Cancer treatment

Treatment information was obtained from both registry and claims data. Surgical treatment data were obtained using the SEER Kentucky Cancer Registry (KCR) surgery codes and claims data from the SEER Medicare Provider Analysis and Review (MEDPAR) and Carrier Claims – National Claims History (NCH) files. Radiation treatment and chemotherapy data were obtained using KCR treatment codes, MEDPAR, NCH, and the SEER outpatient claim files. Furthermore, the SEER Durable Medical Equipment and Prescription Drug Event files were evaluated to determine the use of chemotherapy. A full description of the specific codes used can be found in Tables S1 and S2.

Continue Reading

Variables of interest

Age at the time of diagnosis, race/ethnic information and the stage at diagnosis were obtained from KCR. Race/ethnicity was categorized into 3 groups: White, black, and other. Comorbidity burden on diagnosis was measured using the lung cancer-specific Klabunde Comorbidity Index (KCI).8 The KCI, a continuous comorbidity index based on the comorbid conditions identified by Charlson index that incorporates the diagnostic and procedure data from physician claims, was converted into a categorical variable of increasing severity of comorbidity: no and very low comorbidity (KCI=0), low comorbidity (KCI=1), moderate comorbidity (KCI=2), and high comorbidity (KCI≥3).8 Poverty at the residential census tract level was categorized in 4 groups according to the percentage of residents with an income below the US federal poverty line (<8%, 8%–15%, 15%–30%, or >30%). Education at the residential census tract level was categorized into 4 groups based on the proportion of adult residents (≥25 years old) with less than high school education (<15%, 15%–25%, 35%–35%, and >35%). Geographic origin of the patient was classified into metropolitan versus rural origin and Appalachian versus non-Appalachian origin according to the US Census Bureau definitions.

Outcome variables included treatment status and overall survival (OS) from the time of diagnosis. The treatment status was a dichotomous variable, defined as any treatment (including surgery, radiation, chemotherapy, or a combination of those) versus no treatment. OS was measured from the day of diagnosis to the date of death or the date of last contact.

Related Articles

Statistical analysis

Descriptive analysis and χ2 tests were used to assess the association between treatment status and covariates. Multivariate logistic regression models were used to explore the association between receiving treatment and comorbidity, while accounting for the potential confounding effect of other covariates. The interaction effects and the goodness of fit were examined. The final model only included the significant covariates. Kaplan–Meier estimates and log-rank tests were calculated to examine the survival curves of various treatment and comorbidity combinations. Cox regression survival models were fitted to identify the significant factors associated with survival while controlling for other covariates. All tests were 2-sided with a 0.05 significance level. SAS 9.3 was used for all analysis.