Lead/Presenter: Liberty (Alison) Greene,
COIN - Palo Alto
All Authors: Greene AL (VA Palo Alto Center for Innovation to Implementation, Division of Primary Care and Population Health, Stanford University School of Medicine), Zulman,DM (VA Palo Alto Center for Innovation to Implementation, Division of Primary Care and Population Health, Stanford University School of Medicine), Boothroyd,D (Quantitative Sciences Unit, Biomedical Informatics Research Division Stanford University School of Medicine, VA Palo Alto Center for Innovation to Implementation) Slightam,CA (VA Palo Alto Center for Innovation to Implementation) Gregory,AJ (VA Palo Alto Center for Innovation to Implementation)
This study seeks to evaluate several distinct measures of care fragmentation in high-risk Veterans. We examine their prevalence and variation by patient sociodemographic and clinical characteristics, and assess predictive validity based on their association with acute hospitalizations.
We identified VA patients with at least 2 VA outpatient visits in FY14 and a Care Assessment Needs score for 1-year hospitalization > = 90th percentile (n = 527,089). Using VA, VA-purchased care and Medicare data, we constructed five fragmentation measures: provider count, concentration of care with an empirically defined "usual provider" (UPC), dispersion of care across providers using Bice-Boxerman Continuity of Care Index (COCI) and an adaption of this measure (MMCI). Among patients 65+ with at least one Medicare encounter (n = 42,564), we also constructed a measure of health system fragmentation based on their distribution of visits across VA and Medicare. We examined the distribution of each measure and variation by patient sociodemographic and clinical characteristics. We then used multivariable logistic regression to examine the relationship between fragmentation in FY14 and hospitalization in FY15.
Just over half of patients (52%) were 65+, 30% lived in rural areas, 80% had 5 or more chronic conditions, and 60% had a mental health condition. The median number of outpatient visits was 9 with an interquartile range (IQR) of 6-14. Medians and IQRs of the measures were: provider count 5 (3-8), UPC 0.38 (0.27 - 0.5) , COCI 0.14 (0.07-0.27) , MMCI: 0.50 (0.32 - 0.63). Median health system fragmentation for VA-Medicare dual users was 0.50 (0.26 - 0.73). Patients with mental health utilization and those with more chronic conditions had more fragmented care across all measures; there was minimal variation across other patient characteristics. In multivariable logistic regression models for each fragmentation measure, greater fragmentation was associated with a higher likelihood of hospitalization (p < 0.001).
This study presents five unique measures of care fragmentation in high risk veterans, and illustrates their relationships with future hospitalization.
Measures of care fragmentation may be used to identify and support high-risk Veterans' care coordination needs. This study provides valuable information about fragmentation measures that may inform studies of utilization patterns among high-risk patients.