Study population and data sources

The inclusion criteria were child < 3 years of age, included in the database between 2013 and 2022 in France, and a diagnosis of CHD according to the ICD-10 (codes Q20 to Q26) [21]. Fetal deaths in utero and pregnancy terminations were also included in the study. The exclusion criteria were as follows: patients residing outside France or with an unspecified place of residence, incorrect French diagnosis-related groups, and incorrect patient identifiers. The reference population consisted of children aged < 3 years between 2013 and 2022 with no diagnosis of congenital defects (absence of any ICD-10 Q code).

PMSI-MCO (médecine-chirurgie-obstétrique [Medicine, Surgery, and Obstetrics]) data were collected between 2013 and 2022. Data from the French Overseas Departments or Regions and Overseas Collectivities (DROM-COM [Départements ou Régions français d’Outre-Mer et les Collectivités d’Outre-Mer]) were analyzed separately. Linking was performed with PMSI-HAD (hospitalisation à domicile [home hospitalization]) and mother-child data to increase data reliability.

The CHD diagnoses for ICD-10 codes Q20.0 to Q26.9 were collected, both separately as a single CHD diagnosis and in association with multiple CHD diagnoses. We collected the frequencies of different CHDs.

Medical, social, and demographic data were collected from the patients with CHDs and the reference population. The variables of interest were age at the time of diagnosis, sex, gestational age, birth weight, term of birth, mother’s age at birth, parity, route of birth and extraction, whether the child underwent medically assisted reproduction (MAR), death, and age at death.

The data from the ICD-10 codes investigated were: palliative care decision (Z51.5), chromosomal abnormalities (Q90 to Q99), fetus and newborn affected by maternal factors and complications of pregnancy, labor, and delivery (P00 to P04), disorders related to length of gestation and fetal growth (P05 to P08), maternal gestational diabetes (P70.0), maternal diabetes (P70.1), and neonatal withdrawal symptoms from maternal use of drugs of addiction (P96.1). We retained the diagnosis of transposition of the great arteries (TGA) or common arterial trunk in cases of an associated diagnosis of tetralogy of Fallot (ToF) to eliminate some border pathologies.

Statistical analysis

Categorical variables were presented as absolute numbers and percentages. Continuous variables were presented in classes: ≤ 1 month and 1–36 months for the age of diagnosis; ≤ 30 days, 31–365 days, and > 365 days for the days of life; < 24 weeks, 24–36 weeks, and > 36 weeks for the term of birth; < 500 g, 500–1999 g, 2000–3999 g, and ≥ 4000 g for the birth weight; and ≤ 18 years, 18–30 years, 31–39 years, and ≥ 40 years for the maternal age. The chi-square test was used to compare categorical variables. Because the magnitude of statistical significance is heavily influenced by sample size, comparisons between the two groups were also expressed in terms of the standardized difference score (d, as an absolute number) to provide a more robust and reliable estimation of group divergence [33]. Cohen suggested that d = 0.2, 0.5, and 0.8 represents a small, medium, and large effect size, respectively [34]. Risk factors for CHDs were determined based on their clinical relevance and existing data in the literature using logistic regression. Multiple births were included; therefore, the mothers were included more than once in these situations. A random effects term was introduced to account for these situations. The analyses were performed using the secure platform of the Agence Technique de l’Information Hospitalière. Data extraction and statistical analyses were performed using SAS Guide Enterprise version 8.2. Standardized difference scores were calculated using a specific SAS macro [35].

Population

From 2013 to 2022, 88,759 children met the inclusion criteria (83,879 in mainland France and 4,880 in French overseas departments and territories) after excluding erroneous stays and dubious anonymous numbers.

The population without CHD included 8,078,838 children (7,739,840 from metropolitan France and 338,998 from DROM-COM). The reference comparison population without congenital malformations according to ICD-10 codes comprised 7,535,862 children (7,218,952 from mainland France and 316,910 from DROM-COM).

Epidemiology of CHDs

The incidence of CHDs was 10.7 per 1,000 births in mainland France and 14.2 per 1,000 births in French overseas departments and territories.

The incidence of CHDs for each ICD-10 code in mainland France is shown in Table 1 (incidence of CHDs in DROM-COM in the S1 Table). In cases of association, each code was counted individually; therefore, certain categories of CHDs exceeded 100%. For example, cardiac septal defects accounted for 111.38% of multiple diagnoses, as the codes for ventricular septal defects (VSDs) and atrial septal defects (ASDs) were counted separately, although some children could have had both.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Incidence of congenital heart defects in metropolitan France from 2013 to 2022 in the PMSI-MCO (French Medical Information System Program in Medicine, Surgery and Obstetrics) database according to the ICD-10 (International Statistical Classification of Diseases– 10th Revision).

https://doi.org/10.1371/journal.pone.0298234.t001

Table 2 shows the frequencies of certain associations in mainland France. The most frequent associations were between VSD and ASD (6.05% of CHD diagnoses), Coarctation of aorta (CoA) and VSD (1.73%), and TGA and VSD (1.46%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Associations between congenital heart defect diagnoses in metropolitan France from 2013 to 2022 in the PMSI-MCO (French Medical Information System Program in Medicine, Surgery and Obstetrics) database.

https://doi.org/10.1371/journal.pone.0298234.t002

Comparison of CHD

Table 3 compares children in metropolitan France with CHDs and those without congenital defects (data from the DROM-COM in the S2 Table). A strong relationship of CHD with prematurity (d = 0.86), maternal arterial hypertension (d = 0.26), and twinning (d = 0.35) was observed. Death occurred more frequently among patients with CHD (d = 0.59). Sex and maternal age did not differ between children with CHD and children without congenital malformations in the univariate analysis (d = 0.01 and d = 0.15, respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Frequency of perinatal data between children with congenital heart defects and children without congenital malformation in metropolitan France from 2013 to 2022.

https://doi.org/10.1371/journal.pone.0298234.t003

N: number; P00.0: Fetus and newborn affected by maternal hypertensive disorders; P01.5: Fetus and newborn affected by multiple pregnancy; P05.0: Light for gestational age; P70.0: Syndrome of infant of mother with gestational diabetes; P70.1: Syndrome of infant of a mother with diabetes; P96.1: Neonatal withdrawal symptoms from maternal use of drugs of addiction; d: standardized differences score.

In the multivariate analysis, maternal age > 40 years was a risk factor for CHD (OR 1.310, p < 0.001) (Fig 1 for metropolitan France; data in S1 Fig for DROM-COM). Similarly, male sex and pregnancy resulting from MAR were associated with an increased risk of CHD (OR, 1.057 and 1.115, respectively; p < 0.001), whereas multiparity had a protective effect against CHD (OR 0.921, p < 0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Results of multivariate analysis in metropolitan France.

Children with congenital heart defects are compared with the reference population from 2013 to 2022. OR: odds ratio; CI: confidence interval.

https://doi.org/10.1371/journal.pone.0298234.g001

Interpretation of results

To our knowledge, this is the first study to calculate the incidence of CHDs in France using the PMSI database. Its aim was not to replace the registries but to study the CHD diagnoses in France using another reference system to increase its completeness.

Studies suggest that the PMSI database is an appropriate data source for epidemiological studies [22–26]. Our methodology using the PMSI database seems correct because we found the same incidences as those reported in the literature [2, 4, 16, 17]. Therefore, a comparison between populations with CHD and those without congenital malformations is legitimate. We excluded children with other congenital malformations from the reference population to avoid confounding factors between the data studied. It is important to note that the number of children varied according to the secondary criteria studied, because information was not always available.

The PMSI database provided an exhaustive list of CHD diagnoses in France, as coding is compulsory and is carried out at the time of diagnosis. This would not have been possible with registries [18]. The population studied was large, increasing the relevance of the comparison between the CHD and unaffected populations on numerous criteria, which the PMSI database also makes possible.

However, unlike registries, what we gained in exhaustiveness may be lost in diagnostic precision. ICD-10 codes are not very precise (e.g., Q24.9 is "congenital malformation of the heart, unspecified"). The PMSI database is also at risk of coding errors when not carried out by specialists or by poorly trained healthcare professionals, albeit improving [25, 28, 36]. ICD-10 coding is compulsory; however, some ICD-10 codes have no impact on the cost of hospital stays and thus may not be coded. We included children under 3 years of age to limit incomplete data. The PMSI database does not allow searching for more than 10 years back in time, therefore if we had included older children or even adults, recovering the data would not have been possible. This should not have decreased our exhaustiveness because CHDs are diagnosed early in a child’s life and even more so if they are severe [4]. We chose the place of diagnosis and not the place of birth or residence to establish a CHD diagnosis in France. This choice appeared to have the lowest risk of child misallocation. We were unable to distinguish between prenatal and postnatal CHD diagnoses.

DROM-COM is separate from the metropolitan population because the populations are different and access to diagnostic and care facilities is sometimes more difficult [37]. Overall, we found the same differences between the populations and similar risk factors as those observed in the mainland population.

Comparisons with the literature

The incidences found in this study were fairly identical to those reported in the literature for each type of CHD, particularly TGA, ToF, and CoA [2, 4]. However, we have observed more cases of patent ductus arteriosus. This difference could be due to the title of the coding; ductus arteriosus is coded as soon as it is present, particularly in premature babies, where it is extremely common without always impacting the child [5, 38]. However, if all diagnoses of patent ductus arteriosus were excluded (i.e., exclusion in excess), the incidence of CHD would be 8.5%, remaining fairly identical to other studies [3–5].

This partly explains why premature birth and low birth weight, which may be associated with CHD, emerged as risk factors for CHD. However, prematurity appears to be a risk factor for CHD independent of ductus arteriosus. Chu et al. showed an overall increase in the incidence of CHD in children born prematurely between 25 and 32 weeks of gestation, with a high percentage of severe malformations, excluding children with a persistent ductus arteriosus [39].

Much of our data are consistent with what has been described previously. CHDs are a major cause of early death in children [1]. Most diagnoses are made before 1 month of life, including antenatal diagnoses [17]. There were more caesarean sections in cases of CHD, with no differences depending on the type of CHD [40, 41]. MAR is associated with a higher risk of CHD [42]. In twins, there was an increased risk of CHD [43].

Some discrepancies between our data and those published in literatures contribute to the discussion of the inconsistency of the literature. In our study, maternal age > 40 years was a risk factor in multivariate analysis, whereas maternal age > 35 years was sometimes described as a risk factor but not always [11, 29, 44, 45]. Similarly, male individuals were slightly more at risk of CHD in our study, which is consistent with the results of some studies but not others [11, 45–47].

Some of the associations described were either absent or weak. We observed fewer chromosomal abnormalities in children with CHD; however, they remained more numerous than those in the general population [48, 49]. One explanation could be the choice of using ICD-10 codes Q90 to Q99, which does not include all chromosomal and genetic abnormalities but is the easiest to research. This deviation could also be due to the lack of precision in the ICD-10 codes. Similarly, maternal diabetes did not appear to be a risk factor for CHD in our study; however, we only studied diabetes declared during pregnancy, unlike other authors [50].

According to our results, there were no excess mortality in cases of TGA, ToF, or CoA when these were considered serious CHDs, contrary to what has been previously described [51–53]. This may be explained by improved diagnosis and management of these diseases, which have become standardized.

Contrary to our study, in which multiparity appeared to be protective, Feng et al. showed in their meta-analysis an increased risk of CHD in repeated pregnancies. However, their results were affected by the inclusion of heterogeneous studies and showed an absence of statistical difference when directly comparing primiparous and multiparous women [54].

Improved data collection

The accuracy of the PMSI database could be improved using the ICD– 11th Revision published in 2018 by the WHO, which codes CHDs using a simplified version of the International Pediatric and Congenital Cardiac Code [55, 56]. However, this precise CHD classification has not yet been applied in France.

It seems worthwhile to use the data from the PMSI to feed the registers. Despite these limitations, registers remain a valid solution that avoids most coding accuracy problems [18]. The database already exists and is classified into most registries. In a French study led by the HAS on registers, it was suggested that existing registers should work together more closely to populate a common database rather than setting up a single register (which would incur more cost and human resources and would lack proximity).