Table 4. Adjusted ORs of Severe Maternal Morbidity Associated With Neuraxial Analgesia for Vaginal Delivery in the Main and Sensitivity Analyses (New York Hospitals, 2010-2017).

Analysis	No neuraxial analgesia			Neuraxial analgesia			Risk difference, % (95% CI)a	Adjusted OR (95% CI)
	Women, No.	Events, No.	Incidence, % (95% CI)	Women, No.	Events, No.	Incidence, % (95% CI)
Main analysis
Inverse propensity score weighting with stabilized weightsb	304 734	4427	1.45 (1.38 to 1.53)	219 308	2727	1.24 (1.17 to 1.32)	−0.21 (−0.30 to −0.12)	0.86 (0.82 to 0.90)
Sensitivity analyses
Inverse propensity score weighting with stabilized weights truncated at 1% and 99%c	292 572	4151	1.42 (1.38 to 1.53)	209 770	2629	1.25 (1.17 to 1.32)	−0.17 (−0.24 to −0.09)	0.90 (0.85 to 0.94)
Propensity score matchingd	96 407	1170	1.21 (1.14 to 1.28)	96 407	1035	1.07 (1.01 to 1.14)	−0.14 (−0.22 to −0.06)	0.91 (0.83 to 0.99)
Handling of missing values for anesthesia care with multiple imputations and inverse propensity score weighting with stabilized weightse	382 685	5322	1.39 (1.33 to 1.45)	331 258	4087	1.23 (1.18 to 1.28)	−0.16 (−0.22 to −0.09)	0.91 (0.87 to 0.95)

Abbreviations: OR, odds ratio; SMM, severe maternal morbidity.

^a 95% CI estimated using bootstrap with replacement (B = 2000) and the percentile method.

^b OR estimated using a weighted logistic regression and further adjustment for 3 variables with persistent imbalance after weighting: (1) hospital proportion of racial and ethnic minority women, (2) hospital proportion of women who were safety net patients, and (3) hospital proportion of neuraxial analgesic or anesthetic techniques in deliveries.

^c OR estimated using a weighted logistic regression and further adjustment for 5 variables with persistent imbalance after weighting: (1) race and ethnicity, (2) health insurance type, (3) hospital proportion of racial and ethnic minority women, (4) hospital proportion of women who were safety net patients, and (5) hospital proportion of neuraxial analgesic or anesthetic techniques in deliveries.

^d OR estimated using a conditional logistic regression and further adjustment for 3 variables with persistent imbalance after matching: (1) hospital proportion of women with comorbidity index for obstetric patients of 1 or more, (2) hospital coding intensity, and (3) year of delivery. Matching used the nearest-neighbor approach with a caliper of 0.2 and 1 case matched to 1 control.

^e OR estimated using a weighted logistic regression. No persistent imbalance was observed after weighting.