Projektdetaljer
Beskrivelse
Background Specific diagnoses are referred to as ambulatory care sensitive conditions (ACSC) where hospital admission may be preventable.
Objectives We aimed to develop and internally validate machine learning (ML) models for prediction of hospital contacts due to ACSC within 365 days among individuals aged 50+. We focus on the model’s ability to identify those that had hospital contacts with ACSC and target the prognostic models to a sensitivity of at least 80 %.
Methods We will use the Lolland-Falster Health Study (LOFUS) and include 10,154 individuals aged 50+ participating between 2016 and 2020. LOFUS will be linked to six Danish national registries. We will create 134 features covering information on lifestyle, sociodemographic factors, laboratory results, diagnoses, pharmacological treatment, use of hospital and general practitioners. Extreme gradient-boosting (XGBoost) and Elastic Net regularization (EN) will be trained to predict ACSC in 80% of the population. We will use two sets of data for model construction 1) Data from LOFUS and the national registries and 2) Data from the national registries. The models trained among individuals aged 50+ will be tested within the same age group (20% of the population), and subsequently in a test set only including individuals aged 65+.
Objectives We aimed to develop and internally validate machine learning (ML) models for prediction of hospital contacts due to ACSC within 365 days among individuals aged 50+. We focus on the model’s ability to identify those that had hospital contacts with ACSC and target the prognostic models to a sensitivity of at least 80 %.
Methods We will use the Lolland-Falster Health Study (LOFUS) and include 10,154 individuals aged 50+ participating between 2016 and 2020. LOFUS will be linked to six Danish national registries. We will create 134 features covering information on lifestyle, sociodemographic factors, laboratory results, diagnoses, pharmacological treatment, use of hospital and general practitioners. Extreme gradient-boosting (XGBoost) and Elastic Net regularization (EN) will be trained to predict ACSC in 80% of the population. We will use two sets of data for model construction 1) Data from LOFUS and the national registries and 2) Data from the national registries. The models trained among individuals aged 50+ will be tested within the same age group (20% of the population), and subsequently in a test set only including individuals aged 65+.
Status | Igangværende |
---|---|
Effektiv start/slut dato | 01/01/24 → … |
Fingerprint
Udforsk forskningsemnerne, som dette projekt berører. Disse etiketter er oprettet på grundlag af de underliggende bevillinger/legater. Sammen danner de et unikt fingerprint.