Abstract

Background

Abortion-related complications are difficult to measure due to lack of standardized definitions and limited available data. We describe the proportion of abortive events that result in a documented complication in Mexico鈥檚 public sector hospitals.

Methods

We used ICD-10 codes from Mexico鈥檚 hospital discharge system (2018-2022), Subsistema Automatizado de Egresos Hospitalarios (SAEH), to describe abortive events admitted to hospitals: complications for excessive bleeding, infection, embolism, and unspecified; patient socio-demographic and clinical characteristics; and municipality-level structural vulnerability. We estimate complications by pregnancy duration, describe types of complications, identify characteristics associated with the presence of a complication using multuvariable regression, and calculate complication rates (proportion of abortive event that result in a complication treated in a public sector hospital per 1,000 women of reproductive age) by state in 2022.

Findings

There were 399,405 abortive events that received hospital-based care in Secretaria de Salud (SS) hospitals between 2018-2022. Ninety-two percent had no complication reported. The adjusted predicted probability of a complication was higher among patients at鈥�>鈥�13听weeks鈥� gestation (8.9%; 95% CI 8.1鈥�9.7%) compared with鈥夆墹鈥�13听weeks (6.6%; 95% CI 6.0鈥�7.2%). Higher parity, care at a tertiary hospital, and high marginalization at place of residence were positively associated with presence of a complication. States with higher complication rates are primarily in the central and southern regions.

Conclusions

In Mexico, 92% of patients who seek care for all abortive events (induced, spontaneous, post-abortion) in SS hospitals have no complications. Marginalized patients are more likely to have a complication and to seek care at later pregnancy durations. Routinely conflating care-seeking and complications leads to overestimates of the risk of abortion.

Introduction

Globally, indirect methods are used to estimate the proportion of abortions that have complications by using post-abortion care seeking, reported by providers, as a proxy [1, 2]; however, few studies rely on large samples or empirical clinical data to estimate the proportion of abortion cases that result in complications [3]. Lack of standardized definitions, limited available data, and stigma all hinder the measurement of abortions that end in complications [4]. Research has also documented that abortion clients seek post-abortion care in health facilities for reasons other than complications, such as confirmation of a complete abortion or reassurance about normal bleeding [5]. Therefore, the act of seeking care in a health facility is likely not always synonymous with experiencing a complication.

Lack of data on the proportion of abortions that have complications hinders the targeting of interventions to improve safety and may overstate the risks of abortion, especially for self-managed medication abortion, which may present in health facilities as post-abortion care or spontaneous abortion. Self-managed abortion, where individuals terminate their pregnancy with medications without formal clinical supervision [6], has been shown to be safe and effective [7], and has recently been incorporated into global abortion care standards at鈥�<鈥�13听weeks鈥� gestation, provided individuals have access to accurate information, quality medications, and health system referrals if needed [8]. Health facility-based treatments for abortive events (e.g. spontaneous, induced or incomplete) are the same 鈥� uterine evacuation 鈥� and are part of standard obstetrics and gynecology care [9]. Therefore, whether an abortive event is induced, spontaneous, or post-abortion care is not relevant to the treatment plan.

In Mexico, induced abortion laws are determined at the state level: nationally, abortion is legal in case of rape, but state legislation varies on the existence and implementation of other indications, such as health of the pregnant person and/or the fetus, or other exceptions [10]. Recent federal supreme court rulings have declared that state- (2021) and federal- (2023) level criminalization of induced abortion in the penal code is inconsistent with the Mexican constitution and to date, 13 states have decriminalized induced abortion [11]. Additionally, the court stated that care for induced abortion to the extent of current state law, as well as spontaneous and post-abortion care should be available to all who present at public sector hospitals. Federal health policy (the Programa de Aborto Seguro) [12] supports states in implementing safe abortion and post-abortion care services and provides clinical guidance (Lineamiento T茅cnico para la atenci贸n del Aborto Seguro en M茅xico) [13].

Research has focused on the Interrupcion Legal de Embarazo (ILE) program in Mexico City [14, 15], which began providing induced abortion services in the public sector in 2007 [16]. National health information systems are rare in low- and middle- income countries. In 2010 Mexico implemented the Automated Subsystem of Hospital Discharges, (Subsistema Automatizado de Egresos Hospitalarios) or SAEH of the Federal Ministry of Health (Secretar铆a de Salud, SS). These data are a national census of all SS public hospital admissions and have been used to examine abortion-related mortality [17] and abortion later in pregnancy (between 13鈥�24听weeks, both spontaneous and induced) [18]. However, there are no published data on complications associated with admissions for abortive events; such data are essential to understand whether care-seeking for spontaneous, induced or post-abortion care always implies a complication. These data can also inform national debates around the relative safety of self-managed and in-facility abortion care (spontaneous and induced) and support ongoing quality improvement efforts in Mexico.

We leverage SAEH data to estimate the proportion of all abortive events (spontaneous, induced, post-abortion care) that have a documented complication across all of Mexico鈥檚 public sector (SS) hospitals. We describe complications as a proportion of all events by pregnancy duration (+/- 13 weeks), and identify individual, clinical, and municipality of residence characteristics associated with the presence of a complication. We also calculate complication rates by state.

Methods

Data and variables

We used 2018鈥�2022 data from Mexico鈥檚 hospital discharge system, SAEH (Subsistema Automatizado de Egresos Hospitalarios), which includes all 45,825 hospitals in Mexico鈥檚 32 states that are operated by state and federal ministries of Health (Secretar铆a de Salud, SS) (see Supplemental Table S1 for data sources). These facilities largely serve individuals covered by public insurance programs for low-income and informal sector workers or without health insurance [19]. Hospitals must submit data monthly, and a reduced anonymized dataset is made publicly available. Our data do not include primary care level facilities, nor the publicly managed systems for formal sector and state employees (IMSS and ISSSTE), or private-sector facilities including hospitals, private physicians鈥� offices or NGO clinics. We chose to focus on public sector hospitals that serve the most structurally vulnerable populations for this analysis. Abortion care is not routinely provided in the primary care level outside Mexico City鈥檚 ILE program, so we do not include primary care facilities and only analyze hospitalizations. Consequently, our focus is on complications for abortive events seen in hospitals, not on incidence of complications overall, or the incidence of induced abortion. Our study followed the STROBE guidelines for observational studies.

We first extracted all abortive events: admissions that included International Classification of Diagnostic Codes Version 10 (ICD-10) codes O02-O08 and Z303 (which is sometimes used to code for legal induced abortion) for reason for admission (Supplemental Table S2). This analysis focused on abortion-related complications; therefore, we excluded molar and ectopic pregnancies, ICD-10 codes O00 and O01 since they are pathological pregnancy codes. We included years 2018-2022; the SAEH system has undergone several rounds of changes to data collection, prohibiting inclusion of earlier years.

Our primary outcome is the existence of a complication and the type of complication. We identified complications using the fourth digit of ICD-10 codes O03-O07 and include: excessive bleeding, infection, embolism, and unspecified complication as well as all O08 codes. No complication reported includes ICD-10 codes (O3-O7) fourth digit for none as well as codes for abnormal products of conception (O02), and menstrual extraction (Z303) (see Fig.听1 for variable construction). We also identified abortive event procedure type (aspiration, sharp curettage, and medication; collapsed into medication and procedural for modeling).

Fig. 1

Abortive event complication variable construction from ICD-10 codes

This figure displays the ICD10 codes and how they are categorized to make up our complication variable

We collapsed pregnancy duration in weeks鈥� gestation as less than or equal to 13听weeks and over 13听weeks as our key independent variable. We also included additional information available in the discharge record: patient age in 5听year age categories from 10 to 50鈥�+鈥�(collapsed into 10听year categories for modeling); whether the patient speaks an indigenous language (yes/no; the Mexican government鈥檚 standard measure of indigenous ethnicity); total number of pregnancies and births (0, 1, 2鈥�4, 5鈥�9, 10鈥�+); hospital care level (secondary, which provide specialized medical care or tertiary, which provide advanced treatments and care for complex conditions according to the federal government classification); patient municipality (administrative unit akin to a county) of residence; and state of residence (collapsed into regions, per standard classifications [20] for modeling: North, Central, Mexico City, and South).

At the patient municipality level, we merged in socio-economic indicators from the census acquired from multiple publicly available sources (see table Supplemental Table S1 for data sources). We used population size (<鈥�15听K, 15鈥�99听K, 100听K鈥�+); a multidimensional wealth index (grado de marginaci贸n), a standard measure used by the Mexican government, in quintiles, that includes measures of education, income, household materials, and the proportion of rural population (collapsed as 鈥渓ess marginalized鈥� or the bottom two quintiles and 鈥渕ore marginalized鈥� containing the top three quintiles); a binary measure of municipality-level education (鈥渕ore educated鈥� or鈥�<鈥�40% of 15听year old residents and older lack a basic education and 鈥渓ess educated鈥� or鈥夆墺鈥�40% of 15听year old residents and older lack a basic education); and municipality-level adolescent fertility rate in dichotomous categories (鈥渓ow鈥� or鈥�<鈥�75 births per 1000 women aged 15鈥�19 and 鈥渉igh鈥濃�夆墺鈥�75 births per 1000 women aged 15鈥�19); this cut-off approximated the median.

We described missing data for all variables and present a heatmap of missing variables by year in Supplemental Figure S1. Type of abortion procedure represents the variable with the most missing values, followed by weeks鈥� gestation. In order to retain these variables in our analyses, we included 鈥渕issing鈥� as a category in all models (described below). All other variables had less missing data (<鈥�10%).

Analysis

The unit of analysis is the abortive event. First, we described patient sociodemographic, clinical, hospital, and municipality-level characteristics by weeks鈥� gestation at abortive event (鈮も��13听weeks,鈥�>鈥�13听weeks). Next, we described types of abortion procedure and complications reported for hospital-based care for abortive events by weeks. To illustrate the stability of these data over time, we show procedures and complications by year (2018-2022) in Supplemental Table S3. We also described the abortive event ICD-10 codes by presence or absence of a reported complication. We conducted cross-tabulations of variables by our outcome (complication reported vs no complication reported) to help select variables for model-building (Supplemental Table S4). We then built a logistic regression model with the dichotomous outcome complication reported (yes/no) adjusted for weeks鈥� gestation, age, indigenous language, parity (previous number of births), type of abortion procedure, level of care, municipality population size, marginalization, adolescent fertility rate, region, and year. To improve the interpretability of our results, we calculated the absolute adjusted probability of a reported complication, using average marginal effects. Finally, we calculated state-level abortion complication rates (complications among those who sought care in a public sector hospital for an abortive event per 1,000 women of reproductive age, 15-45听years) and created a heat map of this complication rate by state in 2022. The number of women of reproductive age used as the denominator comes from the National Population Council's demographic projection data (Supplemental Table S1). All analyses were conducted in R (R Core Team, 2023) and the heat map was created with geographical data from DIVA-GIS (Supplemental Table S1). This analysis was deemed non-human subjects by the Oregon Health and Science University Institutional Review Board.

Results

In total, there were 399,405 abortive events that received hospital-based care in a Secretaria de Salud (SS) hospitals between 2018-2022. Three quarters of these events took place at鈥夆墹鈥�13听weeks of gestation (n鈥�=鈥�294,142; 74%) although 12% of data is missing for this variable. The largest age category represented for the patient is 20-24听years old (27%) and few patients spoke some indigenous language (2.5%). Around half (54%) of patients reported 2-4 pregnancies; one quarter (25%) indicated 2-4 previous births (Table听1).

Table 1 Patient characteristics of public hospital-based care for abortive events by pregnancy duration, Mexico 2018鈥�2022

At the hospital care-level, most care occurred at secondary facilities (76% overall). At the municipality level, patients were similar by pregnancy duration; however, slightly more patients at鈥�>鈥�13听weeks鈥� gestation came from rural municipalities (<鈥�15听K residents; 6.4%) compared to patients at鈥夆墹鈥�13听weeks of gestation (5.9%, p鈥�<鈥�0.001). Similarly, more patients at鈥�>鈥�13听weeks鈥� gestation came from more marginalized municipalities (20%) compared to patients at鈥夆墹鈥�13听weeks of gestation (18%, p鈥�<鈥�0.001). Overall, most (40%) patients seen in hospitals for abortive events resided in the central region of the county, followed by the south (32%), the north (23%), and Mexico City (5.1%) (Table听1).

Of all patients who sought care for abortive events in SS public-sector hospitals, the overwhelming majority (92%) had no reported complication. The proportion of abortive events with a complication was stable over time (2018-2022); additional temporal data are in Supplemental Table S4. The majority of abortive events were procedural (75%) with over half of these procedures conducted by sharp curettage (51%) and the remaining by aspiration (24%). Less than 1% of abortive events occurred with medications (misoprostol or other), although there was 24% missing data for this variable (Table听2). A higher proportion of abortive events at鈥�>鈥�13听weeks of gestation (9.3%) had a reported complication, compared to鈥夆墹鈥�13听weeks of gestation (6.7%; p鈥�<鈥�0.001). The proportion of cases that were treated using sharp curettage were higher for patients at鈥�>鈥�13听weeks of gestation (61%) compared to patients at鈥夆墹鈥�13听weeks of gestation (50%;鈥�<鈥�0.001) (Table听2). We show ICD-10 codes by presence of a complication in Table听3.

Table 2 Procedures and complications reported for hospital-based care for abortive events by pregnancy duration, Mexico 2018鈥�2022

Table 3 Abortive event ICD-10 codes by reported complication, Mexico 2018鈥�2022

In multivariable analyses, the adjusted predicted probability of a reported complication was higher among patients at鈥�>鈥�13听weeks of gestation (8.9%; 95% CI 8.1鈥�9.7%) compared to patients at鈥夆墹鈥�13听weeks of gestation (6.6%; 95% CI 6鈥�7.2%; Table听4). Higher parity and care at a tertiary hospital were positively associated with presence of a complication. At the municipality level, patients who came from more marginalized municipalities (top 3 quintiles) had higher odds of a reported complication compared to less marginalized (bottom 2 quintiles) and patients from municipalities where residents were on average less educated (鈮モ��40% of residents aged 15 or older lacked basic education) had higher odds of a reported complication compared to municipalities where residents were more educated. See Table听4 for full model results.

Table 4 Multivariable logistic modeling results of a reported complication of abortive events from patients who sought hospital care, Mexico 2018鈥�2022

Finally, the national rate of abortive events that were seen in public sector SS hospitals (cases attended for all abortive events, with or without complications, not incidence) was 11.3 per 1000 women of reproductive age in 2022; however, the rate of abortive events with complications documented in SS public sector hospitals was 0.12 per 1,000 women of reproductive age (data not shown). We show state-level complication rates for 2022 in Fig.听2. Most states show low complication rates; states with higher rates are located primarily in the central and southern regions (Queretaro, Puebla, Michoac谩n, and Chiapas).

Fig. 2

Rate of abortive event complications at public hospitals per 1000 women of reproductive age, Mexico 2022

This heat map illustrates the rate of abortive event complications in public sector hospitals per 1000 women of reproductive age in each state

Discussion

Using individual-level public sector SS hospital data from Mexico over 5听years, we showed that the overwhelming majority (92%) of patients who received hospital-based care for an abortive event had no complications reported. We found the adjusted predicted probability of a reported complication to be lower (6.6%; 95% CI 6-7.2%) for patients at鈥夆墹鈥�13听weeks鈥� gestation compared to at鈥�>鈥�13听weeks鈥� (8.9%; 95% CI 8.1-9.7%). Nationally, in 2022, we found the rate of all abortive events attended in public sector SS hospitals to be 11.3 per 1000 women of reproductive age; it is important to clarify that, in this study, we do not intend to directly or indirectly estimate overall incidence of abortion. The corresponding national rate of complications attended in public sector SS hospitals was 0.12 per 1000 women of reproductive age; similarly, this is not an estimate of incidence of all abortion complications; it is rather the rate of complications among cases seen in public hospitals.

Our data lead us to argue that hospitalizations for abortive events should not be routinely counted as an indicator of abortion complications. We show that most individuals who seek hospital-based care do not have a complication and that complications are associated with different markers of marginalization. However, 7.6% of patients did have documented complications, which is higher than would be expected given the well-documented safety of abortion [7, 21], and lower rates documented in hospital settings in the United States (2.1%) [22] and from a population-based study in Canada (<鈥�1%) [23]. Our finding of a higher probability of complications at 13听weeks or more supports previous Latin American literature [3], although existing literature focuses only on samples with complicated hospitalized cases.

Our estimate of complications (0.12/1000 women of reproductive age) is much lower than previous work in Mexico (8.1/1000 reproductive age women in 2009), which used all hospital admissions with an abortive event ICD-10 code as an indicator of the presence of a complication, resulting in a higher numerator; previous work has not examined ICD-10 codes for complications specifically [2]. Our analysis improves upon the methodology for estimating complications by examining the fourth digit of the ICD-10 codes where a fourth digit of 鈥�4鈥� or 鈥�9鈥� specifically indicates 鈥渨ithout complication,鈥� and therefore cases with these codes applied should be excluded from any calculations of complications. [2]

Misoprostol is safe and effective [7] and is used more widely in Latin America compared to other regions with reported high rates of reported complications [2, 24]. We hypothesize that misoprostol is likely an important determinant of the low complication rates we observe among abortive events admitted to public SS hospitals in Mexico; however, we are unable to identify which of the abortive event cases that present for care in public hospitals are self-managed induced abortions because ICD-10 classification was not designed to identify self-managed abortions. WHO abortion care guidelines state that routine follow-up care following an uncomplicated procedural or medication abortion is not required if the person has adequate information about when to seek care if needed [8]. However, some people who self-manage their abortion using misoprostol are told to seek care or wish to seek confirmation of complete pregnancy termination in a health facility. As the visibility of self-managed abortion continues to increase worldwide [25], more individuals who experience safe medication abortions may seek care in facilities for reassurance or confirmation of completion, making care-seeking as a proxy for complications a problematic metric of abortion complications.

Marginalized patients are more likely to have a reported complication and also seek care at later pregnancy durations, when greater barriers to care exist [26]. Our results are consistent with this, both at the individual-level, with individuals with high parity and who spoke some indigenous language (both markers of marginalization) and at the municipality-level, where individuals who come from municipalities that are more marginalized and have, on average, lower level of education were more likely to have reported complication. This is also consistent with previous work in Mexico that showed that public sector utilization of abortion later in pregnancy was higher for more marginalized patients [18]. Financial and logistical barriers, fear of stigma or prosecution, as well as seeking care for later gestational stage abortions can compound and extend delays [27]. Delays in care can additionally lead to denial of services due to legal restrictions on gestational limits [28]. We found that patients from rural municipalities had lower odds of a reported complication when compared to those from municipalities with鈥�>鈥�100听K residents which may be a marker of reduced access to care given that more secondary and tertiary-level facilities are present in urban areas.

Our findings also show that almost one fifth (16.4%) of the abortive events were coded as spontaneous abortions; we suspect that this code is applied when there is relative clinical certainty of a miscarriage. This proportion is higher than the commonly used estimates at the international level; but it corresponds to reported incidence of miscarriages in the USA [29]. We are not able to interpret with certainty the codes 鈥渙ther鈥� or 鈥渦nspecified鈥� abortion, which represent 60% of all abortive events. Anecdotical, qualitative information suggests that both codes are applied when people seek care for a suspected incomplete, self-managed abortion, when health providers and facilities prefer not to get involved (and not to involve the patients) in legal reporting. However, the treatment for spontaneous and induced abortion 鈥� uterine evacuation 鈥� is the same regardless of the diagnosis, and all health facilities that provide obstetrics and gynecology care should provide this essential service. We also show that overall, half of the abortive procedures in this study were treated by sharp curettage, a procedure that the WHO specifically recommends against, in favor of vacuum aspiration or multiple misoprostol doses [8]. Previous work in Mexico shows that persisting high use of sharp curettage in post-abortion care is driven by problems of availability and maintenance of supplies, lack of training and unfavorable attitudes toward newer, recommended technologies, and stigma [30].

Strengths of this study include the use of individual-level data over 5听years, and we use the 4th digit of the ICD-10 code to provide a more granular picture of the indication for care seeking. We also include weeks鈥� gestation and rich contextual-level data to assess sociodemographic factors. However, this study has limitations to keep in mind when interpreting the finding. First, variable coding practices may limit data quality and we are unable to interpret the causes and indications of 鈥渦nspecified鈥� abortion. Second, we are not able to distinguish between spontaneous, induced, and unspecified abortion in our data. However, the objective of the study is not to use data to indirectly estimate incidence of induced abortion, but to analyze the prevalence and type of complications in all abortion cases. Technically, the care delivered (uterine evacuation), is the same regardless of induced or spontaneous abortion. New, methodologies will be needed to estimate incidence of abortion and abortion complications in Mexico and Latin America given the wide use of misoprostol and the documented safety of self-managed medication abortion [7]. Third, complications can occur due to unsafe abortion, delays in seeking care, or from post-abortion care (i.e. sharp curettage) received in the hospital. In this study, we are unable to distinguish between these causes, in particular whether the complication is from the induced abortion or from the post-abortion care. Fourth, despite having a large, national dataset, state-level complication rates rely on small numbers which may result in instable estimates in some cases.

Conclusions

In Mexico, 92% of patients who seek care for abortive events (induced, spontaneous, post-abortion) in Secretaria de Salud (SS) public hospitals do not experience complications. Marginalized patients are more likely to experience a complication and seek care at later pregnancy durations. Routinely conflating care-seeking and complications leads to overestimates of the risk of abortion.