Emre DEMİR, Engin YILDIRIM, Serdal Kenan KÖSE, Ömer Faruk AKMEŞE

An interactive web application for propensity score matching with R shiny; example of thrombophilia

Aim: The aim of this study was to develop a new web-based R Shiny package that calculates propensity score using manyalgorithms such as logistic regression, machine learning, and performs matching analysis with balance evaluation. In addition, itwas aimed to explain the process of matching analysis on a real data set by comparing the number of live births between those withmethylenetetrahydrofolate reductase (MTHFR) homozygous mutations and those without mutations in women hospitalized due toabortion in the gynecology and obstetrics clinic.Material and Methods: The web-based application was developed using R shiny. The “matchIt” library was used for matchinganalysis and PS prediction. The “cobalt” library was used to evaluate balance and generate plots.Results: The abortion variable, which was statistically significantly different in the groups before matching (p=0.010), was similar inthe groups after matching (p=0.743). In addition, when the descriptive statistics and p values of the other variables were examined,it was seen that almost full balance was achieved after matching and the confounder variables were similar distributed in groups.After matching analysis, it was determined that the result variable “livebirths” did not show statistically significant difference in thegroups (p=0.864).Conclusion: In this study, we developed an interactive web application for matching analysis based on propensity score. It is thoughtthat this application will facilitate the studies of the researchers.

PDF

___

1. Rubin DB. Estimating causal effects of treatments in randomized and nonrandomized studies. J Educational Psychology 1974;66:688-701.
2. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:41-55.
3. Rosenberger WF, Lachin JM. Randomization in clinical trials: theory and practice. New York: John Wiley 2002; 1-133
4. Kaspar EÇ, Bekiroğlu N, Genceli M. Propensity score in observational studies and an application in medical sciences. Turkiye Klinikleri J Biostat 2010;2:1-10.
5. Gökmen D, Alkan A, Bakırarar B, et al. Bilimsel araştırma yöntemleri. Ankara: Ankara Üniversitesi Basımevi 2018;1-61.
6. Hill HA, Kleinbaum DG. Bias in observational studies. In: Armitage P, ed. Encyclopedia of Biostatistics. 2nd edition. Chichester: John Wiley and Sons 2005;323- 30.
7. Hoffmeister H, Szklo M, Thamm M. Bias in observational studies. Epidemiological Practices in Research on Small Effects. Berlin: Springer 1998;59- 60.
8. D’agostino RB JR. Tutorial in biostatistics: propensity score methods for bias reduction in the comparıson of a treatment to a non-randomized control group. Statistics in Medicine 1988;17:2265-81.
9. Demir E. Development of new propensity score estimation models with machine learning algorithms for optimal matching analysis in non-randomized clinical trials. Ph.D. dissertation, Ankara University, Ankara 2019.
10. Demir E, Köse SK. Development of new propensity score estimation models with machine learning algorithms for optimal matching analysis in non-randomized clinical trials and an interactive web application with R shiny. XXI. National and IV. International Biostatistics Congress 2019;149-62.
11. Mccaffrey DF, Ridgeway G, Morral AR. Propensity score estimation with boosted regression for evaluating causal effects in observational studies. Psychological Methods 2004;9:403-25.
12. Setoguchi S, Schneeweiss S, Brookhart MA, et al. Evaluating uses of data mining techniques in propensity score estimation: a simulation study. Pharmacoepidemiology and drug safety 2008;17:546- 55.
13. Lee BK, Lessler J, Stuart EA. Improving propensity score weighting using machine learning. Statistics in Medicine 2010;29:337-46.
14. Pirracchio R, Petersen M, Van Der Laan M. Improving propensity score estimators’ robustness to model misspecification using super learner. American Journal of Epidemiology 2014;181:108-19.
15. Stuart EA. Matching methods for causal ınference: a review and a look forward. Statistical Science 2010;25:1-21.
16. Ho D, Imai K, King G, et al. Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Political Analysis 2007;15:199-36.
17. Greifer, N. Covariate balance tables and plots: A guide to the cobalt package. https://mran.microsoft. com/snapshot/2017-11-12/web/packages/cobalt/ vignettes/cobalt_basic_use.html, access date 01.07.2019.
18. Sekhon, JS. Multivariate and propensity score matching software with automated balance optimization: the matching package for R. Journal of Statistical Software 2011;42:1-52.