Taking the Next Steps in Epidemiological Strategies to Predict Childhood Respiratory Disease

Berger, Daria Olena (2023). Taking the Next Steps in Epidemiological Strategies to Predict Childhood Respiratory Disease (Unpublished). (Dissertation, Universität Bern, Graduate School for Health Sciences)

Text Berger_Daria_PhD_Thesis.pdf - Other Restricted to registered users only Available under License BORIS Standard License. Download (15MB) | Request a copy

Epidemiology is the study of health and disease states1. Several epidemiological strategies exist that help researchers to predict the risk of health and disease states in various populations. These strategies, such as clinical prediction modelling and population screening, have applications in early diagnosis which leads to improvements in short and long-term health outcomes, reduction in inappropriate treatment and avoidance ofmisdiagnoses2–4.

This thesis adds new knowledge to two specific epidemiological strategies used in the prediction of childhood respiratory disease, namely: clinical prediction modelling in childhood asthma and newborn screening for cystic fibrosis.

Clinical prediction modelling helps researchers and clinicians to identify children at risk for future asthma. Modelling can also predict asthma control and the risk of future asthma attacks. However, despite the availability of the TRIPOD guidelines for reporting of clinical prediction models since 20155, few prediction models have been adequately developed and validated. Additionally, even fewer have been externally validated in a paediatric respiratory clinic setting to assess potential clinical use.

The Predicting Asthma Risk in Children (PARC) tool is one clinical prediction model that aims to predict preschool aged children at future risk of asthma. This model has been developed and internally validated using advanced statistical methods and externally validated in several population-based cohorts. However, it has yet to be externally validated in a clinical setting, where prediction tools can be used to inform clinician decision making. Additionally, clinical prediction models for asthma attacks and asthma control in children aged >5 years with a diagnosis of asthma have similar limitations in that few have been adequately developed and validated and none have been validated in clinical settings.

Screening has a well-established role in identifying populations at risk of developing a certain disease. In epidemiological terms, screening is considered to a play a role in the early detection and secondary prevention of disease. Newborn screening for cystic fibrosis, the most common autosomal recessive disease to affect children globally, has been increasingly accepted as a part of the gold standard of care for people with this severe, debilitating disease2. However, cystic fibrosis newborn screening (CF NBS)programmes need constant evaluation to ensure that their performance is optimised6.

AIMS
The primary aim of my research was to perform an external validation of the PARC tool in a paediatric respiratory outpatient clinic cohort, the Swiss Paediatric Airway Cohort (SPAC), to assess performance and consider clinical applications. A secondary project for which I was a co-author, aimed to examine the SPAC for predictors that could be used in the development of a model for asthma attacks and asthma control in children with a diagnosis of asthma aged over 5 years old. The secondary aim of my research was to evaluate the CF NBS for Switzerland and provide an update on CF NBS across Europe.

RESULTS

CLINICAL PREDICTION MODELS: PUBLICATION I AND II
I contribute new knowledge by performing an external validation of a childhood asthma clinical prediction tool, the Predicting Asthma Risk in Children (PARC) tool in a clinical cohort, the Swiss Paediatric Airway Cohort (SPAC)(Publication I). The aim of this study was to externally validate the PARC tool in a clinical cohort. In summary, we found that the PARC tool performed less well in a clinical setting as assessed by area under the receiver operating statistic curve (AUC), scaled Brier’s score and Nagelkerke’s R2when compared to the original performance of the model in its developmental cohort at a PARC score of 4,while the positive predictive value (PPV), negative predictive value (NPV) and sensitivity and specific also performed less well. The output of this work has been published as a 1st author manuscript in Paediatric Pulmonology.

Publication II, (Ardura-Garcia, Cet. al.).contributes new knowledge on predictors for asthma control and attacks and is included in this thesis as an example to describe the development of a clinical prediction model using a paediatric respiratory clinic cohort, the SPAC. The aims of this work were to assess the predictors for suboptimal asthma control and asthma attacks using a clinical cohort. Results of this work indicated that predictors for asthma control and asthma attacks were different and need to be considered when designing future clinical prediction models. The output of this work has been submitted for publication as a co-author manuscript in Clinical and Experimental Allergy and is currently under revision.

CYSTIC FIBROSIS NEWBORN SCREENING:PUBLICATIONS III, IVAND V
In Publications III and IV I evaluate the cystic fibrosis newborn screening programme for Switzerland over a 10-year period, contributing new longitudinal knowledge on its performance. The aims of this study were to evaluate the performance of the algorithm over its10-year period as well as consider strategies to optimise its performance. The evaluation showed that the positive predictive value of the screening algorithm has been steadily declining despite a stable incidence rate of cystic fibrosis cases. We suggest strategies to improve the PPV for the future screening algorithm. The output of this evaluation is a report that has been published for the Federal Office of Public Health (BAG) in Switzerland for which I am the 1st author. A manuscript with an updated analysis is currently in the process of being written for which I am currently the planned 1st author. A German language version of this report has also been published on the Swiss Medical Forum for which I am listed as a co-author.

In Publication V I present results of a survey and evaluate the performance of cystic fibrosis newborn screening programmes across national and regional sites in Europe, contributing new knowledge on the current state of CF NBS. The aims of this study were to perform a survey of the current state of CF NBS in Europe and assess performance. The results of this work indicate that a growing number of sites across Europe are using CF NBS but that there is a large range of screening strategies, data collection and data maintenance methods. The output of this evaluation is a shared 1st co-author manuscript that has been published in the Journal of Cystic Fibrosis.

I conclude that epidemiological strategies continue to play an important role in childhood respiratory disease risk prediction. In my work on clinical prediction models in asthma, we conclude that external validation in a clinical cohort performs less well than in a population-based setting but may have applications in clinical practice. In Publication II we show that asthma attack and asthma control predictors differ. Evaluation of CF NBS in Swiss and European contexts found that programmes must continue to optimise their performance and focus on rigorous database management to be able to accurately assess key CF NBS performance parameters

Interest & Impact

Downloads

5 since deposited on 17 Jul 2023
5 in the past 12 months

Citations

5 Citations in Web of Science ®
6 Citations in Scopus

Search

in Google Scholar™

Services

Actions (login required)

Edit item

Actions (login required)

Edit item