Recording blood pressure readings in elderly patients’ charts

Selection of practices

Lists of family physicians in the Ottawa (n = 592) and Hamilton (n = 335) regions of Ontario were generated, and 242 family physicians (200 from Ottawa and 42 from Hamilton) were randomly selected and invited to come for an assessment of their eligibility for participation in the Community Hypertension Assessment Trial (CHAT).12 The CHAT was a paired cluster randomized controlled trial evaluating the effectiveness of an invitational cardiovascular health promotion program held in local pharmacies and staffed by specially trained peer health educators. The program emphasized BP control and management. Of the physicians selected for assessment, 25% (60/242) met the initial eligibility criteria. Eligible physicians were those who had non-academic, full-time, regular family practices in terms of patient population and case mix and who were able to generate a roster of regular patients aged 65 and older. Only 1 physician per group practice was eligible to participate in the study. In total, 47% (28/60) of eligible family physicians agreed to participate. The primary reasons for not participating were being too busy and having no interest in the project.

Selection of patients

In the 28 participating practices, patients who were community-dwelling, 65 years old and older, and regular patients of the family physician and who had visited the office at least once during the previous 12 months were eligible. The health records of 55 eligible patients from each practice were randomly selected for review using random number generator software.

Data collection and quality assurance

From May 2003 to January 2004, 7 trained abstractors (4 in Ottawa and 3 in Hamilton) reviewed the records using a pretested standardized clinical data extraction form. To train the abstractors, 59 health records from practices not participating in the CHAT were reviewed by at least 2 abstractors. During this training exercise, interobserver reliability was very good (κ scores ranged from 0.66 to 1.00, with an average of 0.88) for all variables except smoking status and family history of cardiovascular disease (κ scores of 0.47 and 0.55, respectively).

Patients were considered to be diagnosed with hypertension if any of the following notations was found in their health records: diagnosis of “hypertension,” or “high blood pressure,” or “high BP” in the context of a diagnosis rather than based on 1 elevated reading or ambiguous notations such as “query,” “probable” or “probably,” and “monitor.”

Patients’ demographic characteristics, cardiovascular risk factors, current antihypertensive medication profiles, number of visits to family physicians during the 12-month review period, and number of BP recordings were obtained from the health records. If no BP was recorded during the 12-month review period, the date of last BP measurement was noted as far back as 24 months before the date of the review. All participating physicians completed a baseline questionnaire on their sociodemographic and practice characteristics.

The study was reviewed and approved by the research ethics boards of the SCO Health Service and Hamilton Health Sciences and the Faculty of Health Sciences at McMaster University in Hamilton.

Data analysis

Blood pressure recording was defined as a reading at an office visit recorded in patients’ charts during the previous 12 months. The associations between physician characteristics (sex, years since graduation from medical school, practice type, and method of payment) and BP recording were examined using univariate logistic regression analysis. Years since graduation from medical school were calculated by subtracting age at graduation from age at audit; the median number of years since graduation, 24, was used to dichotomize the variable. Three types of family practices were identified: solo, group sharing office space only, and group sharing a patient roster. Two types of payment methods were identified: fee-for-service and capitation, which included payment through health service organizations and primary care networks.

The associations between patient characteristics (age, sex, smoking status, family history of cardiovascular disease, diagnosis of hypertension or taking antihypertensive medication, diabetes, heart disease including peripheral vascular disease and aortic aneurysm, stroke or transient ischemic attack, retinopathy, nephropathy, number of visits to a family physician during the past 12 months) and BP recording were examined using univariate logistic regression analysis. To avoid multicolinearity, a composite variable including patients with diagnosed hypertension or patients taking antihypertensive medications was created. A composite variable including heart disease, peripheral vascular disease, and aortic aneurysm was also constructed.

Patient and physician characteristics significant at P < .1 in univariate logistic regression analysis were retained for a full multivariate logistic regression model to examine their association with BP recording. A reduced multivariate model included only those variables that were statistically significant at P < .05 in the full multivariate model. Odds ratios (OR), 95% confidence intervals (CI), and 2-tailed P values were calculated for each variable in the univariate and multivariate logistic regression models. Analyses were conducted using SAS 9.1.3. All univariate and multivariate analyses accounted for clustering of patients within practices using the PROC GENMOD program.