How well do ICD-9 physician claim diagnostic codes identify confirmed pertussis cases in Alberta, Canada? A Canadian Immunization Research Network (CIRN) Study

Abstract

BackgroundRates of Bordetella pertussis have been increasing in Alberta, Canada despite vaccination programs. Waning immunity from existing acellular component vaccines may be contributing to this. Vaccine effectiveness can be estimated using a variety of data sources including diagnostic codes from physician billing claims, public health records, reportable disease and laboratory databases. We sought to determine if diagnostic codes from billing claims (administrative data) are adequately sensitive and specific to identify pertussis cases among patients who had undergone disease-specific laboratory testing.MethodsData were extracted for 2004–2014 from a public health communicable disease database that contained data on patients under investigation for B. pertussis (both those who had laboratory tests and those who were epidemiologically linked to laboratory-confirmed cases) in Alberta, Canada. These were deterministically linked using a unique lifetime person identifier to the provincial billing claims database, which contains International Classification of Disease version 9 (ICD-9) diagnostic codes for physician visits. We examined visits within 90 days of laboratory testing. ICD-9 codes 033 (whooping cough), 033.0 (Bordetella pertussis), 033.1 (B. parapertussis), 033.8 (whooping cough, other specified organism), and 033.9 (whooping cough, other unspecified organism) in any of the three diagnostic fields for a claim were classified as being pertussis-specific codes. We calculated sensitivity, specificity, positive (PPV) and negative (NPV) predictive values.ResultsWe identified 22,883 unique patients under investigation for B. pertussis. Of these, 22,095 underwent laboratory testing. Among those who had a laboratory test, 2360 tested positive for pertussis. The sensitivity of a pertussis-specific ICD-9 code for identifying a laboratory-confirmed case was 38.6%, specificity was 76.9%, PPV was 16.0%, and NPV was 91.6%.ConclusionICD-9 codes from physician billing claims data have low sensitivity and moderate specificity to identify laboratory-confirmed pertussis among persons tested for pertussis.Electronic supplementary materialThe online version of this article (doi:10.1186/s12913-017-2321-1) contains supplementary material, which is available to authorized users.

Under-reporting of pertussis in Ontario: A Canadian Immunization Research Network (CIRN) study using capture-recapture

“…These explanations are speculative and highlight the need for a validation study of the administrative data to examine what is in the complete clinical record of the cases [ 28 ]. The underlying purpose of OHIP data is physician reimbursement, not disease surveillance, and while the use of OHIP and other un-validated administrative data have potential to be useful in identifying exposure and disease outcomes, significant limitations of such data have been described [ 29 – 31 ] and indicate that caution is advised. OHIP data are likely to include both false positive pertussis cases, which lead to over-estimation in capture-recapture analysis, and false negative cases, which are less of an issue with this method unless they are systematically different in some way from cases identified in other sources.…”

Accuracy of health administrative data to identify cases of reportable travel or migration-related infectious diseases in Ontario, Canada

“…While the utility and validity of health administrative data for chronic disease surveillance has been extensively studied worldwide, the use of these data for infectious diseases is still in its infancy [ 23 , 31 ]. As the potential of these data expand, growing to encompass electronic health records and population-level integrated laboratory databases, characterization of the limitations and strengths of health administrative data for identifying infectious disease cases is needed.…”

Cannabis use disorder and adverse cardiovascular outcomes: A population‐based retrospective cohort analysis of adults from Alberta, Canada

“…For each service, up to three diagnostic codes are assigned using the International Classification of Diseases, 9th revision (ICD‐9). Claims Data facilitates physician remuneration but has been commonly used for health research studies [39–41]. The Alberta Provincial Population Registry [38] contains basic demographic information, including date of death and geographic information on all Alberta residents since 1993. The Pharmaceutical Information Network (PIN) [42] is a pharmacy‐based drug information system implemented in Alberta in 2008, which records all prescriptions filled in pharmacies within the province.…”