A review of the use of propensity score methods with multiple treatment groups in the general internal medicine literature
Bibliographic record
Abstract
BACKGROUND: Propensity score (PS) methods with two treatment groups (e.g., treated vs. control) is a well-established technique for reducing the effects of confounding in nonrandomized studies. However, researchers are often interested in comparing multiple interventions. PS methods have been modified to incorporate multiple exposures. We described available techniques for PS methods in multicategory exposures (≥3 groups) and examined their use in the medical literature. METHODS: A comprehensive search was conducted for studies published in PubMed, Embase, Google Scholar, and Web of Science until February 27, 2023. We included studies using PS methods for multiple groups in general internal medicine research. RESULTS: The literature search yielded 4088 studies (2616 from PubMed, 86 from Embase, 85 from Google Scholar, 1671 from Web of Science, five from other sources). In total, 264 studies using PS method for multiple groups were identified; 61 studies were on general internal medicine topics and included. The most commonly used method was that of McCaffrey et al., which was used in 26 studies (43%), where the Toolkit for Weighting and Analysis of Nonequivalent Groups (TWANG) method and corresponding inverse probabilities of treatment weights were estimated via generalized boosted models. The next most commonly used method was pairwise propensity-matched comparisons, which was used in 20 studies (33%). The method by Imbens et al. using a generalized propensity score was implemented in six studies (10%). Four studies (7%) used a conditional probability of being in a particular group given a set of observed baseline covariates where a multiple propensity score was estimated using a non-parsimonious multinomial logistic regression model. Four studies (7%) used a technique that estimates generalized propensity scores and then creates 1:1:1 matched sets, and one study (2%) used the matching weight method. CONCLUSIONS: Many propensity score methods for multiple groups have been adopted in the literature. The TWANG method is the most commonly used method in the general medical literature.
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 | 0.029 | 0.097 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.005 | 0.007 |
| Bibliometrics | 0.035 | 0.035 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.004 | 0.004 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.005 | 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".