Abstract

Background

Unsafe abortion remains a significant driver of maternal morbidity and mortality. We aimed to assess the severity of abortion complications among post-abortion care (PAC) clients in Sierra Leone and factors associated with more severe complications.

Methods

We applied the Prospective Morbidity Survey (PMS) among PAC patients and their providers in 142 facilities in Sierra Leone between August and October 2021. We administered a total of 522 patient PMS surveys and 513 provider PMS surveys. To assess post-abortion complication severity, we utilized a five-level severity classification system ranging from mild to maternal death. We conducted bivariate and multivariable tests to assess factors associated with the severity of complications. Dependent variables included demographic characteristics and delays to care reported by PAC patients.

Results

Overall, 36% of PAC patients had mild complications, 30% moderate, 27% severe, and 8% near-miss. One person died. Many women experienced delays to care, particularly in realizing care was needed, deciding to seek care, and arriving at the health facility. Controlling for facility level, the risk of experiencing a severe or near-miss complication or death was not significantly associated with the patient鈥檚 sociodemographic characteristics, except age and number of pregnancies; compared to adolescents 19 and under, PAC patients aged 20鈥�24 had significantly lower risk of a severe/near-miss complication or death, while PAC patients with 2鈥�4 pregnancies had significantly higher risk of a severe/near-miss complication or death compared to PAC patients experiencing their first pregnancy. Delays in accessing care were significantly associated with having more severe complications; patients were more likely to suffer the most severe complications if they had to wait longer than one hour to be attended to, or did not receive complete treatment within 12听h.

Conclusions

Compared to regional and global estimates, the burden of abortion-related complications in Sierra Leone is high. The recent effort to liberalize abortion law is promising; this potential legal reform must be paired with the expansion of safe abortion services to reduce abortion-related morbidity and mortality in the country. In the absence of legal change, our analysis also emphasizes the need to ensure PAC services are always free of cost, that women are aware of the availability and legality of PAC, and that facilities have the human and material resources needed to handle these cases.

Background

Maternal morbidity and mortality remain pervasive public health challenges in many countries. Across sub-Saharan Africa (SSA), the World Health Organization鈥檚 (WHO) 2020 estimates report a maternal mortality ratio (MMR) of 536 maternal deaths per 100,000 live births [1]. Unsafe abortion is one of the leading causes of maternal mortality [2], and the Western Africa region has one of the highest rates of abortion-related deaths in the world, at 225 deaths per 100,000 abortions [3].

Despite recent improvements, Sierra Leone continues to have a high MMR; for every 100,000 live births in the country, an estimated 443 women die [1]. In 2015, the WHO estimated that the lifetime risk of maternal mortality in Sierra Leone was 1 in 17 [4]. Abortion in Sierra Leone is highly restricted and only legally permitted when the woman鈥檚 life is in danger [5]. A recent nationally representative mortality survey estimated that about 7% of maternal deaths in the country are attributable to unsafe abortion [6]. However, this proportion is likely underestimated as it is based on maternal death registrations, which are often characterized by challenges in quality and incompleteness. Sierra Leone鈥檚 performance on other sexual and reproductive health indicators likely contributes to these high rates of maternal mortality; for one, Sierra Leone has one of the highest rates of teenage pregnancy in the world: according to the 2019/2020 Sierra Leone demographic health survey, 21% of girls between the ages of 15 and 19 have begun childbearing (either pregnant or have given birth), and unintended pregnancies are common [7]. Additionally, an estimated 25% of women have an unmet need for modern contraception and this proportion is even higher among young women and adolescents [8]. In 2019, an estimated 17% of births in Sierra Leone were either unwanted or mistimed, and 51% of unintended pregnancies ended in induced abortion [7].

While complications are more likely to occur in induced abortions that are performed using non-recommended methods, miscarriage can also lead to adverse health outcomes [9]. Thus, understanding severe and near-miss complications related to induced and spontaneous abortion alike can be particularly helpful in elucidating delays to care and advocating for safe abortion care. Documenting the severity of abortion-related complications is especially important in understanding the causes of rarer outcomes like maternal deaths, pinpointing specific drivers of poor outcomes, and strengthening programs targeted at improving in-facility quality of care. Further, such information can provide the rationale for legislative changes to improve access to safe abortion care and post-abortion care (PAC). In 2015, lawmakers in Sierra Leone successfully passed a Safe Abortion Bill, which would have improved access to safe and legal abortion and PAC. However, the bill ultimately failed to become law following strong objections from religious leaders in the country [10]. Since then, the Ministry of Health and Sanitation (MoHS) in Sierra Leone initiated renewed efforts to formulate the Safe Motherhood and Reproductive Health Bill, which aims to improve access to family planning and contraceptive services, as well as expand grounds to access safe abortion care services within the formal health system [11]. Further, the MoHS and other stakeholders have developed comprehensive PAC clinical guidelines (2022), with training of providers in health facilities ongoing.

Despite the need for data on the severity of post-abortion complications, little is known about these outcomes in Sierra Leone. The most recent study on the severity of abortion-related complications in the country was conducted in 2011, which estimated that 21% of patients had moderate or severe post-abortion complications [12]. However, that study has several limitations that limit its usefulness in the current environment; the study utilized a small sample of only 21 public facilities (district and regional-level facilities), used now outdated severity categorizations, and relied solely on medical records which are often incomplete or missing altogether. These limitations may have resulted in this previous study underestimating the true severity of post-abortion complications in Sierra Leone. As such, there is a need for robust and up-to-date evidence on the severity of abortion-related complications in Sierra Leone.

The purpose of this study is to provide nationally representative evidence on the severity of abortion-related complications in Sierra Leone. We do this by analyzing data from a nationally representative sample of health facilities that provide PAC in Sierra Leone and classifying post-abortion complications into five categories: mild, moderate, severe, near miss, and death. Patients included in the study could be experiencing pregnancy loss from an induced abortion, where an intervention was done to intentionally end the pregnancy, or from a spontaneous abortion (aka a miscarriage). Results from this analysis can be used as evidence in support of ongoing reforms in the country, as well as to evaluate their effectiveness in the future.

Methods

Data sources and sampling

Data used in this analysis are part of a larger study that used the Abortion Incidence Complications Method (AICM) to understand the provision of PAC and estimate the incidence of unintended pregnancy and induced abortion in Sierra Leone [13, under review]. We conducted the AICM study from August to September 2021. Components of this study included a nationally representative survey of health facilities (HFS), as well as a Prospective Morbidity Survey (PMS). The PMS consists of interviews with consenting PAC patients and their providers about complications resulting from spontaneous and induced abortions, delays in care, and treatment received over a 30-day observation period at each health facility included in the study. The PMS has been widely used in other AICMs; the PMS used for this study was adapted from the version used by Madziyire et al. [14] to align with the health system structure and provider profiles in Sierra Leone (Supplemental File 1).

The study health facility samples were drawn from a masterlist of 1,440 health facilities provided by the Ministry of Health on January 31, 2021. First, we selected a nationally representative sample of health facilities to participate in the HFS. Fifty-four facilities were excluded from the masterlist because they were deemed non-eligible for participation for being specialized facilities (e.g., psychiatric or neurological hospitals and others). Of the remaining 1,386 eligible facilities, we stratified facilities by region and facility level and randomly sampled health facilities for the HFS. We therefore included all national referral, regional and district-level facilities, as well as all Marie Stopes International (MSI), and Planned Parenthood Association of Sierra Leone (PPASL) facilities in the study. Additionally, we selected 15% of community health centers (CHCs), maternal and child health posts (MCHPs) and community health posts (CHPs), resulting in a final HFS sample of 442 facilities.

The PMS facility sample is a sub-sample of the HFS sample and was also designed to be nationally representative. We randomly sampled a total of 302 health facilities to participate in the PMS. Of all the sampled facilities, 247 sent healthcare providers for the PMS training to serve as interviewers in their facilities; these providers collected data over 30 consecutive days. A total of 142 facilities reported PAC cases during the period of observation. To account for the sampling strategy and non-response in our analysis, we created facility-level weights by generating composite weights. We obtained these composite weights by multiplying the sampling weights by the facility-level non-response weights. This approach allowed us to adjust appropriately for the complexities associated with our sampling design and the potential biases introduced by non-response.

At each sampled facility, interviews were conducted with both PAC patients and their healthcare providers. All patients presenting for PAC at selected facilities during the study period were eligible for inclusion, regardless of whether the abortion was induced or spontaneous. All stable PAC patients were approached with information about the study and administered informed consent; parental/guardian consent was waived in the case of minors due to their status as emancipated minors and the possible harm of disclosure to third parties. PAC patients could decline to participate in the patient survey but give consent to their provider to answer questions about their care. In the rare case that a patient was unable to provide informed consent because she died or was in a coma for the duration of the study period, study staff interviewed the patient鈥檚 provider only. Sierra Leone鈥檚 Ministry of Health and Sanitation Ethics and Scientific Review Committee reviewed and approved the study protocol. Further approvals were obtained from the district health offices and each facility. The African Population and Health Research Center (APHRC) also reviewed and approved the study.

Data collectors recorded a total of 605 PAC patients throughout data collection. Of these, 522 (86%) consented to be interviewed and completed a patient interview. Forty-four (7%) consented but did not complete their interviews, 30 (5%) refused, and 8 interviews (1%) were not conducted or completed for an unspecified reason. One person died before she could be interviewed. Of the 522 patients interviewed for the PMS, only two (<鈥�1%) did not consent to data collectors interviewing their providers. As such, our final sample size was 514 provider interviews (all women who consented, plus the one maternal death), and 513 patient interviews.

Measures

To assess post-abortion complication severity, we applied the classification system used by Madziyire et al. (2018) in their study on abortion complications in Zimbabwe [14]. The Madziyire classification builds on the original conceptualization of complication severity presented by Rees et al. (1997) [15] by incorporating the WHO near-miss criteria to better distinguish severe complications and categorizes severity as either mild, moderate, severe, near-miss, or death. The analyses presented in this paper use the same criteria as those used by Madziyire et al. with some exceptions due to limitations in our data: (1) to be classified as 鈥渘ear-miss鈥� in the Zimbabwe study, a blood transfusion of two units was required. However, the number of units of blood transfused was not captured in the Sierra Leone PMS tool. As such, we removed blood transfusion from the near-miss criteria to avoid overestimating near-miss by counting cases requiring less than two units of blood; (2) because hemoglobin levels were not collected, we were unable to determine severe anemia. Due to the high levels of anemia in Sierra Leone [16], we removed anemia from the criteria for severe complication, with the exception of 10 cases that were classified as having anemia, hemorrhage and blood transfusion. Similar to Madziyire et al.鈥檚 classification, we expanded the definition of normal temperature to more accurately capture very low temperatures and avoid categorizing patients with very low temperature and no other clinical symptoms as having severe complications. Severity classifications were mutually exclusive, meaning that a patient cannot be classified under more than one category; women were categorized in the highest level of severity for which they met the criteria. Figure听1 displays the indicators we used to classify cases into each of the five categories for this analysis.

Fig. 1

Criteria for classification of abortion-related morbidity

We also collected information on sociodemographic characteristics of PAC patients, including age, education, marital status, religion, and place of residence. We collected household characteristics similar to those in the Demographic Household Survey (DHS) to measure wealth [17]. We calculated wealth differently for rural versus urban women, given the variation in household characteristics by place of residence. The wealth variable describes the wealth of our sample compared to DHS-derived cut points. We collected information on PAC patients鈥� reproductive health histories, including total number of pregnancies, pregnancy intention, pregnancy avoidance, and self-reported induced abortion. We defined trimester using clinician-reported estimates of gestational weeks based on the last menstrual period, with first trimester defined as pregnancies less than 13 weeks and second trimester 13 to 27 weeks.

We also collected information on delays to care experienced by patients accessing PAC, modified from the Three Delays model originally applied to maternal mortality programs by Maine et al. [18]. Our five delays fell into two groups: non-health system delays and health facility-based delays. Women were coded as experiencing a non-health system delay in care if they took more than 24听h to realize care was needed from the onset of symptoms, took more than 24听h to decide to seek care at a health facility once they realized care was needed, or took more than six hours to arrive at the health facility after they decided to seek care. The second set of delays are those related to the care provided by the health facility. Waiting more than one hour to be attended to after arrival was classified as a delay in care, as well as the receipt of complete care requiring longer than 12听h. In addition, we also assessed the reasons for non-health system delays in care. All women were asked about reasons for delays in seeking care from a doctor or health facility (Delay 2) while only respondents who said it took days or weeks to get to the facility were asked about their reasons for delays in arriving at the health facility (Delay 3). These were asked as open-ended questions and interviewers coded responses into response categories. The top three most common reasons for these two delays are presented in this paper; all possible reasons are listed in Supplemental Table 1. We did not ask about reasons for delay in time to realizing healthcare was needed (Delay 1).

Analyses

We first present descriptive statistics for selected sociodemographic and reproductive health characteristics of the PAC patients in our final sample. Next, we present the distribution of post-abortion complication severity among our sample. Since more severe complications are likely to be treated at higher level facilities, we applied facility-level weights to our severity estimates to account for the complex sample design and to account for variation in PAC severity based on facility level. Next, we present data on women鈥檚 experiences with delays in seeking and obtaining PAC and conduct bivariate chi-square tests of association between these factors and the facility level at which the women received care.

Finally, we investigated whether sociodemographic characteristics and delays to care are associated with complication severity. To do this, we collapsed severity into a two-category variable (mild/moderate morbidity vs. severe/near-miss complications or death). Since the proportion of patients with severe outcomes was greater than 10%, we used Poisson regression models with robust variance estimation to assess the relationship between each sociodemographic characteristic or delay to care with having the most severe outcomes, controlling for facility level [19]. We calculated incidence rate ratios with 95% confidence intervals to assess whether each characteristic was associated with more severe complications.

Results

Unweighted sociodemographic characteristics of the women in our sample are presented in Table听1. Women presenting to facilities for PAC ranged in age from 11 to 45 years, with most (59%) between the ages of 20 and 29. One in five (20%) were adolescents 19 or under. Half of our sample (50%) had some or completed secondary education, and approximately one in four (24%) reported having no education. Slightly more than half of our sample lived in a rural area of Sierra Leone (56%), while 44% lived in an urban setting and 1% were from outside Sierra Leone. Marital status was fairly evenly distributed across our sample, with 28% being single, divorced, separated, or widowed, 37% in a partnership and 36% legally married. Of those reporting a religion, a majority (64%) identified as Muslim. Respondents in our sample were wealthier when compared to the national distribution of wealth, with 76% falling in the wealthy or wealthiest quintiles. Most respondents (61%) were interviewed at primary care facilities, with 25% and 14% at secondary and tertiary facilities, respectively. A large majority (79%) were identified by their clinicians as being in the first trimester of pregnancy.

Respondents were also asked about their reproductive histories and pregnancy intentions. For one-third (32%) of our sample, the pregnancy that brought them to the facility was their first. When asked about pregnancy intentions for the index pregnancy, only 43% reported wanting the pregnancy when they became pregnant. Conversely, 31% reported that they desired the pregnancy later, and 21% did not want it at all. Approximately 43% of respondents were taking measures to avoid or delay pregnancy at the time they became pregnant. More than one in four respondents (29%) reported having done something to intentionally end the pregnancy that led them to seek care.

Table 1 Characteristics of the sample (N鈥�=鈥�513)*

We observed high levels of abortion-related morbidity in our sample (Table听2); 27% of women were classified as having severe abortion complications and 8% were classified as near-miss. One person died. Nearly one in four PAC patients (23%) had septic abortion, and 36% had clinical signs of infection. In approximately one-third of cases (35%), the provider reported evidence of a foreign body or mechanical injury. Having a septic abortion, clinical sign of infection, or foreign body/mechanical injury can be used as a proxy for potentially unsafe abortion as these complications are more common in cases of unsafe abortion; a majority of our sample (63%) was reported to have at least one of these clinical indications.

Table 2 Severity incidence (N鈥�=鈥�514)

Substantial proportions of women experienced non-health system-based delays in care seeking (Table听3). Slightly less than half of PAC patients (46%) took 24听h or more to realize care was needed after the initial onset of complication symptoms. Once patients realized care was needed, 32% took more than 24听h to decide to seek care, and 35% took six hours or longer to arrive in the health facility after deciding to seek care. Delays in receiving care once arriving at the facility were somewhat less common; 26% of patients had to wait more than one hour to be attended to by facility staff, and 28% received complete treatment 12听h or more after first being attended to (all non-weighted percentages). The only delay with significant differences across facility levels was the delay between time to arrival in the facility after deciding to seek care (p鈥�<鈥�0.05), with smaller proportions of patients reporting this delay in tertiary facilities compared to primary facilities. The most common reason women provided for experiencing a non-health system delay (deciding to seek care and arriving at a health facility) was a lack of money (43% and 34%, respectively). Women also cited fears that someone would find out that they were experiencing a post-abortion complication causing delays in deciding to seek care (14%) and arriving at a health facility (25%). Reasons for delays did not differ significantly by facility level with the exception of delays in arriving at the health facility due to lack of money (p鈥�<鈥�0.05), with larger proportions of women reporting this delay in primary facilities.

Table 3 Women鈥檚 experiences with delays to obtaining post-abortion care, overall and by facility level (N鈥�=鈥�513)*

After controlling for facility level, the risk of experiencing a severe complication, near-miss complication, or death was not significantly associated with the patient鈥檚 sociodemographic characteristics, with the exception of age and number of pregnancies. Compared to adolescents 19 and under, PAC patients between the ages of 20 and 24 had significantly lower risk of a severe/near-miss complication or death (aIRR 0.59; 95% CI: 0.38鈥�0.94); compared to PAC patients experiencing their first pregnancy, patients with 2鈥�4 total pregnancies had significantly higher risk of a severe/near-miss complication or death (aIRR 1.45; 95% CI: 1.03鈥�2.06) (Table听4). Our results show that delays in accessing care were significantly associated with having more severe complications. Controlling for facility level, patients were significantly more likely to suffer the most severe complications if they had to wait longer than one hour to be attended to at the health facility (aIRR 1.78; 95% CI: 1.20鈥�2.63), or did not receive complete treatment within 12听h (aIRR 1.76; 95% CI: 1.18鈥�2.64).

Table 4 Adjusted incidence rate ratios for relationship between each sociodemographic and delay to care variable and post-abortion complication severity, adjusting for facility level

Discussion

Our study shows high levels of abortion-related morbidity among PAC patients in Sierra Leone compared to other countries in the region. Approximately 64% of women were classified as either having moderate or more severe complications, and just under 10% of PAC patients experienced a near-miss complication or death. Comparing the study estimates to those from other countries and settings highlights the elevated rates of abortion-related morbidity in Sierra Leone. The severity of abortion complications in Zimbabwe (40% moderate or severe) [14] and Malawi (27% moderate or severe) [20] are much lower than our estimates for Sierra Leone. A recent WHO multi-country study on abortion-related complications across 11 Sub-Saharan African countries estimated that approximately 1.9% of cases were near-miss and 7.0% were potentially life threatening. While the criteria used for these classifications differ slightly from ours, these rates are much lower than this study鈥檚 near-miss and severe complication estimates for Sierra Leone (7.5% and 26.5%, respectively) [21].

This study鈥檚 national estimates for Sierra Leone are similar to those reported in conflict-affected settings in Nigeria (74% moderate and severe) and Central African Republic (CAR) (68% moderate and severe), where poorer health outcomes due to the breakdown of health systems can be expected [22]. Further, severity estimates from Nigeria and CAR are based on data from two referral hospitals which can treat and therefore receive more severe cases, meaning our national Sierra Leone estimates remain high in comparison. Finally, our results are similar to those reported in Kenya (77% moderate or severe) and in the Democratic Republic of the Congo (62% moderate or severe); however, these studies utilized 3- and 4-category severity classifications with broader inclusion criteria that likely overestimated severe morbidity [23, 24].

One explanation for the high levels of moderate and more severe abortion complications observed in this study is that Sierra Leone may also have a high incidence of unsafe abortion [6]. A recent 2021 study estimated an induced abortion incidence rate of 44.2 per 1,000 for Sierra Leone, although that study did not estimate what proportion of those abortions were unsafe [13]. However, evidence suggests that contexts with more restrictive abortion laws, such as Sierra Leone, also have higher incidence of severe abortion complications [22, 23, 25]. The reasons for this are likely that restrictive laws often lead people to pursue clandestine and unsafe abortions in the absence of safe options, and that abortion-related stigma causes barriers to accessing and providing abortion care. Evidence from this study supports both of these hypotheses. First, while it is often difficult to distinguish between post-abortion cases due to spontaneous versus induced abortions [26], complications such as septic abortion or infection (while potentially present in miscarriage), are much more common among unsafe induced abortions [27, 28]. Further, evidence of a foreign body or mechanical injury is almost certainly a result of unsafe abortion procedures. In this study, one in four post-abortion cases were septic, and approximately one third had signs of infection or evidence of a foreign body or mechanical injury. Providers reported at least one of these indicators in 64% of cases, suggesting that unsafe abortion may be responsible for a sizable proportion of the cases observed in this study. It is not surprising that this number is substantially higher than the 29% of PAC patients reporting induced abortion, given that induced abortion is often underreported in self-reported data [29, 30]. Second, fears that others would find out about their health problem was cited by women as a reason for both delaying seeking care and arriving at the health facility, signaling how abortion stigma may pose an obstacle to seeking PAC care.

A related explanation for the high levels of moderate and more severe complications in Sierra Leone is that women experiencing abortion-related complications may be seeking care outside of formal healthcare facilities, especially given legal restrictions that often beget fears of stigmatization and criminalization. It is possible that women with mild complications may prefer to self-treat or seek treatment from traditional or informal providers, only presenting to a facility if complications worsen. There is some evidence to support this hypothesis in our data: nearly one-quarter of our sample stated they went somewhere else first before coming to the health facility. Of these, 16% visited a traditional healer and 23% went to a pharmacy or drug store. However, it is not clear that this type of treatment pattern is more common in Sierra Leone than in other settings, and as such would not explain the higher levels of severe post-abortion complications observed in this study compared to other countries.

In addition to documenting the overall levels of the severity of post-abortion complications in Sierra Leone, we investigated whether certain sociodemographic characteristics were associated with having more severe complications, as this type of information can be useful in targeting policies and services to women experiencing higher levels of risk. However, our analysis indicated that most of the selected demographic or pregnancy indicators were not associated with having more severe complications. The two exceptions to this were age and number of pregnancies; PAC patients 20鈥�24 years old were significantly less likely to experience a severe or near-miss outcome or death when compared to PAC patients 19 years or younger, while PAC patients reporting 2鈥�4 total pregnancies had a higher risk of a severe of near-miss outcome when compared to PAC patients experiencing their first pregnancy. However, the fact that the other age and pregnancy bands did not show significant differences in severity weakens any conclusions about these characteristics and complication severity. Our lack of significant findings differs from other studies, which have found that numerous sociodemographic and pregnancy characteristics are associated with increased severity, including second trimester pregnancy, rural residence, being single, having less schooling, and being poor [14, 23, 24]. These studies generally had larger sample sizes than ours; it is possible that small cell sizes once we controlled for facility level may have obscured meaningful differences in this outcome by sociodemographic and pregnancy characteristics.

Investigating delays to care can help elucidate ways to minimize the occurrence of these delays as well as the risk of experiencing more severe complications. Overall, delays in care were common throughout women鈥檚 care trajectories. We found that larger proportions of women in primary care facilities reported experiencing delays in arriving at the health facility as compared to women accessing care at higher level facilities. This is likely because lower-level facilities predominate in rural areas where transportation options are limited. In addition, lack of money was cited as a major reason across both non-health system delays to care for which we probed the reason for delay, highlighting the close link between poverty and poor health outcomes that has been exhaustively documented [31,32,33,34]. Financial insecurity not only disincentivizes care-seeking where care is not free or is associated with other costs (including paying for medication, blood transfusions, or other unofficial costs), but may also mean that people are unable to pay for transport, miss work or lose wages, find childcare, or be away from their homes for an extended period. Indeed, nearly half of our sample said they or other members of their household lost income because of their health problem. While the Sierra Leonean free healthcare policy stipulates free care for pregnant women, lactating mothers, and children under the age of five, knowledge of this policy may be lacking, and may be inequitably implemented for PAC in rural settings where direct supervision is infrequent. Providers may also be more likely to request payment in situations where the abortion is perceived to be induced [35], and this can be a disincentive for such patients to seek care early. Finally, among respondents who delayed in seeking care from a doctor or health facility upon realizing they needed medical attention, nearly one-in-five said that one reason for delaying care-seeking was because they did not think the health problem was serious; providing education to women experiencing pregnancy complications on when care should be sought could help get patients to care faster and reduce the likelihood of complications becoming severe in the absence of treatment.

After controlling for facility level, we found that patients who experienced facility-based delays to care (delays in being initially seen; delays in receiving complete treatment) had a significantly higher risk of experiencing a severe outcome. As such, the findings from this analysis emphasize the need for investments in the Sierra Leonean health system and the importance of ensuring health facilities have both the material and human resources to respond to complications. Delays in the receipt of complete care are likely associated with severe outcomes due to the intensive care required to treat severe abortion-related complications, leading to longer stays in the facility. While safeguarding the health of pregnant people in Sierra Leone should be sufficient argument to invest in minimizing abortion-related complications, the argument also stands that mitigating the risk of severe complications means shorter stays and fewer costs to the health system, as is supported in existing literature [36,37,38].

Limitations

Our study has several limitations. For one, our results are weighted at the facility level but not at the individual level; as such, these findings are not nationally representative and should be interpreted with caution. Despite this, our moderate and severe complication estimates are high, even compared to referral hospitals in conflict-affected settings with higher maternal mortality ratios than in Sierra Leone. Data collectors did not maintain caseload tracking sheets, meaning we are unable to make any conclusions regarding the proportion of PAC cases at sampled facilities that we were able to interview. It is possible that some mild cases, which tend to require less time in the facility, were missed and therefore led us to overestimate the severity of abortion complications. Furthermore, 14% of PAC patients approached declined to be interviewed. Since we are not sure the reasons for their refusal, this could bias our results in either direction.

Additionally, criteria used by Madziyire et al. was able to identify severe anemia through data on hemoglobin levels and used the units of blood transfused to distinguish more severe cases of hemorrhage. The fact that our survey tools did not capture data on hemoglobin levels and the number of units of blood transfused limited our ability to comprehensively apply these same criteria. Classifying all anemic cases or blood transfusions into severe or near-miss categories would have vastly overestimated severe complications, but excluding massive blood transfusion and severe anemia from the criteria weakens our ability to meaningfully separate out more severe complications, and may have led to misclassifications and even underestimations of severe and near-miss cases. Another limitation is that severity estimates were based on data collected via provider interviews, and this data could be subject to reporting bias. We attempted to mitigate this by encouraging providers to review patient charts while providing complications data.

Additionally, we did not attempt to distinguish between spontaneous and induced abortions in our analyses. Nonetheless, we have several reasons to believe that our sample likely includes many PAC patients accessing care due to induced abortion, including the high incidence of severe complications in our study, the large number of cases with at least one indication of potentially unsafe abortion, the one-in-four PAC patients who self-reported trying to induce an abortion, the restrictive abortion legal environment, and the high unmet need for contraception in Sierra Leone. Lastly, as a facility-based study, our analysis only captures those patients with complications severe enough to bring them to the health facility, but not so severe such that they died before accessing care. This may skew our results to fewer mild complications and deaths than in the population at-large.

Conclusions

Our study shows high incidence of moderate and severe abortion-related complications among our sample of PAC patients in Sierra Leone. Evidence shows that abortion restrictions are associated with increased complication severity. Clearly, liberalizing the abortion law, and ensuring that safe abortion services are readily accessible, is a key intervention that would help to reduce the incidence of abortion-related complications and alleviate the strain of these complications on already under-resourced health facilities. In the absence of legal change, interventions focused on reducing the incidence of unintended or mistimed pregnancy, including access to contraception for those with an unmet need, are important. Furthermore, our analysis of complication severity and delays to care also emphasizes the need to ensure PAC services are always free of cost (including associated or unofficial costs), to educate women on the availability and legality of PAC, and to invest in the human and material resources facilities need to handle these cases appropriately and on-time. More research is needed on the severity of abortion-related complications of unsafe abortion broadly (not limited to in-facility PAC cases), and its population-level impacts.