This is a cross-sectional, comparative study of 156 adolescents and young adults (80 CHD and 76 Healthy) recruited via flyers or provider referrals from University of California Los Angeles (UCLA) and Children’s Hospital Los Angeles (CHLA) pediatric cardiology clinics and non-hospital based cardiology clinics in Southern California. This study was carried out with Institutional Review Board approval from both UCLA and CHLA. We included adolescents and young adults with CHD between the age of 14 and 21 years, who have undergone surgical palliation requiring CPB. CHD participants were excluded, if they had isolated coarctation of the aorta or patent ductus arteriosus (not requiring CPB), previous head injury (e.g., concussion, stroke) severe developmental delay precluding active study participation and self-reporting (e.g., cerebral palsy, severe hypoxic-injury, or genetic syndrome associated with cognitive delays). If eligible, either a same day or future appointment was made to participate in the study.

Healthy controls were recruited from local high schools and community flyer. This control group was recruited instead of using normative data to capture the high prevalence of Hispanic ethnicity in the City of Los Angeles. Participants were screened by self-report and excluded for any chronic medical or psychiatric conditions or any previous head injury. If eligible, controls were matched to a CHD participant for age (±2 years) and gender, and an appointment was made to participate in the study either at their home, public library, or research office.

After parental permission and assent were obtained from participants under age 18, and informed consent was obtained from participants aged 18 and over, all study procedures were performed with the adolescent and test administrator in a private room. The test administrators were two trained graduate research assistants who met qualifications for administration with interrater agreement of 98% for the Wide Range of Assessment Memory and Learning second edition (WRAML2). The WRAML2 core and optional subtests were completed on all subjects (approximately 1 h) followed by self-administered questionnaires (approximately 15–20 min). Demographic and clinical data were obtained from a limited medical record review which included age, gender, ethnicity, type of CHD, number and type of surgical procedures, first surgery performed <30 days of life, cyanosis at birth, presence of genetic syndrome, attention deficit disorder (ADD) and/or hyperactivity (ADHD), and use of remedial educational services.

Memory was measured using the WRAML2. This administered test is a highly reliable and valid measure of memory and learning abilities in subjects from age 5 to 90 years (19). The broad age range of the WRAML2 was important in instrument selection because of the need to cover adolescents and young adults compared to using two different measures (i.e., child and adult). The WRAML2 is a comprehensive test that measures an overview of memory function which consists of verbal and visual memory, attention/concentration, working memory, and memory recognition. The core battery consists of six subtests [story memory, verbal learning, design memory, picture memory, finger window (short-term memory of a visual sequential pattern), numbers/letters (digit-span format using both numbers and letters)] that when combined yield a general memory index (GMI) score (mean 100, SD ± 15) (19). Additional optional subtests performed were working memory and memory recognition yielding the general memory recognition index (GMR) score (mean 100, SD ± 15). The GMI measures immediate recall and the GMR measures delayed recall. The WRAML2 or GMI has been used in previous cognitive studies in children with and without CHD (20–22). The alpha reliabilities for the core subtests range from 0.85 to 0.94 (GMI 0.93) (19).

Perceived health status was measured using the Short-Form-36 Health Survey Version 2 (SF-36v2). The SF-36 was originally developed to measure perceived health status. However, many researchers incorrectly use the terms health-related quality of life or quality of life in reference to what the SF-36 measures (23). This self-reported questionnaire consisting of 36 Likert questions with 8 health concepts that assess perceived health status in: (1) physical function, (2) the physical function limitation as result of physical injury, (3) the role due to emotional problems, (4) energy and vitality, (5) mental health, (6) social function, (7) physical pain, and (8) general health (24). Two summary scores are calculated from the eight domain scores which include a physical component summary (PCS) and mental component summary (MCS). Summary scores range from 0 to 100 with higher scores indicating better health status (24). The SF-36 has been used in other studies in CHD (18, 25–27). Correlations between the SF-36 and other health status measures ranged from 0.51 to 0.82 in mental health, and 0.52 to 0.85 in physical health (24).

Anxiety levels were measured using the Beck anxiety inventory (BAI) (28). The self-reported BAI is a 21-item questionnaire that measures the severity of common anxiety symptoms. Responses are rated on a 4-point (0–3) Likert-type scale (0 = “not at all bothered” to 3 = “severely bothered”) with a scores ranging from 0 to 63. Higher scores indicate greater anxiety severity (0–21 as low anxiety to 36–63 as severe anxiety). The BAI has been used in previous studies in CHD with a Cronbach’s alpha of 0.93 (29).

Depressive symptoms were measured using the Patient Health Questionnaire Depression Module (PHQ-9). The self-reported PHQ-9 is a 9-items questionnaire, responses are rated on a 4-point (0–3) Likert-type scale (0 = “not at all bothered” to 3 = “bothered nearly every day”), with scores ranging from 0 to 27. Higher scores indicate greater depression severity (0–4 = no depression to greater than 20 = severe depression) (30). The PHQ-9 is a widely used measure of depression severity in the CHD population with a Cronbach’s alpha range of 0.86–0.89 (18, 31).

Excessive sleepiness was assessed using the Epworth sleepiness scale (ESS) (32). The self-reported ESS is a 9-item questionnaire, responses are rated on a 4-point (0–3) Likert scale (0 = “no chance of falling asleep” to 3 = “high chance of falling asleep”), with scores ranging from 0 to 24. Higher scores on the ESS indicate more excessive sleepiness (≥10 is considered positive). The ESS is a widely used instrument with established reliability and validity with a Cronbach’s alpha of 0.86 (32).

Self-efficacy was measured using the general self-efficacy (GSE) scale. The GSE is a 10-item questionnaire used to assess the belief in one’s own ability to cope with difficult demands in life (33). The self-reported responses are rated on how true the statement is for the person on a 4-point Likert-type scale (1 = “not at all true” to 4 = “exactly true”) with scores ranging from 10 to 40. Higher scores indicate greater self-efficacy with a Cronbach’s alpha range from 0.76 to 0.90 (33).

Characteristics of the sample are presented as means with SD or medians with range for continuous variables. Subjects were classified into two groups (CHD and Healthy). Variables were examined for normality and outliers. The continuous data had non-normal distributions (per Shapiro–Wilks tests of normality), and groups were compared using non-parametric statistics consisting of the Mann–Whitney U test for all continuous variables and Chi-squared for all categorical variables. Spearman’s rho correlation coefficients between all predictor and the outcome variable GMI were examined. Only predictors with significant correlations (p < 0.05) to the outcome variable GMI were entered into the multivariable analysis. The stepwise logistic regression model was performed for the binary variable of GMI ≤85 and >85 (1 SD below the expected population mean 100). The software identified the sequence of entry of covariates into the statistical model. All analyses were conducted using the Statistical Package for the Social Sciences version 23.0 (IBM; Somers, NY, USA).

Demographic characteristics of the CHD and healthy control groups are summarized in Table 1. No statistically significant differences in age, gender, ethnicity, and education emerged between groups. However, ethnicity showed a trend toward significance (p = 0.06) with 41 and 39% Hispanic ethnicity in the CHD and control group, respectively. Public insurance was higher in the CHD group (68%) compared to controls (10%) and the incidence of having ADD or ADHD (21%) compared to controls (4%). Use of remedial education services was identified in 24% of the CHD group only. Clinical characteristics of the CHD group are summarized in Table 2. Fifty percent of the sample had complex CHD, first surgery <30 days of age (58%), median of 2 previous surgeries, median 14 years from last surgery, taking 1–2 medication (48%), and New York Heart Association class I (64%).

Comparison of memory, anxiety, depression, sleepiness, self-efficacy, and health status between CHD and healthy controls is summarized in Table 3. In the CHD group, 50% scored 1 SD below the normal for GMI and 8% were 2 SD below the normative of 100 compared to 4 and 0% in the controls, respectively. Furthermore, GMI scores 1 SD below the normal were greater in males [n = 26 (62%)] vs. females [n = 9 (32%)]. Median GMI scores between CHD and healthy controls were 85 vs. 108, p < 0.001, respectively. All subgroups of the GMI and GRI showed statistically significant differences between CHD and controls except for visual memory.

The CHD group had 68 and 50% with greater than or equal to mild anxiety and depression symptoms, respectively, compared to 32 and 25% of the healthy controls. Median anxiety and depression scores between CHD and healthy controls were statistically significant (15 vs. 6; p = < 0.001 and 6 vs. 2; p = 0.002), respectively. Males had more anxiety and depressive symptoms compared to females.

Median sleepiness scores were statistically significant in the CHD group compared to the healthy controls (10 vs. 8; p = 0.016), respectively. Excessive sleepiness (abnormal ≥10) was identified in 32% of the CHD group compared to 27% of healthy controls with no gender differences identified.

Median self-efficacy scores were lower in the CHD group compared to healthy controls (28 vs. 34; p = 0.006), respectively. Low self-efficacy (scores <30) was identified in 51% of the CHD group compared to 19% in the healthy control group. Lower self-efficacy was associated with worse anxiety, depression, sleepiness, and perceived health status. Males had lower self-efficacy than females in the CHD group.

Perceived health status is summarized in two component scores (physical and mental). Median scores in physical were statistically significant in the CHD group compared to healthy controls (49 vs. 58; p < 0.001) with no difference identified in the mental component scores. In the eight subscales, all were statistically significant except for bodily pain, energy/fatigue, and mental health.

The list of covariates associated with GMI included in the multivariate model for the total cohort is summarized in Table 4. Gender, number of surgeries, pacemaker, cyanosis at birth, first surgery <30 days of life, number of medications, defect severity, self-efficacy, sleepiness, anxiety, depression, and physical and mental health status were all entered in the stepwise logistic regression if the variables were significant on the bivariate analyses for memory deficits. The final multivariate logistic regression model for GMI is presented in Table 5. Male gender, number of surgeries, anxiety, and self-efficacy were independent predictors which explain approximately 45–64% of the variance for memory deficits using Cox and Snell R² or −2 log likelihood, respectively.

Our findings are important for providers, patients, and parents to be awareness of potential risk factors associated with memory deficits in adolescents and young adults living with CHD. Unfortunately, certain variables such as male gender and number of surgeries are non-modifiable while anxiety and self-efficacy are amenable to change. More research has suggested that family and maternal factors (i.e., mental health, education, socioeconomic status) may play a more significant role than disease or surgical variables related to behavioral/psychosocial outcomes. However, there are very few reported intervention studies to support psychosocial adjustment in children with CHD and their families (47, 48). Interventions focused on maternal well-being and optimizing parenting skills that focus on self-efficacy and reducing anxiety or worries can be helpful in early childhood development. Other intrinsic or patient factors not included in this study (i.e., hypoxic–ischemic brain injury) that can contribute to memory deficits may benefit from future pharmacological strategies to protect or promote neurogenesis [i.e., thiamine or green tea extract supplementation (49–51) and statins (52)]. Nonetheless, special focus on males with CHD to encourage participation in future interventions is warranted.

Our study should be interpreted in light of some limitations. Although the logistic regression model explained almost half of the variance, other factors not accounted for in this study are making a contribution to memory deficits. Other studies have identified the contribution of maternal factors and socioeconomic status to impact cognitive and neurodevelopmental outcomes (38) which were not measured in this study. However, we did find that the CHD group had a higher prevalence of public vs. private insurance compared to controls. Parents suspicious of a memory problem in their child might have been more motivated to participate, whereas parents of children who perform well at school or parents not wanting to potentially find another problem with their child may have decline disproportionately to participate. Our sample was heterogeneous with the majority of CHD diagnoses classified as moderate to severe and cannot be generalized to simple/less complex forms of CHD. The sample size prohibited further group comparison related to defect severity. The numbers of participants with genetic syndromes or ADD/ADHD in both groups may be higher as testing was not performed and only identified via medical chart review in CHD participants or parental self-report in healthy controls. Primary care chart reviews were not performed on healthy controls as this would be difficult or almost impossible due to inconsistent medical follow-up during adolescents creating potential sample bias. Generalizability of our findings can be challenging given the high proportion of Hispanic participants which is quite different from other cardiac centers. Furthermore, most CHD participants had their last surgical procedure over a decade ago and did not have the advantages in improved surgical technique and medical management of the current era. Neuroimaging was not performed for this study, so the incidence of congenital or acquired brain injury is unknown. However, neuroimaging is currently being performed by our research team in the single ventricle participants in relation to brain structures that affect memory, and we are hopeful that this will provide clarification on the relationships between brain structure and cognitive status in these subjects.