Validation of an EMR algorithm to measure the prevalence of ADHD in the Canadian Primary Care Sentinel Surveillance Network (CPCSSN)

Rachael Morkem; Kenneth Handelman; John A Queenan; Richard Birtwhistle; David Barber

doi:10.1186/s12911-020-01182-2

Validation of an EMR algorithm to measure the prevalence of ADHD in the Canadian Primary Care Sentinel Surveillance Network (CPCSSN)

BMC Med Inform Decis Mak. 2020 Jul 20;20(1):166. doi: 10.1186/s12911-020-01182-2.

Authors

Rachael Morkem¹, Kenneth Handelman², John A Queenan³, Richard Birtwhistle⁴, David Barber⁵

Affiliations

¹ Research Associate, Centre for Studies in Primary Care, Queen's University, Kingston, Ontario, Canada. rachael.morkem@dfm.queensu.ca.
² Psychiatrist, Centre for Integrative Mental Health, Oakville, Ontario, Canada.
³ Senior Epidemiologist, Centre for Studies in Primary Care, Queen's University, Kingston, Ontario, Canada.
⁴ Professor of Family Medicine and Public Health Sciences, Centre for Studies in Primary Care, Queen's University, Kingston, Ontario, Canada.
⁵ Network Director and Assistant Professor, Centre for Studies in Primary Care, Queen's University, Kingston, Ontario, Canada.

Abstract

Background: Building and validating electronic algorithms to identify patients with specific disease profiles using health data is becoming increasingly important to disease surveillance and population health management. The aim of this study was to develop and validate an algorithm to find patients with ADHD diagnoses within primary care electronic medical records (EMR); and then use the algorithm to describe the epidemiology of ADHD from 2008 to 2015 in a Canadian Primary care sample.

Methods: This was a cross sectional time series that used data from the Canadian Primary Care Sentinel Surveillance Network (CPCSSN), a repository of primary care EMR data. A sample of electronic patient charts from one local clinic were manually reviewed to determine the positive predictive value (PPV) and negative predictive value (NPV) of an ADHD case-finding algorithm. In each study year a practice population was determined, and the algorithm was used to measure an observed prevalence of ADHD. The observed prevalence was adjusted for misclassification, as measured by the validity indices, to obtain an estimate of the true prevalence. Estimates were calculated by age group (4-17 year olds, 18 to 34 year olds, and 35 to 64 year olds) and gender, and compared over time.

Results: The EMR algorithm had a PPV of 98.0% (95% CI [92.5, 99.5]) and an NPV of 95.0% (95% CI [92.9, 98.6]). After adjusting for misclassification, it was determined that the prevalence of patients with a clinical diagnosis of ADHD has risen in all age groups between 2008 and 2015, most notably in children and young adults (6.92, 95% CI [5.62, 8.39] to 8.57, 95% CI [7.32, 10.00]; 5.73, 95% CI [4.40, 7.23] to 7.33, 95% CI [6.04, 8.78], respectively). The well-established gender gap persisted in all age groups across time but was considerably smaller in older adults compared to children and young adults.

Conclusion: Overall, the ADHD case-finding algorithm was found to be a valid tool to assess the epidemiology of ADHD in Canadian primary care practice. The increased prevalence of ADHD between 2008 and 2015 may reflect an improvement in the recognition and treatment of this disorder within primary care.

Keywords: Algorithm; Attention deficit-hyperactivity disorder; Prevalence; Validation.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Adolescent
Adult
Algorithms
Attention Deficit Disorder with Hyperactivity*
Canada
Child
Child, Preschool
Cross-Sectional Studies
Electronic Health Records*
Female
Humans
Male
Middle Aged
Prevalence
Primary Health Care
Sentinel Surveillance
Young Adult

Grants and funding

6020024/Shire Pharma Canada ULC/International