Unsupervised Item Response Theory Models for Evaluating Differential Item Functioning in Epilepsy

Arimoro, Olayinka Imisioluwa

Unsupervised Item Response Theory Models for Evaluating Differential Item Functioning in Epilepsy

dc.contributor.advisor	Sajobi, Tolulope
dc.contributor.advisor	Josephson, Colin
dc.contributor.author	Arimoro, Olayinka Imisioluwa
dc.contributor.committeemember	Lix, Lisa
dc.contributor.committeemember	Patten, Scott
dc.contributor.committeemember	Wiebe, Samuel
dc.contributor.committeemember	Kim, Sunmee
dc.contributor.committeemember	Bobawsky, Kirsten
dc.date	2024-11
dc.date.accessioned	2024-06-10T16:30:26Z
dc.date.available	2024-06-10T16:30:26Z
dc.date.issued	2024-06-07
dc.description.abstract	Depression is a highly prevalent comorbid condition among patients with epilepsy, and it is associated with poorer quality of life. In epilepsy management, validated patient-reported multi-item questionnaires are useful for screening for depression. However, the validity of these screening instruments may be threatened by heterogeneity in how patients with the same underlying health status interpret and respond to questions about their health and quality of life, a phenomenon also known as differential item functioning (DIF). Heterogeneity may arise due to differences in an individual's demographic, disease, clinical characteristics, or experience with treatment or quality of care received, or other unknown factors. The presence of DIF in patient-reported outcome measures (PROMs) could lead to measurement biases that threaten the validity of inferences on PROM scores to inform clinical and healthcare decisions. Advanced statistical and latent variable methods have been developed to test for DIF. However, most of them assume that the individual characteristics associated with DIF are known a priori. These methods are not suitable to test for DIF when there are no clear prior hypotheses. To address this challenge, polytomous tree-based item response theory (IRTree) has been developed to test for DIF when the patient characteristics associated with DIF are not known a priori. This study investigates the statistical properties of the polytomous IRTree model that combines the features of a latent variable model and model-based recursive partitioning methods to test for DIF in PROM data when patient characteristics associated with DIF are not known a priori. Specifically, the objectives are to (1) apply the polytomous IRTree model to test for DIF in depression PROMs in people with epilepsy and (2) evaluate the Type I error and statistical power rates of the polytomous IRTree model for testing for DIF under a variety of data analytic simulation conditions. The study's objectives were addressed using a combination of computer simulations and real-world data analysis. Data were obtained from 1,576 adults with epilepsy seen in the outpatient clinics of the Calgary Epilepsy Program in Calgary, Alberta. The polytomous IRTree model was used to test for DIF in Neurological Disorders Depression Inventory for Epilepsy scale items. The IRTree model identified four subgroups defined by interactions among age, sex, and employment status as exhibiting DIF. Subgroup 1 were unemployed patients ≤26 years old, subgroup 2 were unemployed patients > 26 years, subgroup 3 were employed females, while subgroup 4 were employed male patients. Monte Carlo simulations were used to assess the statistical properties of the polytomous IRTree model with respect to Type I error and statistical power to detect DIF. The IRTree model provided good control of Type I error and high statistical power to detect DIF especially when the Bonferroni correction was applied, sample size was reasonable (minimum sample size of 500), and the number of nuisance variables were minimal. This work contributed to the literature on the performance of IRTree models. In conclusion, the polytomous IRTree model with Bonferroni correction is a promising method for evaluating DIF in potentially heterogeneous populations. We recommend that the selection of potential explanatory variables for IRTree is guided by an understanding of the response processes of the respondents in the population being studied.
dc.identifier.citation	Arimoro, O. I. (2024). Unsupervised item response theory models for evaluating differential item functioning in epilepsy (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.
dc.identifier.uri	https://hdl.handle.net/1880/118923
dc.identifier.uri	https://doi.org/10.11575/PRISM/46520
dc.language.iso	en
dc.publisher.faculty	Graduate Studies
dc.publisher.institution	University of Calgary
dc.rights	University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission.
dc.subject	differential item functioning
dc.subject	patient-reported outcome measures
dc.subject	recursive partitioning
dc.subject	epilepsy
dc.subject	partial credit model
dc.subject.classification	Education--Health
dc.subject.classification	Biostatistics
dc.title	Unsupervised Item Response Theory Models for Evaluating Differential Item Functioning in Epilepsy
dc.type	master thesis
thesis.degree.discipline	Medicine – Community Health Sciences
thesis.degree.grantor	University of Calgary
thesis.degree.name	Master of Science (MSc)
ucalgary.thesis.accesssetbystudent	I do not require a thesis withhold – my thesis will have open access and can be viewed and downloaded publicly as soon as possible.