Ajay K. Israni

Missing data are often imputed for analysis, but a single imputation may be inaccurate when performing feature selection in mining data. Feature selection procedures applied to multiply imputed data demonstrate this phenomenon and suggest that multiple imputation is an important adjunct to knowledge discovery.

Background: Three million people in the United States are estimated to have chronic kidney disease (CKD). Management of these CKD patients in the outpatient primary care clinic setting has not been well studied. Hypothesis: Primary care management of CKD can be assessed and opportunities for improvement can be identified. Methods: Management of CKD based on available published literature and guidelines was assessed in a single primary care site of an academic hospital with 23,000 annual visits and 8,300 patients. Charts of patients seen between October 1, 1997 and March 25, 1999 with an elevated SCr ≧1.7 mg/dl on two separate measurements at least 6 months apart were reviewed for predefined indicators of CKD management. Results: Assessment identified several aspects of CKD management to be suboptimal: control of blood pressure, use of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, assessment of proteinuria, and renal consultation. Better management was found with respect to hemoglobin A1c measurement for diabetic patients. In general, CKD care was similar for diabetic and non-diabetic patients. CKD management was also similar regardless of level of creatinine clearance (≧50 vs. 50–30 vs. ≤30 ml/min). Conclusion: CKD care can be measured in an outpatient academic primary care clinic and opportunities to improve were identified.

With the growing acceptance of hypertension programs and the recent subspecialty certification in hypertension, information on the successful blood pressure control in patients referred to hypertension programs is critical. We examined the records of a single practitioner based hypertension program (RRT) at a university hospital outpatient hypertension practice in whom subjects were seen on at least two occassions between 1995 and 2000. Using a predetermined survey tool a random sample of 81 subjects were included for this analysis. The subjects were 58±14 (Mean±SD) years old, 51% were female, 52% were white, 33% AA, 4% hispanic and 10% were diabetic. Seated blood pressure in the dominant arm at the first visit (average of two readings) was 170±33/81±17 mm Hg. At the final visit 50 of 81 patients (62%) achieved a systolic blood pressure <140 mm Hg. The most significant demographic factor in those who did not achieve the systolic goal was age. The mean age was 64±11 in those did not achieve goal systolic blood pressure vs 55±15 in those who did achieve goal systolic blood pressure; p = 0.004).