Study design
We performed a retrospective chart abstraction study involving patients with first-time rheumatology referrals in Ontario, Canada, which has a publicly-funded single payer healthcare system, where access to rheumatologists is dependent upon referrals. During the study period, there was no policy implementation regarding rheumatology referrals in this setting.
Data sources
We used the Electronic Medical Record Administrative data Linked Database (EMRALD), which is comprised of electronic medical records (EMRs) from PCPs (using the same EMR software system) throughout Ontario [13]. Patient-level EMRs contain all PCP encounters, current and past medical histories, laboratory test results, prescriptions, referral letters, diagnostic tests, as well as information related to care received elsewhere and reported to the practice (such as consultation letters). Encounters with rheumatologists were verified by linking with provincial physician billing claims from the Ontario Health Insurance Plan claims history database. These data sets are linked using unique, encoded patient and physician identifiers, and are securely held and analyzed at ICES (www.ices.on.ca).
Participants
We studied 168 PCPs, of whom 32 practiced in rural locations, 39 in suburban areas, and 97 urban areas. The mean duration of EMR use in our sample was 5 years (range 2–25). Our sample of PCPs was slightly younger with mean (range) age of 47 (28–69) years, in comparison to all Ontario PCPs [with a mean (range) age of 52 (27–79) years]. Our PCP study population also comprised more females (56% vs. 41% for all Ontario PCPs). The mean number of years in practice was 15 for our PCP participants in comparison to 19 years for all Ontario PCPs [14].
Drawing upon primary care EMRs of 268,854 patients with at least 2 years of EMR data, we identified 2430 patients with a first-time referral to a rheumatologist between 2000 and 2013.
Data abstraction
We reviewed each patient’s record to categorize patients by their principal diagnosis or clinical impression associated with the referral. Diagnoses and clinical impressions provided by the rheumatologist where used to categorize patients. Patients were categorized into 6 main rheumatic and musculoskeletal diseases (RMDs): systemic inflammatory rheumatic diseases, osteoarthritis, regional musculoskeletal (MSK) syndromes, chronic pain, osteoporosis, and miscellaneous referrals. Systemic inflammatory rheumatic diseases were further defined. Patients where categorized according to the most serious complaint when they carried multiple RMD diagnoses.
Based on review of prior studies, we performed standardized data abstraction to ascertain the quality and completeness of referral letters, such as providing reasons for referral, relevant medical and family histories, and diagnostics tests. We also abstracted details of symptoms provided on the referral letter according to whether the symptoms were present, absent, or not reported. We also determined whether patients had diagnostic imaging performed within the 3 months preceding referral. Rheumatology consultation letters following the referral were reviewed to abstract details on whether rheumatologists were providing information related to diagnoses and general management plans.
Three trained medical abstractors performed the chart abstraction. We performed double data abstraction on an initial 10% sample of charts, whereby the data for each patient were abstracted a second time by the same abstractor and once by a different abstractor. To ensure good agreement, we required κ scores for inter- and intra-rater reliability to exceed 0.85 before commencing full data abstraction. For all patients, an independent abstractor (J.W.) also performed double data abstraction related to assigning patients to diagnostic categories.
Statistical analysis
Descriptive statistics were used to describe patients and the contents of the letters, stratified according to diagnostic category. We assessed the frequency of general details provided on referral letters (patient history and laboratory results), details of symptoms provided on the referral letter, actual diagnostic imaging performed on the patient in contrast to what was reported on the referral letter, and details and timeliness of consultation letters.
Analyses were performed on coded data using SAS, version 9.2, and Microsoft SQL Server 2012.