Linked claims and medical records for cancer case management†
Evaluation of mammography abnormalities
Presented in part at the American Society of Clinical Oncology, Atlanta, Georgia, June 2–6, 2006.
Failure to obtain the requisite follow-up of patients with abnormal mammograms may delay cancer diagnosis and impact outcome. Up to 20% of women do not receive timely recommended follow-up. The current study tested the accuracy of the linkage of payer claims and clinical data to identify the appropriate treatment for patients with abnormal mammograms.
Electronic medical records in a staff model practice that was affiliated with a single health payer were scanned to identify the Breast Imaging Reporting and Data System (BI-RADS®) code for all mammograms. For each woman with a BI-RADS code 0, 3, 4, or 5 mammogram, the payer claims were searched for follow-up breast procedures (imaging, biopsy, and surgery) occurring within 2 months for BI-RADS code 0, 4, and 5 mammograms and within 8 months for BI-RADS code 3 mammograms. For women with >1 abnormal mammogram during the study period, only follow-up for the first abnormal mammogram was examined. The medical records of cases defined by claims as not having recommended follow-up care were reviewed to determine the accuracy of claims analyses.
A total of 17,329 women underwent 23,721 mammograms between January 1, 2001 and December 31, 2003. BI-RADS codes 0, 3, 4, or 5 occurred in 1490 mammograms (6.3%). Among 1206 women with a first abnormal mammogram who were eligible for claims follow-up, 16% did not receive recommended follow-up care. Medical record review demonstrated that the claims search accurately identified follow-up care in 97% of these cases.
Administrative claims supplemented with BI-RADS data were found to accurately identify the follow-up care of patients with abnormal mammograms. Case management using this method may assist physicians in ensuring that all patients receive appropriate care. Cancer 2007. © 2007 American Cancer Society.
The omission of potentially life-saving treatment is a serious quality problem in cancer care.1 Case management systems to monitor care may help busy clinicians ensure that patients receive recommended treatment.2–5 Such a structured approach is more easily implemented by integrated health systems that coordinate services for a defined population. The majority of care in America is administered in nonintegrated practice settings. There is a need for real-time care management programs that include patients regardless of the venue of care.
Case management systems are limited by the availability of community-wide clinical data. A rarely tapped source of patient-centered data is payer administrative claims. Claims identify all care for persons covered by a payer regardless of where the care was delivered. When matched with key clinical information to provide a clinical reference, claims data provide a complete picture of treatment. The majority of studies of claims data in cancer care have been retrospective reviews of care patterns using linkage to cancer registries.6, 7 Prospective monitoring of claims is used internally by payers, but to our knowledge there has been little research performed to date to examine its utility in assisting physicians ensure patients do not miss required care. Because claims are submitted to the payer within days to weeks, they may be useful in case management for events that occur within that time frame.
One practice area in which case management may help is the diagnostic evaluation of mammographic abnormalities. Failure to perform essential follow-up studies may delay the diagnosis of breast cancer and negate the benefits of screening.8–10 Several studies have shown that as many as 20% to 30% of women with breast lesions detected by mammography fail to obtain appropriate or timely follow-up imaging or biopsy.11 Although this may affect young, minority, and underserved women to a greater extent,12 the omission of follow-up has occurred in all populations studied, including patients in leading managed care organizations.
The results of all mammograms performed in the U.S. are coded with a number from 0 to 6 using the Breast Imaging Reporting and Data System (BI-RADS®) of the American College of Radiology.13, 14 The code denotes the overall finding (normal or suspicious) of the mammogram (Table 1). The code correlates with the likelihood that a finding is cancer. In 1 mammography registry series, this chance was estimated by the percentage of women that developed cancer within 1 year and was found to be 0.6% of women with a BI-RADS code 0 mammogram, 0.8% of women with a BI-RADS code 3 mammogram, 14% of women with a BI-RADS code 4 mammogram, and 46% of women with a BI-RADS code 5 mammogram.15 For each code, there is a recommended pathway for follow-up. Additional follow-up with immediate or 6-month follow-up imaging (BI-RADS codes 0 and 3, respectively) is required for approximately 6% to 8% and biopsy (BI-RADS codes 4 and 5) is required for 1% to 2% of screening mammograms.15 The imaging and biopsy studies required for the assessment of abnormal mammograms are readily identified by specific Current Procedural Terminology (CPT) codes used in claims.
Table 1. Breast Imaging Reporting and Data Standards (BI-RADS) System13
|0||Incomplete||Additional mammogram and/or ultrasound immediately||0.6%|
|3||Probably benign||Short-term follow-up; repeat imaging in 6 mo||0.8%|
|6||Known cancer||Used for staging and assessment of response†||NA|
The current study reports the accuracy and feasibility of the linkage of payer claims with BI-RADS codes from the mammogram as a strategy to identify women who miss the appropriate follow-up care of abnormal mammograms. It further identifies the frequency of missed follow-up for patients with abnormal mammograms in an insured population. This strategy may be useful as the core element in a program to ensure the appropriate follow-up of patients with abnormal mammograms.
MATERIALS AND METHODS
The current study was conducted by Roswell Park Cancer Institute (RPCI) in collaboration with Lifetime Health, Inc, a staff model medical practice, and its associated payer, Univera Healthcare, Inc. The study met all applicable privacy and auditing requirements and was approved by the Institutional Review Board of the RPCI. The study database was maintained in a secure information technology environment. Investigators were blinded to the identity of patients and providers through the use of a study link-back identifier in the study database.
The study cohort included women insured by Univera Healthcare who underwent screening or diagnostic mammography at a Lifetime Health facility between January 1, 2001 and December 31, 2003. There are approximately 22,000 patient encounters per month at Lifetime Health. Approximately 60% to 80% of the patients at Lifetime Health during the study period were covered by Univera Healthcare, a health insurance company that covers approximately 175,000 individuals within the 8-county area of the western New York region. Lifetime Health performs mammograms at 3 of its 6 office locations. Medical records at Lifetime Health are maintained in an electronic health record that includes radiology reports.
Lifetime mammogram reports provide the BI-RADS result using the term “BI-RAD” in the report. The BI-RADS code for each mammogram was identified by text searching the mammogram report in the electronic health record for key phrases including the term “BI-RAD.” Applying this method, all mammograms were successfully identified and coded. The accuracy of this method is supported by the distribution of BI-RADS codes reported in results, and by 100% concordance of BI-RADS classifications on a review of selected records.
For each woman who underwent a mammogram and who was covered by Univera, all medical claims data were copied to a study database for the period of 6 months prior to the date of the mammogram through July 7, 2004. Claims data elements for each patient included insurance coverage start and stop dates, year of birth, codes for primary and specialty providers, date of service, and the BI-RADS code from the index mammogram. Each claim line also included date of service, CPT codes, International Classification of Diseases codes (ICD-9), and place of service and provider codes.
For each patient with a BI-RADS code 0, 3, 4, or 5 mammogram during the study period, the dates of subsequent imaging and breast-related procedural events were identified from claims. For patients with >1 BI-RADS code 0, 3, 4, or 5 mammogram, only follow-up care from the first abnormal mammogram was reviewed. Any events that occurred on a single day were classified as a single breast imaging or breast biopsy event. Breast imaging was identified by CPT codes 76090 through 76092 and 76645. Breast imaging events were classified as mammography or ultrasound. Breast biopsy events included any needle or surgical breast procedure in the claims. Breast biopsy was identified by CPT codes 76095 through 76098, 19110 through 19103, and 10110 through 19160.
Care was defined as appropriate for BI-RADS code 0 mammograms if the time between the mammogram and the date of additional imaging was within 60 days. For a BI-RADS code 3 mammogram, care was considered appropriate if follow-up breast imaging was performed within 8 months of the abnormal mammogram, or if a breast biopsy was performed during this interval. Biopsy was considered appropriate for BI-RADS code 3 mammograms because biopsy could be prompted by clinical findings of a mass or discharge or serious patient anxiety that cannot be identified from claims. For BI-RADS code 4 and 5 mammograms, care was considered appropriate if biopsy occurred within 60 days of the abnormal mammogram.
For the purpose of this study, namely, determining the accuracy of claims for assessing follow-up, patients were censored if the length of time of available claims was shorter than that specified as appropriate care. This occurred when Univera Healthcare coverage for a patient terminated before the required follow-up time period, or in the case of patients who had a BI-RADS code 3 index mammogram in November or December of 2003 and did not have claims available as of July 7, 2004 (the study end). Because the objective of the current study was to determine the accuracy of claims in defining follow-up care, women who had mammograms at Lifetime Health and were not covered by Univera, or who were covered by Univera Healthcare and were patients at Lifetime but had their screening mammograms performed outside the Lifetime offices were not included in the analysis of follow-up.
For every patient who did not receive the guideline recommended care within the specified time, the Lifetime medical records (total of 149) were reviewed manually to determine whether any treatments occurred that were not captured in the claims. If the missed follow-up was documented in the record, any reason identified for this was recorded. Because the main objective of the study was to identify the accuracy of follow-up care from claims and not to ascertain reasons for missed follow-up, and due to privacy concerns and the number of years passed since the event, the study did not include direct patient contact to elucidate these reasons.
Between January 1, 2001 and December 31, 2003, there were 23,721 mammograms performed on 17,340 women at Lifetime. The BI-RADS code for each mammogram is shown in Table 2. Overall, 1490 mammograms (6.3%) were coded as BI-RADS 0, 3, 4, or 5. Among these, 1319 were the first mammogram among women with Univera coverage. Of these, 113 (8%) had truncated claims due to insurance termination before the required time frame for follow-up. This occurred among 13% of women with BI-RADS code 3 mammograms by 8 months, and in 4% of women with BI-RADS codes 4 and 5 mammograms by 2 months. Although in a case management program assessing the follow-up of these cases would be important to do by means other than claims, they were excluded from this test of the accuracy of claims for identifying follow-up testing.
Table 2. Frequency of BI-RADS® Codes in 23,721 Mammograms Among 17,340 Female Patients of Lifetime Health Inc Between January 1, 2001 and December 31, 2003
The study cohort included 1206 women. The follow-up imaging and biopsy information for each BI-RADS code is shown in Table 3. Overall, 189 of 1206 women (16%) did not receive the recommended care in the defined timeframe. Among 381 women with a BI-RADS code 0 mammogram, 75% underwent additional imaging, 13% underwent a breast biopsy, and 12% received no additional breast services within 2 months. Among 576 women with BI-RADS code 3 mammograms, 71% underwent additional imaging, 10% underwent a breast biopsy, and 19% received no additional breast services within 8 months. Of the 214 women with a BI-RADS code 4 and the 35 women with a BI-RADS code 5 mammogram, 22% and 14%, respectively, did not undergo a biopsy within 2 months.
Table 3. Follow-up Among 1206 Women With First BI-RADS Codes 0, 3, 4, and 5 Mammograms
|BI-RADS Code 0||2 mo||381||105||105||76||286 (75%)||49 (13%)||46 (12%)|
|BI-RADS Code 3||8 mo||576||295||22||94||411 (71%)||60 (10%)||105 (19%)|
|BI-RADS Code 4||2 mo||214||2||5||6||13 (6%)||168 (78%)||33 (15%)|
|BI-RADS Code 5||2 mo||35||0||0||0||0 (0%)||30 (86%)||5 (14%)|
|Total||—||1206||402||132||176||710 (59%)||307 (25%)||189 (16%)|
The accuracy of the claims data to identify follow-up care was determined by reviewing the medical records of women who did not have claims for the recommended follow-up procedures. Overall, for women with BI-RADS code 3, 4, and 5 mammograms, medical record review demonstrated that the claims accurately identified the absence of follow-up procedures in 144 of 149 cases (97%) (Table 4). Follow-up procedures were performed in the recommended time frame in 5 women but there was no claim for the procedure.
Table 4. Accuracy of Claims for Defining Follow-up as Determined by Medical Chart Review for BI-RADS Codes 3, 4, and 5 Mammograms and Circumstances of Missed Follow-up
|3||105||2||103 (98%)||26 (25%)||65 (62%)||–||8 (7%)||4 (4%)|
|4||39||2||37 (95%)||17 (44%)||8 (20%)||6 (15%)||5 (13%)||1 (2%)|
|5||5||1||4 (80%)||1 (20%)||1 (20%)||–||–||2 (40%)|
|Overall||149||5||144 (97%)||44 (30%)||74 (50%)||6 (4%)||13 (9%)||7 (5%)|
Chart review identified the clinical circumstances for nearly all the patients who did not receive the recommended follow-up (Table 4). The most common reasons were patient refusal or loss to follow-up in 44 cases, and follow-up occurring after the prescribed time frame in 74 cases. A second opinion led to recoding of the mammogram from BI-RADS code 4 to BI-RADS code 3 for 6 women who therefore did not go on to have the biopsy recommended for the BI-RADS code 4 mammogram (the claims accurately identified that biopsy was not performed). In 13 cases, although the BI-RADS code was correctly identified from the mammogram's overall conclusion in the electronic health record, the narrative text of the mammogram report reflected a normal mammogram requiring annual screening. This indicates a clerical error on the part of the individual dictating or entering the BI-RADS code into the mammogram conclusion. If the 13 miscoded mammograms are excluded from analysis of missed follow-up, the rate of missed follow-up decreases from 16% to 14%. In 7 cases, medical record review could not identify the reasons for missed follow-up.
Practice-based interventions that assist physicians, such as provision of care pathways and reminders, may help to reduce variations in care and the omission of required treatments. These interventions may be especially useful for events such as abnormal mammograms that occur infrequently in an individual physician's practice. Although case management interventions may be more readily implemented in large, coordinated healthcare organizations with integrated information systems, the majority of care is provided by physicians dispersed throughout the community who are not part of a health network. Moreover, specialist providers may unintentionally exclude the primary care provider from critical patient care information. A community-wide, practice-based quality improvement intervention has the potential to support physicians, improve communication between the physician and patient, and ensure that patients receive appropriate care.
To our knowledge, there are few data sources that capture community-wide outpatient treatment information. Payer claims are 1 potentially rich source of clinical information. Although claims are primarily intended for accounting purposes and may contain errors because of clerical or data entry mistakes or the omission of claims submission, they may be useful in quality management. Claims are used by managed care organizations for simple quality indicators such as the percentage of women who receive mammography under the Health Plan Employer Data and Information Set (HEDIS) program.16 Claims are also used by some payers as physician reminders for annual mammography, colonoscopy, and other primary care services. The use of claims in the analysis of cancer care is limited by the absence of key clinical information, especially the stage of the cancer. The linkage of claims to cancer registry data to provide stage data, most notably the linkage of Medicare claims to the National Cancer Institute's Surveillance Epidemiology, and End Results (SEER) registry, is useful for the retrospective analysis of patterns, disparities, and quality in cancer care.6, 7, 17–19 To our knowledge to date, there are few programs that have used claims for care management. Prospective linkage of claims to key clinical data readily available in electronic format may allow such case management programs.
Failure to obtain the appropriate follow-up testing for screening-detected breast abnormalities delays the diagnosis of breast cancer and may negate the effects of screening. A number of studies have demonstrated that as much as 30% of women with an abnormal mammogram (BI-RADS codes 0, 3, 4, or 5) do not receive the recommended follow-up care.11 This affects women in all social strata irrespective of insurance coverage, and is more pronounced among ethnic minorities and uninsured women.12 A system to monitor the management of patients with abnormal mammograms could greatly improve or even eliminate this problem.
The current study tested the accuracy of linking claims data to BI-RADS code data to identify the follow-up care of patients with abnormal mammograms. It demonstrated that claims linked with BI-RADS code data accurately identified the follow-up treatment for women with abnormal mammograms and that appropriate follow-up testing did not occur in 16% of cases. This was confirmed by medical record review. Therefore, this is an attractive target for community-wide evaluation and intervention, and an appealing method with which to identify missed care as part of a case management program.
This method should be applicable in a community-wide program linking claims for 1 or more payers with BI-RADS data from mammography providers. All mammography centers collect BI-RADS data in an electronic format to meet quality and auditing requirements for federal reimbursement under the Mammography Quality Standards Act (MQSA).20 This code could be submitted with the claim to the payer for such a case management program. The mammogram claim with attached BI-RADS coding data could be matched with subsequent claims to identify follow-up.
Although there are administrative and legal issues to address, payers are increasingly eager to leverage their data for the purpose of quality improvement. One concern is that payers may be unwilling to conduct such programs because of potential liability; however, case management systems are not intended to supplant the physician's responsibility to the patient but rather to support physicians in fulfilling this responsibility. Many payers and provider groups already engage in case management programs and accept the potential liability for the quality of care administered to their enrollees.21–24 The cost of the case management system is another concern. Although an important question, this study was not designed to define the cost or cost/benefit of such a system. Because this system leverages data already collected in electronic format, the cost should not be prohibitive and the savings in care for patients with earlier stage cancers may offset this cost.
Another legal concern is protection of patient privacy. Payer contracts with providers and patients allow for the collection of data for quality assessment. They do so to apply HEDIS measures and to review other practice parameters such as prescription drug usage patterns and the use of screening tests. This method does not differ materially from these data uses. The aggregation of data among payers poses additional privacy concerns and will require appropriate blinding of identity so that all identified information is channeled through the respective payer.
For the purpose of demonstrating the accuracy of the linkage of claims and BI-RADS data, we censored from analysis those women not continuously covered by the payer. For those with BI-RADS code 4 and 5 mammograms, for which immediate biopsy is indicated, this was 4% of the women. Women who lose or transfer care and insurance may be most at risk of missing follow-up care. A claims-based method also excludes women without insurance and women who have insurance coverage with a payer not participating in the case management system. Secondary systems to collect follow-up data regarding these vulnerable populations will be an important component of a case management program. This can be based on existing programs in mammogram centers that are required as part of MQSA auditing requirements. The claims-based system may assist mammogram centers for the majority of cases with claims coverage, and provide an enhanced system for monitoring the follow-up of women without insurance coverage.20
The time frames allowed for follow--up in the current study were chosen as a reasonable expectation and ones in which claims should be available for review without jeopardizing cancer outcomes. One group recommended using time to biopsy after a BI-RADS code 4 or 5 mammogram as a measure of the quality of care, with the recommended time of <2 weeks elapsing from mammogram to biopsy.25 This short period may be difficult to apply in a case management system because of the time required to submit and process claims. Using electronic transmission, most claims are submitted quickly to minimize accounts receivable, but expecting all claims to be processed in a 2-week time frame may be unreasonable. The time frames in the current study allow for appropriate workup to occur, account for scheduling issues, and allow time for the patient to seek second opinions, while not jeopardizing cancer outcomes.
A limitation of this method is that claims do not distinguish failure to obtain follow-up, in which a patient declines therapy or when a second opinion changes the recommendation. However, in the context of a case management program, these circumstances are readily identified when the provider is notified. Such events are of sufficiently low frequency as to not constitute an undue provider burden. Channeling this information through the provider also protects patient autonomy because it is the personal provider who may best ascertain patient preference and decisions. Another limitation of this method is that claims do not provide the findings of clinical examination, making it impossible to evaluate the overuse of biopsy procedures. Findings such as a palpable breast mass, suspicious nipple discharge, or serious patient anxiety are appropriate indications for biopsy, even in the setting of a BI-RADS code 0, 1, 2, or 3 mammogram. In the current study, breast biopsy was performed within the time specified for additional imaging in 12% and 10%, respectively, of women with BI-RADS code 0 and 3 mammograms.
This methodology may be applicable to other screening modalities and cancer therapies that require follow-up or treatment in the time frame of weeks to months. Examples in cancer care in which claims could be effectively linked to key clinical data for case management include the administration of stage-appropriate adjuvant systemic and radiation therapy, compliance with hormonal therapy (by monitoring pharmacy claims for prescription refills), and adherence with recommended physician follow-up visits and testing.
The current study demonstrated the accuracy and potential use of linking claims and selected, readily available clinical data. A prospective intervention to determine whether a case management system based on this linkage will reduce the frequency of missed follow-up care of patients with abnormal mammograms is in progress and expansion to multiple payers in a large geographic region is also in development. Such a case management system has the potential to ensure that all patients receive appropriate and timely care.