Severity and management of postabortion complications in Kenya: results from a nationally representative cross-sectional survey

Abstract

Introduction

Complications from induced abortion are an avoidable cause of maternal morbidity and mortality. A nationally representative 2012 study estimated that 37% of postabortion care patients in Kenya experienced severe complications. This study provides updated estimates of postabortion complication management and severity in Kenya.

Methods

Data were collected on 4666 postabortion care cases in a nationally representative sample of health facilities (n=367). Abortion-related complications were classified as mild, moderate, potentially life-threatening (PLTC) or a severe maternal outcome (SMO). Adjusted Poisson regression was used to estimate the association between patient characteristics and experiencing a PLTC/SMO.

Results

1.4% (95% CI 0.5% to 3.8%) of cases were classified as a SMO, 16.4% (95% CI 11.8% to 22.3%) as a PLTC, 28.5% (95% CI 21.5% to 36.8%) as moderate and 53.6% (95% CI 45.1% to 62.0%) as mild. Medical vacuum aspiration/electric vacuum aspiration was the most common uterine evacuation method (65.2%, 95% CI 54.0% to 74.9%), and 11.7% of cases (95% CI 7.1% to 18.5%) did not receive any uterine evacuation procedure. Experiencing a PLTC/SMO was associated with primary-level schooling or less (adjusted incidence rate ratio, aIRR 1.86, 95% CI 1.05 to 3.29), inability to meet household needs (aIRR 1.43, 95% CI 1.16 to 1.76), food insecurity (aIRR 1.65, 95% CI 1.21 to 2.24), using a non-recommended abortion method (aIRR 2.56, 95% CI 1.81 to 3.61) and seeking care from another source before arriving at the current facility (aIRR 2.06, 95% CI 1.14 to 3.73).

Conclusions

Comparing the results of this analysis to the 2012 study reveals that the proportion of postabortion care patients experiencing the most severe outcomes has reduced over time and patients with mild complications more than doubled (from 23% in 2012 to 54% in 2023). This shift is likely a result of expanded primary care and increased access to medication abortion from informal sources. Despite this decline, severe postabortion complications remain an avoidable risk to women who induce abortions. Factors associated with these complications, including use of non-recommended abortion methods and economic vulnerability, underscore the need for interventions that enhance awareness around safe abortion methods and increase access to affordable healthcare.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Previous work has demonstrated that legal restrictions around safe abortion care are associated with increased risk of negative health outcomes among women who seek abortions clandestinely.

• Despite the highly restrictive legal environment, induced abortion is common in Kenya, and a nationally representative study from 2012 documented high levels of abortion-related morbidity and mortality.

• While it is difficult to measure the impact of recent increases in access to medication abortion on women’s health outcomes, many have hypothesised that this increased access could reduce the prevalence of severe health outcomes among women who induce abortions in legally restrictive settings.

WHAT THIS STUDY ADDS

• Over the past 10 years, the proportion of women experiencing the most severe postabortion complications in Kenya has decreased (37.1% in 2012 to 17.8% in 2023) while the proportion of women who received care for mild complications more than doubled (22.9% to 53.6%), and this evidence supports the hypothesis that the increased availability of medication abortion may be contributing to the lessened burden of postabortion complications in settings with restrictive abortion laws.

• Several factors were shown to be associated with increased risk of experiencing one of the most severe outcomes, including indicators of economic vulnerability (low levels of education, inability to meet household needs, food insecurity) and delays in accessing care.

• Severe postabortion complications remain an avoidable risk to women who induce abortions, and the findings from this study underscore the need for interventions that enhance awareness around safe abortion methods and increase access to affordable healthcare.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The results of this study demonstrated how restricting access to safe abortion care can add additional and avoidable pressures on healthcare systems. While these results are specific to Kenya, it is likely that similar phenomena are occurring in other settings where abortion is highly legally restricted and/or people do not have adequate access to information about abortion processes or resources.

• There is a need to design and implement interventions aimed at increasing people’s knowledge about safe abortion care, providing virtual access to information needed, triage systems after inducing an abortion and training pharmacists or community health promoters to increase high-quality provision of medication abortion services.

Introduction

Induced abortion is a common experience worldwide. Recent global estimates suggest that 73.3 million induced abortions occur each year, corresponding to an incidence rate of 39 per 1000 women of reproductive age (WRA) annually.¹ When abortions are provided in adherence to clinical guidelines (the use of a WHO-recommended method appropriate to gestational age, provided by someone with necessary training, guidance and/or skills), they are associated with low levels of negative health outcomes.² However, access to appropriate, high-quality abortion care is limited in contexts where abortion is legally restricted. A study using data from 2010 to 2014 found that the proportion of abortions that are considered ‘safe’ (ie, aligned with the WHO guidelines) is substantively higher in settings where abortion is broadly accessible versus highly legally restricted (87.4% vs 25.2%, respectively).³

When access to safe abortion care is legally restricted, individuals may need to resort to traditional or harmful methods to induce abortions, as most people did not have the knowledge or financial means to access safer providers and/or methods.⁴ A study using data from 2009 to 2020 estimated that 8% of maternal deaths in sub-Saharan Africa, where abortion is largely restricted, were attributable to complications from abortion (7.7%, 95% CI 5.6% to 11.0%).⁵ Further, a meta-analysis that covered research from 1988 to 2014 estimated that at least 10.5% of abortion-related complications treated in health facilities in legally restrictive settings resulted in death or a ‘near-miss’ case (near-miss cases are those where the woman nearly died but survived the abortion complication after receiving emergency medical intervention).⁶ However, more recent evidence suggests that the rise in access to medication abortion (MA) may be reducing the number of women who experience negative health events after an induced abortion.⁷,¹¹ A recent study conducted in 11 African countries found that only 2.3% of postabortion care (PAC) patients were classified as near miss or death, while more than 90% had moderate or mild complications, with vaginal bleeding being the most common symptom.¹²

In Kenya, while abortion remains highly legally restricted,^{13 14} many women induce abortions outside the formal healthcare system; a 2023 study estimated that the 1 year induced abortion incidence rate was 57.3 per 1000 women aged 15–49, and 55% of unintended pregnancies ended in an induced abortion.¹⁵ However, little is known about the safety of these abortions. The 2019 maternal mortality ratio (MMR) in Kenya was 355 deaths per 100 000 live births, which is more than double the global estimate,¹⁶ and a 2012 study documented high levels of abortion-related morbidity and mortality, with 37% of PAC patients experiencing severe complications.¹⁷ That said, it is possible that recent changes in the abortion landscape may have influenced the abortion method mix in Kenya,^{18 19} as well as the severity of abortion-related complications.

This study aims to provide nationally representative estimates of the severity and management of postabortion complications in Kenya. In addition, we investigate whether sociodemographic characteristics or delays to care are associated with experiencing more severe outcomes. The resulting evidence is essential for policy formulation and programming aimed at improving the health of women and girls in Kenya.

Methods

Sampling and data collection

Data come from a 2023 Prospective Morbidity Study (PMS) conducted in Kenya, which was implemented as part of a larger study on induced abortion.¹⁵ The PMS included four data collection tools: (1) a survey of patients who received PAC (patient survey), (2) a survey of clinicians who provided care to each patient (provider survey), (3) a medical records review for each patient (MRR) and (4) a tally sheet that tracks the number of patients who received care each facility during the data collection period.

To identify PMS facilities, we employed a two-stage sampling design (table 1). In the first stage, we created a nationally representative sample of health facilities in Kenya capable of providing PAC (n=658). (A detailed explanation of this sampling strategy is provided elsewhere.¹⁵) In stage 2, we first removed facilities from the sample that reported not providing PAC (n=7) and level II facilities (n=125) due to low monthly caseloads. Next, we selected a random sample of health facilities that successfully participated in the HFS to participate in the PMS (100% of level VI and V facilities, 75% of level IV facilities and 55% of level III facilities, n=431). All sampled facilities were invited to send clinical staff to participate in a week-long, in-person training (In some cases, larger facilities with high caseloads sent more than one person to be trained). Eligible staff provided some aspect of PAC in their facilities, including medical doctors, clinical officers and nurses/midwives. In total, 367 facilities (85%) participated in the study. Reasons for non-participation (n=64) included inability to send a provider to the training or a trained provider subsequently transferring to a different facility.

Table 1. PMS facility and patient sampling and response rates.

	Facility sampling data			Patient sampling data
	Facility sampling frame	Sampling fraction	Participating facilities	Total PAC patients during study period	Patients with any survey data	Number excluded from analysis	Analytic sample	Patients with all survey data	MRR only	Provider survey only
	n	% (n)	% (n)	n	% (n)	n	n	% (n)	% (n)	% (n)
Overall	526	n/a (431)	85 (367)	5222	89.4 (4666)	561	4105	81.6 (3282)	9.7 (479)	8.8 (344)
Level
Level III	116	55 (63)	81 (55)	305	77.4 (236)	6	230	83.4 (194)	7.4 (17)	9.2 (19)
Level IV	348	75 (306)	84 (256)	2413	88.8 (2143)	207	1936	80.0 (1548)	11.5 (223)	8.6 (165)
Level V	59	100 (59)	90 (53)	2338	91.1 (2129)	321	1808	78.1 (1419)	13.5 (234)	8.4 (155)
Level VI	3	100 (3)	100 (3)	166	95.2 (158)	27	131	93.0 (121)	3.3 (5)	3.7 (5)

MRR, medical records review; n/a, not available; PAC, postabortion care; PMS, Prospective Morbidity Study.

Data collection took place over 30 days (Due to the large number of study facilities, data collection was conducted in two rounds (round 1: June–July 2023, round 2: August–September 2023), during which data collectors attempted to recruit all women who presented with any signs or symptoms of complications of pregnancy loss up to 28 weeks. Patients aged 15+ provided verbal consent for the patient and provider surveys. Patients <15 were not eligible for the patient survey but could consent to have their providers share their clinical information. Patient and provider surveys were administered face-to-face, and responses were electronically recorded using SurveyCTO on Android tablets. To ensure the most severe cases were captured, ethical approval was granted to interview the providers of patients who could not consent due to coma or death. MRRs were conducted for all patients who received care during the data collection period, regardless of their participation in the patient/provider survey.

The patient survey and provider surveys were adapted from previous PMS studies.²⁰,²² The patient survey collected information on sociodemographic characteristics, reproductive health history, delays to accessing care and PAC family planning. The provider survey collected clinical data on the severity and management of complications. The MRR was adapted from the Abortion-related Morbidity and Mortality in Fragile and Conflict Settings (AMoCo) study.²³ It was included in this study to validate initial provider reports, supplement cases of incomplete or missing provider survey data (including laboratory measurements), and document final diagnoses. Data collectors conducted the MRRs after patients had been discharged. Tally sheets documented the total number of women who presented for PAC during data collection, the number successfully surveyed and the number who left without being approached (either because they were discharged or referred to a higher-level facility for care).

Overall, 5222 women presented for PAC during the data collection period and at least one survey was implemented for 4666 patients (89.4%) (table 1). Patients with threatened abortion (vaginal bleeding with closed cervix), ectopic pregnancy (implantation of a fertilised egg outside the uterus), molar pregnancy (development of hydatidiform mole from an abnormal fertilised egg) or post 28 weeks gestation did not meet the study definition of PAC and were excluded from analysis (n=561). The final N was 4,105, among which 81.6% of cases (n=3282) have data from all three surveys 9.7% (n=479) have MRR data only and 8.8% (n=344) have patient and/or provider data alone.

Classifying postabortion complication severity

We adapted methodologies employed by the World Health Organization Multi-Country Survey on Abortion-related Morbidity (WHO-MCS-A) and AMoCo studies to create severity classifications.^{12 23} Patients were classified into four severity categories: mild complications, moderate complications, potentially life-threatening complications (PLTCs) and severe maternal outcomes (SMOs) (table 2). SMOs include death and cases classified as ‘near-miss’, defined as experiencing dysfunction at least one of the following organs/systems: cardiovascular, respiratory, renal, hepatic, neurologic, uterine or coagulation/haematologic. PLTC cases are those with severe systemic infection, generalised peritonitis, severe haemorrhage, perforated uterus or other intra-abdominal injuries. Moderate complications include abdominal syndrome, uterine infection and severe vaginal bleeding. Mild complications are any abnormal physical examination findings, including vaginal bleeding, open cervix, uterine tenderness, abnormal vital signs or abnormal mental status. Detailed definitions of how indicators were operationalised using clinical data are provided in online supplemental table S1.

Table 2. Severity definitions.

Classification	Signs and symptoms
Mild complications	Any abnormal physical examination findings on initial assessment (abnormal vital signs, appearance, mental status, abdominal examination, gynaecological examination, any bleeding)
Moderate complications (requires ≥1 criterion)	Severe vaginal bleeding Abdominal syndrome Uterine infection
Potentially life-threatening complications (requires ≥1 criterion)	Severe systemic infection Generalised peritonitis Uterine perforation Other intra-abdominal perforation Severe haemorrhage
Severe maternal outcomes (requires ≥1 criterion)	Death Near-miss cases Neurological dysfunction (eg, coma, paralysis) Cardiovascular dysfunction (eg, shock, cardiac arrest) Respiratory dysfunction (eg, acute cyanosis, severe bradypnoea, intubations) Renal dysfunction Coagulation dysfunction (eg, failure to clot) Uterine dysfunction (eg, uterine rupture, hysterectomy) Hepatic dysfunction

^*Severity definitions adapted from WHO-MCS-A and subsequent adaptations done in AMoCo.^{12 21}

AMoCo, Abortion-related Morbidity and Mortality in Fragile and Conflict Settings; WHO-MCS-A, World Health Organization Multi-Country Survey on Abortion-related Morbidity.

Data from the MRR and provider survey were used to construct the severity indicator. We deferred to the MRR for cases with data from both sources (n=3282), as the MRR was specifically designed to generate the necessary indicators for the WHO-MCS-A and AMoCo severity definitions. Further, the MRR was likely a more accurate representation of clinical outcomes due to the timing of the data collection and the reliance on written records. Exceptions to MRR deference were cases that were inconsistently coded between the MRR/provider survey as a PLTC or SMO (n=64, 2.0%). In these instances, clinically trained study members conducted case reviews to determine whether the complication manifested in that particular patient. For cases missing the MRR (n=344), indicators from the provider survey were used.

Other indicators

The provider survey collected information on the management of postabortion complications. Uterine evacuation methods included medical vacuum aspiration (MVA)/electric vacuum aspiration (EVA), manual evacuation/dilation and evacuation (D&E), dilation and curettage (D&C), MA alone and no evacuation method. Provider types included general physicians/OBGYNs, nurses/midwives, and clinical officers. Gestational age was categorised as first versus second trimester (<14 weeks vs 14–28 weeks). We also measured length of stay (outpatient, 1 night, 2+nights), any receipt of pain medication during treatment, the receipt of PAC family planning counselling, and whether the patient left with a family planning method.

The patient survey provided sociodemographic characteristics, including age, marital status, educational attainment and two indicators of economic vulnerability: whether the patient’s household income was adequate to take care of all food/health needs in the past month and whether the patient went at least one full day without eating in the past month. We also assessed whether each patient had ever given birth and if their current pregnancy was unintended.

Data from the patient and provider surveys were used to identify induced abortions. Patients were asked whether they did anything to intentionally end their pregnancy and, if so, which methods they used. Providers were asked if they thought the patient was seeking care after an induced abortion and their assessment of the abortion method. Provider reports were only included if the patient told the provider directly about their abortion or if there were clear clinical signs (ie, evidence of mechanical injuries). We created a three-category indicator for induced abortions: no evidence, induced abortion using recommended methods only (MVA/EVA and/or MA), and induced abortion using any non-recommended methods (ie, traditional methods, inserting something sharp into the vagina, ingesting caustic substances, etc).

The patient survey assessed delays to care, including when the respondent realised they needed care, how long after symptom onset they arrived at the health facility, and whether they sought care somewhere else before the current facility. Patients also reported whether they experienced a delay in their decision to seek care and/or travelling to the facility. These delays could be due to an inability to recognise the need to seek care, family/personal issues (eg, concerns about paying for care, partner or other family member disagreement, fear of others finding out), and structural reasons (eg, not knowing where to go for care, lack of transportation, blocked access due to conflict, natural disaster or roadwork).

Analysis

We constructed probability weights to account for the two stages of health facility sampling. We then adjusted these weights to account for facility-level non-response. All results are weighted using these composite design weights.

We first present weighted descriptive statistics for key study indicators, including patient sociodemographic and reproductive health characteristics, delays to care, distribution of severity outcomes, and management of postabortion complications. Next, we use χ² tests to determine whether the management of postabortion complications differed by complication severity.

We also construct a dichotomous indicator for the most severe complications (SMO/PLTC vs moderate and mild complications) to assess the relationship between key patient indicators and the most severe complications. We use Poisson regression to estimate adjusted incidence rate ratios (aIRRs) with 95% CIs and robust standard errors for the risk of experiencing a SMO/PLTC for each of the key study indicators, adjusting for facility-level. We also conduct sensitivity tests by investigating whether there were differences in key sociodemographic characteristics for women who experience delays to care using bivariate tests for statistical significance. All analyses were conducted using Stata V.18.²⁴

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

Results

The majority of patients were aged 25 or older (25–34=41.8%, 95% CI 38.7% to 45.0%, 35–49=15.6%, 95% CI 11.2% to 21.2%) and 13.6% (95% CI 9.8% to 18.4%) were under the age of 20 (table 3). Most patients were married or living with a partner (78.6%, 95% CI 71.7% to 84.2%). Approximately one in three had a primary-level education or less (33.3%, 95% CI 24.6% to 43.4%), 36.7% (95% CI 29.4% to 44.5%) completed secondary school and 30.0% (95% CI 27.2% to 33.0%) had post-secondary education. Almost one-third of patients (29.9%, 95% CI 20.0% to 42.1%) reported being unable to meet all household needs in the past 30 days, and about one-fifth (22.8%, 95% CI 17.3% to 29.5%) reported experiencing food insecurity. Most women (65.6%, 95% CI 60.6% to 70.3%) had previously given birth, and 44.5% (95% CI 35.7% to 53.7%) reported that their current pregnancy was unintended. Overall, there was evidence of an induced abortion conducted using only recommended methods for 21.8% of patients (95% CI 15.4% to 30.1%), and another 6.7% of patients (95% CI 3.9% to 11.2%) displayed evidence of seeking care after an induced abortion conducted with any non-recommended methods.

Table 3. Sociodemographic characteristics, reproductive health indicators and the severity of abortion-related complications among postabortion care patients (n=3626).

	%	95% CI
Socioeconomic characteristics
Age*
<20 years	13.6	9.8 to 18.4
20–24 years	29.1	25.1 to 33.4
25–34 years	41.8	38.7 to 45.0
35–49 years	15.6	11.2 to 21.2
Marital status
Never married	16.6	10.5 to 25.1
Married/living together	78.6	71.7 to 84.2
Divorced/separated/widowed	4.8	3.0 to 7.7
Education
Primary or less	33.3	24.6 to 43.4
Secondary	36.7	29.4 to 44.5
Post-secondary	30.0	27.2 to 33.0
Household income in the last month
Unable to take care of all food, health, other personal needs in past month	29.9	20.0 to 42.1
At least a full day without food in past month	22.8	17.3 to 29.5
Reproductive health behaviours
Has previously given birth*	65.6	60.6 to 70.3
Current pregnancy unintended	44.5	35.7 to 53.7
Evidence of induced abortion
None	71.5	62.0 to 79.4
Yes, using recommended methods only	21.8	15.4 to 30.1
Yes, using any non-recommended methods	6.7	3.9 to 11.2
Severity of postabortion complications*
Severe maternal outcomes	1.4	0.5 to 3.8
Potentially life-threatening complications	16.4	11.8 to 22.3
Moderate	28.5	21.5 to 36.8
Mild	53.6	45.1 to 62.0
Delays to care
Knew help was needed within 24 hours of symptom onset	70.2	60.2 to 78.5
Arrived at health facility within 24 hours of symptom onset	59.6	46.0 to 71.9
Current health facility not first source of medical care	20.6	13.7 to 29.8
Self-reported delays: decision to seek care	23.0	17.1 to 30.2
Self-reported delays: travel to the facility	15.4	9.7 to 23.5
Reasons for delays in seeking care and/or travelling†
Patient did not think their health problem was serious	44.9	31.5 to 59.0
Family/personal reasons	61.4	51.6 to 70.4
Structural reasons	39.6	22.5 to 59.7

^*Data on age, prior birth history and severity of postabortion complications available for full sample (n=4105).

^†Multiple responses allowed.

Overall, we estimate that 1.4% (95% CI 0.5% to 3.8%) of PAC patients experienced a SMO (table 3). This includes five women who died and eight who were in a coma (see online supplemental table S2 for distributions of individual sign or symptoms comprising each severity category.) Approximately 16.4% (95% CI 11.8% to 22.3%) of patients were classified as experiencing PLTC. Moderate complications comprised 28.5% (95% CI 21.5% to 36.8%) of cases, and more than half of PAC patients had mild complications (53.6%, 95% CI 45.1% to 62.0%). Results of the sensitivity test revealed almost no change in the distribution of severity outcomes after removing clinical data from the provider survey (SMO=1.5%, PLTC=15.9%, moderate=28.6%, mild=53.9%, online supplemental table S3).

More than two-thirds of patients (70.2%, 95% CI 60.2% to 78.5%) reported realising that medical care was needed within 24 hours of exhibiting symptoms, and 59.6% (95% CI 46.0% to 71.9%) arrived at the health facility in this time frame (table 3). About one-fifth (20.6%, 95% CI 13.7% to 29.8%) sought care from another source prior to arriving at the current facility, including a different health facility or provider, pharmacy or traditional healer/birth attendant. One in four patients (23.0%, 95% CI 17.1% to 30.2%) felt their decision to seek care was not made in time, and 15.4% (95% CI 9.7% to 23.5%) reported delays while travelling to the facility. Among this group, patients often cited multiple reasons for experiencing delays, including the belief that their complications were not serious (44.9%, 95% CI 31.5% to 59.0%), familial/personal reasons (61.4%, 95% CI 51.6% to 70.4%) and structural reasons (39.6%, 95% CI 22.5% to 59.7%).

Most PAC cases were treated with MVA/EVA (65.2%, 95% CI 54.0% to 74.9%), and 18.9% (95% CI 10.1% to 32.8%) were treated with MA (table 4). Few cases were treated with D&E or a manual evacuation method (2.3%, 95% CI 1.4% to 4.0%) and D&C was similarly rare (1.9%, 95% CI 1.4% to 2.6%). The remaining 11.7% of cases (95% CI 7.1% to 18.5%) did not receive any uterine evacuation procedure, most commonly because the patient had expelled all the products of conception prior to arriving at the facility (online supplemental table S4). About half of patients (51.5%, 95% CI 38.9% to 64.0%) were treated by clinical officers, followed by nurses/midwives (28.9%, 95% CI 20.8% to 38.6%) and general doctors/obstetrician-gynecologists (OBGYNs) (19.6%, 95% CI 6.9% to 44.5%). Approximately 74.7% of patients (95% CI 68.8% to 79.8%) were in their first trimester of pregnancy, half were treated as outpatients (49.8%, 95% CI 33.0% to 66.7%) and 80.4% (95% CI 73.6% to 85.8%) received pain medication during their uterine evacuation. While almost all women received family planning counselling (92.3%, 95% CI 82.8% to 96.8%), less than half (43.5%, 95% CI 31.5% to 56.2%) left the facility with a contraceptive method. There were no statistically significant differences in postabortion management indicators by complication severity apart from length of stay, with the proportion of women with the most severe outcomes who were admitted for two or more nights being almost double that of women experiencing mild complications (38.8% vs 20.9%) (p=0.043).

Table 4. Management of postabortion complications, overall and by complications severity (n=3540).

	Overall		Mild	Moderate	PLTC or SMO	P value
	%	95% CI	%	%	%
Method of evacuation
MVA/EVA	65.2	54.0 to 74.9	62.2	70.9	64.7	0.516
MA	18.9	10.1 to 32.8	20.8	17.1	16.2
D&E/other manual	2.3	1.4 to 4.0	1.8	2.7	3.3
D&C	1.9	1.4 to 2.6	1.9	1.6	2.2
No evacuation procedure	11.7	7.1 to 18.5	13.2	7.7	13.6
Provider type
General doctor/OBGYN	19.6	6.9 to 44.5	20.0	15.7	25.3	0.142
Nurse/midwife	28.9	20.8 to 38.6	32.4	23.7	27.1
Clinical officer	51.5	38.9 to 64.0	47.6	60.6	47.6
Gestational age*
First trimester	74.7	68.8 to 79.8	77.1	78.8	62.0	0.161
Second trimester	25.3	20.2 to 31.2	22.9	21.2	38.0
Number of nights in facilities
0	49.8	33.0 to 66.7	55.9	45.0	38.6	0.043
1	25.2	14.8 to 39.5	23.2	30.5	22.6
2+	25.0	13.7 to 41.0	20.9	24.5	38.8
Patient received pain medication	80.4	73.6 to 85.8	74.0	87.3	87.8	0.096
Patient received contraceptive counselling*	92.3	82.8 to 96.8	92.6	92.8	89.9	0.166
Patient left facility with a method of contraception*	43.5	31.5 to 56.2	39.7	45.1	52.4	0.205

^*Data on gestational age, receipt of contraceptive counselling and receipt of contraceptive method available for full sample (n=4105).

D&C, dilation and curettage; D&E, dilation and evacuation; EVA, electric vacuum aspiration; MA, medication abortion; MVA, medical vacuum aspiration; OBGYN, obstetrician-gynecologist; PLTC, potentially life-threatening complications; SMO, severe maternal outcomes.

Figure 1 displays the aIRRs and 95% CIs for the relationships between the most severe complications (SMO/PLTC) and key sociodemographic characteristics, reproductive health factors, and delays to care (see online supplemental table S5 for aIRRs, 95% CI and p values.) After adjusting for facility level, there were no statistically significant relationships between complication severity and age or marital status. However, women with the lowest levels of education (primary school or less) were 1.86 times more likely to experience the most severe complications compared with women with a post-secondary education (95% CI 1.05 to 3.29). Being unable to afford all household needs in the past month (aIRR 1.43, 95% CI 1.16 to 1.76) or experiencing food insecurity in the last month (aIRR 1.65, 95% CI 1.21 to 2.24) were associated with experiencing the most severe complications.

Figure 1. Factors associated with experiencing the most severe postabortion complications (severe maternal outcomes or potentially life-threatening complications versus moderate or mild complications)^§. §All models control for facility level. aIRR, adjusted incidence rate ratio. *The reference group for age is <20 years. **The reference group for education level is post-secondary education. ***The reference group for induced abortion by method is did not induced.

Experiencing a PLTC/SMO was more likely among women who had previously given birth (aIRR 1.37, 95% CI 1.17 to 1.62) and among women who reported that their pregnancy was unintended (aIRR 1.72, 95% CI 1.48 to 2.01). Women with evidence of an induced abortion performed using any non-recommended method were 2.56 times more likely to experience a SMO/PLTC (95% CI 1.81 to 3.61). Increased risk of SMO/PLTC was observed for women who felt their decision to seek care was not taken soon enough (aIRR 2.03, 95% CI 1.28 to 3.23), for women who sought care from another source before arriving at the current facility (aIRR 2.06, 95% CI 1.14 to 3.73), and for women who reported experiencing a delay while travelling to the facility (aIRR 1.85, 95% CI 0.91 to 3.77).

In sensitivity analyses, there were no significant differences between those who experienced delays to care by age, marital status, education level or any of the reproductive health outcomes (online supplemental table S6). However, the two indicators of economic vulnerability were significantly associated with experiencing a delay accessing care. Being unable to take care of all household needs in the past month was more common among patients who reported the current facility was not their first source of medical care (44.5% vs 26.1%, p=0.012), for patients who reported a delay in deciding to seek treatment (43.8% vs 25.5%, p<0.001), and for patients who experienced a delay while travelling to the health centre (44.9% vs 26.7%, p=0.061). Women who reported going at least a full day without food in the past month also more commonly reported experiencing delays to care (current facility not first source of care: 32.8% vs 20.3%, p=0.059; delayed decision to seek care: 32.2% vs 19.7%, p=0.003; delay during travel to facility: 37.3% vs19.6%, p=0.007).

Discussion

The results from this study offer important insights into maternal morbidity related to spontaneous and induced abortion in Kenya. Approximately 18% of women who sought PAC in 2023 experienced either a PLTC or a SMO. This finding represents a dramatic decrease over the past decade in Kenya; results from a nationally representative study conducted in 2012 estimated that 37.1% of PAC patients in Kenya experienced severe complications, more than double the proportion from this study.¹⁷ At the same time, the 2012 study found that only 22.9% of patients had mild complications, compared with the 53.6% estimated in this analysis (see online supplemental table S7 for full comparison).

One likely contributor to the increased proportion of PAC patients with mild complications observed in the study are recent efforts by the Kenyan Ministry of Health to expand access to primary healthcare facilities, which includes adopting models designed to increase coordination and efficiency, encouraging task-shifting and increasing uptake of Kenya’s Social Health Insurance Fund.^{25 26} As such, women with mild complications may be more likely to seek treatment from a formal health facility, as opposed to forgoing treatment or seeking care using traditional remedies and/or non-health system providers. Some empirical evidence supports this hypothesis; even though the proportion of women experiencing severe complications decreased over the past decade, the PAC treatment rate (number of PAC cases per 1,000 WRA in Kenya) increased from 16.0 per 1000 in 2012 to 22.0 per 1000 in 2023.^{15 27} Further, women at later gestational ages are more likely to require clinical interventions to avoid severe morbidity or death, and the proportion of PAC patients in their second trimester decreased from 37% in 2012 to 25% in 2023.¹⁷ Expanded access would likely reduce delays to care, which would reduce the risk of severe morbidity.^{20 28 29} Some evidence from this study supports this claim; while less than half of women in 2012 (46%) accessed care within a 48 hour timeframe,¹⁷ 60% of patients in this study were able to access care within 24 hours of symptom onset.

It is also possible that the changing abortion landscape is contributing to the shifting composition of abortion-related complication severity in Kenya. While little is known about trends in MA use in Kenya over time, recent work suggests that MA is the most commonly used abortion method in Kenya, with most women accessing MA from pharmacies, drug shops or other non-formal health system sources.^{30 31} While research has shown that MA abortions from these sources are safe with little evidence of severe health complications,^{10 11 32} mystery client studies have documented that pharmacists/other sellers often do not provide adequate information about what to expect after taking the medications or how to identify possible complications.³³,³⁸ The combination of these circumstances is likely resulting in increased PAC seeking for physical symptoms that are, in reality, expected parts of the process of inducing an abortion. Recent research from similar settings has documented that people seek care during/after an abortion for a variety of non-clinical reasons, most commonly to confirm that the abortion process has completed or for reassurance from a medical professional.^{39 40} The increased proportion of women who access PAC with mild complications in Kenya could be a reflection of this phenomenon.

It is difficult to differentiate between induced and spontaneous abortions using this study’s data, which limits our ability to determine whether women experiencing mild complications were seeking care after an MA-induced abortion. This is because women often underreport induced abortion in surveys,³⁹,⁴¹ and the clinical presentations of pregnancy loss due to spontaneous and induced abortions are often indistinguishable.⁴² As such, this study’s finding that 29% of PAC cases had evidence of an induced abortion is almost certainly an underestimate. That said, some of this study’s results provide support for the hypothesis that the large proportion of women experiencing mild complications is due, at least in part, to the rise in informal access to MA in Kenya. Among PAC cases where induced abortions were positively identified, more than three-quarters involved any use of MA. Further, no uterine evacuation procedure was conducted in approximately 12% of all PAC cases, most commonly because the patient had already expelled all products of conception prior to arriving at the facility. These cases potentially reflect women who had a successful medical abortion and sought PAC for non-clinically indicated reasons (ie, concern about bleeding due to inadequate information, desire to confirm the pregnancy termination, etc). It is important to note that the average cost of treating PAC in the public health facilities in Kenya is estimated at approximately US$58 per case (~US$5.1 million annually on PAC), while the cost of providing MA for abortion services is estimated to be much lower.^{43 44} As such, efforts to expand access to high-quality safe abortion care (either through legal reform or programmes aimed at training pharmacists or community health promoters), as well as interventions for MA users designed to provide virtual access to information needed to triage symptoms, can alleviate pressure on the health system and save valuable government resources.

Despite these positive trends in severity, PLTCs and SMOs remain a concern. A recent study estimated that 3 04 159 women received PAC in Kenya in 2023.¹⁵ Applying this study’s estimated proportion that 17.8% of all PAC cases result in a SMO/PLTC, this would correspond to approximately 54 000 women experiencing one of these most severe outcomes each year (see online supplemental table S7). Unlike maternal morbidity and mortality associated with childbirth or later gestational age miscarriage, many (if not most) cases of severe abortion-related complications are avoidable; induced abortions conducted by trained providers using recommended methods are extremely safe,^{45 46} and timely access to appropriate treatment can mitigate the severity of complications associated with miscarriage.⁴⁷ However, aspects of the Kenyan context facilitate increased risk for experiencing severe complications. This study found that delays to care were associated with experiencing severe complications, and that economically vulnerable women were more likely to experience a delay to care. A recent study revealed that only one in four referral-level facilities in Kenya meet the criteria for providing comprehensive PAC, commonly because facilities could not provide surgical care.⁴⁸ As such, continued efforts to improve the capacity of the Kenyan health system to provide PAC are needed, as well as interventions aimed at decreasing delays to care.

Further, legal restrictions limit women’s access to safe abortion care. Among identifiable cases of induced abortion in this study, non-recommended method use was associated with experiencing a SMO/PLTC, while women who used recommended methods did not experience an increased risk. Further, economically vulnerable women were also shown to face an elevated risk of experiencing severe complications. Liberalising Kenya’s abortion law would help reduce the occurrence of severe complications, although persisting abortion-related stigma and structural barriers to healthcare would likely still result in inequitable access to safe abortion services. As such, targeted efforts aimed at enhancing awareness of safe abortion methods and empowering individuals to make informed decisions about their reproductive health will reduce the burden of abortion-related morbidity among the most vulnerable women and girls in Kenya.

Limitations

It is possible that our results are not representative of the underlying population. We did not include level II facilities in our sampling strategy due to their low monthly patient volume. While most instances of severe complications were likely captured by our study, as women seeking care at level II facilities would be referred to higher-level facility for treatment, the proportion of moderate or mild complications may be larger than reported in this study. Further, women who died in the community or did not seek care for their complications were not included in this analysis. We were also unable to include high-profile private facilities in the study sample due to permission issues, meaning our study may have missed a specific profile of women who seek services in these facilities. While non-response rates and missingness were low, patterns in missing data were not necessarily random. In some cases, women were not approached due to provider availability, high caseload volumes, or short duration of clinical care. MRRs were typically missing due to interviewer error or poor record keeping. Further, while patient age and severity classification are available for the full sample, other key sociodemographic characteristics and reproductive history indicators are missing for the 10% of cases with MRR data only. The extent to which the cases with missing data systematically differed from those in the final analysis may have unduly biased our estimates.

There are some differences in how the previous 2012 severity study classified complication severity and the methods used in this study, although the impact of these differences likely does not negate our ability to compare the findings over time. Most importantly, the 2012 study included all cases with a temperature >37.9°C in the ‘severe’ category. In this study, patients with uterine infection (the definition of which includes temperature >37.9°C) but no other more severe indicators were classified as having moderate complications. However, reclassifying these cases (n=24) into the PLTC group would only increase estimated severe outcomes from 17.8% to 18.4%. There are no differences between the two studies’ classification of mild cases. If anything, our study was likely better able to identify the most severe cases due to our use of multiple survey instruments and more precise clinical indicator definitions.

There are also important limitations in our classification of mild complications. As noted above, while some patients in this category may have required medical intervention, it is likely that a meaningful proportion of women classified with mild complications were presenting with the intended effects of MA. Unfortunately, existing tools and routinely collected indicators are insufficient for differentiating between clinically necessary PAC and other types of PAC-seeking. In many cases, it is difficult to determine which patients would have eventually expelled the products of conception without medical intervention; and, in the case of induced abortion, it is difficult to determine patients’ motivations for care-seeking due to abortion-related stigma and underreporting. As the availability of MA increases, particularly in legally restrictive settings where women may not have access to other types of supportive services and information, misclassification of mild complications (vs follow-up care seeking) will continue to be a problem for the field.

Finally, given our inability to identify the true proportion of PAC cases in the sample due to induced abortion, this study cannot estimate the contribution of induced abortion to maternal mortality or morbidity in Kenya. The results related to induced abortion should be interpreted with caution.

Conclusions

The distribution of abortion-related complications in Kenya has changed over the last decade, with the proportion of women experiencing mild complications increasing and the proportion with the most severe abortion-related complications decreasing. Factors that may be contributing to this trend are efforts by the Kenyan government to increase access to primary care as well as increased availability of MA. Given the difficulty of differentiating between patients with mild complications and those seeking follow-up care after MA, future work is needed to investigate whether better tools and indicators can be designed to disentangle and identify these different types of PAC cases.

That said, a meaningful proportion of women still experience a severe complication, which underscores the need to improve access to high-quality safe abortion care, PAC, and postabortion family planning. Further, the results of this study reveal that the burden of the most severe complications was inequitably shouldered by economically vulnerable women. As such, interventions aimed at increasing access to care and information about safe abortion practices that are tailored to the unique needs and realities of economically vulnerable women are necessary to reduce the burden of abortion-related morbidity and mortality in Kenya.

Data availability statement

Data are available on reasonable request.