Methods for inference under Assessment Not at Random in electronic health records data
Bibliographic record
Abstract
Analyses of observational longitudinal data, such as those from electronic health records (EHR) databases, are becoming increasingly important in informing healthcare decision-making. However, analyzing EHR data can be challenging because the assessments are often irregularly spaced, with frequency and timing related to the patient's health status. Existing methods account for irregular assessment times under the assumption that assessment at any given time is conditionally independent of the outcome at that time, given the observed history -- known as Assessment at Random (AAR). However, there is a lack of methodology for handling scenarios where latent, potentially time-varying factors influence both the outcome and assessment processes, known as Assessment Not at Random (ANAR). This situation can occur in relapsing-remitting diseases, where patients may experience relapses between scheduled assessments and require as-needed assessments for urgent treatment. This thesis presents methodology for first characterizing the irregular assessment process and evaluating the plausibility of AAR, and then, for cases where there is evidence of departures from AAR, both semiparametric and parametric approaches to accommodate ANAR are proposed. The semiparametric approach involves positing a specific form of dependence between the unobserved outcome and assessment process, and having the strength of that relationship governed by a set of sensitivity parameters, which are varied over a grid of plausible values. The exploration of parametric approaches begins with examining when standard mixed effects models on the outcome alone are sufficient, and then shows how joint models of the outcome and assessment process can be specified when needed. These joint models accommodate a form of ANAR where the two processes are linked through subject-specific, time-invariant random effects. The last methodological component is a Bayesian joint model incorporating the relapse process is developed, which has the potential to provide useful prognostic information that is not available when using standard mixed effects models. The use of the proposed methodology is illustrated through an application to a real-world pediatric study on juvenile dermatomyositis.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.012 | 0.017 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.003 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".