Chronic Respiratory Diseases
Conditions
Keywords
lung function, trajectory, predictions
Brief summary
Currently, it remains unclear how to manage serial lung function measurements in a clinical setting. The investigators aimed to tackle this problem by developing a machine learning (ML) model that can accurately predict population and individual lung function trajectories. These predictions would enable the investigators to identify positive or negative deviations, thereby revealing unexpected disease patterns. A prospective validation is needed that includes data on mortality, hospitalisations, emergency-room visits and patient-reported outcomes. Within this study, the goal is to validate the ML model with the data collected from this observational study.
Detailed description
The objective of this study is to explore the clinical value of models predicting longitudinal lung function patterns in individuals with chronic respiratory diseases across Belgium. 1. The investigators will assess the accuracy of individualised lung function prediction models in a multicentre lung function dataset with prospective clinical and lung function follow-up. 2. The investigators will evaluate important health outcomes, step-up in care, patient-reported outcomes in individuals identified with an expected and unexpected observed trajectory as compared to the predicted population and individualised trajectory. The hypothesis is that patients with an unexpected decline in lung function will have worse health outcomes, such as a higher mortality rate and more hospitalisations, compared to patients with an expected lung function pattern. The investigators hypothesise to observe better health outcomes and lower mortality rates in patients with an unexpectedly positive lung function evolution compared to patients with an expected negative lung function pattern. Individuals will be recruited from 4 Belgian Hospitals (UZ Leuven, UZ Antwerpen, AZ Delta, ZOL Genk). Based on the annual rate of pulmonary function testing in these hospitals, a sample size of 1.000 participants per centre is anticipated within one year of inclusions, resulting in a total sample size of 4.000 patients. All available historical lung function data of included individuals will be retrieved from the individuals medical file. Additionally, the individual will be prospectively followed for 2 years where all lung function data will be collected.
Interventions
None listed
Sponsors
KU Leuven
AZ Delta
University Hospital, Antwerp
Ziekenhuis Oost-Limburg
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Above 18 years old * Diagnosed with a chronic respiratory disease and followed up in one of the participating Belgian hospitals * Performed a complete lung function test (spirometry, body plethysmography and diffusion capacity) at baseline * Have at least 3 historical spirometry measurements over a minimal time window of 2 years prior to inclusion * Planned routine follow-up within standard clinical care in one of the participating hospitals
Exclusion criteria
* Patients who have had a lung transplantation * Patients not being able to give consent to participate
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Accuracy of lung function predictions (FEV1) | at 1 and 2-year follow-up | Proportion of correct and incorrect FEV1 predictions compared to the observed measure |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Differences in clinical outcomes between correct and incorrect lung function predictions (FEV1) | at 1 and 2-year follow-up | Differences between patients with correct and incorrect individual lung function predictions for FEV1 on clinical endpoints (such as mortality, hospitalisations, frailty, health status and step-up in care |
| Accuracy of lung function predictions | at 1 and 2-year follow-up | Proportion of correct and incorrect lung function predictions (FVC, TLC, RV/TLC, DLCO) compared to the observed measure |
| Differences in clinical outcomes between correct and incorrect lung function predictions | at 1 and 2-year follow-up | Differences between patients with correct and incorrect individual lung function predictions for FVC, TLC, RV/TLC, DLCO on clinical endpoints (such as mortality, hospitalisations, frailty, health status and step-up in care) |
| Identifying the minimal needed to make predictions | after 2 years | Minimal number of tests/length of follow-up required for optimal predictions |
| Performance of ML-based predictions compared to linear regression analysis | at 1 and 2-year follow-up | Comparison of the ML-based predictions for individual and population lung function changes with predictions based on linear regression on individual historical data |
| Overall description of population | baseline, 1 and 2-year follow-up | Sociodemographic information, health status, comorbidities, frailty, disease labels, interventions and prognosis of individuals with a chronic respiratory disease |
Countries
Belgium
Contacts
CONTACTMarieke Wuyts
PRINCIPAL_INVESTIGATORWim Janssens
UZ/KU Leuven
Outcome results
None listed