Methods: We used monthly patient-level discharge data for hospitalizations from the Healthcare Utilization Project (HCUP) database for 20 states between 2007-2014. We selected 9 mental health outcomes, which include the Clinical Classifications Software (CCS) categories using ICD-9 codes. The mental health categories included anxiety disorders, alcohol-related disorders, mood disorders, substance-related disorders, suicide and intentional self-inflicted injury, and schizophrenia and other psychotic disorders. We created three additional aggregate categories: alcohol or substance-related disorders, anxiety or mood disorders, and all mental health. We used the United States Drought Monitor (USDM) for our drought exposure metric to determine the severity of drought for a given month during the study period. USDM has 5 drought categories, no drought, D0, D1, D2, D3, and D4. We created three drought categories from this, no drought (no drought, D0), moderate drought (D1, D2), and severe drought (D3, D4). To compare different drought metrics, we selected the Standardized Precipitation Evapotranspiration Index (SPEI) at the 6-month timescale, as this timescale is likely to capture potential health impacts from drought conditions. SPEI uses a continuous numeric scale to measure drought from -3 (driest conditions) to 3 (wettest conditions). We converted this into equivalent USDM categories and followed the same procedure as discussed for creating three categories from USDM. For this analysis, we ran a two-stage hierarchical regression model. The first stage determined the county-level association between drought and mental health hospital visits, using a population threshold of 50,000 residents per county. 352 out of 1017 counties were included using this criterion. In the second stage, we performed a meta-analysis to determine an overall effect estimate of drought's impact on mental health visits. We performed sensitivity tests to determine the impact on the effect estimates based on urbanicity, lagged drought effects, economic performance, and social vulnerability.
Results: We observe a dose-response relationship with the mental health outcomes increasing with the severity of the drought measure. When comparing severe drought to no drought, the all mental health category showed a 4.35% increase in hospitalizations. The statistical significance of these associations was more common across the mental health outcomes for SPEI compared to USDM. We did not observe any difference between using non-lagged and lagged measures of drought. We observed a larger impact of drought in areas with lower social vulnerability, and in urban areas. Anxiety disorders showed a 5.72% increase in hospitalizations. This value increased to 6.10% in counties with lower social vulnerability, and 6.98% in urban counties. We did not observe any difference in the association across counties identified as higher or lower than the mean US GDP levels for counties.
Conclusions: The analysis of the data suggests that severe drought conditions show correlation with increased hospital visits for mental health conditions, especially anxiety and substance-related disorders. This fills an important gap in the literature where there has been limited quantification of the mental health impacts of drought in the U.S. Limitations of this study include having limited data due to gaps in the HCUP dataset and the month-level temporal scale. Some counties, particularly those in the Midwest and Northeast, have very few months that reach severe drought conditions. This introduces some volatility into the analysis, where the effect estimates can vary widely due to such limited data points during severe drought. Further analysis will be needed to more precisely quantify the impact of drought on mental health hospital visits in areas where severe drought is rarer, and assess how impacts may vary based on age, sex, race, and other social determinants of health.
The findings and conclusions in this presentation have not been formally disseminated by the Centers for Disease Control and Prevention and should not be construed to represent any agency determination or policy.

