Who can provide effective and safe termination of pregnancy care? A systematic review*

Abstract

Background Unsafe termination of pregnancy is a major contributor to maternal morbidity and mortality. Task sharing termination of pregnancy services between physicians and mid-level providers, a heterogeneous group of trained healthcare providers, such as nurses, midwives and physician assistants, has become a key strategy to increase access to safe pregnancy termination care.

Objectives To systematically review the evidence to assess whether termination of pregnancy services by nonphysician providers can be performed safely and effectively.

Search strategy We searched the Cochrane Central Register of Controlled Trials, MEDLINE, MEDLINE in process and other nonindexed citations and POPLINE.

Selection criteria We included randomised controlled trials (RCTs), as well as clinical studies, using study designs that compared efficacy, safety and acceptability of termination of pregnancy services by physicians versus other provider groups.

Data collection and analysis Two reviewers independently extracted the data, and we performed a meta-analysis where appropriate using RevMan. Quality assessment of the data used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.

Main results We identified five controlled studies comprising 8908 women undergoing first-trimester surgical termination of pregnancy (one RCT and three prospective cohort studies) and medical termination of pregnancy (one RCT). The mid-level provider group included midwives, nurses, auxiliary nurse midwives and physician assistants trained in termination of pregnancy services. Safety and efficacy outcomes, including incomplete termination of pregnancy, haemorrhage, injury to the uterus or cervix, did not differ significantly between providers.

Author’s conclusions Limited evidence indicates that trained mid-level providers may effectively and safely provide first-trimester surgical and medical termination of pregnancy services. Data are limited by the scarcity of RCTs and biases of the cohort studies.

Introduction

Multiple barriers may prevent access to safe abortion care (termination of pregnancy or management of miscarriage). Even when legal, access to safe services is often restricted by the lack of trained providers,¹ particularly in nonurban settings. Many countries have policies that limit the provision of services to physicians or physician specialists (obstetrician gynaecologists), thereby limiting the number of available providers and further restricting women’s access.

Improving maternal health is one of the eight Millennium Development Goals,² and an estimated 13% of all maternal deaths are caused by unsafe termination of pregnancy.³ Task shifting termination of pregnancy care services to mid-level providers (MLPs), a heterogeneous group of various nonphysician healthcare providers, such as nurses, midwives and physician assistants, has become a key strategy to achieve this goal.⁴ Task shifting can broaden the provision of safe termination of pregnancy to the primary care level and thus has the potential to expand access to services.

To assess whether the provision of termination of pregnancy services by providers other than physicians could be performed with similar levels of safety, efficacy and acceptability as when performed by physicians, we undertook a systematic review of this topic.

Methods

The current review has been prepared in accordance with the PRISMA guidelines⁵ and included a review protocol.

Criteria for considering studies for this review

Criteria covered the areas of participants, interventions, comparisons, outcomes and study design (PICOS) as outlined below.

Types of study

Randomised controlled trials (RCTs) on this topic were anticipated to be scarce and many of the safety outcomes are rare. In order to increase the evidence base addressing this issue and to increase the generalisability to various settings, we included clinical studies using the following study designs: RCTs, cluster RCTs, nonrandomised cluster-controlled trials, controlled before and after studies, interrupted time series⁶ and cohort studies.

Types of participant

Women undergoing medical or surgical termination of pregnancy or receiving treatment for miscarriage at any gestational age of pregnancy.

Types of intervention

Termination of pregnancy or management of a miscarriage provided by either physician providers or MLPs. MLP for this review is defined as any formally trained nonphysician provider, and comprises a heterogeneous group depending on the cadre of provider, country and setting.

Types of outcome measure

The primary outcomes were: (1) efficacy: complete termination of pregnancy; correct assessment of gestational age; (2) safety: complications, their recognition and management. The secondary outcomes were: (1) training needed for competency; (2) acceptability and satisfaction; (3) costs.

Search methods for the identification of studies

Electronic searches

An electronic literature search of the Cochrane Central Register of Controlled Trials (third quarter, 2011), MEDLINE (1946 to October 2011), MEDLINE in process, other nonindexed citations (6 October 2011) and POPLINE (1927 to October 2011) without language preference and using terms that include abortion, termination of pregnancy services and healthcare providers.

Searching other sources

We contacted researchers in the field to seek information on unpublished or ongoing studies that we might have missed or that may have been unavailable with the electronic search. Reference lists of articles retrieved were hand searched.

Data collection and analysis

Selection of studies

The primary reviewer (RMR) evaluated the articles identified from the literature search. A second reviewer (NK) evaluated the articles in question for inclusion and, where needed, additional information was sought from the study’s author. Studies under consideration were evaluated based on methodology without consideration of the results. We included studies examining manual vacuum aspiration, electric suction aspiration or dilation and curettage. We considered as medical termination of pregnancy those regimens that used mifepristone/misoprostol, misoprostol alone or methotrexate/misoprostol.

Data extraction and management

Two reviewers (RMR and NK) independently extracted data using standard extraction forms. Data collected included PICOS, funding information and, if applicable, randomisation, allocation and blinding and loss to follow-up. Participant characteristics included number, demographics and gestational age. Discrepancies were resolved by consensus of the reviewers.

Assessment of the risk of bias in the included studies

Two reviewers (RMR and NK) independently graded the quality of the included studies using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system that incorporates in its assessment the study design, limitations, inconsistency, indirectness and imprecision.⁷ The quality of evidence was classified into one of four categories: high (further research is very unlikely to change our confidence in the estimates of the effects), moderate (further research is likely to greatly affect our confidence in the estimates of the effects and may change the estimates), low (further research is very likely to greatly affect confidence in the estimates of the effects and is likely to change the estimates) or very low (any estimate of effect is very uncertain).

For RCTs, we assessed bias by:

• 1

  Assigning a quality score for concealment of allocation:⁸

  • (a)

    adequate concealment of allocation;

  • (b)

    unclear whether adequate concealment of allocation;

  • (c)

    inadequate concealment of allocation;

  • (d)

    allocation concealment not used.

• 2

  Blinding of participants, clinicians and investigators.

• 3

  Protection against exclusion bias.

• 4

  Appropriate analysis of data.

Any discrepancies between reviewers were addressed by a joint re-evaluation of the original article. Funnel plots were used to assess publication bias.

Measures of treatment effect and assessment of heterogeneity

We used RevMan5.1 software for data analysis. Effects of dichotomous outcomes are presented as risk ratios with a 95% confidence interval, and were considered to be statistically significant if the confidence interval did not include 1.0. We examined for heterogeneity using the I² statistic.⁹ If I² > 75% (considerable heterogeneity), we used a random effects model in the analysis.

Dealing with missing data

For RCTs, we aimed to use an intention-to-treat analysis. However, in one study,¹⁰ the need for post-randomisation exclusion of women who turned out to be ineligible was part of the study design. We excluded these women from the analysis, but included women who were lost to follow-up after undergoing a termination of pregnancy. Depending on the study and outcome, the denominator of the cohort studies included women who underwent the intervention or only those who were followed up. We did not impute any outcomes, and thereby may have slightly mis-estimated the true proportion of complications. Loss to follow-up was considered in the GRADE assessment.

Data synthesis

Using RevMan5.1 software for data analysis, and consistent with our protocol, we divided the studies into two comparisons.

• 1

  Comparison 1: surgical termination of pregnancy.

• 2

  Comparison 2: medical termination of pregnancy.

Within each comparison, RCTs and prospective cohort studies were considered separately. The only RCT on surgical termination of pregnancy showed significant heterogeneity between the two countries in which it was performed. Therefore, we combined them in a meta-analysis using a random effects model. We summarised but did not perform meta-analyses of the cohort studies because of their heterogeneity and other limitations.

Results

Results of the search

Our search strategy (Appendix S1) identified a total of 806 articles (Figure 1).

Figure 1

Study selection.

Studies were excluded after screening the title and abstract or the full-text article (n = 41) based on study design, lacking a comparison group or not assessing the outcomes of interest. Ultimately, we identified five studies that met all inclusion criteria. Of note, we did not identify studies comparing the provision of active miscarriage management or costs of services. Each study investigated only one method of termination of pregnancy; there were no identified studies in which providers chose the method.

Included studies

Five studies with a total of 8908 subjects were included in this review with the following interventions and comparison groups.

Comparison 1: surgical termination of pregnancy. One RCT¹¹ (2894 women) compared MLPs, mostly midwives and one doctor assistant, with physicians performing surgical termination of pregnancy up to 12 weeks at various clinics in South Africa (mean gestational age, 7.7 weeks) and Vietnam (mean gestational age, 6.1 weeks). Three prospective cohort studies (4910 women) compared MLPs – nurses in India¹² (1089 women) and physician assistants in two US studies^13,14 (3821 women) – with physicians. In the Indian clinics, women with pregnancies up to 10 weeks of gestation (mean gestational age, 8.6 weeks) and, in the USA, up to 15 weeks of gestation (>80% <10 weeks) were included. All studies were conducted in a setting in which emergency care in the case of complications was accessible.

Comparison 2: medical termination of pregnancy. One RCT¹⁰ (1104 women) compared nurses and auxiliary nurse midwives with physicians in several Nepalese hospitals using 200 mg mifepristone, followed by 800 μg vaginal misoprostol 2 days later. Women with pregnancies up to 9 weeks (mean, 6.6 weeks) were included. The study required that the women resided within 90 minutes of the study site to guarantee access to emergency care if needed.

The Indian study¹² compared newly trained MLPs with newly trained physicians. The other studies included previously trained pregnancy termination providers;^10,11,13,14 however, in Warriner et al.,¹¹ physicians had more years of experience than did MLPs. Table 1 presents the characteristics of the included studies.

Table 1. Characteristics and findings of the included studies

Reference	Study design	Participants	Country; setting; time of study or data collection	Intervention	Critical outcomes	Number of women analysed: n (%)		Effect: RR (95% CI)	Comments
						Mid-level providers (MLPs)	Physicians	MLP versus physician
Surgical termination of pregnancy (TOP)
Warriner et al.^11	Randomised controlled trial, equivalence trial; computer-generated randomisation, sealed opaque envelopes for allocation concealment, no blinding	1160 South African and 1734 Vietnamese women, surgical TOP up to 12 weeks of gestation	South Africa and Vietnam; multicentre (8 urban and peri-urban Marie Stopes International); September 2003–June 2004 South Africa, May–December 2003 Vietnam	1) MLP 2) Physician; manual vacuum aspiration (MVA); 2-week follow-up	Overall complications	18/1453 (1.2)	10/1441 (0.69)	3.1 (0.2–59.3)	Provider experience (MLP/physician): South Africa 4/7 years; Vietnam 7/10 years. 25 providers; MLP: 6 South Africa (2 midwives), 8 Vietnam (1 doctor assistant, 7 midwives). Undergo standardised and government-accredited training in abortion; doctors: 6 South Africa, 5 Vietnam
					Incomplete TOP	16/1453 (1.1)	8/1441 (0.56)	3.0 (0.2–42.0)
					Haemorrhage, injury to uterus or cervix	0/1453 (0)	0/1441 (0.0)	No estimate
					Infection	2/1453 (0.1)	1/1331 (0.1)	1.7 (0.2–12.5)
Freedman et al.^13	Prospective cohort study (next available provider)	2458 women, surgical TOP up to 12 weeks of gestation	USA; freestanding clinic in Vermont; January 1981–December 1982	1) Physician assistant (PA) 2) Physician (next available provider); MVA or electric vacuum aspiration (EVA); 4-week follow-up	Overall complications	35/1285 (2.7)	36/1173 (3.1)	0.9 (0.6–1.4)	Experienced providers; PAs undergo structures training and a national certification examination, they are state licensed to practice with physician supervision; trainees were excluded; only 94% of all women undergoing a procedure participated in the study (no explanation). Some outcomes are not broken down by provider type
Goldman et al.^14	Prospective cohort study (PAs and physicians were in different clinics)	1363 women, surgical TOP up to 14 weeks of gestation	USA; 2 freestanding clinics, 1 in New Hampshire (physician providers) and 1 in Vermont (PA providers); July 1996–October 1997	) PA 2) Physician; MVA or EVA; 2-week follow-up	Overall complications	12/546 (2.2)	19/817 (2.33)	1.0 (0.5–1.9)	Experienced providers; PAs undergo structures training and a national certification examination, they are state licensed to practice with physician supervision; 3 PAs and 3 physicians all with minimum of 5 years of experience in surgical abortion; PAs gave routine postoperative 14 days of antibiotics, physicians not; 28.5% loss to follow-up
					Ongoing pregnancy	1/546 (0.2)	0/817 (0.0)	4.5 (0.2–110.0)
					Incomplete TOP	7/546 (1.3)	3/817 (0.4)	3.5 (0.9–13.4)
					Haemorrhage	5/546 (0.9)	1/817 (0.1)	7.5 (0.9–63.9)
					Injury to uterus	1/546 (0.2)	0/817 (0.0)	4.5 (0.2–109.9)
					Injury to cervix	0/546 (0.0)	0/817 (0.0)	No estimate
					Infection	2/546 (0.4)	16/817 (2.0)	0.2 (0.04–0.8)
					Missed ectopic pregnancy	1/546 (0.2)	0/817 (0.0)	4.5 (0.2–110.0)
Jejeebhoy et al.^12	Prospective cohort study (just one provider type was assigned to each facility at a given point in time)	1089 women, surgical abortion up to 10 weeks of gestation	India (Bihar and Jhakhand); five clinics of the NGO Janani; July 2009–January 2010	1) Nurses 2) Physicians; clinical exam; MVA; 1-week follow-up with pelvic exam	Overall complications	6/433 (1.4)	4/432 (0.9)	1.5 (0.4–5.3)	Newly trained providers; a ‘verifier’ repeated all pelvic exams and confirmed appropriate inclusion of woman
					Inaccurate assessment of gestational age	21/555 (3.8)	14/534 (2.6)	1.4 (0.7–2.8)
					Ongoing pregnancy; haemorrhage; injury to uterus or cervix, infection, missed ectopic pregnancy	0/433 (0.0)	0/432 (0.0)	No estimate
					Incomplete TOP	5/433 (1.2)	4/432 (0.9)	1.3 (0.3–4.6)
					Failure to assess complete TOP	3/433 (0.7)	3/432 (0.7)	1.0 (0.2–4.9)
					Overall acceptability	425/433 (98.2)	422/432 (97.7)	1.0 (0.99–1.02)
					Acceptability to women of counselling about contraception	433/433 (100)	431/432 (99.8)	1.00 (1.00–1.01)
Medical termination of pregnancy
Warriner et al.^10	Randomised controlled equivalence trial; computer-generated randomisation, sealed opaque envelopes for allocation concealment, no blinding	1104 women, medical abortion up to 9 weeks (63 days) of gestation, no more than 90-minute journey from the study clinic	Nepal; multicentre (5 district hospitals); April 2009–March 2010	1) MLP 2) Physician;counselling; clinical exam; medical abortion with 200 mg oral mifepristone followed by 800 μg vaginal misoprostol 2 days later; 3-hour hospital monitoring after misoprostol; 2-week follow-up	Serious complication requiring blood transfusion or hospitalisation	0/0 (0.0)	0/0 (0.0)	No estimate	Newly trained providers (3-day training course); all providers MVA trained. MLP: 8 staff nurses (3-year degree), 3 auxiliary nurse midwives (18 months of training) all MVA trained; doctors: 6 OB&GYN, 3 GP, 5 doctors (with Bachelor of Medicine and Bachelor of Surgery degree)
					Ongoing pregnancy	0/542 (0.0)	5/535 (0.9)	0.1 (0.0–1.6)
					Incomplete TOP	14/542 (2.6)	15/535 (2.8)	0.9 (0.5–1.9)
					Haemorrhage	0/542 (0.0)	0/535 (0.0)	No estimate

Quality of evidence

The quality of evidence of all included studies was graded using the GRADE system.⁷ A panel of international experts reviewed and prioritised our proposed outcomes of interest, and the critical outcomes are summarised in Table 1 (for complete GRADE tables, see Tables S1 and S2). We found no serious inconsistency or indirectness among the studies. RCTs were rated as low to moderate quality, whereas prospective cohort studies ranged from very low to low quality for the outcomes of interest. The most frequent reason for downgrading the evidence was imprecision because of a wide confidence interval (related to the small number of studies or to the heterogeneity of the results¹¹) and small number of events (related to the rare occurrence of complications). No studies met the criteria for upgrading. In cohort studies, high loss to follow-up also occurred.

Risk of bias in the included studies

Inclusion and exclusion criteria were clearly defined^10,11 for protection against exclusion bias.

Incomplete data occurred in 3.6% of women with surgical termination of pregnancy,¹¹ either because they withdrew before the procedure or because they were lost to follow-up after the procedure. In the RCT on medical termination of pregnancy by Warriner et al.,¹⁰ 6.3% of women had incomplete data (including the women who were excluded post-randomisation as they were found to be ineligible based on the provider examination).

We did not find evidence of reporting bias. According to the funnel plots, studies included in this review are symmetrically distributed, indicating that publication bias is unlikely. Published results corroborate observational study results, such as programme evaluations on the safety and efficacy of MLPs providing termination of pregnancy services.

Effects of intervention

Results are stratified by surgical versus medical termination of pregnancy. All effect sizes are presented in Table 1 and in the forest plot of overall complications (Figure 2).

Figure 2

Forest plot of overall complications in surgical termination of pregnancy.

Comparison 1: surgical termination of pregnancy

RCT

None of the safety and efficacy outcomes (overall complications, incomplete abortion necessitating re-aspiration, haemorrhage (>500 ml), injury to the uterus, injury to the cervix or infection) occurred frequently or showed a statistically significant difference between provider groups (Table 1). Among the 2894 women randomised to MLPs (mostly midwives or doctor assistants) and physicians, 18 and 10, respectively, experienced complications [risk ratio (RR), 3.1; 95% confidence interval (95% CI), 0.2–59.3] (Table 1, Figure 2].¹¹

Prospective cohort studies

Overall complications, ongoing pregnancy, incomplete termination of pregnancy, haemorrhage, injury to the uterus, injury to the cervix and missed ectopic pregnancy showed no differences between provider groups in all studies (Table 1). Overall complications occurred in 112 (53 in the nurses and physician assistant group, 59 in the physician group) of 4686 women.^12–14 Only infection (18 of 1363 women) was less common among procedures performed by physician assistants than physicians (two versus 16).¹⁴ In addition, one study compared the assessment of gestational age and completion of termination of pregnancy, both via bimanual examination by an experienced physician, who repeated the examination to verify the findings of the MLP, which did not differ between provider groups.¹²

Comparison 2: medical termination of pregnancy

RCT

Among 1077 women, none experienced a serious complication (haemorrhage requiring blood transfusion or condition necessitating hospitalisation).¹⁰ None of the safety and efficacy outcomes [serious complication, ongoing pregnancy or incomplete abortion (retained products of conception possibly requiring aspiration)] showed a significant difference between provider types (Table 1). Of note, there were five ongoing pregnancies, all of which were in the physician group, and 29 incomplete abortions (14 in the MLP group [staff nurses and auxiliary nurse midwives] and 15 in the physician group). MLPs chose to discuss <2% of cases with a physician, and referred <1% for termination of pregnancy care elsewhere.

Discussion

Summary of main results

Two RCTs and three prospective cohort studies were included in this review which compared various formally trained provider groups with physicians providing surgical^11–14 and medical¹⁰ termination of pregnancy.

Comparison 1: surgical termination of pregnancy

The RCT¹¹ and prospective cohort studies^12–14 showed that various MLP groups (nurses, midwives, doctor assistants, physician assistants) have effectively and safely provided surgical termination of pregnancy using manual or electric vacuum aspiration in clinic settings in South Africa, Vietnam, India and the USA. Overall complications in the RCT¹¹ resulted in RR = 3, favouring physicians, but the difference was not significant. The experience of physicians – 7 and 10 years in South Africa and Vietnam, respectively – compared with that of MLPs – 4 and 7 years, respectively, may have contributed to this finding, as a lack of experience is a known risk factor for complications.¹⁵ The heterogeneity of the data is probably related to varying provider experience and the rarity of outcomes.

Complications were rare in all included studies; infection was the only complication that differed between provider groups, favouring the MLP group, in one study.¹⁴ This result is confounded by different practices in antibiotic prophylaxis: MLP patients routinely received 1 week of antibiotics, whereas physicians’ patients received either the antibiotic medication or only a prescription.

Comparison 2: medical termination of pregnancy

One RCT¹⁰ did not identify any significant differences between efficacy, safety outcomes or women’s satisfaction when comparing nurses and auxiliary nurse midwives with physicians using mifepristone and misoprostol.

Overall completeness and acceptability of the evidence

Studies have been conducted in various countries and settings: developing and developed countries, clinics and hospitals. This is a strength of the data, as it increases the generalisability, as does the inclusion of various groups of MLPs, including nurses, midwives, auxiliary nurse midwives and physician assistants.

Limitations of the evidence include the scarcity of RCTs, making the data imprecise. Data of prospective cohort studies are limited by imprecision, as well as by biases inherent to the study design. Studies on the surgical termination of pregnancy included women up to 15 weeks of gestation; however, women actually recruited had an average gestational age of 7 weeks (RCT) and <10 weeks (cohort studies), leaving few data for the later first trimester. Mortality has been shown to increase with increasing gestational age, particularly above 8 weeks.¹⁶ Women undergoing surgical termination of pregnancy had access to care for complications and women undergoing medical termination of pregnancy resided within 90 minutes of the study site. The extrapolation of these data to care in the second trimester (medical or surgical) or to settings without access and referral for emergency care options may not be appropriate. No comparative study assessed active management of miscarriage.

Agreement or disagreement with other studies or reviews

The incidence of complications observed in the included studies is uncommon and consistent with other publications.^15–17 MLPs have been evaluated in cross-sectional studies and programme evaluations, usually without direct comparison with physicians. Although the evidence of these reports is subject to even more bias and of lower quality than the evidence included in this review, our findings are consistent with the conclusion that trained MLPs can safely provide surgical and medical termination of pregnancy. Providers used manual or electric vacuum aspiration (as opposed to sharp curettage) as is recommended in the literature.^18,19 Access to emergency care was available in all studies, and has been generally recommended for termination of pregnancy services.^15,17,19

Conclusions

Implications for practice

Limited evidence suggests that MLPs can be trained to provide first-trimester surgical (using vacuum aspiration) and medical (mifepristone and misoprostol) termination of pregnancy services as safely and effectively as physicians. A learning curve probably affects complication risk, particularly in the provision of surgical services.

Implications for research

This review highlights the importance of further studies to address task sharing in termination of pregnancy care, such as the safety of providing surgical termination of pregnancy beyond 10 weeks of gestation, the use of other medical termination of pregnancy methods in the first or second trimesters, the safety and efficacy of post-termination of pregnancy care and management of miscarriage, implications for training for competency for different groups of providers and the impact of increasing access to safe termination of pregnancy services.

Disclosure of interests

R-MR, DB and NK have nothing to disclose.

Contribution to authorship

R-MR developed the protocol for this review and conducted all steps of the review, including the development of the search strategy, identification of the studies for inclusion, data extraction, analysis and interpretation, as well as writing the manuscript.

DB contributed to data analysis and interpretation, as well as to the manuscript, during her time at the Department of Reproductive Health and Research, World Health Organization.

NK also contributed to all aspects of this review. She was the second author who independently extracted the data, contributed to the analyses and their interpretation, as well as to the manuscript.

Details of ethics approval

Ethics review and approval were not necessary for the appropriate conduct of this systematic review.

Funding

Resources from the Department of Reproductive Health and Research at the World Health Organization and from the Oregon Health & Science Family Planning Fellowship Fund supported the staff time for the completion of this review.

Acknowledgements

We thank Metin Gulmezoglu and Marit Johansen for support in the development of the search strategy.