Improving Longitudinal Outcomes, Efficiency and Equity in the Care of Patients with Congenital Heart Disease

Abstract

Background:

Longitudinal follow-up, resource utilization, and health disparities are top congenital heart research and care priorities. Medicaid claims include longitudinal data on in-patient, out-patient, emergency, pharmacy, rehabilitation, home health utilization, and social determinants of health—including mother-infant pairs.

Objectives:

The New York Congenital Heart Surgeons Collaborative for Longitudinal Outcomes and Utilization of Resources (CHS-COLOUR) linked robust clinical details from locally held state and national registries from 10 of 11 NY congenital heart centers to Medicaid claims, building a novel, state-wide mechanism for longitudinal assessment of outcomes, expenditures, and health inequities.

Methods:

We included all children <18 undergoing cardiac surgery in the Society of Thoracic Surgeons Congenital Heart Surgery Database and/or the NY State Pediatric Congenital Cardiac Surgery Registry from 10 of 11 NY centers, 2006–2019. Data were linked via iterative, ranked deterministic matching on direct identifiers. Match rates were calculated and compared. Proportions of the linked cohort trackable over 3-, 5-, and 10-years were described.

Results:

Of 14,097 registry cases, 59% (n=8,322) reported Medicaid use. Of these, 7,414 were linked to NY claims, at an 89% match rate. Of matched cases, we tracked 79%, 74%, and 65% of children over 3-, 5-, and 10-years when requiring near continuous Medicaid enrollment. Allowing more lenient enrollment criteria, we tracked 86%, 82%, and 76%. Mortality over this time was 7.7%, 8.4%, and 10.0%. Manual validation revealed ~100% true matches.

Conclusions:

This establishes a novel state-wide data resource for assessment of longitudinal outcome, health expenditure, and disparities for children with congenital heart disease.

Condensed Abstract:

We linked state and national clinical registry data to Medicaid claims, 2006–2019, via iterative, ranked deterministic matching. Of 14,097 registry cases, 59% (n=8,322) used Medicaid. We linked 89% of Medicaid cases (n=7,414). Of matched cases, we tracked 79%, 74%, and 65% of children over 3-, 5-, and 10-years, requiring near continuous Medicaid enrollment. Allowing more relaxed enrollment, we tracked 86%, 82%, and 76%. Mortality over this time was 7.7%, 8.4%, and 10.0%. Manual validation revealed ~100% true matches. This establishes a novel state-wide data source for assessment of longitudinal outcome, health expenditure, and disparities for patients with congenital heart disease.

Introduction

Congenital heart defects (CHDs) are the most common and resource intensive birth defects managed in the United States (US), affecting ~40,000 births per year in the US (1). Yet there are currently limited data on long-term outcomes and health expenditures for these children. The Centers for Disease Control and Prevention, US Congress, American College of Cardiology, and American Heart Association have all defined longitudinal follow-up, health expenditures, and health equity for patients with CHD as top research and patient care priorities (2–4). Due to marked heterogeneity in disease subtypes and treatments among CHD patients, the power of single-center studies is limited. Multi-center data are siloed in diagnostic or procedural registries or in-patient databases or are the product of individual investigations (5). Administrative data may lack clinical precision, as billing codes for this population are not physiology-based. Further, data on costs and value typically rely on cost-to-charge ratio-based costs, which are highly influenced by hospital accounting (6).

Nationwide, approximately 50% of children with chronic diseases—including ~50% of children who undergo congenital heart surgery—are insured by Medicaid (7). Medicaid claims include longitudinal data on all billed services related to in-patient, out-patient, emergency room, pharmacy, rehabilitation, and home health service utilization and reimbursement, including those related to mother-infant pairs. The New York Congenital Heart Surgeons Collaborative for Longitudinal Outcomes and Utilization of Resources (CHS-COLOUR) was formed to bring together the leadership and locally held clinical registry data from 10 of the 11 congenital heart surgical centers operating in New York State since 2006, health services researchers, health economists, and the New York Department of Health. Linking New York Medicaid claims data from the State Department of Health via direct patient identifiers to granular clinical details from CHS-COLOUR clinical registry data, we built a novel population-based mechanism for longitudinal assessment of risk-adjusted outcomes, resource utilization, health expenditures, and health disparities across the lifespan.

Methods

Data Sources

NY Congenital Heart Surgeons Collaborative for Longitudinal Outcomes and Utilization of Resources (CHS-COLOUR):

CHS-COLOUR brings together the leadership and locally held clinical registry data from the Society of Thoracic Surgeons Congenital Heart Surgery Database (STS CHSD) and the New York State Pediatric Congenital Cardiac Surgery Registry (NYS-PCCS) from 10 of the 11 congenital heart surgical centers with Certificates of Need to operate in NY State since 2006. STS CHSD is the largest congenital heart surgery clinical registry in the world. The STS CHSD registry is a voluntary registry that captures >90% of all US congenital heart operations, with high-level clinical granularity, including detailed information on patient demographics, hospital, surgeon, 110 anatomically specific cardiac diagnoses, 187 procedures / procedural subtypes, 22 comorbid conditions / preoperative risk-factors (including a history of prior cardiac surgeries), elective / urgent / emergent case status, bypass / cross-clamp / circulatory arrest times, 11 major postoperative morbidities, length-of-stay, and 30-day and operative mortality (8). The NYS-PCCS registry is a mandatory registry managed by the NY State Department of Health. NYS-PCCS has captured 100% of NYS congenital heart operations in patients <18 years of age since 1991. CHS-COLOUR members submit in-patient data to the STS and to NY State on all covered procedures, but maintain ownership of local data. CHS-COLOUR sites share with the Collaborative locally-held data from each center’s in-house STS CHSD and NYS-PCCS registry software, including direct patient identifiers on a quarterly, on-going basis. Dates of initiation of STS CHSD participation varied by site, whereas all sites submitted to NYS-PCCS from 1991 onward. Including data from both clinical registries serves two purposes for the Collaborative: 1) Including the NYS-PCCS allows the collaborative to capture data from all cases in NYS since 1991, which can be used in future analyses; 2) including the STS CHSD expands the generalizability of the approach, as the STS CHSD would be available nationwide for future collaboratives. One NY center declined to participate.

NY Medicaid Claims/Social Determinants of Health:

NY Medicaid Claims include longitudinal data on all billed services associated with in-patient, out-patient, emergency room, pharmacy, rehabilitation, home healthcare, and educational services, including direct patient identifiers, mother-infant pairs, healthcare reimbursement, and out of pocket expenses, for both fee-for-service and managed Medicaid patients, whether these services occurred in or outside of the State. In partnership with the NY Department of Health, we have built a robust NY Medicaid data system, including 14-years of cleaned and tested data, inclusive of >6,000 individual data fields and direct patient identifiers, and linked to Census Bureau data, data from over a dozen state and local social programs, and New York Vital Statistics and Medicaid death index data for patients who die while enrolled (9) (Central Illustration).

Central Illustration. CHS-COLOUR Data Repository:

The New York Congenital Heart Surgeons Collaborative for Longitudinal Outcomes and Utilization of Resources (CHS-COLOUR) brings together New York Medicaid claims, locally held state and national clinical registry data, and Census Bureau and social programming data to establish a novel state-wide data resource for assessment of longitudinal outcome, health expenditure, and disparities for children with congenital heart disease. A. CHS-COLOUR Data Sources. B. CHS-COLOUR Overall and Medicaid Match Rates for 14-years of CHS-COLOUR data. Overall match rate describes the percentage of total registry cases able to be linked to Medicaid claims for the procedure of record. Medicaid match rate describes the percentage of Medicaid cases able to be linked to claims. Additional matches describe the percentage of additional cases who could not be matched to Medicaid claims for that specific procedure, but for whom data were found in Medicaid (either before or after the index procedure).

Study Population

We included all children <18 years of age who underwent heart surgery from 10 of the 11 congenital heart surgical centers with Certificates of Need to operate in NY State, 2006 – 2019. STS CHSD and NYS-PCCS registry data were used for case finding. We excluded children <2.5 kilograms who underwent isolated patent ductus arteriosus closure, in keeping with analytic standards for congenital heart research and public reporting. We also removed any non-NYS residents, as residents of other states are not eligible for NYS Medicaid.

Statistical Analyses

Data Linkage:

We tested a multi-pass iterative deterministic matching algorithm on direct patient identifiers between data collected in each center’s in-house STS CHSD or NYS-PCCS registry software and Medicaid claims files (10,11). Available patient identifiers included: child’s first and last names, mother and father’s last names, alternate or prior names (e.g., in cases of name change), dates of birth, hospital names, and dates of admission, surgery, and discharge. The Medicaid claims universe was restricted to patients who had an in-patient admission before their 18^th birthday at a NY State hospital, with cardiac surgery-related International Classification of Disease-9 (ICD-9) or ICD-10 procedure codes to minimize false positives. The algorithm then proceeded through an iterative series of matches: (1) through a pass that required exact matches on the identifiers, (2) through several passes that permitted fuzzy matches (e.g., minor name and date typographical errors), and (3) through several passes expanding the universe of potential claims matches to those that did not have a qualifying cardiac surgery procedure on the claim but were linked via exact match on all dates and identifiers. In full, the algorithm included seven passes, summarized in Table 1. Potential matches were ranked bypass, with the highest-ranking match retained. Matches whose highest rank fell at or below the fifth pass (i.e., imperfect or “fuzzy” matches on names or missing qualifying cardiac procedure code) and whose facility identifier did not match between datasets were manually reviewed as potential false positives. Match accuracy from the iterative deterministic algorithm was also compared with a fully probabilistic algorithm available via a SAS application (Link King v6.4) (11).

Table 1.

Iterative Deterministic Matching Algorithm Used in Medicaid-Registry Match

Iterative Matching Scheme		Cumulative % of all linked cases
1	Exact match on all fields	65.5
2	Fuzzy match on date of birth	66.0
3	Fuzzy match date of service	67.1
4	Use of alternate, prior, or family names	76.0
5	Fuzzy match last or first name (~1 -letter typo or substrings and nicknames) and dates, exact DOB	98.0
6	Exact match on all fields, but no qualifying cardiac procedure	99.5
7	Exact match on names or family/alternate name and DOB, fuzzy match on dates, but no qualifying cardiac procedure	100

The initial overall match rate was calculated as the percentage of total cardiac cases performed able to be linked to the Medicaid claims files. The initial Medicaid match rate was calculated as the percentages of patients listed in the registry data as having Medicaid insurance who were able to be matched to Medicaid claims data.

Some surgical cases in the registry data linked to Medicaid claims, but did not have Medicaid listed as payor in the registry. These cases were manually reviewed. Payor was updated in the clinical registries as appropriate and match rates recalculated.

To account for the possibility that there were additional, unmatched cases with improperly coded payor as well, sensitivity analyses were performed, assuming a conservative 80% detection rate of miscoded Medicaid cases. These cases were added to the denominator to calculate a lower bound for the estimated Medicaid match rate.

Medicaid Cohort Representativeness

Linked records were, by definition, Medicaid enrollees. Thus, they were expected to differ in a number of ways from those who did not match (and were presumed non-Medicaid). To establish the degree to which patients in the linked Medicaid cohort were or were not representative of all children undergoing congenital heart surgery across NY State, we first compared proportions of patients operated on at each center in the most complete linked Medicaid cohort versus the rest of the clinical registry (the non-linked, non-Medicaid patients from the STS CHSD/NYS-PCCS). We then compared patient demographics, procedure risk categories (by Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery Mortality Score), preoperative risk factors, lengths-of stay, and operative mortality rates, between patients in the linked Medicaid cohort vs. the rest of the clinical registry. Preoperative risk factors and operative mortality were defined using standard STS CHSD definitions (8). We also described two additional groups: those who were unmatched to Medicaid claims but reported Medicaid enrollment in the registry (potential false negatives), and those who matched to Medicaid enrollment files at some point in the 14-year study period, but not at the exact time of their cardiac procedure. Standard summary statistics and univariable analyses were used for all comparisons.

Longitudinal Follow-up

The ability to track patients over time using the linked Medicaid data was assessed by calculating the proportion of months each patient was enrolled in Medicaid during the 3-, 5-, and 10-years post procedure and then calculating the proportion of patients able to be tracked across each time period. If a patient died during the follow-up period, they were included as “tracked” for the remainder of that window. Churning on and off of Medicaid or disenrollment from the program is a known phenomenon for Medicaid enrollees, due to administrative barriers, changes to family income, or movement across state lines. We tested two definitions of follow-up to account for noncontinuous enrollment. The first definition required patients be enrolled in Medicaid for at least 80% of months in a given follow-up period (“near-continuous” enrollment) or until death. The second required patients only be enrolled in Medicaid for at least 50% of months in each follow-up period or until death. Descriptive statistics were used to describe common patterns for loss to follow-up; “churners” were defined as patients who cycled on and off of Medicaid at least three times during the follow-up period; “disenrollees” were defined as patients who left Medicaid at some point following their procedure and never returned.

To evaluate the full spectrum of the utility of the data repository, we also assessed our ability to detect prior Medicaid enrollees (patients in the registry who previously had Medicaid claims, but were not enrolled in Medicaid at the time of their procedures), future enrollees (patients who enrolled in Medicaid within one year of their procedure), and patients with incomplete claim filing or multiple payors at the time of surgery (patients enrolled in Medicaid as of the month of their registry-listed surgery, but who did not have a Medicaid in-patient claim for the procedure in question).

Linkages and analyses were performed using SAS version 9.4 (SAS Institute, Inc. Cary, NC). This study was approved by the Western Institutional Review Board, with waiver of informed consent.

Results

Overall Match Rates

In total, the registry contained 14,079 admissions from 11,713 patients under the age of 18, living in New York State at the time of their procedure, January 1, 2006 – December 31, 2019. Overall and Medicaid match rates are summarized in the Central Illustration. Of the 14,079 cases, the registry records listed Medicaid as payor at the time of procedure in 53% (7,479) of admissions. Using iterative deterministic matching, 6,571 of these records were matched to Medicaid in-patient claims, for an initial Medicaid match rate of 87.9%. Medicaid match rates by center ranged from 85.0 to 92.9%. The initial overall match rate (ability to match to all pediatric cardiac cases across New York State) was 46.7% (36.0% to 65.2% by center). In the registry, 843 cases (7% of total) that did not have Medicaid initially listed as payor also linked to Medicaid in-patient claims for their procedure (typically listed as “other payor,” “HMO,” or “self-pay”). Upon manual adjudication, 100% of these cases were determined to have had Medicaid as their insurance at the time of their procedure, for a total Medicaid cohort of 8,322. Adding these cases to the numerator and denominator of the match rates resulted in a Medicaid match rate of 89.1% (7,414 matches from 8,322 expected cases; 85.5% to 94.0% by center) and an overall match rate of 52.7% (7,414 matched from 14,079 total cases; 45.5% to 67.8% by center).

Allowing for the possibility of greater payor miscoding among unmatched cases (20% miscoding), resulted in an estimated lower bound for the Medicaid match rate of 86.9% (7,414 matches from 8,533 expected Medicaid cases), 84.8 to 90.5% by center.

Matching Algorithm Performance

The majority of Medicaid matches (65.5%) were found via exact match on all key linking variables (first and last names, date of birth, and admit, discharge, or procedure date). The inclusion of family members’ last names and alternate or prior last names contributed an additional 8.9 percent of matches; fuzzy matches (allowing for minor typos, alternate spellings, dates of birth with month and day flipped, or use of common first name placeholders like “Baby Boy” Smith) accounted for an additional 22.0 percent. All other passes of the algorithm, including expanding the universe to those who did not have a qualifying cardiac procedure claim, in total accounted for less than 5% of matches. Upon manual review, 5 cases (<0.1%) were determined to be false positives. The probabilistic matching algorithm (Link King for SAS) performed similarly in sensitivity, but at the expense of additional false positive matches.

Medicaid Cohort Representativeness

Children in the matched Medicaid cohort differed from children in unmatched (largely non-Medicaid) cohort in a number of key ways. Differences are described in Table 2. In brief, children in the Medicaid cohort were more likely to be younger, to be Hispanic or Black non-Hispanic, to have other congenital anomalies, and to require higher risk surgical procedures. Children in the Medicaid cohort were not more likely to have other, non-genetic preoperative comorbidities. Non-risk adjusted (crude) lengths-of-stay were longer and non-risk adjusted mortality was higher among patients in the Medicaid cohort.

Table 2.

Demographic and Clinical Comparison of Matched Medicaid and Unmatched Cases^*

	Matched	Unmatched	p-value
Age at procedure			<0.001
≤60 days	1,958 (26.4)	1,707 (25.6)
61 days – <1 year	2,581 (34.8)	1,884 (28.3)
1 year – 12 years	2,383 (32.1)	2,305 (34.6)
13 years – 18 years	492 (6.6)	769 (11.5)
Race/Ethnicity			<0.001
White Non-Hispanic	2,755 (37.2)	4,497 (67.5)
Black Non-Hispanic	1,611 (21.7)	718 (10.8)
Hispanic	1,882 (25.4)	668 (10.0)
Other	1,014 (13.7)	572 (8.6)
Unknown	152 (2.1)	210 (3.2)
Prematurity in Neonates	221 (14.0)	189 (13.8)	0.202
Other Congenital Anomaly	418 (8.9)	244 (5.7)	<0.001
Other Preoperative Comorbid Condition			<0.001
Preoperative Mechanical Circulatory Support	70 (0.9)	60 (0.9)
Preoperative Shock Persistent at Surgery†	45 (1.0)	37 (0.9)
Preoperative Mechanical Ventilation	1,352 (18.2)	1,065 (16.0)
Preoperative Renal Dysfunction / Dialysis	55 (0.7)	50 (0.8)
Procedure Risk Category (STAT)			<0.001
1	1,725 (31.6)	1,738 (36.0)
2	1,552 (28.4)	1,334 (27.7)
3	671 (12.3)	571 (11.8)
4	1,152 (21.1)	915 (19.0)
5	262 (4.8)	171 (3.5)
No STAT Category	105 (1.9)	95 (2.0)
Postoperative Length of Stay (days)	7.0 (4.0–15.0)	6.0 (4.0–11.0)	<0.001
Operative Mortality	278 (3.8)	185 (2.8)	0.003

^*Numbers are presented as n (%) or median (interquartile range).

^†Preoperative shock are available in the Society of Thoracic Surgeons-Congenital Heart Surgery registry only, not in the New York-Pediatric Congenital Cardiac Surgery registry.

Long-term Follow-up

Requiring near continuous enrollment, we are able to track 78.9% of children in the matched Medicaid cohort over 3 years, 73.8% over 5 years, and 64.9% over 10 years post procedure. Allowing more lenient enrollment requirements, we are able to track 86.1% of children in the matched Medicaid cohort over 3 years, 82.1% over 5 years, and 76.2% over 10. In total, 7.7%, 8.4%, and 10.0% of children died over these same 3-, 5-, and 10-postoperative years.

The majority of children in the matched Medicaid cohort who were lost to follow-up were classified as “disenrollees” (children who were continuously enrolled in Medicaid up until the time at which they were no longer enrolled). “Churning” (enrolling and disenrolling in Medicaid three or more times during the follow-up period) was rare, representing only 0.2% of the total matched Medicaid cohort at 3-years, 0.7% at 5-years, and 2.8% at 10-years (Table 3).

Table 3.

The Proportion of Children in The Linked Medicaid Database Able to Be Tracked over Time

Follow-up*	3-year	5-year	10-year
Requiring ≥50% Medicaid Enrolled	86%	82%	76%
Requiring Near Continuous Enrollment	79%	74%	65%
Mortality	7.7%	8.4%	10.0%
Types of Disenrollment
“Disenrollees” (mean # months tracked in each time period)	17% (13mo)	20% (19mo)	24% (29mo)
“Churners” (mean # months tracked in each time period)	0.2% (21mo)	0.7% (36mo)	2.8% (72mo)
Number of Cases Eligible †	5,685	4,536	1,938

^*These categories are not mutually exclusive; patients who died within each time-period were included in the percent followed.

^†To be eligible for follow-up, patients’ procedures had to occur at least 3-, 5-, or 10- years prior to Dec 31, 2019.

There were 1,998 additional cases in the registry file (14.2% of cases) that did not have in-patient Medicaid claims corresponding with the procedure in the registry, but that did have claims at some other point between 2006 and 2019. These additional cases could be useful for some, but not all, longitudinal analyses. Of these cases, 13.5% (270) were prior enrollees—enrolled in Medicaid prior to their cardiac procedure, but disenrolled by the time of surgery; 17.1% (341) enrolled in Medicaid within 12 months of their registry-listed surgery; 35.8% (715) enrolled in Medicaid more than one year after their registry-listed surgery, and 33.6% (672) were enrolled in Medicaid the month of their registry-listed surgery, but did not have a Medicaid in-patient claim for this procedure, indicating either incomplete claim filings or the presence of multiple payors for some cases.

Discussion

Our analysis successfully linked over half of all pediatric cardiac surgeries from state-wide registry data to Medicaid claims across a 14-year period, establishing a novel population-based mechanism for longitudinal assessment of risk-adjusted outcomes, resource utilization, health expenditures, and health disparities for patients with congenital heart disease. The methodology utilized for this linkage was highly efficacious, with nearly 90% Medicaid case capture and virtually no false matches. Claims data were available for 82% of linked patients at 5-years and 76% at 10-years. Further, two thirds of children in the registry state-wide were on Medicaid at some point during the study period.

The Centers for Disease Control and Prevention, United States Congress, The American College of Cardiology, and The American Heart Association have all defined longitudinal follow-up and health expenditures for patients with CHD as top research and patient care priorities (2–4). Yet there is currently no way to track the full scope of CHD across the lifespan. The federal government has recently invested heavily in population-based surveillance. Substantial advancements have been made even within just the last few years, through partnerships with state Birth Defects Registries and others (12–16). Others have begun to describe long-term mortality by linking National Death Index to registries such as the Pediatric Cardiac Care Consortium (17–19). To our knowledge, however, this linkage has formed the first and largest population-based data repository of congenital heart patients that includes robust diagnostic and procedural clinical details, comprehensive health expenditure data, and socioeconomic data in a longitudinal form. Further, our Medicaid data include mother-infant pairs, which will allow us to assess care from the fetal period through adulthood. Such longitudinal data allow for the identification of areas with large variation in practice, outcomes, and payments, in order to prioritize targets for future interventions. The Congenital Cardiac Care Map presented in Figure 1 delineates example targets for interventions across the lifespan that could be studied in the linked dataset.

Figure 1. Congenital Cardiac Care Map.

Map delineates potential targets for interventions across the lifespan able to be studied in the linked dataset.

Previously, investigators within adult medicine explored and paved the way for linkage between clinical registries and payor data. The prototype was piloted by Potosky and colleagues for cancer; they developed The Surveillance, Epidemiology and End Results (SEER)-Medicare linked database in 1993, linking over 94% of SEER participants diagnosed at ≥65 years of age via deterministic matching (20). The Society of Thoracic Surgeons demonstrated that linkages were feasible in adult cardiac surgery registry and Medicare data via probabilistic matching on indirect identifiers (21–23). In 2010, Boscoe and colleagues established the scientific premise of utilizing state Medicaid data for longitudinal outcomes assessment, linking Medicaid data to a New York cancer registry (10). This team has explored its use in the study of long-term end-of-life care for young adult cancer patients (24). In each case, these linkages have resulted in rigorous and influential research on long-term outcomes for patients, building understandings of barriers to care, outcomes and resource optimization, and health inequities (25–28).

Successful linkage of NY State Medicaid data to both locally-held state and national registry data establishes the scientific premise for the creation of a multi-state STS CHSD data linkage. There is additional potential for utilization of similar linkages to facilitate the study of longitudinal outcomes and expenditures outside of pediatric cardiology, particularly in other chronic diseases in children that suffer from similarly high variabilities in care, patient heterogeneity, and small populations within patient subgroups. The STS CHSD is the largest congenital heart surgery clinical registry in the world, with highly granular clinical detail. Linking STS CHSD data with national claims—Medicaid and private payor—expands our potential to assess variation in practice, outcomes, and payments, and to prioritize targets for future interventions on a national level.

This work also establishes a state-wide, multi-disciplinary collaborative across 10 of the 11 surgical centers within NY State and the NY Department of Health. The wide range of expertise within CHS-COLOUR, including pediatric cardiologists, congenital heart surgeons, health economists, health policy experts, health services researchers, and policy makers, shows the promise of data sharing in research and quality initiatives. Furthermore, expertise from this diverse group, coupled with data focused on some of the most vulnerable congenital heart patients, will be used to facilitate the identification of modifiable social determinants of health and will focus on the creation of practical, sustainable plans for action, and translation of knowledge into changes in care.

Limitations

There are limitations to our linkage. First, our work thus far is limited to the Medicaid population. It is known that patients on Medicaid—and particularly Black and Hispanic patients and children from lower income neighborhoods within the Medicaid population—are at particular risk of mortality and higher resource requirements after cardiac surgery (7). Further, Medicaid eligibility in NYS becomes more stringent (requires a lower income) after the first year of life and then again as children move into adulthood. Thus, longitudinal follow-up in our cohort is biased to the lowest income patients. While focusing on this population does not allow us to compare with a private payor or higher income cohorts, it allows us to focus specifically on potential barriers to care within this higher-risk population. We are working to expand this linkage across other states and to encompass private payor data. Second, as with any registry or administrative dataset, there can be coding errors. Incorporation of registry data addresses concerns about administrative data’s lack of cardiac diagnosis and procedure specificity, while utilizing the strengths of administrative data for longitudinal follow-up. Coding errors, however, may still exist. One area in particular that may influence our match rate is the registry coder’s ability to distinguish Medicaid from Children’s Health Insurance Program (CHIP) or managed care plans whose names include the name of a private payor. Most STS CHSD and NYS-PCCS clinical registry data managers in the Collaborative have clinical backgrounds, making them well suited for congenital heart data abstraction. They may be less well versed in assigning insurance status. The Medicaid claims data we used do not contain CHIP claims. It is possible that miscoding in the registry of CHIP patients as Medicaid may have resulted in a lower than true match rate. It is, therefore, possible that our true Medicaid match rate was greater than 89%. Miscoding or incomplete billing could also happen within the administrative data. It is also possible that patients might undergo procedures captured on secondary insurance or after leaving Medicaid. Future addition of private payors to this linkage will address this gap. Third, not all aspects of care and outcomes are available in this longitudinal data source. Medicaid data include details on all billed services associated with in-patient, out-patient, emergency room, pharmacy, rehabilitation, home healthcare, and educational services, including direct patient identifiers, mother-infant pairs, healthcare reimbursement, and out of pocket expenses, for both fee-for-service and managed Medicaid patients, whether they occur in or outside of the State. These data will not, however, include clinical laboratory values, echocardiographic findings, results of other advanced imaging modalities or catheterizations, pathology, etc. Future linkages will be needed to address these gaps. Finally, while it is unknown the degree to which expansion of Medicaid eligibility in New York State might influence long-term follow-up for the CHS-COLOUR cohort, data have generally shown minimal migration as the result of Medicaid expansion(29) and New York and neighboring states have maintained reasonably generous and similar eligibility thresholds for both children and adults over time. That said, effects might weigh differently for sicker patients with congenital heart disease who both know they have medical needs and who might qualify for Medicaid as they transition to adulthood based on disability status.

Conclusions

In conclusion, we have established a novel collaborative and data source for population-based assessment of longitudinal outcomes, resource requirements, and health inequities for children with congenital heart disease. This work establishes a model for assessment of heterogeneous chronic diseases across the lifespan, focused on a vulnerable pediatric population.

Clinical Perspectives.

Competency in Systems-Based Practice:

Federal agencies and professional organizations have identified longitudinal follow-up and assessment of resource utilization for children with congenital heart disease as top research and clinical priorities. Statewide linkage of Medicaid claims and clinical registry data permits longitudinal studies of outcomes and expenditures for a large proportion of children with congenital heart disease across inpatient, outpatient, emergency department, pharmacy, rehabilitation, and home care settings.

Translational Outlook:

Further efforts are needed to integrate a national network data resource for more complete assessment of longitudinal outcome, health expenditure, and disparities for children with congenital heart disease.

Abbreviations:

CHD

Congenital heart defect

US

United States

CHS-COLOUR

New York Congenital Heart Surgeons Collaborative for Longitudinal Outcomes and Utilization of Resources

STS CHSD

Society of Thoracic Surgeons Congenital Heart Surgery Database

NYS-PCCS

New York State Pediatric Congenital Cardiac Surgery Registry

ICD-9

International Classification of Disease-9

SEER

Surveillance, Epidemiology and End Results

CHIP

Children’s Health Insurance Program