Determinants Associated With Longitudinal Adherence to Annual Lung Cancer Screening: A Retrospective Analysis of Claims Data

Abstract

Objective: Lung cancer screening (LCS) efficacy is highly dependent on adherence to annual screening, but little is known about real-world adherence determinants. We used insurance claims data to examine associations between LCS annual adherence and demographic, comorbidity, health care usage, and geographic factors.

Materials and methods: Insurance claims data for all individuals with an LCS low-dose CT scan were obtained from the Colorado All Payer Claims Dataset. Adherence was defined as a second claim for a screening CT 10 to 18 months after the index claim. Cox proportional hazards regression was used to define the relationship between annual adherence and age, gender, insurance type, residence location, outpatient health care usage, and comorbidity burden.

Results: After exclusions, the final data set consisted of 9,056 records with 3,072 adherent, 3,570 nonadherent, and 2,414 censored (unclassifiable) individuals. Less adherence was associated with ages 55 to 59 (hazard ratio [HR] = 0.80, 99% confidence interval [CI] = 0.67-0.94), 60 to 64 (HR = 0.83, 99% CI = 0.71-0.97), and 75 to 79 (HR = 0.79, 99% CI = 0.65-0.97); rural residence (HR = 0.56, 99% CI = 0.43-0.73); Medicare fee-for-service (HR = 0.45, 99% CI = 0.39-0.51), and Medicaid (HR = 0.50, 99% CI = 0.40-0.62). A significant interaction between outpatient health care usage and comorbidity was also observed. Increased outpatient usage was associated with increased adherence and was most pronounced for individuals without comorbidities.

Conclusions: This population-based description of LCS adherence determinants provides insight into populations that might benefit from specific interventions targeted toward improving adherence and maximizing LCS benefit. Quantifying population-based adherence rates and understanding factors associated with annual adherence are critical to improving screening adherence and reducing lung cancer death.

Keywords: Adherence determinants; cancer screening; lung cancer; patient adherence.

Figure 1.. Study cohort.

Study population was derived as described in the text. Four individuals had multiple exclusion criteria. Censored individuals had 10-18 months of follow-up since index CT but no claim for a second LDCT and hence could not be classified as adherent or non-adherent. The duplicate claims are explained by the procedure (LDCT) and professional fee (radiology interpretation) being billed separately to insurance.

Figure 2.. Time between index and second LDCT.

Defining adherence as having a second LDCT 10-18 months after the index CT scan captures 87% of returning individuals. Claims 3-9 months after the index LDCT were assumed to be diagnostic studies for monitoring pulmonary nodules.

Figure 3.. Screening adherence by year of index LDCT.

(A) Adherence to first annual LDCT by year of index claim. *For 2017, individuals who could not be classified due to data maturity are not included. ^†Overall numbers do not include unclassifiable individuals screened in 2017 or any individuals screened in 2018. (B) Adherence to the second annual LDCT by year of index claim. Only individuals who were adherent to the first annual screening are included.

Figure 4.. Variables associated with lung cancer screening adherence.

Analysis was performed as described in Methods. After adjustment for other covariates, female sex, ages 70-74, and commercial insurance (compared to Medicare Advantage) did not have a statistically significant relationship with screening adherence. ^† There was a significant interaction between number of outpatient visits and CCI score, therefore these results are presented in Figure 5.

Figure 5.. Influence of comorbidities and heath care utilization on screening adherence.

After adjustment for other covariates, increased screening adherence is seen with increasing number of outpatient visits across all comorbidity burdens.

Figure 6.. Interaction between residence and insurance at index LDCT.

There is a statistically significant reduction in adherence for rural residents with commercial and Medicare Advantage insurance, however this is likely driven by the small number of rural residents with each insurance type (commercial n = 85, Medicare Advantage n = 67, Medicare FFS n = 372, and Medicaid n = 90).

Figure 7.

(A) Insurance type by residence. (B) Association between comorbidity burden and health care utilization.