Validating the 8 CPCSSN Case Definitions for Chronic Disease Surveillance in a Primary Care Database of Electronic Health Records

Chronic disease surveillance requires the on-going collection of valid health information from large geographic areas. Primary data collection is time consuming and expensive, thus secondary data has been used for surveillance. Many countries routinely collect health data from hospital discharges, called hospital discharge abstract data. Such data captures sick patients and the majority of patients with chronic diseases are managed or cared at out-patient settings. Thus surveillance relying on hospital discharge abstract data alone hugely underestimates disease burden. In Canada, physicians submit claims for services they provide to be remunerated regardless service locations, including inpatient, outpatient, emergency, daily surgery, homecare, etc. The submitted claims contain information on the reason for visit (i.e. diagnosis and procedure). Using the hospital discharge abstract data and physician claims, Canadian investigators reported hypertension surveillance (1) and are monitoring other chronic diseases.

Unfortunately, the large ICD (International Classification of Disease)-coded outpatient data is not available in many countries. Electronic medical records (EMR) are potential sources for chronic disease surveillance because EMR has been widely implemented in primary care settings. The remaining technical question is how to extract the texted data from EMR and if the extracted data is valid enough to be used for surveillance. Canadian Primary Care Sentinel Surveillance Network (CPCSSN) collects data from 475 primary care practitioners, representing Canadian national primary care patients. The CPCSSN validation study led by Williamson et al. clearly demonstrated that the EMR data is valid for chronic disease surveillance. They validated eight chronic disease extraction algorithms through reviewing original EMRs and reported high statistical estimates. For example, hypertension has sensitivity of 84.9%, specificity of 93.5%, positive predictive value of 92.9% and negative predictive value of 86.0%. These values are very similar with the statistical estimates for hypertension (2), which were defined using hospital discharge abstract data and physician claims (sensitivity 75%, specificity 94%, positive predictive value 81%, negative predictive value 92%). This CPCSSN validation study is fundamental and gives researchers confidence to use the invaluable CPCSS EMR data for surveillance, health service research and epidemiological studies.

References

1. Robitaille C, Dai S, Waters C, Loukine L, Bancej C, Quach S, Ellison J, Campbell N, Tu K, Reimer K, Walker R, Smith M, Blais C, Quan H. Diagnosed hypertension in Canada: incidence, prevalence and associated mortality. CMAJ. 2012 Jan 10;184(1):E49-56.

2. Quan H, Khan N, Hemmelgarn BR, Tu K, Chen G, Campbell N, Hill MD, Ghali WA, McAlister FA; Hypertension Outcome and Surveillance Team of the Canadian Hypertension Education Programs. Validation of a case definition to define hypertension using administrative data. Hypertension. 2009 Dec;54(6):1423-8.

Competing interests: None declared