PC2080: Does Total Cholesterol Predict Mental Health Disease and Suicide Risk in Pediatric Patients?

Background: Mental Health (MH) disorders among children are rising. In 2005, Youth Risk Behavior Survey revealed that 8.4% of high school students in the US attempted suicide in the past year. Serum lipid levels have been suggested as a biomarker for aggressive behavior, depression and increased risk of suicide. In adult studies, the association between low serum cholesterol and neuropsychiatric disorders including suicidality has been described. Amongst pediatric and adolescent population, this data is limited and with conflicting results. Universal lipid screening for children was incorporated into AAP guidelines for well-child visits in 2014 with the focus on identifying children with high risk for dyslipidemia. However, if low cholesterol can predict MH diagnosis, it could be a valuable screening tool and help mitigate disease progression.

Methods: Data warehouse (DW) queried the records of patients aged 9-17 years old who had TC done and seen within the children's hospital system between 2011-2019. Children with BMI < 19, eating disorders and with no general consent were excluded. State death registry was also queried to identify children with suicide. Reference level used for T.C in children 2-17 years was < 170(desirable/low), 170-199(borderline high) and >200 (high). ICD-9 and 10 codes for MH diagnosis were used in the dataset. Chi-square and ANOVA analysis was performed to compare discrete and continuous variables between patients with and without MH diagnosis or suicidality. Linear and/or logistic regression model was used to test the effect of low cholesterol level and other confounding factors on MH diagnosis and suicidality.

Results: DW query yielded 64417 encounters of TC levels. In our data set, there were 53067 values < 170 and 11350 >170. MH Diagnosis were seen in 78.4% with low TC and 86.1% with TC >170. To further explore this bimodal distribution, by employing logistic regression, we regressed patient age, gender, TC level and TC level squared on the presence of a MH diagnosis within 5 years of the TC measurement for 49247 cholesterol tests (Table 1). Using a significance level of .05, we found significant effects for all 4 variables including the squared term for total cholesterol which demonstrates a clear non-linear trend (Figure 1). The hit rate was constructed around the mean of the distribution and resulted in a 3.9% increase in prediction. In our dataset, TC levels did not predict suicidality.

Conclusion: The results suggest a non-linear inverse relationship in the trend of TC level and MH diagnosis over a period of time depicting the importance of trend rather than specific cut-offs for TC levels. Further prospective studies are needed to explore the association of low TC and its association with MH diagnosis in order to use TC level as a biomarker for its predictability.