training phase and were therefore not included in this analysis. As a result of a protocol violation at 1 site, 79 patients in the physician group were excluded. Follow-up data were available for 69.5% of patients, and follow-up rates were nondifferential between provider groups. Patients who did not return the follow-up survey were retained in the analytic sample because we found that they contacted the facility when they did experience a complication (n = 41), which we also discovered via medical chart abstraction, suggesting a low likelihood of missing complications among this group. Additionally, in a sensitivity analysis, complication incidence and risk differences were similar when we excluded patients who did not return the RESEARCH AND PRACTICE 456 | Research and Practice | Peer Reviewed | Weitz et al. American Journal of Public Health | March 2013, Vol 103, No. 3 follow-up survey. Patients without follow-up data were more likely to have no insurance, have fewer risk factors, be multigravida, and be at less than 5 weeks gestation than were those with follow-up data (P < .05; not shown). The final analytic sample size was 11 487; of these procedures, 5812 were performed by physicians and 5675 were performed by NPs, CNMs, or PAs. Patient Characteristics The majority of women in both groups had had 3 or more pregnancies; no previous cesarean deliveries, miscarriages, or induced abortions; and no history of medical risk factors (Table 1). Women in the NP---CNM---PA group were more likely to be younger (P < .01), less likely to be Asian than White (P < .01), and more likely to be non-Hispanic Black than White (P < .03). Women were similar on all other sociodemographic characteristics across provider groups. Outcomes Overall, complications were rare (Table 2). Out of 11 487 aspiration abortions, 1.3% (n = 152; 95% CI = 1.11, 1.53) resulted in a complication; 1.8% of NP-, CNM-, and PA-performed aspirations and 0.9% of physicianperformed aspirations resulted in a complication. The majority of complications (146/152, or 96%) were minor (1.3% of all abortions) and included cases of incomplete abortion (n = 9 among physicians, n = 24 among NPs, CNMs, and PAs), failed abortion (n = 7 among physicians, n = 11 among NPs, CNMs, and PAs), bleeding not requiring transfusion (n = 2 among NPs, CNMs, and PAs), hematometra (n = 3 among physicians, n = 16 among NPs, CNMs, and PAs), infection (n = 7 among physicians, n = 7 among NPs, CNMs, and PAs), endocervical injury (n = 2 among physicians, n = 2 among NPs, CNMs, and PAs), anesthesia-related reactions (n = 1 among physicians, n = 1 among NPs, CNMs, and PAs), and uncomplicated uterine perforation (n = 3 among NPs, CNMs, and PAs). We classified complications without clear etiology but accompanied by patient symptoms as symptomatic intrauterine material (n = 16 among physicians, n = 24 among NPs, CNMs, and PAs). We classified 11 minor complications as “other”; 4 were from physician-performed procedures TABLE 1—Baseline Characteristics of Patient Study Participants by Provider Tested positive for an STI 3.6 3.4 .77 Continued RESEARCH AND PRACTICE March 2013, Vol 103, No. 3 | American Journal of Public Health Weitz et al. | Peer Reviewed | Research and Practice | 457 (1 urinary tract infection, 1 possible false passage, 1 probable gastroenteritis, 1 unspecified allergic reaction), and 7 were from NP-, CNM-, or PA-performed procedures (1 fever of unknown origin, 1 intrauterine device---related bleeding, 3 sedation drug errors, 1 inability to urinate, 1 vaginitis). Only 6 major complications occurred (3 in each provider group), which included 2 uterine perforations, 3 infections, and 1 hemorrhage. We found no difference in risk of major complications between provider groups: 0.001% (95% CI = –0.08, 0.09). The overall unadjusted risk difference for total complications between NPs, CNMs, and PAs and physicians was 0.87% (95% CI = 0.45, 1.29). The risk difference in immediate complications (n = 9 for physicians; n = 20 for NPs, CNMs, and PAs) was 0.20% (95% CI = 0.01, 0.38); for delayed complications (n = 43 for physicians; n = 80 for clinicians), it was 0.67% (95% CI = 0.29, 1.10). Abortions by NPs, CNMs, and PAs were 1.92 (95% CI = 1.36, 2.72) times as likely to result in a complication as those performed by physicians after adjusting for potential confounders (see table available as a supplement to the online version of this article at http:// www.ajph.org). Among the propensity scorematched sample, complications were 2.12 (95% CI = 1.33, 3.37) times as likely to result from abortions by NPs, CNMs, and PAs as by physicians. The corresponding risk differences were 0.70% (95% CI = 0.29, 1.10) in overall complications between provider groups in the adjusted model and 0.83% (95% CI = 0.33, 1.33) in the propensity score---matched sample. The estimated 95% CIs for risk differences in unadjusted, adjusted, and propensity score--- matched analyses all fell well within the predetermined margin of noninferiority, and therefore complication rates from aspiration abortions performed by recently trained NPs, CNMs, and PAs were statistically no worse than those from those performed by the more experienced physician group (Figure 2). DISCUSSION In 2008, 1.21 million abortions took place in the United States, with more 200 000 (18%) in the State of California.2