR (WZB), was included for analysis. Finally, for provider-referred clients, a variable that categorized the type of institution clients were referred from (hospital, provider practice, community health center, outpatient clinic, and human services/others) was included to examine its impact on client quit outcome.
Data Analysis Pearson chi-square tests for categorical variables and t tests for continuous variables were used to identify significant differences between clients with and without follow-up in each referral population (Table 1). Wald chi-square tests for categorical variables and t-tests for continuous variables were used to identify significant differences between provider-referred and self-referred clients across demographic, smoking behavior, service utilization rates, and quit outcomes (Table 2). Only those with complete follow-up data were included for analysis (n¼3,267).
394
Song et al / Am J Prev Med 2014;47(4):392–402 Massachusetts Helpline clients who enrolled between November 2007 and June 2011 n=11,582
Clients with missing data for any covariate were excluded n=1,228 (10.6%)
Clients with follow-up data
Clients without followup data
n=3,267 (31.6%)
n=7,087 (68.4%)
Provider-referred clients with complete follow-up data
Self-referred clients with complete followup data
n=2,737
n=530
Used for baseline comparison (Table 1)
Used for data analysis (Table 2 and Table 3)
Figure 1. Exclusion criteria for final data set selection. Note: This analysis excluded clients who came through the quitline between November 2007 and June 2011 as part of a free patch promotion (n¼30,276). These individuals largely did not participate in counseling and are not representative of the average quitline caller.
To measure the odds of quitting associated with the use of quitline services and determine key predictors of quit success, two separate multivariate logistic regression models were used, one for each referral population (Table 3). In each model, the service use variables predicted client quit status, after adjusting for other covariates. Only the provider-referred model included the use of a quitline-supplied NRT variable and referral institution type variable as covariates. To test whether variation existed among clients referred from different institutions within each institution type, a mixed-effects model was constructed for the providerreferred population that treated each unique referral site as a cluster. The two referral populations were not analyzed together because of the many differences that existed between them. Unlike selfreferred clients, provider-referred clients were eligible for a 2-week supply of nicotine patches from the quitline, may have received a physician intervention prior to referral, and may have had higher prevalence of physical and mental health comorbidities. Altogether, these and other factors may differentially influence the two populations’ quit outcomes. The sampling and evaluation completion rates determined the proportion of clients in the provider-referred and self-referred populations with complete follow-up evaluation data (41% and 14%, respectively). To account for any selection bias resulting from differences between those with and without follow-up, and to improve the overall representativeness of the sample population, an inverse probability of treatment weights (IPTW) propensity score method was applied. Logistic regression was used to estimate the propensity scores for client follow-up status,
predicted from nine covariates including demographics, smoking characteristics, and referral population that were collected for everyone. The propensity score weight was then calculated as the inverse of the propensity score for each client (IPTW).16 This weight was applied to the Wald chi-square test comparisons between the two referral populations and to the two logistic regression models. Data cleaning and manipulation were performed in Microsoft ACCESS (Redmond, WA) and SAS, version 9 (SAS Institute Inc., Cary NC). All statistical testing, including hypothesis tests and propensity score regression analyses, was performed in SAS.
Results Comparison of Client Characteristics by FollowUp Status In both referral groups, compared to those followed up with, those not followed up with were more likely to be younger than age 35 years and less likely to be older than age 55 years (Table 1). The population without follow-up was also more likely to have a high school education or less. In both referral groups, compared to those with follow-up, the population without follow-up was more likely to have public insurance or no insurance. Self-referred clients without follow-up on average smoked one additional cigarette per day compared to those with follow-up (Table 1). www.ajpmonline.org
395
Song et al / Am J Prev Med 2014;47(4):392–402
Table 1. Comparison of characteristics between clients with and without follow-up, n (%) Provider-referred (n¼6,682) No follow-up (n¼3,945)
Variables
Follow-up (n¼2,737)
Self-referred (n¼3,672) p-valuea
No follow-up (n¼3,142)
Follow-up (n¼530)
p-valuea
DEMOGRAPHIC VARIABLESb Gender
0.1576
0.7710
Male
1,661 (42.1)
1,200 (43.8)
1,266 (40.3)
210 (39.6)
Female
2,284 (57.9)
1,537 (56.2)
1,876 (59.7)
320 (60.4)
o0.0001
Age group (years)
o0.0001
18–34
901 (23.1)
415 (15.3)
842 (27.1)
100 (19.0)
35–54
2,064 (52.3)
1,441 (52.6)
1,547 (49.2)
257 (48.5)
936 (24.0)
851 (31.4)
722 (23.2)
168 (32.0)
Z55 Race
0.0745
White
2,966 (76.0)
2,094 (77.4)
Black
440 (11.3)
Other
495 (12.7)
0.6966 2,620 (84.2)
441 (84.0)
319 (11.8)
204 (6.6)
39 (7.4)
294 (10.9)
287 (9.2)
45 (8.6)
o0.0001
Education
0.0095
High school or less
2,398 (61.5)
1,523 (56.3)
1,648 (53.0)
246 (46.9)
College or beyond
1,503 (38.5)
1,184 (43.7)
1,463 (47.0)
279 (53.1)
o0.0001
Insurance None
280 (7.1)
148 (5.4)
Public
2,400 (60.9) 1,265 (32.1)
Private SMOKING BEHAVIOR
o0.0001 449 (14.3)
41 (7.7)
1,553 (56.7)
1,622 (51.6)
265 (50)
1,036 (37.8)
1,071 (34.1)
224 (42.3)
b
Plan to quit
0.9326
0.3205
Currently quit
499 (12.6)
346 (12.6)
266 (8.6)
57 (10.9)
Within 1 week
787 (20)
530 (19.4)
974 (31.3)
164 (31.2)
Within 1 month
1,572 (39.9)
1,088 (39.8)
1,060 (34.1)
166 (31.6)
After 1 month or don’t know
1,074 (27.5)
759 (28.2)
811 (26.1)
138 (26.3)
Time to first cigarette (minutes)
0.4014
0.3660
r30
3,185 (81.7)
2,188 (80.9)
2,609 (83.9)
432 (82.3)
430
716 (18.4)
519 (19.2)
502 (16.1)
93 (17.7)
17.9 (10)
17.5 (9.5)
Cigarettes per day (M [SD])
0.0980
20.9 (11)
19.7 (10.8)
0.0370
Note: Boldface indicates statistical significance. a Pearson χ2 tests were used for comparison of categorical variables and two-sample t tests were used for continuous variables. b This information was collected at time of intake.
Comparison of Client Characteristics by Referral Population Compared to self-referred clients, provider-referred clients were more likely to be male, aged 35–54 years, October 2014
of black or other race, have a high school education or less, and have public insurance. Self-referred clients were more likely to be aged 18–34 years and have either no insurance or private insurance (Table 2). Self-referred
396
Song et al / Am J Prev Med 2014;47(4):392–402 a
Table 2. Weighted comparison of characteristics, service use, and quit outcomes between provider-referred and selfreferred clients with follow-up Variables
Provider-referred (n¼2,737)
Self-referred (n¼530)
p-valueb
DEMOGRAPHIC VARIABLESc Gender
0.0234
Male
1,200 (43.8)
210 (39.6)
Female
1,537 (56.2)
320 (60.4)
Age group (years)
0.0036
18–34
415 (15.3)
100 (19.0)
35–54
1,441 (52.6)
257 (48.5)
851 (31.4)
168 (32.0)
Z55 Race
0.0006
White
2,094 (77.4)
441 (84.0)
Black
319 (11.8)
39 (7.4)
Other
294 (10.9)
45 (8.6)
Education
0.0005
High school or less
1,523 (56.3)
246 (46.9)
College or beyond
1,184 (43.7)
279 (53.1)
None
148 (5.4)
41 (7.7)
Public
1,553 (56.7)
265 (50)
1,036 (37.8)
224 (42.3)
Insurance
0.0025
Private c
SMOKING BEHAVIOR
o0.0001
Plan to quit Currently quit
346 (12.6)
57 (10.8)
Within 1 week
530 (19.4)
165 (31.1)
1,088 (39.8)
168 (31.7)
759 (28.2)
138 (26.3)
Within 1 month After 1 month or don’t know Time to first cigarette (minutes)
0.3340
r30
2,188 (80.9)
432 (82.3)
430
519 (19.2)
93 (17.7)
Cigarettes per day (M [SD])
17.5 (9.5)
19.7 (10.8)
o0.0001
USE OF QUITLINE SERVICES o0.0001
Self-help materiald Any
1,472 (53.8)
338 (63.8)
None
1,265 (46.2)
192 (36.2) o0.0001
Counseling sessions Z5 4 3
125 (4.6) 91 (3.3) 108 (4)
51 (9.6) 29 (5.5) 31 (5.8) (continued on next page)
www.ajpmonline.org
397
Song et al / Am J Prev Med 2014;47(4):392–402 a
Table 2. Weighted comparison of characteristics, service use, and quit outcomes between provider-referred and selfreferred clients with follow-up (continued) Variables 2
Provider-referred (n¼2,737)
Self-referred (n¼530)
136 (5)
29 (5.5)
1
406 (14.8)
105 (19.8)
0
1,871 (68.4)
285 (53.8)
Two-weeks of NRT patch
d,e
N/A
Any
1,067 (39)
N/A
None
1,670 (61)
N/A
Levels of services used
N/A
Any counseling and NRT patch
341 (12.5)
N/A
Any NRT patch only
726 (26.5)
N/A
Any counseling only
525 (19.2)
N/A
1,145 (41.8)
N/A
No counseling or NRT patch PROVIDER REFERRAL INSTITUTION TYPE
c
N/A
Provider practice
492 (18)
N/A
Community health center
287 (10.5)
N/A
Outpatient clinic
462 (16.9)
N/A
Human services/others
222 (8.1)
N/A
1,274 (46.6)
N/A
Hospitals ADDITIONAL QUIT AIDS USE
p-valueb
d
Additional nicotine replacementf Any None
0.0003 813 (29.7)
195 (36.8)
1,924 (70.3)
335 (63.2) o0.0001
Chantix Any None
361 (13.2)
138 (26)
2,376 (86.8)
392 (74)
Wellbutrin, Zyban, or Buproprion Any None
0.0504 235 (8.6)
33 (6.2)
2,502 (91.4)
497 (93.8)
d
QUIT OUTCOME Quit status Quit Not quit
0.0004 549 (20.1)
139 (26.2)
2,188 (79.9)
391 (73.8)
Note: Boldface indicates statistical significance. a Hypothesis test for each variable was weighted using the inverse probability of treatment weights propensity score weighting method. b Wald χ2 test were used for comparison of categorical variables and two-sample t-tests were used for continuous variables. c This information was collected at time of intake. d This information was collected at the time of the 6–8-month follow-up evaluation. e Some provider-referred clients did not receive patches owing to medical ineligibility or voluntary opt-out. Among 2,737 provider-referred clients, 50% (n¼1,342) self-reported to receiving patches from the quitline. f Additional nicotine replacement includes use of gum, inhaler, lozenge, and/or patches (for provider-referred clients, this indicates patches used outside of the free 2-week supply that may have been provided by the quitline). NRT, nicotine replacement therapy patches provided by the quitline.
October 2014
398
Song et al / Am J Prev Med 2014;47(4):392–402
Table 3. Odds of quitting associated with service use and key predictors of smoking abstinence,a AOR (95% CI) Provider-referred (n¼2,737)
Self-referred (n¼530)
p-value
Predictorsb
p-value
DEMOGRAPHIC VARIABLES Age group (years) Z55
1.65 (1.28, 2.13)
o0.0001
1.72 (0.82, 3.58)
0.1496
18–34
1.45 (1.16, 1.80)
0.0010
1.09 (0.60, 1.99)
0.7698
35–54
ref
ref
ref
ref
USE OF QUITLINE SERVICES Self-help material Any None
1.25 (1.06, 1.46)
0.0062
ref
1.62 (1.01, 2.61)
0.0458
ref
ref
o0.0001
Counseling sessions Z5
3.63 (2.68, 4.92)
o0.0001
4.46 (2.23, 8.92)
4
2.63 (1.84, 3.77)
o0.0001
4.15 (1.60, 10.74)
0.0034
3
2.08 (1.48, 2.93)
o0.0001
0.83 (0.34, 2.03)
0.6815
2
1.29 (0.92, 1.82)
0.1403
1.63 (0.69, 3.88)
0.2671
1
1.01 (0.80, 1.28)
0.9202
0.62 (0.33, 1.16)
0.1312
0
ref
ref
ref
N/A
N/A
N/A
N/A
Two weeks of NRT patch Any None
1.99 (1.60, 2.48)
o0.0001
ref
Levels of services used Any counseling and NRT patch
3.45 (2.60, 4.58)
o0.001
N/A
N/A
Any NRT patch only
2.19 (1.72, 2.80)
o0.001
N/A
N/A
Any counseling only
1.73 (1.40, 2.14)
o0.001
N/A
N/A
N/A
N/A
No counseling or NRT patch
ref
PROVIDER REFERRAL TYPE Provider practice
1.40 (1.13, 1.73)
0.0023
N/A
N/A
Community health center
1.14 (0.86, 1.50)
0.3684
N/A
N/A
Outpatient clinic
1.09 (0.87, 1.36)
0.4439
N/A
N/A
Human services/others
0.78 (0.56, 1.08)
0.1413
N/A
N/A N/A
Hospitals
ref
ref
N/A
Currently quit
6.84 (5.26, 8.90)
o0.0001
3.82 (1.83, 8.00)
0.0004
Within 1 week
1.76 (1.38, 2.23)
o0.0001
0.72 (0.39, 1.32)
0.2860
SMOKING BEHAVIOR Plan to quit
(continued on next page)
www.ajpmonline.org
399
Song et al / Am J Prev Med 2014;47(4):392–402 a
Table 3. Odds of quitting associated with service use and key predictors of smoking abstinence, AOR (95% CI) (continued) Provider-referred (n¼2,737) p-value
Predictorsb Within 1 month After 1 month or don’t know
Self-referred (n¼530)
1.26 (1.02, 1.56)
0.0309
p-value 1.18 (0.66, 2.09)
ref
ref
ref
430
1.50 (1.24, 1.82)
o0.0001
2.03 (1.17, 3.52)
r30
ref
ref
ref
0.5802 ref
Time to first cigarette (minutes) 0.0117 ref
ADDITIONAL QUIT AIDS USE Additional nicotine replacementc Yes
1.48 (1.16, 1.90)
No
ref
0.0016
2.60 (1.52, 4.46)
0.0005
ref
Chantix Yes
2.21 (1.63, 3.00)
No
ref
o0.0001
5.93 (3.22, 10.92)
o0.0001
ref
Wellbutrin, Zyban, or Buproprion Yes
2.85 (1.84, 4.42)
No
ref
o0.0001
2.53 (1.14, 5.62)
0.0228
ref
Note: Boldface indicates statistical significance. a Both logistic regression models predicted quit status (quit or not quit) at the 6–8-month follow-up and were weighted using the inverse probability of treatment weights propensity score method. b Other covariates included in both models are race, gender, education level, and insurance type. These covariates were not significant at α¼0.05 for either referral model and therefore are not presented in the table. c Additional nicotine replacement includes use of gum, inhaler, lozenge, and/or patches (for provider-referred clients, this indicates patches used outside of the free 2-week supply that may have been provided by the quitline). NRT, nicotine replacement therapy patches provided by the quitline.
clients were also more likely to be ready to quit within a week, whereas provider-referred clients were more likely to be ready to quit within a month. There was no significant difference in time to first cigarette, although self-referred clients on average smoked two more cigarettes per day than provider-referred clients (Table 2).
Service Use and Quit Outcomes by Referral Population Thirty-nine percent of provider-referred clients self-reported using the 2-week quitline-supplied NRT (Table 2). Compared to self-referred clients, provider-referred clients were less likely to use self-help materials (54% vs 64%) and any counseling sessions (32% vs 46%). They were also less likely to use additional NRT (30% vs 37%) and Chantix (13% vs 26%). Twenty percent of provider-referred clients quit compared to 26% of self-referred clients who quit. October 2014
Odds of Quitting Associated with Service Use Provider-referred clients who read the self-help materials had 1.2 times the odds of quitting compared with those who did not (Table 3). The effect of counseling approximated a dose–response relationship, where increased number of sessions was positively associated with higher odds of quitting. Clients who used any of the 2-week quitline-supplied NRTs had twice the odds of quitting than those who did not. Provider-referred clients who used only counseling and no quitline-supplied NRT experienced 1.7 times the odds of quitting, and those who only used quitline-supplied NRT and no counseling experienced 2.2 times the odds of quitting compared to clients who used neither. However, clients who used counseling together with NRT experienced the greatest success, having 3.4 times the odds of quitting compared to those who used neither service (Table 3). Self-referred clients who used self-help materials had 1.6 times the odds of quitting compared to non-users. Compared to provider-referred clients who, on average,
400
Song et al / Am J Prev Med 2014;47(4):392–402
saw significantly improved outcomes after three counseling sessions, self-referred clients saw improved outcomes only after four sessions (Table 3).
Predictors of Abstinence Among ProviderReferred Clients Among provider-referred clients, those aged 35–64 years had the lowest odds of quitting. By comparison, those aged 18–34 years and 65 years and older had 1.5 and 1.6 times the odds of quitting, respectively (Table 3). Readiness to quit was also an important predictor of quit outcome. Clients ready to quit within a week and within a month had 1.8 times and 1.3 times, respectively, the odds of quitting compared to those who planned to quit after one month or “didn’t know.” Those “currently quit” at time of enrollment had 6.8 times the odds of quitting. However, more than 50% of these clients had relapsed by the time of the 6–8 month follow-up (data not shown). Tobacco dependence was another important predictor; those who were able to wait more than 30 minutes before smoking their first cigarette had 1.5 times the odds of quitting compared to those who first smoked within 30 minutes after waking (Table 3). Those who used additional quit aids and medication experienced greater success, with clients who used nicotine replacement aids having 1.5 times the odds of quitting than non-users, those who used Chantix having 2.2 times the odds of quitting than non-users, and those who used Wellbutrin, Zyban, or Buproprion having 2.8 times the odds of quitting than non-users (Table 3). Finally, private practice–referred clients had 1.4 times the odds of quitting compared to hospital-referred clients. No other difference between comparison pairs was significant. Results from the mixed-effects model showed that the different referral institutions clustered within an institution type explained little of the remaining variation in residuals after accounting for the other independent variables in the model (variance component¼0.007, p¼0.372).
Predictors of Abstinence Among Self-Referred Clients Self-referred clients who could wait more than 30 minutes before smoking their first cigarette had twice the odds of quitting compared to those who smoked within 30 minutes after waking (Table 3). Self-referred clients who used nicotine replacement aids had 2.6 times the odds of quitting than non-users; those who used Chantix had 5.9 times the odds of quitting than non-users; and those who used Wellbutrin, Zyban, or Buproprion had 2.5 times the odds of quitting than non-users (Table 3).
Discussion This study found that provider-referred smokers who used quitline services had greater odds of quitting than those who did not. Those who used a combination of counseling and NRT had the greatest odds of quitting. Overall, provider-referred clients were more likely to be non-white, have less education, and have public insurance and were less ready to quit than self-referred clients. Provider-referred clients, having recently visited a provider, may also have been more likely to have physical or mental health comorbidities. All of these characteristics are associated with poorer cessation outcomes,17–20 and this was observed in lower rates of service utilization and quit success among provider-referred clients. These results generate a number of questions about quitline and provider roles under a provider-referral model. First, how might a quitline address the socioeconomic, motivational, and comorbid health concerns more often associated with provider-referred patients? Findings from the literature indicate that lower-SES smokers may need to receive more treatment content and have access to more intensive pharmacotherapy.21,22 Although offering medication or more services may not be feasible for public quitlines, any quitline could incorporate scripts that inform clients of available cessation coverage by insurance plans and encourage clients to utilize their benefits. To address readiness to quit among providerreferred clients, quitlines could aim to minimize the wait time between provider referral and actual service provision to the client. Increased wait time before receipt of service can increase the room for ambivalence for any smoker, but may be especially detrimental for provider-referred smokers who are less ready to make a quit attempt.23 Finally, to address disease comorbidities that are often associated with provider-referred smokers, several quitlines have piloted the integration of disease and risk behavior management into their quitline counseling.24,25 However, more research is needed regarding the effectiveness and sustainability of these programs. Second, what is a provider’s role in a referral model, both in assessing patient readiness to quit and in preparing patients for quitline services? In reality, not all patients have received an evidence-based intervention or are ready to quit at the time of enrollment in quitline services. When patients who are not ready to quit are referred, many are lost in the callback process and do not engage with the quitline at all. The Massachusetts reach rate for referred patients using three callback attempts is approximately 40%, and 50% using five callback attempts (data not shown). Similar rates have been observed by other quitlines.11,14,26 The difficulty reaching provider-referred clients highlights the critical need for effective and frequent www.ajpmonline.org
Song et al / Am J Prev Med 2014;47(4):392–402
provider training, outreach, feedback reporting, clinical champions on site, and systems support to ensure that providers can better assess and equip patients for quitline participation to improve the quality of referrals.27
Limitations The low proportion of clients in both populations with complete follow-up data raises concerns about potential selection bias that may affect the representativeness of the study sample to the overall population. A propensity score weighting method was used to minimize this effect. Results from the weighted regression models and unweighted regression models (not shown) were very similar. Any difference between the study sample and those lost to follow-up likely did not affect the study results and should not account for the associations that were observed. Therefore, the observed associations likely reflect the true associations in the population. The lower response rate among self-referred compared to provider-referred (47% vs 50%, p¼0.0174) clients may potentially have yielded an overestimation of the quit rate for self-referred clients, as those with poorer outcomes are more difficult to contact for follow-up.28 However, after adjustment with propensity score weights to minimize non-response bias, the difference in quit rates between the two groups remained significant. Another potential limitation is that quitline services were not allocated in a randomized setting. Inclusion of other covariates such as physical and mental health status may further help correct for any bias in the estimate of service effects. However, the models did include many key covariates, and strong c statistics and adjusted R2 values for both models (c¼0.75, adjusted R2¼0.24 for provider-referred clients; c¼0.81, adjusted R2¼0.39 for self-referred clients) suggest reasonable goodness of fit.
Conclusions Overall, provider-referred clients who utilized quitline services had higher odds of quitting compared to those who did not, and one in five provider-referred clients who used the quitline quit smoking. However, this quitline model also has limitations marked by lower enrollment, lower utilization, and poorer outcomes compared to self-referred smokers who are offered similar services. Changes in quitline infrastructure and protocols, including academic detailing and reduced wait times, can be made to improve enrollment, utilization, and outcomes for this special population. This study was supported in part by a cooperative agreement from CDC’s American Reinvestment and Recovery Act (ARRA) program and the Communities Putting Prevention October 2014
401
to Work (CPPW) program. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the USDHHS or CDC. The authors would like to thank Dr. Wenjun Li for providing help with data analysis, Mark Paskowsky for providing comments on earlier drafts of this paper, and the staff at both the Massachusetts Smokers’ Helpline and the Center for Tobacco Treatment Research and Training at University of Massachusetts Medical School for the quality of their work. No financial disclosures were reported by the authors of this paper.
References 1. Lichtenstein E, Glasgow RE, Lando HA, Ossip-Klein DJ, Boles SM. Telephone counseling for smoking cessation: rationales and metaanalytic review of evidence. Health Educ Res 1996;11(2):243–57. 2. Hopkins DP, Briss PA, Ricard CJ, et al. Reviews of evidence regarding interventions to reduce tobacco use and exposure to environmental tobacco smoke. Am J Prev Med 2001;20(2S):S16–S66. 3. Stead LF, Perera R, Lancaster T. Telephone counselling for smoking cessation. Cochrane Database Syst Rev 2006(3):CD002850. 4. Stead LF, Perera R, Lancaster T. A systematic review of interventions for smokers who contact quitlines. Tob Control 2007;16(1S):Si3–Si8. 5. Fiore M, Jaen CR, Baker TB. Treating tobacco use and dependence: 2008 update. Rockville MD: USDHHS, 2008. 6. Warner DD, Land TG, Rodgers AB, Keithly L. Integrating tobacco cessation quitlines into health care: Massachusetts, 2002–2011. Prev Chronic Dis [serial online] 2012;9:E133. www.ncbi.nlm.nih.gov/pmc/articles/PMC3468303/. 7. Rothemich SF, Woolf SH, Johnson RE, et al. Promoting primary care smoking-cessation support with quitlines: the QuitLink randomized controlled trial. Am J Prev Med 2010;38(4):367–74. 8. Zhu SH, Lee M, Zhuang YL, Gamst A, Wolfson T. Interventions to increase smoking cessation at the population level: how much progress has been made in the last two decades? Tob Control 2012;21(2): 110–8. 9. Greenwood DA, Parise CA, MacAller TA, et al. Utilizing clinical support staff and electronic health records to increase tobacco use documentation and referrals to a state quitline. J Vasc Nurs 2012;30(4):107–11. 10. Borland R, Balmford J, Bishop N, et al. In-practice management versus quitline referral for enhancing smoking cessation in general practice: a cluster randomized trial. Fam Pract 2008;25(5):382–9. 11. Mahabee-Gittens EM, Gordon JS, Krugh ME, Henry B, Leonard AC. A smoking cessation intervention plus proactive quitline referral in the pediatric emergency department: a pilot study. Nicotine Tob Res 2008;10(12):1745–51. 12. Fortmann SP, Killen JD. Who shall quit? Comparison of volunteer and population-based recruitment in two minimal-contact smoking cessation studies. Am J Epidemiol 1994;140(1):39–51. 13. McBride CM, Curry SJ, Grothaus LC, et al. Use of self-help materials and smoking cessation among proactively recruited and volunteer intervention participants. Am J Health Promot 1998;12(5):321–4. 14. Willett JG, Hood NE, Burns EK, et al. Clinical faxed referrals to a tobacco quitline: reach, enrollment, and participant characteristics. Am J Prev Med 2009;36(4):337–40. 15. Gordon JS, Andrews JA, Crews KM, Payne TJ, Severson HH. The 5A’s versus 3A’s plus proactive quitline referral in private practice dental offices: preliminary results. Tob Control 2007;16(4):285–8. 16. Leslie S, Thiebaud P. Using propensity scores to adjust for treatment selection bias. Proceedings of the SAS Global Forum, Statistics and
402
17. 18.
19.
20. 21.
22.
Song et al / Am J Prev Med 2014;47(4):392–402
Data analysis. Cary NC: SAS Institute Inc, 2007. www2.sas.com/ proceedings/forum2007/184-2007.pdf. Glassman AH, Helzer JE, Covey LS, et al. Smoking, smoking cessation, and major depression. JAMA 1990;264(12):1546–9. Nides MA, Rakos RF, Gonzales D, et al. Predictors of initial smoking cessation and relapse through the first 2 years of the Lung Health Study. J Consult Clin Psychol 1995;63(1):60–9. DiClemente CC, Prochaska JO, Fairhurst SK, Velicer WF, Velasquez MM, Rossi JS. The process of smoking cessation: an analysis of precontemplation, contemplation, and preparation stages of change. J Consult Clin Psychol 1991;59(2):295–304. Coleman T. Motivation, physical activity and smoking cessation. Patient Educ Couns 2010;79(2):141–2. Sheffer CE, Stitzer M, Landes R, Brackman SL, Munn T, Moore P. Socioeconomic disparities in community-based treatment of tobacco dependence. Am J Public Health 2012;102(3):e8–e16. Piper ME, Cook JW, Schlam TR, et al. Gender, race, and education differences in abstinence rates among participants in two randomized smoking cessation trials. Nicotine Tob Res 2010;12(6):647–57.
23. Baca CT, Yahne CE. Smoking cessation during substance abuse treatment: what you need to know. J Subst Abuse Treat 2009;36(2): 205–19. 24. Celestino P, Krupski L. NYS smokers’ quitline implementing a brief alcohol use intervention to boost quit success. North American Quitline Consortium Conference; 2012 Aug 13–14; Kansas City MO. 25. Bush T, Beebe L, Levine M, Hsu C. Smoking, weight and outcomes in quitline research. North American Quitline Consortium Conference; 2012 Aug 13–14; Kansas City MO. 26. Shelley D, Cantrell J. The effect of linking community health centers to a state-level smoker’s quitline on rates of cessation assistance. BMC. Health Serv Res 2010;10:25. 27. Sheffer MA, Baker TB, Fraser DL, Adsit RT, McAfee TA, Fiore MC. Fax referrals, academic detailing, and tobacco quitline use: a randomized trial. Am J Prev Med 2012;42(1):21–8. 28. An L, Betzner A, Luxenberg ML, Rainey J, Capesius T, Subialka E. Measuring quit rates. Phoenix AZ: Quality Improvement Initiative, North American Quitline Consortium, 2009.
www.ajpmonline.org
