Bayesian modelling of imperfect ascertainment methods in cancer studies
Bibliographic record
Abstract
Tumour registry linkage, chart review and patient self-report are all commonly used ascertainment methods in cancer epidemiology. These methods are used for estimating the incidence or prevalence of different cancer types in a population, and for investigating the effects of possible risk factors for cancer. Tumour registry linkage is often treated as a gold standard, but in fact none of these methods is error free, and failure to adjust for imperfect ascertainment can lead to biased estimates. This is true both if the goal of the study is to estimate the properties of each ascertainment type, or if it is to estimate cancer incidence or prevalence from one or more of these methods. Although rarely applied in the literature to date, when cancer is ascertained by three or more methods, standard latent class models can be used to estimate cancer incidence or prevalence while adjusting for the estimated imperfect sensitivities and specificities of each ascertainment method. These models, however, do not account for variations in these properties across different cancer sites. To address this problem, we extend latent class methodology to include a hierarchical component, which accommodates different ascertainment properties across cancer sites. We apply our model to a data set of 169 lupus patients with three ascertainment methods and eight cancer types. This allows us to estimate the properties of each ascertainment method without assuming any to be a gold standard, and to calculate a standardized incidence ratio for cancer for lupus patients compared to the general population. As our data set is small, we also illustrate the effects as more data become available. We show that our model produces parameter estimates that are substantially different from the currently most popular method of ascertainment, which uses tumour registry data alone.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.107 | 0.247 |
| Meta-epidemiology (narrow) | 0.002 | 0.004 |
| Meta-epidemiology (broad) | 0.005 | 0.005 |
| Bibliometrics | 0.005 | 0.007 |
| Science and technology studies | 0.002 | 0.005 |
| Scholarly communication | 0.007 | 0.007 |
| Open science | 0.009 | 0.005 |
| Research integrity | 0.006 | 0.007 |
| Insufficient payload (model declined to judge) | 0.007 | 0.001 |
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 source (direct Gemma or distilled Codex), 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".