<p><strong>BACKGROUND: </strong>Surface mining has become a significant method of coal mining in the Central Appalachian region of the eastern United States alongside the traditional underground mining. Concerns have been raised about the health effects of this surface mining, particularly mountaintop removal mining where coal is mined upon steep mountaintops by removing the mountaintop through clearcutting forests and explosives.</p>

<p><strong>METHODS: </strong>We used a control group design with a pretest and a posttest to assess the associations of surface mining in Central Appalachia with low birth weight and other adverse birth outcomes. The pretest period is 1977-1989, a period of low surface mining activity. We consider three posttest periods: 1990-1998, 1999-2011 and 2012-2017, with 1999-2011 as the primary analysis and the other periods as secondary analyses. Surface mining in Central Appalachia increased after 1989, partly resulting from the Clean Air Act Amendments of 1990 which made surface mining more financially attractive. For the primary analysis, we fit a logistic regression model of the primary outcome (low birth weight, &lt;2500 grams) on dummy variables for county and year; individual level maternal/infant covariates (maternal race, maternal age, infant sex and whether birth was a multiple birth); and the amount of surface mining during the year of the birth in the maternal county of residence.</p>

<p><strong>RESULTS: </strong>Our analysis sample consisted of 783,328 infants -- 482,284 infants born from 1977-2017 to women residing in substantial surface mining activity counties and 301,044 infants born from 1977-2017 to women residing in matched control counties. Compared to the pre-period of low surface mining from 1977-1989, for the primary analysis post-test period of 1999-2011, there was an estimated relative increase in low birth weight in surface mining counties compared to matched control counties that was not statistically significant (odds ratio for a 5 percentage point increase in area disturbed by surface mining: 1.07, 95% confidence interval (0.96, 1.20), p-value: .22). For the secondary analysis post-test period of 1990-1998, there was no increase (odds ratio: 0.91, 95% confidence interval: (0.74, 1.13), p-value: .41). For the secondary analysis post-test period of 2012-2017, there was a statistically significant relative increase (odds ratio: 1.28, 95% confidence interval: (1.08, 1.50), p-value: .004). Qualitatively similar results were found for the outcomes of very low birth weight, preterm birth and small-for-gestational age.</p>

<p><strong>CONCLUSIONS: </strong>We examined the hypothesis that surface mining activity in Central Appalachia contributes to low birth weight using an observational study. We found evidence in secondary analyses that surface mining was associated with low birth weight in the 2012-2017 time period and potentially beginning in the early to mid 2000's. Evidence for an association was not found prior to 2000. A potential explanation for this pattern of association is that surface mining caused an increase in low birth weight but its onset was delayed. Future research is needed to clarify the findings and if replicated, identify the mechanism necessary to mitigate the impacts of mining on adverse birth outcomes.</p>

<p>Two problems that arise in making causal inferences for nonmortality outcomes such as bronchopulmonary dysplasia (BPD) are unmeasured confounding and censoring by death, ie, the outcome is observed only when subjects survive. In randomized experiments with noncompliance and no censoring by death, instrumental variable (IV) methods can be used to control for the unmeasured confounding. But, when there is censoring by death, the average causal treatment effect cannot be identified under usual assumptions but can be studied for a specific subpopulation by using sensitivity analysis with additional assumptions. However, evaluating the local average treatment effect (LATE) in observational studies with censoring by death problems while controlling for unmeasured confounding is not well studied. We develop a novel sensitivity analysis method based on IV models for studying the LATE. Specifically, we present the identification results under an additional assumption and propose a three-step procedure for the LATE estimation. Also, we propose an improved two-step procedure by simultaneously estimating the instrument propensity score (ie, the probability of instrument given covariates) and the parameters induced by the assumption. We show with simulation studies that the two-step procedure can be more robust and efficient than the three-step procedure. Finally, we apply our sensitivity analysis methods to a study on the effect of delivery at high-level neonatal intensive care units on the risk of BPD.</p>