Using Mixed Integer Programming for Matching in an Observational Study of Kidney Failure After Surgery
The DOI matches its Crossref record; 40 of the 40 references it cites were found in a registry or an index.
Abstract
This article presents a new method for optimal matching in observational studies based on mixed integer programming. Unlike widely used matching methods based on network algorithms, which attempt to achieve covariate balance by minimizing the total sum of distances between treated units and matched controls, this new method achieves covariate balance directly, either by minimizing both the total sum of distances and a weighted sum of specific measures of covariate imbalance, or by minimizing the total sum of distances while constraining the measures of imbalance to be less than or equal to certain tolerances. The inclusion of these extra terms in the objective function or the use of these additional constraints explicitly optimizes or constrains the criteria that will be used to evaluate the quality of the match. For example, the method minimizes or constrains differences in univariate moments, such as means, variances, and skewness; differences in multivariate moments, such as correlations between covariates; differences in quantiles; and differences in statistics, such as the Kolmogorov–Smirnov statistic, to minimize the differences in both location and shape of the empirical distributions of the treated units and matched controls. While balancing several of these measures, it is also possible to impose constraints for exact and near-exact matching, and fine and near-fine balance for more than one nominal covariate, whereas network algorithms can finely or near-finely balance only a single nominal covariate. From a practical standpoint, this method eliminates the guesswork involved in current optimal matching methods, and offers a controlled and systematic way of improving covariate balance by focusing the matching efforts on certain measures of covariate imbalance and their corresponding weights or tolerances. A matched case–control study of acute kidney injury after surgery among Medicare patients illustrates these features in detail. A new R package called mipmatch implements the method.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
The central role of the propensity score in observational studies for causal effects
ROSENBAUM, P. R., & RUBIN, D. B. (1983) · Biometrika · doi:10.1093/biomet/70.1.41 · Cited by 31,241
Statistical Methods for Rates and Proportions
Fleiss, J. L., Levin, B., & Paik, M. C. (2003) · Wiley Series in Probability and Statistics · doi:10.1002/0471445428 · Cited by 18,646
The Hungarian method for the assignment problem
Kuhn, H. W. (1955) · Naval Research Logistics Quarterly · doi:10.1002/nav.3800020109 · Cited by 13,120
Matching Methods for Causal Inference: A Review and a Look Forward
Stuart, E. A. (2010) · Statistical Science · doi:10.1214/09-sts313 · Cited by 5,478 · cited 9×
“We observe an improvement in the balance of most of the covariates; however, the means of eGFR, APACHE, risk, dysfunction, and failure are still quite far apart.”
InfluentialRosenbaum, P. R., & Rubin, D. B. (1985) · The American Statistician · doi:10.2307/2683903 · Cited by 4,774 · cited 1×
“Fourth, as pointed out by Rosenbaum and Rubin (1985), despite the complexity of some matching algorithms, very simple methods can be used to analyze the outcomes, and typical analysis of matched samples do not require the parametric assumptions of most regression methods (Lu et al. 2011).”
Journal of the American Statistical Association
Fildes, R. (1989) · International Journal of Forecasting · doi:10.1016/0169-2070(89)90106-4 · Cited by 3,874
Causal Inference without Balance Checking: Coarsened Exact Matching
Iacus, S. M., King, G., & Porro, G. (2012) · Political Analysis · doi:10.1093/pan/mpr013 · Cited by 3,563 · cited 1×
“Note that in the current example, there are no propensity scores because it is a case–control study (Joffe and Rosenbaum 1999).”
Assessing Kidney Function — Measured and Estimated Glomerular Filtration Rate
Stevens, L. A., Coresh, J., Greene, T., & Levey, A. S. (2006) · New England Journal of Medicine · doi:10.1056/nejmra054415 · Cited by 2,952 · cited 1×
“A matched case–control study of acute kidney injury after surgery among Medicare patients illustrates these features in detail.”
Large sample properties of matching estimators for average treatment effects
Abadie, A., & Imbens, G. W. (2006) · Econometrica · doi:10.1111/j.1468-0262.2006.00655.x · Cited by 2,908
Introduction to linear optimization
Bertsimas, D., & Tsitsiklis, J. N. (1997) · Cited by 2,705
Causal effects in nonexperimental studies: Reevaluating the evaluation of training programs
Dehejia, R. H., & Wahba, S. (1999) · Journal of the American Statistical Association · doi:10.1080/01621459.1999.10473858 · Cited by 2,395
Design of Observational Studies
Rosenbaum, P. R. (2010) · Springer Series in Statistics · doi:10.1007/978-1-4419-1213-8 · Cited by 1,150
Bayesian Nonparametric Modeling for Causal Inference
Hill, J. L. (2011) · Journal of Computational and Graphical Statistics · doi:10.1198/jcgs.2010.08162 · Cited by 1,139 · cited 1×
“The results in Hollander, Pledger, and Lin (1974) and Lam and Longnecker (1983) suggest that matching the treated and controls units closely on these covariates tends to make statistical tests less conservative.”
Gladwell, M. (2000) · Cited by 1,123
Diamond, A., & Sekhon, J. S. (2013) · Review of Economics and Statistics · doi:10.1162/rest_a_00318 · Cited by 1,102 · cited 1×
“The problem is that this iterative process usually involves a considerable amount of guesswork, and it is common to see that balancing some covariates worsens the balance of other covariates (Diamond and Sekhon 2005; Hill 2011; Iacus, King, and Porro 2012).”
Invited Commentary: Propensity Scores
Joffe, M. M., & Rosenbaum, P. R. (1999) · American Journal of Epidemiology · doi:10.1093/oxfordjournals.aje.a010011 · Cited by 1,048 · cited 1×
“Note that in the current example, there are no propensity scores because it is a case–control study (Joffe and Rosenbaum 1999).”
Causal Effects in Non-Experimental Studies: Re-Evaluating the Evaluation of Training Programs
Dehejia, R., & Wahba, S. (1998) · National Bureau of Economic Research · doi:10.3386/w6586 · Cited by 805
Wiley encyclopedia of operations research and management science
(2011) · Choice Reviews Online · doi:10.5860/choice.49-0052 · Cited by 781
Optimal Matching for Observational Studies
Rosenbaum, P. R. (1989) · Journal of the American Statistical Association · doi:10.1080/01621459.1989.10478868 · Cited by 715 · cited 2×
“For instance, with this method, it is possible to minimize or constrain differences in univariate moments, such as means, variances, and skewness; differences in multivariate moments, such as correlations between covariates; differences in quantiles; and differences in statistics, such as the K–”
Rubin, D. B. (1979) · Journal of the American Statistical Association · doi:10.1080/01621459.1979.10482513 · Cited by 705
Optimal Full Matching and Related Designs via Network Flows
Hansen, B. B., & Klopfer, S. O. (2006) · Journal of Computational and Graphical Statistics · doi:10.1198/106186006x137047 · Cited by 703 · cited 1×
“The problem is that this iterative process usually involves a considerable amount of guesswork, and it is common to see that balancing some covariates worsens the balance of other covariates (Diamond and Sekhon 2005; Hill 2011; Iacus, King, and Porro 2012).”
Full Matching in an Observational Study of Coaching for the SAT
Hansen, B. B. (2004) · Journal of the American Statistical Association · doi:10.1198/016214504000000647 · Cited by 605
The prognostic analogue of the propensity score
Hansen, B. B. (2008) · Biometrika · doi:10.1093/biomet/asn004 · Cited by 371 · cited 1×
“Sixth, using existing inferential techniques is straightforward if matching is done without replacement (Hansen and Klopfer 2006).”
A new algorithm for the assignment problem
Bertsekas, D. P. (1981) · Mathematical Programming · doi:10.1007/bf01584237 · Cited by 292 · cited 2×
“An often faster algorithm is the auction algorithm proposed by Bertsekas (1981).”
MIP: Theory and Practice — Closing the Gap
Bixby, E. R., Fenelon, M., Gu, Z., Rothberg, E., & Wunderling, R. (2000) · IFIP Advances in Information and Communication Technology · doi:10.1007/978-0-387-35514-6_2 · Cited by 258 · cited 1×
“However, during the last years there have been important advances in linear programming (LP), which is the facilitating technology for solving MIPs (Bixby et al. 2000; Bertsimas and Weismantel 2005).”
Optimal multivariate matching before randomization
Greevy, R. (2004) · Biostatistics · doi:10.1093/biostatistics/5.2.263 · Cited by 205 · cited 2×
“Along these lines, Greevy et al. (2004) showed that covariance adjustment is more efficient when the covariates are more nearly balanced.”
Planning and Analysis of Observational Studies.
Sykes, R. E., & Cochran, W. G. (1984) · Contemporary Sociology · doi:10.2307/2067916 · Cited by 194
Bixby, R., & Rothberg, E. (2007) · Annals of Operations Research · doi:10.1007/s10479-006-0091-y · Cited by 193
Li, Y. P., Propert, K. J., & Rosenbaum, P. R. (2001) · Journal of the American Statistical Association · doi:10.1198/016214501753208573 · Cited by 191 · cited 2×
“…solution space are needed with the branch-and-bound algorithm, and, on the other hand, to avoid the slow convergence rates of the addition of too many constraints, which presents numerical difficulties for solvers with the cutting plane algorithm (Bixby and Rothberg 2007; Linderoth and Lodi 2010).”
Optimal Nonbipartite Matching and Its Statistical Applications
Lu, B., Greevy, R., Xu, X., & Beck, C. (2011) · The American Statistician · doi:10.1198/tast.2011.08294 · Cited by 181 · cited 2×
“Fourth, as pointed out by Rosenbaum and Rubin (1985), despite the complexity of some matching algorithms, very simple methods can be used to analyze the outcomes, and typical analysis of matched samples do not require the parametric assumptions of most regression methods (Lu et al. 2011).”
Planning and Analysis of Observational Studies
Cochran, W. G. (1983) · Wiley Series in Probability and Statistics · doi:10.1002/9780470316542 · Cited by 166
Rosenbaum, P. R., Ross, R. N., & Silber, J. H. (2007) · Journal of the American Statistical Association · doi:10.1198/016214506000001059 · Cited by 161 · cited 2×
“In the spirit of Rosenbaum, Ross, and Silber (2007) and Yang et al. (2012), it is also possible to match with fine balance and near-fine balance by adding the constraints ∑ t∈T ∑ c∈C at,c1{xc,p=b} = m ∑ t∈T 1{xt,p=b} ∀b ∈ B (11) and ∣∣∣∣∣ ∑ t∈T ∑ c∈C at,c1{xc,p=b} − m ∑ t∈T 1{xt,p=b} ∣∣∣∣∣ ≤ ξ ∀b ∈…”
Medical and Financial Risks Associated With Surgery in the Elderly Obese
Silber, J. H., Rosenbaum, P. R., Kelz, R. R., Reinke, C. E., Neuman, M. D., Ross, R. N., Even‐Shoshan, O., David, G., Saynisch, P. A., Kyle, F. A., et al. (2012) · Annals of Surgery · doi:10.1097/sla.0b013e31825375ef · Cited by 92 · cited 1×
“The match in the upper-right plot (match 1) was obtained through the classical assignment algorithm, the match in the lower-left plot (match 4) used mixed integer programming to constrain the differences in means of all the covariates to zero, and the match in the lower lower-right plot (match 5)…”
Yang, D., Small, D. S., Silber, J. H., & Rosenbaum, P. R. (2012) · Biometrics · doi:10.1111/j.1541-0420.2011.01691.x · Cited by 77 · cited 4×
“When fine balance is not feasible, it is possible to approximate it with near-fine balance (Yang et al. 2012).”
InfluentialMatching and thick description in an observational study of mortality after surgery
Rosenbaum, P. R. (2001) · Biostatistics · doi:10.1093/biostatistics/2.2.217 · Cited by 65
Robustness of the Wilcoxon Test to a Certain Dependency Between Samples
Hollander, M., Pledger, G. W., & Lin, P. (1974) · The Annals of Statistics · doi:10.1214/aos/1176342623 · Cited by 58 · cited 1×
“The problem is that this iterative process usually involves a considerable amount of guesswork, and it is common to see that balancing some covariates worsens the balance of other covariates (Diamond and Sekhon 2005; Hill 2011; Iacus, King, and Porro 2012).”
Zubizarreta, J. R., Reinke, C. E., Kelz, R. R., Silber, J. H., & Rosenbaum, P. R. (2011) · The American Statistician · doi:10.1198/tas.2011.11072 · Cited by 51 · cited 3×
“When exact matching is not feasible, it might still be desirable to approximate it with near-exact matching (see Rosenbaum 2010, sec 9.2 for an introductory discussion on near-exact matching, and Zubizarreta et al. 2011 for using it to match for several sparse nominal covariates).”
InfluentialSelf-citationPlanning and Analysis of Observational Studies.
Smeeton, N., Cochran, W. G., Moses, L. E., & Mosteller, F. C. (1984) · Journal of the Royal Statistical Society Series D (The Statistician) · doi:10.2307/2988241 · Cited by 24
Estimating Anesthesia Time Using the Medicare Claim
Silber, J. H., Rosenbaum, P. R., Even‐Shoshan, O., Mi, L., Kyle, F. A., Teng, Y., Bratzler, D. W., & Fleisher, L. A. (2011) · Anesthesiology · doi:10.1097/aln.0b013e31821d6c81 · Cited by 23 · cited 1×
“We used data from the U.S. Medicare population, from the Obesity and Surgical Outcomes Study (OBSOS; Silber et al. 2011, 2012).”
RASMUSSEN, E. A., Turner, J. C., Ninomiya, K., & Renfrew, I. A. (2003) · Cambridge University Press eBooks · doi:10.1017/cbo9780511524974.004 · Cited by 15
Research works printed but not linked: 6
- Bertsimas , D. and Weismantel , R. 2005 . Optimization Over Integers , Belmont, MA : Athena Scientific & Dynamic Ideas .
- ——— . 2007 . “Flexible, Optimal Matching for Observational Studies,” . R News , 7 : 18 – 24 .
- 10.1093/biomet/70.2.510 · doi:10.1093/biomet/70.2.510
- 10.1080/00031305.1985.10479383 · doi:10.1080/00031305.1985.10479383 · Paper page
- Snedecor , G. W. and Cochran , W. G. 1980 . Statistical Methods , Ames, IA : Iowa State University Press .
- Yoon F. 2009 . “Entire Matching and Its Application in an Observational Study of Treatments for Melanoma ” . Ph.D. dissertation Department of Statistics The Wharton School University of Pennsylvania.
Cited by
Stable weights that balance covariates for estimation with incomplete outcome data
Zubizarreta, J. R. (2015) · Journal of the American Statistical Association · doi:10.1080/01621459.2015.1023805 · Cited by 369
Matching Methods for Causal Inference with Time‐Series Cross‐Sectional Data
Imai, K., Kim, I. S., & Wang, E. H. (2023) · American Journal of Political Science · doi:10.1111/ajps.12685 · Cited by 349
Anesthesia Technique, Mortality, and Length of Stay After Hip Fracture Surgery
Neuman, M. D., Rosenbaum, P. R., Ludwig, J. M., Zubizarreta, J. R., & Silber, J. H. (2014) · JAMA · doi:10.1001/jama.2014.6499 · Cited by 303
Self-citationGordon, B. R., Zettelmeyer, F., Bhargava, N., & Chapsky, D. (2019) · Marketing Science · doi:10.1287/mksc.2018.1135 · Cited by 287
“12Researchers have recently developed more sophisticated methods for estimating causal effects (Imai and Ratkovic 2014), including those that blend insights from operations research (Zubizarreta 2012, Zubizarreta 2015) and machine learning (Athey, Imbens, and Wager forthcoming).”
Machine Learning in Policy Evaluation: New Tools for Causal Inference
Kreif, N., & DiazOrdaz, K. (2019) · Oxford Research Encyclopedia of Economics and Finance · doi:10.1093/acrefore/9780190625979.013.256 · Cited by 57
Metadata
- Authors
- Zubizarreta, J. R.
- Venue
- Journal of the American Statistical Association
- Publisher
- Informa UK Limited
- Published
- 2012-12
- Volume
- 107
- Issue
- 500
- Pages
- 1360-1371
- ISSN
- 0162-1459 (ISSN-L)
- 1537-274X
- OpenAlex
- W2027739215
- Cited by
- 219
- References
- 40
- FWCI
- 8.09
- Type
- journal article
- Language
- en
- Open access
- closed
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.