Neonatal mortality inequalities in Peru, 2007–2021: an ecological joinpoint trends analysis

Abstract

Background

Neonatal disorders remain a leading cause of loss of healthy life years worldwide, second only to COVID-19 in 2021, although most neonatal deaths are preventable. The neonatal mortality rate (NMR), a key indicator of the 2030 Sustainable Development Agenda, varies widely within and between countries, reflecting social conditions that shape neonatal survival. This study examined the magnitude and temporal trends of ecosocial inequalities in Peru’s NMR from 2007 to 2021, their relationship with selected social determinants, and changes in the epidemiological profile of neonatal deaths.

Methods

An ecological study was conducted using data from Peru’s 25 regions (2007–2021). Temporal trends in NMR and inequalities along a social gradient defined by monetary poverty, unmet basic needs, and food insecurity were analyzed. Absolute and relative inequalities were measured using the slope index of inequality (SII) and concentration index (CIx). Inflection points in trends were identified with joinpoint regression, and monotonic associations between NMR (and its inequalities) and contextual variables were assessed using Spearman’s rank correlation. Changes in the epidemiological profile of neonatal deaths were evaluated with the Chi-square test.

Results

Peru’s NMR declined from 10.3 to 8.8 deaths per 1,000 live births between 2007 and 2021, with the steepest reduction around 2010–2014. Most regions experienced decreases, except Huancavelica, Pasco, and Puno. Cross-regional inequalities showed a persistent pro-rich pattern, indicating survival disadvantages in regions with higher unmet basic needs. National NMR trends correlated positively with unmet basic needs and monetary poverty and negatively with current health expenditure per capita. During the COVID-19 years, inequalities narrowed as NMR fell in poorer regions and rose in richer ones. The epidemiological profile shifted toward a higher proportion of deaths from extreme prematurity and low birth weight.

Conclusions

From 2007 to 2021, Peru achieved a decline in neonatal mortality, but pro-rich regional inequalities persisted, and the burden remained concentrated in highland regions. During the COVID-19 period, the downward trend continued while inequalities narrowed, in parallel with reductions in poverty and increases in health expenditure. These ecological findings highlight the importance of monitoring health inequalities alongside national averages to support accountability toward the SDG commitment to “leave no one behind.”

Supplementary Information

The online version contains supplementary material available at 10.1186/s12939-025-02731-9.

Background

Neonatal disorders continue to be the leading cause of loss of healthy life years (DALYs) globally, surpassed only by COVID-19 in 2021 [1]. At least 80% of the burden of neonatal mortality –i.e., the death of a live-born infant within 28 days of birth, regardless of birth weight or gestational age– is concentrated in three preventable and treatable conditions: complications due to prematurity, intrapartum-related conditions (including birth asphyxia and trauma), and neonatal infections [2]. The neonatal mortality rate (NMR) –i.e., the number of neonatal deaths per 1,000 live births (lb)– varies significantly between countries and regions, indirectly reflecting the adverse impact of comparatively less favorable health and social conditions on neonatal survival depending on where the birth occurs.

In Latin America and the Caribbean (LAC) –and in Peru in particular– substantial progress has been observed in reducing child mortality, including neonatal mortality, especially since 2000 with the momentum generated by the Millennium Development Goals (MDGs) [3]. According to estimates from the United Nations Inter-agency Group for Child Mortality Estimation (IGME), the median number of deaths among children under 5 declined by 58.4% in LAC and 62.1% in Peru between 2000 and 2021, corresponding to an average annual reduction of -3.4% and − 4.3% in the under-five mortality rate (U5MR), respectively. Additionally, the median number of neonatal deaths decreased by 52.1% in LAC and 53.0% in Peru, equivalent to an average annual reduction of -2.7% and − 3.3% in the NMR, respectively [4].

Because the pace of NMR reduction is slower than that of U5MR –mainly due to greater success in preventing post-neonatal deaths from diarrhea, pneumonia, measles, and other infectious causes [5]– neonatal mortality has gained greater relevance in efforts to improve early-age survival. This has been reflected since 2015 in the global priority given to reducing NMR in the 2030 Agenda for Sustainable Development and its third goal: to ensure healthy lives and promote well-being for all at all ages (SDG 3) [6]. Target 3.2 of SDG 3 aims to end preventable neonatal deaths, with all countries striving to reduce NMR to at least 12 per 1,000 lb. At the regional level, the Sustainable Health Agenda for the Americas (SHAA) proposes reducing NMR to less than 9 per 1,000 by 2030 across all population groups, including those at higher risk [7].

The high priority of NMR, reflected in its considerable presence in specialized literature and international agendas, contrasts with the limited available evidence on the magnitude and trends of inequalities in NMR between and within countries. This limited evidence may explain the absence of explicit targets for reducing inequalities in NMR, despite the commitment to “leave no one behind” declared in the 2030 Agenda [6], and the strong calls from health equity derived from the works and reports of the WHO Commission on Social Determinants of Health [8] and the PAHO Commission on Equity and Health Inequalities in the Americas [9]. Tracking changes in NMR over time according to the relative social position of population groups and territories is a critical element of health intelligence to inform equity-oriented decision-making and to ensure accountability in reducing inequalities in neonatal survival [10]. This equity-focused approach requires consideration of social determinants such as poverty, investment in health systems, baseline levels of neonatal mortality, and the structure of causes of neonatal death, among other relevant factors [10–12]. The availability of longitudinal, subnationally disaggregated data on these dimensions in Peru presents a unique opportunity to explore the national evolution of NMR and its intranational inequalities, thereby generating potentially useful information to guide health policy toward achieving SDG 3.2 equitably in the country. In this regard, the objective of this study is to analyze the magnitude and patterns of ecosocial inequalities in neonatal mortality (NMR) over time, in relation to national NMR trends and selected social determinants, their temporal correlation, potential inflection points, and changes in the epidemiological profile of neonatal death. To do so, our study focuses on the most recent 15-year historical series—covering the second half of the MDG period and the first half of the SDG period, including the first two years of the COVID-19 pandemic.

Methods

Study design and setting

This is an ecological time-trend study using administrative data at the national and subnational levels, covering the 25 territorial regions of Peru from 2007 to 2021.

Study variables and data sources

Our primary study outcomes were three: the neonatal mortality rate (NMR), absolute inequality in NMR (slope index of inequality, SII), and relative inequality in NMR (concentration index, CIx). We considered five contextual variables: monetary poverty incidence (MP), unmet basic needs incidence (UBN), caloric deficit incidence (as a proxy for food insecurity, FI), per capita current health expenditure (CHEpc), and current health expenditure as a percentage of the gross domestic product (CHE%gdp). All variables were aggregated at the country/year level; NMR, MP, UBN, and FI were also disaggregated at the region/year level. Additionally, a set of six variables characterizing the epidemiological profile of neonatal death was available from individual records (cause of death, sex, place of birth, place of death, prematurity, and birth weight).

NMR was estimated from death certificates registered in Peru’s mortality database (PMD), available through the National Unified Health Information Repository (REUNIS) [13]. To indirectly assess the completeness of national registration during the study period, additional NMR estimates were gathered from the Demographic and Health National Survey (ENDES) by the National Institute of Statistics and Informatics (INEI) [14], the 2021 Global Burden of Disease Study by the Institute for Health Metrics and Evaluation (IHME) [15], and the United Nations Inter-agency Group for Child Mortality Estimation (IGME) [4].

Absolute inequality in NMR (SII) and relative inequality in NMR (CIx) were calculated directly from PMD’s NMR values and MP, UBN, and FI estimates for each subregion/year, as generated by the Roundtable for the Concerted Fight Against Poverty (MCLCP) [16], as described below. Country/year values for current health expenditure (CHEpc and CHE%gdp) were obtained from the National Health Accounts available in the World Health Organization’s Global Health Expenditure Database [17].

Data on the epidemiological profile of neonatal death were obtained from individual records of neonatal deaths captured in the National Perinatal and Neonatal Epidemiological Surveillance Subsystem (SSVEPN) of the Ministry of Health of Peru [18], available only for the period 2012–2021.

NMR inequality analysis

We separately ranked Peruvian regions from the most to the least disadvantaged according to MP, UBN and FI –i.e., the social stratifiers– each year to construct a gradient of relative social position. Two summary measures of NMR inequality were then calculated: the slope index of inequality (SII) and the concentration index of inequality (CIx). The SII –a measure of absolute inequality (i.e., the excess neonatal deaths per 1,000 live births across the social gradient)– was computed by regressing the NMR on the cumulative live birth population fractional rank, selecting the mean squared error (MSE)-based best fit over four regression models explored, including Poisson, and negative binomial. The CIx –a measure of relative inequality (i.e., the disproportionality between the population share across the social gradient and the corresponding share of neonatal deaths)– was computed by fitting a Lorenz concentration curve to the observed cumulative relative distribution of the population, ordered by social stratifiers, and the observed cumulative distribution of neonatal deaths, and numerically integrating the area under the curve [10, 19]. Because the Neonatal Mortality Rate (NMR) is a rate variable bounded at zero but not by a fixed upper limit, corrections developed for binary or proportional variables [20]—such as those proposed by Wagstaff or Erreygers—are not applicable. Accordingly, the standard Concentration Index was determined to be the sufficient and appropriate measure for this analysis [19]. For descriptive purposes, we also calculated quartile distribution gradients and inequality gaps at the extremes for selected study variables over time. As neonatal mortality is an adverse health outcome, a negative SII value indicates a negative slope, where mortality decreases as socioeconomic position improves. A negative CIx value indicates that mortality is disproportionately concentrated among the socioeconomically disadvantaged. In this context, both negative values are interpreted as a pro-rich pattern of inequality.

Trend analysis

Segmented linear regression was used to identify statistically significant inflection points, or joinpoints, in the 2007–2021 time trends of NMR, NMR inequalities (SII, CIx), and contextual variables. Our analysis required two different approaches based on the nature of the variable. For positive-only variables (NMR, MP, UBN, FI, CHEpc, CHE%gdp), variables were log-transformed to stabilize variance. For these log-linear models, serial correlation was assessed using the Durbin-Watson test, and a first-order autocorrelation error model was fitted when indicated. Trend segments were evaluated by calculating their Annual Percent Change (APC), and the overall trend was summarized using the Average Annual Percent Change (AAPC), with statistical significance determined by its 95% confidence intervals [21, 22]. A different approach was necessary for the inequality metrics (SII and CIx), as these metrics can assume negative values, making a log-transformation mathematically inappropriate. For these variables, a standard linear model was fitted. Consequently, the trend for these segments is reported not as an APC, but as the linear Slope of the trend line, and its statistical significance was assessed via a p-value. For both model types, the Bayesian Information Criterion (BIC) was used to guide model selection via the Data-Driven Selection (DDS) method, and a grid search for 0 to 2 joinpoints was used based on the number of available data points (15 years) and the software’s guidance [23].

Correlation analysis

To assess whether a monotonic correlation –its direction and strength– existed between the trends of the study’s main variables (NMR) and contextual variables, Spearman’s ρ rank correlation was used.

Frequency analysis

To determine whether the epidemiological profile of neonatal deaths differed statistically over time, the chi-square test of independence was used.

Trend analyses were conducted using Joinpoint [23]; all other analyses were performed in R [24].

Results

The NMR showed a downward trend during the study period (AAPC: -1.07%, 95% CI: -1.44%; -0.70%), falling from 10.3 to 8.8 per 1,000 live births between 2007 and 2021 (Fig. 1A), representing a cumulative 6,055 avoided neonatal deaths compared to the baseline NMR (Table 1). Two inflection points were identified in the national NMR trend, with the 2010–2014 segment being the only one with a statistically significant reduction in national NMR (APC: -3.71%, 95% CI: -5.63; -2.06) (Fig. 1B). Based on these results, the neonatal death surveillance data were grouped into three segments: before 2015, 2015–2019, and 2020–2021 (COVID-19 years) to analyze changes in the epidemiological profile. A statistically significant change (p < 0.001) was observed in this profile, with a proportional increase in prematurity as a cause of death, institutional births, institutional deaths, extreme prematurity, and extremely low birth weight, with the most notable increases occurring during the two pandemic years (Table 2). These differences were also observed in the annual series analysis from 2012 to 2021 (Additional file 1: Table S1).

Fig. 1.

Neonatal mortality rate time trends and joinpoint regression output; Peru, 2007–2021. (A) Estimated Neonatal Mortality Rate (NMR) from various sources. (B) Joinpoint regression analysis of the NMR estimated by PMD. PMD: Peruvian Mortality Database. UN-IGME 2023: UN Inter-agency Group for Child Mortality Estimation. IHME GBD 2021: IHME Global Burden of Disease 2021 Study. ENDES: Demographic and Family Health Survey of Peru. NMR: Neonatal Mortality Rate. CI: Confidence interval. APC: Annual Percent Change. AAPC: Average Annual Percent Change. (*) Statistically significant

Table 1.

Basic indicators selected for national-level analysis and neonatal mortality rates’ corresponding spearman’s ρ rank correlation coefficients; Peru, 2007–2021

Basic indicator	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021		ρ
Live births	619,312	611,440	598,800	584,445	571,427	562,799	559,531	559,587	561,514	563,855	565,157	565,815	566,797	567,512	567,369		-
Neonatal deaths	6,365	6,498	6,280	6,075	5,927	5,555	5,138	5,088	5,111	4,856	5,378	5,263	5,020	5,037	5,001		-
Avoided neonatal deaths (from 2007 baseline)	0	-214	-126	-68	-54	229	613	663	660	939	430	552	805	796	830		-
Neonatal mortality rate	10.3	10.6	10.5	10.4	10.4	9.9	9.2	9.1	9.1	8.6	9.5	9.3	8.9	8.9	8.8		1.00
Monetary poverty incidence (%)	39.3	36.2	34.8	31.3	27.8	25.8	23.9	22.7	21.8	20.7	21.7	20.5	20.1	30.1	25.9		0.650
Unmet basic needs incidence (%)	30.3	28.9	26.8	23.9	23.3	21.6	20.3	19.7	19.4	18.7	18.0	16.6	16.0	16.6	16.1		0.822
Food insecurity incidence (%)	26.3	30.0	29.1	27.3	27.7	27.4	26.5	26.7	23.9	25.3	25.6	25.6	26.8	32.7	32.1		0.228
CHE per capita	158.6	191.4	209.2	241.4	264.5	309.3	321.0	332.2	310.6	311.3	332.5	362.8	368.0	399.6	455.7		-0.793
CHE as GDP%	4.4	4.5	5.0	4.7	4.5	4.7	4.7	4.9	4.9	4.9	4.8	5.1	5.1	6.4	6.7		-0.664

CHE: Current health expenditure. GDP: Gross Domestic Product

Table 2.

Epidemiological characteristics of neonatal deaths, by summary study periods; Peru, 2012–2021

Epidemiological characteristics	2012–2014 (n = 10,879)	2015–2019 (n = 15,989)	2020–2021 (n = 5,940)	p-value
Cause of death
Prematurity-immaturity	24.2%	26.6%	27.5%	< 0.001*
	(2,627)	(4,247)	(1,630)
Lethal congenital malformation	14.1%	15.4%	15.4%
	(1,527)	(2,461)	(914)
Infections	18.9%	18.6%	16.3%
	(2,052)	(2,974)	(968)
Asphyxia and related causes	20.7%	18.4%	17.0%
	(2,247)	(2,943)	(1,007)
Other causes	22%	21.0%	23.9%
	(2,381)	(3,352)	(1,419)
Sex				0.123
Female	42.7%	43.5%	44.3%
	(4,630)	(6,940)	(2,626)
Male	57.3%	56.5%	55.7%
	(6,220)	(9,026)	(3,305)
Place of birth				< 0.001*
Home birth	13.5%	8.3%	7.4%
	(1,464)	(1,330)	(440)
Institutional birth	86.5%	91.7%	92.6%
	(9,415)	(14,643)	(5,485)
place of death				< 0.001
Community	14.4%	9.8%	7.7%
	(1,568)	(1,573)	(457)
Institutional	85.6%	90.2%	92.3%
	(9,311)	(14,399)	(5,466)
Preterm birth				< 0.001
Not preterm [37 + weeks)	35.4%	31.1%	32.4%
	(3,852)	(4,973)	(1,927)
Late preterm [32–36 weeks)	29.1%	28.5%	26.4%
	(3,164)	(4,551)	(1,566)
Very preterm [28–32 weeks)	18.6%	19.9%	19.6%
	(2,022)	(3,176)	(1,165)
Extremely preterm (< 28 weeks)	16.9%	20.6%	21.6%
	(1,841)	(3,289)	(1,282)
Weight at birth				< 0.001
Normal [2,500 + grams)	34.4%	31.2%	32.8%
	(3,739)	(4,995)	(1,946)
Low weight [1,500–2,499 g)	27.7%	25.8%	25.0%
	(3,016)	(4,124)	(1,483)
Very low weight [1,000–1,499 g)	18.0%	18.2%	18.1%
	(1,959)	(2,907)	(1,073)
Extremely low weight (< 1,000 g)	19.9%	24.8%	24.2%
	(2,165)	(3,963)	(1,438)

(*) Statistically significant

Overall, the three poverty-related contextual variables –MP, UBN, and FI– also showed a downward trend, while the two health expenditure variables –CHEpc and CHE%gdp– showed an upward trend during the period. The joinpoint regression analysis identified significant inflection points, particularly a reversal in the trends of MP and FI in 2018 (Additional file 1: Figure S2). Nevertheless, the national NMR trend was highly positively correlated with UBN (ρ = 0.82) and MP (ρ = 0.65), and negatively correlated with CHEpc (ρ=-0.79) and CHE%gdp (ρ=-0.66) during the study period (Table 1).

Intranational NMR trends between 2007 and 2021 showed that, with the exception of Huancavelica, Pasco, and Puno, all regions of Peru exhibited a downward trend (Figure S1, Additional file 1: Table S2).

Both absolute (SII) and relative (CIx) inequalities in NMR were of moderate magnitude, negative in sign, and generally showed a stationary trend during the study period, except for NMR inequality related to food insecurity, which showed a clear downward trend (Fig. 2, Additional file 1: Table S3). For instance, the excess neonatal mortality across the interregional social gradient defined by UBN incidence –concentrated at the most deprived end– decreased from 6.7 deaths per 1,000 live births (SII − 6.7; 95% CI: -8.6; -4.9) in 2007 to 5.4 (SII − 5.4; 95% CI: -8.6; -2.3) in 2021; the concentration of inequality decreased from − 11.1% (-15.1; -5.0) to -9.6% (-14.7; -2.2), respectively. Statistically significant inflection points were identified only in the trends of absolute and relative inequalities in NMR related to monetary poverty: the SII decreased between 2007 and 2013 (Slope:0.27; p < 0.05) and between 2018 and 2021 (Slope: 0.96; p < 0.05), while inequality in NMR increased between 2013 and 2018 (Slope: -0.97; p < 0.05) (Fig. 2).

Fig. 2.

Joinpoint regression time trends of cross-region neonatal mortality inequalities by equity stratifier; Peru, 2007–2021. SII: Slope Index Inequality. CIx: Health Concentration Index. CI: Confidence interval. (*) Statistically significant (p-value < 0.05)

During the pandemic period, ecosocial inequality in NMR decreased markedly: the SII dropped by 35%, from − 8.6 (95% CI: -12.0; -5.1) excess neonatal deaths per 1,000 live births across the interregional gradient of monetary poverty in 2019 to -5.6 (95% CI: -9.7; -1.5) in 2021. At the extreme of highest monetary poverty, the modeled NMR fell from 14.3 to 13.1 per 1,000 live births (a reduction of -8.4%) between 2019 and 2021, while at the lowest poverty extreme, NMR rose from 5.7 to 7.5 neonatal deaths per 1,000 live births (an increase of 31.6%) (Additional file 1: Figure S3). This shift in the intranational distribution of NMR was also observed in the gradients defined by the other two social stratifiers (i.e., UBN and FI).

Table 3 shows the Spearman’s rho (ρ) rank correlation coefficients between the summary measures of inequality in NMR (i.e., SII and CIx) and the social determinants over time. High monotonicity was observed only between FI-related NMR inequalities and UBN, MP, CHEpc, and CHE%gdp. The correlation was negative with UBN and MP, and positive with CHEpc and CHE%gdp (Additional file 1: Figure S4 and Figure S5). Quartile distribution gradients were consistently present over time in NMR, as well as in all contextual variables and epidemiological characteristics of neonatal deaths (Additional file 1: Table S4 and S5).

Table 3.

Spearman’s ρ coefficients for the rank correlation between neonatal mortality cross-region inequality (absolute and relative) and selected social determinants; Peru, 2007–2021

Social determinants	Absolute inequality in neonatal mortality (SII)			Relative inequality in neonatal mortality (CIx)
	Monetary poverty	Unmet basic needs	Food insecurity	Monetary poverty	Unmet basic needs	Food insecurity
Monetary poverty	0.50 [-0.13; 0.89]	0.02 [-0.54; 0.62]	-0.62* [-0.91; -0.05]	0.13 [-0.52; 0.68]	0.06 [-0.66; 0.65]	-0.68* [-0.96; -0.16]
Unmet basic needs	0.61* [0.08; 0.89]	-0.17 [-0.70; 0.49]	-0.67* [-0.96; -0.17]	0.33 [-0.31; 0.77]	0.14 [-0.50; 0.72]	-0.94* [-0.99; -0.77]
Food insecurity	0.33 [-0.20; 0.73]	0.08 [-0.50; 0.60]	-0.06 [-0.58; 0.47]	-0.06 [-0.60; 0.51]	-0.03 [-0.51; 0.46]	-0.08 [-0.72; 0.50]
CHE per capita	-0.49 [-0.83; 0.02]	0.23 [-0.43; 0.71]	0.64* [0.07; 0.95]	-0.23 [-0.78; 0.44]	-0.07 [-0.63; 0.54]	0.92* [0.73; 0.99]
CHE as GDP%	-0.50* [-0.83; -0.02]	0.38 [-0.18; 0.79]	0.47 [-0.18; 0.87]	-0.25 [-0.77; 0.36]	0.14 [-0.41; 0.67]	0.73* [0.34; 0.93]

(*) Statistically significant. CHE: Current health expenditure. GDP: Gross Domestic Product. SII: Slope Index Inequality. CIx: Health Concentration Index

Discussion

Our study shows slow but sustained progress in reducing the national neonatal mortality rate (NMR) between 2007 and 2021, alongside improvements in meeting basic needs and health spending, and in contrast to the stationary trend of a moderately pro-rich inequality pattern in NMR across Peru’s 25 regions. The epidemiological profile of neonatal death showed a significant increase in the proportion of extremely premature births, extremely low birth weight, and mortality associated with prematurity, especially during the pandemic years. However, the national NMR trend was not altered during the COVID-19 pandemic, and intranational inequality in NMR during the pandemic years decreased due to increased neonatal mortality in the more socially advantaged regions, and a concurrent decrease in the more disadvantaged regions.

The downward trend in NMR in Peru since the beginning of the current millennium has been documented in various studies [25–28], and the country has surpassed both the global (SDG 3.2) and regional (SHAA) targets since 2007 and 2015, respectively. In fact, among the 75 countries prioritized for collectively accounting for 95% of the global burden of maternal, neonatal, and child mortality in 2000 (the so-called Countdown countries), Peru achieved the greatest reduction in NMR during the MDG period (from 26 per 1,000 live births in 1990 to 9 in 2015). This dramatic progress has been attributed to a strong anti-poverty political agenda and a more proactive governmental role in promoting high-impact interventions, such as the implementation of the Maternal-Neonatal Strategic Program, the expansion and creation of the Comprehensive Health Insurance (SIS), with special attention to children under 5 and mothers. Conditional cash transfer programs (JUNTOS), cross-sectoral programs (Results-Based Budgeting), continued food supplementation programs (PRONAA), water and sanitation provision in lagging rural areas, and improvements in women’s education were also introduced [25, 27, 29]. With some nuances, this evidence also extends to infant mortality in general [30].

Supporting this argument, our study found a strong correlation between the decline in NMR and the reduction in monetary poverty and unmet basic needs, as well as the increase in health spending—both per capita and as a proportion of GDP. In contrast, the temporal correlation between NMR and food insecurity was weak, due to its stationary trend until the arrival of COVID-19, when it increased markedly.

However, our study also found that, at the subnational level, by the end of 2021, fourteen of the 25 regions –which account for 42.8% of live births and 55.1% of neonatal deaths in the country– still had NMRs above 9 per 1,000 live births, and seven of them (Amazonas, Apurímac, Ayacucho, Cusco, Huancavelica, Pasco, and Puno, which account for 16.5% of live births and 25.8% of neonatal deaths) had rates above 12 per 1,000 live births. Moreover, these geographic differences form ecosocial gradients, configuring a pro-rich pattern, of moderate intensity, that concentrates the risk and burden of neonatal mortality in the most socioeconomically disadvantaged regions (i.e., with higher monetary poverty, unmet basic needs, and/or food insecurity), whose trend has generally remained stationary over the 15 years evaluated. These findings align with reports from Peru prior to the COVID-19 pandemic: higher NMRs in socioeconomically poorer regions and those with lower educational levels [27, 28, 31]. Similarly, various global studies have documented the presence and persistence of pro-rich patterns of inequality in NMR, both between and within countries, using survey and administrative data [32–35].

Three findings regarding ecosocial inequality trends in NMR in Peru deserve special emphasis. First, inequality in NMR –both absolute and relative– across the interregional gradient defined by food insecurity showed a clear downward trend during the study period and strong monotonicity with the declining trends in unmet basic needs and monetary poverty, and the rising trends in health spending per capita and as a proportion of GDP. This points, on one hand, to caloric deficit incidence as a dimension of chronic poverty [36], as it remained stable during the period –around 25% nationally– and, on the other hand, to the progressive equalization of caloric deficit due to its increasing incidence in regions with lower NMR. Indeed, while in 2007 the incidence of caloric deficit poverty in the quartile of regions with the highest NMR was 38.3% and in the quartile with the lowest NMR was 17.7% (relative gap: 2.2), in 2021 these incidences were 33.8% and 32.9%, respectively (relative gap: 1.0) [16].

Second, during the period of significant national NMR reduction (2010–2014, revealed by trend inflection points), inequality in NMR –both absolute and relative– across the interregional gradient defined by monetary poverty also decreased significantly. This suggests the absence of the ubiquitous phenomenon of elite capture that characterizes the so-called inverse equity hypothesis, whereby average improvements in NMR –resulting from a new intervention or program– initially occur at the expense of the more socially advantaged population segments [37, 38]. In fact, between 2010 and 2014, NMR declined at an average annual rate of -5.7% in the quartile of regions with the highest monetary poverty, compared to -4.5% in the quartile with the lowest monetary poverty. These results suggest an element of social protection and effective targeting in the anti-poverty political agenda, documented elsewhere [27, 28, 39].

Third, ecosocial inequalities in NMR decreased markedly during the first two years of COVID-19 –without altering the downward trend in national NMR. That the national NMR trend remained unchanged during the pandemic is a finding generally consistent with with what has been reported in the literature: despite the enormous mortality burden imposed by COVID-19 globally and especially in Peru [40–43], there is currently no solid evidence of excess neonatal mortality associated with the pandemic [4, 44–47]. However, systematic reviews and meta-analyses on the pandemic’s impact on neonates indicate an increase in premature births due to SARS-CoV-2 infection in pregnant women, comorbidities, and reduced access to care, especially in low-income countries. By the end of 2021, Peru ranked second in the Americas in terms of the number of pregnant women infected with SARS-CoV-2 (55,440 cases), which may have contributed to the increase in premature births and neonatal deaths due to prematurity during the pandemic, as evidenced in our study by a slight increase in neonatal deaths among extremely premature infants [48–52].

The reduction in ecosocial inequalities in NMR during COVID-19 shown by our study is, to the best of our knowledge, an unprecedented finding. With the exception of one study in the United States documenting a disproportionate increase in NMR among American Indian and Alaska Native populations compared to other ethno-racial identity groups during the pandemic [53], we found no published quantitative evidence on the impact of COVID-19 on NMR inequalities. These findings suggest a pivoting pattern in the neonatal mortality gradient, with a decrease in NMR in poorer regions and an increase in less poor ones. Explaining this complex and seemingly counterintuitive phenomenon requires confirmatory studies, particularly on at least two concurrent findings: the possible element of social protection and effective targeting mentioned earlier, and the observed increase in food insecurity in regions with lower NMR during the pandemic period, which may have driven the rise in NMR in more socioeconomically privileged regions, slowed the national NMR trend, and consequently reduced intranational inequality in NMR.

Available evidence points to more proximal determinants related to health services, which are especially sensitive to disruptions caused by lockdowns –both on the supply and demand sides– and other pandemic containment measures [54, 55]. Inequalities in NMR are greater where prenatal care access –and its quality– is more unequal. Evidence also points to more distal determinants, such as higher out-of-pocket spending, high fertility rates, and low maternal education levels [56–59]. Some recent studies also point out to persistent inequalities in antenatal care and essential newborn care disproportionately concentrated in local settings with relatively high neonatal mortality, low access to quality care in facilities, and among indigenous populations, particularly in the Peruvian Amazon [60–62].

Another aspect to consider is the observed change in the epidemiological profile of neonatal death over the years, showing a trend toward universal institutional birth and death, and a predominance of preterm births (< 37 weeks gestation), low birth weight (< 1,500 g), and prematurity as the leading cause of death. These findings, however, should be interpreted with caution, as they come from a neonatal death surveillance system that captures about 60% of the annual mortality burden and tends to overrepresent the profile typical of higher-complexity health services [63]. The finding that 15% of neonatal deaths are related to lethal congenital malformations is concerning, with cardiac and neural tube defects being the most frequent. Peruvian studies show an increase in outpatient visits for congenital heart disease, although overall mortality related to this condition has remained stable. This finding could be associated with the introduction of the Comprehensive Health Insurance (SIS), which has enabled greater access to specialized health services since its creation in 2002 and its expansion in the late 2010s. Additionally, in 2013, two new specialized centers for pediatric congenital heart disease care were established. Data from a congenital malformations surveillance system could help better define the scope and relevance of this issue in the epidemiology of neonatal deaths in Peru.

We found that inequality gaps are evident: in 2021, for example, among the seven regions comprising the quartile with the highest NMR (weighted average: 13.8 per 1,000 live births), 13.8% of births and 18.4% of neonatal deaths occurred at home, while in the quartile with the lowest NMR (weighted average: 6.2 per 1,000 live births), these frequencies were 2.2% and 1.0%, respectively. Moreover, in the high-NMR quartile, 35.1% of neonatal deaths were born with low birth weight and 65.3% were preterm, compared to 50.0% and 73.4%, respectively, in the low-NMR quartile. While the structural determinants of these inequalities are evident, these gaps also reflect more proximal determinants associated with the content and quality of prenatal care [64], which warrant further investigation.

Our study carries the limitations inherent to an ecological design, including the inability to suggest –let al.one establish –individual-level inferences (known as the ecological fallacy), meaning that associations observed at the regional level cannot be assumed to apply to individuals. Furthermore, this design makes it difficult to control for confounding, temporal ambiguity, and data incompleteness. However, our analytical intent is exploratory –not causal– and aimed at describing contextual effects and distributional patterns of neonatal mortality risk across territories as units of analysis, identifying research priorities and promoting policy advocacy for equity in neonatal health [65]. Another important limitation is the low granularity of the data, which may mask inequalities, especially in regions with high urban population concentration, as documented during the COVID-19 pandemic [66]. Similarly, while our study descriptively identified the ‘pandemic pivot’ in inequality, a formal decomposition analysis to explain this trend through its determinants was beyond our descriptive, ecological scope and remains an important area for future research. The quality and completeness of administrative data is another potential limitation, particularly those from the neonatal death surveillance subsystem, which still suffers from significant underreporting –unevenly distributed across regions, concentrated in the most socially disadvantaged departments with the highest NMR, located in the Andean area of the country. Although our study focuses on the 2007–2021 period, future research could extend the analysis to more recent years to explore whether the trends observed during the early COVID-19 period persisted, reversed, or evolved further in the post-pandemic context. In contrast, our study showed very high consistency between national NMR estimates and those generated by other sources, including survey data and internationally recognized estimates [4, 14, 15].

It is imperative to build accountability mechanisms for the commitment to “leave no one behind” on the path to sustainable development adopted globally in the 2030 Agenda and other high-level global and regional agreements. Monitoring only national NMR trends is necessary but insufficient to objectively fulfill that commitment: alongside these average changes, distributional changes in NMR must be tracked –especially those associated with population gradients of relative social position defined by intermediate and structural determinants and their intersectionality [67, 68]. Research on social inequalities in neonatal mortality and survival must be promoted and deepened –using both quantitative and qualitative analytical methods– to clarify the underlying mechanisms of their generation and perpetuation, and the drivers of their change. It is equally essential to strengthen institutional capacities in ministries of health and public agencies for monitoring social inequalities in health, prioritizing the analytical use of individual-level microdata generated by surveys as well as high-quality, granular administrative data –not only to quantify the depth of inequalities but also to assess the impact of health policies, programs, and interventions on their distributive equity.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AAPC

Average Annual Percent Change

APC

Annual Percent Change

BIC

Bayesian Information Criterion

CDC-Peru

National Center for Epidemiology, Prevention and Disease Control, Lima, Peru

CHE%gdp

Current Health Expenditure as a percentage of GDP

CHEpc

Current Health Expenditure per capita

CIx

Concentration index of inequality

COVID-19

Coronavirus Disease 2019

DALYs

Disability-Adjusted Life Years lost

DDS

Data Dependent Selection

ENDES

Encuesta Nacional de Demografía y Salud (Demographic and Health National Survey)

FI

Food Insecurity

GDP

Gross Domestic Product

IGME

United Nations Inter-agency Group for Child Mortality Estimation

IHME

Institute for Health Metrics and Evaluation

INEI

Instituto Nacional de Estadística e Informática (National Institute of Statistics and Informatics)

LAC

Latin America and the Caribbean

MCLCP

Mesa de Concertación para la Lucha Contra la Pobreza (Roundtable for the Concerted Fight Against Poverty)

MDG

Millennium Development Goals

MP

Monetary Poverty

MSE

Mean Squared Error

NMR

Neonatal Mortality Rate

PAHO

Pan American Health Organization

PMD

Peruvian Mortality Database

REUNIS

Repositorio Único Nacional de Información en Salud (National Unified Health Information Repository)

SDG

Sustainable Development Goals

SHAA

Sustainable Health Agenda for the Americas

SII

Slope Index of Inequality

SIS

Seguro Integral de Salud (Comprehensive Health Insurance)

SSVEPN

Subsistema Nacional de Vigilancia Epidemiológica Perinatal y Neonatal (National Perinatal and Neonatal Epidemiological Surveillance Subsystem)

U5MR

Under-5 Mortality Rate

UBN

Unmet Basic Needs

WHO

World Health Organization

Author contributions

JA, AVM, OJM, GS, AS designed the study. JA and OJM were responsible for data quality and curation. AVM, OJM conducted the formal analysis. JA, AVM and OJM wrote the original draft of the manuscript. GS, AS, NO, MR, LTR, OJM, and CM wrote and revised the manuscript. All authors provided feedback and revised the text, tables, graphs, and supplementary information. All authors read and approved the final manuscript.

Funding

Not applicable.

Data availability

These analyses utilized publicly available data collected by third parties. Datasets from INEI, MCLCP and REUNIS are accessible through their respective data repositories.

Declarations

Ethical approval

Not applicable. This manuscript is based on data collected by third parties; therefore, no additional ethical approval was required. The data derived from surveys had already received the necessary ethical approval from the responsible institutions overseeing their implementation in Peru.

Consent for publication

Not applicable.

Disclaimer

The views expressed in this article are solely those of the authors and do not necessarily reflect the official policy or position of their respective institutions.

Competing interests

The authors declare no competing interests.