Can mid‐level providers manage medical abortion after 12 weeks' gestation as safely and effectively as physicians? A non‐inferiority, randomized controlled trial in Addis Ababa, Ethiopia

To determine whether clinical outcomes among clients undergoing medical abortion after 12 weeks' gestation differ by provider cadre.


| INTRODUC TI ON
WHO recommends that abortion services be available at the lowest level of the healthcare system and provided by many cadres of health professionals. 1,2Provision by mid-level providers is a recognized facilitator to access abortion services in the first trimester. 3Comparative studies of the provision of medical and surgical abortion in the first trimester by mid-level providers demonstrate the safety, feasibility, and utility of a policy allowing mid-level provision in disparate countries globally. 46][7] Randomized controlled trials found no differences in incomplete abortion or complication following surgical or medical abortion in the first trimester when performed by mid-level providers or physicians. 6aining midwives, nurses, and other non-physician providers to conduct abortions after 12 weeks is proposed as a strategy to increase access to abortion procedures at later gestations.WHO highlighted expanding provider cadres as an area in which rigorous research is needed. 8Later abortions are a persistent need; about 10% of all abortions take place beyond 12 weeks' gestation globally. 9,10[13][14] In Ethiopia, abortions over 12 weeks' gestation are conducted at facilities by physicians, but there are insufficient providers to meet the current need. 156][17] Expanding services to providers has the potential to increase accessibility, particularly in rural areas.
To contribute to the call for research, we investigated whether outcomes differ among women who receive this care from mid-level providers and those who are cared for by physicians.Michu Clinic, a comprehensive reproductive health clinic at Saint Paul Hospital Millennium Medical College, was an ideal site for our study.Michu is a high-volume clinic (80-100 clients per month) providing abortion services through the second trimester and serving as a teaching and research center for trainees.The study's objective was to investigate whether clinical outcomes for women who underwent mifepristone and misoprostol abortion beyond 12 weeks' gestation differed between those receiving care from trained mid-level providers and those receiving care from physicians.We designed a similar study as those that demonstrated that mid-level providers could provide first-trimester abortions as safely and effectively as physicians.Participants were eligible to participate in the trial if they were: (1) seeking induced abortion at Michu Clinic during the study period, (2) between 13 and 20 weeks' gestation, (3) medically eligible, (4) at least 16 years old, and (5) chose medical abortion, as dilation and evacuation (D&E) was also available.Recruitment took place in the clinic reception before seeing a provider; those seeking abortion who reported a last menstrual period between 13 and 20 weeks prior were approached by a trained enumerator.The enumerator explained the study, obtained informed consent for those interested, and completed a baseline survey to capture socio-demographic characteristics, and randomized the participant to the provider group (mid-level or physician provider).
During intake, the provider determined or confirmed the gestational age of the pregnancy using standard assessments (last menstrual period, bimanual exam, or ultrasound).Anyone determined to be less than 13 weeks or greater than 20 weeks by the provider was ineligible to participate in the study despite having been randomized by the research assistant.Abortion services were available at the clinic beyond 20 weeks for those who exceeded the gestational age for the study.Providers also confirmed medical eligibility: anyone with an allergy to mifepristone or misoprostol, history of more than one hysterotomy, or with chronic hypertension, adrenal failure, or on chronic steroids was excluded.
Participants in both treatment arms received 200 mg mifepristone orally and were sent home with a planned return 1-2 days later for misoprostol administration.Misoprostol was administered in-patient, 400 μg, sublingually every 3 h until fetal and placental expulsion, as per facility.Discharge occurred as early as 2 h after expulsion, or as late as the next morning.In the experimental arm, all the clinical care, including administration of mifepristone and misoprostol, was performed by mid-level providers.In the control abortion, medical abortion, mid-level providers, provider cadre, second-trimester medical abortion arm, the standard of care was maintained: physicians performed all clinical care (including mifepristone administration) and oversight of facility nurses, who administered repeated doses of misoprostol.
The clinical care included pre-abortion assessment, including medical eligibility and assessment of gestational age, administration of medicines, management of the abortion process in the clinic, placental management, and provision of contraception and discharge from the facility.
Trained enumerators entered data using CommCare (Dimagi) on electronic tablets.They collected baseline demographic information and eligibility to participate in the study from the participant using a questionnaire.After randomization, the allocated provider documented on a data collection sheet their medical and physical assessment of the patient, and details of the abortion process, which were entered by the enumerators directly onto tablets.Enumerators monitored completeness of the provider documentation throughout the abortion process, to avoid missing data.Before discharge, female enumerators conducted an interviewer-administered survey with the patient and entered the data directly into the tablet.
0][21][22][23] The time of first misoprostol dose and time of fetal expulsion were recorded by the provider and used to calculate the interval.Secondary endpoints included other time points, including time from mifepristone to fetal expulsion and time from fetal to placental expulsion.Providers recorded the date, time, dose, and route for each misoprostol dose, to calculate the total number of doses received.The incidence of retained placenta (non-expulsion requiring intervention to remove the placenta) was recorded by the provider.The provider also recorded any complications (hemorrhage, blood transfusion, infection or fever requiring treatment, or need to change the procedure).Before discharge, the patient was interviewed by the enumerator: the questions asked them to compare their experience to what they had expected after their initial consultation with the provider, specifically level of pain, amount of bleeding, duration, privacy during consultation and abortion process, and side effects.They were also asked about overall satisfaction, the acceptability of the abortion process, and any decision regarding uptake of post-abortion contraception.
1][22][23][24] We required 88 participants to determine non-inferiority of the mid-level provider group, with a one-sided margin of −1.5 h, assuming 80% power (β = 0.2), and a one-sided confidence level of 97.5% (α = 0.025).We anticipated post-randomization clustering at the provider level; thus, we increased the sample size to a total of 111 patients by including an intraclass correlation (ρ) of 0.026 at the provider level based on previous studies 26  We assessed non-inferiority of the mid-level compared with the physician group in terms of median time to fetal expulsion using unadjusted and adjusted regression models. 27The mid-level provider group would be considered non-inferior to the physician group if the one-sided confidence interval (CI) for the difference in median time to expulsion was within the predetermined non-inferiority margin of −1.5 h after accounting for provider-level clustering. 28For secondary outcomes, we computed crude and adjusted differences in medians using quantile median regression models for continuous outcomes, and computed crude and adjusted risk differences from log relative risks using bivariate Poisson generalised estimating equations (GEE) models with log-link for dichotomous outcomes.Additional analyses included patient demographics, secondary outcomes, and patient satisfaction across provider groups using χ 2 , Wilcoxon-Mann-Whitney, and log-rank tests, as appropriate.All adjusted models included parity and gestational age as prognostic covariates [20][21][22][23][24][25] and accounted for clustering at the provider level.Data were analyzed using Stata (16.1;StataCorp, College Station, TX, USA).

| RE SULTS
Over a period of 8 months of data collection during 2020-2021, interrupted by 5 months during the COVID-19 pandemic, 267 women were assessed for eligibility: 245 (91.8%) consented to participate; 230 (93.9%) were randomized; and after assessment by the allocated provider, 171 (74.3%) remained eligible to participate and were enrolled-81 (47.4%) in the physician and 90 (52.6%) in the mid-level provider group (Figure 1).Sociodemographic and reproductive characteristics of participants did not differ by provider group, suggesting successful randomization (Table 1).Participants were aged 24 on average, half had less than primary education, 135 (78.9%) were from Addis Ababa, more than three-quarters were employed in the prior year, and slightly more than half had never been married.For half of participants, this was their first pregnancy; 108 (63.2%) had no living children and 31 (  methods discussed was 2.1 (0.5) with no difference between groups (P = 0.593).The implant was the most adopted modern method: 45 (55.6%) in the physician group and 62 (68.9%) in the mid-level provider group (P = 0.072).The next most common method was the injectable (n = 26, 15.2%), and 30 (17.5%) received no method of contraception.
Patient experience was largely positive and did not differ by provider group; most reported the pain (n = 107, 62.6%), bleeding (n = 116, 67.8%), duration (n = 107, 62.6%), and side-effects (n = 126, 73.7%) of the abortion process to be about as expected or less than expected, but approximately one-third reported that these were more than expected (pain: n = 64, 37.4%; bleeding: n = 53, 31.0%,duration: n = 63, 36.8%,side effects: n = 44, 25.7%) (Table 3).Participant's reports of the client-provider interaction suggested no significant differences in quality of clinical care provided.A total of 136 (79.5%) patients reported that the abortion method was explained clearly and 142 said that they understood (83.0%), but only two-thirds reported being given the opportu-

| DISCUSSION
Induced abortion beyond 12 weeks' gestation may safely be provided by mid-level providers, although further research in varying settings is needed.We found medical abortion provision by mid-level providers was non-inferior to physicians in terms of time to abortion.Our primary outcome measure-time to fetal expulsion-measures an effective and efficient abortion service which mid-level providers in this study were able to provide.We also found that secondary clinical outcomes-abortion quality of care and patient experience-were comparable between the two provider groups.
Within the abortion service, we found that, although the physician providers used ultrasound assess age at much higher rates than the mid-level providers, the confirmed gestational age by fetal foot length measurement after expulsion revealed that the two provider groups only varied by a small, non-significant amount.Accurate assessment of gestational age is important to ensure appropriate abortion care, including correct misoprostol dosing; however, the recommended misoprostol regimen only differs for pregnancies greater than 24 weeks.We did not observe mis-estimates of pregnancy duration among either group which would have required different dosing or care, but such outcomes would be important to include in future confirmatory studies.
We found a decreased time interval from misoprostol to fetal expulsion among those with higher parity, consistent with the global literature of induced abortion beyond 12 weeks' gestation.1][22][23][24][25] We did not detect an increase in time to expulsion among those with higher gestational ages or among those who were older, likely because of the relatively small number of higher gestational age pregnancies and participants older than 25 years.
The experiences reported by study participants were generally positive regardless of provider group, similar to previous studies in which patients reported services provided by nurses and midwives as equally acceptable and satisfactory as physician providers. 29,30ients and patients report preferring services performed by nurses and midwives, and expanding their role in caring for abortion patients is recognized as a strategy that would improve care. 29,31However, we found that patient experience was comparable between groups, indicating that care at St Paul's was perceived as acceptable by patients regardless of the provider cadre.Some practices at St Paul's did not conform to quality care as recommended by the WHO. 1 One such practice was in the management of the placenta following fetal expulsion: providers intervened after waiting only 30 min for spontaneous expulsion, a protocol typically used after childbirth and not uncommon for abortion beyond 12 weeks in settings where providers also care for obstetric patients.
Prior studies have demonstrated the safety of waiting up to 4 h with no resulting adverse events. 31,32As uterine evacuation interventions may be costly, physically uncomfortable for the woman, and may require the involvement of other staff, additional supplies, and use of procedure rooms, awaiting placental expulsion is a useful practice to introduce into facility protocols.Additionally, most women at St. Paul's received no pain management during the entire abortion process, whereas WHO and professional guidelines recommend using non-steroidal anti-inflammatories at a minimum, and other modalities to be offered, if available. 1,31mitations of this study include that it was performed at a single site in an urban center, which limits its generalizability.Importantly, we introduced mid-level providers to abortion care from 13 to 20 weeks' gestation in a setting where physician back-up was readily available, which may have increased their confidence and willingness to participate in training and provision.Future research should include rural intervention sites to test the applicability of mid-level provider provision remote from physicians doing the same care. 33ditionally, some procedure data, such as pain medicine and contraceptives offered, may have been under-reported in the providers' documentation; in client exit interviews, slightly more patients reported being offered pain medicine and a greater variety of contraceptives than were recorded by the providers.
In conclusion, we found that training mid-level clinicians to provide abortion services from 13 to 20 weeks' gestation, independent of physicians, is feasible and may result in comparable clinical outcomes in this setting.As these results have the potential to impact policy, the facilitators and barriers to expanding provision by mid-level providers should be explored and further confirmatory research in various settings should be conducted.Advantages to policy change could be greater abortion access to those in rural communities and cost-savings on providers and technologies such as ultrasound and uterine aspiration equipment.

2 |
MATERIAL S AND ME THODSWe conducted a randomized controlled non-inferiority trial comparing clinical outcomes between women who received medical abortion between 13 and 20 weeks' gestation from mid-level providers (nurses or midwives) and those who received care from physicians specialized in obstetrics and gynecology (OB/GYN).Clinical training in second-trimester medical abortion by OB/GYN staff, including by two of the study investigators, for the mid-level providers working in the department was conducted over 5 days.A 6-week period of supervision of the nurses and midwives providing second-trimester care followed the training.The trial was initiated after completion of the supervision period and was conducted from February 2020 to March 2021.The Institutional Review Board of St Paul's Hospital Millennium Medical College (PM23/146) approved the study.
and an average cluster size (m) of 11 patients.With rounding, the final sample size was 120, assuming an average of 10 clients recruited from 12 providers (six physicians and six mid-level providers).There was one deviation from the protocol: we enrolled a larger number of participants than the stated sample size.As part of the COVID-19 response, physicians rotated rapidly through the OB/ GYN department, resulting in 10 physicians (rather than six) caring for patients randomized to the control group.To mitigate the impact of informative clustering (smaller number of participants per physician compared with mid-level providers), we kept recruiting patients to the physician group until six of the 10 physicians attended at least 10 patients, in accordance with initial calculations.Randomization occurred at the individual participant level.A simple randomization sequence was generated in Stata SE/16 by the offsite study team and used to create sequentially numbered opaque envelopes for allocation.The randomization group determined whether the participant would receive clinical care from a mid-level or physician abortion provider for their medical intake, physical exam, and subsequent abortion management.The envelope was opened by the facility-based enumerator to assign the treatment group after having obtained informed consent from the participant.

TA B L E 3
nity to ask questions (n = 111, 64.9%) and three-quarters reported receiving information about post-abortion complications (n = 128, Participant experiences with medical abortion care by provider group (n = 171).
Crude and adjusted differences in primary and secondary outcomes by provider group (n = 171).
18.1%) reported a previous abortion.The mean gestational age was 16.6 weeks.Gestational age was assessed differently among the two provider groups: ultrasonography was used for 73 (90.1%)FIGURE 1 CONSORT flow.siondidnotdiffersignificantly between provider groups: 6.6 (±9.1) h amongst the physician and 8.1 h (±13.0) for the mid-level group (Table2).The crude median difference was −1.5 h (95% CI -3.45 to 0.46), while the adjusted median difference was 0.8 h (95% CI -1.15 to 2.66).The lower bound of the CI on the adjusted median difference was within the non-inferiority margin of −1.5 h, suggesting non-inferiority of the mid-level provider versus the physician group when holding gestation and parity constant.The point estimate increased, and the CI narrowed in the adjusted model as median time to fetal expulsion was highly associated with parity; compared with participants with no prior births, those with one prior birth had a shorter median interval by 3.7 h (95% CI -6.61 to 0.12) and those with two or more prior births had a shorter median interval, 4.1 h shorter (95% CI -5.79 to 2.43).Median time to fetal expulsion was not associated with gestational age.TA B L E 1 Participant socio-demographic and reproductive characteristics by provider group (n = 171).Analysis of secondary outcomes showed no significant differences between provider groups (Table2).The median time from mifepristone to fetal expulsion was 52.7 h in the physician group compared with 51.7 h in the mid-level provider group (adjusted median difference 1.8 h, 95% CI -2.01 to 5.67), and the median time from fetal expulsion to placental expulsion was 0.3 h in the physician group versus 0.2 h in the mid-level provider group (adjusted median sion occurred in the physician provider group; she received mifepristone 36 h prior and was scheduled to return for misoprostol but expelled the pregnancy at home.Mid-level providers successfully managed most cases independently; in only 8 (8.9%) cases did they consult with the OB/GYN during their management.These eight patients received more misoprostol (mean 6.9 doses, ±7.2), were more likely to have a retained placenta (n = 5 [62.5%] vs n = 13 [15.9%]among the remaining midlevel provider group and n = 24 [29.6%]among the physician group; P < 0.001), and had a higher incidence of complications (n = 4 [50.0%]vs n = 4 [4.9%]among the remaining mid-level provider group, and n = 5 [6.2%] among the physician group; P < 0.001).Complications among these four patients included three hemorrhages, one of whom required a transfusion, and one woman needing to switch procedure to a D&E. in the physician group (n = 16, 19.8%) received ibuprofen compared with the mid-level provider group (n = 30, 33.3%) (P = 0.046).Other analgesic (diclofenac, indomethacin, tramadol) use did not differ by provider group.Most patients received contraceptive counseling (n = 161, 94.2%) before discharge, with no difference between provider groups (P = 0.336).The mean number of contraceptive TA B L E 2