Bayesian latent class analysis produced diagnostic accuracy estimates that were more interpretable than composite reference standards for extrapulmonary tuberculosis tests
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
BACKGROUND: Evaluating the accuracy of extrapulmonary tuberculosis (TB) tests is challenging due to lack of a gold standard. Latent class analysis (LCA), a statistical modeling approach, can adjust for reference tests' imperfect accuracies to produce less biased test accuracy estimates than those produced by commonly used methods like composite reference standards (CRSs). Our objective is to illustrate how Bayesian LCA can address the problem of an unavailable gold standard and demonstrate how it compares to using CRSs for extrapulmonary TB tests. METHODS: We re-analyzed a dataset of presumptive extrapulmonary TB cases in New Delhi, India, for three forms of extrapulmonary TB. Results were available for culture, smear microscopy, Xpert MTB/RIF, and a non-microbiological test, cytopathology/histopathology, or adenosine deaminase (ADA). A diagram was used to define assumed relationships between observed tests and underlying latent variables in the Bayesian LCA with input from an inter-disciplinary team. We compared the results to estimates obtained from a sequence of CRSs defined by increasing numbers of positive reference tests necessary for positive disease status. RESULTS: Data were available from 298, 388, and 230 individuals with presumptive TB lymphadenitis, meningitis, and pleuritis, respectively. Using Bayesian LCA, estimates were obtained for accuracy of all tests and for extrapulmonary TB prevalence. Xpert sensitivity neared that of culture for TB lymphadenitis and meningitis but was lower for TB pleuritis, and specificities of all microbiological tests approached 100%. Non-microbiological tests' sensitivities were high, but specificities were only moderate, preventing disease rule-in. CRSs' only provided estimates of Xpert and these varied widely per CRS definition. Accuracy of the CRSs also varied by definition, and no CRS was 100% accurate. CONCLUSION: Unlike CRSs, Bayesian LCA takes into account known information about test performance resulting in accuracy estimates that are easier to interpret. LCA should receive greater consideration for evaluating extrapulmonary TB diagnostic tests.
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.
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 | 0.006 | 0.074 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 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 it