Exhibit 3.

Estimated Effect Of A Consultation With A Palliative Care Team Within Two Days Of Admission On Total Direct Hospital Costs For Patients With Advanced Cancer, By Elixhauser Comorbidity Score At Admission

	Sample			Utilization summary			Primary results
Comorbidity score	Compariso n group	Treatme nt group	All patie nts	Proporti on in treatmen t group	Mean direct costs per patient	Mean LOS per patient (days)	Estimated mean treatment effecta	95% CI	Implied mean savingb
0–1	207	28	235	12%	$8,440	7.4	−$1,775	−5,093, 1,544	18%
2–3	276	86	362	24	8,528	7.0	−2,321*	−3,869, −773	22
4 or higher	230	79	309	26	10,030	8.2	−3,515*	−5,949, −1,081	32

SOURCE Authors’ analysis. NOTES We regressed total direct costs against a binary intervention variable, the independent variables listed in Exhibit 1, and fixed-effects variables to control for hospital site, applying subsample-specific propensity score weights in all cases. Further details are available in the Appendix (see Note 25 in text). Of the 906 matched patients, 98 had an advanced cancer that was not lymphoma, metastatic cancer, or a solid tumor. Of these, 49 (50 percent) had an advanced cancer diagnosis but no comorbidity on the Elixhauser comorbidity scale and thus had an Elixhauser comorbidity score of 0, as explained in the text. The other 49 had an Elixhauser comorbidity score of 1–8 (median: 2), with their cancer diagnosis unaccounted for in the score. Costs are in 2011 dollars. Significance refers to the estimated mean treatment effect.LOS is length-of-stay. CI is confidence interval.

^aThe estimated effect on total direct hospital cost of moving a patient from the comparison arm to the treatment arm, holding all other values constant.

^bSavings in total direct hospital costs resulting from the consultation. Further details on how these savings were calculated are available in the Appendix (see Note 25 in text).

^*p < 0.01