All-Payer Claims Databases (APCDs) present an opportunity to close gaps in the United States' fragmented health data infrastructure, allowing for a more complete assessment of patient-centered outcomes. APCDs collect health care claims from multiple payers with the goals of assessing the value and affordability of health care and improving cost transparency. APCDs are administered at the state level and include medical, pharmacy, and dental claims, and eligibility and provider files from most public and private payers.1,2 In 2022, 25 states had either a voluntary APCD or a statutorily mandated APCD in which covered payers are required to submit claims; 6 other states are in the process of developing APCDs.3 The No Surprises Act implemented in 2022 increased federal support to enhance state APCDs.4 APCDs have spurred the interest of researchers because, in some states, these datasets longitudinally track patients across insurers, as well as identify patients with multiple insurance plans.3 They also track patients across health systems and facilities to capture the full array of utilization and costs. Depending on data availability, it may be possible to link individuals insured under the same plan, which could allow health care assessments at the household level. In the absence of APCDs, researchers often rely on single payers (eg, Medicare5,6 and Medicaid7) or proprietary datasets at the institution or health system level that are limited in their ability to track patients across payers and providers to create a comprehensive picture of health care use and costs.8,9

APCDs, however, are limited to people who are insured, making it impossible to track care provided to people with the least access to health care or those who experience gaps in coverage.10 Race and ethnicity are incomplete in claims.11–13 The absence of these data hinders health equity assessments on treatment, costs, and outcomes. The National Association of Health Data Organizations found that only 28% of records have usable race information.3,14 In addition, self-insured employers regulated under the Employee Retirement Income Security Act (ERISA) are not required to report to the APCD. As with other claims data, APCDs lack information on biomarkers, detailed clinical information beyond what can be captured by procedure and diagnosis codes, and laboratory values that are important for determining disease onset and severity. Without this information, the ability to assess the quality of care a patient receives can be limited.

Linking APCDs with disease registries that track incidence, clinical details, and vital status over time may overcome some of these limitations. Cancer registries, which meet standards set by the North American Association of Central Cancer Registries,15 are an excellent example of how APCDs can be enhanced through linkages. Cancer registries have patient and tumor-level data on the date of diagnosis, disease stage, and vital status. Patient data are nearly complete on variables, such as race and ethnicity, and offer granular classifications.16 However, treatment data are incomplete in registries, especially beyond the first course of treatment,16 and can systematically vary by region, cancer site, insurance payer, and reporting source.17 Because registries are population-based, linkages between APCDs and registries allow for assessments of people missing in APCDs. Although this does not solve the problem of missing data, these assessments may identify whether the missingness is systematic across geographic regions (eg, urban, rural, and frontier), patient characteristics (eg, age and race/ethnicity), or census tract characteristics (eg, high or low-income) that inform researchers about potential biases. Registries also have address information that allows researchers to construct area-level indices of social determinants of health, such as the Area Deprivation Index, the Social Deprivation Index (SDI), or the Social Vulnerability Index.18 These indices can be augmented by individual race/ethnicity in the Colorado Central Cancer Registry (CCCR) and insurance data in the APCD.

This paper uses the linked Colorado APCD and CCCR data as a prototype for building a data infrastructure centered on APCD linkages. In Colorado, insurers covering over 1000 lives are mandated to submit claims to the APCD. Colorado is one of 5 states where ERISA plans submit claims voluntarily, with ~25% of Colorado’s ERISA-covered plans submitting claims.3 Colorado creates longitudinal identifiers for each patient to allow tracking of patients over time and across insurance payers. Linking APCDs to cancer or other registries may be applicable to other states and conditions where registry and cohort data exist and can be linked to claims data.9,19–21

We describe the process of linking both data sources over multiple years and assess the challenges of using these data while demonstrating the potential role of APCDs in supporting patient-centered economic outcomes research. Prior published studies have discussed treatment completeness in cancer registry and APCD linkages, highlighting the role of the APCD in augmenting information beyond the initial course of treatment.22,23 Therefore, our discussion focuses on less explored topics, specifically, a comparison of patient characteristics in both data sources to highlight the advantages of complementing information in each source, a description of insurance coverage captured in the APCD, and a comparison of out-of-pocket costs that play a critical role in economic outcomes.

METHODS Data Sources and Linkage

The Center for Improving Value in Health Care (CIVHC), a not-for-profit organization, is authorized by the state to collect and administer the Colorado APCD.24 The Colorado APCD includes medical claims and dates of service from commercial health plans (large group, small group, and individual), Traditional Fee-for-Service Medicare, Medicare Advantage, and Colorado’s Medicaid Program. The CCCR is managed by the Colorado Department of Public Health and Environment.25 Claims from private payers and Medicaid submitted to the APCD have a delay of 3–6 months, whereas the CCCR data submissions are completed 18–24 months after the end of the calendar year. Therefore, the linked data will be a minimum of 2–3 years old once the data are ascertained, linked, and validated. To augment these data, our Data Use Agreement (DUA) permitted the inclusion of variables that allowed us to code Geographic Underserved Areas26 at the census tract and/or county level using the American Community Survey. Using these data, we constructed the SDI based on where individuals live.27 We also included variables for whether a person resides in a frontier and remote (FAR) area based on FAR codes.28

The linkage was performed by the CCCR registrar using personal identifying data from CIVHC. Registry authorizing legislation allowed them to collect additional data on people who were in the CCCR. The CCCR used a probabilistic approach with Social Security Number, date of birth, last name, first name, and sex as identifiers in both datasets. Of the potential matches, 31.5% were partial matches that were manually reviewed. The manual review proceeded in stages, first reviewing partial matches with only one field not matching (70.5% of partial matches), then partial matches with two missing fields, and so on. Most partial matches (80.5%) had only partial or missing Social Security Number but agreed on the other 4 identifiers. Our dataset covers adults aged 21 years and older diagnosed with cancer in Colorado from 2012 to 2017. Of 146,884 patients with a cancer diagnosis during the period, 136,613 patients were successfully linked (93%), with near-perfect linkage rates for those enrolled in Medicare and Medicaid. Because the linkage was performed using 7 years of data, a patient could be linked at any point in time but not necessarily at the time of diagnosis. We restricted the sample to linked patients with information in the APCD at the month of diagnosis (70.8% of linked patients) and complete age and sex data, leaving a total of 91,883 patients in the analytical dataset. Additional details on the linkage process are available elsewhere.22 An updated linkage, including 2022 data, is underway.

Assessments

In both data sources, we evaluated patients’ characteristics, insurance status, and out-of-pocket payments to assess how the data sources complement each other and to assess the extent to which these data can be used in future patient-centered outcomes research and health equity evaluations. We examined patients’ race/ethnicities (White, non-Hispanic; Black, non-Hispanic; Hispanic; Other; and Unknown), sexes, and age categories (21–40, 41–60, 61–80, and 80 or older) at the time of the first tumor diagnosis per patient during 2012–2017. Each of these characteristics is associated with cancer treatment and outcomes.29–31 To understand the distribution of patient characteristics and the degree of missingness among the variables, we compared the CCCR and APCD, treating registry data as the gold standard for demographic information. We used only categories that are common between data sources, but registry data contained more detailed patient information than the APCD. For example, registry data are based on self-report and include sex codes for “other (hermaphrodite)” and “transexual,” and race and ethnicity classifications are also more granular.32 We report χ2 tests comparing the distribution of these characteristics between the two sources.

For the assessment of insurance coverage, we used only the APCD because cancer registries are limited in their ability to classify patient insurance status.33 With the APCD, insurance at any point in time is determined by enrollment information submitted by payers. We provided descriptive data based on payer type (public and private), coverage (medical and pharmacy), and plan type (fee-for-service, managed care). We also examined the number of patients with multiple insurance plans, and the extent to which the primary payer in the APCD agreed with the primary payer identified in the CCCR 1 month after diagnosis. To assess insurance agreement, we excluded plans not captured in the APCD but captured in the CCCR, such as Veteran Affairs plans and other insurance for armed forces and Indian Health Services but retained patients who were regarded as having no insurance or with unknown insurance status in the registry. We excluded dental plans from the analysis. Last, we noted the number and percentage of patients with <3 and 6 months of coverage after the month of diagnosis and those who changed plans within 12 months of diagnosis among patients who survived for 3, 6, and 12 months.

We also assessed geographic data, such as whether the individual lived in FAR. FAR is defined by the Economic Research Service and is based on population size and geographic remoteness, characterized by how long it takes to travel by car to the edges of nearby urban areas.34 There are 4 FAR levels, but due to small sample sizes, we grouped patients by whether they lived in a FAR area (yes/no). We did not have information a priori on which of the area-level indices were informative for predicting cancer outcomes. We selected the SDI.27 SDI is a composite measure of area-level deprivation based on 7 demographic characteristics collected in the American Community Survey and used to quantify the socioeconomic variation in health outcomes. The components are the percentage of people living in poverty, the percentage with <12 years of education, the percentage of families that are single-parent households, the percentage of households in a rented housing unit, the percentage of households in an overcrowded housing unit, the percentage of households without a car, and the percentage unemployed adults under 65 years of age. The SDI is defined at different geographic areas, including Census ZIP Code Tabulation Areas but not ZIP codes. We used a crosswalk to convert APCD postal ZIP codes into ZIP Code Tabulation Areas.35

Last, we estimated out-of-pocket costs for patients who survived 6 months after diagnosis, including the month of diagnosis. Out-of-pocket costs are an important predictor of the financial burden associated with a cancer diagnosis.36,37 Out-of-pocket costs were calculated as the sum of copays, deductibles, and/or applicable coinsurance. For Medicaid and dual enrollees, deductibles and coinsurance amounts were not included. These items were, at times, incorrectly reported in the APCD before 2017.38 CIVHC requires insurers with capitated plans to submit their equivalent fee-for-service schedule for the services provided39 alongside out-of-pocket charges.

RESULTS

Table 1 reports demographic characteristics for the 91,883 patients who were linked and had APCD data at the month of diagnosis by data source (67%). Age and sex are nearly identical, which is expected, as both identifiers were used for the probabilistic linkage, although sex was mostly used in the manual review of potential matches. The APCD claims do not capture race and ethnicity well, with 53.0% of race and ethnicity coded as “unknown” (31.7%) or “missing (21.4%).” The APCD only correctly captures 31% of the Hispanic population reported by the registry. Marital status, which can be an important predictor of cancer treatment and survival,40,41 was reported for 95% of the patients in the CCCR. Most patients resided in Colorado (93%) and did not live in a FAR area (89.9%). About 31% of patients lived in an area with the greatest level of deprivation as measured by the SDI.

TABLE 1 - Patient Characteristics by Data Source, 2012–2017, N = 91,883 Characteristic CCCR APCD P Sex  Female 44,648 (48.59) 44,617 (48.56) 0.8850  Male 47,235 (51.41) 47,266 (51.44) — Age at the month of diagnosis* (y)  21–40 4132 (4.50) 4138 (4.50) 0.9504  41–60 19,198 (20.89) 19,272 (20.97) —  61–80 52,755 (57.42) 52,755 (57.42) —  80+ 15,798 (17.19) 15,718 (17.11) — Race/ethnicity  White/non-Hispanic 75,211 (81.86) 30,284 (32.96) <0.0001  Black/non-Hispanic 3083 (3.36) 1676 (1.82) —  Hispanic 10,378 (11.29) 3208 (3.49) —  Other 1819 (1.98) 7993 (8.70) —  Unknown 1392 (1.51) 29,099 (31.67) —  Missing 0 19,623 (21.36) — Marital status  Missing 4859 (5.33) NA —  Not married or partnered 40,219 (43.77) NA —  Married or partnered 46,765 (50.90) NA — Colorado residents  Not in Colorado 0 1108 (1.21) <0.0001  Colorado 85,762 (93.34) 90,575 (98.58) —  Missing 6121 (6.66) 200 (0.22) — Patients who lived in FAR areas (FAR by Zip code)†  In FAR areas NA 7109 (7.74) —  Not in FAR areas NA 82,603 (89.90) —  Missing NA 2171 (2.36) — Quintile of 2015 SDI (SDI by Colorado Zip code)‡  1 (1–15) NA 21,020 (22.88) —  2 (16–32) NA 18,840 (20.50) —  3 (33–53) NA 22,107 (24.06) —  4 (54–99) NA 28,235 (30.73) —  Missing ZIP code or outside Colorado NA 1681 (1.83) —

Data for the first tumor between 2012 and 2017. Data presented as number of patients and percentage.

*Age at the month of diagnosis category for the APCD column was calculated based on the date of diagnosis (only month and year available) from CCCR and the date of birth from APCD.

‡We calculated the quantiles of the 2015 Colorado Social Deprivation Index at the ZIP code level. Higher SDI equates to higher levels of disadvantage.

APCD indicates All-payer Claims Database; CCCR, Colorado Central Cancer Registry; FAR, frontier and remote; NA, not available; SDI, Social Deprivation Index.

Appendix Table A1 (Supplemental Digital Content 1, https://links.lww.com/MLR/C700) reports insurance coverage at the month of cancer diagnosis (first tumor) based on APCD enrollment information. Most patients have pharmacy benefits (70.1%), and a small portion of patients have pharmacy plans without medical plans (4.6%). These patients may have medical coverage in a plan that does not report to the APCD. We identified Medicare beneficiaries with supplemental plans (14.3%). Approximately 23% of patients were enrolled in Medicare Advantage plans, which is lower than enrollment statewide.42 Just over a third (36.7%) of patients with private plans had managed care plans. The agreement between the APCD and CCCR in plan ascertainment is relatively low (69.2%), which highlights the importance of relying on APCD data for accurate insurance identification. After a cancer diagnosis, few patients became uninsured or switched insurance to a plan outside of the APCD. At 3 and 6 months, only 2.7% and 5.0%, respectively, no longer had evidence of insurance coverage in the APCD. Of those who survived at least a year and had continuous enrollment in the APCD, 5.3% switched insurance plans during the 12 months after diagnosis.

Table 2 reports out-of-pocket expenses for patients by insurance type. The mean out-of-pocket spending for patients with medical-only plans was $3074, whereas those with pharmacy benefits in addition to medical benefits had an average of $1950. Among Traditional Medicare enrollees, patients with a supplemental plan had on average $3568 lower out-of-pocket expenses than those without a supplemental plan or part D coverage. We note that 82.0% of patients with Traditional Medicare and supplemental coverage did not have out-of-pocket charges, suggesting they had complete coverage.

TABLE 2 - Out-of-pocket Patient Costs, 6 Months After Cancer Diagnosis Including the Month of Diagnosis, Based on APCD Claims, 2012–2017, N = 91,874 Insurance coverage month of diagnosis Mean (SD) Median (IQR) All plans  Medical-only 3074 (18516) 549 (0, 2611)  Medical + pharmacy 1950 (8257) 479 (31, 2200)  Pharmacy-only 387 (1637) 150 (39, 385) Medicare (N = 33,806, 36.80%)  Traditional part A and/or B, no supplement 4015 (20682) 1426 (233, 3830)  Traditional part A and/or B, supplement 447 (1774) 0 (0, 1050)*  Traditional part A and/or B, D† 2369 (17381) 241 (61, 741)  Medicare Advantage plan 1321 (1672) 711 (186, 1781) Private (N = 18,064, 19.66%)  Medical-only 2326 (7472) 258 (0, 2778)  Medical and pharmacy 3855 (7166) 2690 (996, 5165)  Pharmacy-only 201 (499) 30 (0, 163)  Single private plan 3608 (7250) 2350 (644, 4957)  Two or more private plans 2663 (3225) 1879 (343, 3533)  Fee-for-service 4656 (8294) 3386 (1315, 5860)  Managed care 2808 (2851) 1880 (726, 4032)  Self-funded or unknown 2648 (9694) 712 (0, 3452)

Data presented in United States dollars. All patients were continuously enrolled in the same plan for 6 months including the month of diagnosis. Patients who enrolled in dental plans only were excluded (n = 9). We did not include patients enrolled in Medicaid due to data reporting errors.

Out-of-pocket costs defined as deductibles, copays, and coinsurance.

*90th percentile; 81.92% of patients have $0 out-of-pocket expenses reported.

†With or without supplemental plans.

APCD indicates All-Payer Claims Database; IQR, interquartile range.

Out-of-pocket payments were lower for patients enrolled in Medicare Advantage plans ($1321) compared with those enrolled in Traditional Medicare without supplemental plans ($4015) and those with part D coverage ($2369). Among patients privately insured, those enrolled in managed care and self-funded plans had lower mean out-of-pocket expenditures ($2808 and $2648, respectively) than those enrolled in fee-for-service plans ($4656). Appendix Tables A2 (Supplemental Digital Content 2, https://links.lww.com/MLR/C701) and A3 (Supplemental Digital Content 3, https://links.lww.com/MLR/C702) report out-of-pocket expenses by FAR and SDI.

DISCUSSION

We highlighted the use of APCD and cancer registry linkages to improve the data available for health economics studies of patient-centered outcomes. Key advantages of registry data include better measurement and classification of race, ethnicity, and sex, allowing investigators a more disaggregated analysis of treatment, outcomes, and expenditures beyond what is possible with claims data alone. Registry and claims data contain geographic information that allows for geographic assessments and incorporation of social determinants of health index, although we caution that such an index must be validated with the outcomes of interest as they were designed with different intentions. Registry data also capture important diagnosis information, including stage and tumor characteristics, that are necessary for clinical and economic assessments of equitable treatment and outcomes. APCD data substantially improve registry data because, as we demonstrate, insurance information is more accurately captured by the APCD as it is based on enrollment information submitted by payers. APCD data also capture out-of-pocket expenses across multiple payers, including for patients covered by Traditional Medicare who have supplemental plans, a distinct advantage over other linkages, such as a SEER-Medicare.43

There are limitations to accessing and using APCDs. In Colorado, acquiring the APCD for research can be more expensive than the cost of acquiring datasets like SEER-Medicare. There is no published uniform pricing. In some cases, it is possible to obtain a “scholarship” from the Colorado Department of Health Care Policy and Financing. Each request is reviewed by a Data Release and Review Committee that evaluates non-public CO APCD data release requests.”44 Without funding, however, it would be challenging to purchase the data or maintain the linkage. The data are restricted to the minimum amount needed to conduct the research and are governed by a DUA that restricts data elements provided and data reuse. Our DUA did not allow for the inclusion of detailed plan information (eg, high deductible plan) or the ability to link people covered by the same plan (ie, family members). These data are available from CIVHC; however, a strong case for their release is needed and requires the approval of the review committee.

Patients younger than age 65 with private plans that do not submit to the APCD or who were uninsured at the time of diagnosis were more likely to have missing claims data at the month of diagnosis, highlighting one of the challenges of gaps in coverage and the limitation of voluntary claims submission by ERISA-covered plans. In our sample, ~33% were missing coverage at the month of diagnosis. With APCD data, and especially in the non-Medicare eligible population, it is not possible to distinguish a person who loses insurance coverage from a person who transitions to a plan not captured in the APCD. Finally, as with other claims datasets, researchers must check for data inconsistencies, some of which may be due to retrofitting claims from multiple payers to fit the common data elements of the APCD.

Linked APCD and registry data are state-specific, and findings from research using these data may not generalize to other states. For example, Colorado is one of the few states that does not enroll Medicaid-insured patients in a managed care plan. Instead, Colorado uses Accountable Care Collaboratives that function like a fee-for-service plan with enhanced care coordination.45 Colorado has expanded Medicaid under the Affordable Care Act. In addition, Colorado’s racial and ethnic composition is unlike many states, with 22% and 4% of its population identifying as Hispanic and African American, respectively.46 Finally, Colorado has many rural and frontier counties where the population is sparse. Once linked to the CCCR, we often find that cell sizes for these areas are small, which may jeopardize patient privacy if reported, especially if cross-tabulated with other characteristics. Cancer registries and APCDs tend to be coordinated by state agencies and focused on data for patients in a single state. Therefore, researchers cannot track patients who cross state lines to receive treatment. Although existing software to conduct probabilistic linkages greatly facilitate record linkages, regulations governing APCDs and cancer registries vary by state, and the process for linking both data sources may differ regarding the safe transmission of personal identifiers. However, privacy-preserving record linkages have been developed to avoid the transmission of personal identifiers. These methods encrypt personal identifiers before performing probabilistic linkages.47

Linked APCD and cancer registry data offer an exciting opportunity for patient-centered, cancer health economics research. Assessments in care delivery and health equity can be made by racial and ethnic groups, and by geography. In addition, comparisons of out-of-pocket costs allow for more nuanced assessments of patients’ financial burdens. As more insurance plans shift medical costs to patients through high deductible plans,48 the financial burden increases and contributes to “financial toxicity.”49 Patients who experience financial burdens have more comorbid conditions and higher mortality.10 The detailed insurance data afforded by the APCD-CCCR linkage will allow for the decomposition of these out-of-pocket costs where our research team can identify patients who are most financially vulnerable, despite having health insurance coverage.

Cancer registry and APCD databases provide detailed patient information, and the two types of data complement each other in important ways. Efforts to implement APCDs over recent decades support the expansion of comprehensive patient-centered economic research, despite the complex and highly fragmented health care system in the United States. Coordinated efforts are needed to develop and sustain multistate or national data infrastructure efforts to fill key gaps in patient-centered research. A network of linked cancer registries and APCD data could serve as a prototype for patient-centered, health equity research for other diseases. Partnerships between researchers and state agencies can address an important research gap in evaluating patient-centered outcomes. As noted by Tsui et al,50 adapting models of shared data analyses, such as those used in distributed research networks,51 can facilitate such research and expedite our understanding of outcomes beyond a single state.

ACKNOWLEDGMENT

The authors thank Sara Kitchen for administrative support. Also, reviewers and attendees at this symposium for their helpful comments.

REFERENCES 1. McCarthy D. Part 1: How States Establish an APCD and Make It Functional. The Commonwealth Fund; 2020. Accessed February 28, 2023. https://www.commonwealthfund.org/sites/default/files/2020-12/McCarthy_State_APCDs_Part1_Report_v2.pdf 2. McCarthy D. Part 2: The Uses and Benefits of State APCDs. The Commonwealth Fund; 2020. Accessed February 28, 2023. https://www.commonwealthfund.org/sites/default/files/2020-12/McCarthy_State_APCDs_Part2_v2.pdf 3. The Office of the Assistant Secretary for Planning and Evaluation (ASPE) at the U.S. Department of Health and Human Services. 2022. Accessed February 28, 2023. https://aspe.hhs.gov/sites/default/files/documents/96f34fd0474b3da4884836c4341f1bbe/Linking-State-Health-Care-Data.pdf 4. Blewett LA, Mac Arthur NS, Campbell J. The future of state All-Payer Claims Databases. J Health Polit Policy Law. 2023;48:93–115. 5. Warren JL, Klabunde CN, Schrag D, et al. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002;40(suppl 8):Iv-3–18. 6. Bradley CJ, Given CW, Luo Z, et al. Medicaid, Medicare, and the Michigan Tumor Registry: a linkage strategy. Med Decis Making. 2007;27:352–363. 7. Nadpara PA, Madhavan SS. Linking Medicare, Medicaid, and cancer registry data to study the burden of cancers in West Virginia. Medicare Medicaid Res Rev. 2012;2:E1–E25. 8. Penberthy L, McClish D, Pugh A, et al. Using hospital discharge files to enhance cancer surveillance. Am J Epidemiol. 2003;158:27–34. 9. Bradley CJ, Entwistle J, Sabik LM, et al. Capitalizing on central registries for expanded cancer surveillance and research. Med Care. 2022;60:187–191. 10. Yabroff KR, Reeder-Hayes K, Zhao J, et al. Health insurance coverage disruptions and cancer care and outcomes: systematic review of published research. JNCI. 2020;112:671–687. 11. Escarce JJ, Carreón R, Veselovskiy G, et al. Collection of race and ethnicity data by health plans has grown substantially, but opportunities remain to expand efforts. Health Aff. 2011;30:1984–1991. 12. Ng JH, Ye F, Ward LM, et al. Data on race, ethnicity, and language largely incomplete for managed care plan members. Health Aff. 2017;36:548–552. 13. Nerenz DR, Carreon R, Veselovskiy G. Race, ethnicity, and language data collection by health plans: findings from 2010 AHIPF-RWJF survey. J Health Care Poor Underserved. 2013;24:1769–1783. 14. NAHDO. Current and Innovative Practices in Data Quality Assurance and Improvement. National Association of Health Data Organizations; 2019. Accessed February 27, 2023. https://www.apcdcouncil.org/publication/current-and-innovative-practices-data-quality-assurance-and-improvement 15. Weir HK, Johnson CJ, Mariotto AB, et al. Evaluation of North American Association of Central Cancer Registries’(NAACCR) data for use in population-based cancer survival studies. J Natl Cancer Inst Monogr. 2014;2014:198–209. 16. Noone AM, Lund JL, Mariotto A, et al. Comparison of SEER treatment data with Medicare claims. Med Care. 2016;54:e55–e64. 17. Bradley CJ, Liang R, Jasem J, et al. Cancer treatment data in central cancer registries: when are supplemental data needed? Cancer Inform. 2022;21:11769351221112457. 18. Trinidad S, Brokamp C, Mor Huertas A, et al. Use of area-based socioeconomic deprivation indices: a scoping review and qualitative analysis: study examines socioeconomic deprivation indices. Health Aff. 2022;41:1804–1811. 19. Schatzkin A, Midthune D, Subar A, et al. The national institutes of health-American association of retired persons (NIH-AARP) diet and health study: power to detect diet-cancer associations after adjusting for measurement error. Am J Epidemiol. 2001;153:S259–S259. 20. Fraser GE, Jacobsen BK, Knutsen SF, et al. Tomato consumption and intake of lycopene as predictors of the incidence of prostate cancer: the Adventist Health Study-2. Cancer Causes Control. 2020;31:341–351. 21. Buttorff C, Wang GS, Tung GJ, et al. APCDs can provide important insights for surveilling the opioid epidemic, with caveats. Med Care Res Rev. 2022;79:594–601. 22. Perraillon MC, Liang R, Sabik LM, et al. The role of all-payer claims databases to expand central cancer registries: experience from Colorado. Health Serv Res. 2022;57:703–711. 23. Hashibe M, Ou JY, Herget K, et al. Feasibility of capturing cancer treatment data in the Utah all-payer claims database. JCO. 2019;3:1–10. 24. CIVCH. Center for improving value in health care. Accessed November 7, 2022. https://www.civhc.org/ 27. Butler DC, Petterson S, Phillips RL, et al. Measures of social deprivation that predict health care access and need within a rational area of primary care service delivery. Health Serv Res. 2013;48:539–559. 28. Long JC, Delamater PL, Holmes GM. Which definition of rurality should i use?: the relative performance of 8 federal rural definitions in identifying rural-urban disparities. Med Care. 2021;59(suppl 5):S413–s419. 29. Morris AM, Rhoads KF, Stain SC, et al. Understanding racial disparities in cancer treatment and outcomes. J Am Coll Surg. 2010;211:105–113. 30. Du XL, Lin CC, Johnson NJ, et al. Effects of individual‐level socioeconomic factors on racial disparities in cancer treatment and survival: findings from the National Longitudinal Mortality Study, 1979‐2003. Cancer. 2011;117:3242–3251. 31. Zavala VA, Bracci PM, Carethers JM, et al. Cancer health disparities in racial/ethnic minorities in the United States. Br J Cancer. 2021;124:315–332. 32. NAACCR. Data standards & data dictionary, volume II. 2022;version 23. Accessed November 11, 2022. https://www.naaccr.org/data-standards-data-dictionary/ 33. Sherman RL, Williamson L, Andrews P, et al. Primary payer at DX: issues with collection and assessment of data quality. J Registry Manag. 2016;43:99–100. 34. USDA. Frontier and remote area codes. Accessed December 1, 2023. https://www.ers.usda.gov/data-products/frontier-and-remote-area-codes 35. UDS Mapper. Project background. Accessed February 21, 2023. https://udsmapper.org/about/ 36. Iragorri N, de Oliveira C, Fitzgerald N, et al. The out-of-pocket cost burden of cancer care—a systematic literature review. Curr Oncol. 2021;28:1216–1248. 37. Fu SJ, Rose L, Dawes AJ, et al. Out-of-pocket costs among patients with a new cancer diagnosis enrolled in high-deductible health plans vs traditional insurance. JAMA Network Open. 2021;4:e2134282. 38. HCPUF. Policy statement: billing health first colorado members for services. colorado department of health care policy and financing. Accessed February 10, 2023. https://hcpf.colorado.gov/policy-statement-billing-medicaid-members-services 40. Lee S, Ma C, Zhang S, et al. Marital status, living arrangement, and cancer recurrence and survival in patients with stage III colon cancer: findings from CALGB 89803 (Alliance). Oncologist. 2022;27:e494–e505. 41. Chen Z-H, Yang K-B, Zhang Y-z, et al. Assessment of modifiable factors for the association of marital status with cancer-specific survival. JAMA Network Open. 2021;4:e2111813–e2111813. 42. Freed M, Biniek J, Damico, A TN. Medicare Advantage in 2022: Enrollment Update and Key Trends. Kaiser Family Foundation; 2022. Accessed November 8, 2022. https://www.kff.org/medicare/issue-brief/medicare-advantage-in-2022-enrollment-update-and-key-trends/ 43. National Cancer Institute. Measures that are limited or not available in the data (SEER, SEER-Medicare). Accessed November 8, 2022. https://healthcaredelivery.cancer.gov/seermedicare/considerations/measures.html#6 44. CIVCH. Data Release Review Committee. Accessed November 7, 2022. https://www.civhc.org/data-release-review-committee/ 45. McConnell KJ, Renfro S, Chan BK, et al. Early performance in Medicaid accountable care organizations: a comparison of Oregon and Colorado. JAMA Intern Med. 2017;177:538–545. 46. U.S. Census Bureau. Quick Facts. Colorado: Census Bureau; 2022. Accessed November 8, 2022. https://www.census.gov/quickfacts/CO 47. Mirel LB, Resnick DM, Aram J, et al. A methodological assessment of privacy preserving record linkage using survey and administrative data. Stat J IAOS. 2022;38:413–421. 48. Kullgren JT, Cliff BQ, Krenz CD, et al. A survey of Americans with high-deductible health plans identifies opportunities to enhance consumer behaviors. Health Aff. 2019;38:416–424. 49. Zafar SY, Aber