Carrilero et al. BMC Pediatrics (2020) 20:358 https://doi.org/10.1186/s12887-020-02253-z RESEARCH ARTICLE Open Access Comorbidity patterns and socioeconomic inequalities in children under 15 with medical complexity: a population-based study Neus Carrilero1,2,3, Albert Dalmau-Bueno1 and Anna García-Altés1,4,5* Abstract Background: Children with medical complexity (CMC) denotes the profile of a child with diverse acute and chronic conditions, making intensive use of the healthcare services and with special health and social needs. Previous studies show that CMC are also affected by the socioeconomic position (SEP) of their family. The aim of this study is to describe the pathologic patterns of CMC and their socioeconomic inequalities in order to better manage their needs, plan healthcare services accordingly, and improve the care models in place. Methods: Cross-sectional study with latent class analysis (LCA) of the CMC population under the age of 15 in Catalonia in 2016, using administrative data. LCA was used to define multimorbidity classes based on the presence/ absence of 57 conditions. All individuals were assigned to a best-fit class. Each comorbidity class was described and its association with SEP tested. The Adjusted Morbidity Groups classification system (Catalan acronym GMA) was used to identify the CMC. The main outcome measures were SEP, GMA score, sex, and age distribution, in both populations (CMC and non-CMC) and in each of the classes identified. Results: 71% of the CMC population had at least one parent with no employment or an annual income of less than €18,000. Four comorbidity classes were identified in the CMC: oncology (36.0%), neurodevelopment (13.7%), congenital and perinatal (19.8%), and respiratory (30.5%). SEP associations were: oncology OR 1.9 in boys and 2.0 in girls; neurodevelopment OR 2.3 in boys and 1.8 in girls; congenital and perinatal OR 1.7 in boys and 2.1 in girls; and respiratory OR 2.0 in boys and 2.0 in girls. Conclusions: Our findings show the existence of four different patterns of comorbidities in CMC and a significantly high proportion of lower SEP children in all classes. These results could benefit CMC management by creating more efficient multidisciplinary medical teams according to each comorbidity class and a holistic perspective taking into account its socioeconomic vulnerability. Keywords: Medical complexity, Comorbidity, Child, Health inequalities, Socioeconomic factors, Administrative data, Latent class analysis * Correspondence: annagarciaaltes@gmail.com 1 Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS), Barcelona, Spain 4 CIBER de Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Carrilero et al. BMC Pediatrics (2020) 20:358 Page 2 of 10 Background status and a severity level group, using administrative data. Childhood is widely recognised as one of the population The higher the GMA score, the greater the individual’s groups that warrants special care and attention, even more medical complexity. To construct GMA score, comorbid- so when they suffer chronic comorbidities and severe limi- ity and severity information is gathered automatically from tations – known as children with medical complexity the Catalan Health Surveillance System (CHSS) database, (CMC) [1], one of the most vulnerable populations. Studies for present and previous years. Each person in contact differ regarding the prevalence of CMC status, ranging with the Catalan health system has a GMA score; this between 0.4% [2] and 0.7% [3] of total child population, scoring is used to stratify the population for the purposes although it is rising, given the continuous increase in their of health planning [24, 25]. It is more accurate and yields survival rates [4–8]. less variability than other health risk tools, such as Clinical Children in this population group have complex acute Risk Group (CRG) [25], and has been approved by the and chronic conditions, numerous and varied comorbid- World Health Organisation [26] (see Additional file 1 for ities (from cerebral palsy to congenital heart defects or further details). According to the GMA percentiles, the cancer), a broad range of mental health and psychosocial population is distributed in relation to clinical complexity needs, major functional limitations, and a higher rate of (P50 very low risk, P75 low risk, P85 moderate risk, P90 high mortality [1, 2, 6, 9]. They are under the continuous care risk, P99 very high risk, P99,5 extreme risk). of multiple paediatric specialists and require access to We identified the CMC population based on the chil- specialised care units6. As such, the CMC status indi- dren included in the top 0.5% of GMA scores (P99,5). cates a child with intensive use of healthcare services This criteria was applied since: 1) stratification tools and special health and social needs [10, 11]. Although have proven useful in determining CMC [2, 3, 21, 27, they represent a small proportion of the population, 28]; 2) this is the highest level of complexity indicated CMC account for a substantial proportion of healthcare by the GMA; 3) previous studies in Catalonia have found costs [3], and impact on other externalities such as family that 0.3% of the population were CMC [18]; and 4) resources, psychological stress, and social exclusion [12–15]. concordance with the prevalence of CMC in other popu- Previous studies have examined socioeconomic pos- lation studies [2, 3]. As a comparative group, we used ition (SEP) [16] and ethnic inequalities [17] in CMC the remainder of the child population (non-CMC), [11], and found that the prevalence of life-limiting con- representing 99.5% of that population. ditions is higher in non-white and the most deprived CMC in England [7]. In Catalonia, low-SEP children are Data twice as likely to be CMC than those at the highest We used two main sources of data: 1) The central registry socioeconomic level [18, 19]. However, a study con- of insured persons was used to obtain the reference popu- ducted in Wales did not find an association between lation (as of January 1, 2016) based on their income level, mortality rates in paediatric intensive care units and employment status, and Social Security benefits; 2) the SEP, despite noting an increase in the most vulnerable CHSS database includes detailed information on sociode- categories, especially among some ethnic groups [17]. mographic characteristics and medical diagnoses at an in- With few exceptions [2], the research to date has dividual level in all contacts with primary care, emergency focused on CMC with diseases within intensive care care, mental healthcare, long-term care services. All the units, where accessible data elements are often restricted historical comorbidities are updated if they are relevant, to the hospital setting [4, 6–8, 20, 21]. A wider approach and it includes the whole population of Catalonia, since all is essential in order to obtain evidence that can guide citizens are granted universal health coverage. the coordination of healthcare resources targeted to the different CMC profiles more efficiently [1]. Variables The aim of this study is to describe more accurately te The main outcome variable is the different classes pathologic patterns of CMC (by clustering health diseases obtained by grouping patients with similar patterns of [22, 23]) and their socioeconomic inequalities in order to comorbidity. Comorbidities for all CMC were gathered better manage better their needs, plan healthcare services from all the diagnoses registered and updated from 2014 accordingly, and improve the care models in place. to 2016. Diagnoses were coded using the Agency for Healthcare Research and Quality’s Clinical Classification Methods Software (CCS) [29]. From a list of 184 relevant CCS, we Study population grouped them into disease categories in order to facili- We selected the CMC individuals from the population tate information management. For each different CCS, it of Catalonia under the age of 15 in 2016 (1,189,325). was only counted once in each individual. To obtain CMC individuals were identified by GMA score [24], a consistent and clinically relevant patterns of association, risk tool which classifies each individual into a health and to avoid spurious relationships that could bias the Carrilero et al. BMC Pediatrics (2020) 20:358 Page 3 of 10 results, we considered only diagnosis categories with a classes) based on a number of observed indicators, in this prevalence of > 1%. Finally, 57 disease categories were in- case, the 57 disease categories. cluded, covering 90.6% of all diseases (see Additional file 2). To determine the optimal number of latent classes to For the exposure variable, the SEP of each child was fit the data, we used the Bayesian Information Criterion measured based on economic information relating to (BIC) and Akaike’s Information Criterion (AIC). An one of their parents or guardians, including: employment overall χ2 statistic was used to assess the model [32]. status, individual income, and the receipt of welfare We compared candidate models and applied substantive assistance. SEP was grouped into three categories: low interpretability and clinical judgement (i.e., do the clas- (no member of the household employed or in receipt of ses defined by a given model possess a clinical signifi- welfare support from the government, and an income < cance or meaning?). After selecting a latent class model, €18,000/year, considered at risk of poverty [30]); middle we assigned each participant to his or her ‘best-fit’ class, (guardian employed with an income <€18,000); and high meaning the class for which the participant had the (guardian employment, with income >€18,000). highest computed probability of membership. Age was categorised based on clinical criteria for chil- Subsequently we describle age, SEP, and GMA distribu- dren’s growth (0–1, 2–4, 5–11, 12–14) and used as the tion in each class found in the LCA analysis by sex. Bivarate covariate, and sex was used as the stratification variable. analysis was conducted to determine differences between boys and girls – a proportion test and Chi-square test, and T-tests or Mann-Whitney U tests were carried out. Finally, Statistical analysis regression logistic models were used to examine the rela- A descriptive analysis of both the CMC and non-CMC pop- tionship between class membership and SEP with confi- ulations was carried out. Bivariate analysis was conducted to dence intervals at 95% (CI95%) and their p-values. determine differences between CMC and non-CMC groups All the analyses were carried out for boys and girls, according to sex, age, SEP, and GMA; proportion tests and separately. For all tests, the accepted significance level Chi-square tests (for categorical variables) and a T-test or was 0.05 and adjusted by age. LCA was performed using Mann–Whitney U (for continous variables) test were carried the poLCA package [33] and R statistical software, out depending on variable distribution. version 3.3.1 [34], for conducting all analyses. Next, we used latent class analysis (LCA) [31] to classify CMC into patterns of comorbidity according to their dis- Results tribution of disease categories. The objective of LCA is to Characteristics of the CMC population classify individuals from an apparently heterogeneous The main characteristics of the CMC (0.5%) and non- population into more homogenous subgroups (latent CMC (99.5%) populations are described in Table 1. Both Table 1 Characteristics of children under 15 by population (CMCa and non-CMC b) and sex in Catalonia, 2016 Boys Girls CMC Non-CMC CMC Non-CMC N % N % P Valued N % N % P Valued Frequency 3480 58.4 609,015 51.4 <.001 2470 41.6 574,360 48.6 <.001 Age (years) <2 878 25.2 70,465 11.6 <.001 604 24.4 66,591 11.6 <.001 2 to 4 916 26.3 113,705 18.7 636 25.8 107,284 18.7 5 to 11 1256 36.1 303,127 49.8 848 34.3 285,247 49.6 12 to 14 430 12.4 121,718 19.9 382 15.5 115,238 20.1 SEP Low 440 12.7 53,360 8.8 <.001 317 12.9 50,261 8.8 <.001 Middle 2030 58.5 321,762 52.8 1428 58.1 303,039 52.8 High 1001 28.8 233,197 38.4 713 29.0 220,437 38.4 GMAc (score) 16.7 (15.1–20.5) 2.3 (0.8–4.1) <.001 16.7 (15.0–20.0) 2.1 (0.7–3.8) <.001 Note: GMA Morbidity Adjusted Group, SEP Socioeconomic Position. Low (none member of the household employed, receiving welfare support from the government and an income < 18,000€/year), Middle (employed and an income < 18,000€/year), High (employed and an income > 18,000€/year) a Children Medically Complex population = top 0.5% of GMA score of all entire population under 15 b Non Children Medically Complex population = 99.5% bottom of GMA score of all entire population under 15 Values are absolute numbers (percentages) for categorical variables. cMedian (IQR) d P Value χ2 test for categorical variables and Mann-Whitney U-test for continuous variables. Differences between CMC and Non-CMC populations according to sex groups. α = 0.005 Carrilero et al. BMC Pediatrics (2020) 20:358 Page 4 of 10 populations contained a higher proportion of boys (CMC boys, 43.2% girls), short gestation, low birth weight, and 58.5% versus non-CMC 51.1%) than girls. foetal growth retardation (38.3% boys, 39.7% girls), and Almost a quarter of CMC were in the two first years other congenital anomalies (36.8% boys, 35.2% girls) of life (25.2% boys and 24.4% girls); compared with the were the most frequent diseases. non-CMC population; this rate was 2.2 times higher in Respiratory class (Fig. 2d): included 1814 (30.5% of the boys and 2.1 times higher in girls. Approximately 50% of CMC). It shows an accumulation of individuals aged CMC of both sexes were aged under five, compared with between years 1 and 6. The most prominent diseases around 30% of non-CMC; the rate was 69.7% higher in were chronic obstructive pulmonary disease and bron- boys and 65.7% higher in girls (Table 1). chiectasis (64.2% boys, 62.1% girls), respiratory failure, In terms of SEP, 71.1% of CMC (6.6% of non-CMC) insufficiency and arrest (54.7% boys, 53.6% girls), and had at least one parent with an annual income of less asthma (53.7% boys, 47.4% girls). than €18,000 (low and middle SEP). Low SEP had a prevalence of 12.8% in the CMC group (12.7% in boys SEP inequalities and 12.9% in girls) compared to 8.8% in non-CMC in SEP inequalities in the four clusters are displayed in both boys and girls; it is 44.5% higher in boys and 46.6% Table 2 and Fig. 3. There were SEP inequalities in all higher in girls in CMC than in the non-CMC group. clusters, for both sexes. From higher to lower OR in one of both sexes, neurodevelopment class showed an associ- Comorbidity classes of CMC ation with low SEP ([OR, 2.3; CI95%, 1.7–3.1 in boys] The smallest BIC and AIC values were obtained for the and [OR, 1.8; CI95%, 1.2–2.6 in girls]) compared to the 4-class and 5-class candidate models. (see Additional file 3 high SEP category, congenital and perinatal class ([OR, for statistical values); after applying clinical criteria and 2.1; CI95%, 1.5–2.8 in girls]) and [OR, 1.7; CI95%, 1.3– χ2 value, we selected the 4-class model. The four classes 2.3 in boys]), followed by respiratory class ([OR, 2.0; were labelled based on which conditions exhibited more CI95%, 1.6–2.6 in girls] and [OR, 2.0; CI95%, 1.7–2.5 in prevalence: oncology, neurodevelopment, congenital and boys]), and finally the oncology class ([OR, 2.0; CI95%, perinatal, and respiratory. 1.7–2.5 in girls] and [OR, 1.9; CI95%, 1.6–2.3 in boys]). Prevalences of all disease categories in each class are summarised in Additional file 4. Upper respiratory Discussion disease, infection, gastrointestinal disorders, fractures Four different comorbidity classes among the CMC were and injuries, and ear, eye, and skin disorders were highly identified. All of them showed SEP inequalities, therefore present in all classes. the more disadvantaged children represent a higher pro- The characteristics of the classes are summarised in portion of the CMC group. In both populations (CMC Table 2 and their distribution during childhood is shown and non-CMC) and sexes, > 60% of children were low in Fig. 1. The SEP and age distribution of each obtained and middle SEP. This finding highlights the fact that class is summarised in Table 2 and Fig. 1. Figures 2a,b,c, children are subject to inequalities from the very begin- d display the most prevalent diseases (> 20%) in each of ning of their lives [35]. the four classes. In this study, all CMC classes shared common diseases Oncology class (Fig. 2a): includes 2141 children (36.0% – specifically gastrointestinal disorders, respiratory dis- of the CMC). Distribution was highest up to five years. eases, and trauma – as in other population studies [20, There was a high proportion of oncological and related 21]. These diseases account for the major causes of hos- diseases: malignant cancer (23.7% boys, 24.7% girls), pitalisation rates together with congenital anomalies, leukaemia (12.3% boys, 10.8% girls), cancer of the brain and cardiovascular and oncological diseases [7, 21]. and nervous system (6.4% boys, 7.1% girls), and haem- All classes had a higher proportion of boys, up to 56%. atological disorders (36.1% boys, 35.0% girls). This result is consistent with the highest vulnerability in Neurodevelopment class (Fig. 2b): includes 818 chil- boys aged up to five years; male foetuses mature slower dren (13.7% of the CMC). Distribution is fairly constant than female foetuses do and, after birth, males experi- from 3 years and aupwards. Among the most prevalent ence more perinatal issues [36]. This also coincides with diseases were developmental disorders (72.9% boys, the maximum prevalence of the congenital and perinatal, 71.0% girls), other nervous system disorders (65.1% boys, and respiratory classes (99.1 and 76.1%, respectively). 63.7% girls), epilepsy and convulsions (57.1% boys and The oncology class contained 36.0% of all CMC and 61.8% girls), and paralysis (37.7% boys and 39.1% girls). was predominated by individuals aged up to 5 years. Congenital and perinatal class (Fig. 2c): includes 1177 Their characteristics were more heterogeneous and children (19.8% of the CMC). Distribution is mainly up showed a higher comorbidity profile. Although this class to 4 years old. Perinatal trauma (84.1% boys, 74.7% girls), included almost all the individuals with malignancies, in- cardiac and circulatory congenital anomalies (43.9% dividuals with mental health or endocrine disorders were Table 2 Socioeconomic characteristics of each comorbidity cluster among the CMC and association between SEP by cluster, in Catalonia, 2016 Carrilero et al. BMC Pediatrics Oncologya Neurodevelopmenta Congenital and perinatala Respiratorya N = 2141 36.0% N = 818 13.7% N = 1177 19.8% N = 1814 30.5% Boys Girls P Value Boys Girls P Value Boys Girls P Value Boys Girls P Value N % N % N % N % N % N % N % N % Frequency 1200 56.1 941 44.0 0.004 501 61.3 317 38.8 0.083 668 56.8 509 43.3 0.182 1110 61.2 704 38.8 0.005 (2020) 20:358 Ages (years)b <2 8 0.7 7 0.7 0.214π 38 7.6 23 7.3 0.126π 552 82.6 400 78.6 0.195π 280 25.2 174 24.7 0.569π 2 to 4 95 7.9 84 8.9 145 28.9 86 27.1 109 16.3 106 20.8 566 51.0 361 51.3 5 to 11 729 60.8 529 56.2 258 51.5 151 47.6 7 1.1 3 0.6 262 23.6 165 23.4 12 to 14 368 30.7 321 34.1 60 12.0 57 18.0 0 0.0 0 0.0 2 0.2 4 0.6 SEPb High 159 13.3 134 14.3 0.749π 69 13.9 40 12.8 0.297π 69 10.4 55 10.9 0.534π 143 12.9 88 12.5 0.9443π Middle 664 55.4 507 54.2 299 60.2 176 56.2 408 61.3 323 63.7 658 59.3 422 60.1 Low 375 31.3 295 31.5 129 26.0 97 31.0 189 28.4 129 25.4 308 27.8 192 27.4 d SEP OR (CI) High 1 1 1 1 1 1 1 1 Middle 1.3 (1.2–1.5) 1.3 (1.1–1.5) 1.6 (1.3–2.0) 1.3 (1.0–1.7) 1.3 (1.1–1.6) 1.5 (1.3–1.9) 1.4 (1.2–1.6) 1.5 (1.2–1.7) Low 1.9 (1.6–2.3) 2.0 (1.7–2.5) 2.3 (1.7–3.1) 1.8 (1.2–2.6) 1.7 (1.3–2.3) 2.1 (1.5–2.8) 2.0 (1.7–2.5) 2.0 (1.6–2.6) GMAc (score) 16.8 (15.2, 20.6) 16.9 (15.1–20-6) 0.555γ 20.9 (16.8, 26.2) 19.5 (16.2–26.5) 0.182γ 17.6 (15.4, 21.5) 16.9 (15.2, 20.2) 0.058γ 15.8 (14.7, 17.4) 15.8 (14.7, 17.4) 0.821γ Note: GMA Morbidity Adjusted Group, SEP Socioeconomic Position. Low (no member of the household employed, receiving welfare support from the government, and an income < 18,000€/year), middle (employed and an income < 18,000€/year), high (employed and an income > 18,000€/year) a Proportion of each comorbidity class (%): (num of individuals in class/num individuals in all CMC)*100 b Values are absolute numbers and percentages in each class for categorical variables. cMedian (IQR) for continuous variables P Value for categorical variables: π χ2 Test. For continuous variables: γ Mann-Whitney U-test. α = 0.005 d Odds Ratio, adjusted by age. CI, 95% confidence intervals of the odds ratio Page 5 of 10 Carrilero et al. BMC Pediatrics (2020) 20:358 Page 6 of 10 Fig. 1 Proportion of comorbidity classes among CMC by age and sex, in Catalonia, 2016 also highly represented. This pattern is consistent with The respiratory class includes mainly pulmonary dis- other studies that emphasise that when a CMC matures eases. In accordance with the natural development of he or she could develop more than one comorbidity as a most respiratory diseases, its prevalence was highest in result of their main pathology [20], and often this new the mid-age range, and it accounted for 30% of all the comorbidities are related to mental disorders [37]. CMC. Risk factors known to be related to SEP inequal- The neurodevelopment class includes two related ities in this class are: exposure to air pollution [46], in types of diseases: nervous system disorders such as utero exposure to tobacco [47], maternal stress [45], and paralysis and epilepsy, and congenital, perinatal, and de- low weight at birth and prematurity [48]. generative anomalies. Their prevalence remains steady The age distribution of each class showed the ages of as the child grows older; they have a chronic, cumulative maximum expression of each of the patterns (see Fig. 1). profile due to the difficulty of healing, and they may be It should be noted that diseases are not static, and prog- precursors of future complications in other systems [38]. nosis may mean that individuals move from class to The aetiology of nervous system anomalies may be class. related to SEP inequalities, such as exposure to certain All CMC classes showed SEP inequalities, thus corrob- environmental factors [39], maternal stress during preg- orating the previous analyses carried out in Catalonia nancy, or adverse gestational and delivery outcomes [40, [18, 19] and elsewhere [7, 16]. SEP inequalities were 41]. All these events occur in the prenatal and perinatal similar across the four classes, which highlights the lack period but their impact may emerge at a later stage. This of economic support in accessing the best development class showed the highest median GMA, since the prog- and care that these children and their family experience. nosis and development of the pathology entail a high Our study denotes that family’s SEP is related to CMC. risk and large use of healthcare resources. This fact could impact on their development and hinder- The congenital and perinatal disease class comprises ing these children from achieving their potential. Having mainly adverse birth outcomes and congenital anomalies to care for CMC may also further negatively affect fam- in diverse body systems, especially heart defects in con- ilies’ economic position and health [12, 14, 15] and may, cordance with the principal incidence of congenital in turn, affect the CMC. In contrast, families with more anomalies in other populations [42]. The maximum economic resources are able to provide more active prevalance of the congenital and perinatal class was ob- stimulation, alternative treatments, and an environment served in the first two years of life, due to the congenital that is safer and more conducive to maintaining good aetiology. In this short time, SEP influences the child health in childhood. This phenomenon has been termed mainly via the mother: maternal behaviour during preg- the “buffering effect of income” in chronic conditions nancy has been identified as a risk factor [43–45]. It [38]. On the other hand, these results denote that in- should also be noted that advances in perinatal care have equities are already established in the first years of life, increased the likelihood of survival for extremely suggesting that there is a pattern of causality as indicated preterm infants, who are mostly included in this class. by different studies of highly disabling diseases [16]. The Carrilero et al. BMC Pediatrics (2020) 20:358 Page 7 of 10 Fig. 2 a Most prevalent diseases (> 20%) in the oncology class. b Most prevalent diseases (> 20%) in the neurodevelopment class. Abreviations: Chronic obstructive pulmonary disease and bronchiectasis. Other hereditary and degenerative nervous system conditions. c Most prevalent diseases (> 20%) in the congenital and perinatal class. Abreviations: Short gestation; low birth weight; and foetal growth retardation. Chronic obstructive pulmonary disease and bronchiectasis. d Most prevalent diseases (> 20%) in the respiratory class. Abreviations: Chronic obstructive pulmonary disease and bronchiectasis idea that, since conception, SEP inequalities are an im- explained by reverse causality alone. This indicates that portant factor determining the developmental origin of the issue warrants further research. According to differ- different diseases, is increasingly gaining more evidence, ent experts on CMC [11], other social determinants of establishing that the mother and the family environment health such as ethnicity, immigration status, or geo- are key to the production of disease [49]. Ensuring this graphical isolation, influence CMC’s health outcomes. pattern in all diseases is challenging, but our results, es- Our data does not allow for deeper insights on ethnicity; pecially in regard to the youngest CMC, cannot be this should be explored further in future research as Carrilero et al. BMC Pediatrics (2020) 20:358 Page 8 of 10 Fig. 3 Odds ratio between socioeconomic position (SEP) and each comorbidity class among CMC by sex. Catalonia, 2016.*. *Models were adjusted by age. Odds ratio and 95% Confidence Interval. High SEP was the reference category some studies have identified it as a factor having more The health status data is based on the use of public health- influence than SEP [7, 17]. care resources, since data from private healthcare providers was not available. Even so, the bias is presumably very low, as CMC patients require highly specialised care and, for this Study strengths and limitations reason, are mainly treated in the public healthcare system. Identifying CMC at a population level is not straightfor- The main strength of this population-based study is ward. There is no specific agreed criteria for definingthe the use of robust individual administrative data, like CMC population, and all of the proposed criteria have similar studies and databases [2, 6, 7]. Another advan- present limitations. The GMA, like other classification tage is that it includes all the children in Catalonia and systems, was originally created for the whole population thus provides a realistic view of the current health status (children and adults). Nevertheless, Clinical Risk Groups of the population beyond hospital-based care. Hospital- based on the same principle and have been successfully based studies do not address outpatient utilisation of used to identify CMC populations [2, 3, 50, 51]. Further- services and do not reflect the highest-risk patients, as more, clinical diagnoses across the historical healthcare they are treated in specialised units; meanwhile, our contacts have been considered as it is recommend [28], study sheds light on all the comorbidities adjacent to the providing a more realistic approach to the health status CMC population with a more chronic profile. of the children. Because of the limited data available on income, we Conclusion were unable to obtain a more detailed segmentation of Our findings have demonstrated the existence of differ- the SEP variable. This was especially true in the case of ent patterns of comorbidities in CMC and a high pro- the high SEP category, which included a wide range of portion of lower socioeconomic children in all classes. income levels. Further segmentation of this category This result could benefit CMC management by enabling would have given a more accurate approximation of the the creation of more efficient multidisciplinary teams SEP gradient. However, parental income is the SEP indi- according to each comorbidity class and informing a cator that most directly measures the family’s material holistic perspective taking into account the socioeco- resources. With other indicators, such as labour, income nomic vulnerability this population faces. has a ‘dose-response’ association with health [52]. Some Daily life for CMC and their families is not only com- studies have used maternal education or an ecological plex from the perspective of healthcare; every area of life deprivation index as a proxy for SEP [7, 16, 17]; the is complex. Child health and family health are two sides present study goes further by using population-based of the same coin. Introducing policies to support both individual income data, which is more directly related to their health and financial situation will have implications the material resources. beyond children’s health itself. Carrilero et al. BMC Pediatrics (2020) 20:358 Page 9 of 10 Supplementary information Author details 1 Supplementary information accompanies this paper at https://doi.org/10. Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS), Barcelona, 1186/s12887-020-02253-z. Spain. 2Department of Experimental and Health Sciences (DCEXS), Universitat Pompeu Fabra, Barcelona, Spain. 3Institut de Recerda de l’Hospital de la Santa Creu i Sant Pau (IR Sant Pau), Barcelona, Spain. 4CIBER de Additional file 1. Adjusted morbidity groups. Description of data: Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain. 5Institut Details of the adjusted morbidity groups construction. d’Investigació Biomèdica (IIB Sant Pau), Carrer de Roc Boronat, 81-95, 08005 Additional file 2. List of the Clinical Classifications Software (CCS) for Barcelona, Spain. ICD-9-MC. included in each disease category (covering 90.6% of all the disease events). Description of data: Clinical codes included in each dis- Received: 26 May 2020 Accepted: 21 July 2020 ease category. Additional file 3. LCA statistics. Description of data: LCA statistics for all the models used. It included Chisq Chi_square goodness of fit, Bayesian Information Criterion, Akaike’s Information Criterion, Log_likelihood, References Consistent Alkaike’s Information Criterion and Likelihood Ratio chi-square. 1. Simon TD, Mahant S, Cohen E. Pediatric hospital medicine and children Additional file 4. Prevalences of all disease categories by sex for each with medical complexity: past, present, and future. Curr Probl Pediatr comorbidity class among the CMC in Catalonia, 2016. Description of data: Adolesc Health Care. 2012;42:113–9. Prevalences of all the disease categories for each of the comorbidity 2. Neff JM, Sharp VL, Muldoon J, Graham J, Popalisky J, Gay JC. Identifying and classes obtained in the LCA. This data shows the frequencies and classifying children with chronic conditions using administrative data with percentages of each disease category by sex for each of the classes the clinical risk group classification system. Ambul Pediatr. 2002;2:71–9. obtained. 3. Cohen E, Berry JG, Camacho X, Anderson G, Wodchis W, Guttmann A. Patterns and costs of health care use of children with medical complexity. Pediatrics. 2012;130:e1463. Abbreviations 4. Burns KH, Casey PH, Lyle RE, Mac BT, Fussell JJ, Robbins JM. Increasing CMC: Children with medical complexity; SEP: Socioeconomic position; prevalence of medically complex children in US hospitals. Pediatrics. 2010; Catalan acronym GMA: The Adjusted Morbidity Groups; CHSS: Catalan Health 126(4):638–46. Surveillance System; CRG: Clinical Risk Group; CCS: Clinical Classification 5. Cohen E, Kuo DZ, Agrawal R, Berry JG, Bhagat SKM, Simon TD, et al. Software; LCA: Latent Class Analysis; BIC: Bayesian Information Criterion; Children with medical complexity: an emerging population for clinical and AIC: Akaike’s Information Criterion research initiatives. Pediatrics. 2011;127:529–38. 6. Fraser LK, Parslow R. Children with life-limiting conditions in paediatric Acknowledgements intensive care units: a national cohort, data linkage study. Arch Dis Child. We thank Emili Vela (from the Catalan Health Service) for generously sharing 2018;103:540–7. his knowledge of GMA; Juan José García García, head of the paediatric 7. Fraser LK, Miller M, Hain R, Norman P, Aldridge J, McKinney PA, et al. Rising service at Sant Joan de Déu Barcelona Hospital, for his insights regarding the national prevalence of life-limiting conditions in children in England. data and the results; Elisenda Martinez, for her support in data analysis; and Pediatrics. 2012;129:e923–9. Cristina Colls and Dolores Ruiz Muñoz, for their assistance and professional 8. Rennick JE, Childerhose JE. Redefining success in the PICU: new patient knowledge. populations shift targets of care. Pediatrics. 2015;135:e289–91. 9. Feudtner C, Christakis DA, Connell FA. Pediatric deaths attributable to Authors’ contributions complex chronic conditions: a population-based study of Washington state, NC had full access to all the data in the study and vouches for the integrity 1980-1997. Pediatrics. 2000;106(1 Pt 2):205–9. of the data and the accuracy of the data analysis. Study concept and design: 10. Van Der Lee JH, Mokkink LB, Grootenhuis MA, Heymans HS, Offringa M. AGA. Acquisition, analysis, and interpretation of data: NC and ADB. Drafting Definitions and measurement of chronic health conditions in childhood: a of the manuscript: NC. Critical revision of the manuscript for important systematic review. JAMA. 2007;297:2741–51. intellectual content: NC and AGA. Study supervision: All authors. All authors 11. Barnert ES, Coller RJ, Nelson BB, Thompson LR, Chan V, Padilla C, et al. have read and approved the manuscript. Experts' perspectives toward a population health approach for children with medical complexity. Acad Pediatr. 2017;17:672–7. Authors’ information 12. Thomson J, Shah SS, Simmons JM, Sauers-Ford HS, Brunswick S, Hall D, et al. This work has been conducted within the framework of the PhD in Financial and social hardships in families of children with medical Biomedics of the University Pompeu Fabra. complexity. J Pediatr. 2016;172:187–93 e1. 13. Seltzer RR, Henderson CM, Boss RD. Medical foster care: what happens Funding when children with medical complexity cannot be cared for by their This work was supported by the Industrial Doctorates Plan of the Catalan families? Pediatr Res. 2016;79:191–6. Government. The founder had no role in the design of the study and 14. Arthur JD, Gupta D. “You can carry the torch now”: a qualitative analysis of collection, analysis, and interpretation of data and in writing the manuscript. parents’ experiences caring for a child with trisomy 13 or 18. HEC Forum. 2017;29:223–40. Availability of data and materials 15. Woodgate RL, Edwards M, Ripat JD, Borton B, Rempel G. Intense parenting: The data that support the findings of this study are not publicly available a qualitative study detailing the experiences of parenting children with due to the presence of personal information that could compromise complex care needs. BMC Pediatr. 2015;15:1–15. research participants’ privacy. The anonymised and unidentified data will be 16. Spencer NJ, Blackburn CM, Read JM. Disabling chronic conditions in accessible to the research staff of the research centres accredited by the childhood and socioeconomic disadvantage: a systematic review and meta- Research Centres of Catalonia (CERCA) institution, SISCAT agents, and public analyses of observational studies. BMJ Open. 2015;5:e007062. university research centres, as well as the same health administration. 17. Parslow RC, Tasker RC, Draper ES, Parry GJ, Jones S, Chater T, et al. Epidemiology of critically ill children in England and Wales: incidence, Ethics approval and consent to participate mortality, deprivation and ethnicity. Arch Dis Child. 2009;94:210–5. Not applicable. 18. Observatori del Sistema de Salut de Catalunya. Desigualtats socioeconòmiques en la salut i la utilització de serveis sanitaris públics en la Consent for publication població de Catalunya. Barcelona (Spain): Agència de Qualitat i Avaluació Not applicable. Sanitàries de Catalunya. Departament de Salut. Generalitat de Catalunya; 2017. [http://observatorisalut.gencat.cat/web/.content/minisite/ Competing interests observatorisalut/ossc_crisi_salut/Fitxers_crisi/Salut_crisi_informe_2016.pdf]. None declared. Accessed on date 10 Oct 2019. Carrilero et al. BMC Pediatrics (2020) 20:358 Page 10 of 10 19. García-Altés A, Ruiz-Muñoz D, Colls C, Mias M, Martín BN. Socioeconomic 41. Himmelmann K, Ahlin K, Jacobsson B, Cans C, Thorsen P. Risk factors for inequalities in health and the use of healthcare services in Catalonia: cerebral palsy in children born at term. Acta Obstet Gynecol Scand. 2011;90: analysis of the individual data of 7.5 million residents. J Epidemiol 1070–81. Community Health. 2018;72:871–9. 42. Mai CT, Riehle-Colarusso T, O’Halloran A, Cragan JD, Olney RS, Lin A, et al. 20. Crow SS, Undavalli C, Warner DO, Katusic SK, Kandel P, Murphy SL, et al. Selected birth defects data from population-based birth defects surveillance Epidemiology of pediatric critical illness in a population-based birth cohort programs in the United States, 2005-2009: featuring critical congenital heart in Olmsted county, MN. Pediatr Crit Care Med. 2017;18:e137–45. defects targeted for pulse oximetry screening. Birth Defects Res Part A - Clin 21. Gold JM, Hall M, Shah SS, Thomson J, Subramony A, Mahant S, et al. Long Mol Teratol. 2012;94:970–83. length of hospital stay in children with medical complexity. J Hosp Med. 43. Stothard KJ, Tennant PWG, Bell R, Rankin J. Maternal overweight and obesity 2016;11:750–6. and the risk of congenital anomalies. JAMA. 2009;301:636. 22. Barnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. Epidemiology 44. Martínez-Frías ML, Bermejo E, Rodríguez-Pinilla E, Frías JL. Risk for congenital of multimorbidity and implications for health care, research, and medical anomalies associated with different sporadic and daily doses of alcohol education: a cross-sectional study. Lancet. 2012;380:37–43. consumption during pregnancy: a case-control study. Birth Defects Res Part 23. Hesketh KR, Fagg J, Muniz-Terrera G, Bedford H, Law C, Hope S. Co- A Clin Mol Teratol. 2004;70:194–200. occurrence and clustering of health conditions at age 11: cross-sectional 45. Wright RJ, Visness CM, Calatroni A, Grayson MH, Gold DR, Sandel MT, et al. findings from the millennium cohort study. BMJ Open. 2016;6:e012919. Prenatal maternal stress and cord blood innate and adaptive cytokine 24. Monterde D, Vela E, Clèries M. Los grupos de morbilidad ajustados: nuevo responses in an inner-city cohort. Am J Respir Crit Care Med. 2010;182:25–33. agrupador de morbilidad poblacional de utilidad en el ámbito de la 46. Li S, Williams G, Jalaludin B, Baker P. Panel studies of air pollution on atención primaria. Aten Primaria. 2016;48:674–82. children’s lung function and respiratory symptoms: a literature review. J 25. Dueñas-Espín I, Vela E, Pauws S, Bescos C, Cano I, Cleries M, et al. Proposals Asthma. 2012;49:895–910. for enhanced health risk assessment and stratification in an integrated care 47. McEvoy CT, Spindel ER. Pulmonary effects of maternal smoking on the fetus scenario. BMJ Open. 2016;6:e010301. and child: effects on lung development, respiratory morbidities, and life 26. Cerezo J, Arias C. Population stratification: A fundamental instrument used long lung health. Paediatr Respir Rev. 2017;21:27–33. for population health management in Spain. In: Good Practice Brief. 48. Skromme K, Vollsæter M, Øymar K, Markestad T, Halvorsen T. Respiratory Regional Office of Europe. World Health Organization (WHO); 2018. [http:// morbidity through the first decade of life in a national cohort of children www.euro.who.int/__data/assets/pdf_file/0006/364191/gpb-population- born extremely preterm. BMC Pediatr. 2018;18:102. stratification-spain.pdf?ua=1]. Accessed on date 10 Oct 2019. 49. Wallack L, Thornburg K. Developmental origins, epigenetics, and equity: 27. Neff JM, Sharp VL, Muldoon J, Graham J, Myers K. Profile of medical charges moving upstream. Matern Child Health J. 2016;20:935–40. for children by health status group and severity level in a Washington state 50. Loh W, Tang M. The epidemiology of food allergy in the global context. Int health plan. Health Serv Res. 2004;39:73–89. J Environ Res Public Health. 2018;15(9):2043. 28. Berry JG, Hall M, Cohen E, O’Neill M, Feudtner C. Ways to identify children 51. Berry JG, Hall M, Hall DE, Kuo DZ, Cohen E, Agrawal R, et al. Inpatient with medical complexity and the importance of why. J Pediatr. 2015;167: growth and resource use in 28 children’s hospitals: a longitudinal, multi- 229–37. institutional study. JAMA Pediatr. 2013;167:170–7. 29. Healthcare Cost and Utilization Project – HCUP A Federal-State-Industry 52. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Davey SG. Indicators of Partnership in Health Data Clinical Classifications Software (CCS); 2015. socioeconomic position (part 1). J Epidemiol Community Health. 2006;60:7–12. [http://www.ahrq.gov/data/hcup]. Accessed on date 10 Nov 2017. 30. Idescat. Enquesta de condicions de vida. Llindar de risc de pobresa segons Publisher’s Note la composició de la llar. Barcelona (Spain): Idescat-Institut d'Estadística de Springer Nature remains neutral with regard to jurisdictional claims in Catalunya. [Internet]. [https://www.idescat.cat/pub/?id=ecv&n=7623]. published maps and institutional affiliations. Accessed on date 04 May 2019. 31. Hagenaars JA, McCutcheon AL. Applied latent class analysis. Cambridge: Cambridge University Press; 2002. 32. Kongsted A, Nielsen AM. Latent class analysis in health research. J Physiother. 2017;63:55–8. 33. Linzer D, Lewis J. Package “poLCA”. Polytomous variable Latent Class Analysis. August 29, 2016. In: The Comprehensive R Archive Network. The R Project for Statistical Computing [Internet]. [https://cran.r-project.org/https:// cran.r-project.org/web/packages/poLCA/poLCA.pdf]. Accessed on date 04 May 2019. 34. RStudio. Open source & professional software for data science teams [Internet]. Boston, MA: RS Studio [https://rstudio.com]. Accessed on date 18 Apr 2018. 35. Instituto UAM-UNICEF de Necesidades y Derechos de la Infancia y la Adolescencia (IUNDIA). Pobreza y exclusión social de la infancia en España. Madrid: Ministerio de Sanidad y Política Social; 2009. [https:// observatoriodelainfancia.vpsocial.gob.es/productos/pdf/ pobrezaExcInfEspana.pdf]. Accessed on date 04 Apr 2020. 36. DiPietro JA, Voegtline KM. The gestational foundation of sex differences in development and vulnerability. Neuroscience. 2017;342:4–20. 37. Nelson LP, Gold JI. Posttraumatic stress disorder in children and their parents following admission to the pediatric intensive care unit: a review. Pediatr Crit Care Med. 2012;13:338–47. 38. Case A, Lubotsky D, Paxson C. Economic status and health in childhood: the origin of the gradient. Am Econ Rev. 2002;92:1308–34. 39. González-Alzaga B, Lacasaña M, Aguilar-Garduño C, Rodríguez-Barranco M, Ballester F, Rebagliato M, et al. A systematic review of neurodevelopmental effects of prenatal and postnatal organophosphate pesticide exposure. Toxicol Lett. 2014;230:104–21. 40. Kelly R, Ramaiah SM, Sheridan H, Cruickshank H, Rudnicka M, Kissack C, et al. Dose-dependent relationship between acidosis at birth and likelihood of death or cerebral palsy. Arch Dis Child Fetal Neonatal Ed. 2018;103:F567–72.