Table 3.

Inference Subpopulation (SubPOP), FIRST Sample (SAMP), and Subpopulation‐Weighted Sample (SubPWS) Comparisons

Variable	SubPOP vs. SAMP		SubPOP vs. SubPWS
	|SMD|	Rubin Criterion 3 σ 2 SUBPOPPRESID/σ 2 SAMPRESID a	|SMD|	Rubin Criterion 3 σ 2 SUBPOPRESID/σ 2 SUBPWSRESID a
Propensity score (logit scale)	2.30b	—	0.11	—
Program type
Academic	0.35	1.03	0.17	1.21
Community‐based	0.28	1.02	0.09	0.90
Military	0.16	5.83a	0.17	1.57
Number of PGY5 slots	0.42	0.80	0.23	1.44
Any ACGME accreditation issue	0.22	0.80	0.16	0.90
Research‐oriented	0.37	1.41	0.00	0.84
2013 mean ABSITE scores	0.20	1.30	0.11	1.05
Annual hospital admissions (1000s)	0.45	1.59	0.05	2.48a
Type of control
Not‐for‐profit	0.21	1.04	0.12	1.02
For‐profit	0.16	1.40	0.13	1.24
Other	0.13	1.03	0.06	1.05
COTH membership	0.49	1.05	0.01	0.95
Percent Medicaid discharges	0.13	2.43a	0.14	1.57
Full‐time RN/bed	0.04	1.33	0.02	1.14
Any transplant services	0.55b	1.16	0.08	1.06
Level 1 trauma designation	0.49	1.20	0.06	1.35
Census division
1—New England	0.31	0.82	0.08	0.46a
2—Mid‐Atlantic	0.27	1.11	0.02	0.89
3—East North Central	0.01	0.93	0.11	0.75
4—West North Central	0.16	1.13	0.10	2.47a
5—South Atlantic	0.01	0.99	0.11	0.95
6—East South Central	0.11	1.09	0.00	1.33
7—West South Central	0.09	1.09	0.11	1.31
8—Mountain	0.06	1.20	0.06	1.35
9—Pacific	0.04	1.04	0.07	1.22

Additional tests: Rubin (2001) Criterion 2: ratio of variance of propensity score between groups should be within (0.5, 2) range. This ratio was 1.00 between SubPOP and SAMP and 17.23 between the SubPOP and SubPWS. Tipton's Generalizability Index (β) was 0.80 (SubPOP vs. SAMP). According to Tipton (2014): β < 0.50 “low”; 0.50 ≤ β < 0.80 “medium”; 0.80 ≤ β < 0.90 “high”; 0.90 ≤ β “very high” generalizability of sample.

^aRubin (2001) Criterion 2: ratio of variance of residuals should be within (0.5, 2) range.

^bRubin (2001) Criterion 1: |SMD|s > 0.5 SDs may indicate distributional dissimilarity between groups.