Severity and management of postabortion complications among women in Zimbabwe, 2016: a cross-sectional study.

OBJECTIVES: Abortion complications cause significant morbidity and mortality. We aimed to assess the severity and factors associated with abortion complications (induced or spontaneous), and the management of postabortion care (PAC) in Zimbabwe.
DESIGN: Prospective, facility-based 28 day survey among women seeking PAC and their providers.
SETTING: 127 facilities in Zimbabwe with the capacity to provide PAC, including all central and provincial hospitals, and a sample of primary health centres (30%), district/general/mission hospitals (52%), private (77%) and non-governmental organisation (NGO) (68%) facilities. PARTICIPANTS: 1002 women presenting with abortion complications during the study period. MAIN OUTCOME MEASURES: Severity of abortion complications and associated factors, delays in care seeking, and clinical management of complications.
RESULTS: Overall, 59% of women had complications classified as mild, 19% as moderate, 19% as severe, 3% as near miss and 0.2% died. A median of 47 hours elapsed between experiencing complication and receiving treatment; many delays were due to a lack of finances. Women who were rural, younger, not in union, less educated, at later gestational ages or who had more children were significantly more likely to have higher severity complications. Most women were treated by doctors (91%). The main management procedure used was dilatation and curettage/dilatation and evacuation (75%), while 12% had manual vacuum aspiration (MVA) or electrical vacuum aspiration and 11% were managed with misoprostol. At discharge, providers reported that 43% of women received modern contraception.
CONCLUSION: Zimbabwean women experience considerable abortion-related morbidity, particularly young, rural or less educated women. Abortion-related morbidity and concomitant mortality could be reduced in Zimbabwe by liberalising the abortion law, providing PAC in primary health centres, and training nurses to use medical evacuation with misoprostol and MVA. Regular in-service training on PAC guidelines with follow-up audits are needed to ensure compliance and availability of equipment, supplies and trained staff. © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2018. All rights reserved. No commercial use is permitted unless otherwise expressly granted.

This nationally representative study covered all provinces in Zimbabwe, included all central and provincial hospitals which have a high postabortion care (PAC) caseload, and avoided concerns encountered in retrospective studies assessing abortion-related morbidity due to missing patient records.

Our revised morbidity criteria reduce potential overestimation of severity by removing unreliable stand-alone criteria such as fever and tachycardia.

We were unable to distinguish between induced and spontaneous abortions.

Information on women with mild complications that resolved spontaneously, severe complications leading to death outside the facility, or any other case of PAC occurring outside of a facility were not captured.

Introduction

Unsafe abortion remains an important cause of maternal morbidity and mortality.1 Globally, between 2010 and 2014, about 25.1 million unsafe abortions occurred annually, largely (97%) in low/middle-income countries.2 Nearly 22 000 women died due to unsafe abortions in 2014,3 and many more suffered serious injuries. Approximately 12% of maternal deaths globally are attributed to abortion (this includes ectopic pregnancies).4 Of the estimated 6.2 million unsafe abortions in Africa yearly, one-third occur in Eastern Africa, where Zimbabwe is located.2

In Zimbabwe, abortion is highly restricted and permitted only in cases of rape, incest, when the mother’s life is at risk, or when the child may be born with serious mental or physical disabilities.5 Restrictive abortion laws are not associated with lower levels of abortion,6 but are associated with increased abortion-related morbidity and mortality. Zimbabwe failed to meet the 2015 Millennium Development Goal of reducing the maternal mortality ratio (MMR) by 75%. In fact, while most countries experienced declines in maternal mortality,7 Zimbabwe’s MMR increased from 450 per 100 000 live births in 19908 to 651 in 2015.9 Estimates of maternal mortality attributable to abortion complications in Zimbabwe range from 6% to 23%, although these estimates come from older studies with methodological limitations.10 11

One approach to reducing abortion-related morbidity and mortality involves improving access to and quality of postabortion care (PAC); such efforts are ongoing in Zimbabwe. National guidelines for comprehensive PAC have been in place since 2001, and were updated in 2014.12 These guidelines emphasise medical management of abortion complications with misoprostol, preference for manual vacuum aspiration (MVA) over dilatation and curettage (D&C) for first trimester abortions, and provision of family planning services.12 However, in many settings, health practitioners do not always adopt or maintain use of efficacious, cost-effective innovations.13 Furthermore, during the past decade, Zimbabwe has undergone economic stagnation,14 potentially affecting health delivery systems, including PAC provision. Access to PAC may also be limited due to stigma, costs and other factors leading to delays in seeking care.15 No recent studies have examined postabortion complications in Zimbabwe. We conducted a national survey to assess the severity and management of postabortion complications, and to understand the factors associated with experiencing severe complications.

Methods

We employed the Prospective Morbidity Methodology to collect information from PAC patients and their providers on complications from spontaneous and induced abortions treated in a health facility. This methodology was developed by WHO,16 modified by Ipas,17–21 and further refined by the Guttmacher Institute.22 23 We conducted this study in conjunction with a project estimating the incidence of induced abortion in Zimbabwe; those methods and results are provided elsewhere.24

Data collection

We conducted a facility-based, prospective survey for 28 days between August and September 2016 among women seeking PAC in Zimbabwe. We compiled a comprehensive list of the 245 facilities in Zimbabwe with the capacity to provide PAC24 (online supplementary table 1). We sampled facilities stratified by province and level, selecting all central (n=5) and provincial (n=8) hospitals, and identified a random sample of primary health centres (30%), district/general/mission hospitals (52%), private facilities (77%) and NGO facilities (68%; includes for-profit and not-for-profit facilities). Overall, we selected 133 facilities, of which 127 participated, resulting in a facility-level response rate of 95%.

All women presenting with incomplete, inevitable, missed, complete or septic abortion during the study period were eligible for inclusion. We could not distinguish between induced or spontaneous abortions, which are often clinically indistinguishable to providers. Furthermore, women often under-report induced abortion due to stigma and fear of being reported to police; for example, 52% of adolescent girls in Zimbabwe believe an unmarried woman seeking treatment in public facilities for postabortion complications will be reported to police.25 Since information reported by both women and providers may not reliably distinguish between induced and spontaneous abortions, results presented are for all PAC patients. We note that complications from induced abortions may be more severe than those from spontaneous abortions.26

Data were collected by one to two nurses in each facility who coordinated with facility PAC providers to track participants. Once the patient was treated and in a stable condition, the interviewer sought informed consent to conduct a face-to-face interview with her, as well as her consent to separately interview her healthcare provider about her case and review her medical file. All women seeking PAC in each facility were recorded in a tracking form, including women who were near misses and too ill to be interviewed and women who died before being interviewed. Among an unweighted total of 1018 eligible patients, 1002 were interviewed (98%) and 986 consented for the provider interview (97%) (online supplementary table 1). Study staff and Ministry of Health and Child Care provincial Reproductive Health Officers supervised data collection.

Key variables

Outcome: postabortion complication severity

We developed a five-level classification system of postabortion complication severity: mild, moderate, severe, near miss or death (box 1). These classifications were adapted from the original criteria proposed by Rees et al which has been used in prior studies.18 20–22 27 We expanded our criteria to include the adapted WHO near-miss criteria28 for a low/middle-income country context.29 30 Near-miss cases have similar morbidities to cases that result in death but occur more frequently and survive because of the treatment they receive. They are therefore useful to assess quality of care for abortion-related emergencies and to understand circumstances around abortion-related deaths.31 These modifications aimed to improve the objectivity of the clinical criteria and overall reliability and content validity. For the severe category and above, we shifted away from standalone clinical signs (eg, fever and tachycardia) which may lead to overestimation of severity. Furthermore, we removed ‘evidence of a foreign body’ as a sole criterion for severe complications, as this may not indicate severe morbidity, and is based on subjective provider reports, which may be affected by provider stigma and restrictive abortion laws. Severity classifications were mutually exclusive, and women were classified into the highest level of severity for which they met the criteria.

Mild morbidity (requires all criteria)

Temperature 35.1°C–38.9°C with no clinical signs of infection*

No system or organ failure†

Systolic blood pressure ≥90 mm Hg

Haemorrhage not requiring any transfusion

Moderate morbidity (requires ≥1 criterion)

Temperature 37.3°C–38.9°C

Clinical signs of infection*

No organ or system failure†

No sign of shock‡

Haemorrhage not requiring any transfusion

Severe morbidity (requires ≥1 criterion)

Temperature ≥39°C or ≤35°C and a clinical sign of infection§

Sepsis/septicaemia with no signs of septic shock‡

Pelvic abscess or pelvic peritonitis with no signs of shock‡

Clinical anaemia without haemorrhagic shock‡

Uterine perforation without laparotomy or repair of perforated uterus, repair of gut perforation, hysterectomy

Near-miss (requires ≥1 criterion)

Haemorrhagic shock‡

Septic shock‡

Generalised peritonitis

Uterine perforation with laparotomy or repair of uterine perforation, repair of gut perforation or hysterectomy

Organ/system failure†

Massive blood transfusion¶

Death

Loss of the life of a woman as a result of an abortion complication

Severity classifications were mutually exclusive. Due to data quality concerns with temperature recordings not aligning with expected clinical symptoms, we expanded the definition of normal temperature (35.1°C–37.2°C) to more accurately capture very low temperatures, which are a sign of shock or infection. To be classified as having severe morbidity, a patient had to have a very low or very high temperature along with a clinical sign of infection, or any of the other criteria. This prevents patients with only recorded low temperature and no other symptoms from being inaccurately captured as a severe case. A normal temperature (ie, a mild morbidity category) was imputed for three cases with missing temperature and who also had no other clinical symptoms. There were four other cases that did not fall into any of the categories based on their clinical criteria. Three cases had low temperature (<35°C) but no other signs of complications so the medical doctors on the study team determined these cases should be classified as mild morbidity. One case had tachycardia, stayed in the hospital for more than 24 hours and was given oral and IV antibiotics but had normal temperature and no other complications. The medical doctors on the team determined this case should be classified as moderate morbidity.

*Clinical signs of infection can include: fever >37.3°C and abdominal/uterine tenderness with or without foul smelling vaginal discharge or pelvic abscess or pelvic peritonitis.

†System or organ failure can include: liver failure or renal failure or cardiac arrest/failure or respiratory distress syndrome or coma or disseminated intravascular coagulopathy.

‡Shock can manifest as: a persistent systolic blood pressure ≤80 mm Hg alone or a persistent systolic blood pressure ≤90 mm Hg with a pulse rate at least 120 bpm, and restlessness, reduced consciousness, cold clammy peripheries, requiring administration of IV fluids.

§For severe, the clinical sign of infection also includes sepsis or pelvic abscess or pelvic peritonitis, or uterine perforation.

¶Massive blood transfusion refers to replacement of ≥2 units of blood.

IV, intravenous.

Covariates

We collected information on sociodemographic and reproductive health-related characteristics and about various delays that may occur in obtaining care for postabortion complications, including reasons for those delays. Furthermore, we inquired about amount of income lost to the respondent or her household as a result of experiencing health problems (excluding the actual costs of treatment or transportation). Variables that merit additional explanation are described below.

Trimester

This was measured using clinician-estimated gestational age in weeks (first trimester: 1–12 weeks, second trimester: 13–27 weeks, third trimester: 28 weeks term). Under 5% of unweighted cases (47/1002) were missing the clinician’s estimate; for 36 we used women’s self-reported gestational age, and the remaining 11 were missing.

Wealth

We constructed wealth indicators among our participants comparable to the national-level wealth quintiles using data from the 2015 Zimbabwe Demographic and Health Survey (DHS).9 We performed principal components analysis using information on assets (eg, car, bicycle, refrigerator, etc) and household characteristics (type of water source, toilet facility and roofing material) available in both the DHS and our questionnaire. We generated factor weights for each asset or characteristic,32 and used them to calculate individual wealth scores for each person in the DHS. We split those wealth scores into relative quintiles (poorest, poor, medium, wealthy, wealthiest) and noted the cut points for each quintile. Next, we applied those factor weights to variables in our survey data, constructed individual wealth scores and classified women into a quintile, using the DHS-derived cut points.9 We conducted this procedure separately for urban women and for rural women, since wealth indicators vary substantially by place of residence. Therefore, our final wealth indicator defines each wealth quintile differently based on urban/rural residence.

Analysis

We performed all analyses in Stata V.14.1. We conducted descriptive analyses to determine frequencies for categorical variables and calculated medians (or means) for continuous variables. We applied facility-level weights and calculated standard errors taking into account the complex sample design, including adjusting for stratification by province and facility level, clustering of women at the facility level, and facility non-response, and applying a finite population correction.

To examine factors associated with increasing severity of postabortion complications, we conducted ordinal regression, using a three-level severity variable (collapsed from the original five levels to avoid estimation problems due to small cell sizes). The levels were defined as: (1) mild complications, (2) moderate complications or (3) severe complications including near miss or death. We assessed variables demonstrated in prior analyses to be significantly associated with abortion severity (urban/rural residence, marital status, educational level, pregnancy duration),27 and variables we hypothesised may be associated with severity (age, parity, facility level, wealth status, delays in access to care). Variables with a P≤0.25 in bivariate analysis were considered for inclusion in a multivariate model. We assessed for collinearity and confirmed that our model did not violate the proportional odds assumption, and used a planned backward block stepwise regression approach.33

Results

Women presenting with postabortion complications in our study ranged from 15 to 47 years old, with adolescents (ages 15–19) accounting for 12% of PAC patients (online supplementary table 2). Most were in union (80%), had partial or complete secondary schooling (71%) and were not formally employed (63%). More rural women were adolescents compared with urban women (17% rural vs 8% urban), less likely to have attended university (19% urban vs 6% rural) and more likely to not be formally employed (77% rural vs 54% urban). Women in our study were wealthier when compared against the national distribution of wealth, with 37% classified in the ‘wealthiest’ quintile, according to place of residence. This was more evident among rural women; nearly half (47%) of rural women were classified in the ‘wealthiest’ quintile. Women reported that most pregnancies resulting in complications for which they were seeking care were wanted at the time of pregnancy (70%), while 15% were wanted later and 15% were not wanted. Most women (65%) were in the first trimester of pregnancy. The largest proportion of patients (40%) were managed in district, general or mission hospitals (24% urban vs 65% rural); an additional 29% were served in central hospitals.

Out of a weighted total of 1282 women, 59% had mild morbidity, 19% had moderate morbidity, 19% had severe morbidity, 3% were classified as near miss and 0.2% participated in the survey but later died (online supplementary figure 1). There were two additional deaths and 12 near misses recorded in facility tracking forms. Since no data could be collected on these patients for verification, these near misses and deaths are not included in this severity distribution.

Women reported that it took a median time of 47 hours from the time of experiencing complications until receiving complete treatment (table 1). This includes all health-seeking delays: realising care was needed, deciding to seek care, arriving at a facility, being attended to and completing treatment. The median self-reported delay in realising care was needed was 8 hours (range: <1–2688 hours). The median delay in deciding to seek care after realising it was needed was 2 hours (range: <1–1008). The most common reasons for this delay included lack of money (41%), partner or family member making the decision (13%), lack of transportation (13%) or distance to the facility (12%). The median delay in arriving to a health facility after deciding to seek care was 1 hour (range: <1–672). Women who received care in primary health centres reported the longest median delay (4 hours) in arriving to a health facility. The most common reasons for this delay included lack of money (63%), lack of transportation (24%) or distance to facility (16%). All women attending primary health centres reported being delayed in arriving at the health centre due to a lack of money. The median delay in being attended to after arriving at a health facility was 0.5 hours (range: <1–504), generally due to non-availability of a nurse or doctor (37%) especially in the central/provincial and district hospitals (37% and 41%, respectively). The longest overall delay (median 11 hours, range: <1–840 hours) occurred between being attended to and receiving complete treatment, with the longest delays in district hospitals (median 17 hours) and the shortest delays in primary health centres (median 1 hour). Nearly half (49%) sought care elsewhere before arriving at the current facility, with the majority of these women seeking but not receiving complete care from primary health centres (61%). Among those reporting lost income, the average loss (excluding costs of treatment or travel) was US $90.82. This average was higher in private and NGO facilities (US$280.27) and lowest in primary health centres (US$41). Average lost income was higher for urban (US$104.04) versus rural women (US$76.71).

Table 1

Experiences related to seeking postabortion care, Prospective Morbidity Survey

	Total		Facility type where care was ultimately received				P value
	Weighted N	Hours	Central and provincial hospitals	District hospitals	Primary health centre	Private and NGO facilities
Time delays in seeking care (hours)*
Median delay in realising healthcare was needed	1210	8	11	5	2	24	0.00
Minimum		0	0	0	0	0
Maximum		2688	2688	1344	168	672
Median delay in decision to seek care after realising care was needed	1221	2	2	2	2	6	0.06
Minimum		0	0	0	0	0
Maximum		1008	1008	672	168	672
Median delay in arriving to initial health facility after deciding to seek care†	1216	1	1	2	4	1	0.98
Minimum		0	0	0	0	0
Maximum		672	504	336	168	672
Median delay in being attended to after arriving at health facility	1193	1	1	1	0.1	0.3	0.33
Minimum		0	0	0	0	0
Maximum		504	504	96	7	24
Median delay in receiving complete treatment after being first attended to	1206	11	12	17	1	2	0.72
Minimum		0	0	0	0	0
Maximum		840	840	168	2	504
Reasons for delay in deciding to seek care after realising healthcare was needed‡	Weighted N	%
Did not have money	77	41	56	34	–	16	0.00
Partner or family member decides	24	13	13	16	–	4	0.05
Lack of transportation	24	13	10	19	–	0	0.01
Distance to facility	22	12	9	18	–	0	0.02
Reasons for delay in arriving to health facility§	Weighted N	%
Did not have money	41	63	55	63	100	67	0.21
Lack of transportation	15	24	18	47	0	0	0.05
Distance to facility	10	16	5	38	0	0	0.04
Reasons for delay in receiving care at facility¶	Weighted N	%
No doctor or nurse available	105	37	37	41	0	10	0.01
Many patients in line for care	82	29	31	23	100	42	0.04
Did not have money	40	14	14	16	0	6	0.35
Sought care at another facility prior to receiving care at current facility	Weighted N	%
	636	49%	58%	46%	15%	37%	0.00
Of those who reported lost income, amount of income lost to household due to health problems (does not include costs of treatment or travel) (in USD)**	Weighted N	USD
Average loss of income for all respondents	320	90.82	60.16	94.54	41.00	280.27	0.06
Average loss of income for urban respondents	165	104.04	66.02	97.36	–	318.25	0.00
Average loss of income for rural respondents	155	76.71	47.23	93.16	41.00	13.12	0.10

*These are self-reported time of delays from respondents. Two respondents had total delays greater than their estimated gestational age, and therefore were set to missing.

†This question asks for their initial visit to a health facility, not necessarily the facility where they were interviewed/received treatment. It can be interpreted as time from decision to seek care to access to healthcare system.

‡Out of 186 respondents. These are multiple response questions so they will not add up to 100 and the table does not present all responses. There was no data for individuals who sought care at primary health centres regarding reasons for delay in deciding to seek care after realising healthcare was needed.

§Out of 65 respondents. These are multiple response questions so they will not add up to 100 and the table does not present all responses.

¶Out of 120 respondents who said it took longer than a reasonable time to be attended to at facility. These are multiple response questions so they will not add up to 100 and the table does not present all responses.

**75% of respondents did not report a loss of income and are therefore not included in this total. If these respondents are imputed as US$0 for loss of income, the overall average loss of income is US$22.59.

NGO, non-governmental organisation.

In multivariate analysis, rural women had 122% higher odds (adjusted OR (adjOR) 2.22, 95% CI1.70 to 2.91) of severe morbidity (vs low or moderate morbidity), or of moderate/severe morbidity (vs low morbidity), holding other factors in the model constant (table 2). In other words, a rural woman had over two times the odds of increasingly severe morbidity from complications of abortion versus a similar urban woman. Women older than 30 were significantly less likely (27% lower odds) than women aged 15–19 to have increasingly severe morbidity (adjOR 0.73, 95% CI 0.55 to 0.98). Women not in union had 63% higher odds of increasingly severe morbidity compared with women in union (adjOR 1.63, 95% CI 1.29 to 2.04). University education was protective, conferring 54% lower odds of increasingly severe morbidity as compared with having no schooling or any primary education (adjOR 0.45, 95% CI 0.31 to 0.65). Having children increased the odds of experiencing increasingly severe morbidity by 68% (1–2 children adjOR 1.68, 95% CI 1.33 to 2.13; 3+ children adjOR 1.68, 95% CI 1.13 to 2.49). Women in their second trimester of pregnancy had 31% higher odds of increasingly severe morbidity compared with women in their first trimester (adjOR 1.31, 95% CI 1.01 to 1.71).

Table 2

Crude and adjusted ORs* (and 95% CIs) for the relationship between sociodemographic or abortion-related characteristics and severity of abortion complications, among women receiving postabortion care, Prospective Morbidity Survey

Characteristic	Crude OR (95% CI)	P value	Adjusted OR† (95% CI)	P value
Residence
Urban	1.00 (reference)		1.00 (reference)
Rural	2.33 (1.82 to 3.00)	0.00	2.22 (1.70 to 2.91)	0.00
Age
15–19	1.00 (reference)		1.00 (reference)
20–29	0.92 (0.69 to 1.22)	0.55	0.93 (0.73 to 1.18)	0.52
30+	0.72 (0.56 to 0.94)	0.01	0.73 (0.55 to 0.98)	0.04
Marital status‡
In union	1.00 (reference)		1.00 (reference)
Not in union	1.30 (1.03 to 1.64)	0.03	1.63 (1.29 to 2.04)	0.00
Educational level
None or any primary	1.00 (reference)		1.00 (reference)
Any secondary schooling	0.67 (0.51 to 0.88)	0.01	0.94 (0.72 to 1.24)	0.67
University or more	0.25 (0.18 to 0.36)	0.00	0.45 (0.31 to 0.65)	0.00
No of living children
None	1.00 (reference)		1.00 (reference)
1–2	1.26 (1.01 to 1.57)	0.04	1.68 (1.33 to 2.13)	0.00
3+	1.37 (1.04 to 1.80)	0.03	1.68 (1.13 to 2.49)	0.01
Estimated gestational age
First trimester	1.00 (reference)		1.00 (reference)
Second trimester	1.42 (1.08 to 1.87)	0.01	1.31 (1.01 to 1.71)	0.04
Facility where postabortion care was received
Primary health centre	1.00 (reference)		–
District hospitals	1.50 (0.45 to 5.00)	0.50	–	–
Provincial and Central hospitals	0.72 (0.22 to 2.30)	0.57	–	–
Private and NGO facilities	0.49 (0.15 to 1.61)	0.24	–	–
Relative wealth quintile
Poorest	1.00 (reference)		1.00 (reference)
Poor	1.17 (0.86 to 1.59)	0.32	1.09 (0.83 to 1.43)	0.51
Medium	0.86 (0.60 to 1.24)	0.42	0.84 (0.59 to 1.20)	0.32
Wealthy	0.90 (0.67 to 1.21)	0.48	0.88 (0.65 to 1.20)	0.42
Wealthiest	0.84 (0.64 to 1.09)	0.19	0.80 (0.63 to 1.02)	0.08
Time between deciding to seek care and arrival at a facility§
No delay (wait <2 hours)	1.00 (reference)		–
Less than a day	1.28 (1.03 to 1.60)	0.03	–	–
1+ days	0.90 (0.57 to 1.43)	0.66	–	–

*Model is an ordinal logistic regression where the outcome is three levels: mild complications; moderate complications; severe complications or near miss or death.

†Bivariate and multivariate models are restricted to cases with no item non-response for any of the variables in the table. The weighted N for the multivariate model is 1232.

‡In union indicates currently married or living together; not in union indicates never married, with partner and not living together, or separated/divorced/widowed.

§Refers to the initial facility the respondent went to, which might not be the same facility where the respondent ultimately received care.

–Not included in final regression.

NGO, non-governmental organisation.

D&C/dilatation and evacuation (D&E) was the most common PAC procedure (75%), except in primary health centres (which do not have the capacity to provide this service) (table 3). Only 11% of clients had medical evacuation using misoprostol, although this was 100% among clients in primary health centres. Doctors performed the majority of procedures (91%) and most participants received antibiotics (97%), pain medication (78%) and intravenous (IV) fluids (67%). While 92% of patients were counselled about contraception at discharge, providers reported that 43% of participants received modern contraception on discharge, with larger proportions receiving methods in primary health centres and private or NGO facilities (61% in each), and smaller proportions receiving methods in central and provincial hospitals (26%).

Table 3

Treatment and services received by postabortion care clients, Prospective Morbidity Survey

	Total		Facility type				P value
	Weighted N	%	Central and provincial hospitals (%)	District hospitals (%)	Primary health centre (%)	Private and NGO facilities (%)
Patient stayed in facility >24 hours	577	46	43	61	0	20	0.00
Main procedure used in management of patient’s condition*							0.00
D&C/D&E†	760	75	74	78	0	67
Manual/electric vacuum  aspiration	125	12	17	6	0	14
Misoprostol	113	11	9	12	100	17
Oxytocin	21	2	0	4	0	2
Procedure performed primarily by:							0.00
Doctor‡	960	91	99	83	0	87
Nurse/midwife/ clinical  officer	93	9	1	17	100	13
Received intravenous fluids	848	67	70	71	36	46	0.01
Antibiotics provided	1232	97	96	99	85	98	0.01
Pain medication provided	960	78	79	75	64	91	0.08
Contraceptive services
Patient counselled on contraception at discharge							0.00
Yes	1154	92	88	96	100	91
No	76	6	10	2	0	8
Not discharged yet	25	2	2	2	0	1
Patient received modern contraception at discharge§							0.01
Yes	535	43	26	55	61	61
No	650	52	67	40	39	37
Don’t know	69	6	7	5	0	2

*Out of women who obtained procedures (weighted n=1018).

†D&C/D&E includes: dilatation and curettage (D&C) (12%), evacuation by sharp curettage (50.5%), digital evacuation (1.5%) and forceps evacuation (10.4%).

‡Includes Obstetrician and Gynaecologist (34%) and medical officers/GPs/senior resident medical officers (57%).

§The definition of ‘modern’ contraception was not included in the question; so it is possible that providers may have interpreted this term slightly variably.

D&C, Dilation and curettage, D&E, Dilatation and evacuation; GPs, general practitioners; NGO, non-governmental organisation.

Discussion

Main findings

About 40% of Zimbabwean women experiencing abortion complications are classified as having moderate or more severe complications. The proportion with severe or near-miss morbidity (21%) is similar to recent studies done in Malawi (21%)27 and in Kenya (37%).34 However, proportions of severe cases in those studies were likely overestimated with the older criteria, and therefore the proportion of severe/near-miss cases in Zimbabwe is high in comparison. We identified several characteristics associated with a greater likelihood of increasingly severe abortion complications, including being young, rural, not in union, less educated, having children or at a later gestational age. Our findings are similar to those from a study in Malawi, which also reported greater risk of abortion-related morbidity among rural women and those not in union.27 In Zimbabwe, PAC is not offered in most primary health centres, which are more accessible to rural women than higher level facilities. Expanding provision of comprehensive PAC services in rural areas, especially ensuring access for adolescents,9 may help address this inequity. Empowerment of young women, including through educational opportunities and health literacy, may play a role in maximising health and reducing unintended pregnancy and recourse to unsafe abortion.35

Financial constraints represented the most common reason for delays in care seeking. Despite the Ministry of Health and Child Care national policy that PAC should be free in public facilities; women still pay for transport, additional service fees and other expenses. The costs of seeking PAC may be inaccessible for many, and women in our sample (ie, those who successfully sought and received PAC) were wealthier than the national wealth distribution in Zimbabwe, suggesting potential selection of wealthier individuals into receipt of these services. Similarly, facility access for deliveries in Zimbabwe also increases progressively by wealth (61% for women in the lowest quintile; 95% for women in the highest).9 As lower income women are more likely to have less safe abortions than richer women, enhancing accessibility and affordability of PAC services for poorer women is essential to decrease maternal mortality. Women seeking care in primary health centres reported the longest median delay in arriving at a facility; these facilities tend to be more remote, particularly in rural areas, and most likely require that women have funding for and access to transportation. The median delay to receiving complete treatment (11 hours) could be considerably reduced if medical evacuation with misoprostol is adopted in all facilities. Surgical evacuation of the uterus is usually done at set times to allow organisation of operating theatres, while medical evacuation can be offered immediately on diagnosis of incomplete abortion. One-hird of clients first sought care at primary health centres but needed to go to higher level facilities to receive PAC; the lack of PAC capacity at many primary health centres further delays management and potentially increases the severity of complications, including the possibility of death. Enabling primary health centres to provide PAC using misoprostol or MVA would reduce costs and improve accessibility, decongest higher level facilities and presumably reduce maternal mortality.

A Zimbabwean pilot study found that using misoprostol for PAC reduced referral rates in primary health centres (from 98% to 10%) and rural/mission hospitals (from 48% to 3%), while maintaining 96% efficacy.36 However, in our study, most PAC cases were managed with D&C/D&E and few were managed with MVA/electrical vacuum aspiration (12%) or misoprostol (11%). Using medical evacuation with misoprostol is considerably less expensive than surgical evacuations while MVA is safer and results in less perioperative blood loss.37 Zimbabwe lags behind other countries like Malawi27 and Kenya34 in adopting MVA. Furthermore, although most clients in our study were managed by medical doctors, training midwives and nurses to use misoprostol and MVA, when appropriate and as allowed under current law, will improve PAC availability, shorten delays and consequently reduce severity, particularly in settings without doctors, like primary health centres. Receipt of IV fluids by 67% of patients suggests potential overuse, as this is generally only necessary for women with moderate or greater severity. Conversely, we observed a potential underuse of analgesics; all patients receiving PAC should receive this,37 but only 78% did. We recommend in-service training of clinicians in the national PAC guidelines and regular audits to check whether clinicians are following the guidelines, or are experiencing stock-outs which prevent them from doing so.

Strengths

Our study was nationally representative, involving all provinces in Zimbabwe and sampling all central and provincial hospitals. Prospective data collection in facilities avoided concerns encountered in retrospective studies assessing abortion-related morbidity due to missing patient records. Using a patient interview, provider interview and case notes improved data accuracy. We achieved a high response rate in sampled facilities (95%). The revised morbidity criteria reduced potential overestimation of severity by removing unreliable stand-alone criteria such as fever and tachycardia. In addition, our comparison of treatment provided to national PAC guidelines and consideration of women’s delays to care are relevant to policies to improve the clinical management and quality of PAC.

Limitations

Our study has several limitations. First, we were unable to distinguish between induced and spontaneous abortions, and as noted above, complications from induced abortions may be more severe than those from spontaneous abortions.26 We asked women if they had done anything to interfere with the pregnancy, but acknowledgement was extremely rare (4%), potentially owing to fear of legal repercussions. We did not consider this self-reported information to be sufficiently reliable. Second, as a facility-based study, information on women with mild complications that resolved spontaneously, severe complications leading to death outside the facility, or any other case of PAC occurring outside of a facility were not captured. Third, although we used prospective data collection, we were unable to gather information on 2 women who died and 12 women classified as near misses in tracking forms. Thus, the most severe cases may still be underestimated.

Generalisability of results

We collected data from over half of all PAC-providing facilities in Zimbabwe, including all higher level facilities, and our results are nationally representative. Regarding generalisability of our findings beyond Zimbabwe, facility-based measures of abortion morbidity are influenced both by abortion safety and women’s access to treatment, which varies across contexts. Our findings on the severity of postabortion complications are likely generalisable to other countries facing similar resource constraints, and with similar levels of abortion safety. The impact of recent declines in Zimbabwe’s healthcare system on abortion morbidity are difficult to quantify, but should be considered in applying these findings elsewhere. Furthermore, little is known about levels of misoprostol access for termination of pregnancy across the region, including Zimbabwe. If misoprostol use, and therefore abortion safety, is different in Zimbabwe compared with other countries, this may also limit generalisability. However, our findings on the characteristics associated with severity and the delays in access to care are likely generalisable to other countries in the region, where access to care is shaped by geography and resources.27 34 In addition, Zimbabwe lags behind its regional counterparts in rolling out MVA as the standard of care for surgical evacuation of the uterus.27 34

Conclusion

In Zimbabwe, abortion-related morbidity and concomitant mortality could be reduced by liberalising the abortion law, providing PAC in primary health centres, and training nurses to use medical evacuation with misoprostol and MVA. Regular in-service training on PAC guidelines should be done with follow-up audits to ensure compliance and availability of equipment, supplies and trained staff. Besides in-service training as a class-based approach, coaching and mentorship are necessary for performance improvement of health providers.38 Efforts are needed to reduce unintended pregnancy and unsafe abortion among those more likely to have severe abortion-related complications, including adolescents and women in rural areas, those with less education or those not in union. Further research should address motivations and barriers for providers to adopt evidence-based best practice for comprehensive PAC.