- Title Information
- Title
- A Computationally Efficient Algorithm for Producing Risk Score Models with Applications to Tuberculosis Diagnosis and Treatment Adherence
- Type of Resource (primo)
- dissertations
- Name:
Personal
- Name Part
- Eglinton, Hannah Jones
- Role
- Role Term:
Text
- creator
- Name:
Personal
- Name Part
- Paul, Alice
- Role
- Role Term:
Text
- Advisor
- Name:
Personal
- Name Part
- Hogan, Joseph
- Role
- Role Term:
Text
- Reader
- Name:
Corporate
- Name Part
- Brown University. Department of Biostatistics
- Role
- Role Term:
Text
- sponsor
- Origin Information
- Copyright Date
- 2024
- Physical Description
- Extent
- 3, 50 p.
- digitalOrigin
- born digital
- Note:
thesis
- Thesis (Sc. M.)--Brown University, 2024
- Genre (aat)
- theses
- Abstract
- Risk score models are simple scoring systems that map patient characteristics to the probability of an outcome occurring. These models are popular with clinicians because they are easy to memorize and can be quickly calculated by hand. Risk score models can be created by rounding the estimated coefficients from a logistic regression model, though rounding can reduce the performance of the models. We introduce a new cyclical coordinate descent algorithm to estimate integer risk score models, expanding on recent work that has aimed to directly solve for the maximum likelihood with integer constraints. By offering an associated R package, we aim to foster wider accessibility and utilization in the medical research community. In a simulation study, our algorithm demonstrates comparable performance to the current state-of-the-art methods while being substantially more efficient. Further, we highlight our method with two applications in tuberculosis (TB) research. First, we develop a risk score model for TB diagnosis in sub-Saharan Africa that shows higher validation AUC than previous rounding methods. Second, we develop a novel model for TB treatment non-adherence of adolescents in Peru. Our risk score model identifies key characteristics influencing non-adherence, aligning with previous qualitative research findings. This study showcases the effectiveness and efficiency of our algorithm in constructing integer risk score models.
- Subject (fast)
(authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/01158499")
- Topic
- Tuberculosis
- Subject
- Topic
- Interpretable Machine Learning
- Subject
- Topic
- Risk Score Models
- Language
- Language Term (ISO639-2B)
- English
- Record Information
- Record Content Source (marcorg)
- RPB
- Record Creation Date
(encoding="iso8601")
- 20240507