06.03.2013 Views

mmpc - National Indian Health Board

mmpc - National Indian Health Board

mmpc - National Indian Health Board

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Methods<br />

on five clinical dimensions: 1) Duration of the condition (acute, recurrent, or chronic); 2) Severity of<br />

the condition (e.g., minor and stable versus major and unstable); 3) Diagnostic certainty (symptoms<br />

versus documented disease); 4) Etiology of the condition (infectious, injury, or other); and 5)<br />

Specialty care involvement (e.g., medical, surgical, obstetric, hematology). The Aggregated Diagnosis<br />

Groups are then weighted and added so that a single ACG morbidity (risk) score is obtained by each<br />

individual.<br />

The ACG system provides an external standard sample of privately insured population in a managed<br />

care plan for reference, but the system recommends use of an internal reference group to<br />

standardize for unadjusted differences between Medicaid populations and provider claims and<br />

privately insured managed care data. We used the random sample of non-Hispanic White Medicaid<br />

recipients living in the same counties as the AIAN as the morbidity reference group.<br />

The software is not recommended for use without a 12 month continuous claim history for<br />

recipients, and therefore we did not apply it to recipients with less than 12 months enrollment. We<br />

did not use the feature of the software that uses pharmacy data because we found that IHS AIAN in<br />

the Medicaid data had much lower pharmacy payments than other which is likely due to access to<br />

medications through their <strong>Indian</strong> healthcare providers, which Other AIAN and Whites do not have<br />

(see Prescription Drug FFS data in the Findings section).<br />

Statistical Adjustments<br />

To adjust for the effects of determinants on mean total payments per recipient, we constructed<br />

Generalized Linear Mixed Models of SAS statistical software (version 9.2) using the log-likelihood<br />

function, assuming the gamma distribution of the dependent variable (Medicaid payments). We<br />

have previously published studies using these methods with Medicaid claims data (Wong et al,<br />

2006), including using the ACG risk adjustment (Korenbrot, Kao & Crouch, 2009).<br />

Medicaid Enrollee Exclusions were: 1) enrollees with no known months of eligibility, 2) enrollees<br />

with longterm or institutionalized care, 3) enrollees with ‘Restricted’ rather than ‘Full’ Medicaid<br />

benefits, and 4) enrollees with no paid claims of $1 or more. Recipients remaining were then<br />

grouped for analytical modeling as in Table 9:<br />

15

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!