Center for Improving Value in Health Care
May 8, 2014 | 0 comments | Posted by
Data, Information, All Payer Claims Database
(The third installment in an ongoing series dedicated to helping Coloradans
better understand health care claims data available on www.cohealthdata.org)
Previous articles in this series attempted to explain risk adjustment and burden of illness concepts and demonstrate how these concepts apply to Total Cost of Care Compared to Expected (C2E) reports available on the Colorado APCD website. This third installment explains how risk adjustment and burden of illness concepts apply to Utilization C2E reports and provides examples of how this information can be used to better understand potential reasons for observed variation in the use of health care services in Colorado.
Before we dive into analysis of the Utilization C2E reports, it will be helpful to briefly review the risk adjustment and burden of illness score assignment processes:
Let’s also take a moment to briefly summarize some key features of expected values:
Utilization reports available on www.cohealthdata.org show the amount of health care services used per thousand insured people on an annual basis. The actual or observed value is generated by dividing the total number of medical services delivered in a particular category by the size of the relevant population group expressed in thousands of people. Utilization is reported statewide and by county and three-digit Zip Code geographic groupings, and by age range and gender. Per thousand population rates are a standard measure of medical services utilization and are widely reported. Although utilization rates per thousand are a high-level measure, they are helpful in making meaningful comparisons across regions and relative to statewide or national averages.
In the Rate per Thousand columns of the table above, we can see the actual or observed utilization rates for all current payers for Hospital Admissions, Outpatient Visits, ER Visits, Professional Claims, Ancillary Claims and Rx Scripts for 2012. Some interpretation of what is included or reflected in these categories is probably in order:
In the last column of the table (near the top) we see a statewide rate of 60 hospital admissions per thousand people in 2012 and an average illness burden score of 1.00. Recall that the statewide illness burden score is set or normalized to a value of one to facilitate comparisons by county and three-digit Zip Code. For the individual counties reflected in the table, the corresponding values are Pitkin (35), Denver (49) and Pueblo (94). Based on the observed values alone, one might be tempted to conclude that the rate of inpatient admissions is very low Pitkin, somewhat higher in Denver (but still well below the statewide value) and very high in Pueblo County. Compared to Expected (C2E) rates, based on average burden of illness scores and expected values, provide additional information that is essential to better understanding variation in the observed rates of hospital admissions.
Based on analysis of Colorado APCD data, the observed rate for Denver County is 49 hospital admissions per 1,000 insured people. While this is 18.3 percent lower than the statewide rate of 60, it is only five percent below expected (C2E) when the average illness burden score of 0.87 is explicitly considered. In other words, this lower rate of hospital admissions is only slightly less than expected given that Denver residents are generally healthier than the statewide population. Similarly, Pitkin County had 35 hospital admissions per thousand people in 2012 – 28.6 percent lower than Denver (49) and 41.7 percent lower than the state (60) – but only four percent lower than expected (C2E) based on a very low illness burden score of 0.76. To complete the example, Pueblo County had a hospital admission rate of 94 per thousand people – 57 percent higher than the statewide rate (60) – but only 12 percent above expected (C2E) based on an illness burden score of 1.34. Although the rate of hospital admissions in Pueblo County is considerably higher than the state, this is not unexpected given the underlying health status of the population based on average burden of illness score.
The bottom line is that while observed variation in utilization rates across counties may appear large at first glance, these differences look different when considered in light of the underlying health status of people living in those areas. Observed differences in utilization may also be due to a variety of other factors including the number and percentage of people covered by Medicaid vs. private insurance and whether people receive health care services in the county where they live. While reports available on www.cohealthdata.org show variation in cost and utilization across regions, they do not specifically address the many potential reasons for the observed differences.
The overall utilization report discussed here provides additional and similar information for the Outpatient Visits, ER Visits, Professional Claims, Ancillary Claims and Rx Scripts categories, including results stratified by age range and gender groupings. Additional utilization reports, not discussed here, provide even greater levels of detail specific to the Inpatient, Outpatient and ER categories.
If you have questions related to risk adjustment, burden of illness scores or compared to expected measures as applied to the Colorado APCD, please contact Jonathan Mathieu, Director of Data and Research, at email@example.com.
Share on: Digg, Facebook, Reddit, Delicious, Twitter
© Copyright. Center for Improving Value in Health Care. All rights reserved.
950 S. Cherry Street, Suite 208, Denver, Colorado 80246 | 720-583-2095 (main phone) | 720-549-9189 (fax) | contact us