Health Plan Switching and Attrition Bias in the RAND Health Insurance Experiment
Bibliographic record
Abstract
One of the most influential “lessons” of the RAND Health Insurance Experiment (HIE) is that cost sharing can reduce hospitalizations by about a quarter, with no effect on health for the average adult. In an earlier paper in this journal, I suggested that a portion of this reduction is due to participants becoming ill and dropping out of the experiment in order to switch to their preexperiment insurance coverage and thus avoid paying the cost-sharing amount. The sixteenfold higher voluntary attrition rate in the cost-sharing arms provides compelling evidence in support of this alternative explanation. Evidence is also provided by the finding in Manning, Duan, Keeler (1993) that the predicted number of hospitalizations among those who dropped out of the coinsurance arms was significantly greater by 34.5 percent than the actual number of hospitalizations, suggesting that participants anticipate hospitalizations and leave the experiment before incurring the cost-sharing payment. Still more evidence is provided by the finding that those (cost-sharing) participants with greater incomes, instead of being more likely to be hospitalized, as greater income usually implies, were less likely to be hospitalized than poor participants. This finding is consistent with their having better preexperiment insurance coverage than poor participants and therefore being more likely to have an incentive to drop out. This inpatient attrition bias makes it dangerous to rely on this lesson of the HIE, because it is not clear that hospitalizations were actually reduced by one-quarter, much less that if such a reduction actually had occurred, there would be no health consequences.
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 | 0.004 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| 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".