Abstract 102: Impact of Using Older Data on the Accuracy of Cardiovascular Risk Scores
Bibliographic record
Abstract
Background: Recent scores to predict atherosclerotic cardiovascular disease, ASCVD, have been found inaccurate, with some concern that risk scores become inaccurate with time, as changing demographics may also change ASCVD risk. A partial solution to the timeliness problem could be creating risk scores using electronic health records (EHR). EHR-based risk scores can be easily updated and tested for changes over time. They can also use more variables than traditional risk scores. This could capture clinical change that would otherwise be missed. We hypothesized that ASCVD risk scores change over time, but this could be minimized with more robust risk scores. To test this, we looked at change of two EHR-based risk score over three follow-up periods and using different statistical techniques to design the risk scores. Methods: Data sources: VA national data linked to Medicare and the National Death Index. Population: 3 overlapping cohorts from 2002, 2006, and 2009. Each consisted of all active VA patients aged 45-80 who had no documented history of CVD, clinical heart failure or loop diuretic use at baseline. Prediction models 1. VARS-ASCVD: uses the same variables as traditional risk scores, but all variables were re-calibrated to our population. 2. VARS-EHR: Uses 41 predictor variables and more interaction effects. Outcome variables: First occurrence of fatal or nonfatal ASCVD during 5 years of follow-up. Analysis: We looked at the accuracy of risk scores developed in 2002 on patients in 2006 and 2009. The discrimination of the risk scores (the ability to distinguish between those who do and do not develop an event), was evaluated with C-statistisic. The calibration (how closely the predicted probabilities reflect true risk) was evaluated with the Hosmer-Lemeshow Goodness of Fit statistic (GoF). Results: Each cohort had at least 1.4 million participants. Between the 3 cohorts the rate of diabetes mellitus increased from 201% to 27% and statin use increased from 25% to 45% of the population. The VARS-ASCVD risk scores for men developed in 2002 had the same discrimination of 0.67 in 2006 and 2009, but in women fell from 0.77 to 0.72 then increased to 0.74. The goodness of fit worsened. Using the VARS-EHR model, discrimination stayed similar in men and women. The GOF worsened, but by substantially less. Conclusions: ASCVD risk prediction tools become poorly calibrated over fairly short time periods. For effective use, they must be updated regularly.
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 | 0.006 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.002 | 0.004 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 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".