What are the important morbidities associated with paediatric cardiac surgery? A mixed methods study

Objectives Given the current excellent early mortality rates for paediatric cardiac surgery, stakeholders believe that this important safety outcome should be supplemented by a wider range of measures. Our objectives were to prospectively measure the incidence of morbidities following paediatric cardiac surgery and to evaluate their clinical and health-economic impact over 6 months. Design The design was a prospective, multicentre, multidisciplinary mixed methods study. Setting The setting was 5 of the 10 paediatric cardiac surgery centres in the UK with 21 months recruitment. Participants Included were 3090 paediatric cardiac surgeries, of which 666 patients were recruited to an impact substudy. Results Families and clinicians prioritised: Acute neurological event, unplanned re-intervention, feeding problems, renal replacement therapy, major adverse events, extracorporeal life support, necrotising enterocolitis, postsurgical infection and prolonged pleural effusion or chylothorax. Among 3090 consecutive surgeries, there were 675 (21.8%) with at least one of these morbidities. Independent risk factors for morbidity included neonatal age, complex heart disease and prolonged cardiopulmonary bypass (p<0.001). Among patients with morbidity, 6-month survival was 88.2% (95% CI 85.4 to 90.6) compared with 99.3% (95% CI 98.9 to 99.6) with none of the morbidities (p<0.001). The impact substudy in 340 children with morbidity and 326 control children with no morbidity indicated that morbidity-related impairment in quality of life improved between 6 weeks and 6 months. When compared with children with no morbidities, those with morbidity experienced a median of 13 (95% CI 10.2 to 15.8, p<0.001) fewer days at home by 6 months, and an adjusted incremental cost of £21 292 (95% CI £17 694 to £32 423, p<0.001). Conclusions Evaluation of postoperative morbidity is more complicated than measuring early mortality. However, tracking morbidity after paediatric cardiac surgery over 6 months offers stakeholders important data that are of value to parents and will be useful in driving future quality improvement.


Conclusions
Evaluation of post-operative morbidity is inherently complex, but offers stakeholders considerably more information about paediatric cardiac surgery outcomes than the current metric of 30-day mortality.

Strengths and limitations of this study
 The study identifies early paediatric cardiac surgical morbidities that are considered to be important by both lay people and clinicians.  The study prospectively measures the incidence of the defined morbidities in children undergoing heart surgery across five centres.  Based on data collected, the study explores the risk factors for paediatric cardiac surgery morbidity statistically.  Within a subset of patients, the impact of morbidities on patients and families are measured over 6 months in terms of quality of life, days spent at home rather than in hospital and hospital costs.

Study registration
The study has ethical approval from London City Road Research Ethics Committee (14-LO-1442).

Conflicts of interest
None declared by authors.

Background
Over 5000 paediatric cardiac surgery procedures are performed in the UK each year and early survival has improved to over 98% since comprehensive national audit commenced in 2000. 1 2 Most stakeholders including clinicians, commissioners and users believe that whilst these early survival rates remain an important safety measure, it is has become imperative to explore a broader range of measures for outcome in this complex field of practice. Whilst there has been considerable research related to measuring, understanding and reducing perioperative mortality for paediatric cardiac surgery, [3][4][5] , much less attention has been focussed on surgical morbidities.
Morbidity is defined as a state of being unhealthy, or of experiencing an aspect of health that is 'generally bad for you'. In this project, by 'surgical morbidity' we mean a defined aspect of ill health associated with a specific operation. In our study, we aimed to identify which surgical morbidities present the greatest burden on patients and health services following paediatric cardiac surgery. Views may differ between professionals and non-professionals as to what the term morbidity exactly refers to and which surgical morbidities are most important. 6 Therefore we set out to combine patient and carers' perspectives with those of professional groups in defining a prioritised list of morbidities for prospective evaluation.
Over four years we: 1. Identified and defined nine surgical morbidities following paediatric heart surgery, taking into account views from patients, carers, psychologists, nurses and clinicians, which together captured important aspects of the clinical and health-economic burden;

Design and Setting
Our multidisciplinary mixed methods study took place across five UK paediatric cardiac surgery centres representing a range of programme sizes, which care for half of all patients nationally.

Participants
The participants were children under 17 years old with congenital heart disease (CHD).

1: Which morbidities?
Between January 2014 and September 2015 we reviewed existing literature, ran three family focus groups and undertook a family online discussion forum moderated by the Children's Heart Federation (a user group). Transcripts were thematically analysed and the resultant themes, together with the literature, helped to inform a long list of candidate surgical morbidities to consider. A multi-disciplinary group with patient and carer representation then ranked and selected a list of nine key morbidities using the nominal group technique and secret voting (6). This 'selection panel' 7 was informed in turn by clinical views on definitions and feasibility of routine monitoring for each candidate morbidity provided by an independent 'definition panel' 8 as reported in previous publications.

Incidence of morbidity Measurement
Between October 2015 and June 2017, we measured morbidity incidence in consecutive surgical admissions across the study sites. Morbidities were attributed to the immediately preceding cardiac surgery and defined within the same hospitalisation, except for unplanned re operation which was defined as an unanticipated procedure within 30-days (see 8 for details) and mediastinitis which could be identified post-discharge by the operating surgeon (part of post-surgical infection morbidity see 8 for details). Data were regularly checked for completeness, clinical congruence and accuracy and, a three month sample of data was independently validated against the national audit data.
As for the UK audit of 30-day mortality, 9 all procedures more than 30 days apart on the same patient were included in the morbidity analysis. As a secondary outcome, we checked each patient's survival status at 6 months after first appearance in the dataset, based on National Congenital Heart Diseases Audit (NCHDA) 2   We used multilevel logistic regression analysis to explore the role of pre-operative, patientlevel case mix factors on the occurrence of any morbidity versus no morbidity, accounting for multiple procedures within patients. Clinical risk factor groups were derived from the finer diagnostic coding, based on previous peer reviewed research by our group. 9 10 A univariate model to predict risk of any morbidity vs none was fitted for each risk factor.
The estimated Odds ratios (OR) are presented along with 95% confidence intervals (CI). All factors significant on univariate analysis (p<0·1) were considered in a multivariable model. We used multiple imputation by chained equations to account for missing data. The imputation model included all risk factors considered in the univariate analysis, which we assumed included all predictors of missingness (missing data are indicated in our results).
The multivariable models were derived by fitting a regression model for all significant predictors and estimates were combined using Rubin's rules. 11 The secondary outcome of survival at 6 months was compared between morbidity groups using logistic regression. All analyses were performed in Stata 14. 12

Impact of morbidity Recruitment
For our study of impact, we attempted to recruit all patients with morbidity (as defined 8 ) where at least one parent spoke reasonable English, and the family were resident in the UK.
Where feasible, for each case we recruited a patient with no morbidity, matched on treating centre, age and univentricular status. Recruitment ran between October 2015 and June 2017 for all but one site where it stopped after 6 months due to resource constraints. The recruitment strategy led to a balanced sample in terms of morbidity patients and those with no morbidity (see results for detail), however we could not match every patient with a control.
We included the available data on children who subsequently died.

Measures of impact
We evaluated the impact of morbidities over the 6 months immediately following surgery on the following outcomes:

a) Quality of life and psychological burden on children and parents
We used age-specific measures at 6 weeks and 6 months following surgery: The PedsQL subscales (physical and psychosocial) and total scores, which range from 0 and 100, a higher score indicating better quality of life. PedsQL scores in a normal healthy population vary by age; expected scores for infants 0-12 months (which encompasses the median age for our cohort) are mean physical 84·98 (SD 9·45), mean psychosocial 80·47 (SD 12·64) and mean total 82·47 (SD 9·95). [13][14][15] For their parents, we used the PHQ-4, which comprises four questions, measuring anxiety and depression in adults. Individual items are scored from zero to three. Scores ≥three for the first two questions suggest anxiety, and scores ≥three for last two questions suggest depression. 16 In the normal population, 4·8% have scores suggestive of anxiety and 6·6% have PHQ-4 scores suggestive of depression. 17 We used mixed effects regression models for PedsQL/PHQ-4 results to compare the impact of morbidity (any versus none) on outcome. All models were adjusted for clustering within matched pairs and significant covariates associated with incidence of morbidity. The inclusion of hospital site made negligible difference to the results and is not included in our results.
We used multiple imputation, by chained equations to account for missing data for those patients known to be alive at 6 weeks and 6 months respectively.

b) Days at home over 6 months as an additional measure of disruption to family life
Length of inpatient stay following the index procedure was defined as the number of days between the operation and the date of discharge from the tertiary centre. Hospital stays that exceeded the 6-month follow-up period were censored after 183 days. All subsequent hospital admissions were captured and the total days spent in hospital including tertiary and secondary care admissions was subtracted from 183 days to generate a value for days at home by 6 months. Children who died in hospital were assigned a value of zero for this outcome.
We used quantile regression to estimate the effects of individual morbidities in terms of differences in medians since the measure was negatively skewed. Models were adjusted as described above. Given the high level of data completeness imputation was not used.

c) Costs of the index hospitalisation
We calculated the cost of the inpatient stay in the tertiary centre following the index surgical procedure as a key indicator of the economic burden to the hospital provider (  Costs were calculated as the summation of days in intensive care unit (ICU), days on the ward, and all operating procedures within the defined period. For ICU stays, data were recorded on the level of care the patient received each day (Enhanced Care, High Dependency, High Dependency Advanced, ICU Basic, ICU Basic Enhanced, ICU Advanced, ICU Advanced Enhanced and ECLS). 18 Unit costs of ICU stays were daily costs for each level of care (ranging from £870 to £5440) applied to the number of days spent receiving that level of care. Unit costs per day for ward stays were obtained from the highest recruiting centre and varied according to the age of the patient (<two years, £904; ≥two years, £680).
Unit costs for surgical procedures were costs per minute supplied by study sites; these varied by site and so we used the value for the site that recruited the largest number of patients to this study (£13·39 per minute), which was similar to the mean value across all sites.
We modelled the relationship between morbidities and cost of index hospitalisation, adjusted for significant covariates associated with hospital costs. Costs were skewed and to account for this we used a generalised linear model (GLM) with gamma family and log link, which has been recommended for modelling positively skewed data. 19 Missing outcome values were imputed using an iterative Markov chain Monte Carlo procedure based on multivariate normal regression, generating 20 imputed datasets.

Patient sample size
We anticipated that between 3000 and 3300 surgical patients would be admitted across the five sites during the study period. 2 We assumed that a clinically relevant difference in quality of life between matched pairs corresponded to a mean difference in quality of life score of at least 0.5 standard deviations. 20 To detect such a difference for PedsQL responses at 5% significance with 80% power requires a minimum of 32 matched pairs. Allowing for a 10% loss to follow-up rate, we aimed to recruit 36 matched pairs for each patient with morbidity, giving 80% power to detect a significant effect for any morbidity with a prevalence of at least 1·5%.

Patient and public involvement
The patient or user perspective lies at the core of the study methodology. Specifically, a key goal of the study was to consider the views of patients and parents when measures of morbidity are selected for future audit and benchmarking, in particular since emphasis may potentially differ between professionals and parents / patients. Patient and family representatives from the Children's Heart Federation (CHF) were involved in aspects of the study design including design and running of the focus groups and the online discussion forum.
Representatives of the CHF participated in the facilitated nominal group meetings to select morbidities for inclusion in the incidence and impact studies.
A member of the CHF and a parent representative sat on the project steering group.

1: Which morbidities?
Our final selected 7  We treated children with more than one of these events as a distinct group referred to as 'multi-morbidity'. Recognising ECLS as a very severe event, which nearly always occurs with other morbidities, we treated ECLS as standalone morbidity (and always excluded from the multiple morbidity group). [21][22][23][24][25] There was some divergence between the views of clinicians and families about the fundamental issue of what the important morbidities linked to paediatric cardiac surgery are.
Health professionals from tertiary centres tended to prioritise clearly clinical issues related to the heart (e.g. use of ECLS and re operation) whereas parents placed greater emphasis on holistic outcomes for their child (e.g. feeding difficulties and child developmental problems). 7

Descriptive results
We collected data on 3248 cardiac procedures and analysed 3090, after excluding 99 cardiac re-operations (a morbidity outcome) and 59 procedures that did not meet inclusion criteria.
The incidence of individual morbidities in isolation and overall for each is shown in Figure 1.  The figure shows the distinction between occurrence as a single morbidity and occurrence in combination with other morbidities, with 95% confidence intervals.

Significant risk factors for morbidity
As shown in Table 1, the most important risk factor for occurrence of 'any morbidity' was young age, particularly neonates, who were approximately five times as likely to experience a morbidity compared to children over one year old. Children with more complex heart disease or children whose operation included a bypass time longer than 90 minutes were almost twice as likely to experience morbidity. Other important independent risk factors were: palliative or staged procedures, the presence of functionally univentricular heart and pre-procedure severity of illness factors (e.g. pre-procedure mechanical ventilation or shock).

Table 2 Summary of risk factors by any morbidity outcome
The weight was missing in 87 (2·8%) procedures and weight for age was set to missing for an additional 99 (3·2%) patients in whom the value was >5 SD from the normative mean. Other data were complete. Values include imputed data.
Cardiac diagnosis group A) hypoplastic left heart syndrome, Truncus arteriosus, pulmonary atresia intact septum, B) Functionally univentricular heart, pulmonary atresia ventricular septal defect, C) transposition of the great arteries all types, interrupted aortic arch, totally anomalous pulmonary venous connection, D) patent ductus arteriosus, tricuspid valve anomalies, acquired heart disease, complete atrioventricular septal defect, E) Tetralogy of Fallot, mitral valve anomalies, isolated aortic stenosis, aortic regurgitation, aortic arch obstruction, subaortic obstruction, ventricular septal defect, atrial septal defect. Patient survival was highest and patient hospital stay was shortest for those with no morbidity (see Table 2). The 6-month survival was significantly lower for patients with a single morbidity, for those with 'ECLS' and for 'multi-morbidity' patients, compared with patients with no morbidity P < 0·001 for all. Table 2 Secondary outcomes of hospital stay and survival at 6 months by morbidity The number (%) of morbidities was based on individual surgical episode, as was the length of stay, whereas survival was calculated per patient. Data are presented based on morbidity in isolation, multi-morbidity and ECLS. Length of stay data was missing for 9 patients and Life status was missing in 16 patients who were not included.

Impact of morbidity Patient sample
We recruited 60% of eligible families, amounting to 666 children of whom 340 (51%) had at least one morbidity. Of these, 19 died within 30 days and 39 died within 6 months. We pair matched 558 patients and 108 patients unmatched of whom 61 had morbidity; all of the patients were included in the statistical analysis.

Quality of life and anxiety -depression outcomes
For the 6-weeks data a comparison between patients with missing versus non missing data indicated the only difference was in the proportions with severity of illness factors and additional cardiac risk factors. Importantly there was no difference in the proportion of patients with or without morbidity. At the 6 months mark there was no difference in any baseline factor or morbidity occurrence. All PedQL scores were lower than in a healthy population (see methods [13][14][15] and Table 3). In terms of our study objective to assess the impact of morbidity; the total PedsQL scores were significantly reduced for children with any morbidity at 6 weeks in comparison to children with no morbidity (case mix adjusted score reduction of -5·2 (95% CI -8·3, -2·2) P<0·01).
On a positive note, this difference narrowed to small non-significant reduction in total score by 6 months. Physical scores were affected more than psychosocial scores, and were still significantly impaired at 6-months.

Overview of findings
This paper presents the results of the first prospective multi-centre study of paediatric cardiac surgery morbidity over 6-months after operation. The results are novel and indicate that evaluation of morbidity adds considerably to the current lone metric of 30-day survival.
We note the excellent outcomes in children who had no morbidities: their survival at 6months post operation was 99·3% and their median length of stay was 8 days, suggesting that the morbidities we selected do capture most of the complication-related adverse outcomes for this context.
Based on our impact sub study, 'ECLS' and 'multiple morbidities' are particularly important adverse outcomes -they are associated with high mortality, high resource use and lower quality of life for surviving patients. Future routine monitoring and public reporting for these should be considered by centres and the national audit.
In our engagement work, patients' families consistently told us that outcomes affecting the child as a whole over the medium or longer term are very important to them. We thus included morbidities such as feeding problems and acute neurological events but note that those types of outcome were the most complicated to capture, and we intend to publish additional material on these topics.

Limitations
Definition, measurement, reporting and interpretation of morbidity, is much more complex than for mortality and requires significant staff resources and clinician buy in. Some morbidities are more challenging to measure and to interpret (see specific morbidities). The number of individual morbidities was low in some of our analyses limiting interpretation.

Context
There have been previous attempts at reporting of paediatric cardiac surgery morbidity outcomes: two single centre studies generated an aggregate 'Morbidity Index' by assigning subjective weights to post-operative complications 26 27 and the Society of Thoracic Surgery in the USA developed a 'Morbidity Score' based on data from their multi-centre registry. 28 Condensing diverse morbidities into a single score loses information and recent work on

Specific morbidities
Neurodevelopmental problems are common in children with CHD undergoing surgery 35 however we detected acute neurological events (ANE) in only 2·1% of patients: it appears that peri-operative 'ANE' represents the tip of the iceberg and hence the scope of surveillance for child neurodevelopment in CHD needs to extend well beyond the peri-operative period.
Despite their great important to families, we note that participants reported subjectivity in the collection of data on 'feeding problems'; hence these were difficult to capture consistently.
Additional research may help determine the best way to alleviate the impact of feeding problems in CHD.
Our measure of renal failure was the need for renal support, and although relatively easy data to collect this may not be the optimal method to capture this morbidity, given that there are differences in practice between clinicians. 36 Further research may help us to understand the best approach to manage post-operative renal injury in CHD.

Feeding back to stakeholders
We used study data collected during the course of the morbidity project, to co-develop parent and carer information resources, showing what the morbidities are and how their incidence and the length of stay may vary based on the complexity of the child's condition. Parents told us that it helps to know that; firstly they are not alone in facing a complication, secondly clinical teams have seen complications before and know how to deal with them, and thirdly that it is better as a parent to 'be prepared'. Furthermore, parents indicated that information Within the scope of this project, a new Excel tool has been developed and piloted, which enables clinical teams to benchmark and report the local rates of morbidities with a quality assurance goal, such as in a mortality and morbidity conference. Furthermore, we have kept in close contact with the National Congenital Heart Diseases Audit (NCHDA) and the Clinical Reference Group for CHD services. NCHDA has already started to collect five of the nine morbidities within the nationally mandated dataset, using the definitions that we developed.  Figure 1 Incidence of morbidities The figure shows the distinction between occurrence as a single morbidity and occurrence in combination with other morbidities, with 95% confidence intervals.

Abstract
Objectives: Given the current excellent early mortality rates for paediatric cardiac surgery, stakeholders believe that this important safety outcome should be supplemented by a wider range of measures. Our objectives were to prospectively measure the incidence of morbidities following paediatric cardiac surgery and to evaluate their clinical and health-economic impact over 6-months.

Design:
The design was a prospective, multi-centre, multidisciplinary mixed methods study.

Setting:
The setting was five of the ten paediatric cardiac surgery centres in the United Kingdom with 21 months recruitment.

Participants:
The participants were children under the age of 17 years with heart diseases.

Results:
Families and clinicians prioritized: Acute neurological event, unplanned re-intervention, feeding problems, renal replacement therapy, major adverse events, extracorporeal life support, necrotising enterocolitis, postsurgical infection, and prolonged pleural effusion or chylothorax.
Follow up for impact of morbidity amongst 666 patients over 6 months indicated that:  Morbidity related impairment in health related quality of life improved between 6 weeks and 6 months.

Conclusions:
Evaluation of post-operative morbidity is more complicated than measuring early mortality.
However, tracking morbidity after paediatric cardiac surgery over 6 months offers stakeholders important data that is of value to parents and will be useful in driving future quality improvement.

Strengths and limitations of this study
 The study prospectively measured the incidence of defined morbidities that are considered to be important by both lay people and clinicians, in children undergoing heart surgery across five UK centres.  Based on data collected, the study explored the risk factors for paediatric cardiac surgery morbidity statistically.  Within a subset of patients, the study explored the impact of the selected morbidities on patients and families over 6 months, in terms of quality of life, days spent at home rather than in hospital and hospital costs.

Study registration
The study has ethical approval from London City Road Research Ethics Committee (14-LO-1442).

Conflicts of interest
None declared by authors.

Data sharing
For data to be made available specific ethical and HRA approval would be required.

Background
Over 5000 paediatric cardiac surgery procedures are performed in the UK each year and early survival has improved to over 98% since comprehensive national audit commenced in 2000. 1 2 Most stakeholders including clinicians, commissioners and users believe that whilst these early survival rates remain an important safety measure, it is has become imperative to explore a broader range of measures for outcome in this complex field of practice. Whilst there has been considerable research related to measuring, understanding and reducing perioperative mortality for paediatric cardiac surgery, 3-5 much less attention has been focussed on surgical morbidities.
Morbidity is defined as a state of being unhealthy, or of experiencing an aspect of health that is 'generally bad for you'. In this project, by 'morbidity' we mean a defined aspect of ill health associated with a specific operation. In our study, we aimed to identify which morbidities present the greatest burden on patients and health services following paediatric cardiac surgery. Views may differ between professionals and non-professionals as to what the term morbidity exactly refers to and which surgical morbidities are most important. 6 Therefore we set out to combine patient and carers' perspectives with those of professional groups in defining a prioritised list of morbidities for prospective evaluation.
Over four years we: 1. Identified and defined nine morbidities following paediatric heart surgery, taking into account views from patients, carers, psychologists, nurses and clinicians, which together captured important aspects of the clinical and health-economic burden; 2. Measured incidence of the defined morbidities in the UK patient population and in subgroups defined by case complexity; 3. Evaluated the impact of defined morbidities on quality of life and estimated their clinical and health economic burden.

Design
The design was a prospective, multi-centre, multidisciplinary mixed methods study. Selection of morbidities was based on consensus methods. Morbidity incidence was evaluated with a prospective cross sectional study. Measurement of morbidity impact entailed a prospective, case-matched, longitudinal study.

Setting
Five UK paediatric cardiac surgery centres representing a range of programme sizes, which care for half of all patients nationally.

Participants
The participants were children under 17 years old with congenital heart disease (CHD).

Which morbidities?
Between January 2014 and September 2015, we reviewed existing literature, ran three family focus groups and undertook a family online discussion forum moderated by the Children's Heart Federation (a user group). Transcripts were thematically analysed and the resultant themes, together with the literature, helped to inform a long list of candidate surgical morbidities. A multi-disciplinary group, with patient and carer representation, then ranked and selected a list of nine key morbidities using the nominal group technique and secret voting. 7 This 'selection panel' 7 was informed in turn by clinical views on definitions and feasibility of routine monitoring for each candidate morbidity provided by an independent 'definition panel' 8 as reported in previous publications.

Incidence of morbidity
Between October 2015 and June 2017, we prospectively measured instances of morbidity within all consecutive surgical admissions across the study sites. Appendix 1 8 provides the details of the criteria followed to define each individual morbidity, including the timeframe for diagnosis. Morbidities were attributed to the immediately preceding cardiac surgery and defined within the same hospitalisation. The only exceptions were unplanned re operation, which was defined as an unanticipated procedure within 30-days and mediastinitis, which could be identified post-discharge by the operating surgeon. 8 Data were regularly checked for completeness, clinical congruence and accuracy and, a three month sample of data was independently validated against the national audit data. As for the UK audit of 30-day mortality, 9 all procedures more than 30 days apart on the same patient, were included in the morbidity analysis. As a secondary outcome, we checked each patient's survival status at 6 months after first appearance in the dataset, based on the individual centres National Congenital Heart Diseases Audit (NCHDA) 2 data and local hospital records.

Impact of morbidity Recruitment
Recruitment to the impact study ran between October 2015 and June 2017 for all but one site, where it stopped after 6 months due to resource constraints. We attempted to recruit all patients with morbidity from the wider population, when at least one parent spoke reasonable English, and the family were resident in the UK. When feasible, for each morbidity case, we recruited a patient with no morbidity, matched on treating centre, age and univentricular status. The recruitment strategy led to the recruitment of a sample that was evenly balanced between patients with morbidity and those with no morbidity, however we were not able to find a match for every morbidity-patient. In order to account for the widest possible spectrum of important outcomes we included the available data on children who subsequently died.

Measures of impact
We evaluated the impact of morbidities over the 6 months following surgery, based on the following outcomes:

a) Quality of life and psychological burden on children and parents
We used age-specific measures at 6 weeks and 6 months following surgery. These were the PedsQL subscales (physical and psychosocial) and total scores. These scores range from 0 and 100, a higher score indicating better quality of life. PedsQL scores in a normal healthy population vary by age; expected scores for infants 0-12 months (which encompasses the median age for our cohort) are mean physical 84·98 (SD 9·45), mean psychosocial 80·47 (SD 12·64) and mean total 82·47 (SD 9·95). [10][11][12] For the parents, we used the PHQ-4, which comprises four questions, measuring anxiety and depression in adults. Individual items are scored from zero to three. Scores ≥three for the first two questions suggest anxiety, and scores ≥three for last two questions suggest depression. 13 In the normal population, 4·8% have scores suggestive of anxiety and 6·6% have PHQ-4 scores suggestive of depression. 14

b) Days at home over 6 months as an additional measure of disruption to family life
Length of inpatient stay following the index procedure was defined as the number of days between the operation and the date of discharge from the tertiary centre. Hospital stays that exceeded the 6-month follow-up period were censored after 183 days. All subsequent hospital admissions were captured and the total days spent in hospital including tertiary and secondary care admissions was subtracted from 183 days to generate a value for days at home by 6 months. Children who died in hospital were assigned a value of zero for this outcome.

c) Costs of the index hospitalisation
We calculated the cost of the inpatient stay in the tertiary centre following the index surgical procedure as a key indicator of the economic burden to the hospital provider (measured in 2016/17 UK£).
Costs were calculated as the summation of days in intensive care unit (ICU), days on the ward, and all operating procedures within the defined period. For ICU stays, data were recorded on the level of care the patient received each day (Enhanced Care, High Dependency, High Dependency Advanced, ICU Basic, ICU Basic Enhanced, ICU Advanced, ICU Advanced Enhanced and ECLS). 15 Unit costs of ICU stays were daily costs for each level of care (ranging from £870 to £5440) applied to the number of days spent receiving that level of care. Unit costs per day for ward stays were obtained from the highest recruiting centre and varied according to the age of the patient (<two years, £904; ≥two years, £680).
Unit costs for surgical procedures were costs per minute supplied by study sites; these varied by site and so we used the value for the site that recruited the largest number of patients to this study (£13·39 per minute), which was similar to the mean value across all sites.

Patient sample size
We anticipated that between 3000 and 3300 surgical patients would be admitted across the five sites during the study period. 2 We assumed that a clinically relevant difference in quality of life between matched pairs corresponded to a mean difference in quality of life score of at least 0.5 standard deviations. 16 To detect such a difference for PedsQL responses at 5% significance with 80% power requires a minimum of 32 matched pairs. Allowing for a 10% loss to follow-up rate, we aimed to recruit 36 matched pairs for each patient with morbidity, giving 80% power to detect a significant effect for any morbidity with a prevalence of at least 1·5%.

Analysis of risk factors for morbidity
Based on the whole sample of cardiac surgeries across the five sites over the study period, we explored risk factors for occurrence of morbidity. Clinical risk factor groups were derived from the finer diagnostic coding, based on previous peer reviewed research by our group. 9 17 We used multilevel logistic regression analysis to explore the role of pre-operative, patientlevel case mix factors on the occurrence of any morbidity versus no morbidity, accounting for multiple procedures within patients. Firstly a univariate model to predict risk of any morbidity vs none was fitted for each risk factor. The estimated Odds ratios (OR) are presented along with 95% confidence intervals (CI). Then all factors significant on univariate analysis (p<0·1) were considered in a multivariable model. We used multiple imputation by chained equations to account for missing data. The imputation model included all risk factors considered in the univariate analysis, which we assumed included all predictors of missingness. We indicate missing data in our results. The multivariable models were derived by fitting a regression model for all significant predictors and estimates were combined using Rubin's rules. 18

Secondary outcome
The secondary outcome of survival at 6 months was calculated for all patients who had cardiac surgery at the sites within the study period. Unadjusted 6-month mortality rates were compared between morbidity groups using logistic regression.

Analysis of the impact of morbidity
In the patients recruited to the impact study, we analysed the 'impact outcomes' over 6 months post-surgery.
We used mixed effects regression models for PedsQL/PHQ-4 results to compare the impact of morbidity (any versus none) on outcome. All models were adjusted for clustering within matched pairs and significant covariates associated with incidence of morbidity. We used multiple imputation, by chained equations to account for missing data for those patients known to be alive at 6 weeks and 6 months respectively.
We used quantile regression to estimate the effects of individual morbidities in terms of differences in median days at home by 6 months, since the measure was negatively skewed.
All models were adjusted for clustering within matched pairs and significant covariates associated with incidence of morbidity. Given the high level of data completeness, imputation was not used. The inclusion of hospital site made negligible difference to either of these analyses, and not included in the statistical models.
We modelled the relationship between morbidities and cost of index hospitalisation, adjusted for significant covariates associated with hospital costs. Costs were skewed and to account for this we used a generalised linear model (GLM) with gamma family and log link, which has been recommended for modelling positively skewed data. 19 Missing outcome values were imputed using an iterative Markov chain Monte Carlo procedure based on multivariate normal regression, generating 20 imputed datasets. The hospital site was included as a significant covariate in the health economic model.
All analyses were performed in Stata 14. 20

Patient and public involvement
The patient or user perspective lies at the core of the study methodology. Specifically, a key goal of the study was to consider the views of patients and parents when measures of morbidity are selected for future audit and benchmarking, in particular since emphasis may potentially differ between professionals and parents / patients. Patient and family representatives from the Children's Heart Federation (CHF) were involved in aspects of the study design including design and running of the focus groups and the online discussion forum.
Representatives of the CHF participated in the facilitated nominal group meetings to select morbidities for inclusion in the incidence and impact studies.
A member of the CHF and a parent representative sat on the project steering group.

1: Which morbidities?
Our final selected 7 and defined 8 list of included morbidities were: Acute neurological event, unplanned re-intervention, feeding problems, renal replacement therapy, major adverse events, extracorporeal life support (ECLS), necrotising enterocolitis, post-surgical infection and prolonged pleural effusion or chylothorax.
We treated children with more than one of these events as a distinct group referred to as 'multi-morbidity'. Recognising ECLS as a very severe event, which nearly always occurs with other morbidities, we treated ECLS as standalone morbidity (and always excluded from the multiple morbidity group). [21][22][23][24][25] There was some divergence between the views of clinicians and families about the fundamental issue of what the important morbidities linked to paediatric cardiac surgery are.
Health professionals from tertiary centres tended to prioritise clearly clinical issues related to the heart (e.g. use of ECLS and re operation) whereas parents placed greater emphasis on holistic outcomes for their child (e.g. feeding difficulties and child developmental problems). 7

Incidence of morbidity Descriptive results
We collected data on 3248 cardiac procedures and analysed 3090, after excluding 99 cardiac re-operations (a morbidity outcome) and 59 procedures that did not meet inclusion criteria.
The incidence of individual morbidities in isolation and overall for each is shown in Figure 1.

Figure 1 Incidence of morbidities
The figure shows the distinction between occurrence as a single morbidity and occurrence in combination with other morbidities, with 95% confidence intervals.

Significant risk factors for morbidity
As shown in Table 1, the most important risk factor for occurrence of 'any morbidity' was young age, particularly neonates, who were approximately five times as likely to experience a morbidity compared to children over one year old. Children with more complex heart disease or children whose operation included a bypass time longer than 90 minutes were almost twice as likely to experience morbidity. Other important independent risk factors were: palliative or staged procedures, the presence of functionally univentricular heart and pre-procedure severity of illness factors (e.g. pre-procedure mechanical ventilation or shock).

Table 1Summary of risk factors by any morbidity outcome
The weight was missing in 87 (2·8%) procedures and weight for age was set to missing for an additional 99 (3·2%) patients in whom the value was >5 SD from the normative mean. Other data were complete. Values include imputed data.
The results of the multiple logistic regression model for occurrence of 'any morbidity', are expressed as odds ratio for 'any morbidity' by the stated factor, adjusted for age band, low weight, cardiac diagnostic category, functionally univentricular heart, acquired comorbidity, congenital comorbidity, severity of illness indicators, additional cardiac risk factors, specific procedure group, bypass time category.

Secondary outcomes
Patient survival was highest and patient hospital stay was shortest for patients with none of the morbidities (see Table 2). The 6-month survival was significantly lower for patients with a single morbidity, for those with ECLS and multiple morbidities, compared with patients with none of the morbidities P < 0·001 for all. Table 2 Secondary outcomes of hospital stay and survival at 6 months by morbidity The number (%) of morbidities was based on individual surgical episode, as was the length of stay, whereas survival was calculated per patient. Data are presented based on morbidity in isolation, multi-morbidity and ECLS. Length of stay data was missing for 9 patients and Life status was missing in 16 patients who were not included.

Impact of morbidity Patient sample
We recruited 60% of eligible families, amounting to 666 children of whom 340 (51%) had at least one morbidity. Of these, 19 died within 30 days and 39 died within 6 months. We pair matched 558 patients and 108 patients unmatched of whom 61 had morbidity; all of the patients were included in the statistical analysis.

Quality of life and anxiety -depression outcomes
For the 6-weeks data, a comparison between patients with missing versus non missing data indicated the only difference was in the proportions with severity of illness factors and additional cardiac risk factors. Importantly there was no difference in the proportion of patients with or without morbidity. At the 6 months mark there was no difference in any baseline factor or morbidity occurrence.
All PedQL scores were lower than in a healthy population (see methods [10][11][12] and Table 3). In terms of our study objective to assess the impact of morbidity; the total PedsQL scores were significantly reduced for children with any morbidity at 6 weeks in comparison to children with none of the morbidities (case mix adjusted score reduction of -5·2 (95% CI -8·3, -2·2) P<0·01). On a positive note, this difference narrowed to small non-significant reduction in total score by 6 months. Physical scores were affected more than psychosocial scores, and were still significantly impaired at 6-months.
All parents experienced higher rates of both anxiety and depression than a healthy population (see methods and Table 3). The parents of children with morbidity were around 57% more likely to experience anxiety; and around 77% more likely to experience depression, at 6 weeks post operation than parents of children without a morbidity. PHQ-4 scores had improved by 6 months, and although there remained a higher chance of both anxiety and depression with morbidity, the difference narrowed and was not significant.  Parental anxiety and depression was derived from PHQ-4 responses of 481 parents at 6 weeks (26% missing) and 394 parents at 6 months (37% missing).
Values include imputed data.
The differences in each outcome are adjusted for age band, low weight, cardiac diagnostic category, functionally univentricular heart, acquired comorbidity, congenital comorbidity, severity of illness indicators, additional cardiac risk factors, specific procedure group, bypass time category. Significantly increased (P<0·05) costs were also found for unplanned re-intervention, feeding problems, renal support and prolonged pleural effusion. The remaining morbidities, representing the lowest number of patients per category, showed non-significant differences. We note the excellent outcomes in children who had none of the morbidities: their survival at 6-months post-operation was 99·3% and their median length of stay was 8 days, suggesting that the morbidities we selected do capture most of the complication-related adverse outcomes for this context.
Based on our impact sub study, occurrence of ECLS and multiple morbidities are particularly important adverse outcomes -they are associated with high mortality, high resource use and lower quality of life for surviving patients over 6 months after surgery. Future routine monitoring and public reporting for these should be considered by centres and the national audit.
In our engagement work, patients' families consistently told us that outcomes affecting the child as a whole over the medium or longer term are very important to them. We thus included morbidities such as feeding problems and acute neurological events but note that those types of outcome were the most complicated to capture, and we intend to publish additional material on these topics.

Limitations
We acknowledge that the definition, measurement, reporting and interpretation of morbidity, is much more complex than for mortality. This activity requires significant staff resources and clinician buy in, in particular because some morbidities are challenging to measure and to interpret. The number of individual morbidities was low in some of our analyses, limiting interpretation. We have not captured every single morbidity that exists after children's heart surgery.

Context
There have been previous attempts at reporting morbidity after paediatric cardiac surgery, however these have in general captured events within a hospital setting, whereas our study captures outcomes out to 6 months post-surgery. Two single-centre studies generated an aggregate 'Morbidity Index' by assigning subjective weights to post-operative complications. 26 27 The Society of Thoracic Surgery in the USA, developed a 'Morbidity Score', based on data from their multi-centre registry. 28 We note that condensing diverse morbidities into a single score may lead to loss of information. Moreover, recent work on using graphical methods to routinely monitor a range of morbidities 29 highlighted the complexity of graphically summarising multiple morbidities. 30 The Pediatric Cardiac Critical Care Consortium (PC⁴) was set up in 2009, and provides partner sites with access to real-time,

Future practice and research
Neurodevelopmental problems are common in children with CHD undergoing surgery 35 however we detected acute neurological events (ANE) following only 2·1% of procedures. : Therefore it appears that peri-operative 'ANE' represents the tip of the iceberg, and hence the scope of surveillance for child neurodevelopment in CHD needs to extend well beyond the peri-operative period.
Despite their great importance to families, we note that participants reported subjectivity in the collection of data on 'feeding problems'; hence these were difficult to capture consistently. Additional research may help determine the best way to measure this important morbidity and to elucidate approaches to alleviate the impact of feeding problems in CHD.
Our measure of renal failure was the need for renal support, and although relatively easy data to collect this may not be the optimal method to capture this morbidity, given that there are differences in practice between clinicians. 36 Further research may help us to understand the best approach to manage post-operative renal injury in CHD.
Our results perhaps unsurprisingly indicate that parents of children who suffer morbidities experience higher rates of anxiety and depression post-operatively. This emphasises the importance of supporting parents during this phase.
We used study data to co-develop parent and carer information resources, showing what the morbidities are and how their incidence and the length of stay may vary based on the complexity of the child's condition. Parents told us that it helps to know that; firstly they are not alone in facing a morbidity, secondly clinical teams have seen morbidities before and know how to deal with them, and thirdly that it is better as a parent to 'be prepared'.
Furthermore, parents indicated that information about impact such as nearly all children who experience a morbidity and recover will have a similar quality of life to children who did not experience a morbidity by the six month mark, was very helpful to know.
Within the scope of this project, a new Excel tool has been developed and piloted, which enables clinical teams to benchmark and report the local rates of morbidities with a quality Includes new abnormality in any of the following: -Electroencephalogram.
-Brain scan (either computerised tomography or magnetic resonance).
-Clinical evaluation (Seizures or movement disorder, focal neurological signs, generalised neurological signs, altered conscious level including even brain death.) The treatment protocol is variable depending on the type of neuromorbidity.
Specialist consultation with a neurologist, a full evaluation of any brain injury and neurodevelopmental follow up would be a minimum.

Feeding problems
A diagnosis of postoperative feeding problems should be considered during recovery after surgery and prior to discharge from the specialist centre either to home or to secondary care if A child may fail to feed normally following paediatric cardiac surgery for a range of reasons including gastrooesophageal reflux, vocal cord paralysis, oral-motor dysfunction, oral aversion, and neurologic impairment. (2) If for any of these reasons a child is not able to orally The requirement for any feeding support.
Includes via the intravenous route or via an enteral tube. the child is unable to feed normally. The goal is detection of feeding problems which are new post-surgery, and it is recognised that this may be challenging where a child was not fed pre-operatively for cardiac reasons since feeding ability will not have been assessed objectively.
feed or completely orally feed and is tube dependent at discharge from the tertiary centre or at 30 days (if he or she is otherwise clinically stable enough to feed at that time point), then a post-operative feeding problem will be diagnosed.
Excludes feeding support that was present to treat a primary problem diagnosed before the surgery, feeding support related to an episode of necrotising enterocolitis, and feeding support because the child dislikes a special diet.
feeding should be monitored by the clinical care team responsible at each stage of the journey.

Need for renal replacement therapy
Includes renal replacement therapy when initiated as a new support at any time from the start of the postoperative admission to ICU up until 30 days following the primary operation.
The child requires renal replacement therapy (peritoneal dialysis or haemofiltration) for renal failure (oligoanuria of less than 0.5 ml/kg/hour and elevated creatinine level for age) and or fluid overload. In patients where renal support is required alongside extracorporeal life support, the primary morbidity is viewed as extracorporeal life support.
The measurement protocol is simply the presence of (new) renal support. (Excludes renal support on extracorporeal life support). Data on renal biochemistry and urine output will be collected.
Instigation of effective renal replacement therapy.
If recovery of kidney function does not occur within 3 to 5 weeks then consultation with paediatric renal

Major adverse cardiac events or never events
Events within this morbidity may be identified during the tertiary hospital stay (either ward or ICU) following the primary surgery.
These morbidity includes: -Cardiac arrest, where the child receives any chest compressions or defibrillation.
-Chest re-opening on the ICU or ward for any reason.
-Major haemorrhage in the ICU following surgery.
-A 'Never Event' applicable to paediatric cardiac surgery as selected from the 'Never Events' list published for NHS for 2015 (Including wrong site or wrong patient surgery, wrong prosthesis surgery, retained foreign object post procedure, wrong route administration of medication, transfusion or transplantation of main red cell group incompatible blood components or organs, misplaced naso-gastric or oro-gastric tubes, Major haemorrhage is defined as bleeding > 10ml/kg/hr on ICU for 2 consecutive hours.
A 'Never Event' includes the events listed plus harm to the patient, for example: if a naso-gastric tube is misplaced, detected and removed in a timely manner before any harm is done then this is not a 'Never Event'. Conversely, if the misplaced naso-gastric tube is not noted, and feed is given into the bronchus, then this is a 'Never Event'.
All events will results in immediate treatment as part of current practice. bleeding but no radiological changes are included, if a general surgery specialist has seen the child and commenced a course of intravenous antibiotics and parenteral nutrition for five to seven days. Cases of severe necrotising enterocolitis with radiological signs systemic instability and bowel perforation are also included.
treatments deployed, thus enabling the necrotising enterocolitis diagnosis to be graded between 1a and 3b. [25] nutrition, radiological investigation and surgical intervention.

Surgical site infection and bloodstream infection
Surgical site and blood stream infections diagnosed within the hospital admission following surgery or following readmission to the same unit during postoperative recovery, where the treating clinical team assesses the infection to be linked to the recent operation. It is noted that mediastinitis may be detected more than 30 days after cardiac surgery (6) hence this time cut off is not applicable.
Deep surgical site Infection and/or mediastinitis includes any infection of an incised wound that undergoes any re intervention by a surgeon (such as opening of the wound, vacuum dressing), mediastinitis and false aneurysm, independent of culture positivity. [23] Blood stream infection includes both catheter related and non-catheter related. prolonged antibiotic therapy.

Prolonged pleural effusion or chylothorax
Prolonged pleural effusion is a post procedural effusion with duration greater than ten days. Chylothorax is diagnosed from after surgery until discharge from the tertiary hospital.
Either a chylous pleural effusion or significant chylous pericardial effusion or significant chylous ascites or a prolonged non-chylous effusion that necessitates thoracic drainage at least ten days following index cardiac surgery.
Chylous effusions are characterised by milky appearance and a pleural fluid white blood cell count of greater than 1000 cells/μl with lymphocytes greater than 80%. (7) If the child is on normal feeds the triglyceride level in the pleural fluid will be > 1.1 mmol/L or the ratio between the pleural triglyceride level and the serum triglyceride level will exceed 1.

Abstract
Objectives: Given the current excellent early mortality rates for paediatric cardiac surgery, stakeholders believe that this important safety outcome should be supplemented by a wider range of measures. Our objectives were to prospectively measure the incidence of morbidities following paediatric cardiac surgery and to evaluate their clinical and health-economic impact over 6-months.

Design:
The design was a prospective, multi-centre, multidisciplinary mixed methods study.

Setting:
The setting was five of the ten paediatric cardiac surgery centres in the United Kingdom with 21 months recruitment.

Participants:
Included were 3090 paediatric cardiac surgeries, of which 666 patients were recruited to an impact sub study.

Conclusions:
Evaluation of post-operative morbidity is more complicated than measuring early mortality.
However, tracking morbidity after paediatric cardiac surgery over 6 months offers stakeholders important data that is of value to parents and will be useful in driving future quality improvement.

Strengths and limitations of this study
 Our study is unique, given that the morbidity measures selected for study reflect the viewpoints of both lay people and clinicians, whereas previous work has tended to focus on clinical metrics only.  A limitation of our study is that although we were able to identify broad risk factors for morbidity after paediatric cardiac surgery, a larger dataset will be needed to generate more reliable risk adjustment methods.  A strength of our study was that it was multi-centre, prospective, and followed children up for 6 months after their operation, which is a longer time horizon than the usual 30-day period.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Study registration
The study has ethical approval from London City Road Research Ethics Committee (14-LO-1442). The funder had no role in the study.

Conflicts of interest
None declared by authors.

Background
Over 5000 paediatric cardiac surgery procedures are performed in the UK each year and early survival has improved to over 98% since comprehensive national audit commenced in 2000. 1 2 Most stakeholders including clinicians, commissioners and users believe that whilst these early survival rates remain an important safety measure, it is has become imperative to explore a broader range of measures for outcome in this complex field of practice. Whilst there has been considerable research related to measuring, understanding and reducing perioperative mortality for paediatric cardiac surgery, 3-5 much less attention has been focussed on surgical morbidities.
Morbidity is defined as a state of being unhealthy, or of experiencing an aspect of health that is 'generally bad for you'. In this project, by 'morbidity' we mean a defined aspect of ill health associated with a specific operation. In our study, we aimed to identify which morbidities present the greatest burden on patients and health services following paediatric cardiac surgery. Views may differ between professionals and non-professionals as to what the term morbidity exactly refers to and which surgical morbidities are most important. 6 Therefore we set out to combine patient and carers' perspectives with those of professional groups in defining a prioritised list of morbidities for prospective evaluation.
Over four years we: 1. Identified and defined nine morbidities following paediatric heart surgery, taking into account views from patients, carers, psychologists, nurses and clinicians, which together captured important aspects of the clinical and health-economic burden; 2. Measured incidence of the defined morbidities in the UK patient population and in subgroups defined by case complexity; 3. Evaluated the impact of defined morbidities on quality of life and estimated their clinical and health economic burden.

Design
The design was a prospective, multi-centre, multidisciplinary mixed methods study. Selection of morbidities was based on consensus methods. Morbidity incidence was evaluated with a prospective cross sectional study. Measurement of morbidity impact entailed a prospective, case-matched, longitudinal study.

Setting
Five UK paediatric cardiac surgery centres representing a range of programme sizes, which care for half of all patients nationally.

Participants
The participants were children under 17 years old with congenital heart disease (CHD).

Consent
Parents and children participating in focus groups and in the prospective study of morbidity impact provided written informed consent.

Which morbidities?
Between January 2014 and September 2015, we reviewed existing literature, ran three family focus groups and undertook a family online discussion forum moderated by the Children's Heart Federation (a user group). Transcripts were thematically analysed and the resultant themes, together with the literature, helped to inform a long list of candidate surgical morbidities. A multi-disciplinary group, with patient and carer representation, then ranked and selected a list of nine key morbidities using the nominal group technique and secret voting. 7 This 'selection panel' 7 was informed in turn by clinical views on definitions and feasibility of routine monitoring for each candidate morbidity provided by an independent 'definition panel' 8 as reported in previous publications.

Incidence of morbidity
Between October 2015 and June 2017, we prospectively measured instances of morbidity within all consecutive surgical admissions across the study sites. Appendix 1 8 provides the details of the criteria followed to define each individual morbidity, including the timeframe for diagnosis. Morbidities were attributed to the immediately preceding cardiac surgery and defined within the same hospitalisation. The only exceptions were unplanned re operation, which was defined as an unanticipated procedure within 30-days and mediastinitis, which could be identified post-discharge by the operating surgeon. 8 Data were regularly checked for completeness, clinical congruence and accuracy and, a three month sample of data was independently validated against the national audit data.

Impact of morbidity Recruitment
Recruitment to the impact study ran between October 2015 and June 2017 for all but one site, where it stopped after 6 months due to resource constraints. We attempted to recruit all patients with morbidity from the wider population, when at least one parent spoke reasonable English, and the family were resident in the UK. When feasible, for each morbidity case, we recruited a patient with no morbidity, matched on treating centre, age and univentricular status. The recruitment strategy led to the recruitment of a sample that was evenly balanced between patients with morbidity and those with no morbidity, however we were not able to find a match for every morbidity-patient. In order to account for the widest possible spectrum of important outcomes we included the available data on children who subsequently died.

Measures of impact
We evaluated the impact of morbidities over the 6 months following surgery, based on the following outcomes:

a) Quality of life and psychological burden on children and parents
We used age-specific measures at 6 weeks and 6 months following surgery. These were the PedsQL subscales (physical and psychosocial) and total scores. These scores range from 0 and 100, a higher score indicating better quality of life. PedsQL scores in a normal healthy population vary by age; expected scores for infants 0-12 months (which encompasses the median age for our cohort) are mean physical 84·98 (SD 9·45), mean psychosocial 80·47 (SD 12·64) and mean total 82·47 (SD 9·95). [10][11][12] For the parents, we used the PHQ-4, which comprises four questions, measuring anxiety and depression in adults. Individual items are scored from zero to three. Scores ≥three for the first two questions suggest anxiety, and scores ≥three for last two questions suggest depression. 13 In the normal population, 4·8% have scores suggestive of anxiety and 6·6% have PHQ-4 scores suggestive of depression. 14

b) Days at home over 6 months as an additional measure of disruption to family life
Length of inpatient stay following the index procedure was defined as the number of days between the operation and the date of discharge from the tertiary centre. Hospital stays that exceeded the 6-month follow-up period were censored after 183 days. All subsequent hospital admissions were captured and the total days spent in hospital including tertiary and secondary care admissions was subtracted from 183 days to generate a value for days at home by 6 months. Children who died in hospital were assigned a value of zero for this outcome.

c) Costs of the index hospitalisation
We calculated the cost of the inpatient stay in the tertiary centre following the index surgical procedure as a key indicator of the economic burden to the hospital provider (measured in 2016/17 UK£).
Costs were calculated as the summation of days in intensive care unit (ICU), days on the ward, and all operating procedures within the defined period. For ICU stays, data were recorded on the level of care the patient received each day (Enhanced Care, High Dependency, High Dependency Advanced, ICU Basic, ICU Basic Enhanced, ICU Advanced, ICU Advanced Enhanced and ECLS). 15 Unit costs of ICU stays were daily costs for each level of care (ranging from £870 to £5440) applied to the number of days spent receiving that level of care. Unit costs per day for ward stays were obtained from the highest recruiting centre and varied according to the age of the patient (<two years, £904; ≥two years, £680).
Unit costs for surgical procedures were costs per minute supplied by study sites; these varied by site and so we used the value for the site that recruited the largest number of patients to this study (£13·39 per minute), which was similar to the mean value across all sites.

Patient sample size
We anticipated that between 3000 and 3300 surgical patients would be admitted across the five sites during the study period. 2 We assumed that a clinically relevant difference in quality of life between matched pairs corresponded to a mean difference in quality of life score of at least 0.5 standard deviations. 16 To detect such a difference for PedsQL responses at 5% significance with 80% power requires a minimum of 32 matched pairs. Allowing for a 10% loss to follow-up rate, we aimed to recruit 36 matched pairs for each patient with morbidity, giving 80% power to detect a significant effect for any morbidity with a prevalence of at least 1·5%.

Analysis of risk factors for morbidity
Based on the whole sample of cardiac surgeries across the five sites over the study period, we explored risk factors for occurrence of morbidity. Clinical risk factor groups were derived from the finer diagnostic coding, based on previous peer reviewed research by our group. 9 17  We used multilevel logistic regression analysis to explore the role of pre-operative, patientlevel case mix factors on the occurrence of any morbidity versus no morbidity, accounting for multiple procedures within patients. Firstly a univariate model to predict risk of any morbidity vs none was fitted for each risk factor. The estimated Odds ratios (OR) are presented along with 95% confidence intervals (CI). Then all factors significant on univariate analysis (p<0·1) were considered in a multivariable model. We used multiple imputation by chained equations to account for missing data. The imputation model included all risk factors considered in the univariate analysis, which we assumed included all predictors of missingness. We indicate missing data in our results. The multivariable models were derived by fitting a regression model for all significant predictors and estimates were combined using Rubin's rules. 18

Secondary outcome
The secondary outcome of survival at 6 months was calculated for all patients who had cardiac surgery at the sites within the study period. Unadjusted 6-month mortality rates were compared between morbidity groups using logistic regression.

Analysis of the impact of morbidity
In the patients recruited to the impact study, we analysed the 'impact outcomes' over 6 months post-surgery.
We used mixed effects regression models for PedsQL/PHQ-4 results to compare the impact of morbidity (any versus none) on outcome. All models were adjusted for clustering within matched pairs and significant covariates associated with incidence of morbidity. We used multiple imputation, by chained equations to account for missing data for those patients known to be alive at 6 weeks and 6 months respectively.
We used quantile regression to estimate the effects of individual morbidities in terms of differences in median days at home by 6 months, since the measure was negatively skewed.
All models were adjusted for clustering within matched pairs and significant covariates associated with incidence of morbidity. Given the high level of data completeness, imputation was not used.
The inclusion of hospital site made negligible difference to either of these analyses, and not included in the statistical models.  11 We modelled the relationship between morbidities and cost of index hospitalisation, adjusted for significant covariates associated with hospital costs. Costs were skewed and to account for this we used a generalised linear model (GLM) with gamma family and log link, which has been recommended for modelling positively skewed data. 19 Missing outcome values were imputed using an iterative Markov chain Monte Carlo procedure based on multivariate normal regression, generating 20 imputed datasets. The hospital site was included as a significant covariate in the health economic model.
All analyses were performed in Stata 14. 20

Patient and public involvement
The patient or user perspective lies at the core of the study methodology. Specifically, a key goal of the study was to consider the views of patients and parents when measures of morbidity are selected for future audit and benchmarking, in particular since emphasis may potentially differ between professionals and parents / patients. Patient and family representatives from the Children's Heart Federation (CHF) were involved in aspects of the study design including design and running of the focus groups and the online discussion forum.
Representatives of the CHF participated in the facilitated nominal group meetings to select morbidities for inclusion in the incidence and impact studies.
A member of the CHF and a parent representative sat on the project steering group.

1: Which morbidities?
Our final selected 7 and defined 8 list of included morbidities were: Acute neurological event, unplanned re-intervention, feeding problems, renal replacement therapy, major adverse events, extracorporeal life support (ECLS), necrotising enterocolitis, post-surgical infection and prolonged pleural effusion or chylothorax.
We treated children with more than one of these events as a distinct group referred to as 'multi-morbidity'. Recognising ECLS as a very severe event, which nearly always occurs with other morbidities, we treated ECLS as standalone morbidity (and always excluded from the multiple morbidity group). [21][22][23][24][25]  Health professionals from tertiary centres tended to prioritise clearly clinical issues related to the heart (e.g. use of ECLS and re operation) whereas parents placed greater emphasis on holistic outcomes for their child (e.g. feeding difficulties and child developmental problems). 7

Incidence of morbidity Descriptive results
We collected data on 3248 cardiac procedures and analysed 3090, after excluding 99 cardiac re-operations (a morbidity outcome) and 59 procedures that did not meet inclusion criteria.
The incidence of individual morbidities in isolation and overall for each is shown in Figure 1.

Figure 1 Incidence of morbidities
The figure shows the distinction between occurrence as a single morbidity and occurrence in combination with other morbidities, with 95% confidence intervals.

Significant risk factors for morbidity
As shown in Table 1, the most important risk factor for occurrence of 'any morbidity' was young age, particularly neonates, who were approximately five times as likely to experience a morbidity compared to children over one year old. Children with more complex heart disease or children whose operation included a bypass time longer than 90 minutes were almost twice as likely to experience morbidity. Other important independent risk factors were: palliative or staged procedures, the presence of functionally univentricular heart and pre-procedure severity of illness factors (e.g. pre-procedure mechanical ventilation or shock).

Table 1Summary of risk factors by any morbidity outcome
The weight was missing in 87 (2·8%) procedures and weight for age was set to missing for an additional 99 (3·2%) patients in whom the value was >5 SD from the normative mean. Other data were complete. Values include imputed data.
The results of the multiple logistic regression model for occurrence of 'any morbidity', are expressed as odds ratio for 'any morbidity' by the stated factor, adjusted for age band, low weight, cardiac diagnostic category, functionally univentricular heart, acquired comorbidity, congenital comorbidity, severity of illness indicators, additional cardiac risk factors, specific procedure group, bypass time category. Patient survival was highest and patient hospital stay was shortest for patients with none of the morbidities (see Table 2). The 6-month survival was significantly lower for patients with a single morbidity, for those with ECLS and multiple morbidities, compared with patients with none of the morbidities P < 0·001 for all. Table 2 Secondary outcomes of hospital stay and survival at 6 months by morbidity The number (%) of morbidities was based on individual surgical episode, as was the length of stay, whereas survival was calculated per patient. Data are presented based on morbidity in isolation, multi-morbidity and ECLS. Length of stay data was missing for 9 patients and Life status was missing in 16 patients who were not included.

Impact of morbidity Patient sample
We recruited 60% of eligible families, amounting to 666 children of whom 340 (51%) had at least one morbidity. Of these, 19 died within 30 days and 39 died within 6 months. We pair matched 558 patients and 108 patients unmatched of whom 61 had morbidity; all of the patients were included in the statistical analysis.

Quality of life and anxiety -depression outcomes
For the 6-weeks data, a comparison between patients with missing versus non missing data indicated the only difference was in the proportions with severity of illness factors and additional cardiac risk factors. Importantly there was no difference in the proportion of patients with or without morbidity. At the 6 months mark there was no difference in any baseline factor or morbidity occurrence.
All PedQL scores were lower than in a healthy population (see methods [10][11][12] and Table 3). In terms of our study objective to assess the impact of morbidity; the total PedsQL scores were significantly reduced for children with any morbidity at 6 weeks in comparison to children with none of the morbidities (case mix adjusted score reduction of -5·2 (95% CI -8·3, -2·2) P<0·01). On a positive note, this difference narrowed to small non-significant reduction in total score by 6 months. Physical scores were affected more than psychosocial scores, and were still significantly impaired at 6-months.
All parents experienced higher rates of both anxiety and depression than a healthy population (see methods and Table 3). The parents of children with morbidity were around 57% more likely to experience anxiety; and around 77% more likely to experience depression, at 6 weeks post operation than parents of children without a morbidity. PHQ-4 scores had improved by 6 months, and although there remained a higher chance of both anxiety and depression with morbidity, the difference narrowed and was not significant.  Parental anxiety and depression was derived from PHQ-4 responses of 481 parents at 6 weeks (26% missing) and 394 parents at 6 months (37% missing).
Values include imputed data.
Based on our impact sub study, occurrence of ECLS and multiple morbidities are particularly important adverse outcomes -they are associated with high mortality, high resource use and lower quality of life for surviving patients over 6 months after surgery. Future routine monitoring and public reporting for these should be considered by centres and the national audit.
In our engagement work, patients' families consistently told us that outcomes affecting the child as a whole over the medium or longer term are very important to them. We thus included morbidities such as feeding problems and acute neurological events but note that those types of outcome were the most complicated to capture, and we intend to publish additional material on these topics.

Limitations
We acknowledge that the definition, measurement, reporting and interpretation of morbidity, is much more complex than for mortality. This activity requires significant staff resources and clinician buy in, in particular because some morbidities are challenging to measure and to interpret. The number of individual morbidities was low in some of our analyses, limiting interpretation. We have not captured every single morbidity that exists after children's heart surgery.

Context
There have been previous attempts at reporting morbidity after paediatric cardiac surgery, however these have in general captured events within a hospital setting, whereas our study captures outcomes out to 6 months post-surgery. Two single-centre studies generated an aggregate 'Morbidity Index' by assigning subjective weights to post-operative complications. 26 27 The Society of Thoracic Surgery in the USA, developed a 'Morbidity Score', based on data from their multi-centre registry. 28 We note that condensing diverse morbidities into a single score may lead to loss of information. Moreover, recent work on using graphical methods to routinely monitor a range of morbidities 29 highlighted the complexity of graphically summarising multiple morbidities. 30 The Pediatric Cardiac Critical Care Consortium (PC⁴) was set up in 2009, and provides partner sites with access to real-time,  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Future practice and research
Neurodevelopmental problems are common in children with CHD undergoing surgery 35 however we detected acute neurological events (ANE) following only 2·1% of procedures. : Therefore it appears that peri-operative 'ANE' represents the tip of the iceberg, and hence the scope of surveillance for child neurodevelopment in CHD needs to extend well beyond the peri-operative period.
Despite their great importance to families, we note that participants reported subjectivity in the collection of data on 'feeding problems'; hence these were difficult to capture consistently. Additional research may help determine the best way to measure this important morbidity and to elucidate approaches to alleviate the impact of feeding problems in CHD.
Our measure of renal failure was the need for renal support, and although relatively easy data to collect this may not be the optimal method to capture this morbidity, given that there are differences in practice between clinicians. 36 Further research may help us to understand the best approach to manage post-operative renal injury in CHD.
Our results perhaps unsurprisingly indicate that parents of children who suffer morbidities experience higher rates of anxiety and depression post-operatively. This emphasises the importance of supporting parents during this phase.
We used study data to co-develop parent and carer information resources, showing what the morbidities are and how their incidence and the length of stay may vary based on the complexity of the child's condition. Parents told us that it helps to know that; firstly they are not alone in facing a morbidity, secondly clinical teams have seen morbidities before and know how to deal with them, and thirdly that it is better as a parent to 'be prepared'.
Furthermore, parents indicated that information about impact such as nearly all children who experience a morbidity and recover will have a similar quality of life to children who did not experience a morbidity by the six month mark, was very helpful to know.
Includes new abnormality in any of the following: -Electroencephalogram.
-Brain scan (either computerised tomography or magnetic resonance).
-Clinical evaluation (Seizures or movement disorder, focal neurological signs, generalised neurological signs, altered conscious level including even brain death.) The treatment protocol is variable depending on the type of neuromorbidity.
Specialist consultation with a neurologist, a full evaluation of any brain injury and neurodevelopmental follow up would be a minimum.  Unplanned re-interventions include procedures that were not intended during the planning phase, follow an initial primary cardiac surgery and result in "substantive alteration to heart" incorporating cardiac bypass, cardiac non bypass, pacemaker placement, interventional catheterisations and also diaphragm plications (which are not related to the heart itself). The definition does not include support or other non-cardiac surgery procedures.
Unplanned return to the operating room or cardiac catheter laboratory within 30 days (Excludes interventional catheters that were planned pre-operatively; excluding delayed chest closure, excluding procedures for bleeding) (Includes diaphragm plication and insertion of pace maker for surgically acquired arrhythmia).
Not applicable. The minimal assessment is cardiovascular evaluation of the repair with echocardiography and tolerance of weaning from life supports.

Feeding problems
A diagnosis of postoperative feeding problems should be considered during recovery after surgery and prior to discharge from the specialist centre either to home or to secondary care if A child may fail to feed normally following paediatric cardiac surgery for a range of reasons including gastrooesophageal reflux, vocal cord paralysis, oral-motor dysfunction, oral aversion, and neurologic impairment. (2) If for any of these reasons a child is not able to orally The requirement for any feeding support.
Includes via the intravenous route or via an enteral tube.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 F o r p e e r r e v i e w o n l y the child is unable to feed normally. The goal is detection of feeding problems which are new post-surgery, and it is recognised that this may be challenging where a child was not fed pre-operatively for cardiac reasons since feeding ability will not have been assessed objectively.
feed or completely orally feed and is tube dependent at discharge from the tertiary centre or at 30 days (if he or she is otherwise clinically stable enough to feed at that time point), then a post-operative feeding problem will be diagnosed.
Excludes feeding support that was present to treat a primary problem diagnosed before the surgery, feeding support related to an episode of necrotising enterocolitis, and feeding support because the child dislikes a special diet.
feeding should be monitored by the clinical care team responsible at each stage of the journey.

Need for renal replacement therapy
Includes renal replacement therapy when initiated as a new support at any time from the start of the postoperative admission to ICU up until 30 days following the primary operation.
The child requires renal replacement therapy (peritoneal dialysis or haemofiltration) for renal failure (oligoanuria of less than 0.5 ml/kg/hour and elevated creatinine level for age) and or fluid overload. In patients where renal support is required alongside extracorporeal life support, the primary morbidity is viewed as extracorporeal life support.
The measurement protocol is simply the presence of (new) renal support. (Excludes renal support on extracorporeal life support). Data on renal biochemistry and urine output will be collected.
Instigation of effective renal replacement therapy.

Major adverse cardiac events or never events
Events within this morbidity may be identified during the tertiary hospital stay (either ward or ICU) following the primary surgery.
These morbidity includes: -Cardiac arrest, where the child receives any chest compressions or defibrillation.
-Chest re-opening on the ICU or ward for any reason.
-Major haemorrhage in the ICU following surgery.
-A 'Never Event' applicable to paediatric cardiac surgery as selected from the 'Never Events' list published for NHS for 2015 (3) (Including wrong site or wrong patient surgery, wrong prosthesis surgery, retained foreign object post procedure, wrong route administration of medication, transfusion or transplantation of main red cell group incompatible blood components or organs, misplaced naso-gastric or oro-gastric tubes, Major haemorrhage is defined as bleeding > 10ml/kg/hr on ICU for 2 consecutive hours.
A 'Never Event' includes the events listed plus harm to the patient, for example: if a naso-gastric tube is misplaced, detected and removed in a timely manner before any harm is done then this is not a 'Never Event'. Conversely, if the misplaced naso-gastric tube is not noted, and feed is given into the bronchus, then this is a 'Never Event'.
All events will results in immediate treatment as part of current practice.

Extracorpore al life support
Extracorporeal life support following surgery and before discharge from the tertiary hospital, including the rare cases when a child was on extracorporeal life support before surgery.
This morbidity is defined by the presence of an extracorporeal life support system connected to the patient following the operation, whether it was placed in the operating theatre or in the intensive care unit, and whether the indication was cardiac arrest, low cardiac output state, poor cardiac function, arrhythmia, residual or recurrent cardiac lesion, pulmonary including pulmonary hypertension or sepsis.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 F o r p e e r r e v i e w o n l y bleeding but no radiological changes are included, if a general surgery specialist has seen the child and commenced a course of intravenous antibiotics and parenteral nutrition for five to seven days. Cases of severe necrotising enterocolitis with radiological signs systemic instability and bowel perforation are also included.
treatments deployed, thus enabling the necrotising enterocolitis diagnosis to be graded between 1a and 3b. [25] nutrition, radiological investigation and surgical intervention.

Surgical site infection and bloodstream infection
Surgical site and blood stream infections diagnosed within the hospital admission following surgery or following readmission to the same unit during postoperative recovery, where the treating clinical team assesses the infection to be linked to the recent operation. It is noted that mediastinitis may be detected more than 30 days after cardiac surgery (6) hence this time cut off is not applicable.
Deep surgical site Infection and/or mediastinitis includes any infection of an incised wound that undergoes any re intervention by a surgeon (such as opening of the wound, vacuum dressing), mediastinitis and false aneurysm, independent of culture positivity. [23] Blood stream infection includes both catheter related and non-catheter related. prolonged antibiotic therapy.

Prolonged pleural effusion or chylothorax
Prolonged pleural effusion is a post procedural effusion with duration greater than ten days. Chylothorax is diagnosed from after surgery until discharge from the tertiary hospital.
Either a chylous pleural effusion or significant chylous pericardial effusion or significant chylous ascites or a prolonged non-chylous effusion that necessitates thoracic drainage at least ten days following index cardiac surgery.
Chylous effusions are characterised by milky appearance and a pleural fluid white blood cell count of greater than 1000 cells/μl with lymphocytes greater than 80%. (7) If the child is on normal feeds the triglyceride level in the pleural fluid will be > 1.1 mmol/L or the ratio between the pleural triglyceride level and the serum triglyceride level will exceed 1.