Prediction of adverse events after catheter-based procedures in adolescents and adults with congenital heart disease in the IMPACT registry

Abstract

Aims

We sought to identify factors associated with major adverse events (MAE) after cardiac catheterization in adolescents and adults with congenital heart disease (CHD), and create the first model to individualize risk discussions in this growing population.

Methods and results

Improving Pediatric and Adult Congenital Treatment (IMPACT), a National Cardiovascular Data Registry, contains congenital catheterization data from over 87 hospitals in the United States. Demographics, pre-procedure, and procedural variables were collected for patients over age 10. Multivariable logistic regression was used to identify significant predictors of MAE, a composite of death, urgent surgery or procedure due to a catheterization complication, transfusion, embolic stroke, tamponade, extracorporeal membrane oxygenation or ventricular assist device placement, and device embolization, malposition or thrombosis requiring surgical intervention. A risk score was built based on the effect sizes of each predictor and validated in a split sample. A MAE occurred in 686 (2.5%) of the 27 293 index procedures meeting inclusion criteria. The independent multivariate predictors of MAE were older age, pre-procedural anticoagulation use, renal disease, lower haemoglobin, lower oxygen saturation, non-elective procedure, higher index procedure risk and having had no prior cardiac procedures. Being underweight or overweight had borderline significance and was added to the model. The C-statistic for the model was robust at 0.787 in the derivation and 0.773 in the validation cohort.

Conclusion

The factors predicting adverse events after cardiac catheterization in adolescents and adults with CHD are different than in the general population. Validation of this model in other national or multi-institutional datasets is the next step.

Introduction

Over 13 million adults in the world are currently living with congenital heart disease (CHD),¹ and this number continues to grow. While cardiac catheterizations are often recommended as part of the diagnostic workup (for evaluation of pulmonary vascular resistance or quantification of shunts, for instance) or treatment,² the rates of and risk factors for complications following these procedures in adolescents and adults with CHD are not well known. Single-centre experiences have reported a wide range of rates of cardiac catheterization complications (2 to 18%) and mortality (0.2 to 2.3%) in children and adults with CHD.^3–6 In adults with acquired heart disease, older age, female sex, very low or high body mass index (BMI), hypertension, and chronic kidney disease are known to be predictors of access site complications and adverse events after coronary interventions,^7,8 but these factors have not been evaluated when determining risk for teenagers, young adults, and adults with CHD. The previous analyses of procedural complications in CHD have been performed in children^9,10 or focused on post-hoc risk adjustment¹¹ using information acquired during the catheterization itself.¹²

The Improving Pediatric and Adult Congenital Treatment (IMPACT) Registry was created by the National Cardiovascular Data Registry of the American College of Cardiology in 2011,^13–15 and contains data from over 77 000 congenital cardiac catheterizations performed at 87 hospitals in the United States. For each case, baseline information, detailed procedural data, as well as post-procedural events within the hospital admission are recorded. We sought to analyse the factors associated with adverse events in adolescents and adults with CHD undergoing a catheterization and to create a first risk stratification tool that can be used to individualize risk discussions prior to the procedure.

Methods

Study population

Hospitals that perform catheterizations on children or adults with CHD were invited to participate in the IMPACT database. Once enrolled, centres report the baseline characteristics and in-hospital outcomes of all their paediatric and their adult congenital cases through a secure online data collection platform. Details on the data elements captured can be found at https://www.ncdr.com/WebNCDR/impact/home/datacollection (27 January 2017). For this study, all patients over the age of 10 who had a diagnostic or interventional catheterization recorded in the IMPACT database between 2011 and the second quarter of 2015 were eligible, and version 1.0.1 of the data collection form was used.

Outcome

The primary composite outcome of major adverse events (MAE) was defined as intra-procedural death or cardiac arrest, death in the hospital, urgent surgery or procedure due to a complication of the catheterization, event requiring extracorporal membrane oxygenation (ECMO) or ventricular assist device (VAD) placement, device malposition or thrombosis requiring surgery, device embolization requiring retrieval by catheterization or surgery, anaemia requiring transfusion after the catheterization, and embolic stroke within 72 h after the procedure. The cause of those events, with the exception of urgent procedure due to a complication of the catheterization, cannot be attributed with certainty to the catheterization in this version of the IMPACT database. A bleeding event requiring a transfusion, for instance, is likely to be dependent on institutional practice as well as the severity of the adverse event; it is however one of the most common serious post-procedural events, and as such was included in the combined endpoint. Sensitivity analyses were also performed with (i) in-hospital mortality alone as the primary outcome; (ii) restriction of mortality events to in-lab mortality only, and exclusion of transfusion events; (iii) addition of haemodialysis to the composite outcome; (iv) exclusion of tamponade, ECMO and left ventricular assist device (LVAD) placement in the absence of other complications; and (v) bleeding endpoints alone, defined as unplanned vascular surgery due to catheterization complication, transfusions, other vascular complications requiring treatment, and bleeding events.

Variables

Baseline variables which could be obtained prior to catheterization were considered for inclusion in the model, since the goal was creation of a tool that could be easily used clinically to predict risk prior to doing the procedure. Baseline demographic variables were screened for inclusion in the model: age; gender; age and gender adjusted body-mass index percentile; history of any genetic syndrome, lung disease, coagulation disorder, diabetes, liver disease, kidney disease, stroke; pre-procedural haemoglobin and oxygen saturation; use of prostaglandins (coded as such in the IMPACT data collection form, and referring to pulmonary vasodilators in most cases) or anticoagulants. The underlying diagnosis severity was entered in the model as diagnostic severity class based on the Bethesda classification^16,17 and in line with previous publications,¹⁷ based on the International Classification of Diseases-9 code of the indication for the procedure. Procedure risk categories were devised based on the CRISP⁹ and C3PO risk categories^11,18 (see Supplementary material online, Table S1). Procedural variables screened were urgency of procedure (categorized as elective/urgent/emergent/salvage) and arterial vs. venous access site. Hospitals reporting procedures on children under age 15 were considered primarily paediatric if they reported less than 25% of their volume in adults over age 18, or combined paediatric and adult centres otherwise; this was crosschecked with their description on the Children’s Hospital Association and American Medical Association websites.

Statistical methods

Summary statistics were performed with the Student’s t-test for continuous variables and the χ² test for categorical variables. Hierarchical multivariable logistic regression with hospital site as a random effect was used to identify the variables significantly associated with MAE. To establish a parsimonious model, a backward selection process using the variables was performed until 90% of the full model R² value was retained.¹⁹ A split-sample validation was performed, with 70% of patients placed in the derivation cohort. Discrimination and calibration in the validation cohort were then tested by computing the C-statistic and calibration slope and intercept, with a slope of 1 and an intercept of 0 indicating perfect calibration. The computed risk (on the logit scale) was used as a predictor variable in the validation cohort, and the slope and intercept were tested on the model.²⁰ The final model is based on the derivation cohort only.

A risk score was created by assigning weighted values to the variables identified by the final model based on their beta coefficients in the regression model, using the lowest absolute beta weight as the base for the point system.²¹ The risk score was then calculated by adding all the individual weighted values. Continuous variables were categorized based on their 25th percentile for ease of use in the final point score. All analyses were performed in SAS version 9.4 (SAS Institute, Cary, North Carolina, USA).

Results

The inclusion criteria were met by 27 293 index procedures performed at 87 hospitals, with age under 10 being the exclusion criterion for the remainder of patients; 19 105 procedures (70%) were included in the derivation cohort and 8188 (30%) in the validation cohort. The majority of patients had a moderate or complex diagnosis and an elective catheterization, and slightly more than half were adolescents (see Table 1). The most common procedure was a combined procedure (48.2%; diagnostic catheterization and an intervention), followed by isolated diagnostic catheterization (34.1%), isolated atrial septal defect (ASD) closure (7.7%), coarctation repair (3.8%), patent ductus arteriorus intervention (2.1%), aortic (1.4%) and pulmonary valve interventions (1.02%). The majority of participating hospitals were primarily paediatric centres (95.6%), but 71% of those reported adult procedures as more than 10% of their catheterization volume. There were fewer than 5% missing values for the variables included in the model. There were no significant differences in baseline characteristics between the derivation and validations cohorts.

Table 1.

Demographics and procedural characteristics

Characteristic	All patients in derivation cohort n = 19 105	Major adverse event n = 485	No major adverse event n = 18 620	P-value
Female sex	8793 (46.0%)	226 (46.6%)	8567 (46.0%)	0.80
Age	21.64 ± 13.88	25.39 ± 18.10	21.54 ± 13.74	<0.001
Age group				<0.001
10–18 years	12 050 (63.07%)	264 (54.43%)	11 786 (63.30%)
19–40 years	4908 (25.69%)	125 (25.77%)	4783 (25.69%)
41–65 years	1790 (9.37%)	70 (14.43%)	1720 (9.24%)
>65 years	357 (1.87%)	26 (5.36%)	331 (1.78%)
Body-mass index category, adjusted for age				0.076
Underweight	1308 (6.9%)	44 (9.2%)	1264 (6.9%)
Normal	10 050 (53.2%)	251 (52.6%)	9799 (53.2%)
Overweight	3776 (20.0%)	80 (16.8%)	3696 (20.1%)
Obese	3753 (19.9%)	102 (21.4%)	3651 (19.8%)
History of at least one prior cardiac catheterization	12 955 (68.31%)	290 (60.29%)	12 665 (68.52%)	<0.001
History of at least one prior cardiac surgery	13 156 (69.86%)	295 (61.46%)	12 861 (70.08%)	<0.001
Number of prior cardiac catheterizations or surgeries				<0.001
None	8575 (44.88%)	267 (55.05%)	8308 (44.62%)
1–3	1785 (9.34%)	39 (8.04%)	1746 (9.38%)
>3	8745 (45.77%)	179 (36.91%)	8566 (46.00%)
Any genetic condition	1367 (7.16%)	49 (10.10%)	1318 (7.08%)	0.01
Down syndrome	305 (1.6%)	8 (1.7%)	297 (1.6%)	0.91
Bethesda diagnosis class				0.007
Simple	3085 (16.48%)	79 (16.81%)	3006 (16.47%)
Moderate	7656 (40.90%)	161 (34.26%)	7495 (41.08%)
Severe	7976 (42.61%)	230 (48.94%)	7746 (42.45%)
History and risk factors
Chronic lung disease	908 (4.77%)	45 (9.34%)	863 (4.65%)	<0.001
Coagulation disorder	264 (1.39%)	17 (3.53%)	247 (1.33%)	<0.001
Diabetes mellitus	682 (3.58%)	31 (6.43%)	651 (3.51%)	<0.001
Hepatic disease	384 (2.02%)	25 (5.19%)	359 (1.93%)	<0.001
Renal insufficiency	992 (5.21%)	49 (10.21%)	943 (5.08%)	<0.001
Stroke	1196 (6.3%)	29 (6.0%)	1167 (6.3%)	0.81
Pre-procedure haemoglobin (mmol/L)	8.18 ± 1.37	7.67 ± 1.59	8.20 ± 1.37	<0.001
Pre-procedure oxygen saturation (%)	96.89 ± 4.55	95.68 ± 6.18	96.92 ± 4.50	<0.001
Pre-procedure medications
Anticoagulants	2229 (11.67%)	122 (25.57%)	2105 (11.31%)	<0.001
Prostaglandinsa	22 (0.12%)	3 (0.62%)	19 (0.10%)	<0.001
Procedural characteristics
Procedure risk category				<0.001
Low risk	15 161 (79.42%)	311 (64.12%)	14 850 (79.82%)
Moderate risk	3538 (18.53%)	150 (30.93%)	3388 (18.21%)
High risk	390 (2.04%)	24 (4.95%)	366 (1.97%)
Procedure status				<0.001
Elective	17 867 (94.10%)	328 (68.05%)	17 539 (94.78%)
Urgent	956 (5.04%)	103 (21.37%)	853 (4.61%)
Emergency	152 (0.80%)	42 (8.71%)	110 (0.59%)
Salvage	12 (0.06%)	9 (1.87%)	3 (0.02%)
Access location				<0.001
Venous	5119 (26.88%)	114 (23.65%)	5005 (26.96%)
Arterial	444 (2.33%)	30 (6.22%)	414 (2.23%)
Both	13 484 (70.79%)	338 (70.12%)	13 146 (70.81%)

Values are n (%) or mean ± SD, as appropriate, unless otherwise noted.

ICU, intensive care unit.

^aCoded as such in the IMPACT data collection form; refer to pulmonary vasodilators in most cases.

A MAE occurred in 485/19 105 procedures in the derivation cohort (2.5%; 686/27 293 in the overall cohort); the incidence of MAE was similar across intervention types, except for pulmonary valve interventions (0/195 procedures with MAE, P = 0.023). There were 8 deaths in the catheterization lab (0.042%), and 104 patients who had an in-hospital death (0.54%), with two-thirds being in a high-complexity Bethesda class, and most having a low risk category procedure. Components of the combined endpoint of MAE and other adverse events after the index procedure are presented in Tables 2 and 3. The majority of patients who suffered tamponade, ECMO or VAD placement also suffered another adverse event (12 of 14 patients with tamponade, 14 of 15 who had initiation of ECMO, and 4 of 8 patients who had VAD placement). Increasing age had a linear relationship with increasing risk of MAE.

Table 2.

Major adverse events in the derivation cohort

Adverse event	All patients
	n = 19 105
In-hospital mortality	104 (0.54%)
Death in the lab	8 (0.04%)
Cardiac arrest	75 (0.39%)
Embolic stroke	8 (0.04%)
Red blood cell or whole blood transfusion	140 (0.73%)
Device embolization requiring device retrieval via catheterization or surgery	65 (0.34%)
Subsequent cardiac catheterization due to a complication	64 (0.34%)
Other vascular complications requiring treatment	51 (0.27%)
Unplanned cardiac surgery	47 (0.25%)
Unplanned other surgery	23 (0.12%)
Unplanned vascular surgery	22 (0.12%)
Device malposition or thrombus requiring surgery	18 (0.09%)
Event requiring ECMO	15 (0.08%)
Tamponade	14 (0.07%)
Event requiring LVAD	8 (0.04%)

Values are n (%).Total number of events exceeds 485 as patients may have had more than one major adverse event.

ECMO, extracorporeal membrane oxygenation; LVAD, left ventricular assist device.

Table 3.

Adverse events not part of composite outcome of MAE in the derivation cohort

Event	All patients in derivation cohort n = 19 105	Major adverse event n = 485	No major adverse event n = 18 620	P-value
Bleeding event	237 (1.24%)	43 (8.92%)	194 (1.05%)	<0.001
Arrhythmia requiring antiarrhythmic medication	122 (0.64%)	47 (9.69%)	72 (0.40%)	<0.001
Arrhythmia requiring cardioversion	113 (0.59%)	32 (6.60%)	81 (0.44%)	<0.001
Arrhythmia requiring temporary pacemaker	28 (0.15%)	10 (2.06%)	18 (0.10%)	<0.001
New requirement for dialysis	13 (0.07%)	7 (1.45%)	6 (0.03%)	<0.001
New valvular regurgitation	10 (0.05%)	6 (1.24%)	4 (0.02%)	<0.001
Arrhythmia requiring permanent pacemaker	8 (0.04%)	4 (0.82%)	4 (0.02%)	<0.001

Values are n (%).

ECMO, extracorporeal membrane oxygenation; LVAD, left ventricular assist device.

Nearly half of patients had not had prior cardiac catheterizations or operations before the index procedure (45.1% of all patients, vs. 9.3% of patients who had 1–3 prior procedures, and 45.6% with >3 prior procedures). Those without prior procedures had lower disease severity (33.5% simple vs. 28.9% complex by Bethesda class, compared with 8.6% vs. 46.2% in those with 1–3 prior procedures, and 1.5% vs. 55% in those with >3 prior procedures, respectively, P < 0.001). They however tended to have more urgent and higher risk procedures (2.3% if no prior procedures vs. 1.3% in those with ≥1 prior procedure, P < 0.001), were older (median age 25.0 ± 16.7, compared with 23.2 ± 15.3 and 18.0 ± 8.6 in those with 1–3 and >3 procedures, respectively, P < 0.001), and had more medical comorbidities (with the exception of renal insufficiency, which was more common in patients with a history of >3 prior procedures).

The significant independent predictors of the combined MAE in the multivariate model were having a non-elective procedure, a higher index procedure risk, having had no prior catheterization or surgeries, older age, pre-procedural anticoagulation use, history of renal disease, lower haemoglobin and lower oxygen saturation (Figure 1 for effect sizes). Being underweight or overweight had borderline significance and was added to the model. The final model beta-coefficients and equation are presented in the Supplementary material online.

Figure 1.

Significant predictors of the combined outcome in the final multivariable model. Odds ratio and 95% confidence interval for the risk factors included in the final model. Cardiac catheterization or cardiac surgeries performed prior to the index admission are considered prior procedures. Effect of haemoglobin presented as drop of 1 g/dL (0.62 mmol/L). BMI, body mass index.

The C-statistic for our model was robust at 0.787 in the derivation and 0.773 in the validation cohort. The model showed good calibration with a calibration intercept of −0.23, P = 0.18, and slope 0.90, P = 0.12 (see Figure 2 for observed/expected plot). The model had adequate fit in both adolescents (ages 10–18) and adults (C-statistic 0.758 and 0.815, respectively), and in the small subset of procedures done in exclusively adult hospitals as well as the centres that also do paediatric procedures (Table 4). It was also similarly predictive in classes of disease severity, procedure complexity and risk level, and across more inclusive endpoint definitions (Table 4).

Figure 2.

Observed/Expected plot for final model. Mean observed major adverse event (MAE) rate and expected risk of each decile of patients is presented by one data point.

Table 4.

Sensitivity analyses in the validation cohort

Subgroup	Procedures (n)	C-statistic for principal model
Validation cohort overall	8188	0.773
Age group
Adolescents (age 10–18)	5087	0.758
Adults (age ≥ 18)	3101	0.815
Hospital type
Adult only	183	0.788
Paediatric and adult	8005	0.772
Bethesda diagnosis severity class
Simple	1332	0.801
Moderate	3463	0.779
Complex	3393	0.770
Isolated diagnostic catheterization	2738	0.775
Procedural risk category
Low procedure risk	6497	0.758
Moderate procedure risk	1520	0.786
High procedure risk	171	0.827
Patient in ICU prior to and after procedure	351	0.768
No planned cardiac surgery after catheterization	8006	0.771
Planned cardiac surgery after catheterization	156	0.770
Predictive ability of main model for specific outcomes
Outcome		C-statistic
Mortality only		0.871
Excluding post-procedural mortality from combined MAE endpoint		0.754
Excluding transfusion events from combined MAE endpoint		0.769
Addition of dialysis to combined MAE endpoint		0.775
Removal of tamponade, ECMO, VAD use to the combined endpoint		0.770
Bleeding complications only		0.774

ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; MAE, major adverse events; VAD, ventricular assist device.

An integer score, designed based on the effect size of each predictor in the multivariable model, was associated with increasing incidence of the MAE, from 0.96% in patients with a score of 0, to 19% in patients with a score of ≥12 points (Figure 3).

Figure 3.

Components of risk score and incidence of MAE by risk group. Major adverse events are defined as a composite of intra-procedural death or cardiac arrest, death in the hospital, urgent surgery or procedure due to a complication of the catheterization, event requiring extracorporal membrane oxygenation (ECMO) or ventricular assist device (VAD) placement, device malposition or thrombosis requiring surgery, device embolization requiring retrieval, anaemia requiring transfusion after the catheterization, and embolic stroke within 72 h after the procedure.

Discussion

In this large registry study with data from 27 293 cardiac catheterizations in adolescents and adults with CHD, we present a set of new risk factors that are strong predictors of major adverse events. This is the first data-driven risk prediction score built to personalize risk discussions and counsel CHD adolescents and adults, a growing population in the cardiac catheterization laboratory. These include potentially modifiable risk factors, and raise mechanistic and physiologic questions. While the risk of a procedure may vary due to patient, procedural, hospital and operator-specific characteristics, especially in a heterogeneous population such as CHD, the risk score had a robust predictive ability across the diverse patient group represented in the IMPACT Registry, with a similarly strong C-statistic to that of other models used in routine practice (C-statistic of 0.77 in our validation cohort, compared for instance to 0.72 reported in the validation of the CATH-PCI registry bleeding model²²).

The independent risk factors for MAE were different than those described in either paediatric population or adults with acquired heart disease. Procedural variables (non-elective and more complex procedure, no history of prior catheterizations or cardiac surgery) had the strongest predictive effect, followed by factors related to patient medical complexity (history of renal disease, pre-procedural anticoagulant use, lower haemoglobin, older age, being underweight) and increased severity of congenital disease (lower oxygen saturation). These factors are predictive of MAE as well as specific endpoints such as in-lab death, bleeding, and need for dialysis. The factors related to medical complexity are newly identified and add to studies in paediatric catheterization, such as the CRISP model,⁹ which found that the most predictive factors for MAE were age under 1 year and under 30 days, weight <5 kg, clinical instability or ECMO, systemic illness or organ failure and physiologic, diagnostic and procedural risk category, and the Congenital Cardiac Catheterization Project on Outcomes (C3PO) analysis,¹² which identified procedural risk category and elevated systemic ventricular end-diastolic pressure as important risk factors. This model is distinct from the risk adjustment model created from the IMPACT database¹¹ by its focus on factors that are exclusively known prior to the procedure, in order to aid with preprocedural planning and discussion with patients of benefits and risks, and its focus on a more homogeneous age group (this analysis did not include neonates and infants, for instance). In addition to some elements also identified as risk factors in the risk-adjustment model (age, renal insufficiency, lower oxygen saturation and procedural risk group), by not including intra-procedural haemodynamics, our risk score additionally identifies 5 novel significant patient and procedural factors that add to our predictive ability.

While anaemia, renal dysfunction and pre-procedural anticoagulation, risk factors for bleeding or mortality after coronary intervention in the non-congenital population,^22–24 are also predictive in CHD adolescents and adults, sex was not a significant predictor in our population, reflecting a sex interaction specific to coronary artery disease that may be less relevant in CHD, though may emerge as the population ages. Data on history of peripheral vascular disease or heart failure, known risk factors for adverse events in adults after coronary intervention,²³ is not currently collected in the IMPACT Registry; addition of these clinical variables in future versions of the model would be expected to further increase its predictive value. Hospital-level characteristics, such as case volume and mix, presence of trainees, and paediatric vs. adult base, may have been associated with more frequent adverse events in these patients, but were not included in this analysis.

The increase in the risk of MAE seen in patients in our study who had not had prior surgeries or procedures has not been previously described. Their higher incidence of MAE was consistent with their observed older age, higher medical complexity, higher procedure risk and higher frequency of urgent or emergency procedures. For instance, patients with large shunts who have not previously undergone reparative procedures are more likely to be cyanotic, have pulmonary hypertension and downstream multiorgan consequences (such as right ventricular failure, arrhythmias, erythrocytosis, hepatic congestion and renal dysfunction). It is likely that patients who had more than one prior procedure had had an earlier diagnosis and remained in care, the latter of which is associated with improved outcomes. Not having had prior procedures or surgeries remained however an independent significant predictor after adjusting for age, procedure risk and status, though may have a residual correlation with medical comorbidities.

Weights assigned to the risk factors based on their effect size in the final regression model were used to create an easily usable risk score to help guide conversations with patient, family, and referring providers. The use of tools that predict individualized procedural risk (such as the PREDICT calculator) have been shown to increase patient engagement and satisfaction, and lower patient anxiety²⁵; a tool specific to CHD would be expected to have similar results. While a risk score is not a replacement for a detailed discussion of benefits and risks among patient, family members, referring physician, interventionalist and multidisciplinary team,² it can be a helpful guide for the conversation. In addition, integration of risk-predictive tools in the procedure preparation and consent process have been shown to improve procedural planning prior to coronary interventions, for instance by increasing the rates of radial access in patients who are at higher risk of access-site bleeding.²⁶

Finally, this analysis also highlights the potentially modifiable patient factors that are associated with adverse events after a catheterization, such as body mass and preprocedural anticoagulation. The higher risk associated with urgent and emergent procedures supports the need for earlier referrals and adherence to guideline recommendations for active surveillance.

Limitations

The principal limitation of our study is that the components of the primary endpoint are not all known with certainty to be related to the catheterization. There is no post-hoc adjudication of events in IMPACT to establish causality, and some events may have been due to planned cardiac surgery post-catheterization. Our model, however, retained good predictive value even in the subset of patients who did not have a planned surgery after catheterization as well as several other sensitivity analyses. It remains crucial to validate and refine this model in other populations, and in individual procedures as the dataset grows. A limitation of any procedural risk model is that some post-procedural adverse events are not related to patient factors, such as stochastic events, hospital or operator-dependent factors. Collinearity between markers of severity of patient illness and procedure risk group or urgency is likely, and may have driven some patient-specific factors out of the model.

In conclusion, we identified risk factors for adverse events after cardiac catheterization procedures in adolescents and adults with CHD in a large registry. The factors predicting risk are different in these patients than in prior published catheterization risk calculators, and this model we developed can be used to individualize the risk and benefit discussion in this special population prior to a catheterization procedure. Validation of the risk score in other national or multi-institutional datasets is the next step.

Funding

This work was supported by the National Institutes of Health [T32HL007604] to ACSS. The manuscript contents are solely the responsibility of the authors and do not necessarily represent the official views of National Institutes of Health.

Conflict of interest: A.C.S.S.: none declared. A.A.: Medtronic Inc.: Research Grants; Edwards Lifesciences: Consultant, Proctor, Research Grant; Siemens Healthcare AX: Consultant; St. Jude Medical: Consultant, Proctor, Research Grants; B. Braun Interventional Systems Inc.: Proctor; pfm medical, Inc.: Research Grant. K.F.K.: none declared. D.N.: consultant work for W.L. Gore and Associates, B. Braun. J.A.: none declared. A.B.B.: none declared.