Abstract
Introduction
Lung cancer is a leading cause of cancer-related death in the United States. Specifically, less than one in five (18.6%) individuals diagnosed with lung cancer will survive 5 years after diagnosis; this is partly due to the late stage of diagnosis for most lung cancer patients. 1 The National Lung Screening Trial examined the efficacy of annual low-dose computed tomography (LDCT) scans with results showing a reduction of lung cancer deaths by 16% to 20% with LDCT screening for 3 consecutive years. 2 In 2013, the US Preventive Services Task Force (USPSTF) recommendation updated lung cancer screening (LCS) to a grade B, 3 and the most recent recommendation remains at that level. 4 The Centers for Medicare and Medicaid Services (CMS) will reimburse LCS for eligible individuals if their health care provider engages in a shared decision-making (SDM) consultation prior to screening. 5 However, uptake and adherence rates have remained low.6,7
Prior work has examined factors related to LCS uptake, focusing on demographic and behavioral factors. Specifically, research has shown that persons who currently smoke are more concerned about lung cancer and more interested in LCS than those who previously smoked. 8 Studies have also shown gender and racial differences in intentions and behavior regarding LCS.9,10 For example, non-Hispanic White individuals may be more likely to undergo screening than other individuals. 10 Less is known about psychosocial factors relating to LCS uptake. One cross-sectional study’s results showed positive associations between four factors and LDCT intentions: perceived accuracy of the LDCT for lung cancer detection, believing that early detection is associated with a better prognosis, perceived high individual risk of lung cancer, and not being afraid of CT scans. 11
The present study is guided by the Ottawa Decision Support Framework, 12 which addresses the decisional needs, outcomes, and support of people who are making health-related choices. In this study, we aimed to examine the association between LCS behavior (completing LCS by the 6-month follow-up) and beliefs about LCS that aligned with information presented in the intervention’s decision aid (importance of early detection, concern about radiation, concern about false alarms, and concern about overdiagnosis). We also included an assessment of anticipated regret if screening was declined and lung cancer was later diagnosed. This variable was included because affective forecasting, which can be defined as an attempt to predict how a decision will make you feel in the future, has been shown as an important factor in patients’ choices for screening and medical treatments. 13 Furthermore, anticipated regret has been shown as a predictor of health intentions and behavior as individuals are motivated to avoid feeling regret for their actions or inaction. 14
Methods
Study Design
This is a secondary analysis of data from the intervention group of a randomized clinical trial conducted with 13 state tobacco quitlines’ clients. The trial was approved by the institutional review board.
Setting
Participants were enrolled from March 2015 to September 2016; they were followed for 6 months (until May 2017). Only participants who were randomized to the intervention were included in the present study. The intervention tested a video-based decision aid in comparison with standard educational materials to assess the effect of the decision aid. 15 Full details of the trial, including its context and methods, are published elsewhere. 16 Data for the present study are drawn from a 1-week assessment, and the screening outcome was collected during the 6-month follow-up. Data were collected via telephone interviews, or by mail if participants were not able to be reached by telephone. In the main study, 235 (of 259) participants randomized to the intervention completed the 1-week follow-up and 218 completed the 6-month follow-up.
Participants
Eligible participants were quitline callers who met LCS eligibility requirements, including being between 55 and 80 years of age, a current or former (within 15 years) smoking history, with a minimum of 30 pack-years; they could read, understand, and write in English; and consented to be included in the trial. Information about comorbidities was not collected.
Primary Outcome
This study’s main outcome reflects whether participants had obtained LCS by the 6-month follow-up point and compares it with LCS intentions (1 = obtained LCS, 0 = did not obtain LCS). This was ascertained by a mailed survey.
Predictors
Independent variables were selected based on observations from prior qualitative research in order to represent factors that may be associated with LCS.17–21
Importance of Detecting Lung Cancer Early
In the 1-week follow-up, participants were asked, “On a scale from 0 to 10, where 10 means extremely important and 0 means not at all important, how important is it to try to find lung cancer early when it is potentially curable?”
Concern About Radiation
In the 1-week follow-up, participants were asked, “On a scale of 0 to 10 [ . . . ], how concerned are you about radiation exposure from lung cancer screening and potential follow-up testing?”
Concern About False Alarms
In the 1-week follow-up, participants were asked, “On a scale of 0 to 10 [ . . . ], how concerned are you that your scan says you have cancer when you do not (in other words, a false-alarm)? This would also mean having additional potentially harmful testing.”
Concern About Overdiagnosis
In the 1-week follow-up, participants were asked, “Some lung cancers may never become life threatening, yet some people may be treated for lung cancer that would never have harmed them, this is called overdiagnosis. On a scale of 0 to 10 [ . . . ], how concerned are you about overdiagnosis?”
Anticipated Regret
In the 1-week follow-up, participants were asked to rate on a 0 to 10 scale, “If you made the decision not to be screened for lung cancer and you were later diagnosed with lung cancer, would you have regrets?”
Practical Considerations
Potential barriers to LCS were assessed in the 1-week follow-up with the following three items that were dichotomized (1 = yes, 0 = no, unsure): 1) “If you wanted to be screened, would you know where to go?” 2) “Do you know if your insurance covers lung cancer screening?” and 3) “If you had to pay for screening would you be able to? Assume a screening scan cost $200?”
Knowledge
Participants’ knowledge was assessed with nine items from the LCS-12 22 and this analysis uses scores from the 6-month follow-up. These items were scored by computing the percentage of questions that were answered correctly. Due to the skew of data, the data were coded into the following dummy variables for analysis: low scores (11.11 [lowest score] to 44.44), average score (55.56, this is the reference group in the regression analysis [there are no scores between 44.44 and 55.56]), and high scores (66.67 to 100 [there are no scores between 55.56 and 66.67]).
Additional Control Variables
Participants’ demographics, including age, educational attainment (i.e., some college or more), and having health insurance, were included in the analysis. Additionally, participants’ intentions to obtain LCS at 1-week assessment following the intervention were included. Three dummy-coded variables to reflect smoking status at the 6-month follow-up (relapsed during study, quit smoking during study, still smoking throughout study, with quit prior to study as the reference variable for the regression analysis) were also included.
Analysis Plan
Data were skewed within some of the predictor variables. Due to the skewness and bimodal distributions of the variables, log transformation was not appropriate for this analysis. 23 In order to account for this, dummy variables were created as described below. This approach allowed us to examine more granular differences within the high and low levels of concern that were seen in responses.
Importance of Detecting Lung Cancer Early
Responses to the item (with ratings of 9 to 10) were highly skewed, and a dummy variable was created for responses of 10, and all other responses (0 to 9) as the reference group for the regression analysis.
Concern About Radiation
Three dummy variables were created with the following groupings: Very Concerned (responses of 9 and 10), Concerned (responses indicating 5 to 8), Semi-Concerned (responses indicating 1 to 4). The reference group in the regression analysis indicated 0 concern.
Concern About False Alarms
Three dummy variables were created with the following groupings: Very Concerned (responses of 9 and 10), Concerned (responses indicating 5 to 8), Semi-Concerned (responses indicating 1 to 4). The reference group in the regression analysis indicated 0 concern.
Concern About Overdiagnosis
Three dummy variables were created with the following groupings: Very Concerned (responses of 9 and 10), Concerned (responses indicating 5 to 8), Semi-Concerned (responses indicating 1 to 4). The reference group in the regression analysis indicated 0 concern.
Anticipated Regret
One dummy variable (very concerned) was created for responses of 10; all lower responses (0 to 9) are in the reference group for the regression analysis.
Statistical Analysis
The data were analyzed in a logistic regression model with three blocks. This reflected the conceptual model that was guided by prior research findings (Figure 1). The predictors for the final model were selected with a seven-step process of purposeful model building. 24 This process was undertaken in order to select only the most important variables. The variables’ correlations as well as VIF from a linear regression model was used in order to diagnose potential multicollinearity as recommended,25,26 with no serious threats detected. The overall model fit was assessed, and standardized residuals examined to detect outliers; four cases that were misclassified were deleted (one had discovered ineligibility for screening while discussing with health care provider; three had been inactive through portions of the study [3-month follow-up] and were missing data). The analyses were conducted in SPSS (v. 24).
Hypothesized model.
Results
Descriptive Data
Participants’ (n = 204, after deleting the 4 outliers) mean age was 61.39 years (SD = 4.92), and the mean pack-years smoking history was 53.61 pack-years (SD = 23.79). The majority of participants were White (69.6%), and more than 50% of the sample had completed some college or higher levels of education. A summary of descriptive statistics is provided in Table 1.
Participant Characteristics
LCS, lung cancer screening.
Reference category in analysis.
LCS Completion by the 6-Month Follow-up
At 1 week following the intervention, a small portion of participants reported low intentions to obtain LCS (n = 24, 11.8%). At the 6-month time point, 62 (30.4%) participants had obtained LCS.
Predictors of Obtained LCS
The regression model had good fit overall (Hosmer-Lemeshow χ2 = 8.14, P = 0.42; Nagelkerke R 2 = 0.37).
Regarding practical barriers, the results showed a positive association between each of the variables that were assessed and obtaining LCS. Specifically, compared with those who did not know where to go, individuals who knew where to go for LCS at 6 months were more likely to have been screened (odds ratio [OR] = 5.67, 95% confidence interval [CI] = 1.56–20.58). Similarly, participants who knew if their insurance covered screening were more likely to have obtained LCS at 6 months compared with those who did not know (OR = 4.73, 95% CI = 2.15–10.41).
High (OR = 0.17, 95% CI = 0.04–0.65) and Moderate (OR = 0.15, 95% CI = 0.05–0.53) levels of concern about overdiagnosis were associated with not obtaining LCS when compared with those who indicated low concern about this risk. In contrast, a high level of anticipated regret is associated with an increased likelihood of obtaining LCS compared with lower levels (OR = 5.59, 95% CI = 1.72–18.10). No other statistically significant association were observed (Table 2).
Factors Associated With Obtaining Lung Cancer Screening
LCS, lung cancer screening.
P < 0.05, **P < 0.01, ***P < 0.001.
Discussion
In this study, we aimed to identify predictors associated with obtaining LCS within 6 months of a decision aid intervention. Approximately one third of participants had obtained LCS by this point. This was lower than pulmonary care settings but higher than some national estimates; further discussion is presented with the main study. 16 Our results show two main areas of association with obtaining LCS. First, we observed that not having practical barriers (i.e., knowing where to go for LCS and knowing if insurance covered LCS) was associated with an increased likelihood of having obtained LCS. Second, beliefs toward LCS were also associated with behavior. Specifically, concern about overdiagnosis was associated with a lower likelihood of having obtained LCS. Conversely, anticipated regret about declining screening and later being diagnosed with lung cancer was associated with about six times higher likelihood of having obtained LCS.
However, responses regarding false alarms, radiation exposure, and believing that finding lung cancer earlier due to screening were not associated with LCS behavior; each of these areas of consideration was included in the decision aid intervention. There has been concern about SDM hindering LCS uptake.27,28 Prior work has shown that use of decision aids and SDM typically results in higher decision quality. 12 Indeed, the main study’s results showed that intervention participants scored higher on knowledge assessments than those in the standard education group. 16 Within the present study, perceived importance of most LCS harms and benefits that are elucidated in the patient decision aid were not significant. Specifically, in the decision aid, overdiagnosis was presented along with the other potential harms of LCS. It was explained as diagnosing a cancer that would not have been life-threatening and that some people may be treated for a cancer that would not have done harm in their lifetime. Anticipated regret was not directly addressed in the decision aid. Rather, viewers were invited to consider what aspects of screening are most important for their situations.
Interestingly, risk factors such as age and pack-year smoking history were not associated with screening. Similarly, screening intention at the 1-week follow-up was also not associated with LCS behavior. These results suggest that a mix of factors are associated with LCS uptake.
Our findings that concern about overdiagnosis align with prior work examining barriers to LCS uptake. For instance, approximately one third of participants in one study reported hesitancy to find out if they had lung cancer. 29 Furthermore, a multilevel examination of barriers to LCS suggests that patient-level fear of diagnosis veracity and provider-level barriers related to limited and misinformation about the screening process and effectiveness hinder uptake. 30
Less work has examined anticipated regret specifically related to LCS. However, a meta-analysis of anticipated regret and health behavior suggests that anticipated regret is associated with both intentions and health behaviors. 14 This aligns with results of a meta-analysis examining affective forecasting in medical screening and treatment decisions. 13 Furthermore, anticipated regret from not engaging in a behavior (“inaction regret”) has been shown as a predictor of stronger intentions and behavior. 14 Similarly, guidance for developing cancer risk messages suggests using loss-based messages to encourage screening behavior based on risk perception research. 31 Our findings align with this perspective.
Finally, prior work has suggested that concerns about practical barriers affect LCS uptake.32,33 Our findings regarding knowing where to go and knowing if insurance covers LCS are in line with this concern. It is noteworthy that most of our assessments were collected during the 1-week follow-up after exposure to the patient decision aid. The timing of data collection would have allowed most participants to have a relatively fresh recollection of information regarding potential harms and benefits of LCS. In contrast, participants who were interested in obtaining screening may not have had enough time to learn about practical considerations such as insurance coverage or screening locations.
Limitations
As is the case with all research, there were some limitations. First, the participants had called quitlines and may have different attitudes toward health-related behaviors such as LCS than other persons with a heavy smoking history.34,35 However, our study may shed light on predictors of LCS among individuals who are considering smoking cessation. Similarly, more than half of the participants had completed some college. However, education was not significantly associated with outcomes in the present study or in the main study. 16 Second, the variables that were included in this analysis were assessed with single items; thus, our measures may be less reliable than if we had access to a full scale. However, the results of our secondary analysis suggest there are associations between individual-level predictors and LCS. Finally, although the participants were followed for 6 months, a longer time period would have been ideal to assess both uptake and subsequent adherence to annual screening.
Conclusion
We identified several factors associated with an increased likelihood of obtaining LCS, including knowing where to go, if insurance covers screening, and anticipated regret about a later diagnosis. Concern about overdiagnosis was negatively associated with obtaining LCS by 6 months. Our results highlight the importance addressing patient concerns such as overdiagnosis and anticipated regret in decision making.
Footnotes
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Financial support for this study was provided, in part, by a cancer prevention fellowship for Kristin Maki, supported by the Cancer Prevention and Research Institute of Texas (CPRIT) Grant Award RP170259, Shine Chang, PhD, Principal Investigator; CPRIT grants RP160674 and RP190210; and by the MD Anderson Cancer Center Support Grant, CA016672, funded by the National Cancer Institute. This study was also supported in part by award CER-1306-03385 from the Patient-Centered Outcomes Research Institute; award P30CA016672 from the National Institutes of Health, National Cancer Institute (Volk) that used the Biostatistics Resource Group, Clinical Protocol and Data Management, and Shared Decision Making Core, and a grant from The University of Texas MD Anderson Cancer Center Duncan Family Institute for Cancer Prevention and Risk Assessment (Volk) that supported the Shared Decision Making Collaborative and Center for Community-Engaged Translational Research. The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report.
Authors’ Note
An earlier version of this article was presented at the 2020 Society for Medical Decision Making Conference.
