Table 2.

F-Statistics Testing Whether Patient Groups Defined by Model Variables Are Related to BCSI Choice for Stage II ESBC Patients in Iowa, 1989–1994^a

	Surgery Choice Model Specification
	Patient Groups per Instrument
	2	4	8	12
Instrument Groups
Area BCSI rate	8.57***	5.19***	3.43**	3.00***
Covariate Groups
Ageb	6.55***	6.49***	6.38***	6.78***
Tumor sizeb	14.67***	15.10***	15.73***	15.52***
Positive nodesb	12.86***	12.76***	13.15***	12.99***
Gradeb	0.98	0.93	0.98	0.94
Tumor locationb	5.77***	5.59***	5.58***	5.52***
Comorbidity indexb	0.96	0.91	0.96	0.76
Povertyb	4.02***	3.78***	3.96***	3.37***
Hospital accessb	0.65	0.55	0.56	1.07
Payerb	3.12***	3.32***	3.30***	3.29***
Yearb	2.46**	2.31**	1.79	2.06*

^***Statistically significant at .99 confidence.

^**Statistically significant at .95 confidence.

^*Statistically significant at .90 confidence.

^aAgainst the null hypothesis that the groups defined by the instrumental variables do not describe a significant portion of treatment variation.

^bBinary variables for age groups (<50, 50–64, 65–69, 70–74, 75–79, 80–84, 85+), tumor sizes (<2 cm, 2–5 cm, 5+cm), positive lymph node involvement, tumor grade groups (1, 2, 3, 4, 9–unknown), tumor location groups (nipple, central portion, upper-inner quad, lower-inner quad, upper-outer quad, lower-outer quad, axillary tail, overlapping lesion, not stated), Charlson comorbidity index (0, 1, 2, 3+), residence zip code poverty percentage (#7, 7–10, 10–13, 13–20,>20), distance from residence to nearest hospital (#2.83, 2.83–9, 9–15, >15), payer (Medicaid, Medicare, Blue Cross/Blue Shield, other private, other government, self-pay), and year of diagnosis (1989, 1990, 1991, 1992, 1993, 1994).