Estimating inverse probability weights using super learner when weight-model specification is unknown in a marginal structural Cox model context

Abstract Background: The effects of combined diabetes and glycemic control strategies on the short-term prognosis in patients with a critical illness are currently ambiguous. The objectives of our study were to determine whether comorbid diabetes affects short-term prognosis and the optimal range of glycemic control in critically ill patients.Methods: We performed this study with the critical care database. The primary outcomes were 28-day mortality in critically ill patients with comorbid diabetes and the optimal range of glycemic control. Association of comorbid diabetes with 28-day mortality was assessed by multivariable Cox regression model with inverse probability weighting. Smooth curves were applied to fit the association for glucose and 28-day mortality.Results: Of the 33,680 patients enrolled in the study, 8,701 (25.83%) had diabetic comorbidity. Cox model with inverse probability weighting showed that the 28-day mortality rate was reduced by 29% (HR=0.71, 95% CI 0.67-0.76) in the group with diabetes in comparison to the group without diabetes. The E value of 2.17 indicated robustness to unmeasured confounders. The effect of the association between comorbid diabetes and 28-day mortality was generally in line for all subgroup variables, significant interactions were observed for glucose on first day, admission type, and use of insulin or not (Interaction P <0.05). A V-shaped relationship was observed between glucose concentrations and 28-day mortality in patients without diabetes, with the lowest 28-day mortality corresponding to the glucose level was 101.75 mg/dl (95% CI 94.64-105.80 mg/dl); whereas in patients with comorbid diabetes, the effect of glucose concentration on 28-day mortality was structurally softer than in those with uncomorbid diabetes. Lastly, of all patients, hyperglycemia had the greatest deleterious effect on patients admitted to CSRU.Conclusions: Our study further confirmed the protective effect of comorbid diabetes on the short-term prognosis of critically ill patients, resulting in an approximately 29% reduction in 28-day mortality. Besides, we also demonstrated the personalized glycemic control strategy for critically ill patients. Lastly, clinicians should be aware of the occurrence and the prompt management of hyperglycemia in critically ill patients admitted to the CSRU.

Download Full-text

Comparison of Parametric and Nonparametric Estimators for the Association Between Incident Prepregnancy Obesity and Stillbirth in a Population-Based Cohort Study

American Journal of Epidemiology ◽

10.1093/aje/kwz081 ◽

2019 ◽

Vol 188 (7) ◽

pp. 1328-1336 ◽

Cited By ~ 4

Author(s):

Ya-Hui Yu ◽

Lisa M Bodnar ◽

Maria M Brooks ◽

Katherine P Himes ◽

Ashley I Naimi

Keyword(s):

Inverse Probability Weighting ◽

Population Based ◽

Parametric Models ◽

Probability Weighting ◽

Inverse Probability ◽

Nonparametric Estimators ◽

Inverse Probability Weights ◽

Highly Sensitive ◽

Prepregnancy Obesity ◽

Index Weight

AbstractWhile prepregnancy obesity increases risk of stillbirth, few studies have evaluated the role of newly developed obesity independent of long-standing obesity. Additionally, researchers have relied almost exclusively on parametric models, which require correct specification of an unknown function for consistent estimation. We estimated the association between incident obesity and stillbirth in a cohort constructed from linked birth and death records in Pennsylvania (2003–2013). Incident obesity was defined as body mass index (weight (kg)/height (m)2) greater than or equal to 30. We used parametric G-computation, semiparametric inverse-probability weighting, and parametric/nonparametric targeted minimum loss-based estimation (TMLE) to estimate the association between incident prepregnancy obesity and stillbirth. Compared with pregnancies from women who stayed nonobese, women who became obese prior to their next pregnancy were estimated to have 2.0 (95% confidence interval (CI): 0.5, 3.5) more stillbirths per 1,000 pregnancies using parametric G-computation. However, despite well-behaved stabilized inverse probability weights, risk differences estimated from inverse-probability weighting, nonparametric TMLE, and parametric TMLE represented 6.9 (95% CI: 3.7, 10.0), 0.4 (95% CI: 0.1, 0.7), and 2.9 (95% CI: 1.5, 4.2) excess stillbirths per 1,000 pregnancies, respectively. These results, particularly those derived from nonparametric TMLE, were highly sensitive to covariates included in the propensity score models. Our results suggest that caution is warranted when using nonparametric estimators to quantify exposure effects.

Download Full-text

Robust Estimation of Inverse Probability Weights for Marginal Structural Models

Journal of the American Statistical Association ◽

10.1080/01621459.2014.956872 ◽

2015 ◽

Vol 110 (511) ◽

pp. 1013-1023 ◽

Cited By ~ 30

Author(s):

Kosuke Imai ◽

Marc Ratkovic

Keyword(s):

Robust Estimation ◽

Structural Models ◽

Marginal Structural Models ◽

Inverse Probability ◽

Inverse Probability Weights

Download Full-text

Early Experience With Neutralizing Monoclonal Antibody Therapy For COVID-19

10.1101/2021.04.09.21255219 ◽

2021 ◽

Author(s):

Mark Jarrett ◽

Warren B. Licht ◽

Kevin Bock ◽

Zenobia Brown ◽

Jamie S. Hirsch ◽

...

Keyword(s):

Monoclonal Antibody ◽

High Risk ◽

Hospital Mortality ◽

Primary Outcome ◽

Cox Model ◽

Adult Patients ◽

Infusion Therapy ◽

Probability Weighting ◽

Inverse Probability ◽

Neutralizing Monoclonal Antibody

ABSTRACTImportanceNeutralizing monoclonal antibody (MAB) therapies may benefit patients with mild to moderate COVID-19 at high risk for progressing to severe COVID-19 and/or hospitalization. Studies documenting approaches to deliver MAB infusions as well as demonstrating their efficacy are lacking.ObjectiveWe describe our experience and patient outcomes of almost 3,000 patients who received MAB infusion therapy at Northwell Health, a large integrated health care system in New York.Design, Setting, and ParticipantsThis is a descriptive study of adult patients who received MAB therapy between November 20, 2020, to January 31, 2021, and a retrospective cohort survival analysis comparing patients who received MAB therapy prior to admission versus those who did not. A multivariable Cox model with inverse probability weighting according to the propensity score including covariates (sociodemographic, comorbidities, and presenting vital signs) was used.Main outcomes and measuresThe primary outcome was in-hospital mortality; additional evaluations included ED utilization and hospitalization within 28 days of a positive COVID-19 test for patients who received MAB therapy.ResultsDuring the study period, 2818 adult patients received MAB infusion. Following therapy and within 28 days of COVID-19 test, 123 patients (4.4%) presented to the ED and were released and 145 patients (5.1%) were hospitalized. These 145 patients were compared with 200 controls who were eligible for but did not receive MAB therapy, and were hospitalized. In the MAB group, 16 (11%) patients met the primary outcome of in-hospital mortality, versus 21 (10.5%) in the control group. In an unadjusted Cox model, the hazard ratio (HR) for time to in-hospital mortality for the MAB group was 1.38 (95% confidence interval [95% CI] 0.696-2.719). Models adjusting for demographics (HR 1.1, 95% CI 0.53-2.23), demographics and Charlson Comorbidity Index (CCI) (HR 1.22, 95% CI 0.573-2.59), and with inverse probability weighting according to propensity scores (HR 1.19, 95% CI 0.619-2.29) did not demonstrate significance. The hospitalization rate was 4.4% for patients who received MAB therapy within 0-4 days, 5% within 5-7 days, and 6.1% within ≥8 days of symptom onset (p-value = 0.15).Conclusions and relevanceEstablishing the capability to provide neutralizing MAB infusion therapy requires significant planning and coordination. While this therapy may be an important treatment option for early mild to moderate COVID-19 in high-risk patients, further investigations are needed to define the optimal timing of MAB treatment in order to reduce hospitalization and mortality.

Download Full-text

Financial inclusion, land title and credit: evidence from China

China Agricultural Economic Review ◽

10.1108/caer-01-2019-0020 ◽

2020 ◽

Vol 12 (2) ◽

pp. 257-273

Author(s):

Meishan Jiang ◽

Krishna P. Paudel ◽

Donghui Peng ◽

Yunsheng Mi

Keyword(s):

Design Methodology ◽

Financial Inclusion ◽

Content Type ◽

Inverse Probability ◽

Rural Credit ◽

Inverse Probability Weights ◽

Land Title ◽

Land Titling ◽

Doubly Robust ◽

Rural Credit Market

Purpose The purpose of this paper is to study land title’s credit effect from a financial inclusion perspective in China. The focus is both small land holding and poor farmers. Formal and informal finances are considered to test their differences in land title’s credit effect. Design/methodology/approach The authors use augmented inverse-probability weights of the doubly robust method to test the effect of land titling on the rural credit market by addressing self-selection, endogeneity and heterogeneity concerns. Findings Results show that the poor, non-poor and small land holders with land titles are willing to borrow more from formal financial institutions. Land titling increases loan accessibility for non-poor and small land holding farmers. As for informal financing, large land holding and non-poor farmers show a decrease in informal lending. Land titling has a financial inclusion effect for some farmers, but poor farmers’ credit restrictions are not entirely solved by land titling. Originality/value This is the first study that focuses on the financial inclusion effect of farm land titling in China.

Download Full-text

Unemployment and Substance Use in Young Adults: Does Educational Attainment Modify the Association?

European Addiction Research ◽

10.1159/000365887 ◽

2014 ◽

Vol 21 (3) ◽

pp. 115-123 ◽

Cited By ~ 20

Author(s):

Maria Melchior ◽

Aude Chollet ◽

Gulizar Elidemir ◽

Cédric Galéra ◽

Nadia Younès

Keyword(s):

Higher Education ◽

Substance Use ◽

Young Adults ◽

Educational Attainment ◽

Community Based ◽

Inverse Probability ◽

Inverse Probability Weights ◽

Parental Characteristics ◽

Level Data ◽

Smoke Tobacco

We studied whether patterns of substance use in relation to unemployment vary depending on educational level. Data come from 1,126 community-based young adults in France (18-35 years of age in 2011) and their parents (TEMPO and GAZEL studies). Tobacco use (≥1 cigarette/day, 22.5% prevalence), nicotine dependence (Fagerström test ≥2, 7.1% prevalence), alcohol use (≥2 units/week, 25.3% prevalence), alcohol abuse (WHO AUDIT ≥7 in women and ≥8 in men, 10.8% prevalence), cannabis use (≥1 time, 16.5% prevalence), and cannabis abuse (CAST ≥2, 5.0% prevalence) were assessed by interview. We conducted logistic regression analyses controlled for inverse probability weights of unemployment, calculated based on demographics, negative life events, health, and juvenile and parental characteristics. Compared to participants who were always employed, those who were unemployed and had no higher education were more likely to smoke tobacco (OR: 2.76, 95% CI: 1.86-4.10), to be nicotine dependent (OR: 5.70, 95% CI: 3.03-10.73), to use cannabis (OR: 2.27, 95% CI: 1.42-3.64), and to abuse cannabis (OR: 3.38, 95% CI: 1.63-7.04). Those who were unemployed and had higher education were especially likely to abuse alcohol (OR: 1.89, 95% CI: 1.16-3.09). Increases in unemployment may impact population levels of substance use, particularly in young adults with low educational attainment.

Download Full-text