A Risk Based Approach to Improving Chronic Kidney Disease Management

Specific Aim 1 We will develop software to calculate the estimated glomerular filtration rates for all patients presenting to primary care physicians randomized to the intervention arm, and identify patients with an estimated glomerular filtration rates in the range of 30 to 59. We will create three electronic alerts that intervention clinicians will receive upon accessing the patient chart based on whether the patient is high risk or low risk. These alerts will focus on recommending overdue laboratory tests (urine protein, blood cholesterol, etc), as well as recommending guideline appropriate medications (ACE inhibitors), and nephrology referral when appropriate.

We will provide self management support materials to patients of primary care physicians randomized to the intervention arm. We will rely on primary care physicians to enroll patients by first recommending referral via the electronic alerts. On a monthly basis, we will identify patient visits during which an alert fired and no referral was placed. We will distribute a list of these patients to each physician via inter-office mail at least every other month. The mailing will ask physicians to return the list indicating which patients should be enrolled in the program, and our project manager will place the referrals. For non-responding physicians, we will follow up with a reminder email. The patient mailings will include recent clinical results and guideline-recommended targets, encouraging patients to become more proactive in the management of their kidney disease. Once a patient is enrolled in the program, they will receive similar mailings with updated personalized data and recommendations every 3 months.

Electronic referrals placed by primary care physicians for management of chronic kidney disease will first be routed to the renal nurse, who will then initiate contact with the patient. A total of two telephone calls followed by a letter will be made to contact the patient. The nephrology visits will occur per standard clinical operations, including evaluation by an attending nephrologist, as well as educational sessions with the renal nurse and nutritionist. We will create new template notes within the electronic record for use by the nephrologists to communicate clinical care recommendations back to the primary care physicians.

Prior to starting the intervention, the study team will travel to each of the 14 health centers to conduct orientation sessions with the primary care physicians. These sessions will provide general information regarding the goals and scope of the upcoming intervention, including demonstrations of the electronic alerts and the self management support outreach program. A similar overview will also be provided to the HVMA Division of Nephrology.

We will randomize approximately 170 physicians into the intervention and control groups. Physicians in the intervention group will receive patient-specific alerts at the time of office visits for patients with Stage 3 kidney disease. Physicians in the control group will not receive active alerts.

Data will all be obtained electronically from automated extracts from the electronic health record. Our study endpoints will be measured at 18 months and are specified according to risk status. The primary endpoints among high risk patients will be 1) the presence of an office visit in nephrology within the prior 12 months, and 2) the use of ACE inhibitors or ARBs for those with hypertension or microalbuminuria. The primary endpoints among low risk patients will include 1) a urine microalbumin result within the prior 12 months, and 2) the use of ACE inhibitors or ARBs for those with hypertension or microalbuminuria. Hypertension will be assessed based on the presence of a most recent blood pressure greater than 130/80 mmHg or a current diagnosis of hypertension on the electronic problem list. A secondary endpoint for both patient groups will be achieving a blood pressure less than 130/80 mmHg. We will also assess primary care physician awareness of chronic kidney disease defined as use of appropriate problem list and encounter diagnosis codes; and rates of annual serum LDL cholesterol, hemoglobin, phosphorous, 25-OH-vitamin D, calcium, and parathyroid hormone testing; as well as rates of LDL cholesterol control (<100 mg/ dL) and anemia management (hemoglobin > 11 g/dL).

We will conduct an assessment of the self management support materials by surveying patients directly to assess 1) their ratings of the delivery of self-management support by our physician practice, 2) awareness of chronic kidney disease and treatment goals, and 3) the utility of the information contained in the patient mailings. These surveys will be conducted at two time points: baseline (first quarterly mailing) and completion of the study (final quarterly mailing). We will conduct the survey at two time points to facilitate an analysis of trends in patient experiences of care.

To assess the impact of the intervention on each of our primary outcomes we will fit hierarchical logistic regression models with random effects for patients within physicians and physicians within centers. We will fit a set of similar models among three subgroups defined by number of patient visits with their primary care physician during the 18 month period (0 visits, 1-2 visits, ≥ 3 visits). We will fit two secondary models that include a third independent variable for race or sex. The effect of the intervention on reducing race or sex-based disparities will be assessed by examining race*group and sex*group interaction terms.

Specific Aim 2 We will use the physician survey to collect data on primary care physician support for electronic reminders and patient self management, and preparedness to manage kidney disease. Physician responses will be collected using 5 point and 4 point Likert scales, and we will examine the distribution of responses for each survey item to create three dichotomous predictor variables of 1) high support for electronic reminders, 2) high support for patient self management, and 3) high preparedness for managing kidney disease. The primary outcomes will be the proportion of electronic alerts accompanied by ordering of 1) the recommended treatment, or 2) referral for patient self management support. We will construct three separate linear regression models for each dichotomous predictor variable defined above, with the proportion of times that the appropriate decision support feature is used as a continuous outcome variable. The dichotomous physician survey outcome measures will be used as the primary independent variables. We will further assess the statistical significance of an interaction term between physician randomization status and each of the three dichotomous survey outcomes in our primary models from Specific Aim 2 to test whether electronic alerts are more or less effective depending on physician reported attitudes.