Factorial Design to Assess Quit-line Connections

Individuals were randomized to one of six conditions. The design was a two-factor, three-by-two factorial design: three levels of financial incentives, and two levels of proactive calls (yes, no). All groups received direct mail and free tobacco treatment services.

The intervention was implemented from June 2017 to October 2017. National Breast and Cervical Cancer Early Detection Programs serve uninsured women with household incomes at or below 250% of the U.S. federal poverty level; subsequently, the sample was strictly female, and the investigators included individuals for whom self-reported smoking status was available (from 2014 to 2017). Smoking status was determined using Sage data, which come from clinics and Sage's call center. Using a uniform measure ("Does the participant currently smoke cigarettes?"), smoking status was based on cigarette use at the time of contact with the Sage program (i.e., clinic visit or contact with the call center). Data on smoking intensity were not available. There were 3,723 smokers who met selection criteria.

All participants received two rounds of direct mail. Mailers consisted of a folded card, which contained messages and graphics rooted in a loss-frame approach. Following previous research, the loss-frame approach was designed to indicate that certain behaviors lead to unhealthy outcomes, and the investigators paired the message with a clear articulation of behavioral steps.

Sage has a call center staffed by patient navigators. When participants called Sage's number, patient navigators followed a script, recorded callers' promotion code, determined callers' desire to participate, and made connections to the quit-line for willing participants. After two rounds of direct mail, patient navigators placed proactive outreach calls to eligible participants. Phone numbers were retrieved from the Sage database. Using a script, navigators reminded participants of the direct mail piece and the program offer. Each participant received one proactive call. One call attempt was made; navigators left a voicemail for participants who were unavailable. Proactive calling lasted three months.

Both the mailer and navigators referred participants to Minnesota's free population-based cessation services. The quit-line is funded and administered by an independent nonprofit organization. Participants can enroll in services either by phone or online. Tobacco users can choose telephone counseling or one or more individual services.

Our primary outcome was phone-only treatment connection, which was defined as a confirmed connection between participants and quit-line staff via three-way calls conducted by patient navigators. For willing participants, patient navigators called quit-line staff and confirmed that a patient would be connected via a three-way call. Navigators remained on the line until communication between participants and quit-line staff was established. Once connected, trained operators from the quit-line were responsible for administering cessation-related services.

A secondary measure captured connections to online services. This measure combined both phone and web connections. The web connection assessed whether participants entered their individual promotion codes (provided via direct mail) on a unique online landing page. Once participants entered their individual codes, they were directly connected to the cessation webpage where they entered their personal information and chose tailored treatment services. Direct transfers to the online services counted as a connection; each code was counted only once. The investigators used the confirmed phone connection as our primary outcome as it was a more conservative measure of connection to tobacco treatment services. The investigators used the measure that combined phone and web connections in secondary analyses.

Intervention components. The investigators used dummy variables to assess levels of financial incentives (for all three measures: 1=yes, 0=no). The proactive calling measure was assessed using a dichotomous measure (1=proactive calling, 0=no proactive calling). The investigators assessed the components' combined effect by generating interaction terms.

Covariates. The investigators examined whether randomized groups differed according to age (continuous measure ranging from 20 to 88), race/ethnicity (five dichotomous measures: white, African American, Hispanic, American Indian, and other races/ethnicities), whether participants spoke Spanish or not (1=yes, 0=no), whether respondents lived with a fellow smoker (1=yes, 0=no), time since last contact with Sage (four dichotomous variables: 2014, 2015, 2016, or 2017), and urban residence (1=metro region, 0=other).

Using a per-protocol approach, the investigators eliminated all individuals with bad addresses. Descriptive statistics were used to compare connection rates and demographics across intervention arms. The investigators ran a logistic regression model that included dichotomous measures for all six possible treatment conditions in order to run post-estimation pairwise comparisons.

Logistic regression is the optimal method for examining unbalanced factorial designs with a binary outcome. The investigators ran four logistic regression models for the primary analysis. The investigators examined: (1) the association between incentive levels and phone-only quit-line connections, (2) the association between proactive calling and phone-only quit-line connections, (3) the direct effects of financial incentives and proactive calling, and the interaction between incentive levels and proactive calling, and (4) the effect of intervention components adjusted for covariates. For the logistic regression models, the main effects for the incentive and proactive call are interpreted as the effect when all other variables equal zero, and the interaction coefficients are interpreted as the test for whether the relationship between proactive calling and connections to the quit-line depended on receiving an incentive. For the secondary analyses, the investigators report in the text any differences between the phone-only measure and the combined measure that combined telephone and web connections.

For cost-effectiveness analysis, the investigators assessed the average cost-effectiveness ratio, which is the increase in costs by each intervention group compared with the no incentive, no call group, divided by the increase in connections within each intervention group compared with the no incentive, no call group. The investigators also examined the incremental cost-effectiveness ratios, which is the change in cost compared to an adjacent condition divided by the change in connections compared to the adjacent condition. The investigators included costs associated with printing, postage, incentives, labor, and the creation and maintenance of the landing page. Costs were in 2017 US dollars and based on receipts and labor hours; hourly rates were based on national averages for respective positions.