Combating unmeasured confounding in cross-sectional studies: evaluating instrumental-variable and Heckman selection models

Bias? Clarifying the language barrier between epidemiologists and economists

In health intervention research, epidemiologists and economists are more and more interested in estimating causal effects based on observational data. However, collaboration and interaction between both disciplines are regularly clouded by differences in the terminology used. Amongst others, this is reflected in differences in labeling, handling, and interpreting the sources of bias in parameter estimates. For example, both epidemiologists and economists use the term selection bias. However, what economists label as selection bias is labeled as confounding by epidemiologists. This paper aims to shed light on this and other subtle differences between both fields and illustrate them with hypothetical examples. We expect that clarification of these differences will improve the multidisciplinary collaboration between epidemiologists and economists. Furthermore, we hope to empower researchers to select the most suitable analytical technique from either field for the research problem at hand.

Impact of Integrating Machine Learning in Comparative Effectiveness Research of Oral Anticoagulants in Patients with Atrial Fibrillation

We aimed to compare the ability to balance baseline covariates and explore the impact of residual confounding between conventional and machine learning approaches to derive propensity scores (PS). The Health Insurance Review and Assessment Service database (January 2012-September 2019) was used. Patients with atrial fibrillation (AF) who initiated oral anticoagulants during July 2015-September 2018 were included. The outcome of interest was stroke/systemic embolism. To estimate PS, we used a logistic regression model (i.e., a conventional approach) and a generalized boosted model (GBM) which is a machine learning approach. Both PS matching and inverse probability of treatment weighting were performed. To evaluate balance achievement, standardized differences, p-values, and boxplots were used. To explore residual confounding, E-values and negative control outcomes were used. In total, 129,434 patients were identified. Although all baseline covariates were well balanced, the distribution of continuous variables seemed more similar when GBM was applied. E-values ranged between 1.75 and 2.70 and were generally higher in GBM. In the negative control outcome analysis, slightly more nonsignificant hazard ratios were observed in GBM. We showed GBM provided a better ability to balance covariates and had a lower impact of residual confounding, compared with the conventional approach in the empirical example of comparative effectiveness analysis.

The Trajectory of Subjective Well-Being: A Partial Explanation of the Marriage Advantage

Although several studies have documented a distinct marriage advantage in well-being, it is still unclear what it is about marriage that renders this benefit. We hypothesize that it is due to factors theorized to accrue to matrimony, such as elevated financial status and specific social psychological supports. We examine the trajectory of subjective well-being for 1,135 respondents from the three-wave 2010 GSS panel survey utilizing linear mixed-effects modeling. We find that about two-fifths of the marriage advantage in subjective well-being is accounted for by a mixture of control variables, finances, and emotional factors, with most of this due to elements that are associated with the marital context. Higher annual income, enhanced interpersonal trust, greater sociability, and less of a sense of loneliness and isolation appear to be responsible for a substantial component of the marital advantage. We further find that the marriage advantage is invariant to both race and gender.

Availability and Affordability of Primary Health Care Among Vulnerable Populations in Urban Kumasi Metropolis: Family Health Perspective

Purpose

Health-related expenditures pose a significant burden on vulnerable populations. This study assessed the availability and affordability of primary health care among disadvantaged populations in urban Kumasi Metropolis, Ghana.

Methods

This study was a descriptive cross-sectional study conducted among multi-level participants of vulnerable populations ≥18 years of age (n=710) constituting the older adults/aged, pregnant women, head porters, sex workers, and other vulnerable groups (people with disabilities and the homeless). Data were collected using a semistructured questionnaire. Poisson regression with robust variance was used to assess the association between vulnerability and access to health care.

Results

There were significant differences in the availability and adequacy of health care among the vulnerable groups studied. Distance to the source of care was >5 km for majority of the vulnerable groups and the average expenditure on a visit to the health facility was GH¢ 27.04 (∼US$ 5.55 as at January 2019). Challenges to health care among the vulnerable groups included monetary (37.9%), stigmatization (18.6%), and staff attitude (25.9%). Head porters and other vulnerable groups were less likely to view health care as affordable compared with older adults. The difference in the perception of health care affordability was, however, explained by sociodemographic characteristic and health care-related factors.

Conclusion

Despite the introduction of a National Health Insurance Scheme in Ghana, this study highlights challenges in health care access among vulnerable populations independent of the type of vulnerability. This suggests the need for stakeholders to adopt other innovative care strategies that may have broader applicability for all populations.

Forgone healthcare and financial burden due to out-of-pocket payments in Bangladesh: a multilevel analysis

BACKGROUND

Ensuring access to health services for all is the main goal of universal health coverage (UHC) plan. Out-of-pocket (OOP) payment still remains the main source of funding for healthcare in Bangladesh. The association between barriers to accessing healthcare and over-reliance on OOP payments has not been explored in Bangladesh using nationally representative household survey data. This study is a novel attempt to examine the burden of OOP payment and forgone healthcare in Bangladesh, and further explores the inequalities in catastrophic health expenditures (CHE) and forgone healthcare at the national and sub-national levels.

METHODS

This study used data from the most recent nationally representative cross-sectional survey, Bangladesh Household Income and Expenditure Survey, conducted in 2016-17 (N = 39,124). In order to identify potential determinants of CHE and forgone healthcare, multilevel Poisson regression was used. Inequalities in CHE and forgone healthcare were measured using the slope index of inequality.

RESULTS

Around 25% of individuals incurred CHE and 14% of the population had forgone healthcare for any reasons. The most common reasons for forgone healthcare were treatment cost (17%), followed by none to accompany or need for permission (5%), and distance to health facility (3%). Multilevel analysis indicated that financial burden and forgone care was higher among households with older populations or chronic illness, and those who utilize either public or private health facilities. Household consumption quintile had a linear negative association with forgone care and positive association with CHE.

CONCLUSION

This study calls for incorporation of social safety net in health financing system, increase health facility, and gives priority to the disadvantaged population to ensure access to health services for all.

Cultural differences in geographic mobility readiness among business management students in Germany and Spain ahead of graduation

Given the increased internationalization of organizations and economies of scale concentrated in urban centers, graduates are often expected to relocate for their first job. Based on Hofstede’s model and Theory of Planned Behavior (TPB), we examine the effects of cultural dimensions (individualism-collectivism and uncertainty tolerance) as well as subjective norms (parents’ and peers’ attitudes towards geographic mobility) on readiness for geographic job-related mobility in samples of German and Spanish business management students ahead of graduation from university. The study involved administering a survey questionnaire to 273 third- and fourth-year business management students of two large universities (one in Germany and another in Spain). Cross-cultural measurement invariance of the measures was confirmed, allowing for comparison of scores across the groups. We found that German students had generally a larger geographical mobility, whose readiness was predicted by parents’ and peers’ perceived attitude and uncertainty tolerance. Readiness for geographic mobility was also higher when social or material incentives are offered, yet geographic mobility readiness for career incentives and for social incentives was predicted by vertical individualism and horizontal collectivism, respectively. This study is one of the first to examine geographic mobility readiness among undergraduate management students in the two countries, who by nature of their training are expected to be mobile. The study also shows the differential effects of sub-dimensions of the Hofstede cultural dimensions.

Associations between older African academics' physical activity, walkability and mental health: a social distancing perspective

This study aimed to assess the moderating influence of neighborhood walkability on the association between physical activity (PA) and mental health among older African academics aged 50 years or more in cities with social distancing protocols in response to the Coronavirus disease 2019 (COVID-19). A total of 905 volunteer academics participated in the study. A hierarchical linear regression analysis was employed to conduct sensitivity analyses and test the study hypotheses. After controlling for sex, education and age, there was a positive association between PA and mental health. Neighborhood walkability moderated the relationship between PA and mental health, which suggests that during the pandemic PA was associated with higher mental health scores in more walkable neighborhoods. The study concludes that PA was beneficial to mental health in the social distancing context and was associated with higher mental health in more walkable neighborhoods, particularly in a social distancing context.

Determinants of healthcare seeking and out-of-pocket expenditures in a "free" healthcare system: evidence from rural Malawi

BACKGROUND

Monitoring financial protection is a key component in achieving Universal Health Coverage, even for health systems that grant their citizens access to care free-of-charge. Our study investigated out-of-pocket expenditure (OOPE) on curative healthcare services and their determinants in rural Malawi, a country that has consistently aimed at providing free healthcare services.

METHODS

Our study used data from two consecutive rounds of a household survey conducted in 2012 and 2013 among 1639 households in three districts in rural Malawi. Given our explicit focus on OOPE for curative healthcare services, we relied on a Heckman selection model to account for the fact that relevant OOPE could only be observed for those who had sought care in the first place.

RESULTS

Our sample included a total of 2740 illness episodes. Among the 1884 (68.75%) that had made use of curative healthcare services, 494 (26.22%) had incurred a positive healthcare expenditure, whose mean amounted to 678.45 MWK (equivalent to 2.72 USD). Our analysis revealed a significant positive association between the magnitude of OOPE and age 15-39 years (p = 0.022), household head (p = 0.037), suffering from a chronic illness (p = 0.019), illness duration (p = 0.014), hospitalization (p = 0.002), number of accompanying persons (p = 0.019), wealth quartiles (p₂ = 0.018; p₃ = 0.001; p₄ = 0.002), and urban residency (p = 0.001).

CONCLUSION

Our findings indicate that a formal policy commitment to providing free healthcare services is not sufficient to guarantee widespread financial protection and that additional measures are needed to protect particularly vulnerable population groups.

Marriage Advantage in Subjective Well-Being: Causal Effect or Unmeasured Heterogeneity?

I investigate whether the marriage advantage in subjective well-being is a true protective effect vs. being attributable to self-selection into (or out of) marriage based on pre-existing mental health. I utilize 1,240 respondents from the GSS panel, a three-wave longitudinal survey collected from 2010-2014. I use a pseudo-treatment approach to informally test for the presence of self-selection. This is followed by a fixed-effect regression analysis to eliminate its influence when estimating the marriage effect. Results support the existence of self-selection: the currently married who in later waves will be exiting marriage are already more distressed than other married respondents in wave 1. And the currently not married who in later waves will be entering marriage are not more distressed in wave 1 than those remaining continuously married. A protective effect is also supported: at any given time, net of self-selection, the currently married are less distressed than the unmarried.

Statistical learning theory for high dimensional prediction: Application to criterion-keyed scale development

Statistical learning theory (SLT) is the statistical formulation of machine learning theory, a body of analytic methods common in "big data" problems. Regression-based SLT algorithms seek to maximize predictive accuracy for some outcome, given a large pool of potential predictors, without overfitting the sample. Research goals in psychology may sometimes call for high dimensional regression. One example is criterion-keyed scale construction, where a scale with maximal predictive validity must be built from a large item pool. Using this as a working example, we first introduce a core principle of SLT methods: minimization of expected prediction error (EPE). Minimizing EPE is fundamentally different than maximizing the within-sample likelihood, and hinges on building a predictive model of sufficient complexity to predict the outcome well, without undue complexity leading to overfitting. We describe how such models are built and refined via cross-validation. We then illustrate how 3 common SLT algorithms-supervised principal components, regularization, and boosting-can be used to construct a criterion-keyed scale predicting all-cause mortality, using a large personality item pool within a population cohort. Each algorithm illustrates a different approach to minimizing EPE. Finally, we consider broader applications of SLT predictive algorithms, both as supportive analytic tools for conventional methods, and as primary analytic tools in discovery phase research. We conclude that despite their differences from the classic null-hypothesis testing approach-or perhaps because of them-SLT methods may hold value as a statistically rigorous approach to exploratory regression. (PsycINFO Database Record

Causal Models for Mediation Analysis: An Introduction to Structural Mean Models

Mediation analyses are critical to understanding why behavioral interventions work. To yield a causal interpretation, common mediation approaches must make an assumption of "sequential ignorability." The current article describes an alternative approach to causal mediation called structural mean models (SMMs). A specific SMM called a rank-preserving model (RPM) is introduced in the context of an applied example. Particular attention is given to the assumptions of both approaches to mediation. Applying both mediation approaches to the college student drinking data yield notable differences in the magnitude of effects. Simulated examples reveal instances in which the traditional approach can yield strongly biased results, whereas the RPM approach remains unbiased in these cases. At the same time, the RPM approach has its own assumptions that must be met for correct inference, such as the existence of a covariate that strongly moderates the effect of the intervention on the mediator and no unmeasured confounders that also serve as a moderator of the effect of the intervention or the mediator on the outcome. The RPM approach to mediation offers an alternative way to perform mediation analysis when there may be unmeasured confounders.

Effect of retention in elementary grades on grade 9 motivation for educational attainment

This study investigated the effect of grade retention in elementary school on students' motivation for educational attainment in grade 9. We equated retained and promoted students on 67 covariates assessed in grade 1 through propensity score weighting. Retained students (31.55%, nretained=177) and continuously promoted students (68.45%, npromoted=384) were compared on the bifactor model of motivation for educational attainment (Cham, Hughes, West & Im, 2014). This model consists of a General factor (student's overall motivation for educational attainment), and three specific factors: student perceived Teacher Educational Expectations, Peer Educational Aspirations, and Value of Education. Measurement invariance between retained and promoted groups was established. Retained students scored significantly higher than promoted students on each specific factor but not on the General factor. Results showed that the retained and promoted students did not significantly differ on the General factor. The retained students had significantly higher scores on each specific factor than those of the promoted students. The results suggested that grade retention may not have the negative effects so widely assumed in the published literature; it is an expensive intervention with minimal evidence of benefits to the retained student.