Table 2.

Comparison of AHR High Performers With All CPC+ Practices at Baseline, 2016

CharacteristicAll CPC+ Practices
(N = 2,888)		AHR High Performers
(n = 14)
Practice site characteristicsa
Practice size, %b
  1-2 PCPs  34    0
  3-5 PCPs  38  29
  ≥6 PCPs  28  71
Attributed Medicare fee-for-service beneficiaries in 2018, mean No.b7101,683
Prior transformation experience, %c  61100
Hospital/system owned (vs independent), %  55  57
Enhanced CPC+ (Medicare and payer partner) payments per NPI in 2018, median (SE), $        42,964 (41,043)        47,559 (43,865)
Beneficiary characteristicsd
Age group, y, %
  ≤64  16  17
  65-74  47  46
  75-84  26  26
  ≥85  12  11
Female, %  59  58
HCC score, mean (SE)e                    1.08 (0.17)                    1.14 (0.08)
Dually eligible for Medicare and Medicaid, %  15  16
Original reason for Medicare enrollment, %
  Age  78  76
  Disability  22  23
  End-stage renal disease    1    1
Race/ethnicity, %
  Black    7    8
  White  86  87
  Hispanic    1    1
  None of above    6    4
Market characteristics
Household income in practice county, median (SE), $54,208 (15,054)53,164 (16,222)
Location, %
  Rural    9  21
  Suburban  15    7
  Urban  76  71
Region, %
  Northeast  29  21
  Midwest  35  29
  South  18  14
  West  18  36
Number of acute care hospital beds per 1,000 population in practice county, %
  1st quartile  26  21
  2nd quartile  26  14
  3rd quartile  26  59
  4th quartile  22  14

  • AHR = acute hospitalization rate; CPC+ = Comprehensive Primary Care Plus; HCC = hierarchical condition category; NPI = national provider identifier; PCP = primary care practitioner.

  • Sources: See Table 1 footnotes.

  • Note: Percentages do not always add to 100 because of rounding.

  • ↵a AHR high performers did not differ from CPC+ practices overall regarding Medicare Shared Savings Program status or CPC+ model track.

  • ↵b A change in AHR in smaller practices could be due to chance from small sample sizes, rather than real change. Very small practices therefore tended not to be identified as AHR high performers because the Bayesian model could not achieve a high level of confidence of a real change based on a small number of attributed Medicare fee-for-service beneficiaries.

  • ↵c See Table 1 footnotes for definition.

  • ↵d Based on Medicare fee-for-service beneficiaries attributed to practices in 2016.

  • ↵e Based on diagnoses in 2015. Score is normalized to a value of 1. Patients more healthy than average will have a score less than 1; those less healthy than average will have a score greater than 1.