Nationally, 92% of abortions take place in the first trimester,7 but Black, uninsured, and lowincome women have less access to this care.6 In California, only 87% of women using state Medicaid insurance obtain abortions in the first trimester.42 Because the average cost of a second-trimester abortion is substantially higher than that of a first-trimester procedure, shifting the population distribution of abortions to earlier gestations would result in safer, less costly care. Increasing the types of health care professionals involved in abortion care is one way to reduce this health care disparity. Our study was designed to examine the effect of removing the physician-only requirement for aspiration abortion provision in California. We found that the care provided by newly trained NPs, CNMs, and PAs was not inferior to that provided by experienced physicians. We estimate that only 1 additional complication would occur for every 120 procedures as a consequence of having an NP, CNM, or PA as the abortion provider. Additionally, the 0.83% risk difference was mainly TABLE 1—Continued Risk factorsg Extreme obesity (BMI > 40 kg/m2 ) 2.3 2.2 .33 Existing chronic illness 5.0 4.9 .72 Placenta previa (16–18 wk) 0.0 0.0 .32 Psychiatric condition 3.3 3.2 .61 Note. BMI = body mass Index; CNM = certified nurse midwife; NP = nurse practitioner; PA = physician assistant; STI = sexually transmitted infection. Physicians had completed a residency in either obstetrics and gynecology or family medicine. Missing data on age (n = 18), patient insurance (n = 35), cesarean delivery history (n = 82), and gravidity (n = 7) were recoded to mean age, no insurance, no history of cesarean delivery, and median gravidity, respectively. Missing data on gestational age by ultrasound (n = 85) were recoded to gestational age by last menstrual period; where those data were also missing, they were recoded to the mean gestational age by ultrasound. For other missing variables, we created a new variable for missing. a P values are based on a significance level of .05 and were calculated using mixed-effects logistic regression for dichotomous variables, mixed-effects multinomial logistic regression for categorical variables, and mixed-effects linear regression models for continuous variables, all of which included random effects for facility. b Data missing for 70 women in the NP–CNM–PA group and 56 in the physician group. c California’s Medicaid program. d Data missing for 11 women in each provider group. e Data missing for 25 women in the NP–CNM–PA group and 20 in the physician group. f Data missing for 17 women in the NP–CNM–PA group and 18 in the physician group. g All risk factor variables are dichotomous (no–yes). “No” is the reference category (not shown in table). TABLE 2—Overall and Major and Minor Complication Rates by Provider Type at 22 California Clinical Facilities: August 2007–August 2011 Physicians (n = 5812) NPs–CNMs–PAs (n = 5675) Total (n = 11 487) Risk Difference Between Provider Groups (n = 11 487) Complication Type Rate/100 (95% CI) No. Rate/100 (95% CI) No. Rate/100 (95% CI) No. 0.45, 1.29) Note. CI = confidence interval; CNM = certified nurse midwife; NP = nurse practitioner; PA = physician assistant. Physicians had completed a residency in either obstetrics and gynecology or family medicine. RESEARCH AND PRACTICE 458 | Research and Practice | Peer Reviewed | Weitz et al. American Journal of Public Health | March 2013, Vol 103, No. 3 the result of higher incidence of minor complications, the majority of which were from diagnoses easily treated and without consequential sequelae. Moreover, on the basis of findings in other studies, we expect this risk difference to narrow further over time.43---45 The comparison of newly trained NPs, CNMs, and PAs with more experienced physician abortion providers suggests that the small difference found would represent the maximum variation in outcomes that might be expected immediately after a policy change. Both provider groups had extremely low numbers of complications, less than 2% overall—well below published rates—and only 6 complications out of 11 487 procedures required hospital-based care. Because the effect size is minimal compared with the published data and within the prespecified margin of noninferiority, we conclude that the difference between the 2 groups of providers is not clinically significant. While the reported odds ratios comparing complication rates from procedures performed by NPs, CNMs, and PAs with those from procedures performed by physicians were statistically significant, these results should be interpreted cautiously. The study was powered specifically for a noninferiority analysis, which necessitated a larger sample size than a superiority analysis would. Therefore the significance we see may be a result of the study being overpowered. These findings support the adoption of policies that increase access to abortion by expanding the number and type of health care professionals who can perform early aspiration abortions. The benefits of expanding access to abortion for California’s women outweigh the small initial difference in risk, particularly because it would likely move many second-trimester abortions into the first trimester, significantly decreasing the overall risk of complications, which increases with gestational age.4 Expanded