Identifying Race/Ethnicity Data via Natural Language Processing Among Women in a Uterine Fibroid Cohort Study
Bibliographic record
Abstract
Background/Aims: Uterine fibroids are associated with morbidity including abnormal bleeding, anemia, pelvic/bladder symptoms and adverse reproductive outcomes. Symptomatic fibroids may affect 25% of women in their late 40s. Race is among the most consistent risk factors known. The Uterine Fibroid Study aims to use automated data to estimate fibroid incidence rates/trends during 2005–2014 in a retrospective cohort of women at Group Health. Race/ethnicity captured from automated structured data has improved yet remains incomplete, particularly with use of retrospective data. Methods: The study included women 18–65 years old without hysterectomy, continuously enrolled with evidence of encounter in 3 years before study entry. Incidence estimates required absence of fibroid history. We collected fibroid diagnoses, demographics and other data from the Virtual Date Warehouse (VDW). VDW demographic race/ethnicity data is sourced from data collected from Group Practice patients, at time of encounter, via entry in the electronic health record. Additionally, Group Health collects race/ethnicity data from breast cancer screening program and tumor registry data. To complement traditional structured race/ethnicity data from VDW, we augmented with race/ethnicity extracted from free-text clinical notes via natural language processing (NLP). Our NLP system used a rule-based dictionary look-up approach to identify common terms used to describe patient race/ethnicity and custom rules to disambiguate race/ethnicity terms that also have other clinical meanings (e.g. the term “white” in “54-year-old white female” as opposed to “Her white blood cell count improved”). We conducted a partial validation of the NLP system in a sample of patients with known structured race/ethnicity data. Results: Prior to amending race/ethnicity data with NLP, in the cohort of 277,821 women, 37.4% had race/ethnicity unknown. Fibroid incidence rates (per 10,000 person-years) were 156 for Hispanics, 133 for whites, 265 for African-Americans, 152 for Asian/Pacific Islanders and 108 for unknown race. NLP work on identifying race/ethnicity in the unknown race group is ongoing and results are pending. Conclusion: Race/ethnicity is an important risk factor for a number of conditions, including uterine fibroids. Improving capture of race/ethnicity from available automated data sources potentially could improve accuracy of research findings and enhance patient care by providing a better understanding of the burden of disease in subgroups of affected patients.
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 | 0.007 | 0.020 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 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".