Steven R Erickson1, Yuan-Nung Tony Lin. 1. University of Michigan College of Pharmacy, 428 Church St., Ann Arbor, MI 48109-1065. serick@med.umich.edu.
Abstract
BACKGROUND: Research has demonstrated that variation in availability and utilization of health care resources exist on a range of scales, from regions of the United States, hospital referral regions, ZIP codes, and census tracts. Limited research using spatial analyses has found that variation in medication adherence exists across census tracts. Using spatial analysis, researchers may be able to effectively analyze geographically dispersed data to determine whether factors such as sociodemographics, local shared beliefs and attitudes, barriers to access such as availability of prescribers or pharmacies, or others are associated with variations in medication adherence in a defined geographic area. OBJECTIVES: To (a) demonstrate that medication adherence may be mapped across an entire state using medication possession ratios and (b) determine whether a geographic pattern of adherence to statins could be identified at the ZIP code level for members of a statewide insurer. METHODS: This study utilized pharmacy claims data from a statewide insurer. Insured statin users were aged greater than 30 years, had at least 1 statin prescription, and were continuously enrolled for the observation year. Patient medication possession ratios (MPR) were derived, which were then aggregated as a mean MPR for each ZIP code. ZIP codes were categorized as higher (MPR greater than 0.80) or lower (MPR less than 0.80) adherence and mapped using Arc GIS, a platform for designing and managing solutions through the application of geographic knowledge. Analysis included a determination of whether the MPRs of higher and lower adherence ZIP codes were significantly different. Hot spot analysis was conducted to identify clustering of higher, midrange, and lower adherent ZIP codes using the GetisORD Gi* Statistic. This test provides z-scores and P values to indicate where features with either high or low values cluster spatially. MPRs for these 3 categories were compared using one-way analysis of variance (ANOVA). RESULTS: Of 1,154 Michigan ZIP codes, 907 were represented by 212,783 insured statin users. The mean statin MPR by ZIP code was 0.79 ± 0.4. The mean MPR for higher adherent ZIP codes was 0.83 ± 0.03 and 0.76 ± 0.03 for lower adherent ZIP codes (P less than 0.001). Significant clustering of ZIP codes by adherence levels was evident from the hot spot analysis. The mean MPR was 0.84 ± 0.04 for high adherence areas, 0.79 ± 0.03 for midrange areas, and 0.74 ± 0.04 for lower adherent areas (overall P less than 0.001). CONCLUSIONS: Significant variations in adherence exist across ZIP codes at a state level. Future research is needed to determine locally relevant factors associated with this finding, which may be used to derive locally meaningful interventions.
BACKGROUND: Research has demonstrated that variation in availability and utilization of health care resources exist on a range of scales, from regions of the United States, hospital referral regions, ZIP codes, and census tracts. Limited research using spatial analyses has found that variation in medication adherence exists across census tracts. Using spatial analysis, researchers may be able to effectively analyze geographically dispersed data to determine whether factors such as sociodemographics, local shared beliefs and attitudes, barriers to access such as availability of prescribers or pharmacies, or others are associated with variations in medication adherence in a defined geographic area. OBJECTIVES: To (a) demonstrate that medication adherence may be mapped across an entire state using medication possession ratios and (b) determine whether a geographic pattern of adherence to statins could be identified at the ZIP code level for members of a statewide insurer. METHODS: This study utilized pharmacy claims data from a statewide insurer. Insured statin users were aged greater than 30 years, had at least 1 statin prescription, and were continuously enrolled for the observation year. Patient medication possession ratios (MPR) were derived, which were then aggregated as a mean MPR for each ZIP code. ZIP codes were categorized as higher (MPR greater than 0.80) or lower (MPR less than 0.80) adherence and mapped using Arc GIS, a platform for designing and managing solutions through the application of geographic knowledge. Analysis included a determination of whether the MPRs of higher and lower adherence ZIP codes were significantly different. Hot spot analysis was conducted to identify clustering of higher, midrange, and lower adherent ZIP codes using the GetisORD Gi* Statistic. This test provides z-scores and P values to indicate where features with either high or low values cluster spatially. MPRs for these 3 categories were compared using one-way analysis of variance (ANOVA). RESULTS: Of 1,154 Michigan ZIP codes, 907 were represented by 212,783 insured statin users. The mean statin MPR by ZIP code was 0.79 ± 0.4. The mean MPR for higher adherent ZIP codes was 0.83 ± 0.03 and 0.76 ± 0.03 for lower adherent ZIP codes (P less than 0.001). Significant clustering of ZIP codes by adherence levels was evident from the hot spot analysis. The mean MPR was 0.84 ± 0.04 for high adherence areas, 0.79 ± 0.03 for midrange areas, and 0.74 ± 0.04 for lower adherent areas (overall P less than 0.001). CONCLUSIONS: Significant variations in adherence exist across ZIP codes at a state level. Future research is needed to determine locally relevant factors associated with this finding, which may be used to derive locally meaningful interventions.
Authors: Jennifer A Rymer; Katherine E Mues; Keri L Monda; Emily W Bratton; Heidi S Wirtz; Ted Okerson; Robert A Overman; M Alan Brookhart; Paul Muntner; Tracy Y Wang Journal: J Am Heart Assoc Date: 2020-04-24 Impact factor: 5.501