All study protocols were approved by Tri-cities Research Ethics Board (THREB), as well as by the Here 24/7 Ethics Committee and Waterloo Regional Police.
Population & Context
Data were captured for the Waterloo-Wellington Local Health Integration Network (LHIN), which is a region in Southern Ontario consisting of Kitchener, Waterloo, and Cambridge, with a total census population of 523 894 (as of November 26, 2020, per Statistics Canada, available at https://www12.statcan.gc.ca/census-recensement/2016/as-sa/fogs-spg/Facts-cma-eng.cfm?LANG=Eng&GK=CMA&GC=541&TOPIC=1). The region is served by three hospitals: Grand River Hospital (GRH), St. Mary’s Hospital (SMH), and Cambridge Memorial Hospital (CMH). All residents of this region with a phone would have access to police services and crisis hotlines.
In the province of Ontario, a state of emergency was declared on March 17, 2020, which directly impacted the Waterloo-Wellington LHIN. The lockdown period included school, university, and playground closures, closure of non-essential businesses, and prohibition of non-essential public gatherings of over five people followed on April 5, 2020. Graduated rollback of restrictions occurred between May 4 and June 2, 2020, with recommendations on mask wearing, social distancing, and limits on large gatherings maintained. Further details on specific restrictions can be found in Appendix 1.
Measures of Mental Health Related ER Visits.
Daily total emergency department visits were collected for GRH, SMH, and CMH between March 5 and September 5, 2020. Anonymous, retrospectively coded National Ambulatory Care Reporting System (NACRS) chart metadata for each of the three hospitals were obtained through the respective Decision Support teams mental health discharge diagnoses: substance related (excluding alcohol), alcohol intoxication, mood related (anxiety, PTSD, depression, and bipolar disorder), psychosis-related (psychosis, bizarre behaviour), situation related (situational disturbance, life crisis, concern for safety, and domestic violence), self harm related, and completion of Form 1 indicating involuntary detention for psychiatric assessment. Patients from all sites were pooled and considered as a single population. We enumerated the daily number of visits for each diagnosis over the study period. The specific ICD-10 diagnostic codes used to produce each mental health category are described in Appendix 2.
Mental Health Related Police Service Use.
Waterloo Regional Police provided population-level data for the observation period and comparison period on police dispatches, as well as Neighbourhood Policing semi-monthly reports. Both sources of data were used to create the following categories: assault, domestic dispute, intoxication, and suicide attempts. We enumerated the daily numbers for each category of police call.
Crisis-Line Use.
Call volumes were obtained from Here 24/7, a call centre which serves as the single access point for mental health, addictions, and crisis services in the Waterloo-Wellington region. Crisis-line call volumes for Waterloo-Wellington are tracked cumulatively, and we enumerated the number of daily calls. Data on call location, presenting issues, and outgoing referral sources (i.e. emergency dispatch) were not available.
Covid Cases.
Publicly available Waterloo Public Health data was used to track progression of the epidemic (as of January 19, 2021, per Region of Waterloo Public Health, available at https://www.regionofwaterloo.ca/en/health-and-wellness/positive-cases-in-waterloo-region.aspx#).
Data Analysis.
Daily total Emergency Department volumes were examined between March 5 and September 5 in both 2019 and 2020. Daily new COVID-19 cases were also examined over this period in 2020. These trends were smoothed using the LOESS (locally estimated scatterplot smoothing) method26, with a 95% confidence interval.
Descriptive analyses of the included mental health diagnostic categories were performed, and trends in the number of diagnoses in each category over time were presented visually using line graphs. Crude values are included in the supplementary materials.
ED visits, mental health diagnoses, police responses, and calls to the Here 24/7 crisis line during the first wave lockdown (March 17 to May 4, 2020) were compared to ED visits, mental health diagnoses, police responses, and crisis calls over the same period in 2019, in order to compare the lockdown period to baseline. Although data was collected from March 5 to September 5, we specifically examined the March 17 to May 4 period to study the effects of lockdown. For each category, we used a univariate quasi-Poisson regression27 with a log link function and year as the only predictor. Quasi-Poisson models were chosen to account for the over-dispersed count data, as the residual deviance was often greater than the degrees of freedom28. Regression coefficients, standard error, t-values, and p-values are reported for each model. Statistical significance was established at p < 0.05. Models were created using the `glm` R function.
Significant changes in ED visits, mental health diagnoses, police responses, and calls to the crisis line from March 5 to September 5, 2020 were examined using changepoint analyses. The changepoint models found significant changes in mean and variance using the Pruned Exact Linear Time (PELT) method29 with a range of penalties between twice the log of the number of observations and 100 times the log of the number of observations. Diagnostic plots comparing the number of changepoints and the penalty values (see Supplementary Materials) were used to find the optimal number of changepoints. We tried to find the “elbow” of these plots, minimizing both the penalty value and the number of changepoints30. The locations of those changepoints and the means of each segment were then calculated. Models were created using the `cpt.meanvar` function from the `changepoint` R package31.
R (version 4.02) was used to calculate all tabulations and statistics. Code is available at https://github.com/alechay/covid19-mh