1
1
Running Head: INTERPERSONAL COMMUNICATION AND SMOKING CESSATION
2
Forthcoming in Journal of Health Communication
3 4 5 6
Interpersonal Communication and Smoking Cessation in the Context of an Incentive-Based Program: Survey Evidence from a Telehealth Intervention in a Low-Income Population
7 8
Michael J. Parks, PhD1
9
Jonathan S. Slater, PhD1,2
10
Alexander J. Rothman, PhD2
11
Christina L. Nelson, MPH1
12 13 14 15
1
Minnesota Department of Health, Saint Paul, MN, USA 2
University of Minnesota, Minneapolis, MN, USA
16 17
Corresponding author: Michael J. Parks, Minnesota Department of Health, 85 East 7th Street,
18
Saint Paul, MN 55164, USA. Email:
[email protected]. Phone: 651-201-5285.
19
Acknowledgements: Project funded through Centers for Disease Control and Prevention
20
(American Recovery and Reinvestment Act; Patient Protection and Affordable Care Act). We
21
thank QUITPLAN® Helpline staff, Shelly Madigan, Sage patient navigators, Janis Taramelli, and
22
Michelle Waste for their efforts. The authors declare that they have no conflict of interest.
23 24
Keywords: smoking cessation; financial incentives; low-income; telehealth; interpersonal communication; population-level interventions
25 26
2
27
Abstract
28
The tobacco epidemic disproportionately affects low-income populations, and telehealth is an
29
evidence-based strategy for extending tobacco cessation services to underserved populations. A
30
public health priority is to establish incentive-based interventions at the population level in order
31
to promote long-term smoking cessation in low-income populations. Yet randomized clinical
32
trials show that financial incentives tend to encourage only short-term steps of cessation, not
33
continuous smoking abstinence. One potential mechanism for increasing long-term cessation is
34
interpersonal communication (IPC) in response to population-level interventions. However, more
35
research is needed on IPC and its influence on health behavior change, particularly in the context
36
of incentive-based, population-level programs. This study used survey data gathered after a
37
population-level telehealth intervention that offered $20 incentives to low-income smokers for
38
being connected to Minnesota’s free quitline in order to examine how perceived incentive
39
importance and IPC about the incentive-based program relate to both short-term and long-term
40
health behavior change. Results show that IPC was strongly associated with initial quitline
41
utilization and continuous smoking abstinence as measured by 30-day point prevalence rates at
42
seven-month follow-up. Perceived incentive importance had weak associations with both
43
measures of cessation, and all associations were non-significant in models adjusting for IPC.
44
These results were found in descriptive analyses, logistic regression models, and Heckman probit
45
models that adjusted for participant recruitment. In sum, a behavioral telehealth intervention
46
targeting low-income smokers that offered a financial incentive inspired IPC, and this social
47
response was strongly related to utilization of intervention services as well as continuous
48
smoking abstinence.
49
3
50
Introduction
51
Tobacco use is the leading preventable cause of mortality and morbidity in the U.S. and
52
abroad (World Health Organization, 2012). Annually, smoking is estimated to be responsible for
53
five million deaths worldwide as it is causally linked to cardiovascular disease and multiple
54
forms of cancer (McAfee, Davis, Alexander, Pechacek, & Bunnell, 2013). The tobacco epidemic
55
disproportionally affects people of low socioeconomic status. Smoking prevalence among adults
56
below the federal poverty level is 28% whereas prevalence at or above poverty level is 17%
57
(CDC, 2014). Disproportionately high smoking rates persist among low-income women (Stewart
58
et al., 2010), and it is estimated that smoking accounts for up to half of male mortality disparities
59
associated with low socioeconomic status in countries such as the U.S. (Jha et al., 2006).
60
Consequently, increasing smoking cessation within low socioeconomic groups can save millions
61
of lives and decrease mortality disparities (Holford et al., 2014; Jha et al., 2006; Thomas et al.,
62
2008).
63
There is a critical need to develop population-level smoking cessation programs for low-
64
income populations that go beyond clinic-based settings as clinics have limited access to low-
65
income populations who underutilize preventive services, are geographically isolated, and
66
inadequately insured (see Chokshi & Farley, 2014; Bryant et al., 2011; Wilson, 1987).
67
Telehealth, or the use of “telecommunications and information technology to provide access to
68
health assessment, diagnosis, intervention, consultation, supervision, education, and information
69
across distance” (Nickelson, 1998: 527), is effective for delivering health services to underserved
70
and low-income populations (Wootton, Jebamani, & Dow, 2005; McBride & Rimer, 1999). Free
71
state telephone tobacco quitlines are exemplars of telehealth’s potential. Free telephone quitlines
4
72
are evidence-based techniques for increasing smoking abstinence rates, particularly in low-
73
income and non-White populations (Stead, Perera, & Lancaster, 2007; Burns, Deaton, &
74
Levinson, 2011; Fiore et al., 2008), but utilization rates are low across the U.S. (Zhu, Lee,
75
Zhuang, Gamst, & Wolfson, 2012).
76
Another promising strategy for extending health services to underserved populations is
77
financial incentives (Oliver, 2009). Designing incentive-based, population-level interventions in
78
order to sustain long-term changes in health behaviors like smoking cessation has become a
79
public health priority, particularly in low-income populations (Blumenthal et al., 2013). This is
80
exemplified by the Affordable Care Act’s section 4108 and the Centers for Medicare and
81
Medicaid Services’ authority to provide grants to states to test the effectiveness of incentives in
82
improving health behaviors such as tobacco cessation (Blumenthal et al., 2013: 497-498).
83
Financial incentives have been shown to increase health enhancing behaviors (e.g., Slater et al.,
84
2005; Gneezy, Meier, & Rey-Biel, 2011), and low-income smokers tend to be responsive to
85
incentives (Bryant et al., 2011). However, incentives tend to be effective for encouraging
86
preventive care that requires only a single activity or “simple” behavior but not “complex”
87
actions that require additional engagement beyond initial intervention services (see Kane,
88
Johnson, Town, & Butler, 2004). Smoking cessation requires prolonged engagement (Prochaska,
89
DiClemente, & Norcross, 1992), and while incentives can influence short-term cessation in the
90
context of a clinical trial (Volpp et al., 2009; Sigmon & Patrick, 2012) less is known about how
91
incentives can encourage or be used to promote long-term cessation in population-level
92
interventions within low-income populations (Blumenthal et al., 2013).
5
93
Research in behavioral economics demonstrates that financial incentives typically have
94
an effect on short-term behavior change through the “direct price effect” that makes incentivized
95
behaviors more attractive (Gneezy et al., 2011). The direct price effect is often explained by its
96
influence on extrinsic motivation to engage in a target behavior and it tends to be less effective
97
for promoting behavior maintenance once incentives are removed. A second type of effect,
98
known as the “indirect psychological effect,” can have both positive and negative influences on
99
incentivized behaviors (Gneezy et al., 2011: 11). Multiple social and psychological factors can
100
alter the effects of a financial incentive presentation in terms of promoting or unintentionally
101
discouraging the incentivized behavior (see e.g., Babcock, Bedard, Charness, Hartman, & Royer,
102
2011).
103
Two specific factors are particularly relevant for population-level interventions, and they
104
have implications for both short-term and long-term behavior change. The first is that the
105
incentive’s effect depends on the context within which it is presented. If presented in a private
106
context within a controlled environment (e.g., randomized clinical trials), social influences are
107
limited, such as informal social control or social image processes. Alternatively, incentives
108
offered in more public settings can activate social influences (Gneezy et al., 2011). Public
109
settings are marked by the presence of others, and this can influence individuals’ behavior by
110
inspiring social responses to incentives instead of strictly monetary responses. For example,
111
social reactions such as interpersonal discussions about the incentive can activate behavior based
112
on social image maintenance or prevailing social norms (Gneezy et al., 2011: 11). A second and
113
related factor is that incentives can interact with social networks. Evidenced by past research on
114
exercise maintenance, incentives can not only increase exercise, but social groups (e.g., friends)
115
who jointly receive incentives are more likely to exhibit long-term behavior maintenance
6
116
(Charness & Gneezy, 2009; Babcock et al., 2011). Addressing the relational context of health
117
behaviors through network-oriented strategies has become a priority for prevention programs
118
(Gest, Osgood, Feinberg, Bierman, & Moody, 2011), and this is especially true for smoking
119
cessation interventions because social dynamics operating within networks strongly influence the
120
cessation process (Christakis & Fowler, 2008; Valente, 2010; Burns, Rothman, Fu, Lindgren, &
121
Joseph, 2014). Yet less is known about how these social mechanisms may influence long-term
122
health behavior change in response to population-level interventions, and particularly incentive-
123
based interventions.
124
In sum, randomized clinical trials demonstrate that incentive-based interventions can
125
encourage short-term behavioral engagement involving an incentivized behavior, and past
126
behavioral economic research implies that financial incentives offered in population-level
127
interventions can potentially encourage social responses and indirect effects due to the context of
128
the intervention. Population-level interventions occur in relatively public settings in conjunction
129
with naturally occurring psychosocial dynamics, and these processes could have implications for
130
both short-term and long-term behavior change. In terms of long-term behavior change such as
131
sustained smoking cessation, a social response to an incentive-based program at the population-
132
level could be a marker of naturally occurring social mechanisms that have inherent capacity to
133
promote sustained behavioral engagement.
134
Past population-level smoking cessation interventions provide empirical support (see e.g.,
135
McAfee et al., 2013; Donovan, Boulter, Borland, Jalleh, & Carter, 2003; van den Putte, Yzer,
136
Southwell, de Bruijn, & Willemsen, 2011). Smoking cessation programs have demonstrated that
137
long-term smoking cessation is indirectly influenced by social reactions measured by
7
138
interpersonal communication about the smoking cessation program (van den Putte, et al., 2011).
139
Supported by communication theories (see Katz & Lazarsfeld, 1955; Southwell & Yzer, 2007),
140
interpersonal communication about a health promotion or prevention program can catalyze
141
mechanisms embedded within social interactions, potentially cultivating intentions to change
142
behavior, activating normative pressure, and making health behavior decisions more cognitively
143
salient (see e.g., Montano & Kasprzyk, 2008; Southwell, 2013). Yet further examination of
144
interpersonal communication and health behavior change in response to population-level
145
interventions remains an important gap in research, particularly in the context of incentive-based
146
programs (van den Putte, et al., 2011; Southwell, Slater, Nelson, & Rothman, 2012).
147
The Current Study
148
This paper reports results of a survey conducted after a population-level intervention that
149
offered financial incentives to low-income smokers for being connected to Minnesota’s free
150
QUITPLAN® Helpline. The survey captured self-report measures of interpersonal
151
communication about the incentive-based program as well as perceived importance of the
152
incentive for the incentivized behavior (i.e., being connected to a quitline). The analysis focuses
153
on the cessation process measured by short-term and long-term behavioral engagement. Using
154
North American Quitline Consortium (NAQC) evidence-based measures (see NAQC, 2011) the
155
cessation process is measured by (1) initial utilization of quitline services, and (2) continuous
156
smoking abstinence measured by 30 consecutive smoke-free days at seven-month follow-up. The
157
main focus of the analysis is on how these two measures of the cessation process relate to both
158
interpersonal communication and perceived incentive importance.
8
159
Based on past research on the effects of interpersonal communication in response to
160
smoking cessation interventions, we hypothesize that interpersonal communication in the context
161
of the current incentive-based telehealth intervention will be related to utilization of smoking
162
cessation services as well as subsequent efforts to quit smoking (i.e., seven-month follow-up 30-
163
day point prevalence rates). This hypothesis is also based on past research demonstrating that
164
incentives presented in more public settings can inspire social and psychological mechanisms,
165
and that these mechanisms are potentially conducive to long-term behavioral engagement. In
166
light of past research on financial incentives and smoking cessation in randomized clinical trials
167
as well as the direct price effect of incentives on incentivized behaviors, we note that incentives
168
are more likely to influence short-term behavior change rather than long-term smoking cessation.
169
Therefore, we hypothesize that perceived incentive importance will relate to initial steps of
170
cessation measured by utilization of quitline services, and that perceived incentive importance
171
will be less likely to impact continuous smoking abstinence.
172 173 174
Method Participants and Procedures Intervention context. From September 2010 to September 2012, a behavioral telehealth
175
intervention was implemented through Minnesota’s National Breast and Cervical Cancer Early
176
Detection Program (NBCCEDP) called “Sage.” Sage serves low-income individuals
177
experiencing health-related disparities. Specifically, Sage provides free cancer screening services
178
to individuals who are ages 40 years or older, have household incomes at or below 250% of the
179
federal poverty level, and are inadequately insured. Unique among NBCCEDPs, Sage has a call
180
center staffed by “patient navigators” (see Freund et al., 2008) who are fluent in English and
9
181
other languages relevant for Sage’s target population. More details on NBCCEDPs and Sage
182
have been published elsewhere (Lee et al., 2014; Slater et al., 2005).
183
Intervention. The current intervention offered a $20 incentive for being connected to
184
Minnesota’s QUITPLAN® Helpline (QL) via a three-way phone call conducted by Sage patient
185
navigators. Relying on participant-initiated phone contact (see Soet & Basch, 1997), we
186
employed two recruitment strategies: (1) a direct mail (DM) campaign and (2) an opportunistic
187
referral with connection (ORC) call transfer system from within the Sage Call Center.
188
Individuals identified as smokers in Sage’s database were recruited using DM mailers designed
189
to prompt cigarette smokers to call Sage’s toll-free phone number. The DM mailers consisted of
190
a folded card with emotionally evocative messages and graphics as well as a small insert card
191
advertising the financial incentive. The mailers were strategically constructed based on past
192
research (see Rothman & Salovey, 1997) and included two rounds of mailings (see Slater et al.,
193
2005).
194
For ORC recruitment, Sage’s patient navigators obtained the smoking status of
195
individuals who called the Sage Call Center for cancer screening information or appointments.
196
Callers identified as smokers were then opportunistically presented the QL referral offer (i.e., a
197
$20 incentive for being connected to the QL). Patient navigators handled phone calls for both
198
intervention groups, and once participants agreed to be transferred, patient navigators put
199
participants on hold and called QL operators. Patient navigators then confirmed with QL
200
operators that an individual from Sage would be connected to the QL via three-way calls. Once
201
QL operators agreed, patient navigators remained on the line until QL operators and intervention
10
202
participants were connected and communication between participants and QL operators was
203
confirmed.
204
QL services included free smoking cessation telephone counseling with a maximum of
205
five sessions within a six-month period. The QL provided self-help materials and free nicotine
206
replacement therapy for those who requested them. A total of 5,420 callers were offered the
207
intervention from the Sage Call Center (DM=870, ORC=4,550), with 2,456 completing QL
208
transfers. Of those transferred, 66% were ORC (N=1,612) compared to DM (N=844).
209
Survey data. Participants who completed QL connections were interviewed via
210
telephone by trained staff at least seven months after each participant’s QL connection in
211
compliance with recommended, evidence-based practices (see NAQC, 2011). All participants
212
who completed QL connections were targeted for the survey; 10 call attempts were made before
213
participants were deemed unresponsive. A total of 1,218 participants completed the survey.
214
Since not all participants were offered QL services after being connected by Sage patient
215
navigators, the current analyses focused on individuals who were offered QL services (N=995)
216
with complete data, comprising an analytic sample of 970. Individuals not offered services were
217
excluded because they were not subsequently asked about smoking cessation in the survey (e.g.,
218
utilization of QL services). No significant differences were found between individuals offered
219
and not offered services in terms of demographic and smoking characteristics, except for age
220
(M=53.7 vs. M=52.2, p=.01). Of those not offered services, about 40% reported it was due to
221
their insurance coverage, and 60% reported miscellaneous reasons associated with program
222
fidelity (e.g., accidental disconnection after patient navigator transfer, reported not receiving a
223
call back from QL).
11
224
Measures. Following past research, two dichotomous outcomes for short- and long-term
225
behavioral steps associated with smoking cessation were employed: (1) QL utilization and (2)
226
being smoke-free for 30 consecutive days at seven-month follow-up (see NAQC, 2011;
227
DiClement et al., 1991). For QL utilization, respondents were asked whether they had used the
228
tools and services that were offered from the QL (1=yes, 0=no). For continuous smoking
229
abstinence (“continuous cessation”), participants were asked if they had not smoked for 30
230
consecutive days (in interviews conducted at least seven months after being connected to the QL)
231
(1=yes; 0=no).
232
Incentive influence was captured by measuring whether participants reported the
233
incentive to be important for their QL connection (1=important, 0=not important).1 Interpersonal
234
communication about the incentive-based program was measured as whether or not participants
235
“told others about the Sage offer that rewards smokers $20 for being connected with a tobacco
236
quitline through the Sage Call Center” (1=yes, 0=no).
237
Background characteristics included age, sex, education, race/ethnicity, and smoking
238
history. Age was measured in continuous years; sex and race/ethnicity were dichotomous
239
measures (1=male, 0=female; 1=white, 0=non-white). Education was an ordinal scale, ranging
240
from 1 (completed 8th grade or less) to 6 (graduate school). Smoking characteristics included a
241
continuous measure of years smoked, whether or not participants smoked on a daily basis 1
Empirically, we tested the credibility of this measure by comparing answers to the question about how important the incentive was for the individual to an additional question that asked how important the incentive was for the Minnesota Department of Health to get smokers to connect to the QL in general. We created a scale out of the two measures, which had a reliability measure of .901. Results using this alternative measure were not different from main analyses that utilized the dichotomous measure, and therefore we only report the results for the dichotomous measure described here.
12
242
(1=yes, 0=no), whether they lived with a smoker (1=yes, 0=no), whether they made a quit
243
attempt in the past with medication (1=yes, 0=no), and whether they were unlikely to contact the
244
QL without the current intervention (1=yes, 0=no). Variables used to assess recruitment
245
differences were (1) a dichotomous measure for DM vs ORC, and (2) a measure of whether the
246
participant had past involvement with Sage (1=Sage, 0=non-Sage). Approximately 65% of the
247
analytic sample was recruited via ORC (ORC=629; DM=341), and 62% had previous contact
248
with Sage.
249
Analytic strategy. The first set of analyses of the analytic sample of surveyed
250
respondents consisted of descriptive statistics and cross-tabulations. These descriptive analyses
251
focused on bivariate relationships for (1) both cessation outcomes and interpersonal
252
communication, and (2) both cessation outcomes and incentive importance. We then explored
253
these relationships more systematically via logistic regression models for each dichotomous
254
outcome. It is important to note that DM and ORC recruitment groups likely differed in their
255
behavioral stage or “readiness” in terms of smoking cessation. This is because the DM group
256
called Sage seeking smoking cessation services whereas the ORC group called regarding
257
unrelated health matters and were unexpectedly offered cessation services. As a result, we ran
258
logistic regression models that included a dummy variable for DM versus ORC group
259
membership, as well as supplementary analyses that consisted of a sub-sample analysis that
260
adjusted for differences between DM and ORC groups in terms of their likelihood of group
261
membership (see Heckman, 1979; Bushway, Johnson, & Slocum, 2007).
262 263
The supplementary analysis consisted of maximum likelihood bivariate probit with selection models (Heckman probit models). In Heckman probit models, recruitment groups were
13
264
used as a bivariate selection criterion (DM vs. ORC)—“outcome” models in Heckman probit
265
models censored unobserved objects from the “selection” model (i.e., ORC individuals). This
266
generated an analysis of QL utilization and 30-day point prevalence quit status for a subgroup
267
potentially more prepared for smoking cessation (i.e., DM) while adjusting for the likelihood of
268
DM membership (see Puhani, 2000; Bushway et al., 2007). Outcome models are presented in
269
table form and within the text, but selection models are not reported (selection models included
270
all variables in outcome models except for an exclusion restriction variable). Multivariate
271
logistic regression and Heckman probit models were run in Stata, version 12. Cluster-adjusted
272
robust standard errors were used in both logistic regression and Heckman probit models in order
273
to account for clustering within patient navigators as well as for heteroscedasticity (Bushway et
274
al., 2007).
275 276 277
Results Participants Participant characteristics are detailed in Table 1. The analytic sample was
278
overwhelmingly low-income female smokers because low-income, inadequately insured women
279
are Sage’s target population. It included a limited number of males primarily as a result of
280
information sharing about the intervention.
281
The sample was primarily White (75%); 15% were African American, 5% were Native
282
American, 2% were Hispanic, and the remaining were other races or ethnicities (3%), providing
283
evidence that the intervention disproportionately reached racial and ethnic minorities in
284
Minnesota (U.S. Census Bureau, 2012), which is also Sage’s target population. Average
14
285
education level was a high school diploma or equivalent. About 12% of the sample did not
286
complete high school and all individuals met income guidelines for receiving Sage services,
287
reflecting the relatively low socioeconomic status of the sample.
288
The majority of participants were intense smokers. Years smoked was normally
289
distributed with an average of 30.94. Over 50% had made a quit attempt with medication in the
290
past, 94% smoked every day, and about 38% lived with a smoker. Most participants were
291
unlikely to have contacted a QL without having been reached with the current intervention
292
(72%).
293 294 295
[TABLE 1 ABOUT HERE] Survey Results Of the surveyed participants who were offered QL services, 643 individuals (66%) were
296
connected to the QL and utilized the services, such as receiving telephone counseling or self-help
297
materials. In terms of 30-day point prevalence quit rates, about 19% reported continuous
298
cessation seven months after QL connection (184 individuals). About 50% engaged in
299
interpersonal communication about the incentive-based program (IPC), and almost 69% noted
300
the incentive was important for their QL connection (see Table 1).
301
QL utilization. The associations between QL utilization and IPC, and between utilization
302
and incentive importance, are displayed in the left-hand portion of Table 2. Interpersonal
303
communication about the incentive-based program was associated with QL utilization (2=32.32;
304
p