Average treatment effect on the treated, under lack of positivity.

 The use of propensity score methods has become ubiquitous in causal inference. At the heart of these methods is the positivity assumption. Violation of the positivity assumption leads to the presence of extreme propensity scoreweights when estimating average causal effects, which affects statistical inference. To circumvent this issue, trimming or truncating methods have been widely used. Unfortunately, these methods require that we pre-specify a threshold. There are anumber of alternative methods to deal with the lack of positivity when we estimate the average treatment effect (ATE). However, no other methods exist beyond trimming and truncation to deal with the same issue when the goal is to estimate theaverage treatment effect on the treated (ATT). In this article, we propose a propensity score weight-based alternative for the ATT, called overlap weighted average treatment effect on the treated. The appeal of our proposed method lies in its abilityto obtain similar or even better results than trimming and truncation while relaxing the constraint to choose an a priori threshold (or related measures). The performance of the proposed method is illustrated via a series of Monte Carlo simulationsand a data analysis on racial disparities in health care expenditures.

Evaluating analytic models for individually randomized group treatment trials with complex clustering in nested and crossed designs.

Many individually randomized group treatment (IRGT) trials randomly assign individuals to study arms but deliver treatments via shared agents, such as therapists, surgeons, or trainers. Post-randomization interactions induce correlations in outcome measures between participants sharing the same agent. Agents can be nested in or crossed with trial arm, and participants may interact with a single agent or with multiple agents. These complications have led to ambiguity in choice of models but there have been no systematic efforts to identify appropriate analytic models for these study designs. To address this gap, we undertook a simulation study to examine the performance of candidate analytic models in the presence of complex clustering arising from multiple membership, single membership, and single agent settings, in both nested and crossed designs and for a continuous outcome. With nested designs, substantial type I error rate inflation was observed when analytic models did not account for multiple membership and when analytic model weights characterizing the association with multiple agents did not match the data generating mechanism. Conversely, analytic models for crossed designs generally maintained nominal type I error rates unless there was notable imbalance in the number of participants that interact with each agent.

A cluster-randomized trial of client and provider directed financial interventions to align incentives with appropriate case management in private medicine retailers: Results of the TESTsmART trial in Lagos, Nigeria.

Malaria remains a major health priority in Nigeria. Among children with fever who seek care, less than a quarter gets tested for malaria, leading to inappropriate use of the recommended treatment for malaria; Artemisinin-based Combination Therapy (ACT). Here we test an innovative strategy to target ACT subsidies to clients seeking care in Nigeria's private retail health sector who have a confirmed malaria diagnosis. We supported point-of-care malaria testing (mRDTs) in 48 Private Medicine Retailers (PMRs) in the city of Lagos, Nigeria and randomized them to two study arms; a control arm offering subsidized mRDT testing for USD $0.66, and an intervention arm where, in addition to access to subsidized testing as in the control arm, clients who received a positive mRDT at the PMR were eligible for a free (fully subsidized) first-line ACT and PMRs received USD $0.2 for every mRDT performed. Our primary outcome was the proportion of ACTs dispensed to individuals with a positive diagnostic test. Secondary outcomes included proportion of clients who were tested at the PMR and adherence to diagnostic test results. Overall, 23% of clients chose to test at the PMR. Test results seemed to inform treatment decisions and resulted in enhanced targeting of ACTs to confirmed malaria cases with only 26% of test-negative clients purchasing an ACT compared to 58% of untested clients. However, the intervention did not offer further improvements, compared to the control arm, in testing rates or dispensing of ACTs to test-positive clients. We found that ACT subsidies were not passed on to clients testing positive in the intervention arm. We conclude that mRDTs could reduce ACT overconsumption in Nigeria's private retail health sector, but PMR-oriented incentive structures are difficult to implement and may need to be complemented with interventions targeting clients of PMRs to increase test uptake and adherence. Trials registration: Clinical Trials Registration Number: NCT04428307. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7816435/ Correction: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9476591/.

Causal inference in the absence of positivity: The role of overlap weights.

How to analyze data when there is violation of the positivity assumption? Several possible solutions exist in the literature. In this paper, we consider propensity score (PS) methods that are commonly used in observational studies to assess causal treatment effects in the context where the positivity assumption is violated. We focus on and examine four specific alternative solutions to the inverse probability weighting (IPW) trimming and truncation: matching weight (MW), Shannon's entropy weight (EW), overlap weight (OW), and beta weight (BW) estimators. We first specify their target population, the population of patients for whom clinical equipoise, that is, where we have sufficient PS overlap. Then, we establish the nexus among the different corresponding weights (and estimators); this allows us to highlight the shared properties and theoretical implications of these estimators. Finally, we introduce their augmented estimators that take advantage of estimating both the propensity score and outcome regression models to enhance the treatment effect estimators in terms of bias and efficiency. We also elucidate the role of the OW estimator as the flagship of all these methods that target the overlap population. Our analytic results demonstrate that OW, MW, and EW are preferable to IPW and some cases of BW when there is a moderate or extreme (stochastic or structural) violation of the positivity assumption. We then evaluate, compare, and confirm the finite-sample performance of the aforementioned estimators via Monte Carlo simulations. Finally, we illustrate these methods using two real-world data examples marked by violations of the positivity assumption.

Longitudinal pathways between maternal depression, parenting behaviors, and early childhood development: a mediation analysis.

Maternal depression is a global public health concern with far-reaching impacts on child development, yet our understanding of mechanisms remains incomplete. This study examined whether parenting mediates the association between maternal depression and child outcomes. Participants included 841 rural Pakistani mother-child dyads (50% female). Maternal depression was measured at 12 months postpartum, parenting behaviors (warmth, stimulation, and harsh parenting) were measured at 24 months, and child outcomes (mental health, socioemotional development, and cognitive skills) were measured at 36 months. Maternal depression predicted increased harsh parenting, child mental health difficulties, and child socioemotional concerns; however, there was little evidence for parenting as a mediator between maternal depression and child outcomes. Sex-stratified results are discussed, and findings are situated in context.

A cluster-randomized trial of client and provider-directed financial interventions to align incentives with appropriate case management in retail medicine outlets: Results of the TESTsmART Trial in western Kenya.

ACTs are responsible for a substantial proportion of the global reduction in malaria mortality over the last ten years, made possible by publicly-funded subsidies making these drugs accessible and affordable in the private sector. However, inexpensive ACTs available in retail outlets have contributed substantially to overconsumption. We test an innovative, scalable strategy to target ACT-subsidies to clients with a confirmatory diagnosis. We supported malaria testing(mRDTs) in 39 medicine outlets in western Kenya, randomized to three study arms; control arm offering subsidized mRDT testing (0.4USD), client-directed intervention where all clients who received a positive RDT at the outlet were eligible for a free (fully-subsidized) ACT, and a combined client and provider directed intervention where clients with a positive RDT were eligible for free ACT and outlets received 0.1USD for every RDT performed. Our primary outcome was the proportion of ACT dispensed to individuals with a positive diagnostic test. Secondary outcomes included proportion of clients tested at the outlet and adherence to diagnostic test results. 43% of clients chose to test at the outlet. Test results informed treatment decisions, resulting in targeting of ACTs to confirmed malaria cases- 25.3% of test-negative clients purchased an ACT compared to 75% of untested clients. Client-directed and client+provider-directed interventions did not offer further improvements, compared to the control arm, in testing rates(RD = 0.09, 95%CI:-0.08,0.26) or dispensing of ACTs to test-positive clients(RD = 0.01,95% CI:-0.14, 0.16). Clients were often unaware of the price they paid for the ACT leading to uncertainty in whether the ACT subsidy was passed on to the client. This uncertainty undermines our ability to definitively conclude that client-directed subsidies are not effective for improving testing and appropriate treatment. We conclude that mRDTs could reduce ACT overconsumption in the private retail sector, but incentive structures are difficult to scale and their value to private providers is uncertain. Trial registration: ClinicalTrials.gov NCT04428307.

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.

SLC38A2 provides proline to fulfill unique synthetic demands arising during osteoblast differentiation and bone formation.

Cellular differentiation is associated with the acquisition of a unique protein signature that is essential to attain the ultimate cellular function and activity of the differentiated cell. This is predicted to result in unique biosynthetic demands that arise during differentiation. Using a bioinformatic approach, we discovered that osteoblast differentiation is associated with increased demand for the amino acid proline. When compared to other differentiated cells, osteoblast-associated proteins, including RUNX2, OSX, OCN, and COL1A1, are significantly enriched in proline. Using a genetic and metabolomic approach, we demonstrate that the neutral amino acid transporter SLC38A2 acts cell-autonomously to provide proline to facilitate the efficient synthesis of proline-rich osteoblast proteins. Genetic ablation of SLC38A2 in osteoblasts limits both osteoblast differentiation and bone formation in mice. Mechanistically, proline is primarily incorporated into nascent protein with little metabolism observed. Collectively, these data highlight a requirement for proline in fulfilling the unique biosynthetic requirements that arise during osteoblast differentiation and bone formation.

Bones have diverse roles in the body, such as supporting weight, allowing movement and protecting internal organs. Regardless of their location, all bones in the body are formed and maintained by specialized cells called osteoblasts. To produce the different components of bone, osteoblasts need a constant supply of amino acids, the building blocks of proteins. If these nutrients are limited, this may lead to weak and fragile bones that can fracture more easily. Naïve cells in the bone, which are yet to have a defined role, also require large amounts of amino acids to develop into fully functioning osteoblasts. Previous studies have found that specific amino acids (like glutamine and asparagine) are particularly important for forming the proteins in bone. However, it was unclear which amino acids are critical for the development of osteoblasts. To investigate, Shen et al. studied naïve cells that had been extracted from the embryos of mice and developed into osteoblasts in the laboratory. They found that the developing osteoblasts produced proteins enriched in proline, and that naïve cells required large amounts of this amino acid as they turned into osteoblasts. Genetic analysis revealed that osteoblasts carry the gene for a protein called SLC38A2, which has been shown to transport proline into other types of cells. Shen et al. then used gene editing tools to delete this transporter from the osteoblasts of mice. The mutated mice could not efficiently produce proline-rich proteins during embryonic development and formed less bone. These findings highlight that proline is important for developing osteoblasts and synthesizing the products of bone. Further research is needed, but it is possible that dietary supplements of proline may be beneficial for maintaining or promoting bone formation in adulthood. This could help individuals that have more fragile bones, such as the elderly or patients with bone diseases, like osteoporosis.

Constrained randomization and statistical inference for multi-arm parallel cluster randomized controlled trials.

A practical limitation of cluster randomized controlled trials (cRCTs) is that the number of available clusters may be small, resulting in an increased risk of baseline imbalance under simple randomization. Constrained randomization overcomes this issue by restricting the allocation to a subset of randomization schemes where sufficient overall covariate balance across comparison arms is achieved. However, for multi-arm cRCTs, several design and analysis issues pertaining to constrained randomization have not been fully investigated. Motivated by an ongoing multi-arm cRCT, we elaborate the method of constrained randomization and provide a comprehensive evaluation of the statistical properties of model-based and randomization-based tests under both simple and constrained randomization designs in multi-arm cRCTs, with varying combinations of design and analysis-based covariate adjustment strategies. In particular, as randomization-based tests have not been extensively studied in multi-arm cRCTs, we additionally develop most-powerful randomization tests under the linear mixed model framework for our comparisons. Our results indicate that under constrained randomization, both model-based and randomization-based analyses could gain power while preserving nominal type I error rate, given proper analysis-based adjustment for the baseline covariates. Randomization-based analyses, however, are more robust against violations of distributional assumptions. The choice of balance metrics and candidate set sizes and their implications on the testing of the pairwise and global hypotheses are also discussed. Finally, we caution against the design and analysis of multi-arm cRCTs with an extremely small number of clusters, due to insufficient degrees of freedom and the tendency to obtain an overly restricted randomization space.

Effect of a lay counselor-delivered integrated maternal mental health and early childhood development group-based intervention in Northern Ghana: a cluster-randomized controlled trial.

BACKGROUND: Caregiver mental health is linked to early childhood development, yet more robust evidence of community-based interventions to prevent maternal depression and optimize socio-emotional development of young children is needed. Objectives of this cluster-randomized controlled trial (cRCT), based in Northern Ghana, are to assess the impact of the lay counselor-delivered, group-based Integrated Mothers and Babies Course and Early Childhood Development (iMBC/ECD) program on (1) the mental health of mothers of children under age 2; and (2) the socio-emotional development of their children. METHODS: This cRCT randomized 32 women's groups - 16 received iMBC/ECD content (intervention) and 16 received general health education content (control). Surveys were administered at baseline, immediate post-intervention, and 8-month post-intervention. The primary outcome was maternal depression [Patient Health Questionnaire (PHQ-9)], and the secondary outcome was child's socio-emotional development [Ages and Stages Questionnaire: Social Emotional (ASQ:SE-2)]. Qualitative interviews with 33 stakeholders were also conducted. RESULTS: In total, 374 participants were enrolled at baseline while pregnant with the index child, 19% endorsing moderate/severe depression. Of these, 266 (71.1%) completed the 8-month post-intervention survey (~19 months post-baseline). There were no significant effects of iMBC/ECD on PHQ-9 and ASQ:SE-2 scores. However, results favored the intervention arm in most cases. iMBC participants were highly satisfied with the program but qualitative feedback from stakeholders indicated some implementation challenges. CONCLUSIONS: This real-world evaluation had null findings; however, post-intervention depression levels were very low in both arms (3%). Future research should examine the potential impact of women's groups on postpartum mental health more broadly with varying content.

Effect of a lay counselor delivered integrated maternal mental health and early childhood development group-based intervention in Siaya County, Kenya: A quasi-experimental longitudinal study.

BACKGROUND: Maternal mental health is linked to early childhood development; yet there is a gap in evidence-based interventions for low-resource settings. This study estimates the impact of 'Integrated Mothers and Babies Course and Early Childhood Development' (iMBC/ECD), a cognitive-behavioral, group-based intervention, on maternal depression and early childhood social-emotional development in Siaya County, Kenya. METHODS: This quasi-experimental study enrolled 417 pregnant women and mothers of children under age 2 across two sub-counties in Siaya County. The intervention area had 193 women in 23 groups implementing iMBC/ECD and the control area had 224 women in 30 groups exposed to ECD only content. Mother/index child dyads were followed for two years. To estimate the causal treatment effect from the non-randomized design, we implemented the propensity score weighting method with inverse probability weights. RESULTS: At baseline, 10.2% of participants endorsed moderate/severe depressive symptoms. At 14-months post-intervention, 7.4% endorsed moderate/severe depression. Overall, iMBC/ECD intervention did not have a significant impact on reducing maternal depression or improving children's social and emotional development. However, sub-group analyses revealed that iMBC/ECD was associated with lowered depressive symptoms among women with no/low education, four or more children and/or no experience of intimate partner violence in the past year. Women with high program attendance (more than half of 14 sessions) also experienced consistently fewer depressive symptoms compared to those with lower attendance. LIMITATIONS: Non-randomized study, sub-group analyses are exploratory. CONCLUSIONS: The iMBC/ECD program may have the potential to improve maternal mental health and early child development for more targeted vulnerable populations.

Incentivizing appropriate malaria case management in the private sector: a study protocol for two linked cluster randomized controlled trials to evaluate provider- and client-focused interventions in western Kenya and Lagos, Nigeria.

BACKGROUND: A large proportion of artemisinin-combination therapy (ACT) anti-malarial medicines is consumed by individuals that do not have malaria. The over-consumption of ACTs is largely driven by retail sales in high malaria-endemic countries to clients who have not received a confirmatory diagnosis. This study aims to target ACT sales to clients receiving a confirmatory diagnosis using malaria rapid diagnostic tests (mRDTs) at retail outlets in Kenya and Nigeria. METHODS: This study comprises two linked four-arm 2 × 2 factorial cluster randomized controlled trials focused on malaria diagnostic testing and conditional ACT subsidies with the goal to evaluate provider-directed and client-directed interventions. The linked trials will be conducted at two contrasting study sites: a rural region around Webuye in western Kenya and the urban center of Lagos, Nigeria. Clusters are 41 and 48 participating retail outlets in Kenya and Nigeria, respectively. Clients seeking care at participating outlets across all arms will be given the option of paying for a mRDT-at a study-recommended price-to be conducted at the outlet. In the provider-directed intervention arm, the outlet owner receives a small monetary incentive to perform the mRDT. In the client-directed intervention arm, the client receives a free ACT if they purchase an mRDT and receive a positive test result. Finally, the fourth study arm combines both the provider- and client-directed interventions. The diagnosis and treatment choices made during each transaction will be captured using a mobile phone app. Study outcomes will be collected through exit interviews with clients, who sought care for febrile illness, at each of the enrolled retail outlets. RESULTS: The primary outcome measure is the proportion of all ACTs that are sold to malaria test-positive clients in each study arm. For all secondary outcomes, we will evaluate the degree to which the interventions affect purchasing behavior among people seeking care for a febrile illness at the retail outlet. CONCLUSIONS: If our study demonstrates that malaria case management can be improved in the retail sector, it could reduce overconsumption of ACTs and enhance targeting of publicly funded treatment reimbursements, lowering the economic barrier to appropriate diagnosis and treatment for patients with malaria. TRIAL REGISTRATION: ClinicalTrials.gov NCT04428307 , registered June 9, 2020, and NCT04428385 , registered June 9, 2020.

Subsidise the test, the treatment or both? Results of an individually randomised controlled trial of the management of suspected malaria fevers in the retail sector in western Kenya.

INTRODUCTION: In many malaria-endemic countries, the private retail sector is a major source of antimalarial drugs. However, the rarity of malaria diagnostic testing in the retail sector leads to overuse of the first-line class of antimalarial drugs known as artemisinin-combination therapies (ACTs). The goal of this study was to identify the combination of malaria rapid diagnostic test (RDT) and ACT subsidies that maximises the proportion of clients seeking care in a retail outlet that choose to purchase an RDT (RDT uptake) and use ACTs appropriately. METHODS: 842 clients seeking care in 12 select retail outlets in western Kenya were recruited and randomised into 4 arms of different combinations of ACT and RDT subsidies, with ACT subsidies conditional on a positive RDT. The outcomes were RDT uptake (primary) and appropriate and targeted ACT use (secondary). Participants' familiarity with RDTs and their confidence in test results were also evaluated. RESULTS: RDT uptake was high (over 96%) across the study arms. Testing uptake was 1.025 times higher (98% CI 1.002 to 1.049) in the RDT subsidised arms than in the unsubsidised groups. Over 98% of clients were aware of malaria testing, but only 35% had a previous experience with RDTs. Nonetheless, confidence in the accuracy of RDTs was high. We found high levels of appropriate use and targeting of ACTs, with 86% of RDT positives taking an ACT, and 93.4% of RDT negatives not taking an ACT. The conditional ACT subsidy did not affect the RDT test purchasing behaviour (risk ratio: 0.994; 98% CI 0.979 to 1.009). CONCLUSION: Test dependent ACT subsidies may contribute to ACT targeting. However, in this context, high confidence in the accuracy of RDTs and reliable supplies of RDTs and ACTs likely played a greater role in testing uptake and adherence to test results.

Propensity score weighting under limited overlap and model misspecification.

Propensity score weighting methods are often used in non-randomized studies to adjust for confounding and assess treatment effects. The most popular among them, the inverse probability weighting, assigns weights that are proportional to the inverse of the conditional probability of a specific treatment assignment, given observed covariates. A key requirement for inverse probability weighting estimation is the positivity assumption, i.e. the propensity score must be bounded away from 0 and 1. In practice, violations of the positivity assumption often manifest by the presence of limited overlap in the propensity score distributions between treatment groups. When these practical violations occur, a small number of highly influential inverse probability weights may lead to unstable inverse probability weighting estimators, with biased estimates and large variances. To mitigate these issues, a number of alternative methods have been proposed, including inverse probability weighting trimming, overlap weights, matching weights, and entropy weights. Because overlap weights, matching weights, and entropy weights target the population for whom there is equipoise (and with adequate overlap) and their estimands depend on the true propensity score, a common criticism is that these estimators may be more sensitive to misspecifications of the propensity score model. In this paper, we conduct extensive simulation studies to compare the performances of inverse probability weighting and inverse probability weighting trimming against those of overlap weights, matching weights, and entropy weights under limited overlap and misspecified propensity score models. Across the wide range of scenarios we considered, overlap weights, matching weights, and entropy weights consistently outperform inverse probability weighting in terms of bias, root mean squared error, and coverage probability.