Using Inverse Probability-Weighted Estimators in Comparative Effectiveness Analyses With Observational Databases

Duke Clinical Research Institute, and Departments of Medicine, Duke University School of Medicine, Durham, North Carolina, USA.
Medical Care (Impact Factor: 3.23). 11/2007; 45(10 Supl 2):S103-7. DOI: 10.1097/MLR.0b013e31806518ac
Source: PubMed


Inverse probability-weighted estimation is a powerful tool for use with observational data. In this article, we describe how this propensity score-based method can be used to compare the effectiveness of 2 or more treatments. First, we discuss the inherent problems in using observational data to assess comparative effectiveness. Next, we provide a conceptual explanation of inverse probability-weighted estimation and point readers to sources that address the method in more formal, technical terms. Finally, we offer detailed guidance about how to implement the estimators in comparative effectiveness analyses.

Download full-text


Available from: Eric Eisenstein, Oct 02, 2015
186 Reads
    • "Secondly, the propensity scores were used with inverse probability of treatment weighting (IPTW). In this method, patients who underwent early surgery were weighted for the reciprocal of the propensity score, and those who underwent late surgery were weighted for the reciprocal of 1 minus propensity score [23]. We used IBM SPSS Statistics version 22.0 (IBM SPSS, Armonk, NY, USA) to conduct all statistical analyses. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The timing of cardiac surgery for infective endocarditis with ischaemic stroke remains controversial. Using a nationwide inpatient database in Japan, we conducted a retrospective observational study. We identified patients aged 20 years or older with ischaemic stroke on admission who were diagnosed with infective endocarditis and underwent cardiac surgery during the initial hospitalization between July 2010 and March 2013. In-hospital mortality and perioperative complications were compared between the early (≤7 days) and late (>7 days) surgery groups using logistic regression analyses with adjustment for propensity scores and inverse probability of treatment weighting. We identified 253 patients who underwent cardiac valve surgery for infective endocarditis with ischaemic stroke on admission. In-hospital mortality rates were 8.6 and 9.5% in the early (n = 105) and late (n = 148) surgery groups, respectively. There were no significant differences in the in-hospital mortality between the early and late surgery groups in the propensity score-adjusted model [odds ratio (OR), 0.95; 95% confidence interval (CI), 0.35-2.54] and inverse probability-weighted model (risk difference, -0.82%; 95% CI, -6.43 to 4.84%). The perioperative complication rates were 42.9 and 37.8% in the early and late surgery groups, respectively, and showed no significant differences in the propensity score-adjusted model (OR, 1.11; 95% CI, 0.63-1.97) and inverse probability-weighted model (risk difference, 1.54%; 95% CI, -7.13 to 10.2%). Early timing of surgery for infective endocarditis patients with ischaemic stroke was not associated with higher in-hospital mortality or complications after admission. Early timing of surgery may not be contraindicated for infective endocarditis patients with ischaemic stroke. © The Author 2015. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.
    Interactive Cardiovascular and Thoracic Surgery 08/2015; DOI:10.1093/icvts/ivv235 · 1.16 Impact Factor
  • Source
    • "Propensity score methods were used for reducing the selection bias on the estimate of the treatment effect in observational studies [19] [20]. Adjustments for the differences between dipyridamole treated and untreated groups were performed by using the inverse probability of treatment weighted (IPTW) estimator [21]. The propensity score for dipyridamole treatment of each patient was obtained by fitting a logistic regression model that included the predictor variable (i.e., dipyridamoletreated or untreated patients) as an outcome and all baseline covariates. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Dipyridamole has been shown to decrease proteinuria and improve renal function progression especially in early chronic kidney disease (CKD) patients with glomerulonephropathy. A combination therapy of dipyridamole with aspirin could prevent second strokes in the general population. Whether these effects of dipyridamole are also true in advanced CKD patients and whether dipyridamole could improve renal outcomes or patient survival is unknown. We retrospectively analyzed an observational cohort of 3074 participants with CKD stage 3–5 from southern Taiwan, of whom 871 (28.3%) had received dipyridamole treatment ≥50 mg/d for ≥3 months and more than half of the observation period. The mean age was 63.6 ± 13.4 years and the mean estimated glomerular filtration rate (eGFR) was 25.5 mL/min/1.73 m2. After inverse probability of treatment weighted adjustment by propensity score, there were no differences between the dipyridamole-treated and untreated groups. Dipyridamole treatment was associated with decreased odds for rapid eGFR decline [odds ratio, 0.755; 95% confidence interval (CI), 0.595–0.958; p = 0.007] and progression of urine protein-to-creatinine ratio (odds ratio, 0.655; 95% CI, 0.517–0.832; p = 0.002). In survival analysis, the dipyridamole-treated group was also associated with a decreased risk for end-stage renal disease (hazard ratio, 0.847; 95% CI, 0.733–0.980; p = 0.011) and all-cause mortality (hazard ratio, 0.765; 95% CI, 0.606–0.971; p = 0.001) but not for cardiovascular events. Our findings demonstrate that dipyridamole treatment is significantly associated with better renal outcomes and patient survival in patients with CKD stage 3–5. Further investigations are warranted to confirm these independent positive effects.
    The Kaohsiung journal of medical sciences 10/2014; 30(12). DOI:10.1016/j.kjms.2014.10.002 · 0.80 Impact Factor
  • Source
    • "Using logistic regression to adjust for between-group differences in baseline characteristics, propensity scores were developed for estimating probability on the basis of patient characteristics such that patients would be selected for BII [22]. Inverse probability weighting (IPW) based on the propensity score was then used as the primary tool to adjust for differences between the 2 treatment groups [23]. This approach was taken to create balance between the treatment groups. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background The question of whether pure metabolic surgery could be used in nonobese patients with type 2 diabetes has been considered. The objective of this study was to assess the comparative effects of the Billroth I (BI) and Billroth II (BII) reconstruction methods on remission of type 2 diabetes in nonobese patients undergoing subtotal gastrectomy for cancer. Methods The charts of 404 patients who underwent radical subtotal gastrectomy for cancer between January 2008 and December 2010 were retrospectively reviewed. From these patients, 49 with type 2 diabetes were included in this study. Diabetes remission rates, the percentage change in fasting plasma glucose levels, glycated hemoglobin levels, body mass index, and fasting total cholesterol levels at 2 years were observed. Outcomes were compared using propensity scores and inverse probability-weighting adjustment that reduced treatment-selection bias. Covariate-adjusted logistic regression models were assessed. Results The 2-year diabetes remission rate for the 23 patients who underwent BI reconstruction was 39.1%, compared with 50.0% for the 26 patients who underwent BII reconstruction. At 2 years, the BII group showed lower glycated hemoglobin levels (BI, 6.4%; BII, 6.1%; P = .003) and had greater percent reductions in their average glycated hemoglobin levels from baseline (BI,−11.6%; BII,−14.5%; P = .043). BII reconstruction was significantly associated with an increased diabetes remission rate (odds ratio, 3.22; 95% confidence interval, 1.05–9.83) in covariate-adjusted logistic regression analysis. Conclusions These propensity score-adjusted analyses of patients who had undergone subtotal gastrectomy indicated that BII reconstruction was associated with increased diabetes remission compared with BI reconstruction during the 2-year follow-up period. This study suggests the possibility of employing the surgical duodenal switch for the treatment of nonobese type 2 diabetes patients.
    Surgery for Obesity and Related Diseases 01/2013; 10(2). DOI:10.1016/j.soard.2013.09.013 · 4.07 Impact Factor
Show more