Factors Associated With Attendance for Cardiac Neurodevelopmental Evaluation

The Cardiac Neurodevelopmental Outcome Collaborative (CNOC) clinical registry was developed to collect neurodevelopmental data for children with CHD to inform best practices for neurodevelopmental care. The registry was built within the same web-based platform as the Pediatric Cardiac Critical Care Consortium (PC⁴) data registry, a quality improvement collaborative with rigorous data auditing for pediatric cardiac intensive care units across North America.^25–27  Integration of the CNOC and PC⁴ registries was an intentional, collaborative effort of Cardiac Networks United²⁸  to facilitate projects spanning cardiac intensive care and neurodevelopmental follow-up. Both registries require training of data entry personnel before data entry can begin. For PC⁴, this includes a 2-day training on the database, data entry process, and data definitions and the requirement to pass a certification examination.²⁷  CNOC has developed a similar half-day training (because of fewer number of variables) followed by a certification examination. Both registries also host regular conference calls with data entry staff to provide clarification on definitions and processes. Institutional Review Board approval, with waiver of consent, was obtained for this project at the University of Michigan Data Coordinating Center and appropriate data use agreements were completed with each contributing site.

Patient-specific inclusion criteria were infant CPB, defined as bypass in the first year of life, date of birth between September 15, 2017 and September 15, 2018, and alive at 11 months of age (making them eligible for a toddler neurodevelopmental evaluation). Infant CPB was chosen as the primary cardiac inclusion criterion because it is a key risk factor for neurodevelopmental disorders in children with CHD.⁵  The birth dates were chosen based on when a toddler’s neurodevelopmental evaluation would have occurred relative to when the CNOC registry became available for data entry (May 2019) and when institutions halted in-person visits because of the coronavirus disease 2019 pandemic (March 2020). To identify eligible patients and to collect sociodemographic and medical predictors of clinic attendance, eligible CNOC centers had to contribute data to the PC⁴ registry by 2017.

The primary outcome was attending a toddler neurodevelopmental evaluation, defined as any visit between 11 and 30 months of age, regardless of whether formal neurodevelopmental testing was performed. This broad age range accounted for practical application of a recommended^5,29  18-month (ie, toddler) visit in clinical practice and across institutional models.

To determine whether a patient attended the neurodevelopmental evaluation, all eligible patients in the PC⁴ registry were cross-referenced with those in the CNOC registry. Because of potential lags in data entry in the new CNOC registry, each participating CNOC center was provided a list of patients at their site who met inclusion criteria but did not have data in the CNOC registry. Centers then queried their institutional data and entered the following information into a standardized data collection form in REDCap (Research Electronic Data Capture^30,31 ): whether any neurodevelopmental evaluation was performed, the age of the child at the visit if it occurred, and the reason if the visit did not occur. Instructions for entering the REDCap data, including what qualified as a neurodevelopmental evaluation, were provided to each center via conference call and in writing. Because PC⁴ was used for all sociodemographic and medical variables (details below), data standards for these variables were not affected by reconciliation of the REDCap and clinical registry data.

Sociodemographic and medical variables were collected from the PC⁴ registry. Sociodemographic variables included the child’s sex, race and ethnicity, insurance type, and household zip code at the time of surgery. Household and cardiac center zip codes were used to estimate the distance patients lived from their center. Distance from center was then categorized into approximate quartiles of the distribution of the data (≤15 miles, 16–45 miles, 45–100 miles, and >100 miles). Institution-specific regulations for sharing private health information within clinical registries limited the availability of household zip codes for some patients.

Medical variables included antenatal diagnosis, gestational age at birth, genetic diagnosis, age at first cardiac surgery, history of stroke, history of clinical or electrographic seizure, history of extracorporeal membrane oxygenation (ECMO), and postoperative length of stay at the time of the first infant CPB operation. Postoperative length of stay was categorized into 2-week intervals based on increased risk of adverse outcome for stays greater than 2 weeks.⁵  The Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery (STAT) mortality score³²  was used to categorize cardiac severity, with higher scores reflecting higher risk of mortality.

Center-specific variables were obtained via a REDCap^30,31  survey developed to optimally capture variations in cardiac neurodevelopmental program design of participating centers. Survey questions addressed the year the program was established, the program’s recommended age for the first neurodevelopmental evaluation, location of the program or clinic in relation to their cardiac center, program referral criteria for neurodevelopmental evaluation (eg, based on AHA and AAP criteria versus only patients with single-ventricle physiology), patient access for neurodevelopmental evaluation at clinics outside of their neurodevelopmental program, and hospital or family-initiated scheduling process for neurodevelopmental evaluation. Hospital-initiated scheduling included automatic scheduling during a hospitalization or outpatient visit or scheduling by a nurse or coordinator who directly contacted the family. Family-initiated scheduling included family response to a letter, developmental screening, or clinician referral.

Patient sociodemographic, medical, and center-specific variables were considered as predictors for attendance for a neurodevelopmental evaluation. Univariable associations compared attendance rates by categories of these predictor variables. Stepwise backward logistic regression based on P value < .05 was conducted to determine independent predictors of attendance for multivariable models. Throughout, generalized estimating equations with exchangeable working correlation for adjustment because of center were used to control for center-to-center variability. Ten-fold cross-validation was performed to assess model diagnostics including sensitivity, specificity, and area under the receiver operating characteristic curve. R version 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria) was used for all analyses.

Of 24 CNOC centers entering data, 16 were eligible for inclusion (Fig 1). Seventy-five percent of centers (12 of 16) used all AHA and AAP high-risk criteria⁵  for referral for neurodevelopmental evaluation, whereas the remainder used targeted criteria within these guidelines (eg, prioritizing single ventricle patients). All centers recommended the first neurodevelopmental evaluation occur by 13 months of age, 94% (15 of 16) were colocated with the cardiology clinic or in a location familiar to the family (ie, same medical campus where surgery was performed), and 75% (12 of 16) used hospital-initiated scheduling.

The median number of patients who underwent infant CPB at each site was 127 (range 78–302) during the study period, with a total of 2385 patients eligible for a toddler neurodevelopmental evaluation (Fig 1). For the entire cohort, 53.5% (1272 of 2379) of patients were male, 44.5% (894 of 2007) were non-Hispanic white, 58.7% (1164 of 1983) had public insurance, and 47.7% (726 of 1522) lived over 45 miles from their cardiac center. An antenatal diagnosis was made in 54.9% (1142 of 2082), 37.4% (892 of 2385) underwent neonatal surgery, and a genetic diagnosis was present in 21.1% (504 of 2384), the majority of which were Trisomy 21. STAT mortality category 4 represented the largest number of patients (712 of 2385, 29.9%) and category 5 the fewest (217 of 2385, 9.1%). Missing data limited the sample size for some variables, which is reflected in the numbers reported above and in Table 1.

The overall attendance rate for a toddler neurodevelopmental evaluation between 11 and 30 months of age was 29.0% (692 of 2385). An additional 196 patients (8.2%) attended a neurodevelopmental evaluation, but the visit was conducted outside of 11 to 30 months. Another 323 (13.5%) had their primary cardiac care outside the institution where the cardiac surgery was performed; thus, attendance could not be verified. For the remainder of the patients, the reason for not attending was unknown by the center. Individual center attendance rates were highly variable, ranging from 7.8% to 54.3%.

Table 1 displays the number who attended the evaluation over the total sample of available data by patient sociodemographic and medical characteristics. Patients who attended a neurodevelopmental evaluation were more likely to have an antenatal diagnosis, undergo neonatal surgery, be in a higher STAT mortality category, undergo ECMO, have a longer postoperative length of stay, have private insurance, and live closer to their cardiac center. Attendance rates also differed for genetic diagnosis categories and for race and ethnicity, although race and ethnicity covaried with insurance type. Figure 2 provides graphical representation of variations in attendance with confidence intervals (CI) by selected medical factors.

Centers with fewer patients undergoing infant CPB and those using a hospital-initiated scheduling process were more likely to have higher attendance rates (Fig 3). For hospital-initiated scheduling, attendance rates did not significantly differ for centers using automatic scheduling compared with those where a nurse or coordinator directly contacted the family to schedule (33.5% versus 37.6%, P = .08). Program location and program referral criteria were also associated with higher attendance (Table 2).

Multivariable models were examined in 2 ways because of the extent of missing data for sociodemographic characteristics. The first multivariable model (Model 1) examined patient medical and center-level characteristics. Antenatal diagnosis, genetic diagnosis, STAT mortality category, postoperative length of stay, and scheduling process for the neurodevelopmental program were all independent predictors for attending a toddler neurodevelopmental evaluation (Table 3). The second multivariable model (Model 2) added the sociodemographic characteristics. In that model, race and ethnicity, insurance type, and distance from the center were associated with attending the evaluation, in addition to the previously identified medical and center factors (Table 3). Across both models, a hospital-initiated scheduling process for the clinic had the highest odds ratio for predicting attendance (Model 1 odds ratio 4.24, 95% CI: 2.74–6.55; Model 2 odds ratio 4.85, 95% CI: 1.80–13.09).

In the last decade there has been significant growth in the number of congenital heart centers with neurodevelopmental follow-up programs, but each program’s design varies because of infrastructure and resources. This is the first multicenter effort to evaluate the attendance rate for cardiac neurodevelopmental evaluation and to compare patient and center level variation, predicting attendance. We found an attendance rate for a toddler neurodevelopmental evaluation of 29.0% (range 7.8% to 54.3% across centers) for patients undergoing CPB in the first year of life. An additional 8.2% of patients were evaluated before or after 11 to 30 months of age. Prior single center studies report 4% to 17% of eligible patients are evaluated in cardiac neurodevelopmental clinics.^21,22  However, these data were published in an earlier era and include different age ranges of follow-up. The attendance rate of 29.0% likely reflects some improvement since these single center studies but may not be indicative of follow-up rates in older children. The factor providing the greatest prediction of attending the neurodevelopmental evaluation was whether the hospital or the family initiated the scheduling of the appointment. This finding emphasizes the importance of infrastructure and programmatic design as a potentially modifiable factor for optimizing cardiac neurodevelopmental follow-up and may inform models of care for other pediatric populations at risk for neurodevelopmental disorders.

Nearly a third of patients with complex CHD who undergo neurodevelopmental evaluation receive a new referral for early intervention, therapy, or ancillary medical services at their first visit.²⁰  Throughout the course of outpatient neurodevelopmental care, over 70% of patients with CHD receive this type of referral and/or a new psychosocial or neurodevelopmental diagnosis.²²  Targeted developmental services and interventions may improve cognitive and motor outcomes for preterm children,¹⁶  enhance academic performance in children with lead exposure,¹⁷  and improve inhibitory control, attention, and cognitive regulatory skills in adolescents with CHD.³³  Yet, compared with preterm children, children with CHD appear to receive fewer therapy services.³⁴  The potential benefits of neurodevelopmental intervention highlight a need to increase the proportion of children with CHD who are evaluated in neurodevelopmental programs.

One modifiable component of cardiac neurodevelopmental follow-up that could improve attendance rates is program design. CNOC centers included in the current analysis were generally well-established centers (based on number of patients undergoing CPB, age of the program, and ability for data entry into the CNOC clinical registry), but the structure of each clinic varied by organizational model and resources.³⁵  Program-specific variation is also reported for Canadian and European cardiac centers, with 40% to 50% having structured cardiac neurodevelopmental programs but variation in surveillance, screening, and evaluation approaches across sites.^36,37  CNOC site level differences provided an opportunity to evaluate specific program components in relation to attendance. The strongest predictor of attendance in this cohort was the clinic’s scheduling process. Although standardization of scheduling across centers may be unrealistic, it seems critical that centers ease scheduling burden for the family. Hospital-initiated scheduling not only reduces burden but provides another interaction with the family to convey the importance of neurodevelopmental follow-up and to coordinate the visit with other clinical appointments. In addition to scheduling process, patient volume and clinic location were also associated with attendance, although only location of the clinic remained in the multivariable model. The univariate association of higher attendance for sites with fewer infant cardiopulmonary bypass cases may reflect catchment area, particularly since patient volume did not remain a predictor in multivariable models including distance from center. Two additional considerations for optimizing return, which were not included in the current study, are incorporating neurodevelopmental care during hospitalization and educational initiatives. Prior data suggest neurodevelopmental surveillance while patients are hospitalized for cardiac surgery may improve follow-up rates.³⁸  Inpatient and outpatient educational efforts geared toward families and healthcare providers could also increase follow-up. Collaborative multicenter approaches aimed at optimizing clinic design while leveraging individual institutional and international experiences will be an essential next step, including understanding resource utilization for hospital-initiated scheduling. Interval surveys of CNOC sites to understand changes in programmatic infrastructure and to incorporate newer, less established programs will be an important component.

Patient-specific factors were also associated with attending a toddler neurodevelopmental evaluation. Similar to prior data,²¹  greater surgical complexity and longer length of stay related to higher attendance. It is also not surprising that antenatal diagnosis was associated with higher attendance, as these patients generally have more complex cardiac disease and have been counseled about neurodevelopment. Further, antenatal diagnosis may relate to access to care and resources available to the family.^23,39  Consistent with this, greater distance from the cardiac center and nonprivate insurance were associated with lower attendance rates. Both of these factors relate to follow-up for pediatric and adult cardiac care in other cohorts, yet healthcare utilization and cost is significantly higher for those who live in rural communities and/or further from their tertiary care.^21,23,24  Minimizing barriers that affect access to care and developing new strategies to limit travel burden, such telehealth visits for initial intake screening, as outlined by the CNOC Telehealth Task Force,^40,41  will likely be essential components for improving follow-up. Race and ethnicity was also associated with attendance rates. However, this relationship likely reflects nonbiological, social and structural factors and differential lived experiences contributing to inequities that were not captured within this data.^42,43  Finally, lower attendance was seen for those with Trisomy 21. This may be because of subspecialty follow-up clinics or other parent resources available for patients with Trisomy 21, potentially making clinic visits seem less relevant for families who have already leveraged neurodevelopmental and related community services. This may also be occurring for other groups of patients, such as those with seizures or stroke who may be followed in neurology clinics and/or by rehabilitation medicine.

The current study has several limitations. First, data were limited to centers contributing to the CNOC and PC⁴ registries. These centers may have the greatest infrastructure and resources for cardiac neurodevelopmental care, so our findings may overestimate attendance rates across newer programs. In addition, a proportion of follow-up evaluations could have been performed at a location not associated with the cardiac center, in specialty clinics, or at other centers closer to where the family lives, which would underestimate attendance rates. Also, completing neurodevelopmental evaluations was challenging during the coronavirus disease 2019 pandemic, especially during lockdown periods. A specific range of birth dates was used to account for this, but the true follow-up rate when a pandemic does not disrupt care may still be underestimated. Finally, the use of REDCap to capture neurodevelopmental visits not entered in the CNOC registry may have introduced the potential for error in estimating follow-up rates. An additional limitation was the use of registry data, which provided a robust sample size but meant specific factors were unavailable, such as primary language spoken, or had higher rates of missing data, such as zip code. Missing sociodemographic data did significantly reduce the sample size for statistical modeling, but data from 1163 patients were still available and provided an adequate sample for analysis. Nonetheless, results should be interpreted in the context of this sample size reduction and the possibility that missing data may not be random. To address this limitation in future analyses, the Cardiac Networks United clinical registries now include a Health Equity Module that uses the Decentralized Geomarker Assessment for Multisite Studies tool, which allows sites to enter information about where a patient lives without directly exchanging protected health information across institutions.⁴⁴  A final limitation is that center factors were specific to those programs included in the current study. As CNOC membership continues to grow and individual programs develop unique strategies for neurodevelopmental care, it will be important to re-evaluate programmatic design to determine whether other factors or new approaches improve attendance rates.

An attendance rate of 29.0% for neurodevelopmental evaluation reflects some improvement compared with previous reports but remains low. Changes to infrastructure and program design and minimizing barriers that affect access to care are essential for improving neurodevelopmental care for children with CHD. These results lay the foundation for more detailed evaluation of variation in cardiac neurodevelopmental programs, and inclusion of additional centers, to inform program design. Evaluation of neurodevelopmental outcomes at a multisite level for those who attend neurodevelopmental appointments will also be important for understanding diagnosis-specific risks that may allow targeted neurodevelopmental care. Finally, a critical component of future efforts should involve health care policy and advocacy to ensure equitable and accessible care for all children with CHD⁴⁵  and collaborative efforts with patient and family groups to understand perceived barriers. This will be important for understanding access for evaluations, as well as variations in region-specific resources such as the availability of early intervention services once a follow-up visit occurs.

We acknowledge the PC⁴ and CNOC data collection teams at the participating centers; the various Cores and Committees of CNOC that provide infrastructure for the CNOC clinical registry; and the University of Michigan Congenital Heart Center and the Heart Institute at Cincinnati Children’s Hospital for their support of Cardiac Networks United.

AAP

American Academy of Pediatrics

AHA

American Heart Association

CHD

congenital heart disease

CNOC

Cardiac Neurodevelopmental Outcome Collaborative

CPB

cardiopulmonary bypass

ECMO

extracorporeal membrane oxygenation

PC⁴

Pediatric Cardiac Critical Care Consortium

REDCap

Research Electronic Data Capture

STAT

Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery