Sarah L Reeves1, Kevin J Dombkowski2, Brian Madden2, Lindsay Cogan3, Shanshan Liu4, Paul B Kirby5, Sara L Toomey6. 1. Department of Pediatrics, Susan B Meister Child Health Evaluation and Research (CHEAR) Center, University of Michigan (SL Reeves, KJ Dombkowski, and B Madden), Ann Arbor, Mich; Department of Epidemiology, University of Michigan (SL Reeves), Ann Arbor, Mich. Electronic address: sleasure@umich.edu. 2. Department of Pediatrics, Susan B Meister Child Health Evaluation and Research (CHEAR) Center, University of Michigan (SL Reeves, KJ Dombkowski, and B Madden), Ann Arbor, Mich. 3. New York State Department of Health, Office of Quality and Patient Safety (L Cogan), Albany, NY; Department of Health Policy Management & Behavior, School of Public Health, University at Albany (L Cogan), Albany, NY. 4. Boston Children's Hospital (S Liu and SL Toomey), Boston, Mass. 5. Commonwealth Medicine Center for Health Policy and Research, University of Massachusetts Medical School (PB Kirby), Quincy, Mass. 6. Boston Children's Hospital (S Liu and SL Toomey), Boston, Mass; Harvard Medical School (SL Toomey), Boston, Mass.
Abstract
BACKGROUND: Measuring quality at varying levels of the health care system requires attribution, a process of determining the patients and services for which each level is responsible. However, it is important to ensure that attribution approaches are equitable; otherwise, individuals may be assigned differentially based upon social determinants of health. METHODS: First, we used Medicaid claims (2010-2018) from Michigan to assess the proportion of children with sickle cell anemia who had less than 12 months enrollment within a single Medicaid health plan and could therefore not be attributed to a specific health plan. Second, we used the Medicaid Analytic eXtract data (2008-2009) from 26 states to simulate adapting the 30-Day Pediatric All-Condition Readmission measure to the Accountable Care Organization (ACO) level and examined the proportion of readmissions that could not be attributed. RESULTS: For the sickle cell measure, an average of 300 children with sickle cell anemia were enrolled in Michigan Medicaid each year. The proportion of children that could not be attributed to a Medicaid health plan ranged from 12.2% to 89.0% across years. For the readmissions measure, of the 1,051,365 index admissions, 22% were excluded in the ACO-level analysis because of being unable to attribute the patient to a health plan for the 30 days post discharge. CONCLUSIONS: When applying attribution models, it is essential to consider the potential to induce health disparities. Differential attribution may have unintentional consequences that deepen health disparities, particularly when considering incentive programs for health plans to improve the quality of care.
BACKGROUND: Measuring quality at varying levels of the health care system requires attribution, a process of determining the patients and services for which each level is responsible. However, it is important to ensure that attribution approaches are equitable; otherwise, individuals may be assigned differentially based upon social determinants of health. METHODS: First, we used Medicaid claims (2010-2018) from Michigan to assess the proportion of children with sickle cell anemia who had less than 12 months enrollment within a single Medicaid health plan and could therefore not be attributed to a specific health plan. Second, we used the Medicaid Analytic eXtract data (2008-2009) from 26 states to simulate adapting the 30-Day Pediatric All-Condition Readmission measure to the Accountable Care Organization (ACO) level and examined the proportion of readmissions that could not be attributed. RESULTS: For the sickle cell measure, an average of 300 children with sickle cell anemia were enrolled in Michigan Medicaid each year. The proportion of children that could not be attributed to a Medicaid health plan ranged from 12.2% to 89.0% across years. For the readmissions measure, of the 1,051,365 index admissions, 22% were excluded in the ACO-level analysis because of being unable to attribute the patient to a health plan for the 30 days post discharge. CONCLUSIONS: When applying attribution models, it is essential to consider the potential to induce health disparities. Differential attribution may have unintentional consequences that deepen health disparities, particularly when considering incentive programs for health plans to improve the quality of care.
Authors: Peter J Huckfeldt; Chris Chan; Samuel Hirshman; Aaron Kofner; Jodi L Liu; Andrew W Mulcahy; Ioana Popescu; Clare Stevens; Justin W Timbie; Peter S Hussey Journal: Rand Health Q Date: 2015-07-15
Authors: Amanda B Payne; Ruth Link-Gelles; Ijeoma Azonobi; W Craig Hooper; Bernard W Beall; James H Jorgensen; Billie Juni; Matthew Moore Journal: Pediatr Infect Dis J Date: 2013-12 Impact factor: 2.129
Authors: Jean L Raphael; Craig L Dietrich; Deborah Whitmire; Donald H Mahoney; Brigitta U Mueller; Angelo P Giardino Journal: Pediatr Blood Cancer Date: 2009-02 Impact factor: 3.167
Authors: James P Guevara; Jeanhee Moon; Elizabeth M Hines; Ettya Fremont; Angie Wong; Christopher B Forrest; Jeffrey H Silber; Susmita Pati Journal: Med Care Res Rev Date: 2013-11-13 Impact factor: 3.929
Authors: Leigh R Warren; Jonathan M Clarke; Sonal Arora; Mauricio Barahona; Naila Arebi; Ara Darzi Journal: World J Gastroenterol Date: 2019-05-07 Impact factor: 5.742