• The Korean Journal of Applied Statistics
• /
• v.36 no.4
• /
• pp.323-335
• /
• 2023

In causal analysis of high dimensional data, it is important to reduce the dimension of covariates and transform them appropriately to control confounders that affect treatment and potential outcomes. The augmented inverse probability weighting (AIPW) method is mainly used for estimation of average treatment effect (ATE). AIPW estimator can be obtained by using estimated propensity score and outcome model. ATE estimator can be inconsistent or have large asymptotic variance when using estimated propensity score and outcome model obtained by parametric methods that includes all covariates, especially for high dimensional data. For this reason, an ATE estimation using an appropriate dimension reduction method and semiparametric model for high dimensional data is attracting attention. Semiparametric method or sparse sufficient dimensionality reduction method can be uesd for dimension reduction for the estimation of propensity score and outcome model. Recently, another method has been proposed that does not use propensity score and outcome regression. After reducing dimension of covariates, ATE estimation can be performed using matching. Among the studies on ATE estimation methods for high dimensional data, four recently proposed studies will be introduced, and how to interpret the estimated ATE will be discussed.

• Communications for Statistical Applications and Methods
• /
• v.30 no.1
• /
• pp.1-19
• /
• 2023

Observational data with missing or incomplete data are common in biomedical research. Multiple imputation is an effective approach to handle missing data with the ability to decrease bias while increasing statistical power and efficiency. In recent years propensity score (PS) matching has been increasingly used in observational studies to estimate treatment effect as it can reduce confounding due to measured baseline covariates. In this paper, we describe in detail approaches to competing risk analysis in the setting of incomplete observational data when using PS matching. First, we used multiple imputation to impute several missing variables simultaneously, then conducted propensity-score matching to match statin-exposed patients with those unexposed. Afterwards, we assessed the effect of statin exposure on the risk of heart failure-related hospitalizations or emergency visits by estimating both relative and absolute effects. Collectively, we provided a general methodological framework to assess treatment effect in incomplete observational data. In addition, we presented a practical approach to produce overall cumulative incidence function (CIF) based on estimates from multiple imputed and PS-matched samples.

• Korean Journal of Social Welfare Studies
• /
• no.37
• /
• pp.171-197
• /
• 2008

This study focuses on the following question; self-supporting training program increases participants' income compare to non-participants who have similar characteristics. This question is based on counterfactual assumption. In other words, this study concentrates on what the outcomes would have been if the participants were to be absent. This study adopts a quasi-experimental design. To overcome previous study's methodological weaknesses, especially selection bias, I applied matching procedure based on a propensity-score matching. Matching process was performed by using 'MatchIt' software. The major findings are as follows From Least Squares Regression analysis, I found the poor's income are significantly different according to age, pre-intervention earning, material status, and participation of training. Since the poor have homogeneous education level, education variable was not statistically significant. From the Simulation Quantities of Interest analysis, I also found that treatment group's expected incomes are lower than control's expected incomes. In other words, participation of training has a negative effect on the participants' earnings.

• Journal of the Korean Data and Information Science Society
• /
• v.27 no.1
• /
• pp.91-100
• /
• 2016

Fuzzy matching is proposed to make propensities of two groups similar with their propensity scores and a way to select control variable to make propensity scores with a process that shows how to acquire propensity scores using logic regression analysis, is presented. With such scores, it was a method to obtain an experiment group and a control group that had similar propensity employing the Fuzzy Matching. In the study, it was proven that the two groups were the same but with a different distribution chart and standardization which made edge tolerance different and we realized that the number of chosen cases decreased when the edge tolerance score became smaller. So with the idea, we were able to determine that it is possible to merge groups using fuzzy matching without a precontrol and use them when data (big data) are used while to check the pros and cons of Fuzzy Matching were made possible.

• Journal of Trauma and Injury
• /
• v.21 no.2
• /
• pp.108-114
• /
• 2008

Purpose: Using a propensity analysis, a recent study reported that blood transfusion might not be an independent predictor of mortality in critically ill patients, which contradicted the results of earlier studies. This study aims to reveal whether or not blood transfusion is an independent predictor of mortality in trauma patients. Methods: A total of three hundred fifty consecutive trauma patients who were admitted to our emergency center from January 2004 to October 2005 and who underwent an arterial blood gas analysis and a venous blood analysis were included in this study. Their medical records were collected prospectively and retrospectively. Using a multivariate logistic analysis, data on the total population and on the propensity-score -matched population were retrospectively analyzed for association with mortality. Results: Of the three hundred fifty patients, one hundred twenty-nine (36.9%) received a blood transfusion. These patients were older (mean age: 48 vs. 44 years; p=0.019) and had a higher mortality rate (27.9% vs. 7.7%; p<0.001). In the total population, the multivariate analysis revealed that the Glasgow coma scale score, the systolic blood pressure, bicarbonate, the need for respiratory support, past medical history of heart disease, the amount of blood transfusion for 24 hours, and hemoglobin were associated with mortality. In thirty-seven pairs of patients matched with a propensity score, potassium, new injury severity score, amount of blood transfusion for 24 hours, and pulse rate were associated with mortality in the multivariate analysis. Therefore, blood transfusion was a significant independent predictor of mortality in trauma patients. Conclusion: Blood transfusion was revealed to be a significant independent predictor of mortality in the total population of trauma patients and in the propensity-score-matched population.

• Journal of Trauma and Injury
• /
• v.37 no.1
• /
• pp.6-12
• /
• 2024

Purpose: Open pelvic bone fractures are relatively rare and are considered more severe than closed fractures. This study aimed to compare the clinical outcomes of open and closed severe pelvic bone fractures. Methods: Patients with severe pelvic bone fractures (pelvic Abbreviated Injury Scale score, ≥4) admitted at a single level I trauma center between 2016 and 2020 were retrospectively analyzed. Patients aged <16 years and those with incomplete medical records were excluded from the study. The patients were divided into open and closed fracture groups, and their demographics, treatment, and clinical outcomes were compared before and after 1:2 propensity score matching. Results: Of the 321 patients, 24 were in the open fracture group and 297 were in the closed fracture group. The open fracture group had more infections (37.5% vs. 5.7%, P<0.001) and longer stays in the intensive care unit (median 11 days, interquartile range [IQR] 6-30 days vs. median 5 days, IQR 2-13 days; P=0.005), but mortality did not show a statistically significant difference (20.8% vs. 15.5%, P=0.559) before matching. After 1:2 propensity score matching, the infection rate was significantly higher in the open fracture group (37.5% vs. 6.3%, P=0.002), whereas the length of intensive care unit stay (median 11 days, IQR 6-30 days vs. median 8 days, IQR 4-19 days; P=0.312) and mortality (20.8% vs. 27.1%, P=0.564) were not significantly different. Conclusions: The open pelvic fracture group had more infections than the closed pelvic fracture group, but mortality was not significantly different. Aggressive treatment of pelvic bone fractures is important regardless of the fracture type, and efforts to reduce infection are important in open pelvic bone fractures.

• The Korean Journal of Applied Statistics
• /
• v.36 no.5
• /
• pp.415-445
• /
• 2023

Oftentimes, the time dependent treatment covariate and the time dependent confounders exist in observation studies. It is an important problem to correctly adjust for the time dependent confounders in the propensity score analysis. Recently, In the survival data, Hade et al. (2020) used a propensity score matching method to correctly estimate the treatment delay effect when the time dependent confounder affects time to the treatment time, where the treatment delay effects is defined to the delay in treatment reception. In this paper, we proposed the Cox model based marginal structural model (Cox-MSM) framework to estimate the treatment delay effect and conducted extensive simulation studies to compare our proposed Cox-MSM with the propensity score matching method proposed by Hade et al. (2020). Our simulation results showed that the Cox-MSM leads to more exact estimate for the treatment delay effect compared with two sequential matching schemes based on propensity scores. Example from study in treatment discontinuation in conjunction with simulated data illustrates the practical advantages of the proposed Cox-MSM.

• The Korean Journal of Applied Statistics
• /
• v.36 no.5
• /
• pp.381-397
• /
• 2023

The most representative design used in clinical trials is randomization, which is used to accurately estimate the treatment effect. However, comparison between the treatment group and the control group in an observational study without randomization is biased due to various unadjusted differences, such as characteristics between patients. Propensity score weighting is a widely used method to address these problems and to minimize bias by adjusting those confounding and assess treatment effects. Inverse probability weighting, the most popular method, assigns weights that are proportional to the inverse of the conditional probability of receiving a specific treatment assignment, given observed covariates. However, this method is often suffered by extreme propensity scores, resulting in biased estimates and excessive variance. Several alternative methods including trimming, overlap weights, and matching weights have been proposed to mitigate these issues. In this paper, we conduct a simulation study to compare performance of various propensity score weighting methods under diverse situation, such as limited overlap, misspecified propensity score, and treatment contrary to prediction. From the simulation results overlap weights and matching weights consistently outperform inverse probability weighting and trimming in terms of bias, root mean squared error and coverage probability.

• Health Policy and Management
• /
• v.25 no.1
• /
• pp.3-10
• /
• 2015

Background: Previous studies showed differences in healthcare utilization among insurance types. This study aimed to analyze the difference in healthcare utilization for percutaneous transluminal coronary angioplasty inpatients by insurance types after controlling factors affecting healthcare utilization using propensity score matching (PSM). Methods: The 2011 national inpatient sample based on health insurance claims data was used for analysis. PSM was used to control factors influencing healthcare utilization except insurance types. Length of stay and total charges were used as healthcare utilization variables. Patients were divided into National Health Insurance (NHI) and Medical Aid (MA) patients. Factors representing inpatients (gender, age, admission sources, and Elixhauser comorbidity index) and hospitals (number of doctors, number of beds, and location of hospitals) were used as covariates in PSM. Results: Tertiary hospitals didn't show significant difference in length of stay and total charges after PSM between two insurance types. However, MA patients showed significantly longer length of stay than that of NHI patients after PSM in general hospitals. Multivariate regression analysis provided that admission sources, Elixhauser comorbidity index, insurance types, number of doctors, and location of hospitals (province) had significant influences on the length of stay in general hospitals. Conclusion: Study results provided evidences that healthcare utilization was differed by insurance types in general hospitals. Health policy makers will need to prepare interventions to influence the healthcare utilization differences between insurance types.

• Genomics & Informatics
• /
• v.17 no.4
• /
• pp.47.1-47.12
• /
• 2019

The achievements of genome-wide association studies have suggested ways to predict diseases, such as type 2 diabetes (T2D), using single-nucleotide polymorphisms (SNPs). Most T2D risk prediction models have used SNPs in combination with demographic variables. However, it is difficult to evaluate the pure additive contribution of genetic variants to classically used demographic models. Since prediction models include some heritable traits, such as body mass index, the contribution of SNPs using unmatched case-control samples may be underestimated. In this article, we propose a method that uses propensity score matching to avoid underestimation by matching case and control samples, thereby determining the pure additive contribution of SNPs. To illustrate the proposed propensity score matching method, we used SNP data from the Korea Association Resources project and reported SNPs from the genome-wide association study catalog. We selected various SNP sets via stepwise logistic regression (SLR), least absolute shrinkage and selection operator (LASSO), and the elastic-net (EN) algorithm. Using these SNP sets, we made predictions using SLR, LASSO, and EN as logistic regression modeling techniques. The accuracy of the predictions was compared in terms of area under the receiver operating characteristic curve (AUC). The contribution of SNPs to T2D was evaluated by the difference in the AUC between models using only demographic variables and models that included the SNPs. The largest difference among our models showed that the AUC of the model using genetic variants with demographic variables could be 0.107 higher than that of the corresponding model using only demographic variables.