PC3045: Comparison of PHQ Screening Results Prior to and Following COVID-19 Pandemic Lockdown

Background: At our institution we have been engaging in routine screening for depression using the patient health questionnaire (PHQ) tool since 2016. Throughout this time, we have amassed a large volume of data. We sought to assess the impact of the COVID-19 lockdown on adolescent mental health by performing a retrospective analysis on patients that were screened in the year prior to and the year after the lockdown. We aimed to identify patient characteristics associated with positive PHQ screenings and evaluate the results using different comparison techniques.

Methods: A SQL-based report was created to abstract individual screening data from the medical record. Patient demographics, insurance information, and presence of behavioral health diagnosis in the patient’s problem list was matched to individual screening results. Screening data was split into two periods March 17, 2019 - March 16, 2020 (pre-lockdown) and March 17, 2020 - March 17, 2021 (post-lockdown. Data analysis was performed using MATLAB R2022a (Natick, MA) and STATA16 (College Station, TX). Matched analysis on patients who were screened in both time periods was performed using McNemar's Chi Square with 95% confidence intervals.

Results: There was a 1.4-fold increased risk of elevated PHQ scores post COVID-19 lockdown (95% CI = 1.3, 1.45). Suicidal ideation in screenings increased from 3.5% pre-lockdown to 5.6% post-lockdown (Table 1). 820 patients with initial high pre-lockdown scores ended with low scores (3.7% of total patients, 54.3% of patients with score ≥10). 87% of patients demonstrated resiliency with low scores in both the pre and post periods. No racial or ethnic differences were noted.

Conclusion: Using matched patient data and aggregated patient data allowed us to determine the true impact of the lockdown as analyzing screenings alone can lead to bias caused by overrepresentation of frequently screened patients and violates statistical analysis assumptions for independent data. Although we did note an increase in suicidal ideation as well as an increase in the proportion of positive screens (score ≥10), we did also find a group whose scores improved during this time which merits further analysis (Figure 1). In addition, it is important to explore the characteristics of those 87% of patients that demonstrated resiliency during this period. Our next step is to attempt to develop a machine learning algorithm to help predict those patients at risk for worsening.