Use the CRF Critical Appraisal Tool
May 19, 2025
Dear Dr. Lesko and Dr. Schisterman,
I would like to provide some editorial comments on Neighborhood eviction trajectories and odds of moderate and serious psychological distress during pregnancy among African American women, which was published in the American Journal of Epidemiology in March 2024.
1. Sample significantly reduced - The authors rely on "women who reported their address 2 years before study enrollment and at study enrollment, and whose address was able to be matched to a latitude and longitude using ArcGIS 10.8.2 (n = 808)". This dramatically reduced the analytic sample from 1,410 to 808. Without further analysis between the total and analytic samples, the study findings cannot be fully trustworthy. This is also related to the study design likely not originally intended for the analysis in the paper because address history would have likely been more fully documented.
2. Census block data - Census data at the block-level can be apt to considerable instability. The margins of error tend to increase with the smaller size of the block. The authors did not address how this might be a threat to internal validity. An unstable variable at the block would introduce considerable noise in the data.
3. Race Essentialism - Because the LIFE study only recruited African American women, the structural effect on moderate psychological distress (MPD) and serious psychological distress (SPD) has analytical limitations. In other words, a more inclusive study (across racial groups) would strengthen the neighborhood effects on MPD and SPD. As I describe in the Critical Race Framework, race essentialism has major limitations. [Study here]. In terms of this study, the authors assume an African American/Black monolith in the Detroit metropolitan area. Regression assumptions on independence among observations would raise important questions. Are observations clustered? Is it feasible that many observations (e.g., city of Detroit proper versus Detroit suburbs) have spatial or temporal dependence? There are significant health outcomes between Detroit and its environs.
In other words, even for shifts between medium to high in preconception to during-pregnancy, would study participants within the city of Detroit have a distinct set of structural changes? In addition, a change in city blocks (e.g., medium to high) may still fall within the same "structure" (e.g., under-resourced area of the city).
4. Covariate Missingness - The authors report 0-10% missingness for covariates. They did not impute or examine the effect that this may have had on data quality - another threat to internal validity.
5. Overcontrolled - My expert judgment is that the authors have considerable risks due to over-controlling. They controlled for age, educational attainment, duration of time lived in the during-pregnancy neighborhood, and a block-group–level neighborhood disadvantage index, which likely inflated the coefficients. There was a lack of citations to justify these decisions - only one citation. The authors also do not report model fit statistics.
6. Noisy Data and Effect on Outcomes - When there are "noisy" variables in a regression, it can inflate race coefficients. I simulated data here [Misclassification Study]. Based on the above, I do not find the data results trustworthy - high attenuated, in my opinion.
7. Practical Significance - If threats to internal validity were assumed to be minimal as discussed above, the practical significance of the study is unclear. Given the lower odds for neighborhood eviction filing and judgment rates for MPD and SPD for high/high, the study results would suggest that maintaining residence in neighborhoods with high eviction/judgment rates would reduce odds for MPD and SPD or switching from High/low should be avoided (Table 6). This does not seem just as a matter of public health practice. Again, I do not believe that the data is trustworthy, but publication in your study could be harmful to public health policy.
Sincerely,
Christopher Williams
Principal Investigator, Critical Race Framework Study
https://www.criticalraceframework.com/home
