Vegetarian diets and risk of nonalcoholic fatty liver disease: An observational study of National Health and Nutrition Examination Survey 2005-2018 using propensity score methods

BACKGROUND

Studies on the association between vegetarian diets and nonalcoholic fatty liver disease (NAFLD) are limited and have inconsistent results. This study aims to explore the association between vegetarian diets and NAFLD and compare the stage of fibrosis between vegetarians and nonvegetarians in a US representative sample.

METHODS

Cross-sectional data from 23,130 participants aged ≥20 years were obtained from the National Health and Nutrition Examination Survey, 2005-2018. Vegetarian status was classified based on two 24-h dietary recalls. We examined the association between vegetarian diets and the risk of NAFLD using the propensity score weighting method.

RESULTS

Vegetarian diets were significantly associated with decreases in hepatic steatosis index (HSI), US fatty liver index and nonalcoholic fatty liver disease fibrosis score with mean differences of -2.70 (95% confidence interval [CI]: -3.69, -1.70), -3.03 (95% CI: -7.15, -0.91) and -0.12 (95% CI: -0.26, -0.01), respectively. While modelling the risk of NAFLD, we estimated that vegetarians were 53% less likely to have NAFLD assessed by HSI (odds ratios [OR]: 0.47; 95% CI: 0.34, 0.65). The effect of vegetarian diets was higher among individuals with lower waist circumferences (OR: 0.20) than among those with higher waist circumferences (OR: 0.53,p interaction ${p}_{\text{interaction}}\,$ = 0.004). However, the association was largely attenuated after adjusting for body mass index and diabetes status. No significant association was identified between vegetarian diets and advanced fibrosis.

CONCLUSIONS

Vegetarian diets were associated with a lower prevalence of NAFLD among US adults, and the association appeared to be stronger in people with lower waist circumferences. Further studies are warranted to replicate our findings.

Analyzing risk factors for post-acute recovery in older adults with Alzheimer's disease and related dementia: A new semi-parametric model for large-scale medicare claims

Nearly 300,000 older adults experience a hip fracture every year, the majority of which occur following a fall. Unfortunately, recovery after fall-related trauma such as hip fracture is poor, where older adults diagnosed with Alzheimer's disease and related dementia (ADRD) spend a particularly long time in hospitals or rehabilitation facilities during the post-operative recuperation period. Because older adults value functional recovery and spending time at home versus facilities as key outcomes after hospitalization, identifying factors that influence days spent at home after hospitalization is imperative. While several individual-level factors have been identified, the characteristics of the treating hospital have recently been identified as contributors. However, few methodological rigorous approaches are available to help overcome potential sources of bias such as hospital-level unmeasured confounders, informative hospital size, and loss to follow-up due to death. This article develops a useful tool equipped with unsupervised learning to simultaneously handle statistical complexities that are often encountered in health services research, especially when using large administrative claims databases. The proposed estimator has a closed form, thus only requiring light computation load in a large-scale study. We further develop its asymptotic properties with stabilized inference assisted by unsupervised clustering. Extensive simulation studies demonstrate superiority of the proposed estimator compared to existing estimators.

An alternative model of maternity care for low-risk birth: Maternal and neonatal outcomes utilizing the midwifery-based birth center model

OBJECTIVE

To assess key birth outcomes in an alternative maternity care model, midwifery-based birth center care.

DATA SOURCES

The American Association of Birth Centers Perinatal Data Registry and birth certificate files, using national data collected from 2009 to 2019.

STUDY DESIGN

This observational cohort study compared key clinical birth outcomes of women at low risk for perinatal complications, comparing those who received care in the midwifery-based birth center model versus hospital-based usual care. Linear regression analysis was used to assess key clinical outcomes in the midwifery-based group as compared with hospital-based usual care. The hospital-based group was selected using nearest neighbor matching, and the primary linear regressions were weighted using propensity score weights (PSWs). The key clinical outcomes considered were cesarean delivery, low birth weight, neonatal intensive care unit admission, breastfeeding, and neonatal death. We performed sensitivity analyses using inverse probability weights and entropy balancing weights. We also assessed the remaining role of omitted variable bias using a bounding methodology.

DATA COLLECTION

Women aged 16-45 with low-risk pregnancies, defined as a singleton fetus and no record of hypertension or cesarean section, were included. The sample was selected for records that overlapped in each year and state. Counties were included if there were at least 50 midwifery-based birth center births and 300 total births. After matching, the sample size of the birth center cohort was 85,842 and the hospital-based cohort was 261,439.

PRINCIPAL FINDINGS

Women receiving midwifery-based birth center care experienced lower rates of cesarean section (-12.2 percentage points, p < 0.001), low birth weight (-3.2 percentage points, p < 0.001), NICU admission (-5.5 percentage points, p < 0.001), neonatal death (-0.1 percentage points, p < 0.001), and higher rates of breastfeeding (9.3 percentage points, p < 0.001).

CONCLUSIONS

This analysis supports midwifery-based birth center care as a high-quality model that delivers optimal outcomes for low-risk maternal/newborn dyads.

Parametric and nonparametric propensity score estimation in multilevel observational studies

There has been growing interest in using nonparametric machine learning approaches for propensity score estimation in order to foster robustness against misspecification of the propensity score model. However, the vast majority of studies focused on single-level data settings, and research on nonparametric propensity score estimation in clustered data settings is scarce. In this article, we extend existing research by describing a general algorithm for incorporating random effects into a machine learning model, which we implemented for generalized boosted modeling (GBM). In a simulation study, we investigated the performance of logistic regression, GBM, and Bayesian additive regression trees for inverse probability of treatment weighting (IPW) when the data are clustered, the treatment exposure mechanism is nonlinear, and unmeasured cluster-level confounding is present. For each approach, we compared fixed and random effects propensity score models to single-level models and evaluated their use in both marginal and clustered IPW. We additionally investigated the performance of the standard Super Learner and the balance Super Learner. The results showed that when there was no unmeasured confounding, logistic regression resulted in moderate bias in both marginal and clustered IPW, whereas the nonparametric approaches were unbiased. In presence of cluster-level confounding, fixed and random effects models greatly reduced bias compared to single-level models in marginal IPW, with fixed effects GBM and fixed effects logistic regression performing best. Finally, clustered IPW was overall preferable to marginal IPW and the balance Super Learner outperformed the standard Super Learner, though neither worked as well as their best candidate model.

Covariate handling approaches in combination with dynamic borrowing for hybrid control studies

Borrowing data from external control has been an appealing strategy for evidence synthesis when conducting randomized controlled trials (RCTs). Often named hybrid control trials, they leverage existing control data from clinical trials or potentially real-world data (RWD), enable trial designs to allocate more patients to the novel intervention arm, and improve the efficiency or lower the cost of the primary RCT. Several methods have been established and developed to borrow external control data, among which the propensity score methods and Bayesian dynamic borrowing framework play essential roles. Noticing the unique strengths of propensity score methods and Bayesian hierarchical models, we utilize both methods in a complementary manner to analyze hybrid control studies. In this article, we review methods including covariate adjustments, propensity score matching and weighting in combination with dynamic borrowing and compare the performance of these methods through comprehensive simulations. Different degrees of covariate imbalance and confounding are examined. Our findings suggested that the conventional covariate adjustment in combination with the Bayesian commensurate prior model provides the highest power with good type I error control under the investigated settings. It has desired performance especially under scenarios of different degrees of confounding. To estimate efficacy signals in the exploratory setting, the covariate adjustment method in combination with the Bayesian commensurate prior is recommended.

Long-Term Effects of a Ketogenic Diet for Cancer

A ketogenic diet has been proposed as a potential supportive therapy for cancer patients, although its long-term influence on survival rates remain controversial. In our previous report, we presented promising results for 37 of 55 patients with advanced cancer enrolled between 2013 and 2018 who remained on a ketogenic diet for at least 3 months. We followed all 55 patients until March 2023 and analyzed the data up to March 2022. For the 37 patients with previously reported promising results, the median follow-up period was 25 (range of 3-104) months and 28 patients died. The median overall survival (OS) in this subset of 37 patients was 25.1 months and the 5-year survival rate was 23.9%. We also evaluated the association between the duration of the ketogenic diet and outcome in all 55 patients, except for 2 patients with insufficient data. The patients were divided into two groups: those who followed the diet for ≥12 months (n = 21) and those who followed it for <12 months (n = 32). The median duration of the ketogenic diet was 37 (range of 12-99) months for the ≥12 months group and 3 (range of 0-11) months for the <12 months group. During the follow-up period, 41 patients died (10/21 in the ≥12 months group and 31/32 in the <12 months group). The median OS was 19.9 months (55.1 months in the ≥12 months group and 12 months in the <12 months group). Following the inverse probability of treatment weighting to align the background factors of the two groups and make them comparable, the adjusted log-rank test showed a significantly better OS rate in the group that continued the ketogenic diet for a longer period (p < 0.001, adjusted log-rank test). These results indicate that a longer continuation of the ketogenic diet improved the prognosis of advanced cancer patients.

Robust causal inference of drug-drug interactions

There is growing interest in developing causal inference methods for multi-valued treatments with a focus on pairwise average treatment effects. Here we focus on a clinically important, yet less-studied estimand: causal drug-drug interactions (DDIs), which quantifies the degree to which the causal effect of drug A is altered by the presence versus the absence of drug B. Confounding adjustment when studying the effects of DDIs can be accomplished via inverse probability of treatment weighting (IPTW), a standard approach originally developed for binary treatments and later generalized to multi-valued treatments. However, this approach generally results in biased results when the propensity score model is misspecified. Motivated by the need for more robust techniques, we propose two empirical likelihood-based weighting approaches that allow for specifying a set of propensity score models, with the second method balancing user-specified covariates directly, by incorporating additional, nonparametric constraints. The resulting estimators from both methods are consistent when the postulated set of propensity score models contains a correct one; this property has been termed multiple robustness. In this paper, we derive two multiply-robust estimators of the causal DDI, and develop inference procedures. We then evaluate their finite sample performance through simulation. The results demonstrate that the proposed estimators outperform the standard IPTW method in terms of both robustness and efficiency. Finally, we apply the proposed methods to evaluate the impact of renin-angiotensin system inhibitors (RAS-I) on the comparative nephrotoxicity of nonsteroidal anti-inflammatory drugs (NSAID) and opioids, using data derived from electronic medical records from a large multi-hospital health system.

Variance estimation for the average treatment effects on the treated and on the controls

Common causal estimands include the average treatment effect, the average treatment effect of the treated, and the average treatment effect on the controls. Using augmented inverse probability weighting methods, parametric models are judiciously leveraged to yield doubly robust estimators, that is, estimators that are consistent when at least one the parametric models is correctly specified. Three sources of uncertainty are associated when we evaluate these estimators and their variances, that is, when we estimate the treatment and outcome regression models as well as the desired treatment effect. In this article, we propose methods to calculate the variance of the normalized, doubly robust average treatment effect of the treated and average treatment effect on the controls estimators and investigate their finite sample properties. We consider both the asymptotic sandwich variance estimation, the standard bootstrap as well as two wild bootstrap methods. For the asymptotic approximations, we incorporate the aforementioned uncertainties via estimating equations. Moreover, unlike the standard bootstrap procedures, the proposed wild bootstrap methods use perturbations of the influence functions of the estimators through independently distributed random variables. We conduct an extensive simulation study where we vary the heterogeneity of the treatment effect as well as the proportion of participants assigned to the active treatment group. We illustrate the methods using an observational study of critical ill patients on the use of right heart catherization.

Predicting academic success of autistic students in higher education

LAYMEN SUMMARY

What is already known about the topic?Autistic youths increasingly enter universities. We know from existing research that autistic students are at risk of dropping out or studying delays. Using machine learning and historical information of students, researchers can predict the academic success of bachelor students. However, we know little about what kind of information can predict whether autistic students will succeed in their studies and how accurate these predictions will be.What does this article add?In this research, we developed predictive models for the academic success of 101 autistic bachelor students. We compared these models to 2,465 students with other health conditions and 25,077 students without health conditions. The research showed that the academic success of autistic students was predictable. Moreover, these predictions were more precise than predictions of the success of students without autism.For the success of the first bachelor year, concerns with aptitude and study choice were the most important predictors. Participation in pre-education and delays at the beginning of autistic students' studies were the most influential predictors for second-year success and delays in the second and final year of their bachelor's program. In addition, academic performance in high school was the strongest predictor for degree completion in 3 years.Implications for practice, research, or policyThese insights can enable universities to develop tailored support for autistic students. Using early warning signals from administrative data, institutions can lower dropout risk and increase degree completion for autistic students.

Superior survival for breast-conserving therapy over mastectomy in patients with breast cancer: A population-based SEER database analysis across 30 years

Introduction

It has been believed that breast-conserving therapy (lumpectomy plus adjuvant radiation, Lum + RT) and mastectomy without radiation (Mast + NoRT) have equivalent survival outcomes. However, there is a need to re-evaluate the role of lumpectomy plus adjuvant radiation due to changed breast cancer management over time. This study aimed to conduct a population-based study that compare long-term oncologic survival outcomes after Lum + RT vs Mast + NoRT.

Methods

The Surveillance, Epidemiology and End Results database was used to identify female breast cancer patients with a primary localized breast cancer diagnosis from 1988 to 2018. The standardized incidence/mortality ratio (SIR/SMR) for breast cancer recurrence (BCR) and breast cancer-specific death (BSD) was estimated by the SEER*Stat program. Cumulative incidences of BCR and BSD were assessed using Gray's method. We evaluated the effects of Lum + RT vs. Mast + NoRT on breast cancer recurrence-free survival (BRFS) and breast cancer-specific survival (BCSS). Fine-Gray competing risk model analyses, propensity score-adjusted Kaplan-Meier analyses and Cox proportional hazards model analyses were applied.

Results

A total of 205,788 women were included in the study. Patients who underwent Lum + RT had higher SIR of BCR (4.14 [95% confidence interval, CI: 3.94-4.34] vs. 1.11 [95% CI: 1.07-1.14]) and lower SMR (9.89 [95% CI: 9.71-10.08] vs. 17.07 [95% CI: 16.82-17.33]) than patients who underwent Mast + NoRT. Lum + RT was associated with higher competing risk of BCR (adjusted hazard ratio [HR]: 1.996, 95% CI: 1.925-2.069, p < 0.001) and lower competing risk of BSD when compared to Mast + RT (adjusted HR: 0.584, 95% CI: 0.572-0.597, p < 0.001). Multivariate Cox regression analysis revealed similar results (adjusted HR after PSW for BRFS: 1.792, 95% CI 1.716-1.871, p < 0.001; adjusted HR after PSW for BCSS: 0.706, 95% CI 0.688-0.725, p < 0.001). These findings persisted in the sensitivity and subgroup analyses.

Discussion

The present study further confirmed superior long-term survival with lumpectomy plus adjuvant radiation over mastectomy independent of patient characteristics including age, race, time period, historic subtype, tumor size, historic grade and stage, indicating that this benefit may result from the treatment itself.

Physical function mediates the effects of sensory impairment on quality of life in older adults: Cross-sectional study using propensity-score weighting

AIMS

To investigate the effect of sensory impairment on quality of life in older adults and to assess the role of physical function as a mediator of the effect of the sensory impairment on quality of life.

DESIGN

A cross-sectional study.

METHODS

Older adults aged ≥65 years (N = 600) were recruited from January 2019 to May 2020. Hearing and visual function were measured with pure-tone audiometry and Snellen visual acuity tests, respectively. Quality of life (World Health Organization Quality of Life Scale Brief Version), physical function (Multidimensional Functional Assessment Questionnaire) and sociodemographic characteristics were reported by participants using interviewer-administered questionnaires. Propensity score weighting analysis was conducted based on generalized propensity scores via multinominal logistic regression for age, gender, education, income, and comorbidities. The difference in the quality of life was tested by applying a one-way analysis of variance. Multiple mediation analysis was conducted to explore the direct, indirect, and total effects of sensory impairment on quality of life through physical function.

RESULTS

After propensity score weighting adjustment, when compared with participants with no sensory impairment, participants with dual sensory impairment had the worst quality of life, followed by visual impairment and then hearing impairment. Physical function statistically significantly mediated the effect of hearing impairment, visual impairment and dual sensory impairment on quality of life in older adults.

CONCLUSION

Our findings demonstrated that the negative effect of the sensory impairment on quality of life in older adults was mediated through physical function.

IMPACT

The convergence of an increasing ageing population and the prevalence of sensory impairment presents a significant global health burden. This study demonstrated that physical function was a mediator of quality of life in older adults. Designing appropriate physical activity interventions for older adults with sensory impairment could serve to enhance physio-psychological health and improve quality of life.

Flexible propensity score estimation strategies for clustered data in observational studies

Existing studies have suggested superior performance of nonparametric machine learning over logistic regression for propensity score estimation. However, it is unclear whether the advantages of nonparametric propensity score modeling are carried to settings where there is clustering of individuals, especially when there is unmeasured cluster-level confounding. In this work we examined the performance of logistic regression (all main effects), Bayesian additive regression trees and generalized boosted modeling for propensity score weighting in clustered settings, with the clustering being accounted for by including either cluster indicators or random intercepts. We simulated data for three hypothetical observational studies of varying sample and cluster sizes. Confounders were generated at both levels, including a cluster-level confounder that is unobserved in the analyses. A binary treatment and a continuous outcome were generated based on seven scenarios with varying relationships between the treatment and confounders (linear and additive, nonlinear/nonadditive, nonadditive with the unobserved cluster-level confounder). Results suggest that when the sample and cluster sizes are large, nonparametric propensity score estimation may provide better covariate balance, bias reduction, and 95% confidence interval coverage, regardless of the degree of nonlinearity or nonadditivity in the true propensity score model. When the sample or cluster sizes are small, however, nonparametric approaches may become more vulnerable to unmeasured cluster-level confounding and thus may not be a better alternative to multilevel logistic regression. We applied the methods to the National Longitudinal Study of Adolescent to Adult Health data, estimating the effect of team sports participation during adolescence on adulthood depressive symptoms.

Estimating the excess burden of pertussis disease in Australia within the first year of life, that might have been prevented through timely vaccination

BACKGROUND

Previous Australian studies have shown that delayed vaccination with each of the three primary doses of diphtheria-tetanus-pertussis-containing vaccines (DTP) is up to 50 % in certain subpopulations. We estimated the excess burden of pertussis that might have been prevented if (i) all primary doses and (ii) each dose was given on time.

METHODS

Perinatal, immunization, pertussis notification and death data were probabilistically linked for 1 412 984 infants born in two Australian states in 2000-12. A DTP dose administered >15 days after the recommended age was considered delayed. We used Poisson regression models to compare pertussis notification rates to 1-year of age in infants with ≥1 dose delayed (Aim 1) or any individual dose delayed (Aim 2) versus a propensity weighted counterfactual on-time cohort.

RESULTS

Of all infants, 42% had ≥1 delayed DTP dose. We estimated that between 39 to 365 days of age, 85 (95% CI: 61-109) cases per 100 000 infants, could have been prevented if all infants with ≥1 delayed dose had received their three doses within the on-time window. Risk of pertussis was higher in the delayed versus the on-time cohort, so crude rates overestimated the excess burden (110 cases per 100 000 infants (95% CI: 95-125)). The estimated dose-specific excess burden per 100 000 infants was 132 for DTP1, 50 for DTP2 and 19 for DTP3.

CONCLUSIONS

We provide robust evidence that improved DTP vaccine timeliness, especially for the first dose, substantially reduces the burden of infant pertussis. Our methodology, using a potential outcomes framework, is applicable to other settings.

The Association of Early Corticosteroid Therapy With Clinical and Health-Related Quality of Life Outcomes in Children With Septic Shock

OBJECTIVES

Corticosteroids are commonly used in the treatment of pediatric septic shock without clear evidence of the potential benefits or risks. This study examined the association of early corticosteroid therapy with patient-centered clinically meaningful outcomes.

DESIGN

Subsequent cohort analysis of data derived from the prospective Life After Pediatric Sepsis Evaluation (LAPSE) investigation. Outcomes among patients receiving hydrocortisone or methylprednisolone on study day 0 or 1 were compared with those who did not use a propensity score-weighted analysis that controlled for age, sex, study site, and measures of first-day illness severity.

SETTING

Twelve academic PICUs in the United States.

PATIENTS

Children with community-acquired septic shock 1 month to 18 years old enrolled in LAPSE, 2013-2017. Exclusion criteria included a history of chronic corticosteroid administration.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Among children enrolled in LAPSE, 352 of 392 met analysis inclusion criteria, and 155 of 352 (44%) received early corticosteroid therapy. After weighting corticosteroid therapy administration propensity across potentially confounding baseline characteristics, differences in outcomes associated with treatment were not statistically significant (adjusted effect or odds ratio [95% CI]): vasoactive-inotropic support duration (-0.37 d [-1.47 to 0.72]; p = 0.503), short-term survival without new morbidity (1.37 [0.83-2.28]; p = 0.218), new morbidity among month-1 survivors (0.70 [0.39-1.23]; p = 0.218), and persistent severe deterioration of health-related quality of life or mortality at month 1 (0.70 [0.40-1.23]; p = 0.212).

CONCLUSIONS

This study examined the association of early corticosteroid therapy with mortality and morbidity among children encountering septic shock. After adjusting for variables with the potential to confound the relationship between early corticosteroid administration and clinically meaningful end points, there was no improvement in outcomes associated with this therapy. Results from this propensity analysis provide additional justification for equipoise regarding corticosteroid therapy for pediatric septic shock and ascertain the need for a well-designed clinical trial to examine benefit/risk for this intervention.

Association between Metformin Use and Risk of Total Knee Arthroplasty and Degree of Knee Pain in Knee Osteoarthritis Patients with Diabetes and/or Obesity: A Retrospective Study

Objectives: We aimed to examine whether metformin (MET) use is associated with a reduced risk of total knee arthroplasty (TKA) and low severity of knee pain in patients with knee osteoarthritis (OA) and diabetes and/or obesity. Methods: Participants diagnosed with knee OA and diabetes and/or obesity from June 2000 to July 2019 were selected from the information system of a local hospital. Regular MET users were defined as those with recorded prescriptions of MET or self-reported regular MET use for at least 6 months. TKA information was extracted from patients’ surgical records. Knee pain was assessed using the numeric rating scale. Log-binomial regression, linear regression, and propensity score weighting (PSW) were performed for statistical analyses. Results: A total of 862 participants were included in the analyses. After excluding missing data, there were 346 MET non-users and 362 MET users. MET use was significantly associated with a reduced risk of TKA (prevalence ratio: 0.26, 95% CI: 0.15 to 0.45, p < 0.001), after adjustment for age, gender, body mass index, various analgesics, and insurance status. MET use was significantly associated with a reduced degree of knee pain after being adjusted for the above covariates (β: −0.48, 95% CI: −0.91 to −0.05, p = 0.029). There was a significantly accumulative effect of MET use on the reduced risk of TKA. Conclusion: MET can be a potential therapeutic option for OA. Further clinical trials are needed to determine if MET can reduce the risk of TKA and the severity of knee pain in metabolic-associated OA patients.

Considerations for Using Multiple Imputation in Propensity Score-Weighted Analysis - A Tutorial with Applied Example

Purpose

Propensity score-weighting for confounder control and multiple imputation to counter missing data are both widely used methods in epidemiological research. Combination of the two is not trivial and requires a number of decisions to produce valid inference. In this tutorial, we outline the assumptions underlying each of the methods, present our considerations in combining the two, discuss the methodological and practical implications of our choices and briefly point to alternatives. Throughout we apply the theory to a research project about post-traumatic stress disorder in Syrian refugees.

Patients and Methods

We detail how we used logistic regression-based propensity scores to produce "standardized mortality ratio"-weights and Substantive Model Compatible-Full Conditional Specification for multiple imputation of missing data to get the estimate of association. Finally, a percentile confidence interval was produced by bootstrapping.

Results

A simple propensity score model with weight truncation at 1st and 99th percentile obtained acceptable balance on all covariates and was chosen as our model. Due to computational issues in the multiple imputation, two levels of one of the substantive model covariates and two levels of one of the auxiliary covariates were collapsed. This slightly modified propensity score model was the substantive model in the SMC-FCS multiple imputation, and regression models were set up for all partially observed covariates. We set the number of imputations to 10 and number of iterations to 40. We produced 999 bootstrap estimates to compute the 95-percentile confidence interval.

Conclusion

Combining propensity score-weighting and multiple imputation is not a trivial task. We present considerations necessary to do so, realizing it is demanding in terms of both workload and computational time; however, we do not consider the former a drawback: it makes some of the underlying assumptions explicit and the latter may be a nuisance that will diminish with faster computers and better implementations.

A Quasi-Experimental Study of the Effects of Pre-Kindergarten Education on Pediatric Asthma

Ensuring access to pre-kindergarten (Pre-K) education remains a pressing policy issue in the United States. Prior research has shown the positive effects that Pre-K has on children’s cognitive development. However, studies on its effects on children’s health outcomes are scarce. This study aimed to investigate the effects of the Pre-K program on pediatric asthma. Children’s individual data from existing research conducted in North Carolina were linked with state Medicaid claims data from 2011–2017. There were 51,408 observations (person-month unit) of 279 children enrolled in Pre-K and 333 unenrolled children. Asthma was identified using the ICD 9/10 codes. A difference-in-differences model was adopted using a panel analysis with three time periods: before, during, and after Pre-K. The explanatory variables were interaction terms between Pre-K enrollment and (a) before vs. during period and (b) during vs. after period. The results indicated that children enrolled in Pre-K had a greater risk of asthma diagnosis during Pre-K (b = 0.0145, p = 0.058). Conversely, in the post-intervention period, the enrolled children had a lower of receiving an asthma diagnosis (b = −0.0216, p = 0.002). These findings indicate that Pre-K may increase the use of asthma-related health services in the short term and decrease the service use after participants leave the program.

Effect of a baby-friendly workplace support intervention on exclusive breastfeeding in Kenya

Exclusive breastfeeding (EBF) during the first 6 months of life is crucial for optimizing child growth, development and survival, as well as the mother's wellbeing. Mother's employment may hinder optimal breastfeeding, especially in the first 6 months. We assessed the effectiveness of a baby-friendly workplace support intervention on EBF in Kenya. This pre-post intervention study was conducted between 2016 and 2018 on an agricultural farm in Kericho County. The intervention targeted pregnant/breastfeeding women residing on the farm and consisted of workplace support policies and programme interventions including providing breastfeeding flexi-time and breaks for breastfeeding mothers; day-care centres (crèches) for babies near the workplace and lactation centres with facilities for breast milk expression and storage at the crèches; creating awareness on available workplace support for breastfeeding policies; and home-based nutritional counselling for pregnant and breastfeeding women. EBF was measured through 24-h recall. The effect of the intervention on EBF was estimated using propensity score weighting. The study included 270 and 146 mother-child dyads in the nontreated (preintervention) group and treated (intervention) group, respectively. The prevalence of EBF was higher in the treated group (80.8%) than in the nontreated group (20.2%); corresponding to a fourfold increased probability of EBF [risk ratio (RR) 3.90; 95% confidence interval (CI) 2.95-5.15]. The effect of the intervention was stronger among children aged 3-5 months (RR 8.13; 95% CI 4.23-15.64) than among those aged <3 months (RR 2.79; 95% CI 2.09-3.73). The baby-friendly workplace support intervention promoted EBF especially beyond 3 months in this setting.

Adjusted logistic propensity weighting methods for population inference using nonprobability volunteer-based epidemiologic cohorts

Many epidemiologic studies forgo probability sampling and turn to nonprobability volunteer-based samples because of cost, response burden, and invasiveness of biological samples. However, finite population (FP) inference is difficult to make from the nonprobability sample due to the lack of population representativeness. Aiming for making inferences at the population level using nonprobability samples, various inverse propensity score weighting methods have been studied with the propensity defined by the participation rate of population units in the nonprobability sample. In this article, we propose an adjusted logistic propensity weighting (ALP) method to estimate the participation rates for nonprobability sample units. The proposed ALP method is easy to implement by ready-to-use software while producing approximately unbiased estimators for population quantities regardless of the nonprobability sample rate. The efficiency of the ALP estimator can be further improved by scaling the survey sample weights in propensity estimation. Taylor linearization variance estimators are proposed for ALP estimators of FP means that account for all sources of variability. The proposed ALP methods are evaluated numerically via simulation studies and empirically using the naïve unweighted National Health and Nutrition Examination Survey III sample, while taking the 1997 National Health Interview Survey as the reference, to estimate the 15-year mortality rates.

Comparing Propensity Score Methods Versus Traditional Regression Analysis for the Evaluation of Observational Data: A Case Study Evaluating the Treatment of Gram-Negative Bloodstream Infections

BACKGROUND

Propensity score methods are increasingly being used in the infectious diseases literature to estimate causal effects from observational data. However, there remains a general gap in understanding among clinicians on how to critically review observational studies that have incorporated these analytic techniques.

METHODS

Using a cohort of 4967 unique patients with Enterobacterales bloodstream infections, we sought to answer the question "Does transitioning patients with gram-negative bloodstream infections from intravenous to oral therapy impact 30-day mortality?" We conducted separate analyses using traditional multivariable logistic regression, propensity score matching, propensity score inverse probability of treatment weighting, and propensity score stratification using this clinical question as a case study to guide the reader through (1) the pros and cons of each approach, (2) the general steps of each approach, and (3) the interpretation of the results of each approach.

RESULTS

2161 patients met eligibility criteria with 876 (41%) transitioned to oral therapy while 1285 (59%) remained on intravenous therapy. After repeating the analysis using the 4 aforementioned methods, we found that the odds ratios were broadly similar, ranging from 0.84-0.95. However, there were some relevant differences between the interpretations of the findings of each approach.

CONCLUSIONS

Propensity score analysis is overall a more favorable approach than traditional regression analysis when estimating causal effects using observational data. However, as with all analytic methods using observational data, residual confounding will remain; only variables that are measured can be accounted for. Moreover, propensity score analysis does not compensate for poor study design or questionable data accuracy.

Laparoscopic vs. open surgery for gastrointestinal stromal tumors of esophagogastric junction: A multicenter, retrospective cohort analysis with propensity score weighting

Objective

Laparoscopic resection is increasingly performed for gastrointestinal stromal tumors (GISTs). However, the laparoscopic approach for GISTs located in the esophagogastric junction (EGJ-GIST) is surgically challenging. This study compares the efficacy of laparoscopic surgery and the open procedure for EGJ-GIST through the propensity score weighting (PSW) method.

Methods

Between April 2006 and April 2018, 1,824 surgical patients were diagnosed with primary gastric GIST at four medical centers in South China. Of these patients, 228 were identified as EGJ-GISTs and retrospectively reviewed clinicopathological characteristics, operative information, and long-term outcomes. PSW was used to create the balanced cohorts.

Results

PSW was carried out in laparoscopic and open-surgery cohorts according to year of surgery, sex, age, body mass index (BMI), tumor size, mitotic rates and recurrence risk. After PSW, 438 patients consisting of 213 laparoscopic (L group) and 225 open surgery (O group) patients were enrolled. After PSW, the following measures in the L group were superior to those in the O group: median operative time [interquartile range (IQR)]: 100.0 (64.5−141.5)vs. 149.0 (104.0−197.5) min, P<0.001; median blood loss (IQR): 30.0 (10.0−50.0)vs. 50.0 (20.0−100.0) mL, P=0.002; median time to liquid intake (IQR): 3.0 (2.0−4.0)vs. 4.0 (3.0−5.0) d, P<0.001; median hospital stay (IQR): 6.0 (4.0−8.0)vs. 7.0 (5.0−12.0) d, P<0.001; and postoperative complications (10.3%vs. 22.7%, P=0.001). The median follow-up was 55 (range, 2−153) months in the entire cohort. No significant differences were detected in either relapse-free survival (RFS) [hazard ratio (HR): 0.372, 95% confidence interval (95% CI): 0.072−1.910, P=0.236) or overall survival (OS) (HR: 0.400, 95% CI: 0.119−1.343, P=0.138) between the two groups.

Conclusions

Laparoscopic surgery for EGJ-GIST is associated with the advantages of shorter operative time, reduced blood loss, shorter time to liquid intake, and shorter length of stay, all without compromising postoperative outcomes and long-term survival.

Improving External Validity of Epidemiologic Cohort Analyses: A Kernel Weighting Approach

For various reasons, cohort studies generally forgo probability sampling required to obtain population representative samples. However, such cohorts lack population-representativeness, which invalidates estimates of population prevalences for novel health factors only available in cohorts. To improve external validity of estimates from cohorts, we propose a kernel weighting (KW) approach that uses survey data as a reference to create pseudo-weights for cohorts. A jackknife variance is proposed for the KW estimates. In simulations, the KW method outperformed two existing propensity-score-based weighting methods in mean-squared error while maintaining confidence interval coverage. We applied all methods to estimating US population mortality and prevalences of various diseases from the non-representative US NIH-AARP cohort, using the sample from US-representative National Health Interview Survey (NHIS) as the reference. Assuming that the NHIS estimates are correct, the KW approach yielded generally less biased estimates compared to the existing propensity-score-based weighting methods.

Multinomial Extension of Propensity Score Trimming Methods: A Simulation Study

Crump et al. (Biometrika. 2009;96(1):187-199), Stürmer et al. (Am J Epidemiol. 2010;172(7):843-854), and Walker et al. (Comp Eff Res. 2013;2013(3):11-20) proposed propensity score (PS) trimming methods as a means to improve efficiency (Crump) or reduce confounding (Stürmer and Walker). We generalized the trimming definitions by considering multinomial PSs, one for each treatment, and proved that these proposed definitions reduce to the original binary definitions when we have only 2 treatment groups. We then examined the performance of the proposed multinomial trimming methods in the setting of 3 treatment groups, in which subjects with extreme PSs more likely had unmeasured confounders. Inverse probability of treatment weights, matching weights, and overlap weights were used to control for measured confounders. All 3 methods reduced bias regardless of the weighting methods in most scenarios. Multinomial Stürmer and Walker trimming were more successful in bias reduction when the 3 treatment groups had very different sizes (10:10:80). Variance reduction, seen in all methods with inverse probability of treatment weights but not with matching weights or overlap weights, was more successful with multinomial Crump and Stürmer trimming. In conclusion, our proposed definitions of multinomial PS trimming methods were beneficial within our simulation settings that focused on the influence of unmeasured confounders.

Comparing pharmacological treatments for cocaine dependence: Incorporation of methods for enhancing generalizability in meta-analytic studies

OBJECTIVES

Few head-to-head comparisons of cocaine dependence medications exist, and combining data from different randomized controlled trials (RCTs) is fraught with methodological challenges including limited generalizability of the RCT findings. This study applied a novel meta-analytic approach to data of cocaine dependence medications.

METHODS

Data from 4 placebo-controlled RCTs (Reserpine, Modafinil, Buspirone, and Ondansetron) were obtained from the National Institute of Drug Abuse Clinical Trials Network (n = 456). The RCT samples were weighted to resemble treatment-seeking patients (Treatment Episodes Data Set-Admissions) and individuals with cocaine dependence in general population (National Survey on Drug Use and Health). We synthesized the generalized outcomes with pairwise meta-analysis using individual-level data and compared the generalized outcomes across the 4 RCTs with network meta-analysis using study-level data.

RESULTS

Weighting the data by the National Survey on Drug Use and Health generalizability weight made the overall population effect on retention significantly larger than the RCT sample effect. However, there was no significant difference between the population effect and the RCT sample effect on abstinence. Weighting changed the ranking of the effectiveness across treatments.

CONCLUSIONS

Applying generalizability weights to meta-analytic studies is feasible and potentially provides a useful tool in assessing comparative effectiveness of treatments for substance use disorders in target populations.

Propensity score weighting analysis and treatment effect discovery

Inverse probability weighting can be used to estimate the average treatment effect in propensity score analysis. When there is lack of overlap in the propensity score distributions between the treatment groups under comparison, some weights may be excessively large, causing numerical instability and bias in point and variance estimation. We study a class of modified inverse probability weighting estimators that can be used to avoid this problem. These weights cause the estimand to deviate from the average treatment effect. We provide some justification for this deviation from the perspective of treatment effect discovery. We show that when lack of overlap occurs, the modified weights can achieve substantial gains in statistical power compared with inverse probability weighting and other propensity score methods. We develop analytical variance estimates that properly adjust for the sampling variability of the estimated propensity scores, and augment the modified inverse probability weighting estimator with outcome models for improved efficiency, a property that resembles double robustness. Results from extensive simulations and a real data application support our conclusions. The proposed methodology is implemented in R package PSW.

Bayesian propensity scores for high-dimensional causal inference: A comparison of drug-eluting to bare-metal coronary stents

High-dimensional data provide many potential confounders that may bolster the plausibility of the ignorability assumption in causal inference problems. Propensity score methods are powerful causal inference tools, which are popular in health care research and are particularly useful for high-dimensional data. Recent interest has surrounded a Bayesian treatment of propensity scores in order to flexibly model the treatment assignment mechanism and summarize posterior quantities while incorporating variance from the treatment model. We discuss methods for Bayesian propensity score analysis of binary treatments, focusing on modern methods for high-dimensional Bayesian regression and the propagation of uncertainty. We introduce a novel and simple estimator for the average treatment effect that capitalizes on conjugacy of the beta and binomial distributions. Through simulations, we show the utility of horseshoe priors and Bayesian additive regression trees paired with our new estimator, while demonstrating the importance of including variance from the treatment regression model. An application to cardiac stent data with almost 500 confounders and 9000 patients illustrates approaches and facilitates comparison with existing alternatives. As measured by a falsifiability endpoint, we improved confounder adjustment compared with past observational research of the same problem.

Generalizability of findings from randomized controlled trials: application to the National Institute of Drug Abuse Clinical Trials Network

AIMS

To compare randomized controlled trial (RCT) sample treatment effects with the population effects of substance use disorder (SUD) treatment.

DESIGN

Statistical weighting was used to re-compute the effects from 10 RCTs such that the participants in the trials had characteristics that resembled those of patients in the target populations.

SETTINGS

Multi-site RCTs and usual SUD treatment settings in the United States.

PARTICIPANTS

A total of 3592 patients in 10 RCTs and 1 602 226 patients from usual SUD treatment settings between 2001 and 2009.

MEASUREMENTS

Three outcomes of SUD treatment were examined: retention, urine toxicology and abstinence. We weighted the RCT sample treatment effects using propensity scores representing the conditional probability of participating in RCTs.

FINDINGS

Weighting the samples changed the significance of estimated sample treatment effects. Most commonly, positive effects of trials became statistically non-significant after weighting (three trials for retention and urine toxicology and one trial for abstinence); also, non-significant effects became significantly positive (one trial for abstinence) and significantly negative effects became non-significant (two trials for abstinence). There was suggestive evidence of treatment effect heterogeneity in subgroups that are under- or over-represented in the trials, some of which were consistent with the differences in average treatment effects between weighted and unweighted results.

CONCLUSIONS

The findings of randomized controlled trials (RCTs) for substance use disorder treatment do not appear to be directly generalizable to target populations when the RCT samples do not reflect adequately the target populations and there is treatment effect heterogeneity across patient subgroups.

Estimating causal effects for multivalued treatments: a comparison of approaches

Interventions with multivalued treatments are common in medical and health research, such as when comparing the efficacy of competing drugs or interventions, or comparing between various doses of a particular drug. In recent years, there has been a growing interest in the development of multivalued treatment effect estimators using observational data. In this paper, we compare the performance of commonly used regression-based methods that estimate multivalued treatment effects based on the unconfoundedness assumption. These estimation methods fall into three general categories: (i) estimators based on a model for the outcome variable using conventional regression adjustment; (ii) weighted estimators based on a model for the treatment assignment; and (iii) 'doubly-robust' estimators that model both the treatment assignment and outcome variable within the same framework. We assess the performance of these models using Monte Carlo simulation and demonstrate their application with empirical data. Our results show that (i) when models estimating both the treatment and outcome are correctly specified, all adjustment methods provide similar unbiased estimates; (ii) when the outcome model is misspecified, regression adjustment performs poorly, while all the weighting methods provide unbiased estimates; (iii) when the treatment model is misspecified, methods based solely on modeling the treatment perform poorly, while regression adjustment and the doubly robust models provide unbiased estimates; and (iv) when both the treatment and outcome models are misspecified, all methods perform poorly. Given that researchers will rarely know which of the two models is misspecified, our results support the use of doubly robust estimation.