Evaluating, assessing and understanding community presentation and disease burden of influenza and influenza-like illness in primary care settings in Alberta
Ian Johnston, Stephanie Garies, Fatemeh Sabet, Boglarka Soos, Jim Dickinson, Neil Drummond
Stephanie Garies at firstname.lastname@example.org
Alberta Health Surveillance and Assessment Branch
Study Completed. Final Report available here.
Influenza and similar illnesses represent a significant burden of disease in Alberta and across Canada as evidenced by morbidity and mortality data from hospitalized patients. However, the true incidence of influenza affecting communities is difficult to estimate, as most patients seeking medical care are wholly managed by their primary care provider and do not attend hospital.
To assess whether it would be feasible to conduct influenza surveillance in community-based, primary care settings, we aimed to develop a case definition for influenza and influenza-like illness. The case definition was validated using flu swab data from TARRANT Viral Watch linked to patient electronic medical record (EMR) information extracted by the Canadian Primary Care Sentinel Surveillance Network (CPCSSN) in southern Alberta. If effective, the definition may be used to further describe the epidemiology of influenza in Alberta and possibly across Canada.
Primary care providers who contribute both to TARRANT and CPCSSN were invited to participate. Flu swab results from 7 contiguous flu seasons (2008/9 to 2014/15) were matched with patient EMR data from CPCSSN. A number of case definitions were created for influenza and influenza-like-illness in the CPCSSN database using various text words and ICD-9 codes in the encounter diagnosis / visit and/or billing table and/or a prescription for a neuraminidase inhibitor. The definition was then applied to the CPCSSN database to ascertain possible cases of flu or ILI and validated using the matched TARRANT viral swabs.
A total of 11 out of 28 providers from 3 clinics agreed to participate. The case definition identified 11,491 suspected influenza encounters from a sample of 10,322 patient records over the 7-year study period. Approximately 68.6% of TARRANT viral swabs matched patient records in the CPCSSN database. Overall, the case definition had a sensitivity of 29% and specificity of 78%, with a positive predictive value of 33% and negative predictive value of 74%, but there was much variation between clinic sites.
The poor performance of the case definition in its current state prohibited its use as a means for influenza surveillance within the CPCSSN database. We identified several contributing factors that limited the usefulness of the case definition for epidemiological purposes, such as missing diagnoses or symptomatic entries in the EMR; the use of viral flu swabs for secondary research purposes; clinic submission of the viral swabs as a group rather than at an individual provider level; viral swabs available for patients not in the CPCSSN database.