By Sarah Nørgaard
Chariman: Zorana Jovanovic Andersen
Opponent: Urs Frey
Opponent: Anders Gorm Pedersen
Childhood illness is very common in otherwise healthy children, particularly in the first 3 years of life. The major part of illness is simple infections that are mild and easily treated, but due to their high frequency, they still constitute a sizeable economic burden to society due to health care utilization and parents’ work absentee. Furthermore, since the immune system is primed in early life there is growing interest in the potential long term effects of early life infections and illness.
The overall objective of this thesis was to describe childhood health on three different levels. 1) the symptoms experienced by the child 2) the costs of illness to society and 3) the consequences of childhood asthma-like symptoms in adulthood.
The work presented in this thesis is based on two cohorts, COPSAC2010 and COPSAC1960. The COPSAC2010 is a prospective mother-child population based cohort of 700 children born in eastern Denmark 2008-2011. Costs of contact with health care providers were required from the Danish National Registries. The COPSAC1960 is recruited from The Copenhagen Perinatal Cohort consisting of children born at Rigshospitalet in Copenhagen from 1959 to 1961. This cohort was linked with registry information of in- and out-patient admissions and prescription medication of respiratory diseases.
In paper I, we investigated 10 common symptoms of illness (cold, cough, breathlessness, wheeze, pneumonia, fever, gastric infections, inflammation of the throat, inflammation of the ear, and eczema) the first three years of life. Episodes of illness were defined and described, a factor model was developed to describe the variation between symptoms the second year of life. A graphical partial correlation network was used to describe correlations over time and between symptoms and also how early exposures such as preterm delivery, parental disease etc. were correlated to this symptom network.
In paper II, we developed the method that was used to investigate the partial correlations in paper I. We used kernels to harmonize heterogeneous data including binary risk factors, longitudinal episode records as well as survival outcomes. We proposed a two-step algorithm to find sparse partial correlation networks with an inner and an outer model illustrated with graphical network based on unfolded kernels and L1 norm penalized inverse covariance estimation.
In paper III, we examined the costs of illness including parent absenteeism the first three years of life. The direct costs were estimated as the sum of all costs of contact with health care providers and the total costs were estimated as the sum of the direct costs and the costs of parent absenteeism.
In paper IV, we investigated how asthma-like symptoms before the age of six years were associated with hospital admissions and prescription medicine for respiratory diseases in adulthood in the COPSAC1960 cohort.
In conclusion, his thesis shows the etiology of common symptoms of illness the first three years of life. Important findings are that respiratory symptoms and eczema are not correlated, and that preterm birth and low gestational age associates with a more than two fold higher number of days with cough. With these analyses we also proposed a new method of exploratory analyses of heterogeneous data via kernelization. The median overall costs of childhood illness were about 11,000 Euros (2017 price level) per child throughout the first three years of life. Gestational age was consequently also the strongest predictor for higher costs, together with cesarean delivery which was associated with 30-45% higher costs of illness. Finally, we showed that asthma-like symptoms the first six years of life were associated with a 1.88 fold amount of hospital admissions and a more than two fold higher use of prescription of long-acting muscarinic antagonists in adulthood.