Setting and design
This study was conducted at two VA Medical Center primary care clinics in 2006. Of the 15 clinicians practicing at the clinics, 13 chose to participate. We observed all consecutive patient visits made to each clinician over the observation period (two to nine weeks during May to August 2006). Patient visits occurred on six to sixteen half-day sessions of clinic and included 37 to 151 patient visits, depending on the clinician.
Data sources
At each visit, research assistants recorded the patient's BP as assessed by the nursing assistant or nurse, and the value and date of the most recent hemoglobin A1c and low density lipoprotein cholesterol (LDL-c) levels as recorded in the medical record. A data collection tool including this information was provided to the clinician prior to seeing each patient. At the conclusion of the visit, the clinicians recorded whether BP medications were intensified. The BP, diabetes, and lipid medications at the time of the visit were also recorded. The Birmingham VA Medical Center's Institutional Review Board approved the study protocol.
Study sample
We observed 946 consecutive patients and include in this analysis the 387 patients who presented with uncontrolled BP. Patients who were seen more than once only had their first visit analyzed. Uncontrolled BP was defined as ≥ 140/90 mmHg, or ≥ 130/80 if the patient had diabetes. Diabetes was defined according to Miller's definition: more than one ICD-9-CM code for diabetes (250.xx) within two years prior to the observed visit, or a prescription for any diabetes medication in the current year [12]. Hyperlipidemia was defined as being on a lipid-lowering medication within two years prior to the observed visit, or an LDL-c ≥ 130 mg/dL at the visit, or ≥ 100 mg/dL if the patient had diabetes.
Dependent variable: BP medications intensified
We defined BP medication as intensified if the clinician indicated, on the data collection form, addition of a new medication or increase in the dose of an existing medication. Clinicians indicated whether they intensified medications in all but 17 patients with uncontrolled high BP. The 17 patients without indication of intensification were classified as not intensified.
Main exposure: uncontrolled CVD risk factors
The main exposures of interest were uncontrolled diabetes and uncontrolled hyperlipidemia. Uncontrolled diabetes was defined as most recent recorded A1c ≥ 7%. Uncontrolled hyperlipidemia was defined as most recent recorded LDL-c ≥ 130 mg/dL, or ≥ 100 mg/dL if the patient had diabetes.
Because patients were consecutive primary care patients, not all had diabetes or hyperlipidemia. Therefore, for each patient we indicated which of the following four mutually exclusive clinical scenarios was present: uncontrolled hypertension and no diagnosed diabetes or hyperlipidemia; controlled diabetes and/or hyperlipidemia; either uncontrolled diabetes or uncontrolled hyperlipidemia; or both uncontrolled diabetes and uncontrolled hyperlipidemia. Because our primary interest was the effect of multiple uncontrolled conditions, patients with known diabetes or hyperlipidemia but no A1c or LDL-C in the record were categorized as having the respective controlled condition. While other clinical scenarios are possible, they occurred too rarely in our sample for separate categorization.
Other independent variables: patient characteristics
Additional variables included age, gender, systolic and diastolic BP values, and the total number of BP, diabetes, and lipid medications prescribed in the 90 days prior to the observed visit. Because so many patients who met the uncontrolled BP criterion were very near the threshold, we also constructed a variable reflecting mmHg of systolic BP above goal.
Analysis
Patient age, gender, total number of medications for hypertension, diabetes, hyperlipidemia, BP level, and mmHg above goal were examined across the four clinical scenarios using Chi-square tests, analysis of variance, and Kruskal-Wallis tests where appropriate. In addition, we examined the proportions of patients whose BP medications were intensified for each patient characteristic, for each risk factor separately, and for combinations of elevated risk factors. Hierarchical logistic regression models were constructed to identify independent associations with BP medication intensification, adjusting for clustering of patients within clinician and for patient factors. The main exposure of interest was a categorical variable representing the four clinical scenarios defined above. The referent category was the scenario where the patient had uncontrolled BP and controlled diabetes and/or hyperlipidemia. All variables were retained in the multivariable model except patient gender, due to small numbers of women. We estimated adjusted odds ratios (OR) and predicted probabilities, each with 95% confidence intervals (CI), from these models. The predicted probabilities of BP medication intensification for each of the four clinical scenarios were calculated from these models using the sample mean age, systolic BP, and sample median number of total medications for hypertension, diabetes, and hyperlipidemia.
Preliminary analyses indicated that intensification rates were strongly associated with systolic BP level, with lower rates for those with lower systolic BP. To examine the robustness of the effect of multiple uncontrolled conditions at even low elevations of systolic BP, we stratified the sample on the median mmHg above the goal, which was 10 mmHg. Therefore, the near-goal sample included individuals with systolic BP of 130 to 139 mmHg if they had diabetes, and 140 to 149 mmHg if they did not. We also examined the effect of restricting the analysis to only patients known to have hyperlipidemia and/or diabetes, again stratifying on the median mmHg above goal (within 10 mmHg). Finally, we conducted a sensitivity analysis that excluded patients without A1c or LDL-c assessed from the analysis, rather than grouping them as controlled.