Mark Weiner1, Judith Long. 1. Center for Health Equity and Research and Promotion, Philadelphia VA Medical Center, PA 19104, USA. mweiner@mail.med.upenn.edu
Abstract
BACKGROUND: Performance assessments help to quantify the level of adherence with practice standards and are often used to measure and compare the quality of care. However, most performance assessments are based on a cross-sectional analysis of patient information, whereas patient care is inherently longitudinal. This discordance could confound the relationship between the performance measure and the true quality of care. OBJECTIVE: The objective of this study was to illustrate differences in performance assessment as measured by a traditional cross-sectional analysis compared with a longitudinal analysis. METHODS: We conducted a cross-sectional and longitudinal analysis of a cohort of diabetic patients in an integrated delivery system having primary care visits and hemoglobin A1c (HBA1c) testing in both 1999 and 2000. RESULTS: In the cross-sectional analysis of 4661 patients, we found a modestly increasing proportion achieved an HBA1c level of <8.0%: 73.1% in 1999 and 75.6% in 2000. Longitudinal analysis, however, suggested that certain subsets of patients were more likely to switch from good to poor control or retain their level of poor control over the 2 years studied. In particular, compared with whites, blacks were 1.76 (95% confidence interval [CI], 1.31-2.37) times as likely to switch from good to poor control and only 0.56 (95% CI, 0.41-0.76) times as likely to switch from poor to good control. Patients aged 35 to 49 were 2.54 (95% CI, 1.79-3.45) times as likely to switch from good to poor and only 0.66 (95% CI, 0.47-0.94) times as likely to switch from poor to good control than patients over age 64 years. CONCLUSIONS: Cross-sectional performance assessments could mask changes in diabetes control among individuals belonging to a cohort and, conceptually, are poorer indicators of care process than longitudinal measures. In addition, longitudinal analyses suggest the influence of patient sociodemographic factors on the performance assessment that should be accounted for when comparing quality of care for diabetes.
BACKGROUND: Performance assessments help to quantify the level of adherence with practice standards and are often used to measure and compare the quality of care. However, most performance assessments are based on a cross-sectional analysis of patient information, whereas patient care is inherently longitudinal. This discordance could confound the relationship between the performance measure and the true quality of care. OBJECTIVE: The objective of this study was to illustrate differences in performance assessment as measured by a traditional cross-sectional analysis compared with a longitudinal analysis. METHODS: We conducted a cross-sectional and longitudinal analysis of a cohort of diabeticpatients in an integrated delivery system having primary care visits and hemoglobin A1c (HBA1c) testing in both 1999 and 2000. RESULTS: In the cross-sectional analysis of 4661 patients, we found a modestly increasing proportion achieved an HBA1c level of <8.0%: 73.1% in 1999 and 75.6% in 2000. Longitudinal analysis, however, suggested that certain subsets of patients were more likely to switch from good to poor control or retain their level of poor control over the 2 years studied. In particular, compared with whites, blacks were 1.76 (95% confidence interval [CI], 1.31-2.37) times as likely to switch from good to poor control and only 0.56 (95% CI, 0.41-0.76) times as likely to switch from poor to good control. Patients aged 35 to 49 were 2.54 (95% CI, 1.79-3.45) times as likely to switch from good to poor and only 0.66 (95% CI, 0.47-0.94) times as likely to switch from poor to good control than patients over age 64 years. CONCLUSIONS: Cross-sectional performance assessments could mask changes in diabetes control among individuals belonging to a cohort and, conceptually, are poorer indicators of care process than longitudinal measures. In addition, longitudinal analyses suggest the influence of patient sociodemographic factors on the performance assessment that should be accounted for when comparing quality of care for diabetes.