Unpacking the Relationship Between Shared Decision-Making and Decisional Quality, Decision to Screen, and Screening Completion in Lung Cancer Screening

Abstract

Objectives.

Lung cancer screening is a complex and individualized decision. To understand how best to support patients in this decision, we must understand how shared decision-making is associated with both decisional and behavioral outcomes.

Methods.

Observational cohort study combining patient survey data with electronic health record data of lung screening-eligible patients who recently engaged in a shared decision-making discussion about screening with a primary care clinician.

Results.

Using multivariable analysis (n=529), factors associated with higher lung cancer screening decisional quality include higher knowledge (OR=1.33, p<.0001), lower perceived benefits (OR=0.90, p=.0004), higher perceived barriers (OR=1.07, p<.0001), higher self-efficacy (OR=1.13, p<.0001), and higher levels of perceiving the discussion was shared (OR=1.04, p<.0001). Factors associated with the patient’s decision to screen include older age (OR=1.12, p=.0050) and higher self-efficacy (OR=1.11, p=.0407). Factors associated with screening completion included older age (OR=1.05, p=.0050), higher knowledge (OR=1.24, p=.0045), and higher self-efficacy (OR= 1.12, p=.0003).

Conclusions.

Shared decision-making in lung cancer screening is a dyadic process between patient and clinician. As we continue to strive for high quality patient-centered care, patient decision quality may be enhanced by targeting key factors such as high-quality knowledge, self-efficacy, and fostering a shared discussion to support patient engagement in lung cancer screening decisions.

1. Introduction

In the U.S., annual lung cancer screening (LCS) with chest low-dose computed tomography (LDCT) is recommended for individuals aged 50 to 80 years with a 20 pack-year smoking history who either currently smoke cigarettes or quit within the past 15 years.¹ However, LCS is a complex and individualized decision where the benefits and risks must be weighed against personal risk factors, comorbidities that may preclude surgical treatment, and available treatment options as well as personal tolerance for uncertainty.^2,3 To date, patient awareness about LCS remains exceptionally low with less than 30% of the screening-eligible patient population aware that screening exists.⁴ Educating eligible patients about the option to screen for lung cancer is critically important if the public health benefit of screening and early detection, which is improved survival in the wake of an earlier lung cancer diagnosis, is to be realized. A key reimbursement requirement of the Center for Medicare and Medicaid Services National Coverage Determination for lung cancer screening is that the decision occur as a result of shared decision making between an individual and an informed clinician.² To understand how best to support patients in the decision to screen, or not, for lung cancer, we must determine how shared decision making (SDM) is associated with patient decisional and behavioral (i.e., screening) outcomes.

In 2015, Medicare made an unprecedented move in mandating documentation of a SDM discussion for LCS reimbursement.² SDM is viewed as the pinnacle of patient-centered care and resonates with the ethical imperative of respect for patient autonomy and engagement.⁵ This unique health policy change created a need for optimal strategies that support patients and clinicians in making shared decisions to screen, or not, for lung cancer. The purposes of this study were to: 1) identify factors associated with a patient’s: a) perception of LCS decisional quality; b) decision to screen; and c) screening completion among those who decided to screen; and 2) determine if, and how, decisional quality is related to the decision to screen and screening completion.

2. Methods

2.1. Study Design

This was an observational cohort study guided by the patient-focused Conceptual Model for Lung Cancer Screening Participation.⁶ We surveyed patients from the Kaiser Permanente Washington (KPWA) health system who underwent SDM for LCS and linked them with patient electronic health record (EHR) data. We used the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement checklist to guide reporting of results.⁷

2.2. Study Setting

Patients who had recently engaged in a SDM discussion about LCS with a primary care clinician were recruited from KPWA health system located in Washington State. KPWA developed smart-sets in 2015 that were integrated into the electronic health record for clinicians to use with potentially eligible patients to confirm eligibility, document the SDM process, and document the patient’s initial decision (opt-in, opt-out, or undecided about screening).⁸ This process allowed the study programmer to rapidly identify patients who had an SDM visit with a clinician about LCS. Institutional review board (IRB) approval was received from the Kaiser Permanente Washington IRB.

Beginning in 2019, the study programmer used KPWA EHR data to identify patients who met USPSTF LCS guideline eligibility criteria and had recently engaged in an SDM conversation with a primary care clinician. In September 2019, USPSTF LCS guidelines included individuals aged 55 to 80 years with a 30 pack-year or greater smoking history who currently smoked or quit within the past 15 years. In response to new scientific evidence, the USPSTF LCS guideline was updated in March 2021 to reduce the minimum screening-eligible age to 50 years and smoking history to 20 pack-years.¹ Study inclusion criteria were therefore updated in April 2021 in response to these revised criteria.

The study programmer ran code daily to identify patients who had an SDM discussion with a primary care clinician documented in the EHR on the prior business day. Each individual was mailed a study invitation letter and information sheet within one week of being identified. Study participation was outlined in the written invitation and included the information sheet. Participants were offered the choice to complete the 30-minute survey online or by telephone. The informed consent form included permission to use relevant data from the participant’s EHR. The information sheet from the recruitment packet was considered part of the informed consent process and provided information about potential risks and benefits of study participation, privacy protection, voluntary participation that would not impact their care, study incentive information, and a phone number to call with questions.

2.3. Study Measures

2.3.1. Study Outcomes

The study outcomes were patient perception of LCS decisional quality, decision to screen, and screening completion. Decisional Quality of the LCS decision was measured using the well-established, multidimensional Decisional Conflict Scale (DCS).⁹ The DCS is a 16-item Likert-response scale that was modified by the investigators for the LCS context,⁹ and has been validated in many health decision contexts, including mammography.^10–13 Despite its name, the DCS measures more than decisional conflict; it encompasses personal perceptions of decision-making quality including: 1) uncertainty in choosing options, modifiable factors contributing to uncertainty (such as feeling uninformed or unclear about personal values and unsupported in decision making), 2) effective decision making including feeling the choice is informed, values-based, and likely to be implemented, and 3) decisional satisfaction. Each item is assessed on a 5-point Likert scale that spans from 0 (strongly agree) to 4 (strongly disagree), resulting in a total score ranging from 0 to 64. This total score is commonly transformed into a standardized score that falls within a 0 to 100 range.^9,12 Decisional quality measured on this numeric scale from which lower scores represent lower decisional conflict and, therefore, higher decisional quality. Scores below 25 have been associated with low decisional conflict. We use this cut-off score to dichotomize this variable in our analysis.⁹ The Decision to Screen was captured categorically in the EHR through smart-sets that clinicians use to confirm screening eligibility, document the SDM discussion, use of a PDA, and the patient’s screening decision. Lung Cancer Screening Completion was assessed prospectively via EHR and coded as “yes” or “no” at six months after the initial SDM conversation.

Antecedent Variables.

Shared Decision-Making. Patients’ perceptions that the screening decision was shared was measured using the validated 9-item Shared Decision-Making Questionnaire (SDM-Q-9).¹⁴ The SDM-Q-9 is a theory-driven instrument that measures the extent to which patients are involved in the process of decision-making from the perspective of the individual making the decision.¹⁴ Lower scores indicate the patient perceived the SDM discussion with the clinician to be minimally shared.¹⁴ Knowledge. Knowledge of lung cancer screening was measured using a 7-item investigator-developed scale used in our prior studies.^15–18 Items with a multiple-choice response format were used to assess knowledge of lung cancer, risk factors, and screening. Correct responses were coded as 1 and incorrect responses as 0 and were summed to indicate the total number of correct responses. Lung Cancer Screening Health Beliefs. Perceived risk of lung cancer, benefits of, barriers to, and self-efficacy for lung cancer screening were measured using four psychometrically validated Likert-response option scales.¹⁵ Perceived Smoking-related Stigma was measured using the 5-item smoking-related subscale of the Cataldo Lung Cancer Stigma Scale,¹⁹ with higher scores indicating higher levels of perceived stigma. Medical Mistrust was measured using the 5-item theoretically-driven psychometrically validated Patient Trust in the Medical Profession Scale with higher scores indicative of higher levels of trust in the clinician.^19,20 Sociodemographic and Health Status Characteristics. Patient age and gender were collected from the EHR. Patient race, ethnicity, education, perceived financial adequacy, general health status, and family history of lung cancer were collected via survey. Patient smoking status was collected from the EHR and verified via survey.

2.4. Data Analysis

We used a complete case analysis (Figure 1). Descriptive statistics were reported using frequency (percentage) and mean (SD), respectively, for categorical and continuous variables. Bivariable tests of association between predictors and dependent variables were modeled using general linear models when the outcomes were continuous, yielding estimates of regression parameters (i.e., coefficients) or least square means for continuous and categorical independent variables, respectively, and using logistic regression models when the outcomes were dichotomous, providing odds ratios (OR) with 95% confidence intervals. Multivariable analyses were performed using those variables that had p-values <0.20 in the bivariable analyses, to determine which independent variables were independently associated with the outcomes of interest. All analytic assumptions were verified and collinearity was assessed. Analyses were performed using SAS v9.4 (SAS Institute, Cary, NC).

Figure 1.

CONSORT Flow Diagram for DECIDE Study: Patient Participants

3. Results

3.1. Participants

A flow diagram of participant inclusion is shown in Figure 1. Overall, the study participation rate (defined as the total number of participants who completed the survey divided by the total number of individuals identified as eligible for the study) was 35.2% and the response rate (defined as the total number of participants who completed the survey divided by the number of participants in the sample that enrolled in the study) was 71%. A total of 529 patients were included in the analyses. The majority of participants (81.1%, n=429) completed the survey after their SDM discussion with a clinician and before completing the screening LDCT of the chest. The mean(+SD) age of participants was 66.4 years(+5.9), 44.4% were women, 87.5% were White , 2.3% were Hispanic, and 34.4% had completed college (Table 1).

Table 1.

Participant Demographic Characteristics (N=529)

Characteristic	n (%)
Age; mean (std)	66.4 (5.9)
Sex
Female	235 (44.4)
Race
White	463 (87.5)
Ethnicity
Hispanic	12 (2.3)
Education
Less than high school	20 (3.8)
High school graduate or equivalent	94 (17.8)
Some college	233 (44.0)
College graduate or higher	182 (34.4)
Perceived Financial Adequacy
Enough money for special things after paying bills	288 (54.4)
Enough money to pay bills, but little else	177 (33.5)
Enough money to pay bills, but have to cut back	48 (9.1)
Difficulty paying bills	16 (3.0)
General Health
Excellent	13 (2.5)
Very good	124 (23.4)
Good	237 (44.8)
Fair	127 (24.0)
Poor	26 (4.9)
Don’t Know	2 (0.4)
Family History of Lung Cancer
Yes	143 (27.0)
Smoking Status
Currently Smokes	233 (44.0)

We examined all variables in the overall sample, and by screening completion outcome (Table 2). Descriptively, the overall sample had moderate knowledge levels about lung cancer risk and screening (total mean score 4.7, SD 1.4; possible range 0-7), high self-efficacy for LCS (total mean score 32.1, SD 4.3; possible range 16-36), fairly high perceived barriers to LCS (total mean score 57.2, SD 8.3; possible range 34–68), low perceived benefits of LCS (total mean score 11.1, SD 3.1; possible range 6–24), moderate perceived risk of lung cancer (total mean score 7.8, SD 1.8; range 3–12), fairly low medical trust (total mean score 10.7, SD 2.8; possible range 5–20), moderate perceived stigma (total mean score 11.0, SD 2.5; possible range 5–20), and moderately high perceptions that the LCS decision was a shared decision (total mean score 65.8, SD 29.9; possible range 0–100). When compared by screening outcome using bivariable tests, there were significant differences in knowledge, self-efficacy, perceived barriers, and perception that the LCS decision was shared. Patients who screened had higher knowledge scores and higher self-efficacy scores compared to those who did not screen (Table 2). Those who screened also had higher perceived barriers to screening and lower perceptions that the decision was shared (Table 2). Decisional quality scores were significantly lower in those who decided to screen and completed screening compared with those who decided not to screen or were undecided; mean=20.2 (SD 14.2) vs. 25.4 (SD 18.3) (p=.0038), respectively (Table 2). We focus on multivariable results, described next in the text. It should be noted that we conducted a sensitivity analysis in which linear models were used with the decisional quality scores (i.e., continuous DCS total scale scores) initially and where we found similar results to the dichotomized DCS scale scores. We are reporting the dichotomous DCS scale score models, given that odds ratios are more intuitively interpreted by clinicians compared to linear regression coefficients.

Table 2.

Health Beliefs, Psychological Characteristics, Decisional Quality, & Perception of Shared Lung Cancer Screening decision

Overall			Screened	Not screened	p
Construct	n (%)	Mean (SD), Range	Mean (SD)	Mean (SD)
Knowledge: Lung Cancer Risk and Screening	529	4.7 (1.4); 0 – 7	4.8 (1.4)	4.3 (1.5)	.0002
Self-Efficacy for Lung Cancer Screening	529	32.1 (4.3); 16 – 36	32.6 (4.0)	30.6 (4.9)	<.0001
Perceived Barriers to Lung Cancer Screening	528	57.2 (8.3); 34 – 68	57.7 (8.4)	55.9 (7.6)	.0331
Perceived Benefits of Lung Cancer Screening	524	11.1 (3.1); 6 – 24	11.0 (3.0)	11.5 (3.3)	.0734
Perceived Risk of Lung Cancer	500	7.8 (1.8); 3 – 12	7.8 (1.8)	8.0 (1.8)	.2790
Medical Trust	521	10.7 (2.8); 5 – 20	10.6 (2.7)	10.8 (3.0)	.4256
Perceived Stigma	525	11.0 (2.5); 5 – 20	11.0 (2.5)	11.0 (2.6)	.8265
Decisional Quality	529	21.5 (15.5); 0 – 75	20.2 (14.2)	25.4 (18.3)	.0038
Perception that Lung Cancer Screening Decision was Shared	511	65.8 (29.9); 0 – 100	57.4 (29.2)	60.9 (31.5)	.0341

3.2. Factors Associated with Lung Cancer Screening Decisional Quality

Overall, patients reported moderate decisional quality (total mean score 21.5, SD 15.5; possible range 0–75). As shown in Table 3, patient factors associated with higher decisional quality in multivariable analysis included higher total knowledge scores (OR=1.33, p<0001), lower perceived benefits of LCS (OR=0.91, p=.0016), higher perceived barriers to LCS (OR=1.07, p<0001), higher self-efficacy for LCS (OR=1.12, p<.0001), and greater perceptions that the screening decision was shared (OR=1.04, p<0001) The odds ratio (OR) for continuous predictors is per a 1-point increase. (Table 3).

Table 3:

Factors Associated with Patient-Perceived Lung Cancer Screening Decisional Quality (N=529)

Variable	Bivariable Model	Multivariable Modela
Gender
Female (n=235)	0.91 (0.65, 1.29); p=.6043
Male (n=294)	Reference
Race
White (n=463)	1.42 (0.85, 2.38); p=.1817	1.25 (0.77, 2.04); p=.3607
Non-White (n=66)	Reference	Reference
Education
Less than high school (n=20)	0.31 (0.12, 0.81); p=.0574
High school graduate or equivalent (n=94)	0.70 (0.42, 1.17); p=.4551
Some college (n=233)	0.63 (0.42, 0.93); p=.8508
College graduate or higher (n=182)	Reference
Perceived Financial Adequacy
Have enough for special things (n=288)	1.38 (0.97, 1.94); p=.0701	1.23 (0.89, 1.70); p=.2204
Barely or Not Enough (n=241)	Reference	Reference
Family History of Lung Cancer
Yes (n=143)	1.30 (0.88, 1.93); p=.1858
No (n=386)	Reference
Smoking Status
Currently Smokes (n=233)	1.14 (0.81, 1.61); p=.4602
Used to Smoke (n=296)	Reference
Age	1.02 (0.99, 1.05); p=.1308	1.01 (0.98, 1.04); p=.5384
Knowledge: Lung Cancer Risk & Screening	1.43 (1.26, 1.63); p<.0001	1.33 (1.18, 1.49); p<.0001
Lung Cancer Screening Health Beliefs
Perceived Risk of Lung Cancer	0.89 (0.80, 0.98); p=.0180	0.94 (0.86, 1.03); p=.2045
Perceived Benefits of Lung Cancer Screening	0.82 (0.77, 0.88); p<.0001	0.90 (0.85, 0.96); p=.0004
Perceived Barriers to Lung Cancer Screening	1.10 (1.07, 1.12); p<.0001	1.07 (1.04, 1.09); p<.0001
Self-Efficacy for Lung Cancer Screening	1.27 (1.21, 1.34); p<.0001	1.13 (1.07, 1.18); p<.0001
Psychological Characteristics
Stigma	0.96 (0.90, 1.03); p=.2702
Mistrust (Trust)	0.84 (0.79, 0.90); p<.0001	0.98 (0.92, 1.05); p=.5612
Perception that Lung Cancer Screening Decision was Shared	1.03 (1.03, 1.04); p<.0001	1.04 (1.03, 1.04); p<.0001

^aMultivariable model includes independent variables that had p<.20 from bivariate models.

Values are OR (95% CI) from logistic regression models of having low decision conflict (DCS <25 group, n=297, 56.1%) versus higher decision conflict (DCS ≥25, n=232, 43.9%). The odds ratio (OR) for continuous predictors is per a 1-point increase.

3.3. Factors Associated with the Patient’s Decision to Screen for Lung Cancer

At SDM discussion with their clinician, 95.3% (n=504 out of 529 total) of patients decided to screen (opted-in), and 4.7% (n=25 out of 529 total) were undecided or decided not to screen (data not shown in tables). In multivariable modelling, older age (OR=1.12, p=.0050) and higher self-efficacy scores (OR=1.11, p=.0407) were associated with opting into LCS; the OR for continuous predictors is per a 1-point increase (data not shown in tables).

3.4. Factors Associated with Lung Cancer Screening Completion

Screening completion rates were high. Among the total sample of participants, 399 (75.4%) participants completed screening; 99% completed screening among those who reported deciding to screen; and screening completion was 0.93% among those who reported they were undecided (data not shown in tables). Multivariable analyses showed that older age (OR=1.05, p=.0050), higher knowledge scores (OR=1.24, p=.0045), and higher self-efficacy scores (OR=1.12, p=.0003) were associated with screening completion; the OR for continuous predictors is per a 1-point increase. (Table 4).

Table 4.

Patient Factors Associated with Screening Completion (N=529)

Variable	Bivariable Model	Multivariable Modela
Age	1.05 (1.01, 1.08); p=.0064	1.05 (1.02, 1.09); p=.0050
Gender
Female (n=235)	1.17 (0.78, 1.74); p=.4461
Male (n=294)	Reference
Race
White (n=463)	1.40 (0.79, 2.46); p=.2495
Non-White (n=66)	Reference
Education
Less than high school (n=20)	1.31 (0.42, 4.13); p=.6407
High school graduate or equivalent (n=94)	1.01 (0.57, 1.80); p=.9625
Some college (n=233)	0.99 (0.63, 1.55); p=.9687
College graduate or higher (n=182)	Reference
Perceived Financial Adequacy
Have enough for special things (=288)	0.84 (0.56, 1.25); p=.3919
Barely or Not Enough (n=241)	Reference
Family History of Lung Cancer
Yes (n=143)	1.48 (0.92, 2.36); p=.1057	1.39 (0.84, 2.31); p=.2044
No (n=386)	Reference	Reference
Smoking Status
Currently smokes (n=233)	1.01 (0.68, 1.51); p=.9580
Used to smoke (n=296)	Reference
Knowledge: Lung Cancer Risk & Screening	1.04 (1.00, 1.09); p=.0323	1.24 (1.07, 1.44); p=.0045
Lung Cancer Screening Health Beliefs
Perceived Risk of Lung Cancer	0.94 (0.84, 1.05); p=.2787
Perceived Benefits of Lung Cancer Screening	0.94 (0.88, 1.01); p=.0741	1.00 (0.93, 1.08); p=.9850
Perceived Barriers to Lung Cancer Screening	1.03 (1.00, 1.05); p=.0338	0.99 (0.96, 1.02); p=.4650
Self-Efficacy for Lung Cancer Screening	1.11 (1.06, 1.16); p<.0001	1.12 (1.05, 1.19); p=.0003
Psychological Characteristics
Stigma	0.99 (0.92, 1.07);p=.8262
Mistrust (Trust)	0.97 (0.90, 1.04); p=.4250
Perception that Lung Cancer Screening Decision was Shared	1.01 (1.00, 1.01)); p=.0348	1.00 (1.00, 1.01); p=.5333

^aMultivariable model includes independent variables that had p<.20 from bivariate models.

Values are OR (95% CI) from logistic regression models of screening completion (n=399, 75.4%) versus non-completion (n=130, 24.6%).

The odds ratio (OR) for continuous predictors is per a 1-point increase.

3.5. Decisional Quality and the Decision to Screen and Screening Completion

For those who decided to screen, participants who then followed through and completed a lung scan reported significantly (p=.0038) better decisional quality (mean=20.21, SD=14.20) compared to those who did not follow through and complete a lung scan (mean=25.36, SD=18.36) (data not shown in tables).

4. Discussion and Conclusion

4.1. Discussion

Consistent with Rennert et al. (2020), an overwhelming majority of study participants who engaged in a SDM discussion with a clinician decided to screen after that discussion and most of those followed through by completing a lung scan,²¹ highlighting the importance of the patient-clinician discussion in complex cancer screening decisions. Findings from this study offer insights into the role of SDM in the patient-clinician encounter related to LCS as to which patient factors are associated with decisional quality, the decision to screen, screening completion, and the relationship between patient decisional quality and screening decision and completion. To date, understanding SDM in LCS has been limited by the use of a proxy for the patient’s decision to screen through the use of EHR data to confirm the presence, or not, of LDCT scan results for the patient. However, SDM in LCS is a process that involves a patient-clinician discussion resulting in a patient decision (i.e., to screen, not to screen, undecided) and a patient behavior (i.e., screening, not screening). The early adoption by the health system to create an EHR-based prompt for clinicians to engage in SDM regarding LCS with their eligible patients and to document the outcome of that conversation allowed the study team a unique real-world laboratory within which to study patients’ perceptions about SDM and the factors associated with the actual decision to screen and the completion of LCS.

Our findings support decisional quality as an important variable in SDM in LCS as higher decisional quality scores were associated with both the decision to get screened and screening completion in this study. As healthcare systems have increased a focus on patient-centered care, decisional quality has emerged as an important component of healthcare. In other studies, decisional quality has also been associated with less decisional anxiety and greater patient satisfaction with their medical care,^10,22 further supporting the importance of patient engagement in healthcare decisions. Ultimately, the individual patient is usually tasked with following through with a healthcare decision (particularly in the context of cancer screening). Because higher decisional quality has the potential to lead to a more satisfying patient engagement experience from the perspective of the individual making the decision, it also has the potential to support recommended cancer screening receipt.

Results of this study have potential to inform tailored intervention development, specifically which variables should be used to tailor messages. It may also be beneficial to tailor messages into patient decision aids to support improvements in decisional quality as a component of an effective patient decision aid. Future research is needed to test this hypothesis. Knowledge, self-efficacy, and perceiving the decision to be shared were associated with higher decisional quality, giving insight into the most salient variables on which to tailor interventions that would support both patients and clinicians in SDM for LCS. For example, tailored interventions delivered prior to the clinical encounter that increase LCS knowledge and empower a patient’s self-efficacy may positively impact the SDM discussion between the patient and their clinician. Equally, clinician targeted interventions that support empathic communication that is perceived by the patient as a “shared” discussion may also lead to higher decisional quality and subsequent screening completion. Unexpectedly, higher decisional quality was associated with decreased perceived benefits of and increased perceived barriers to LCS which is inconsistent with our prior work testing the Conceptual Model for Lung Cancer Screening Participation⁶ where both perceived benefits of and perceived barriers to LCS were associated with LCS behavior in the hypothesized directions. It is possible that benefits and barriers to LCS may be perceived differently among patients who are insured and receive care in an integrated health system as they may not experience similar barriers that individuals outside an integrated health system experience.

Prior studies have also noted the importance of an individuals’ perception of risk as a motivator to screen for multiple cancers, including lung. However, results from this study did not support an association between perceived risk of getting lung cancer and the decision to screen or screening completion. This finding raises the question of how important it may be to address perceived risk as a potentially modifiable intervention target for LCS uptake. This is consistent with our prior findings testing the Conceptual Model for Lung Cancer Screening Participation which did not support a significant association between an individuals’ perception of risk and LCS behavior. While it may be logical to assume that higher perceived risk would be associated with LCS behavior, perceived risk may be moderated by other key variables such as fatalism, worry, or fear. This may also be in part related to the role that stigma may play in this unique patient population.

Trust in the clinician was not associated in this study with LCS completion in contrast with results of our prior qualitative work conducted in the KPWA health system with patients (n=20) who had completed LCS.²³ That study revealed that trust in the referring clinician was one of four key themes that were primary motivators to screen.²³ As noted in our prior work testing the Conceptual Model on Lung Cancer Screening Participation,⁶ both quantitative and qualitative data are critical to understanding the complex phenomenon of LCS behavior more robustly. In this prior work, trust was not significantly associated with screening behavior in the quantitative data but qualitative interviews with screening-eligible participants revealed trust/mistrust as an important variable related to both an individuals’ decision to screen and screening behavior.⁶ Additional research is needed to better understand the concept of trust and how it interacts with SDM.

Lung cancer screening has been recommended by the USPSTF for the past decade, however, screening uptake nationally remains extremely low compared to other types of cancer screening at this same historical stage of implementation (see for example, Chao et al., 2019, Knudsen et al., 2021, Nisbet et al., 2019).^24–26 Results from this study demonstrate that SDM may promote LCS by providing an opportunity to learn about the option to screen from an informed clinician, discussing the benefits and barriers to screening, receiving a personalized recommendation from one’s clinician, and receiving a referral for screening if desired.

4.2. Limitations

Several limitations should be considered when interpreting our study results. The sample was insured, predominantly White and highly educated; similar results may not be seen in more diverse populations. Also, the integrated health system setting may be unique in several ways. KPWA provides health insurance and care to more than 600,000 members.⁸ In addition, KPWA’s lung cancer screening program was established in 2015 and included continuing education to all clinicians and staff which may limit generalizability given the high screening completion rate of our study participants when compared nationally. Although it is important to note that the screening rate in our study reflects those who underwent SDM who were eligible to screen as the denominator, and ultimately decided to screen as the numerator. Other studies estimate the national screening rate in the U.S. is <10-15% among those who screen (numerator) by those who are eligible for lung cancer screening (denominator). We did not directly calculate this rate to be comparable to national data, as we did not capture the entire sample of eligible patients in this study.

Other components of the KPWA program included integration of modules to guide and document SDM discussions and the patient’s decision about whether or not to screen into the EHR. For patients who decide to screen after a SDM discussion with their clinician, the EHR prompts clinicians to order a screening LDCT. In other health systems that do not have LCS integrated into the EHR, or that serve more diverse populations, outcomes may be different. Further, because SDM mandates may alter who receives SDM and how SDM occurs, care should be taken when generalizing findings to other clinical contexts. Additionally, we did not conduct mixed-effects models accounting for clustering of patients within clinicians because only 353 of the 529 patients could be linked to clinician IDs; and the average number of patients per clinician was small (353/106 = 3.33); however, and importantly, the intraclass correlation coefficient (ICC) for the decisional quality scores, estimated in the 353 subset using a mixed-effects model with clinician class specified as a random effect, was slightly less than zero (−0.0238), indicating that the true ICC is near zero and that two patient participants chosen randomly from any same clinician vary as much on decisional quality scores as any two randomly chosen patient participants of the whole population. This suggests that results presented here would have changed negligibly if clinician clustering had been accounted. Finally, the descriptive study design imposes limitations on conclusions that can be drawn regarding causality.

5. Conclusion

Shared decision making in LCS is a dyadic process involving communication between a patient and an informed clinician followed by enactment of a behavior on the part of the patient to follow-through on the decision. As we continue to strive for high quality patient-centered care, decisional quality can support or hinder patient engagement in the healthcare decision-making process. Key factors such as sufficient knowledge secondary to education support, enhancing self-efficacy, and fostering a SDM discussion may contribute to the environmental milieu that leads to improved patient engagement in LCS decisions.

5.1. Practice Implications

Decisional quality is an important variable in patient-clinician discussions about the option to screen, or not, for lung cancer. Because higher levels of decisional quality have the potential to lead to a more satisfying patient engagement experience from the perspective of the individual making the decision, it also has the potential to support recommended lung cancer screening receipt. Higher decisional quality is associated with both the decision to get screened and screening completion, and findings from this study support knowledge of lung cancer risk and screening, self-efficacy for lung cancer screening, and perceiving the decision to be shared were associated with higher levels of patient-perceived decisional quality, giving insight into the most salient variables on which to tailor patient and clinician-facing interventions.

Table 5.

Factors Associated with Decision to Screen for Lung Cancer (N=529)

Variable	Bivariable Model	Multivariable Modela
Age	1.13 (1.05, 1.22); p=.0009	1.12 (1.04, 1.22);p=.0050
Gender
Female (n=235)	0.73 (0.33, 1.62); p=.4364
Male (n=294)	Reference
Race
White (n=463)	1.36 (0.45, 4.09); p=.5863
Non-White (n=66)	Reference
Education
Less than high school (n=20)	>99 (<0.1, >999); p=.9998
High school graduate or equivalent (n=94)	0.59 (0.19,1.80); p=.3507
Some college (n=233)	0.74 (0.28, 1.91); p=.5297
College graduate or higher (n=182)	reference
Perceived Financial Adequacy
Enough for special things (n=288)	0.79 (0.35, 1.79); p=.5684
Barely or Not Enough (n=241)	Reference
Family History of Lung Cancer
Yes (n=143)	0.64 (0.28, 1.49); p=.3044
No (n=386)	Reference
Smoking Status
Currently Smokes (n=233)	1.00 (0.45, 2.25); p=.9963
Used to Smoke (n=296)	reference
Knowledge: Lung Cancer Risk & Screening	0.97 (0.63, 1.51); p=.9092
Lung Cancer Screening Health Beliefs
Perceived Risk of Lung Cancer	0.82 (0.64, 1.05); p=.1099	0.93 (0.69, 1.25); p=.6146
Perceived Benefits of Lung Cancer Screening	0.80 (0.71, 0.92); p=.0011	0.87 (0.74, 1.02); p=.0787
Perceived Barriers to Lung Cancer Screening	1.07 (1.02, 1.12); p=.0068	1.02 (0.95, 1.09); p=.5594
Self-Efficacy for Lung Cancer Screening	1.14 (1.06, 1.23); p=.0008	1.11 (1.00, 1.23); p=.0407
Psychological Characteristics
Stigma	0.89 (0.76, 1.04); p=.1464	0.91 (0.76, 1.08); p=.2677
Mistrust (Trust)	0.93 (0.81, 1.07); p=.2953
Perception that Lung Cancer Screening Decision was Shared	1.00 (0.99, 1.01); p=.8695

^aMultivariable model includes independent variables that had p<.20 from bivariate models.

Values are OR (95% CI) from logistic regression models of decision to opt-in (n=504, 95.3%) versus deciding to not opt-in (n=25, 4.7%).

The odds ratio (OR) for continuous predictors is per a 1-point increase.

Highlights.

• Decisional quality is an important variable associated with higher decisional quality associated with the decision to get screened and subsequent screening completion.

• Because higher levels of decisional quality have the potential to lead to a more satisfying patient engagement experience from the patient’s perspective, it also has the potential to support recommended lung cancer screening receipt.

• Results give insight into the most salient variables – knowledge of lung cancer risk and screening, self-efficacy for lung cancer screening, and perceiving the decision to be shared – on which to tailor patient and clinician-facing interventions.

Availability of Data and Materials

The quantitative datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.