https://orcid.org/0000-0001-7290-3864
Abstract
We estimated trends in total and out-of-pocket (OOP) costs around the first year of diagnosis for privately insured nonelderly adult cancer patients.
We constructed incident cohorts of breast, colorectal, lung, and prostate cancer patients diagnosed between 2009 and 2016 using claims data from the Health Care Cost Institute. We identified cancer-related surgery, intravenous (IV) systemic therapy, and radiation and calculated associated total and OOP costs (in 2020 US dollars). We assessed trends in health-care utilization and cost by cancer site with logistic regressions and generalized linear models, respectively.
The cohorts included 105 255 breast, 23 571 colorectal, 11 321 lung, and 59 197 prostate cancer patients. For patients diagnosed between 2009 and 2016, total mean costs per patient increased from $109 544 to $140 732 for breast (29%), $151 751 to $168 730 for lung (11%) or $53 300 to $55 497 for prostate (4%) cancer were statistically significant. Increase for colorectal cancer (1%, $136 652 to $137 663) was not statistically significant (P = .09). OOP costs increased to more than 15% for all cancers, including colorectal, to more than $6000 by 2016. Use of IV systemic therapy and radiation statistically significantly increased, except for lung cancer. Cancer surgeries statistically significantly increased for breast and colorectal cancer but decreased for prostate cancer (P < .001). Total costs increased statistically significantly for nearly all treatment modalities, except for IV systemic therapy in colorectal and radiation in prostate cancer.
Rising costs of cancer treatments, compounded with greater cost sharing, increased OOP costs for privately insured, nonelderly cancer patients. Policy initiatives to mitigate financial hardship should consider cost containment as well as insurance reform.
Cancer remains one of the most challenging diseases in modern medicine. Despite substantial improvement in survival, with the 5-year relative survival increasing from 49% for patients diagnosed between 1975 and 1977 to 68% for those diagnosed between 2011 and 2017 (1), there is growing concern over the affordability of cancer care as more expensive treatments become the new standard of care (2). National costs of cancer are projected to reach $246 billion based solely on population growth (3).
Prior studies using the Surveillance, Epidemiology, and End Results–Medicare data consistently demonstrated higher costs within the first year of diagnosis and in the last year of life (4-7). However, these macrolevel estimates do not provide sufficient detail to inform policy makers about the components of treatment that drive trends in cost. Warren and colleagues (8) published one of the few comprehensive evaluations of trends in costs incurred for older patients following cancer diagnosis for the 4 most commonly diagnosed cancers: female breast, colorectal, lung, and prostate (BC, CC, LC, PC). They reported considerable variation in utilization and cost trends between 1991 and 2002.
No study has provided a comprehensive evaluation of cancer care costs among privately insured nonelderly adults, and few have reported trends in out-of-pocket (OOP) costs. Using more contemporary data, this study examines the trends in the total and OOP costs of cancer for privately insured nonelderly patients. As cancer-specific treatment patterns may lead to differences in cost trends, we report cost by cancer site and focus on the 4 most common cancers: BC, CC, LC, and PC (9).
Methods
Data Source and Study Cohort
We identified cancers diagnosed between 2009 and 2016 from the Health Care Cost Institute (HCCI). The HCCI data include claims of more than 50 million privately insured individuals annually from 3 of the 5 largest insurers nationwide—Aetna, Humana, and UnitedHealthcare— representing approximately 28% of people with employer-sponsored insurance in the United States (10).
To identify incident cohorts of patients with BC, CC, LC, and PC, we applied a previously published algorithm that required 3 or more claims of the same cancer type on different dates within 90 days of the first claim date indicating cancer diagnosis, designated as the index date (11,12). This algorithm had 88.21, 85.47, 83.18, and 80.48 sensitivity and 98.31, 99.74, 99.84, and 97.37 specificity for BC, CC, LC, and PC, respectively. We limited study cohorts to adults aged 18 to 64 years at diagnosis. To ensure costs were specific to the cancer type under investigation, we excluded patients with other cancers observed within 12 months of the index date. We further required patients to have continuous enrollment 12 months before and after the index date to ensure completeness of claims data (see Figure 1).
Ascertainment criteria of study cohorts. OOP = out-of-pocket.
The study was exempt for approval by the institutional review board at the corresponding author’s institution.
Costs Definition and Determination of Treatment Costs
Following Warren et al. (8), costs of initial cancer treatment included all services incurred from 2 months before the index date of cancer diagnosis to 12 months afterward (8). We aggregated claims from inpatient, physician, and outpatient claims in this 14-month duration and reported costs as total and OOP payment. Total payment was the sum of provider and OOP payments. OOP payment was the sum of coinsurance, copayment, and deductible. We normalized costs to 2020 US dollars using the medical care component of consumer price index (CPI). Our main analysis focused on medical benefit because only 50% of the study cohort purchased both medical and pharmacy benefits from the 3 insurers in the HCCI data. A subgroup analysis was included for the subset of patients with medical and pharmacy benefits.
At any time point, service use and costs were classified into 4 mutually exclusive treatment modalities for each patient: cancer-related surgery, intravenous (IV) systemic therapy, radiation, and other hospitalizations. See Supplementary Table 1 (available online) for codes to identify each treatment. Classification of treatment was based on the following sequence. First, we identified cancer surgeries from inpatient and outpatient claims, allocated these claims to cancer surgeries, and assigned the remaining inpatient claims to other hospitalizations. Next, we assembled claims for IV systemic therapy from the remaining outpatient claims and classified them to IV systemic therapy. We followed the same procedure to assign claims to radiation. We then calculated treatment-related costs by aggregating payment (total and OOP) associated with the patient’s claims incurred on the same date as the claims designated to a treatment modality.
Statistical Analysis
We examined service utilization and cost trends based on the year of diagnosis by cancer site. To determine whether the observed trends were statistically significant, we conducted regression analyses using year of diagnosis as an independent variable. In evaluations of the trend of service use, we conducted separate logistic regressions, each with a binary dependent variable indicating cancer surgery, IV systemic therapy, radiation, and other hospitalizations service use. We evaluated the trend of mean costs using generalized linear models with Gamma family and log link to account for the skewed distribution in cost data and applied quantile regression to examine the trend in median costs (13). All statistical tests were 2-sided, and the statistical significance level was .05. All analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC, USA) and STATA 16.0 (StataCorp LLC, College Station, TX, USA).
Results
The study cohorts consisted of 105 255 BC, 23 571 CC, 11 321 LC, and 59 197 PC patients diagnosed between 2009 and 2016. Table 1 reports the characteristics of each study cohort. Overall, the distribution of insurance type and geographic location was similar across cohorts. A higher proportion of LC and PC patients were aged 55-64 years, and more than 50% of BC and CC patients were younger than age 55 years at diagnosis.
Characteristics of patients in breast, colorectal, lung, and prostate cancer cohorts
| Patient characteristics | Breast (n = 105 255) No. (%) | Colorectal (n = 23 571) No. (%) | Lung (n = 11 321) No. (%) | Prostate (n = 59 197) No. (%) |
|---|---|---|---|---|
| Age, y | ||||
| Younger than 44 | 40 678 (38.7) | 3239 (13.7) | 524 (4.6) | 834 (1.4) |
| 45-54 | 18 537 (17.6) | 8895 (37.7) | 2860 (25.3) | 13 780 (23.3) |
| 55-64 | 46 040 (43.7) | 11 437 (48.5) | 7937 (70.1) | 44 583 (75.3) |
| Sex | ||||
| Female | 105 255 (100.0) | 10 576 (44.9) | 6123 (54.1) | 0 (0.0) |
| Male | 0 (0.0) | 12 994 (55.1) | 5198 (45.9) | 59 197 (100.0) |
| Insurance type | ||||
| HMO | 10 293 (9.8) | 2398 (10.2) | 1276 (11.3) | 6081 (10.3) |
| Othersa | 7449 (7.1) | 1678 (7.1) | 901 (8.0) | 4166 (7.0) |
| POS | 71 763 (68.2) | 16 082 (68.2) | 7443 (65.8) | 40 706 (68.8) |
| PPO | 15 750 (15.0) | 3413 (14.5) | 1701 (15.0) | 8244 (13.9) |
| Region | ||||
| Northeast | 18 686 (17.8) | 3890 (16.5) | 2113 (18.7) | 10 375 (17.5) |
| Midwest | 22 876 (21.7) | 5198 (22.1) | 2722 (24.0) | 13 009 (22.0) |
| South | 48 292 (45.9) | 11 297 (47.9) | 5229 (46.2) | 28 003 (47.3) |
| West | 15 401 (14.6) | 3186 (13.5) | 1257 (11.1) | 7810 (13.2) |
| Patient characteristics | Breast (n = 105 255) No. (%) | Colorectal (n = 23 571) No. (%) | Lung (n = 11 321) No. (%) | Prostate (n = 59 197) No. (%) |
|---|---|---|---|---|
| Age, y | ||||
| Younger than 44 | 40 678 (38.7) | 3239 (13.7) | 524 (4.6) | 834 (1.4) |
| 45-54 | 18 537 (17.6) | 8895 (37.7) | 2860 (25.3) | 13 780 (23.3) |
| 55-64 | 46 040 (43.7) | 11 437 (48.5) | 7937 (70.1) | 44 583 (75.3) |
| Sex | ||||
| Female | 105 255 (100.0) | 10 576 (44.9) | 6123 (54.1) | 0 (0.0) |
| Male | 0 (0.0) | 12 994 (55.1) | 5198 (45.9) | 59 197 (100.0) |
| Insurance type | ||||
| HMO | 10 293 (9.8) | 2398 (10.2) | 1276 (11.3) | 6081 (10.3) |
| Othersa | 7449 (7.1) | 1678 (7.1) | 901 (8.0) | 4166 (7.0) |
| POS | 71 763 (68.2) | 16 082 (68.2) | 7443 (65.8) | 40 706 (68.8) |
| PPO | 15 750 (15.0) | 3413 (14.5) | 1701 (15.0) | 8244 (13.9) |
| Region | ||||
| Northeast | 18 686 (17.8) | 3890 (16.5) | 2113 (18.7) | 10 375 (17.5) |
| Midwest | 22 876 (21.7) | 5198 (22.1) | 2722 (24.0) | 13 009 (22.0) |
| South | 48 292 (45.9) | 11 297 (47.9) | 5229 (46.2) | 28 003 (47.3) |
| West | 15 401 (14.6) | 3186 (13.5) | 1257 (11.1) | 7810 (13.2) |
Others included exclusive provider organization, indemnity, and other health insurance plans. HMO = health maintenance organization; POS = point of service; PPO = preferred provider organization.
Characteristics of patients in breast, colorectal, lung, and prostate cancer cohorts
| Patient characteristics | Breast (n = 105 255) No. (%) | Colorectal (n = 23 571) No. (%) | Lung (n = 11 321) No. (%) | Prostate (n = 59 197) No. (%) |
|---|---|---|---|---|
| Age, y | ||||
| Younger than 44 | 40 678 (38.7) | 3239 (13.7) | 524 (4.6) | 834 (1.4) |
| 45-54 | 18 537 (17.6) | 8895 (37.7) | 2860 (25.3) | 13 780 (23.3) |
| 55-64 | 46 040 (43.7) | 11 437 (48.5) | 7937 (70.1) | 44 583 (75.3) |
| Sex | ||||
| Female | 105 255 (100.0) | 10 576 (44.9) | 6123 (54.1) | 0 (0.0) |
| Male | 0 (0.0) | 12 994 (55.1) | 5198 (45.9) | 59 197 (100.0) |
| Insurance type | ||||
| HMO | 10 293 (9.8) | 2398 (10.2) | 1276 (11.3) | 6081 (10.3) |
| Othersa | 7449 (7.1) | 1678 (7.1) | 901 (8.0) | 4166 (7.0) |
| POS | 71 763 (68.2) | 16 082 (68.2) | 7443 (65.8) | 40 706 (68.8) |
| PPO | 15 750 (15.0) | 3413 (14.5) | 1701 (15.0) | 8244 (13.9) |
| Region | ||||
| Northeast | 18 686 (17.8) | 3890 (16.5) | 2113 (18.7) | 10 375 (17.5) |
| Midwest | 22 876 (21.7) | 5198 (22.1) | 2722 (24.0) | 13 009 (22.0) |
| South | 48 292 (45.9) | 11 297 (47.9) | 5229 (46.2) | 28 003 (47.3) |
| West | 15 401 (14.6) | 3186 (13.5) | 1257 (11.1) | 7810 (13.2) |
| Patient characteristics | Breast (n = 105 255) No. (%) | Colorectal (n = 23 571) No. (%) | Lung (n = 11 321) No. (%) | Prostate (n = 59 197) No. (%) |
|---|---|---|---|---|
| Age, y | ||||
| Younger than 44 | 40 678 (38.7) | 3239 (13.7) | 524 (4.6) | 834 (1.4) |
| 45-54 | 18 537 (17.6) | 8895 (37.7) | 2860 (25.3) | 13 780 (23.3) |
| 55-64 | 46 040 (43.7) | 11 437 (48.5) | 7937 (70.1) | 44 583 (75.3) |
| Sex | ||||
| Female | 105 255 (100.0) | 10 576 (44.9) | 6123 (54.1) | 0 (0.0) |
| Male | 0 (0.0) | 12 994 (55.1) | 5198 (45.9) | 59 197 (100.0) |
| Insurance type | ||||
| HMO | 10 293 (9.8) | 2398 (10.2) | 1276 (11.3) | 6081 (10.3) |
| Othersa | 7449 (7.1) | 1678 (7.1) | 901 (8.0) | 4166 (7.0) |
| POS | 71 763 (68.2) | 16 082 (68.2) | 7443 (65.8) | 40 706 (68.8) |
| PPO | 15 750 (15.0) | 3413 (14.5) | 1701 (15.0) | 8244 (13.9) |
| Region | ||||
| Northeast | 18 686 (17.8) | 3890 (16.5) | 2113 (18.7) | 10 375 (17.5) |
| Midwest | 22 876 (21.7) | 5198 (22.1) | 2722 (24.0) | 13 009 (22.0) |
| South | 48 292 (45.9) | 11 297 (47.9) | 5229 (46.2) | 28 003 (47.3) |
| West | 15 401 (14.6) | 3186 (13.5) | 1257 (11.1) | 7810 (13.2) |
Others included exclusive provider organization, indemnity, and other health insurance plans. HMO = health maintenance organization; POS = point of service; PPO = preferred provider organization.
Trend in Total Costs
Mean total cost exhibited statistically significantly upward trend for BC (P < .001), LC (P < .001), and PC (P = .006). Although the mean total cost was highest for LC, the largest increase was observed in BC, increasing from $109 544 in 2009 to $140 743 in 2016, followed by LC ($151 751 to $168 730) and PC ($53 300 to $55 497). After inflation adjustment, the average annual growth rate was 3.7%, 1.5%, and 0.6% for BC, LC, and PC, respectively (Figure 2, A). Mean total cost of CC, although higher than BC and PC, did not statistically significantly increase over time (from $136 652 to $137 663; P = .09). Similar patterns were observed in median total cost (Figure 2, B).
Total and OOP cost for privately insured nonelderly patients with BC, CC, LC, and PC diagnosed between 2009 and 2016. A) Mean total cost, (B) median total cost, (C) mean OOP cost, and (D) median OOP cost are shown. Panels A and C estimate mean costs using generalized linear models with gamma family and log link. Panels B and D estimate median costs using quantile regression. P values were calculated using Z test for means and t test for medians and represent whether the time trend of costs from regression models was statistically significant. All statistical tests were 2-sided. AGR = average growth rate; BC = breast cancer; CC = colorectal cancer; LC = lung cancer; OOP = out-of-pocket; PC = prostate cancer.
Both mean and median OOP costs have a statistically significantly rising trend for each cancer, with the annual growth rate between 2.0% and 2.7% and between 2.7% to 3.8% for mean (Figure 2, C) and median (Figure 2, D) OOP costs, respectively. For patients with BC, CC, and LC, mean OOP costs increased from low-to-mid $5000 in 2009 to low-to-mid $6000 in 2016, whereas mean OOP costs for patients with PC grew from $3830 to $4604 during the same period. Figure 3 deconstructs mean OOP costs into deductible, copayment, and coinsurance. Deductibles accounted for an increasingly larger proportion of OOP across all 4 cancers, rising from approximately 30% in 2009 to approximately 40% in 2016.
Component of OOP costs for privately insured nonelderly patients with breast, colorectal, lung, and prostate cancer diagnosed between 2009 and 2016. Components of OOP costs for (A) breast cancer, (B) colorectal cancer, (C) lung cancer, and (D) prostate cancer are shown. OOP = out-of-pocket.
Trend in Cancer Surgery
Approximately 90% of patients with BC or CC received surgical treatment (Figure 4, A). The rate of cancer surgery statistically significantly increased for BC (P < .001) and CC (P = .02), decreased for PC (P < .001), and exhibited no consistent pattern for LC (P = .28). Costs of cancer surgeries rose statistically significantly, except for LC, and were highest for CC (annual growth 2.3%, from $45 369 in 2009 to $53 140 in 2016), with the largest growth in surgical cost observed in BC (annual growth 3.9%, from $23 422 to $30 598). Use of minimally invasive surgery (MIS) in the form of robotic-assisted or laparoscopic procedures increased statistically significantly in CC (from 42% in 2009 to 58% in 2016; P < .001) and PC (72% to 87%; P < .001) surgeries (Figure 4, C) but were uncommon in BC and LC. Costs of MIS grew 3.9% and 2.9% for CC and PC, respectively; cost changes of non-MIS were not statistically significant (Figure 4, D; P = .07 for CC and P = .20 for PC).
Utilization and cost trends of cancer-related surgeries for privately insured nonelderly patients with BC, CC, LC, and PC diagnosed between 2009 and 2016. A) Use of cancer surgery, (B) mean cost of cancer surgery, (C) use of minimally invasive surgery among patients with cancer surgeries, and (D) mean cost of miniminally vs nonminially invasive surgery are shown. Panel A estimates the trend of cancer surgery among cancer patients using logistic regressions. Panel B estimates mean costs among patients who received cancer surgeries using generalized linear models with gamma family and log link. Panel C estimates the trend in the use of minimially invasive surgeries among CC and PC patients who had cancer surgery using logistic regressions. Panel D uses generalized linear models with gamma family and log link to estimate mean costs of surgeries for CC and PC patients who received minimally invasive surgeries and those who did not have minimally invasive surgeries. P values were calculated using Z test and represent whether the time trend of costs from regression models was statistically significant. All statistical tests were 2-sided. AGR = average growth rate; BC = breast cancer; CC = colorectal cancer; LC = lung cancer; MIS = minimally invasive surgery; PC = prostate cancer.
Trend in IV Systemic Therapy
The percentage of patients who received IV systemic therapy was 48%, 50%, 63%, and 15% in 2009 for BC, CC, LC, and PC, respectively. The percentage increased statistically significantly for CC, and PC, to 54% and 17% in 2016, but remained relatively stable for BC and declined to 59% for LC (P = .002) (Figure 5, A). The mean cost of IV systemic therapy increased 8.1% annually for BC (from $53 888 in 2009 to $92 418 in 2016; P < .001), 8.7% for PC ($8528 to $15 154; P < .001), and 3.6% for LC ($74 186 to $92 796; P < .001) but decreased from $100 719 to $64 639 (P < .001) for CC (Figure 5, B).
Utilization and cost trends of intravenous systemic therapy for privately insured nonelderly patients with BC, CC, LC, and PC diagnosed between 2009 and 2016. (A) Use of intravenous systemic therapy and (B) mean cost of intravenous systemic therapy are shown. Panel A estimates the trend of intravenous systemic therapy among cancer patients using logistic regressions. Panel B uses generalized linear models with gamma family and log link to estimate mean costs of intravenous systemic therapy among patients who received intravenous systemic therapies. P values were calculated using Z test and represent whether the time trend of costs from regression models was statistically significant. All statistical tests were 2-sided. AGR = average growth rate; BC = breast cancer; CC = colorectal cancer; LC = lung cancer; PC = prostate cancer.
Trend in Radiation Therapy (RT)
The percentage of patients who received RT was 60%, 17%, 42%, and 32% in 2009 for BC, CC, LC, and PC, respectively. The percentage increased statistically significantly for BC and CC, to 62% (P = .003) and 20% (P < .001) in 2016, decreased statistically significantly for PC (29%; P < .001), and remained stable for LC (P = .68) (Figure 6, A). The largest increase in RT costs was in CC, with the mean cost increasing 3.2% annually (from $37 148 in 2009 to $46 185 in 2016; P < .001), followed by LC (2.9% annual increase, $40 906 to $49 535; P < .001), and BC (0.8% annual increase, $38 266 to $40 274; P < .001). For PC, RT costs were reduced 1.4% annually, from $62 708 to $56 658 (P < .001) over the same period (Figure 6, B).
Utilization and cost trends of radiation therapy for privately insured nonelderly patients with BC, CC, LC, and PC diagnosed between 2009 and 2016. A) Use of radiation and (B) mean cost of radiation are shown. Panel A estimates the trend of radiation among cancer patients using logistic regressions. Panel B uses generalized linear models with gamma family and log link to estimate mean costs of radiation among patients who received radiation therapy. P values were calculated using Z test and represent whether the time trend of costs from regression models was statistically significant. All statistical tests were 2-sided. AGR = average growth rate; BC = breast cancer; CC = colorectal cancer; LC = lung cancer; PC = prostate cancer.
Trend in Other Hospitalizations
Among patients with hospitalizations other than cancer surgeries, the percentage of hospitalization statistically significantly declined for all 4 cancers (Figure 7, A). Patients with LC had the highest percentage of other hospitalizations. Mean costs of hospitalizations increased statistically significantly over time, with the highest costs in LC (increasing from $44 577 to $57 756) and the largest annual growth rate in PC (4.5%) (Figure 7, B).
Utilization and cost trends of hospitalizations other than cancer surgeries for privately insured nonelderly patients with BC, CC, LC, and PC diagnosed between 2009 and 2016. A) Use of other hospitalizations and (B) mean cost of other hospitalizations are shown. Panel A estimates the trend of hospitalizations for reasons other than cancer surgery among cancer patients using logistic regressions. Panel B uses generalized linear models with gamma family and log link to estimate mean costs of other hostpitalizations among patients who were hospitalized for reasons other than cancer surgeries. P values were calculated using Z test and represent whether the time trend of costs from regression models was statistically significant. All statistical tests were 2-sided. AGR = average growth rate; BC = breast cancer; CC = colorectal cancer; LC = lung cancer; PC = prostate cancer.
Subgroup Analysis
Subgroup analysis included patients with medical and pharmacy benefits. Characteristics of this subgroup (Supplementary Table 2, available online) were similar to the cohorts in the main analysis; the most noticeable difference is a higher proportion of patients in the subgroup were enrolled in an health maintenance organization. Total and OOP costs were higher than those reported in the main analysis, although the rate of increase in OOP costs was smaller. However, mean, median total, and OOP costs from medical benefits (Panels A2 and B2 in Supplementary Figures 1 and 2, available online) exhibited similar trends as those in the main analysis (Figure 2). When focusing on OOP costs for outpatient prescription drugs, mean and median OOP costs steadily declined over time (Panels A3 and B3 in Supplementary Figure 2, available online). The share of OOP costs for prescription drugs decreased from 12% to 17% in 2009 to 6% to 10% in 2016 (Supplementary Figure 3, available online). The decomposition of mean OOP costs showed an increasing share of deductibles over time, from approximately 30% in 2009 to more than 40% in 2016. With the inclusion of pharmacy benefits, the share of coinsurance was approximately 40% (Supplementary Figure 4, available online), compared with approximately 50% in the main analysis (Figure 3).
Discussion
This study analyzed a large commercial insurance claims database to estimate total and OOP costs for nonelderly adults with BC, CC, LC, and PC diagnosed between 2009 and 2016. Three important conclusions emerge. First, rising cost sharing from private insurance has exacerbated financial burden even in cancers with little or no increase in treatment costs. Second, with few exceptions, costs increased for nearly all treatment modalities; and third, costs were higher despite the reduction in hospitalizations for reasons other than cancer surgeries.
Our study provides new evidence of the growing financial burden for cancer patients. This trend is consistent with the rising health-related financial burden among the privately insured in the general population. HCCI reports showed OOP costs for medical benefits increased 2.9% annually after inflation adjustment, from $458 ($633 in 2020 dollars) in 2009 to $691 ($733 in 2020 dollars) in 2016 (14–16). In comparison, OOP costs for patients with BC, CC, LC, and PC were approximately 8.5, 8.3, 8.6, and 6.1 times higher than those without these diagnoses, respectively. As reported by others, the trend in total costs observed for BC, LC, and PC are associated with increasing OOP payment over time (17–19). Most noticeably, the trend in CC reveals a concerning pattern not noticed previously. Whereas total mean costs per patient remained relatively stable between 2009 and 2016, OOP costs rose at a rate of 2% higher than the rate of medical cost CPI annually. This observation has important policy implications as it suggests cost containment policies alone may not be sufficient to ease cancer patients’ financial hardship if patients are asked to bear an increasingly larger share of health-care cost. Our analysis examining the components of OOP costs showed a consistent pattern across all 4 cancers that deductibles account for an increasing proportion of OOP costs (Figure 3). By 2018, more than 40% of adults with employment-based insurance had high-deductible plans, with the amount of deductible for individual coverage as high as $6750 in some plans (20).
We observed a trend of decreasing cost of IV systemic therapy for CC and radiation for PC. Both trends indicate pricing and reimbursement policies can be effective tools to curb the costs of cancer care. For CC, a major contributor for cost reduction in IV systemic therapy was a generic entry of 2 commonly used chemotherapy agents, oxaliplatin and irinotecan, following their patent expiration in 2007 and 2013, respectively. FOLFOX (leucovorin calcium, fluorouracil, and oxaliplatin) and FOLFIRI (leucovorin calcium, fluorouracil, and irinotecan) have long been the preferred first-line treatment for CC (21). Average sale price for these drugs reduced dramatically as generics became available, by more than 90% for oxaliplatin and 70% for irinotecan from 2009 to 2016. As more than 75% and 11% of CC patients with IV systemic therapy had a chemotherapy regimen containing oxaliplatin and irinotecan, respectively, reduction in drug prices enabled by generic entry appears to have lowered the mean cost of IV systemic therapy in CC. For PC, the use of intensity-modulated radiation therapy grew from 68% in 2009 to 78% in 2016. A reduction in the reimbursement rate of intensity-modulated radiation therapy and the increasing use of hypofractionation of treatment courses (22,23) contributed to the trend of decreasing radiation costs.
Cancer surgeries for patients with LC and PC cancer reported in our study were at least 10 percentage points higher than those reported in Warren et al. (8). Higher rates of cancer surgeries likely reflect more aggressive treatment common among younger patients. Rising cost of cancer surgeries in CC and PC is likely driven by the growing use of MIS. Increase in cost of BC surgeries cannot be explained by MIS, as such procedures are rare in BC. Consistent with national trends, rates of other hospitalizations declined for all 4 cancers (24). The trend suggests that oncology practices are better at managing symptoms and adverse events associated with cancer and its treatment. Nonetheless, costs of other hospitalizations continue to rise, a trend consistent with research that reported substantial growth in hospital prices among the privately insured (25).
Subgroup analysis suggests the increase of OOP costs was not driven by spending on prescription drugs. Although research has found a sharp growth in the prices of oral anticancer medications (11,26), most oral anticancer medications approved before 2017 were for treatment of leukemia, lymphoma, kidney, and some subtypes of lung and breast cancer. Also, few (<3%) patients in our data incurred coinsurance for their oral prescription drugs, which offered some financial protection for patients from having to shoulder a proportion of costs for fast-rising drug prices.
Our study has limitations. First, we relied on a claims-based algorithm to identify incident cohorts and were unable to stratify analyses by cancer stage. However, patterns observed in our study are unlikely driven by changes in cancer stage. Information from the National Cancer Database indicates that the stage distribution among the privately insured, similar to patients included in our study, was stable between 2010 and 2016 with the exception of increasing incidence of distant stage PC (Supplementary Figure 5, available online). Rising distant stage may contribute to the increasing use of IV systematic therapy in PC observed here. Second, the requirement to have continuous insurance coverage for the entire study period excluded patients who died within the first year of diagnosis, potentially underestimating costs because of underrepresentation of late-stage cancer patients who often receive intensive therapy. This issue differentially affected the lung cancer sample, as evident by a larger reduction in sample size after imposing this requirement. Consequently, the proportion of IV systemic therapy and other hospitalizations as well as costs for patients with LC was likely underestimated. Our study period also does not capture costs of late toxicities from treatment and ongoing survivorship care that occur more than 12 months following diagnosis. The assessment of trends is not affected, however. Third, our study underestimates the financial impact of immunotherapy agents because most were approved after 2016 and are substantially more costly. In addition, we identified IV systemic therapy from billing codes for IV anticancer medications. Therefore, costs associated with IV systemic therapy would capture supportive care drugs administered intravenously on the same day as IV anticancer medications but would not reflect oral supportive care medications or those administered on different days. Future research should comprehensively examine the utilization and cost trends of supportive care medications using inclusion and exclusion criteria specifically designed for this purpose. Fourth, OOP costs estimated from claims do not capture over-the-counter medications, nutritional supplements, and complementary and alternative medicines not covered by insurance—all are common among cancer patients (27,28). Lastly, findings, especially treatment patterns, from our study may not generalize to elderly cancer patients or those with public insurance or who are uninsured.
Growth in the OOP costs of cancer care consistently outpaced medical care CPI between 2009 and 2016. By 2016, OOP costs for services covered under medical benefits alone were $6000 or higher for patients with BC, CC, and LC and more than $4500 for PC. The growing prevalence of high-deductible plans and the rising cost-sharing requirements in private insurance plans will continue to exert financial burden. This trend calls for regulations to insurance benefit design. Ultimately, a meaningful reduction in cancer-related financial burden can only be achieved through initiatives that contain rising costs of cancer care on multiple fronts.
Funding
YTS acknowledges funding from the National Cancer Institute (NCI) (R01CA207216, R01CA225646, R01CA225647). CB acknowledges funding from the NCI (P30CA44688). SHG is supported by CPRIT RP160674 and Komen SAC150061. This study was also supported, in part, by NCI P30 CA016672.
Notes
Role of the funder: The National Cancer Institute funded this research but had no role in the design or conduct of the study; analysis or interpretation of the data; review or approval of the manuscript; or the decision to submit the manuscript of the publication.
Disclosures: YTS served as a consultant for a review panel for Pfizer Inc. and an advisory board for AstraZeneca in 2019. KRY serves on the Flatiron Health Equity Advisory Board. All other authors have no conflicts of interest to disclose on the subject of this manuscript. YTS and KRY, who are JNCI Associate Editors and co-authors on this article, were not involved in the editorial review or decision to publish this manuscript.
Author contributions: Conceptualization: YTS, YX, CB, KRY. Resources, funding acquisition, and supervision: YTS. Data curation, software, formal analysis, and project administration: YTS, YX. Methodology and visualization: YTS, YX, CB, KRY. Validation, investigation, writing-original draft and review & editing: YTS, CB, KRY. Critical revision of the article for important intellectual content: YX, SHG, JY. Final approval of the article: YTS, YX, CB, SHG, JY, KRY.
Prior presentation: Preliminary study findings were presented virtually on September 25 at the 2021 ASCO Quality Care Symposium.
Data Availability
The raw/processed data required to reproduce the above findings cannot be shared under the data use agreement between University of Texas MD Anderson Cancer Center and the Health Care Cost Institute.
