Integration of Cancer Screening Data into Routine Cancer Surveillance Systems: A Florida Pilot Project

Abstract

Introduction:

Integration of screening data into routine cancer surveillance systems can create more robust data systems to inform cancer prevention and control activities. Currently, state central cancer registries do not routinely collect breast and cervical cancer screening data as part of state cancer surveillance activities. Florida conducted a pilot study involving: (1) linkage of breast and cervical cancer screening data from the Florida Breast and Cervical Cancer Early Detection Program (FBCCEDP) from 2009 to 2021 to the Florida Cancer Data System (FCDS) database to capture screening data for matched cancer cases in the FCDS; and (2) evaluation of the feasibility of developing a population-based breast and cervical cancer screening surveillance system by capturing electronic screening data from private health care providers.

Methods:

In 2018, the FCDS worked with the Florida Department of Health to identify data partners for the 5-year cancer screening pilot project funded by the Centers for Disease Control and Prevention. Engagement of project partners required extensive review of available screening data; data standards and formatting; data transmission schedules and methods; and processing procedures. The FCDS developed a database to integrate multiple source data sets into a single database whereby linkage to the central cancer registry could be performed.

Results:

The FCDS worked with Suncoast Health Systems, a clinical practice in the Hillsborough region of Florida, and the FBCCEDP to evaluate data availability, standardization of data sets, and data submission schedules for the pilot project. Extensive meetings and data reviews were conducted with both partners in the first phase of the project. The FCDS developed automated data processing procedures to integrate the data into a single cancer screening database and then linked records to the central cancer registry data set.

Discussion:

Registry collaboration with the FBCCEDP and Suncoast team on data quality and standardization has produced positive results. The project required extensive review of data and produced many lessons learned for development of a cancer screening surveillance system. Our pilot project depended on partnership building, commitment to data quality, and consistency in data submission practices.

Introduction

The Florida Cancer Data System (FCDS) is a central cancer registry supported by the National Program of Cancer Registries (NPCR) administered by the Centers for Disease Control and Prevention (CDC). The FCDS was legislatively established by Florida Statute in 1978 and began collecting cancer reports statewide in 1981, becoming a member of the NPCR in 1994.^1,2,3,4 Currently, the FCDS contains approximately 4.6 million unique cancers representing reporting from hospitals, ambulatory surgical centers, radiation treatment facilities, private physician practices, and pathology offices. Linkages to the Florida Bureau of Vital Statistics and hospital discharge records help to improve casefinding, data completeness, and quality. The resulting data set provides hospitals, researchers, public health professionals, and policy makers access to a trove of cancer incidence data. However, lacking from the robust FCDS database are applicable screening data for individuals who have been diagnosed with screen-detectable cancers.

Together, breast and cervical cancer constitute a leading cause of morbidity and death among women in the United States.⁵ In 2019, breast cancer was the most common cancer reported and the second most common cause of cancer death among women in Florida, contributing 19,062 (126.54 per 100,000) new cancer cases and more than 3,084 deaths (19.05 per 100,000).^6,7 Despite documented declines, Florida's cervical cancer and mortality rates remain among the highest in the United States. In 2019, there were 1,082 (9.17 per 100,000) new cases of cervical cancer and 354 (2.6 per 100,000) attributable deaths.^5-7

Early diagnosis of breast cancer through mammog-raphy, and cervical cancer through Papanicolaou (Pap) or human papillomavirus (HPV) testing, can prevent disease progression to advanced or invasive stages. Screening is also associated with improved treatment outcomes, higher survival rates, and reduced mortality rates.⁸ The United States Preventive Services Task Force (USPSTF) recommends mammography screening for women aged 50 to 74 years every 2 years.⁹ The USPSTF also recommends screening for cervical cancer in women aged 21 to 65 years with a Pap test every 3 years or, for women aged 30 to 65 years, with a combination of Pap and HPV tests, or HPV test alone, every 5 years. Provisions in the Affordable Care Act ensure that nongrandfathered private insurance plans, as well as Medicare and Medicaid expansion states, must cover certain preventive services, including breast and cervical cancer screening, with no cost sharing.¹⁰ For women who are low income, uninsured, or underinsured, the Florida Breast and Cervical Cancer Early Detection Program (FBCCEDP) provides breast and cervical cancer screening, follow-up diagnostic services for abnormal results, and treatment-eligibility referral. In 2020, it was estimated that 79.2% of Florida women aged 50 to 74 years received a mammogram in the past 2 years, while 77.1% of those 21 to 65 years had received a Pap test in the previous 3 years.¹¹ These rates are comparable to 2020 national breast (mammography) and cervical cancer screening (Pap test) rates (78.3% and 77.9%, respectively).^12,13

Recognizing that surveillance data systems constitute the basis for the planning, implementation, and evaluation of public health programs, integration of health services data such as screening data into routine cancer surveillance systems can complement incidence, mortality, and survival data.¹⁴ Currently, data collected from women during breast and cervical screenings are not part of the FCDS. Linkage of screening data, for both positive and negative screening results, with the central cancer registry can allow for timely collection of additional information, including who is getting screened, age at screening, number of screenings, screening results, and diagnostic follow-up. Specifically, racial and ethnic information from the screening procedures can help supplement incomplete race data from the registry cancer records. These more comprehensive registry data can be used to inform cancer prevention and control activities such as highlighting common geographical or sociodemo-graphic characteristics of patients who were not screened prior to diagnosis, informing future state cancer screening efforts.

To determine if this comprehensive registry is feasible, we evaluated reporting mechanisms for data availability, data quality, and efficiency in capturing screening data as part of a 5-year award from the CDC. To do this, Florida conducted a pilot study involving: (1) linkage of all breast and cervical cancer screenings from the FBCCEDP for the time period 2009 to 2021 to the FCDS to capture medically-confirmed screening data and self-reported medical history for current cancer patients within the FCDS database; and (2) evaluation of the feasibility of developing a population-based breast and cervical cancer screening surveillance system in Florida by electronically capturing screening data from private health care providers through their electronic health record (EHR) systems.

Methods

Overview of Approach

Florida has adopted a multiphase approach to evaluate and identify an effective mechanism for reporting breast and cervical cancer screening data by (1) Collecting statewide FBCCEP data and regional data from Suncoast Community Health Centers (Suncoast), and (2) linking the consolidated screening database to the FCDS.

Florida Breast and Cervical Cancer Early Detection Program

The FBCCEDP provides breast and cervical cancer screening to women aged 50 to 64 years who are at or below 200% of the federal poverty level and who are underinsured or uninsured. According to the US Census Bureau's Small Area Health Insurance Estimates, approximately 647,000 Florida women between the ages of 50 and 64 years were at or below 200% of the federal poverty level in 2017, with about 25.1% being uninsured.¹⁵ Additionally, women below age 50 years who are symptomatic or have a family history of breast cancer and who meet other eligibility requirements for income and insurance may be screened by the FBCCEDP.^16-18 The FBCCEDP provided over 28,000 screening and diagnostic services to eligible women in fiscal year 2017.¹⁹ The FBCCEDP maintains 16 regional sites serving eligible women from all 67 counties in Florida.

Suncoast Community Health Centers

With support from the FBCCEDP Hillsborough County coordinator, the FCDS established a partnership with Suncoast Community Health Center, a mammography and gynecology center that is a not-for-profit, federally qualified community health center. Accredited by the Accreditation Association for Ambulatory Health Care, Suncoast has served Hillsborough County since 1977. The center provides primary health services to poor, uninsured rural persons and migrant, seasonal agricultural workers in Eastern and Southern Hillsborough County and Lakeland, Polk County. With a network of 10 practices, 3 mobile units, 44 medical providers, and 32 dental providers, Suncoast provided services to 63,250 patients in 2018. The patient population of Suncoast derives from an ethnically diverse population. Further, as a federally qualified community health center, the patient population has disproportionately higher rates of poverty. Approximately 86% of these patients had income at or below 200% of the federal poverty level, with 35% uninsured, 72% from racial and ethnic minority groups, and 12% being agricultural workers.²⁰ Hillsborough County was selected as the initial pilot site given its large size and racial-ethnic and socioeconomic diversity. Hillsborough County is 28% Hispanic, 15.5% non-Hispanic Black, 50% non-Hispanic White, and 4% Asian. Approximately 17% of people in the county are foreign-born, with a majority from Latin America and Asia (67% and 20%, respectively).^15,21 Hillsborough county constitutes a major urban center and captures a significant population in need of cancer screening services.

Data Submission

Florida Breast and Cervical Cancer Early Detection Program. The FBCCEDP collects the personal health history as well as demographic, screening, testing, diagnosis, and treatment data for women who receive breast and cervical cancer screening services. After a series of discussions in biweekly meetings and review of the FBCCEDP patient reporting form, the FBCCEDP agreed to submit many of the screening and diagnostic follow-up data items captured on the patient reporting form along with key demographic data required to facilitate linkage to the registry. As part of the pilot project, the FBCCEDP submitted screening data to the FCDS for services and procedures beginning from the year 2009, which was the earliest year that the program implemented an electronic capture data system, through the year 2021.

Following data review, including quality checks, standardization of data values, and deduplication, the data set was consolidated to unique patient-level records. The range of procedures included both screening and diagnostic procedures as well as consultations. The FCDS worked with the FBCCEDP program to identify the procedure types that are specific to screening procedures versus diagnostic procedures to further define the consolidated screening database. These included mammograms, clinical breast examinations (CBEs), Pap tests, and HPV tests. While CDC did not require or recommend the collection of CBEs as a cancer-screening variable for this feasibility study, we included it as a state-specific item to expand the potential for linkage to the central cancer registry.

Suncoast. It was determined that the best way to extract breast and cervical cancer screening data was through Suncoast's EHR system. Screening data cover Pap tests, HPV tests, CBEs, and mammograms. Quarterly submissions from Suncoast began in January 2019 and continued through the first half of 2022. Each submission reflected the most recent screening dates by patient. For example, a Pap test performed on January 1, 2019, will remain on each report until the next Pap is performed, when it will be replaced with the latest date on the report for each type of screening.

As FCDS received submissions, the data were parsed by procedure type (CBE, HPV, Pap, mammogram) and results from the screening. The FCDS data import module allowed screening results to be updated as needed if new information was reported.

To date, Suncoast has submitted screening data for women who received screening services between January 2019 and January 2022. Some women who received a single screening during this period also had historical screening data for other screenings included in the data set that went as far back as 2009, but most procedures occurred between 2019 and 2022, as those were the years for which the data were extracted.

Consolidated Screening Surveillance Database. The FCDS team began the design of the consolidated database in November 2020. The first step required developing the related procedures to load and update data from Suncoast and FBCCEDP into respective patient and screening tables (Figure 1). These source-specific tables have their own update logic given structural differences between reporting sources. The second step required consolidation logic to incorporate both sources into a consolidated patient and screening table, with text-to-code conversions where applicable for fields such as race, ethnicity, smoking status, and sex. Where possible, we applied standardized values from the North American Association of Central Cancer Registries (NAACCR).

Figure 1.

Consolidated Patient and Screening Database Flowchart

CBE, clinical breast examinations; FBCCEDP, Florida Breast and Cervical Cancer Early Detection Program; FCDS, Florida Cancer Data System; HPV, human papillomavirus; Pap, Papanicolaou.

Each record contains a field to indicate the reporting source. The Florida project team selected 4 procedure types to include in the final consolidated screening database: mammogram, CBEs, Pap test, and HPV test. They included an insert date and last update field to capture the status of each record. The consolidated patient table contains the matched patient identifier and matched date field where records were linked with a patient in the cancer registry tumor database. Tables 1 and 2 show the consolidated fields for the screening and patient tables.

Table 1.

Consolidated Patient Table Variables

C2P__PATIENT_ID

C2P__SOURCE

C2P__NAME_LAST

C2P__NAME_MIDDLE

C2P__NAME_FIRST

C2P__DOB

C2P_SSN

C2P__SEX

C2P__ADDRESS1

C2P__ADDRESS2

C2P_CITY

C2P_ZIP

C2P_STATE

C2P_BMI

C2P__HEIGHT

C2P_WEIGHT

C2P__HEIGHT_TEXT

C2P_WEIGHT_TEXT

C2P__VITAL_STS_DATE

C2P__SMOKING_STATUS

C2P__TOBACCO_STS_DATE

C2P_.INSURANCE

C2P_RACE

C2P_ETHNICITY

C2P__INSERT_DATE

C2P__LAST_UPDATED

C2P_COUNTY

C2P__MATCHED_PATIENT_ID

C2PMATCHED DATE

Table 2.

Consolidated Screening Table Variables

C2S__PATIENT_ID

C2S_SOURCE

C2S__SCREENING_TYPE

C2S__SCREENING_DATE

C2S__SCREENING_RESULT

C2S_JNSERT_DATE

C2SLAST UPDATE

While the FBCCEDP data included both screening and diagnostic procedures, we only included those that fell within the 4 screening criteria: mammogram, clinical breast exams, Pap test, and HPV test. Procedures from Suncoast fell exclusively into 1 of these 4 screening categories.

Cancer registry linkage. A final consolidated FCDS tumor data set was extracted in year 5 of the pilot and included unique female-only patient-level records regardless of diagnosis year; we used the entire FCDS data because some women may have been captured in the registry for any cancer diagnosed prior to or after utilizing the screening services.

The FCDS implemented a probabilistic linkage of the 2 data sets using the R package fastLink with a matching threshold of at least 0.98. The linkage variables consisted of social security number, zip code, date of birth, street number, and phonetic encodings of first and last names. The data set of linked cancer patients resulting from the linkage was subsequently used to extract tumor records for each patient. We analyzed the linked data set to generate frequencies of tumor types by site.

The unique patients from the consolidated patient screening database were linked to the consolidated FCDS patient-level database. The tumor records were further limited to diagnosis years 1981-2019, representing complete incidence cases in the registry. No manual review was performed on the linkage. Furthermore, the fastLink program includes a deduplication procedure.

Results

FBCCEDP

Upon consolidation of the FBCCEDP source data to include only those screenings that fell within the 4 main screening categories, the total number of unique patients totaled 114,713 patients and 331,414 screenings (49% mammogram, 31% CBE, 16% Pap test, and 4% HPV test). Approximately 92% of all patients were between 50 and 64 years at the time of the procedure, 39% were Non-Hispanic White, 22% were Non-Hispanic Black, and 34% were Hispanic.

Suncoast

The Suncoast data set represented a total of 19,308 unique female patients and 53,553 screening procedures. Pap tests comprised most of cancer screening procedures (40%), followed by HPV tests (29%), mammograms (18%), and CBE (13%). Of the 19,308 unique patients, 63% were Hispanic, 26% Non-Hispanic White, and 9% Non-Hispanic Black. Many patients were aged 24-49 years (65%), and 35% of individuals represented the 50 years or older age group. The data extraction process from Suncoast pulls records based on the current years’ screenings, but also includes the date of last screening across screening modalities. Therefore, screenings have been captured that predate the year of data extract.

Consolidated Screening Database

The design and functioning of the consolidated screening database have been effectively automated and handles the unique formats of each data source to produce a standardized analytic data set. There are 2 main analytic tables; patient-level information is consolidated by source into a single patient table, and screening procedures are also consolidated by source (Suncoast and FBCCEDP). The final patient and screening tables combine sources into a single analytic set. Tables 3 and 4 describe the total number of consolidated and unique patient records and screenings combined by source and demographic characteristics.

Table 3.

Consolidated and Unique Patient Records, Suncoast (2019-2022) and FBCCEDP (2009-2021)

	Combined		Suncoast		FBCCEDP
	n	%	n	%	n	%
Total patient records	134,021	100	19,308	100	114,713	100
Race/ethnicity
Hispanic	52,037	39	12,161	63	39,876	35
Non-Hispanic White	49,835	37	4,962	26	44,873	39
Non-Hispanic Black	27,577	21	1,780	9	25,797	22
Non-Hispanic AI/AN/Other*	2,784	2	13	<1	2,771	2
Unknown	1,788	1	392	2	1,396	1
Smoking Status
Never smoker	98,724	74	14,758	76	83,966	73
Current	25,601	19	2,118	11	23,483	20
Former smoker	2,432	2	2,432	13	0	0
Unknown	7,264	5	0	0	7,264	6

AI/AN, American Indian/Alaska Native; FBCCEDP, Florida Breast and Cervical Cancer Early Detection Program.

Table 4.

Consolidated Breast and Cervical Cancer Screening Procedures, Suncoast (2019-2022) and FBCCEDP (2009-2021)

	Combined		Suncoast		FBCCEDP
	n	%	n	%	n	%
Total screening records	384,967	100	53,553	100	331,414	100
Procedure
Mammogram	173,888	45	9,493	18	164,395	50
Clinical breast examination	108,597	28	7,293	13	101,304	30
Pap test	72,822	19	21,365	40	51,457	16
HPV	29,660	8	15402	29	14,258	4
Age at time of screening (y)
<20	53	<1	0	0	53	<1
20-29	7,230	2	6,162	12	1,068	<1
30-39	20,347	5	16,152	30	4,195	1
40-49	25,923	7	14,504	27	11,419	3
50-59	230,524	59	11,592	22	218,932	66
60-69	99,723	26	4,863	9	94,860	29
>70	1,167	<1	280	<1	887	<1

FBCCEDP, Florida Breast and Cervical Cancer Early Detection Program; Pap, Papanicolaou.

Cancer Registry Linkage Results

The results of the tumor linkage for Florida malignant breast and cervical cancers resulted in a total of 5,602 linked patients and 6,006 patient tumors (there can be multiple tumors per patient) diagnosed between 1981 and 2019. The linkage included a de-duplication component, did not use blocking, and did not include manual review. Given that there were known duplicate patients in the consolidated screening patient table, as described previously, the de-duplication was helpful in eliminating many of these instances. The final linked tumor data set excluded out of state diagnosis, and removed tumors diagnosed prior to 1981 and after 2019. The cancer registry was established in 1981 and the most recently published data were for the 2019 diagnosis year. This ensures higher data quality and validation. Therefore, we included a total of 6,006 breast and cervical tumors diagnosed among 5,602 patients in the final analytic data set. Most linked tumors consisted of cancers of the breast (93%) with fewer cervical cancers (7%) (Table 5). It is worth noting that the source of screening data skew toward individuals of lower socio-economic status. Therefore, the distribution of covariates is not representative of all diagnosed breast and cervical cancers in the FCDS.

Table 5.

Linked Tumors by Site with Consolidated Unique Patient Table (n = 6,006), FCDS, 1981-2019

Site	Combined		Suncoast		FBCCEDP
	n	%	n	%	n	%
Breast	5,595	93	113	84	5,482	93
Cervix uteri	411	7	21	16	390	7
Total	6,006	100	134	100	5,872	100

FBCCEDP, Florida Breast and Cervical Cancer Early Detection Program; FCDS, Florida Cancer Data System.

Limitations

There were multiple challenges throughout the pilot screening project that are worth noting. First, establishing data sharing partnerships requires a considerable time commitment and investment in effort to meet with health agencies, private practices, clinicians, and public health officers. While connections to internal and partnered agencies within the Department of Health were more easily facilitated, relationships with private practices and clinicians required more effort given less familiarity with the central cancer registry's legislative mandates. For example, a prospective clinical practice declined to work with the pilot project citing concerns around the releasing of patient identifiable information (PII) and potentially violating patient confidentiality. Although the current state cancer registry legislation and a data use agreement with the health care system allowed the project team to obtain cancer screening data for this feasibility project, there were concerns about whether the legislation covered cancer screenings not linked to a diagnosis or treatment of cancer. Developing a statewide cancer screening surveillance system may require policy revisions, including more specific legislative language that requires reporting of all cancer screening data, with or without a cancer diagnosis and prospectively, to circumvent any barriers to acquiring data in the future.

Secondly, there are no data standards for capturing screening data from the EHR. Each provider and program use different file layouts. The FBCCEDP data set contains multiple patient records if more than 1 procedure was performed, while the Suncoast data set structure contains patient data on a single record with multiple procedures listed in different fields. A statewide roll-out would require a data dictionary for consistent reporting and assimilation to the standard format by each provider. Providers use different EHR software resulting in the various file layouts and formatted data. While some data elements are coded into distinct categories with defined values, others are entered in as free text or without formatting standards. The business flow and logic of collecting and maintaining the data sets in the EHR are often combined with ad hoc data cleaning and recoding, resulting in less standardized data sets. A statewide roll-out would require a data dictionary defining consistent coding standards by each provider.

Third, EHR system upgrades have contributed to delays in data extraction and submission. Going forward, software vendor changes could also represent a potential barrier to receiving timely data from other providers.

Lastly, the COVID-19 pandemic shifted organizational priorities across the state, which directly and indirectly impacted the pilot project in terms of the availability of providers to dedicate efforts toward data review and submission. While it did not stall progress on the project completely, it presented additional challenges that affected project implementation and efficiencies.

Discussion

The 5-year cancer screening surveillance feasibility pilot project presented Florida with a valuable learning experience and provided an opportunity to create a road map for inclusion of cancer screening surveillance data within the central cancer registry. The linked screening and tumor database can be analyzed to address specific screening to diagnosis research questions that may include identification of delayed screenings, time from screening to diagnosis, screening disparities and associated advanced stage tumors, among other inquiries that inform evidence-based policy and program decision making. The addition of other cancer screening modalities such as low dose computed tomography (CT) scans for lung cancer, and colonoscopies for colorectal cancer can be considered in the expansion of this project.

Prior studies that linked breast cancer screening with cancer registry data were conducted over 20 years ago and focused on acquiring data from a screening and provider or health insurance claims database.^22,23 Although these studies demonstrated the feasibility of acquiring breast cancer screening data, the quality and completeness of variables available for linkage with registry varied. Additionally, prior studies concluded that performing cancer screening and registry data linkage was resource and time intensive and posed data privacy and security issues. A recent study by Heins and colleagues linked national cancer registry data with a small percentage (10%) of patient-level primary care EHRs without reliance on a unique identifier; this limited data quality and completeness and may have resulted in false data linkages.²⁴

Integral to the overall completion and success of the feasibility pilot was ensuring that annually proposed planned activities were evaluated, identifying facilitators in accomplishing planned activities and discovering challenges and barriers, which delayed completion of proposed activities and consequently the development of new activities. Important factors for success involve facilitators at the state and county levels to connect the cancer registry with potential data partners. Without these support networks, identifying appropriate data sources and partners would have been extremely difficult. Based on Florida's experience implementing this project, expanding this project to additional registries would require each of the barriers to be fully addressed. A formalized and standardized process for data capture, formatting, and submission, as well as the development of a statewide data dictionary, is vital to planning and implementing a cancer screening surveillance system within an existing central cancer registry infrastructure.