Table 1.

Summary of Analytic Steps

Step number	Step title	Step methods	Result of step
Step 1	Visit frequency groups	Categorize Medicare beneficiaries according to their frequency of PC visits	5 groups of beneficiaries stratified by frequency of PC visits
Step 2	Visit regularity subgroups	Sub-divide each of the 5 groups from Step 1 into “regular” and “irregular” subgroups based on regularity of PC visits	10 subgroups of beneficiaries
Step 3	Probability of being in regular subgroup	Use beneficiary demographic and clinical characteristics to predict the probability of a beneficiary having more than the median regularity of PC visits	Estimated probability (range 0–1) for each beneficiary to be assigned to the regular subgroup
Step 4	Expected site-level proportion of beneficiaries with regular visits	Categorize sites based on whether they had more or fewer beneficiaries in the regular PC subgroup than would be predicted	Calculated proportion (range 0–100%) of beneficiaries at each site, within each PC frequency group, predicted to be in the regular subgroup
Step 5	Site-level comparison of observed and expected	Compare the observed proportion of beneficiaries in the regular subgroup at each site with the expected proportion, to generate an observed minus expected (O-E) score	Each site characterized as having more, fewer, or about the same proportion of beneficiaries in the regular subgroup than predicted
Step 6	Beneficiary-level analyses	Test the beneficiary-level hypothesis that beneficiaries with more regular PC visits would have better outcomesa than beneficiaries with less regular PC visits	Beneficiary-level hypothesis is accepted or rejectedb
Step 7	Site-level analyses	Test the site-level hypothesis that sites with more regular PC visits than predicted would have better outcomesa than sites with less regular PC visits than predicted based upon beneficiary characteristics	FQHC site-level hypothesis is accepted or rejectedc

PC primary care

^aBetter outcomes are defined as fewer ED visits, fewer hospitalizations, and lower total Medicare expenditures

^bIf the hypothesis is accepted, beneficiaries with more regular PC visits had better outcomes than beneficiaries with less regular PC visits

^cIf the hypothesis is accepted, beneficiaries at FQHC sites with more regular PC visits than predicted had better outcomes than beneficiaries at FQHC sites with less regular PC visits than predicted