Childhood Malnutrition and Postwar Reconstruction in Rural El Salvador: Title and subTitle BreakA Community-Based Survey

During the second half of the 20th century, expensive international efforts to relieve suffering and promote peace in countries at war have become common, although doubts have been expressed regarding their effectiveness.^{1 - 3} One such effort occurred in El Salvador, following a 1992 agreement⁴ that ended a 12-year civil war notorious for human rights abuses.^{5 - 9} The Salvadoran war was rooted in economic inequity,^{10 - 12} especially for land distribution.¹³ War exacerbated the chronic problem of hunger.^{14 - 15}

The peace settlement's economic accords, which favored demobilized ex-combatants and landless peasants, attempted to motivate combatants to lay down their arms and to rectify some prewar economic inequities. The international donor community contributed approximately US $200 million per year in reconstruction aid.¹⁶ The original plan called for basic services, such as health and water, to be restored to zones of conflict as soon after January 16, 1992, as possible and for land to be legally redistributed in zones of conflict within 180 days.⁴

The principal elements of reconstruction as originally envisioned—housing, latrines, water, land, and primary health services—would have alleviated, at least in part, the severe resource constraints that had been associated with malnutrition and mortality among children in rural El Salvador.^{17 - 23}

Six years after the peace agreement, we attempted to describe the association between the reconstruction effort and child health in one rural area, using child nutrition as an indicator.²⁴ Our hypotheses, supported by findings of a 1997 pilot study, were that: (1) the prevalence of childhood malnutrition, in the form of stunting, was still high even in communities that had received significant amounts of postwar reconstruction aid; (2) delays in the postwar land transfer process were positively associated with child malnutrition; and (3) delays in providing basic services to the target population were also positively associated with child malnutrition.

The study site was a single municipality in Cuscatlán, El Salvador, whose peasant population had fled around 1980 and began to repopulate in 1986 (Figure 1). Thirty-seven of approximately 50 rural villages were resettled by the end of 1993, in time to be beneficiaries of the postwar reconstruction. Logistic considerations limited our study to 27 of the original 37 villages. Twelve of the 27 villages had existed before the war. The known characteristics of the villages included in the study did not differ from those of villages not included. The study population was defined as children younger than 5 years who lived with their families in any of the 27 villages in January 1998.

Eligible single- or multiple-family households were identified by village health workers. In smaller villages, we attempted to visit each eligible household. In larger villages, we chose 1 or more locations from which the survey teams randomly (by coin toss) selected households to be visited. Overall, 69% of eligible households were surveyed. In only 1 village was the sampling fraction less than 50%.

We also interviewed village governing council members about village history, population, water, sanitation, housing, schools, roads, health, and participation in the postwar land redistribution program (Programa de Transferencia de Tierras [PTT]).

In each household, we administered a questionnaire on land tenure, farming, water, housing, family size and composition, latrines, maternal education, and birth dates of young children. Families who shared living quarters were presumed to share agricultural work.

Children younger than 5 years were weighed (lightly dressed; weight recorded to the nearest 100 g) and measured (standing if >2 years, otherwise recumbent; height recorded to the nearest millimeter) in accordance with international protocols.²⁴ Interviews and anthropometric measurements were carried out by 4 Spanish-speaking physician-investigators (intraclass correlation for interrater reliability >0.99). No family refused interviews. Four children refused to be either weighed or measured.

Height-for-age, weight-for-height, and weight-for-age z scores were calculated using reference curves developed by the National Center for Health Statistics and the Centers for Disease Control and Prevention.²⁵ Stunting, wasting, and underweight were defined as height-for-age, weight-for-height, or weight-for-age z score of less than −2. Two extreme values (z scores of ≤−6 or ≥6) were excluded from subsequent analyses.

We used the χ² test to evaluate crude associations between socioeconomic factors and stunting, our principal outcome of interest, and multivariate logistic regression to estimate adjusted odds ratios (ORs) and their 95% confidence intervals (CIs). Because child outcomes within a given village tend to be correlated, we used a generalized equations estimation approach and reported results based on robust variances. Analyses were carried out using SAS Version 6.12 software (SAS Institute, Cary, NC).

The study protocol was approved by Physicians for Human Rights and by representatives of principal health programs operating in the study region. Informed consent was obtained from parents or other heads of participating households and from village governing council members.

General characteristics of the study villages are described in Table 1.

All villages had received aid from many sources since resettlement. The total cost of this aid is unknown, but land purchase alone probably cost more than US $1 million.¹⁶

The complex land negotiations following the peace settlement required international financing and were behind schedule. Nearly 6 years after the original target date for land transfer, some villages still sought suitable acreage, and some purchases had failed in apparent swindles. The median lag time between the original target date and transfer of land title to village groups was 4 years. A 2-step process in which land was first assigned to groups and then subdivided into individual plots compounded this delay. Much of the redistributed land was of poor quality, as is most land in El Salvador.²⁶

All postresettlement water systems and latrines had been constructed by nongovernmental aid agencies. Only 1 village had had a working water system at the time of repopulation. All villages had trained lay village health workers at some point, but services available were limited by lack of financing, supplies, and supervision.

We visited 458 households, comprising 491 families and 761 children younger than 5 years. Most houses consisted of a single room. Households are described in Table 2.

Each household containing a PTT beneficiary had received or expected to receive a mean of 3.5 hectares of land, of which less than a third was actively cultivated. Principal reasons given for not cultivating all PTT land included lack of funds for seed and fertilizer (agricultural loan interest rates were about 20%), poor land quality, land too far away to work without moving, delays in assignment of individual plots, designation of transferred land as protected forest, lack of able-bodied men in the household, and the physical difficulty of farming when one's heaviest implement is a machete.

Overall prevalences of stunting, wasting, and underweight (shown by age and sex in Table 3) were 32.4%, 1.3%, and 9.2%, respectively.

Crude associations of important socioeconomic factors with stunting are presented in Table 4. Factors found to be strongly associated with stunting were smaller land area under cultivation, delayed first cultivation of PTT land, less crop variety, more children younger than 5 years in the household, lack of piped water, and not tending livestock. Table 5 shows the estimated associations between stunting and amount of PTT land received, amount of PTT land cultivated, date of first cultivation, and access to water, adjusted by child, maternal, household, and village characteristics. Multivariate adjustment did not significantly change direction or magnitude of odds ratios (Table 5).

We did not find a significant association between land area theoretically owned by PTT beneficiaries and stunting, but we did find a strong inverse association between PTT land area cultivated and stunting (OR for stunting per additional hectare of PTT land cultivated, 0.64; 95% CI, 0.44-0.93). Prevalence of stunting increased steadily as cultivated land area decreased, especially below the 2-hectare threshold. Most study children (84.7%) lived in households cultivating 2 hectares or less of PTT land.

We also found that stunting was less prevalent in families who had been able to cultivate their PTT land for longer periods (OR for stunting per additional year of land cultivation, 0.91; 95% CI, 0.87-0.95). Longer periods of PTT land cultivation were strongly associated with greater crop variety and with greater area of land under cultivation, factors that probably mediated the relationship between longer time of cultivation and better nutrition. These 2 factors were therefore not included in the logistic model estimating the association between time since first cultivation and stunting.

Children living in villages that had piped (nonpotable) water for 7 or more years were significantly less likely to be stunted than those living in waterless villages (OR, 2.72; 95% CI, 1.87-3.96). We found that 38.8% of children still lived in villages without piped water. The OR for stunting per additional year of access to piped water was 0.91 (95% CI, 0.87-0.95).

We elected to use stunting as our principal outcome of interest. Stunting is more likely than wasting or underweight to reflect cumulative health insults occurring during the postwar period, and at the population level, a decrease in its prevalence is thought to indicate that social development is benefiting the poor.^{24 ,27 - 31} Stunting, though often unrecognized, may have serious adverse health and economic consequences. It is associated with diminished intellectual capacity and impaired work performance later in life.^{32 - 35}

Stunting is not simply a marker for genetically determined short stature. Linear growth in children younger than 5 years appears to be determined far more by environmental than by genetic factors; thus, the World Health Organization Expert Committee on Physical Status advocates a universal standard of height distribution for young children.²⁴ Though eradication of stunting may require more than 1 generation, partial recovery can be observed in a single generation if adequate nutrition is provided and infections are prevented or controlled.^{36 - 37}

Nearly one third of children measured in this study were stunted. The prevalence of stunting increased steadily with age, as is typical in a population initially protected by breast-feeding then chronically exposed to infectious diseases and hunger. Overall prevalence of stunting resembled what has been reported about Salvadoran children in national surveys undertaken at the beginning of the war (28.5% in 1978), during the war (28.1% overall, 31.5% in rural areas in 1988), and immediately afterward (22.0% overall, 27.3% in rural areas in 1993).^{38 - 39} The prevalences of wasting and underweight were also similar to those previously reported (wasting: 3.9% in 1988, 2.9% in 1993; underweight: 15.0% in 1988, 10.5% in 1993).^{38 - 39} These findings are consistent with known patterns of malnutrition in Latin America, particularly in the early dry season.^{28 ,40}

Lack of maternal education, an important risk factor for stunting elsewhere, was only weakly associated with stunting in this study. However, few women studied had attained the educational levels (7th grade or higher) previously shown to be associated with decreased malnutrition prevalence in El Salvador,³⁹ and the protective effect of maternal education is apparently weaker where stunting is more prevalent.⁴¹

We were also unable to demonstrate any effect of local village health worker programs on stunting rates, although other investigations have suggested positive effects of these programs,⁴² and the primary health interventions they attempted should have been helpful. However, none of the existing programs provided food when child malnutrition was detected. And the potential benefits of sick-child case management, especially in the absence of antimicrobial therapy, may have been overwhelmed by the impact of land, food, and water shortages.⁴³

We did not observe the inverse relationship that we expected between stunting prevalence and amount of land owned by each family except at the highest levels of land ownership (≥5 hectares). The expected inverse relationship was seen, instead, between stunting and the amount of land actually under cultivation (OR for stunting per additional hectare of land cultivated, 0.64; 95% CI, 0.44-0.93). Even small increases in cultivated area were associated with noteworthy reductions in the likelihood of stunting, consistent with other published findings of significantly decreased stunting in families working more vs less than 1.4 hectares.^{17 ,20} It is likely that families with greater land area under cultivation were able to increase both household consumption of food that they grew and the ability to purchase other foodstuffs, although we were not able to distinguish the relative contributions of these 2 mechanisms.

Most study children (60.7%) lived in families cultivating less than 1.4 hectares (a class historically defined as land-poor in El Salvador) despite the PTT. Prior to the war, only 50.7% of economically active Salvadoran agricultural workers were considered land-poor or landless.¹³ The current level of functional landlessness has policy implications. It has been suggested, most famously by Prosterman,^{44 - 45} that the risk of civil strife increases with increases in the proportion of landless peasants in a national population.

The inverse association that we observed between crop variety and stunting may reflect the benefits of crop diversification or the unfavorable economics of corn monoculture (of those families growing only 1 crop, 56% grow corn) in El Salvador at present. The known deficiencies of the maize-based diet may have contributed to stunting in the corn monoculturists, many of whom reported unsuccessful attempts to raise beans (a potential source of complementary protein) for household consumption.

The observed relationship between provision of any piped water and decreased prevalence of stunting probably underestimates the importance of water, given that most piped water was rationed in the study villages, and none at all was potable. The relationship between increased access to water and decreased prevalence of stunting may be mediated in many ways.⁴⁶ For example, water systems probably favored better child health through better personal hygiene, small-scale irrigation of household fruit trees and vegetables (no water supply was adequate for large-scale irrigation projects), and redirection of maternal time from water-carrying to child care.

Our data demonstrate that 1 population deemed a high-priority beneficiary of postwar reconstruction aid still lacked potable water and basic sanitation 6 years after the cease-fire, and redistribution of land titles had not yet led to land cultivation at levels higher than those of the prewar land-poor. We observed associations between stunting and land area under cultivation, crop variety, date of first cultivation of redistribution land, and date of installation of water system (if any). These associations suggest that earlier access to PTT land (through earlier title transfer or through prepurchase agreements with sellers) allowed beneficiaries more time to clear, fence, fertilize, terrace, and plow land—slow processes without heavy farm equipment—thus gradually creating the conditions necessary for larger and more varied harvests.

Malnutrition and poverty are frustratingly multifactorial, and this study did not examine all possible contributing factors, such as recent crop failures due to inadequate rains, fluctuations in market prices of corn and beans, and the "culture of dependence" thought to have derived from refugee camp life. In addition, the cross-sectional design used in this study has inherent weaknesses. Current malnutrition prevalences may represent either improvement or backsliding from unknown baselines. However, knowledge of the overall baseline prevalence of stunting in 1992 would not change the associations found between current stunting and differing present-day patterns of land usage and access to water. The primary limitation of the cross-sectional design is probably the restriction it places on causal interpretation of findings. We suspect that delays in provision of piped water and in awarding of land titles were causally related to malnutrition, but confounding factors may have permitted certain groups to acquire and exploit land and infrastructure projects more expeditiously, while contributing independently to better child health. For example, to acquire a water system, each village had to find and purchase water sources and rights-of-way for pipes, forge alliances with competent aid agencies, haul cement, dig ditches, and organize subsequent maintenance. Communities that had acquired water or land earlier may have been better organized, richer at baseline, or better placed politically. Even so, we do not think that such confounding factors would completely explain the pattern of associations that we found within this rural area even after adjustment for socioeconomic factors.

In some ways, however, the question of causation is moot here. After 6 years of cease-fire and millions of dollars in aid, the majority of families in this study are still subject to exactly the resource constraints that make childhood malnutrition and illness predictable and social stability elusive.

Further research is necessary to determine the generalizability of those findings to other regions of El Salvador and to other countries and is necessary to elucidate the causal connections among land ownership, land use, water, and child health. Better information and surveillance may, where on-the-ground conditions are favorable, help to implement relief and reconstruction projects in ways that permit other beleaguered populations to recover more swiftly after similar crises and that promote social equity instead of perpetuating inequalities.⁴⁷