. 2020 Nov 7;12(21):21730–21746. doi: 10.18632/aging.103978

Table 3. Cox proportional hazard regression with patient mortality as the dependent variable.

Outcomes	Events	Person-year^#	Incidence density*	Crude^$		Model A^&		Model B^%
Outcomes	Events	Person-year^#	Incidence density*	HR	95% CI	HR	95% CI	HR	95% CI
Total cohort
Naïve patients	4,018	110,097.3	36.5	1	-	1	-	1	-
Users	4,540	110,622.9	41.0	1.124	1.08 – 1.17^a	1.183	1.13 – 1.24^a	1.183	1.13 – 1.24^a
Frail participants (Item ≥ 3)
Naïve patients	137	1,646.2	83.2	1	-	1	-	1	-
Users	126	1,707.4	73.8	0.886	0.7 – 1.13	1.08	0.83 – 1.4	1.079	0.83 – 1.41
Non-frail participants (Item < 3)
Naïve patients	3,881	108,451.1	35.8	1	-	1	-	1	-
Users	4,414	108,915.5	40.5	1.12	1.07 – 1.19^a	1.183	1.13 – 1.24^a	1.183	1.13 – 1.24^a
Non-frail participants (Item = 0)
Naïve patients	444	24,638.9	18.0	1	-	1	-	1	-
Users	743	23,874.6	31.1	1.727	1.54 – 1.94^a	1.503	1.33 – 1.69^a	1.503	1.33 – 1.69^a
Non-frail participants (Item = 1)
Naïve patients	2,383	63,425.7	37.6	1	-	1	-	1	-
Users	2,693	63,166.5	42.6	1.134	1.07 – 1.2^a	1.173	1.11 – 1.24^a	1.175	1.11 – 1.24^a
Non-frail participants (Item = 2)
Naïve patients	1,054	20,386.5	51.7	1	-	1	-	1	-
Users	978	21,874.5	44.7	0.866	0.79 – 0.95^b	1.026	0.94 – 1.12	1.024	0.94 – 1.12

aDCSI, adapted diabetes complications severity index; CI, confidence interval; HR, hazard ratio

^#Cumulative follow-up duration

^· per 1000 patient-year

^$ Incorporating opioid use only

^& Incorporating age/gender, lifestyle factors, all comorbidities, aDCSI, all medications, FRAIL item counts, treatment variables, and opioid use; findings without an attached superscript “a” or “b” denote insignificant results (p > 0.05)

^% Incorporating model A components and individual FRAIL item status

^a p < 0.001

^b p < 0.01