Developing a prediction model to identify people with severe mental illness without regular contact to their GP - a study based on data from the Danish national registers
Research output: Contribution to journal › Journal article › Research › peer-review
Standard
Developing a prediction model to identify people with severe mental illness without regular contact to their GP - a study based on data from the Danish national registers. / Naesager, Astrid Helene Deleuran; Damgaard, Sofie Norgil; Rozing, Maarten Pieter; Siersma, Volkert; Møller, Anne; Tranberg, Katrine.
In: BMC Psychiatry, Vol. 24, No. 1, 301, 2024.Research output: Contribution to journal › Journal article › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - JOUR
T1 - Developing a prediction model to identify people with severe mental illness without regular contact to their GP - a study based on data from the Danish national registers
AU - Naesager, Astrid Helene Deleuran
AU - Damgaard, Sofie Norgil
AU - Rozing, Maarten Pieter
AU - Siersma, Volkert
AU - Møller, Anne
AU - Tranberg, Katrine
N1 - © 2024. The Author(s).
PY - 2024
Y1 - 2024
N2 - INTRODUCTION: People with severe mental illness (SMI) face a higher risk of premature mortality due to physical morbidity compared to the general population. Establishing regular contact with a general practitioner (GP) can mitigate this risk, yet barriers to healthcare access persist. Population initiatives to overcome these barriers require efficient identification of those persons in need.OBJECTIVE: To develop a predictive model to identify persons with SMI not attending a GP regularly.METHOD: For individuals with psychotic disorder, bipolar disorder, or severe depression between 2011 and 2016 (n = 48,804), GP contacts from 2016 to 2018 were retrieved. Two logistic regression models using demographic and clinical data from Danish national registers predicted severe mental illness without GP contact. Model 1 retained significant main effect variables, while Model 2 included significant bivariate interactions. Goodness-of-fit and discriminating ability were evaluated using Hosmer-Lemeshow (HL) test and area under the receiver operating characteristic curve (AUC), respectively, via cross-validation.RESULTS: The simple model retained 11 main effects, while the expanded model included 13 main effects and 10 bivariate interactions after backward elimination. HL tests were non-significant for both models (p = 0.50 for the simple model and p = 0.68 for the extended model). Their respective AUC values were 0.789 and 0.790.CONCLUSION: Leveraging Danish national register data, we developed two predictive models to identify SMI individuals without GP contact. The extended model had slightly better model performance than the simple model. Our study may help to identify persons with SMI not engaging with primary care which could enhance health and treatment outcomes in this group.
AB - INTRODUCTION: People with severe mental illness (SMI) face a higher risk of premature mortality due to physical morbidity compared to the general population. Establishing regular contact with a general practitioner (GP) can mitigate this risk, yet barriers to healthcare access persist. Population initiatives to overcome these barriers require efficient identification of those persons in need.OBJECTIVE: To develop a predictive model to identify persons with SMI not attending a GP regularly.METHOD: For individuals with psychotic disorder, bipolar disorder, or severe depression between 2011 and 2016 (n = 48,804), GP contacts from 2016 to 2018 were retrieved. Two logistic regression models using demographic and clinical data from Danish national registers predicted severe mental illness without GP contact. Model 1 retained significant main effect variables, while Model 2 included significant bivariate interactions. Goodness-of-fit and discriminating ability were evaluated using Hosmer-Lemeshow (HL) test and area under the receiver operating characteristic curve (AUC), respectively, via cross-validation.RESULTS: The simple model retained 11 main effects, while the expanded model included 13 main effects and 10 bivariate interactions after backward elimination. HL tests were non-significant for both models (p = 0.50 for the simple model and p = 0.68 for the extended model). Their respective AUC values were 0.789 and 0.790.CONCLUSION: Leveraging Danish national register data, we developed two predictive models to identify SMI individuals without GP contact. The extended model had slightly better model performance than the simple model. Our study may help to identify persons with SMI not engaging with primary care which could enhance health and treatment outcomes in this group.
U2 - 10.1186/s12888-024-05743-x
DO - 10.1186/s12888-024-05743-x
M3 - Journal article
C2 - 38654257
VL - 24
JO - B M C Psychiatry
JF - B M C Psychiatry
SN - 1471-244X
IS - 1
M1 - 301
ER -
ID: 389578144