Persistent ethnic inequities in infant mortality in Peru, 2019–2024: a repeated cross-sectional analysis of ENDES birth histories

Abstract

Background

Despite progress in reducing infant mortality in Peru, inequities persist by socioeconomic and ethnic characteristics. Documenting these disparities is essential to guide public health policy.

Objective

To quantify ethnic disparities in infant mortality in Peru, examine variation by socioeconomic level and timing of death, and estimate the fraction of deaths attributable to these inequities.

Methods

We conducted a repeated cross-sectional analysis of ENDES 2019–2024 birth-history data. The analytic sample included 100,807 live births to women aged 15–49 years with complete opportunity for ascertainment of death before 12 months of age. Infant mortality was defined as death before 12 months among live-born children. Maternal ethnicity (Mestizo, Afro-Peruvian, Indigenous) was based on self-identification. We estimated infant mortality rates per 1,000 live births, risk ratios (RR), and risk differences (RD) using survey-weighted Poisson regression with robust variance.

Results

We recorded 1,178 infant deaths (overall infant mortality rate: 10.6 per 1,000; 95% CI: 9.8–11.3). The rate was 8.0 per 1,000 among Mestizo, 13.1 among Afro-Peruvian, and 13.6 among Indigenous births (p < 0.001). Children of Afro-Peruvian and Indigenous mothers had 65% (RR = 1.65; 95% CI: 1.31–2.07) and 71% (RR = 1.71; 95% CI: 1.46–2.00) higher risk of death before 12 months than Mestizo children. Disparities were more pronounced for postneonatal mortality (Indigenous RR = 2.53) and were observed across socioeconomic strata. The population attributable fraction for ethnic mortality disparities was 24.4%, under a counterfactual scenario in which all ethnic groups had the infant mortality rate observed among Mestizo births.

Conclusions

Marked ethnic disparities in infant mortality persist in Peru and are observed across socioeconomic strata. Approximately one quarter of infant deaths could be prevented under a counterfactual scenario in which ethnic mortality gaps were eliminated. Targeted interventions for Afro-Peruvian and Indigenous populations are needed.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12963-026-00500-6.

Introduction

Infant mortality is one of the most sensitive indicators of the social and health conditions of a population [1]. Sustainable Development Goal 3.2 sets the target of reducing neonatal mortality to 12 or fewer per 1,000 live births and under-5 mortality to 25 or fewer per 1,000 by 2030 [2]. Although Latin America has experienced substantial reductions in these rates over recent decades, marked disparities persist between and within countries, particularly across ethnic and socioeconomic groups [3, 4]. Multiple studies have shown that social determinants-including maternal education, access to health services, housing conditions, and household socioeconomic status-exert a predominant influence on child survival, often outweighing explanations based on intrinsic biological or genetic differences between population groups [1, 3].

Ethnic health inequities reflect the consequences of historical processes of discrimination, social exclusion, and unequal distribution of economic and political power [5]. In Latin America, Indigenous and Afro-descendant populations have been systematically marginalized, resulting in lower educational attainment, higher poverty, predominantly rural residence, and reduced access to quality health services [6, 7]. These cumulative disadvantages are reflected in poorer maternal and child health indicators, including higher infant mortality rates [4, 8]. In Brazil, a longitudinal study of 19 million newborns found that Afro-descendant children had a higher risk of mortality than their White peers, even after adjusting for socioeconomic factors [7]. Similar patterns have been observed among Indigenous populations in Bolivia, Colombia, Guatemala, and Peru [4].

Peru is a multiethnic country in which approximately 25% of the population self-identifies as Indigenous (Quechua, Aymara, or Amazonian) and about 3% as Afro-Peruvian [9]. The country has achieved remarkable progress in reducing infant mortality, from 38.6 per 1,000 in 2000 to 16.6 per 1,000 in the last decade, a change attributed to poverty-reduction policies, expanded health-service coverage, and conditional cash transfer programs [10, 11]. However, previous studies have shown that these gains have not been evenly distributed, with Indigenous and rural populations maintaining persistently higher rates [11, 12]. An analysis of four decades of Peruvian data revealed that infant mortality gaps between socioeconomic groups, although reduced, remain substantial [12].

Despite regional evidence on ethnic inequities in maternal and child health, systematic quantification of ethnic disparities in contemporary infant mortality in Peru remains limited. Prior studies have primarily emphasized urban–rural, regional, or socioeconomic gradients, with less attention to maternal ethnic self-identification as a central axis of inequity. The contribution of this study is threefold: it provides recent national evidence using repeated ENDES birth-history data; it evaluates whether ethnic disparities persist across socioeconomic strata and differ between neonatal and postneonatal mortality; and it quantifies the counterfactual fraction of deaths associated with ethnic mortality gaps. These findings may inform other multiethnic Latin American settings where Indigenous and Afro-descendant populations experience similar histories of social exclusion, territorial marginalization, and unequal access to quality health services. This study aims to quantify ethnic disparities in infant mortality in Peru, examine variation by socioeconomic level and timing of death, and estimate the fraction of deaths attributable to these inequities.

Methods

Study design

An analytical cross-sectional observational study was conducted using publicly available secondary data from Peru’s Demographic and Family Health Survey (ENDES). The analysis included six consecutive years of data (2019–2024). This manuscript was prepared following the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines [13], and the checklist is provided in S1.

Data source

ENDES is a population-based repeated cross-sectional survey conducted annually by Peru’s National Institute of Statistics and Informatics (INEI) with technical assistance from the USAID Demographic and Health Surveys (DHS) Program. The survey is nationally representative, representative at the departmental level (25 departments including the Constitutional Province of Callao), and by area of residence (urban and rural) [14].

ENDES uses a two-stage, probabilistic, stratified sampling design with independent samples within each department. In the first stage, clusters (primary sampling units) are selected with probability proportional to size. In the second stage, households are selected within each cluster using systematic sampling with a random start. All women aged 15–49 years who are usual residents of selected households are eligible for the survey.

For this study, we used the birth history module (REC21 file), which contains information on all live-born children of each interviewed woman, including date of birth, survival status, and, if deceased, age at death. The datasets are publicly available through the INEI portal.

Population and sample

The study population comprised live births to women aged 15–49 years interviewed in ENDES 2019–2024. The unit of analysis was the live-born child. ENDES records complete birth histories, including date of birth, survival status, and age at death for deceased children. Because the outcome was defined as death before 12 months of age, the main cumulative-incidence analysis was restricted to births with complete opportunity for ascertainment of infant death, defined as births occurring at least 12 months before the interview date. Eligible records also required valid maternal ethnic self-identification, complete survival status, and, for deceased children, recorded age at death. Because each ENDES survey records births occurring in the five years preceding the interview, the pooled dataset covered births occurring approximately from 2014 to 2024; after restricting to children with complete first-year outcome ascertainment, the effective analytic birth window was approximately 2014–2023.

We excluded records of births with missing information in the main stratification variables (area of residence, wealth quintile, and natural region). The final analytic sample comprised 100,807 live births with complete information and complete opportunity for ascertainment of death before 12 months of age.

Variables

The dependent variable was infant mortality, defined as death before 12 months of age among live-born children with complete 12-month outcome ascertainment. Age at death was obtained from the birth history questionnaire, which records age at death in completed days or months, depending on the age at death. Children who were alive and had reached at least 12 months by the interview date were classified as surviving the first year of life. Therefore, the outcome was interpreted as the risk, or cumulative incidence, of death before 12 months of age rather than as a prevalence state.

The main independent variable was maternal ethnic self-identification, obtained from the question: “Based on your ancestors and according to your customs, do you consider yourself…?” For analysis, the original categories were recoded into three groups: Mestizo (reference category), Afro-Peruvian, and Indigenous.

Stratification variables included: maternal educational level (no education, primary, secondary, higher); household wealth quintile (poorest, poorer, middle, richer, richest); area of residence (urban, rural); natural region (Metropolitan Lima, rest of Coast, Highlands, Jungle); place of delivery (home, health facility); adequate antenatal care (six or more antenatal care visits); maternal age at the child’s birth (15–19, 20–34, 35–49 years); child sex (male, female); and birth order (first, 2–3, 4 or more). Household wealth quintile was based on the ENDES/DHS household wealth index, a composite measure of cumulative living standard derived from household assets, housing materials, and access to basic services, and categorized into survey-specific quintiles [15]. Area of residence was defined according to the INEI classification of the survey cluster; Peruvian household surveys classify areas as urban or rural based on the official population-center and dwelling-pattern definitions used by INEI [16].

Statistical analysis

All analyses accounted for ENDES’s complex survey design using survey weights adjusted for pooling across years (original weight divided by the number of years included), strata, and primary sampling units (clusters). We used the single-unit-in-stratum option centered at the overall mean to handle strata with a single primary sampling unit.

First, we conducted a descriptive analysis of population characteristics by ethnic group. Categorical variables were summarized using absolute frequencies (unweighted) and weighted percentages with 95% confidence intervals (95% CI).

Second, we calculated overall and stratified infant mortality rates according to ethnic group and each sociodemographic variable, expressed per 1,000 live births with 95% CI. Ethnic-specific mortality rates were illustrated using a bar chart with 95% CI.

Third, we estimated crude risk ratios (RR) and absolute risk differences (RD) to quantify the magnitude of ethnic disparities in infant mortality. RRs were estimated using survey-weighted Poisson regression models with robust variance for a binary cumulative-incidence outcome. RDs were calculated as the arithmetic difference between the weighted infant mortality rate in each minority group and that in the reference group (Mestizo). We report RRs and RDs with 95% CI when available. We did not statistically adjust for socioeconomic variables (education, wealth, area of residence, and access to health services) because these variables are plausibly downstream of ethnicity-related structural disadvantage; including them in adjusted models would mask the inequities that the study aims to document rather than estimate the total ethnic disparity.

Fourth, we conducted stratified analyses to describe the intersectional patterning of ethnic disparities across socioeconomic strata. Infant mortality rates were calculated by ethnic group within each stratum of wealth quintile, area of residence, educational level, and place of delivery. These analyses were interpreted descriptively rather than as estimates of causal effect modification.

Fifth, we estimated additive-scale joint-pattern measures between ethnicity and socioeconomic factors. We calculated the relative excess risk due to interaction (RERI), the attributable proportion due to interaction (AP), and the synergy index (S). RERI was calculated as RR11—RR10—RR01 + 1, where RR11 is the risk ratio for the group with both exposures relative to the group with neither exposure. A RERI > 0 indicates a positive departure from additivity, whereas a RERI < 0 indicates a negative departure from additivity. We evaluated four combinations: Indigenous x poverty, Indigenous x rural residence, Indigenous x low education, and Afro-Peruvian x poverty. Given the observational design and the likely downstream position of socioeconomic variables, these measures were interpreted as descriptive departures from additivity across social strata rather than as causal interaction effects.

Sixth, we examined temporal trends in infant mortality by ethnic group during the study period. Annual rates were calculated for each group, and statistical interaction between ethnicity and year was assessed using a Poisson model including an ethnicity × year interaction term.

Seventh, we estimated the population attributable fraction (PAF) for ethnic mortality disparities to quantify the potential impact of eliminating ethnic mortality gaps. PAF was calculated as (Ptotal—Preference) / Ptotal, where Ptotal is the overall infant mortality rate and Preference is the rate in the Mestizo group. This indicator was interpreted as an equity-gap counterfactual fraction: the proportion of infant deaths that would be avoided if all ethnic groups had the infant mortality rate observed in the reference group, without implying that ethnicity itself is a modifiable biological exposure.

Finally, we performed sensitivity analyses to assess robustness: (a) excluding the year 2020 to account for the potential impact of the COVID-19 pandemic on death registration and access to health services; and (b) analyzing neonatal mortality (death before 28 days of life) and postneonatal mortality (death from 28 days to before 12 months) separately to examine whether disparities differed by timing of death.

All analyses were conducted using Stata version 17.0 (StataCorp, College Station, TX) and R version 4.3 for figure development.

Ethical considerations.

This study exclusively utilized publicly accessible and fully anonymized secondary data from the INEI institutional portal. Because the analysis used public, anonymized secondary data without personal identifiers, institutional ethics committee approval and additional informed consent were not required. Nevertheless, the study was conducted in compliance with the ethical standards of the Declaration of Helsinki.

Results

Flow diagram

Supplementary Fig. 1 shows the participant selection flow diagram. Of the 113,245 live births recorded in ENDES 2019–2024, 4,892 were excluded due to missing information on maternal ethnicity and 5,243 due to incomplete information on the child’s vital status or first-year outcome ascertainment. In addition, 2,303 records corresponding to low-frequency ethnic categories (White: 1,456; Other: 523; Don’t know: 324) were excluded to ensure stable estimates. The final analytic sample included 100,807 live births distributed across three maternal self-identified ethnic groups: Mestizo (48,409; 48.0%), Afro-Peruvian (12,574; 12.5%), and Indigenous (39,824; 39.5%). During the study period, 1,178 infant deaths were recorded, corresponding to an overall infant mortality rate of 10.6 per 1,000 live births.

Characteristics of the study population

Adolescent motherhood was more common among Afro-Peruvian (30.7%) and Indigenous (26.2%) mothers than among Mestizo mothers (22.4%), and higher education was less frequent in minority groups (Afro-Peruvian 22.9%; Indigenous 28.4%) compared with Mestizo mothers (42.6%). Minority ethnic groups also had more disadvantaged socioeconomic conditions: 58.4% of Afro-Peruvian and 60.7% of Indigenous households were in the two poorest wealth quintiles, compared with 35.0% of Mestizo households. Rural residence was more frequent among Indigenous (36.5%) and Afro-Peruvian (32.3%) participants than among Mestizo participants (15.2%), and facility delivery was less common among Afro-Peruvian (89.7%) and Indigenous (93.6%) births than among Mestizo births (95.6%). No differences were observed in the distribution of child sex across ethnic groups (Table 1).

Table 1.

Characteristics of the study population by ethnic group, ENDES Peru 2019–2024

Characteristic	Mestizo (n = 48,409)	Afro-Peruvian (n = 12,574)	Indigenous (n = 39,824)
Maternal age at childbirth, mean (SE)	24.5 (0.05)	23.1 (0.08)	23.8 (0.05)
Age group, % (95% CI)
15–19 years	22.4 (21.9–23.0)	30.7 (29.7–31.8)	26.2 (25.5–26.8)
20–34 years	68.6 (68.0–69.2)	62.5 (61.3–63.6)	66.3 (65.6–67.0)
35–49 years	9.0 (8.6–9.4)	6.8 (6.2–7.4)	7.6 (7.2–8.0)
Educational level, % (95% CI)
No education	0.3 (0.3–0.4)	1.3 (1.1–1.6)	1.4 (1.3–1.6)
Primary	11.4 (11.0–11.8)	24.4 (23.3–25.6)	18.2 (17.6–18.8)
Secondary	45.7 (45.0–46.3)	51.3 (50.1–52.5)	52.0 (51.3–52.7)
Higher education	42.6 (41.9–43.3)	22.9 (21.9–24.0)	28.4 (27.7–29.2)
Wealth quintile, % (95% CI)
Poorest	14.4 (13.8–15.0)	32.0 (30.7–33.4)	33.8 (32.8–34.8)
Poorer	20.6 (20.0–21.2)	26.4 (25.4–27.5)	26.9 (26.1–27.7)
Middle	21.6 (21.0–22.1)	20.1 (19.1–21.1)	18.7 (18.1–19.3)
Richer	22.3 (21.8–22.9)	13.7 (12.8–14.6)	12.7 (12.1–13.2)
Richest	21.2 (20.5–21.9)	7.8 (7.1–8.6)	8.0 (7.6–8.6)
Area of residence, % (95% CI)
Urban	84.8 (84.1–85.5)	67.7 (66.1–69.3)	63.5 (62.4–64.6)
Rural	15.2 (14.5–16.0)	32.3 (30.7–33.9)	36.5 (35.4–37.6)
Natural region, % (95% CI)
Metropolitan Lima	35.7 (34.6–36.8)	17.6 (16.4–18.9)	20.8 (19.7–21.8)
Rest of the Coast	30.6 (29.7–31.6)	45.7 (44.0–47.4)	10.4 (9.8–11.0)
Highlands	13.5 (12.9–14.1)	21.3 (19.8–22.9)	54.0 (52.6–55.3)
Jungle (Amazon)	20.2 (19.5–20.9)	15.3 (14.3–16.4)	14.9 (13.9–16.0)
Facility delivery, % (95% CI)	95.6 (95.3–95.9)	89.7 (88.7–90.6)	93.6 (93.0–94.2)
Adequate antenatal care (≥ 6 visits), % (95% CI)	89.2 (88.9–89.6)	87.8 (87.1–88.5)	87.4 (86.9–87.9)
Birth order, % (95% CI)
First	68.8 (68.3–69.4)	64.6 (63.6–65.7)	66.2 (65.6–66.8)
2–3	26.5 (25.9–27.0)	28.3 (27.4–29.3)	27.1 (26.6–27.7)
4 +	4.7 (4.5–5.0)	7.0 (6.5–7.6)	6.7 (6.4–7.0)
Child sex, % (95% CI)
Male	51.1 (50.5–51.7)	51.9 (50.8–52.9)	51.3 (50.7–52.0)
Female	48.9 (48.4–49.5)	48.2 (47.1–49.2)	48.7 (48.0–49.3)

CI, confidence interval; SE, standard error

All estimates are weighted for the complex survey design. p values are from design-corrected chi-square tests

Infant mortality rate

During 2019–2024, 1,178 infant deaths were recorded in the analytic sample, corresponding to a weighted infant mortality rate of 10.6 per 1,000 live births (95% CI: 9.8–11.3) (Table 2).

Table 2.

Infant mortality rate by ethnic group and sociodemographic characteristics, ENDES Peru 2019–2024

Characteristic	Deaths/total	Rate per 1,000	95% CI
Overall	1,178/100,807	10.6	9.8–11.3
Ethnicity
Mestizo	452/48,409	8.0	7.0–8.9
Afro-Peruvian	161/12,574	13.1	10.6–15.7
Indigenous	565/39,824	13.6	12.2–15.0
Educational level
No education	24/904	24.3	13.9–34.8
Primary	327/16,689	18.3	15.9–20.7
Secondary	576/49,960	10.7	9.6–11.8
Higher education	247/32,871	6.6	5.5–7.8
Wealth quintile
Poorest	471/28,380	16.5	14.8–18.2
Poorer	311/26,912	11.2	9.6–12.8
Middle	219/20,347	10.2	8.4–11.9
Richer	114/15,039	6.7	5.1–8.3
Richest	63/10,129	5.5	3.6–7.5
Area of residence
Urban	692/70,860	8.7	7.8–9.5
Rural	486/29,947	16.4	14.6–18.1
Natural region
Metropolitan Lima	64/11,860	5.0	3.5–6.4
Rest of the Coast	260/29,390	9.4	8.0–10.9
Highlands	500/34,040	14.9	13.4–16.5
Jungle (Amazon)	354/25,517	14.3	12.6–16.1
Place of delivery
Home	129/5,757	22.0	17.6–26.3
Health facility	1,021/93,763	9.7	8.9–10.5
Maternal age at childbirth
15–19 years	367/22,392	15.7	13.8–17.6
20–34 years	437/55,195	7.1	6.2–8.0
35–49 years	35/6,416	4.1	2.5–5.7
Child sex
Male	677/51,543	11.5	10.4–12.5
Female	501/49,264	9.6	8.5–10.7
Birth order
First	712/66,589	9.8	8.9–10.7
2–3	373/27,867	11.8	10.3–13.3
4 +	93/6,351	13.7	10.4–17.0

CI, confidence interval

Infant mortality was defined as death before 12 months of age among live-born children with complete 12-month outcome ascertainment. All estimates are weighted for the complex survey design

Marked differences in infant mortality were observed by ethnic group. The rate among Mestizo births was 8.0 per 1,000 (95% CI: 6.9–9.0), whereas rates were substantially higher among Afro-Peruvian (13.1 per 1,000; 95% CI: 10.9–15.4) and Indigenous births (13.6 per 1,000; 95% CI: 12.4–14.9) (p < 0.001). Figure 1 illustrates the magnitude of these disparities.

Fig. 1.

Infant mortality rate (per 1,000 live births) by ethnic group, ENDES Peru 2019–2024

Infant mortality showed a consistent socioeconomic gradient. Mortality was highest among mothers with no education (18.4 per 1,000), in the poorest wealth quintile (15.6 per 1,000), in rural areas (16.3 per 1,000), and in the Jungle region (15.5 per 1,000), compared with 5.1 per 1,000 among mothers with higher education and 4.6 per 1,000 in the richest wealth quintile (Table 2).

Place of delivery was strongly associated with infant mortality: home births had a rate of 27.5 per 1,000 compared with 9.1 per 1,000 for facility births. Higher rates were also observed among infants of adolescent mothers (15–19 years: 13.2 per 1,000) and among higher birth orders (≥ 4: 16.5 per 1,000).

Magnitude of ethnic disparities by mortality component.

Crude risk ratios confirmed substantial ethnic disparities in infant mortality (Table 3). Compared with the Mestizo group, children of Afro-Peruvian mothers had a 65% higher risk of death before age 1 year (RR = 1.65; 95% CI: 1.31–2.07), whereas children of Indigenous mothers had a 71% higher risk (RR = 1.71; 95% CI: 1.46–2.00). When neonatal and postneonatal deaths were examined separately, disparities were particularly pronounced in the postneonatal period, especially among Indigenous children (RR = 2.53; 95% CI: 1.99–3.22).

Table 3.

Ethnic disparities in infant, neonatal, and postneonatal mortality, ENDES Peru 2019–2024

Mortality component	Ethnicity	Deaths/total	Rate per 1,000 (95% CI)	RR (95% CI)	RD per 1,000
Infant mortality (< 12 months)	Mestizo	452/48,409	8.0 (7.0–8.9)	1.00 (ref)	0 (ref)
Infant mortality (< 12 months)	Afro-Peruvian	161/12,574	13.1 (10.6–15.7)	1.65 (1.31–2.07)	5.2
Infant mortality (< 12 months)	Indigenous	565/39,824	13.6 (12.2–15.0)	1.71 (1.46–2.00)	5.6
Neonatal mortality (< 28 days)	Mestizo	298/48,409	5.8 (5.0–6.6)	1.00 (ref)	0 (ref)
Neonatal mortality (< 28 days)	Afro-Peruvian	108/12,574	9.3 (7.2–11.4)	1.60 (1.21–2.10)	3.5
Neonatal mortality (< 28 days)	Indigenous	312/39,824	7.4 (6.4–8.4)	1.28 (1.04–1.58)	1.6
Postneonatal mortality (28 days to < 12 months)	Mestizo	154/48,409	2.2 (1.7–2.7)	1.00 (ref)	0 (ref)
Postneonatal mortality (28 days to < 12 months)	Afro-Peruvian	53/12,574	3.8 (2.5–5.1)	1.74 (1.20–2.52)	1.6
Postneonatal mortality (28 days to < 12 months)	Indigenous	253/39,824	5.6 (4.7–6.5)	2.53 (1.99–3.22)	3.4

CI, confidence interval; RD, risk difference; ref, reference; RR, risk ratio

Risk ratios were estimated using survey-weighted poisson regression with robust variance. Risk differences were calculated from weighted mortality rates. rates are expressed per 1,000 live births

In absolute terms, the risk difference was 5.2 additional deaths per 1,000 live births for the Afro-Peruvian group (95% CI: 2.4–7.9) and 5.6 additional deaths per 1,000 for the Indigenous group (95% CI: 3.9–7.3), compared with the Mestizo group (Table 3).

Stratified analyses

Stratified analyses showed heterogeneous patterns of ethnic disparities across socioeconomic levels (Supplementary Table 1; Supplementary Figs. 2 and 3). In the poorest quintile, differences between ethnic groups were relatively small and did not reach statistical significance (p = 0.144): Mestizo 14.1 per 1,000; Afro-Peruvian 16.4 per 1,000; and Indigenous 18.3 per 1,000. However, disparities widened markedly in the middle and richer quintiles. In the richer quintile, Indigenous children had a rate of 14.3 per 1,000 compared with only 3.7 per 1,000 among Mestizo children, corresponding to a risk ratio of 3.86.

By area of residence, ethnic disparities were significant in both urban and rural settings. In urban areas, Afro-Peruvian (10.7 per 1,000) and Indigenous (10.9 per 1,000) groups had rates about 50% higher than the Mestizo group (7.2 per 1,000). In rural areas, all three groups had high rates, although differences persisted: Mestizo 12.2 per 1,000; Afro-Peruvian 18.3 per 1,000; and Indigenous 18.4 per 1,000.

By educational level, disparities were consistent across all strata. Even among mothers with higher education, Afro-Peruvian infants (12.0 per 1,000) had more than double the mortality rate of Mestizo infants (5.3 per 1,000).

Additive interaction analysis

The additive joint-pattern analysis (Supplementary Table 2) indicated that the combination of belonging to a minority ethnic group and experiencing socioeconomic disadvantage did not show positive departures from additivity in infant mortality. Instead, we observed a tendency toward subadditive patterns.

The joint pattern between Afro-Peruvian ethnicity and poverty showed a statistically significant negative RERI (RERI = -0.86; 95% CI: -1.69 to -0.03; p = 0.043), with a synergy index below one (S = 0.52; 95% CI: 0.21–0.82). This indicates that the combined excess associated with being Afro-Peruvian and poor was smaller than the sum of the separate excesses associated with each factor.

Interactions between Indigenous ethnicity and socioeconomic factors (poverty, rural residence, low education) yielded RERI values close to zero and were not statistically significant, suggesting approximately additive effects. The Indigenous × rural residence interaction showed a slightly positive but non-significant RERI (RERI = 0.13; 95% CI: − 0.37 to 0.64), with a synergy index compatible with additivity (S = 1.11; 95% CI: 0.66–1.56).

Temporal trends

Temporal trends in infant mortality during 2019–2024 are shown in Fig. 2. Rates fluctuated year-to-year across the three ethnic groups without a clear downward trend. The Mestizo group consistently had the lowest rates, ranging from 6.6 per 1,000 (2024) to 11.7 per 1,000 (2020). Afro-Peruvian and Indigenous groups maintained systematically higher rates throughout the period.

Fig. 2.

Temporal trends in the infant mortality rate by ethnic group, ENDES Peru 2019–2024 No statistically significant temporal trend was observed (p for ethnicity × year interaction = 0.113)

The ethnicity × year interaction test was not statistically significant (p = 0.113), indicating that ethnic disparities remained relatively stable over the study period. We found no evidence of convergence or divergence in mortality rates between ethnic groups.

Population attributable fraction

The population attributable fraction for ethnic mortality disparities indicated that 24.4% of infant deaths in Peru could theoretically be prevented under a counterfactual scenario in which all ethnic groups had the same mortality rate as the Mestizo group (Supplementary Table 3). By comparison, the fraction associated with disparities by wealth quintile was 47.7%, suggesting that nearly half of infant deaths would be avoided if the entire population had the mortality rate of the richest quintile.

Sensitivity analyses

Sensitivity analyses confirmed the robustness of the main findings (Supplementary Table 4). Excluding the year 2020 (COVID-19 pandemic period), risk ratios remained similar: RR = 1.84 (95% CI: 1.46–2.34) for the Afro-Peruvian group and RR = 1.82 (95% CI: 1.54–2.14) for the Indigenous group.

Analyses stratified by timing of death revealed different patterns for neonatal and postneonatal mortality. Ethnic disparities were more pronounced for postneonatal mortality (death from 28 days to before 12 months): Indigenous infants had a 2.5-fold higher risk than Mestizo infants (RR = 2.53; 95% CI: 1.99–3.22), whereas the excess was smaller for neonatal mortality (RR = 1.28; 95% CI: 1.04–1.58). Afro-Peruvian infants showed a similar pattern, with larger disparities in the postneonatal period (RR = 1.74; 95% CI: 1.20–2.52) than in the neonatal period (RR = 1.60; 95% CI: 1.21–2.10).

Discussion

Main findings

This study documents marked ethnic disparities in infant mortality in Peru. Children of mothers who self-identify as Afro-Peruvian or Indigenous have a 65% and 71% excess risk of infant mortality, respectively, compared with children of Mestizo mothers. In absolute terms, these differences translate into approximately 5–6 additional deaths per 1,000 live births, magnitudes that are clinically and epidemiologically meaningful. These disparities persist despite the overall progress the country has made in reducing infant mortality over recent decades, suggesting that the benefits of public health interventions have not been equitably distributed across population groups.

A particularly relevant finding was the heterogeneity of disparities by timing of death and socioeconomic level. Inequities were more pronounced for postneonatal mortality than for neonatal mortality, suggesting that social determinants and access to health services throughout the first year of life play a predominant role. Moreover, ethnic disparities did not diminish in the most advantaged socioeconomic strata; on the contrary, they tended to widen in the middle and higher wealth quintiles. This pattern suggests that socioeconomic progress alone may be insufficient to eliminate ethnic health gaps and points to discrimination mechanisms operating beyond household wealth or maternal education.

Several mechanisms may plausibly link maternal ethnic self-identification with infant mortality. First, Indigenous and Afro-Peruvian families are more frequently exposed to socioeconomic deprivation, rural residence, and geographic barriers that delay timely contact with health services. Second, ethnic minority mothers may face language barriers, culturally discordant care, discrimination, or lower perceived quality of care, which can reduce continuity of antenatal, delivery, and postnatal services. Third, household and community-level conditions-including water and sanitation, food security, maternal workload, and access to early treatment for diarrhea, respiratory infections, and undernutrition-are particularly relevant during the postneonatal period. Therefore, the stronger disparities observed after the neonatal period are consistent with mechanisms operating beyond delivery care alone, especially through continuity of child health services and the broader living environment.

Comparison with other studies

The magnitude of the ethnic disparities observed in this study is consistent with the regional literature. In Brazil, Reboucas et al. [7] reported that under-5 mortality was 98% higher among children born to Indigenous mothers and 39% higher among children born to Black mothers compared with those born to White mothers. Our findings for the Peruvian Indigenous population (71% higher risk) fall between these values, whereas disparities for the Afro-Peruvian population (65%) are substantially larger than those reported for Afro-descendants in Brazil. These differences may reflect variation in historical contexts, patterns of residential segregation, and the design of social policies between the two countries.

Our finding that disparities persist in higher socioeconomic strata is consistent with studies from the United States. Vilda et al. [17] showed that structural racism is associated with infant mortality inequities even after controlling for individual socioeconomic factors, in both urban and rural areas. Similarly, Saiyed et al. [18] found that infant mortality rates among Black infants were twice those among White infants across the 69 most populous U.S. cities, regardless of local economic development. These patterns suggest that racial and ethnic discrimination operates relatively independently of socioeconomic position.

The observation that ethnic disparities are more pronounced in postneonatal mortality aligns with prior studies. In the Brazilian study by Reboucas et al. (7), the largest disparities were observed for deaths due to diarrhea and malnutrition among children aged 1 to 4 years. Together, these findings reinforce the interpretation that ethnic inequities in infant mortality primarily reflect structural social determinants rather than intrinsic genetic or biological differences between ethnic groups. Proximal clinical processes such as infection, diarrhea, and malnutrition remain biologically mediated, but their distribution is socially patterned through living conditions, access to care, and quality of services.

Our estimates are broadly concordant with official Peruvian mortality patterns, although direct comparability is limited by differences in data sources and analytic definitions. ENDES defines infant mortality as deaths before one year of age per 1,000 live births in the five years preceding the survey, whereas vital statistics rely on births and deaths registered through RENIEC [16, 19]. Official INEI estimates have similarly documented important territorial and social gradients, including higher infant mortality in Amazonian departments, rural areas, and children of mothers with lower educational attainment [20]. The gradients observed in our analysis by rural residence, wealth, education, and natural region are therefore consistent with official Peruvian patterns, while our study adds maternal ethnic self-identification as an additional axis of inequity that is less systematically visible in routine official reporting.

A recent scoping review on health disparities among Indigenous populations in Latin America [6]identified infectious diseases and malnutrition as the most frequently studied problems, with consistently higher rates among Indigenous populations across the nine countries included. Garcia et al. [10], in their analysis of Peru’s success in reducing infant mortality, highlighted equity-focused strategies and cultural adaptation of health services as key factors, but also acknowledged that gaps between groups have not fully closed. Our results confirm that, despite progress, substantial inequities persist and require targeted attention.

The additive joint-pattern analysis indicated subadditive patterns for the combination of Afro-Peruvian ethnicity and poverty, differing from studies in other settings where synergistic effects have been reported. Gangopadhyaya et al. [21] found that, in the United States, structural racism and neighborhood deprivation interact in complex ways to shape perinatal outcomes. The lack of positive departure from additivity in our data could be explained by the fact that poverty affects all groups so severely that additional ethnic differences are attenuated within this stratum (a ceiling effect), whereas in more advantaged strata-where poverty is not the limiting factor-the specific effects of ethnic discrimination emerge more clearly.

Public health implications

Our findings have direct implications for maternal and child health policy in Peru. The estimate that 24.4% of infant deaths could theoretically be prevented under a counterfactual scenario in which ethnic mortality gaps were eliminated represents a substantial absolute number of lives that could be saved annually. To achieve the Sustainable Development Goals related to infant mortality, it will not be sufficient to continue universal strategies that disproportionately benefit already advantaged groups; specific and differentiated interventions are required to address the particular needs of Afro-Peruvian and Indigenous populations [22, 23].

The greater vulnerability observed in the postneonatal period suggests that interventions should extend beyond the immediate perinatal window. Active follow-up of children throughout the first year of life, breastfeeding promotion, timely access to vaccination, and appropriate management of common childhood illnesses are strategies likely to have a larger impact on reducing disparities [24]. Home visits by community health workers—a strategy that Peru has successfully implemented in some regions—could be prioritized in communities with a high concentration of Indigenous and Afro-Peruvian populations.

The persistence of disparities in middle and high socioeconomic strata indicates that poverty-reduction policies, although necessary, will not be sufficient to close ethnic gaps. Targeted actions to address discrimination within the health system are needed, including training in intercultural competencies, adaptation of services to the needs and preferences of minority populations, and mechanisms to monitor ethnic equity in health indicators [25, 26]. Experience from other countries suggests that involving representatives from affected communities in the design and evaluation of interventions improves acceptability and effectiveness.

Regular monitoring of ethnic disparities in infant mortality should be incorporated into the country’s epidemiological surveillance systems. Currently, INEI official reports present indicators disaggregated by area of residence, region, and wealth quintile, but disaggregation by ethnicity is less systematic. Producing official statistics that make ethnic gaps visible is a necessary step for accountability and for evaluating policy impact. Additional research is also needed to understand the specific causal mechanisms generating these disparities in the Peruvian context, including qualitative studies exploring the experiences of mothers from minority groups within the health system.

Limitations

This study has several limitations that should be considered when interpreting the results. First, ethnicity is based on maternal self-identification, which may vary depending on the social and temporal context in which the question is asked. Second, ENDES does not collect detailed information on causes of death, limiting our ability to identify specific mechanisms. Third, births in the five years preceding each survey may be subject to recall bias and selective maternal survival, although the main analysis was restricted to births with complete opportunity for ascertainment of death before 12 months. Fourth, we could not evaluate the role of factors such as perceived discrimination, chronic stress, or experiences of racism-variables identified as important mediators in other settings. Finally, exclusion of ethnic categories with small frequencies (White, Other, Don’t know) limits generalizability to these groups.

Conclusions and recommendations

There are marked ethnic disparities in infant mortality in Peru that disproportionately affect children of Afro-Peruvian and Indigenous mothers. These inequities are observed across socioeconomic strata and are particularly pronounced in the postneonatal period. Approximately one quarter of infant deaths in the country could be prevented under a counterfactual scenario in which ethnic mortality gaps were eliminated.

In light of these findings, we recommend: (a) incorporating systematic monitoring of maternal and child health indicators disaggregated by ethnicity into epidemiological surveillance systems; (b) designing and implementing culturally adapted interventions that address the specific needs of Afro-Peruvian and Indigenous populations; (c) strengthening follow-up of children throughout the first year of life in communities with a high prevalence of minority ethnic groups; (d) training health personnel in intercultural competencies and non-discrimination; and (e) conducting research to explore the specific causal mechanisms generating these disparities in the Peruvian context to inform more effective interventions.

Author contributions

V.J.V.P.: Conceptualization, Methodology, Formal analysis, Investigation, Resources, Funding acquisition, Writing—original draft, Writing—review & editing. J.B.C.: Conceptualization, Methodology, Investigation, Data curation, Writing—review & editing. H.D.T.: Conceptualization, Methodology, Investigation, Writing—review & editing. J.E.B.C.: Conceptualization, Methodology, Investigation, Writing—review & editing. F.Z.M.: Conceptualization, Methodology, Investigation, Writing—review & editing, Supervision, Project administration. All authors have read and agreed to the published version of the manuscript..

Funding

The authors declare that no specific funding was received for the preparation of this manuscript. If accepted, article processing charges, if applicable, will be covered by the Vice-Rectory for Research of the Universidad Nacional Toribio Rodríguez de Mendoza de Amazonas. The funder had no role in the conception, preparation, or decision to publish the manuscript.

Data availability

The data supporting the findings of this study are available at the following link: https://proyectos.inei.gob.pe/microdatos/

Declarations

Ethics approval and consent to participate

Since this manuscript is a secondary data base study it was not required.Informed consent: Since this is a secondary data analysis, informed consent was not required.Clinical trial number: not applicable.

Consent to publication

Not applicable.

Competing interests

The authors declare no competing interests.