PH784: Final Exam

Author

Koichiro Shiba

Published

November 11, 2025

Introduction

Read the following article from JAMA Pediatrics as if you were a peer-reviewer of this paper.

Note that I chose this paper for this assignment not because it is low quality with many weaknesses—in fact, this is a well-written and high-quality study. However, no study is perfect. The gaps you identify can present good opportunities for future research. This is how each individual (imperfect) study contributes to science.

Briefly, this study examined the relationships between two widely used measures of neighborhood socioeconomic environment and asthma incidence among children at three different life stages (birth, infancy, and early childhood). The neighborhood measures included the Social Vulnerability Index (SVI) and Child Opportunity Index (COI) at the census tract level. Their main findings are shown in Figure 3 (i.e., adjusted incidence rate ratios, IRRs). In this exam, you will focus on results concerning COI.

Based on what you have learned in the course, critically evaluate the article by answering the questions below. Work alone and do not discuss this exam with classmates.

Tip
  • Avoid providing unrelated responses and focus on answering the questions you are asked
  • Each question asks you to discuss multiple points, so make sure you address all of them

Q1

Describe how the authors operationalized the outcome (“asthma incidence”) and discuss the validity of what they assessed as a measure of asthma incidence. If there is a validity concern, explain how the lack of validity may influence their main findings—does it underestimate or overestimate the true effect? Justify your response.

Your answer

Q2

Briefly describe how the COI scores were calculated. Then, explain how they defined and operationalized the “neighborhood” environment exposure based on the COI scores. Discuss the pros and cons of their approach and suggest at least one alternative approach to operationalize neighborhood exposure, including its pros and cons.

Your answer

Q3

Provide the causal interpretation of the adjusted association with asthma incidence for the “very high” COI in early childhood (upper right panel of Figure 3): IRR = 0.80 (95% CI: 0.68-0.94).

Your answer

Q4

Describe the specific exchangeability assumption for causal identification required in this study. Use specific constructs rather than providing a generic description of the assumption. Then, discuss whether you think the assumption is reasonable. Do not just say there may be unmeasured confounders due to the observational nature of the study. First, consider what covariates they controlled for and WHEN they were measured, and discuss whether the control was appropriate. Second, list one potential unmeasured confounder and qualitatively discuss the bias the confounder may induce.

Your answer

Q5

Describe the specific consistency assumption for causal identification required in this study. Use specific constructs rather than providing a generic description of the assumption. Then, describe the hypothetical intervention they are conceptualizing and discuss whether you think the consistency assumption is reasonable. Justify your answer.

Your answer

Q6

Suppose the causal identification assumptions are reasonably met. Instead of the causal estimand the authors used in this paper (i.e., your response to Q3), 1) suggest an alternative causal estimand based on a stochastic intervention for the COI and 2) explain the potential advantages of the new estimand.

Your answer

Q7

Although the authors reported they found no multiplicative effect heterogeneity by sex, race and ethnicity, and rurality, they did not provide the numeric results (which they should have). In general, regression analysis with a product term (e.g., \(race \times COI\)) is statistically underpowered. Suppose they repeated the same study with a much larger sample size and found effect heterogeneity by race and ethnicity where COI is associated with even lower rates of asthma incidence among white group compared to non-white groups. Discuss 1) what might explain this observed effect heterogeneity and 2) the policy implications of the finding (i.e., what insights do you gain from such a finding and how would you use them?).

Your answer