Estimating Incidence of Infectious Diseases in Canada Beyond Surveillance

Abstract

Introduction Public health surveillance provides valuable information about the epidemiologic trends of infectious diseases but is not designed to provide exact estimates of population-level disease incidence. Case reporting to public health surveillance may be impacted due to a variety of reasons: clinical misdiagnosis, sensitivity and specificity of laboratory testing assays producing false negative results, incomplete case information, failure to report to public health, patient access to healthcare resources, etc. Estimating the true incidence of a disease is particularly important when conducting health economic assessments that evaluate the value of public health interventions. Equally important is to investigate sources of bias in reporting by evaluating differences in case ascertainment by sub-group (age, race, socioeconomic status (SES), gender, etc). The aim of my thesis is to explore alternative methods of estimating incidence of infectious disease of public health importance and where possible, explore potential sources of bias. Study 1: Using Seroprevalence to Estimate Incidence of SARS-CoV-2 in Canada For study 1, a systematic literature review was conducted of SARS-CoV-2 seroprevalence studies in Canada. Disease incidence was estimated from seroprevalence studies, which were then grouped by region and age. Estimated incidence was compared to reported cases by calculating under-ascertainment ratios for each study. Under-ascertainment of cases was highest at the beginning of the pandemic, as access to testing was limited early on. Over time, the under-ascertainment declined, perhaps due to increased access to testing, but may have been confounded by reports of reduced assay sensitivity of thresholds to lower titre levels over time. This work was published in July 2022. Study 2: Estimating Incidence of Lyme Disease in Manitoba using Administrative Data Lyme Disease (LD) is a growing public health threat in Canada due to the impact of environmental factors that favour the expansion of Borrelia burgdorferi-infected ticks (Ixodes scapularis) into Canada. It is widely accepted that LD cases are under reported in Canada, but there are no empirical estimates that quantify true incidence of disease. An algorithm developed by the Center for Disease Control, using administrative healthcare billing codes (ICD-9, ICD-10) for Lyme and prescription data, was clinically validated in several endemic Lyme regions within the US. Access to administrative claims databases is province specific with variation in the types of linked datasets available for research. To ensure compliance with the validated CDC protocol, we needed to have physician billing codes that were specific for LD, hospital discharge abstracts, emergency room visits and antibiotic prescriptions. Our investigation was limited to the eastern Canadian provinces, which is the only habitat for I. scapularis in Canada. Manitoba was selected as an ideal province to conduct this study as it met all the database criteria as well as provided access to public health serology test results for LD, which was not available to the CDC in their study. The algorithm was adapted to the Manitoba database and a protocol was submitted to the Manitoba data custodian for extraction. The primary objective was to to estimate the incidence of LD from 2010-2021, overall, by patient characteristics (e.g., age, sex), and by local residential area. Secondary objectives were to describe the demographic and clinical characteristics of LD cases, and to estimate the clinical stage of patients based upon reported symptoms. The yearly incidence of Lyme cases identified using this method were between 5.1 and 11.0 times higher than those reported to public health surveillance during the study period. There were no clear trends in the under-ascertainment over time. Due to the limitations of using administrative billing data, it was not possible to estimate the clinical stage of identified cases.

Study 3: Modeling the impact of elevated LD incidence on population health in Manitoba, Canada LD is an emerging and growing public health threat in Canada due to climate change factors responsible for the rapid expansion of Borrelia burgdorferi-infected ticks (Ixodes scapularis) into densely populated areas. Under-reporting cases to public health and delays early treatment can lead to serious and persistent sequalae. The objective of this study was to construct a cohort LD model to assess the impact of elevated incidence on health-related quality-of-life. The health-related quality-of-life impact of LD in Manitoba over a 10-year period resulted in loss of 13.8 QALYs for the base case (reported incidence) and 73.4 QALYs for the elevated incidence scenario. Our model estimated average direct healthcare costs of $210 for the base case and $1,029 for the elevated incidence scenario, although these costs were taken from an Ontario costing study from reported cases which may underestimate the true cost of treatment of more complicated cases that are untreated during the early clinical stages of LD. Receipt of costing data from Manitoba is expected in June 2024 and will replace the Ontario LD costing estimates. The model was most sensitive to variations in incidence rates and time spent with persistent sequelae, namely arthritis, neurologic sequelae, and Post-treatment Lyme Disease Syndrome (PTLDS). Our model demonstrated that elevated incidence rates that account for under-reporting of LD in Manitoba have a substantial impact on health-related quality-of-life. Conclusion Using two infectious diseases of public health significance, lessons were learned by characterizing their under-ascertainment within the context of access to care, access to testing, and identified equity issues that impact case ascertainment and low disease awareness. The validity of HR-QoL or health economic estimates is questionable in light under-ascertainment. The goal of quantifying under-ascertainment of disease and evaluating the health-related quality-of-life impact provided important insights into burden of disease. Understanding the specific transmission patterns and immune profiles stimulated by exposure were important in designing the studies and interpreting results. Linked administrative datasets are an efficient method of conducting research while preserving patient privacy. Future efforts would benefit from increasing the kind of datasets included in provincial databases, such as inclusion of electronic health records to gather symptom-related data, as well as seroprevalence data from studies conducted by Canada Blood Services or public health. Methods to quantify the level of uncertainty of incidence estimates play an important role in the development and assessment of public health initiatives.