Introduction

Diarrhea remains a leading cause of child morbidity and contributes to 19% of deaths in children under 5 years of age [1]. A plethora of studies have described the epidemiology as well as risk factors associated with diarrheal morbidity and mortality, including poor infant feeding practices, low maternal education, and poverty. In a multi-country analysis, the absence of breastfeeding was associated with an almost sixfold increase in risk of death during early infancy [2]. Although the effect of no breastfeeding decreased with age, low maternal education more than doubled the infant mortality risk associated with not breastfeeding, presumably in part due to inadequate knowledge and resources to prepare a safe substitute for breast milk. Similar results were reported in a study in urban Accra, Ghana [3]. Diarrheal prevalence among infants and preschool-age Ghanaian children was highest among poor households, those with illiterate mothers as well as among those with inadequate hygiene and water quality. Improving water and sanitation services for households is often the focus of interventions to reduce diarrhea. Cairncross et al. analyzed recent systematic reviews to identify the reduction in diarrheal mortality that could be attributed to specific interventions prioritized for scaling up [4]. The reduction in diarrheal mortality was estimated at 43% with proper hand washing, 36% with adequate excreta disposal, and 17% with improved household water quality.

The HIV/AIDS pandemic has added an additional concern for young child survival as HIV infection may compromise a household’s ability to provide adequate child care. Reduced financial resources as well as compromised physical and mental health of caregivers may contribute to poor hygienic environments and suboptimal infant feeding practices; misguided beliefs about disease etiology may contribute to limited use of health care services when children become ill [5]. HIV infection has been reported to be associated with a twofold increase in risk for adult major depressive disorders [6] which may, in turn, be a contributing factor to the observed increase in infant morbidity [7] and mortality from infectious diseases, even among HIV-negative infants [8]. Recent studies have identified maternal mental health as an important determinant of infant health outcomes. Rahman and colleagues reported that postnatal depression (PND) was associated with a three-fold increase in the risk of having five or more diarrheal episodes per year among Pakistani infants [9]. These studies were conducted, however, in non-HIV-affected populations. No known study has examined the association between PND and infant health in HIV-affected populations. The co-existence of HIV infection and maternal depression may synergistically influence infant morbidity outcomes including diarrhea. The present study examined the incidence and number of days ill with diarrhea among Ghanaian infants living in communities affected with HIV and examined the interaction between maternal HIV status and symptoms of PND on infant risk of having diarrhea. We hypothesized that:

  1. (i)

    maternal HIV infection would be associated with infant diarrheal morbidity,

  2. (ii)

    symptoms of PND among mothers would be associated with infant diarrheal morbidity, and

  3. (iii)

    there would be a synergistic effect of maternal HIV infection and symptoms of PND such that infants of women who were affected by both conditions would have a risk of diarrhea that was greater than the sum of the individual risks associated with the two conditions.

Methods

Study Site

This study [Research to Improve Infant Nutrition and Growth (RIING) Project (2004-2008)] was conducted in the Manya and Yilo Krobo districts, Eastern region, Ghana. This region has consistently recorded the highest HIV prevalence; the 2007 prevalence (4.2%) was more than twice the national rate of 1.9%. The Manya Krobo district was the first in the country to benefit from services and support for voluntary counseling and testing (VCT) for HIV in 2001.

Recruitment and Follow-up

Recruitment of pregnant women occurred at prenatal clinics in three hospitals. Pregnant women were recruited and enrolled after VCT. In 2003, under the national VCT program, all pregnant women received counseling on the risk of HIV transmission and were informed about the available services, after which they were invited to be tested for HIV (‘opt-in’ system). In 2005, the national program was changed to an ‘opt-out’ system in which all women were tested after counseling unless they requested not to be.

Women were considered eligible for the study if they (1) were at least 18 years old, (2) were pregnant at the time of enrollment, and (3) had completed pre- and post-VCT, or only pre-VCT if they had refused testing. For those who completed pre- and post-VCT, they were considered eligible if they agreed to have their HIV test results released to the project coordinator. The rapid test ‘Abbott Determine HIV-1/2′ (Abbott Laboratories, Abbott Park, IL, USA) was used at the hospital to test for HIV. Samples from indeterminate results were sent to the Noguchi Memorial Institute for Medical Research (NMIMR) for confirmation with polymerase chain reaction (PCR) analysis. The participants were classified as HIV positive (HIV-P, tested positive for HIV), HIV negative (HIV-N, tested negative for HIV) or unknown HIV status (HIV-U, refused to be tested for HIV). A participant was considered eligible for the postpartum longitudinal surveillance if (1) she gave birth to a live infant, (2) her infant was free of any birth defect that may otherwise hinder feeding, and (3) she was free of any physical condition that may limit her ability to care for her infant. Only the field supervisor had access to women’s HIV status. There was no information available on participants’ CD4 counts and viral loads at the time of entry into the study as these tests were not done routinely by the Ghana Health Services. Antiretroviral therapy was limited to the administration of Nevirapine by Ghana Health Services staff to the mother during labor and to the newborn at birth to prevent mother-to-child transmission of HIV.

Mother–infant pairs were followed from birth up to 12 months. Twice a week, mothers were visited to collect information on infant health, including total number of stools and stool characteristics (e.g., liquid). For each symptom, the mother was asked to recall the events for the day of visit and for each preceding day since she was last visited by the field worker for up to a maximum of 7 days of recall. Infant feeding information was collected with the same methodology. The mother was asked if she breastfed the baby and if the baby received any non-breast milk liquids (water, infant formula, other non-human milk) or semi-solid or solid food.

Symptoms of maternal PND were measured shortly after birth (10.0 ± 0.5 days) and at 6 months (189.0 ± 0.4 days) using the Edinburgh Postnatal Depression Scale (EPDS), a 10-item scale that documents symptoms that occurred over the past 7 days [10]. Each item on the EPDS was scored on a 4-point rating scale that represented the level of occurrence. The Cronbach’s alpha for the EPDS was 0.82. The day on which the scale was administered did not differ among the three HIV groups at either time point (birth: χ2 = 3.42, P = 0.18; 6 months: χ2 = 0.28; P = 0.87).

Maternal stress was measured at enrollment (prenatal) using the Perceived Stress Scale (PSS), a 4-item scale (Cronbach’s alpha = 0.87) used to measure the degree to which situations in a person’s life are perceived as stressful [11]. Each item on the PSS measures the frequency of occurrence. Total scores range from 4 (low stress) to 20 (high stress).

Infant weight was measured at birth, then monthly. Maternal weight was measured monthly and height was measured once at 6 months postpartum. Duplicate weight measurements were made using the Tanita digital weighing scale (Tanita Corporation of America Inc., Arlington Heights, IL, USA) and measured to the nearest 0.1 kg. Maternal height was measured in duplicate to the nearest 0.1 cm using a wooden stadiometer (Shorr Productions; Maryland, USA).

Variables

Diarrhea was defined as passing three or more liquid or semi-liquid stools in a 24 h period [12]. An episode was separated from another by at least two symptom-free days [13]. Incidence rate was calculated as the number of new episodes divided by the number of days at risk (days at risk = (days observed) − (# days ill) + (# episodes)) [14]. Diarrheal incidence and days ill were standardized to 100 days of observation. The main outcome measures included (i) total number of days ill per days observed, (ii) incidence rate of diarrhea, and (iii) risk of having diarrhea on any day during the first 3 months of life.

An infant was considered to be exclusively breastfed (EBF) if since birth the infant received only breast milk (excluding medicines). EBF ceased on the day that the infant received any non-breast milk liquids or semi-solids.

A binary variable was used for maternal PND. Mothers with total EPDS scores below 13 were classified as not showing symptoms of PND whereas those with total EPDS scores 13 and above were classified as showing symptoms of PND [10, 15]. Maternal perceived stress score was analyzed as a continuous variable.

Statistical Analysis

Morbidity was described for four age intervals (0–3, 3–6, 6–9 and 9–12 months) and for the first year of life (0–12 months). The descriptive data are presented as means and standard errors or frequencies and percentages. Chi-square or Fisher’s exact test were used for categorical variables and Analysis of Variance for normally distributed continuous variables. The non-parametric Kruskal–Wallis test was used for non-normally distributed continuous data. The greatest unadjusted effect of PND on illness was in early infancy, therefore the focus of the analysis using multiple logistic regression included data only from 0 to 3 months using the SAS PROC GENMOD procedure. Logistic regression provided a better fitted model than linear regression because of the frequent response of zero or few days ill among our sample. The models were fitted with the assumption that the data were generated from a binomial distribution. Overdispersion, a condition that occurs when the variance is larger than expected under a given model, was checked and adjusted for (using the D-scale syntax) whenever it occurred. The initial model to explain risk of diarrhea included maternal (parity, marital status, prenatal stress score), infant (EBF status, low birth weight), and household (primary water source, cooking fuel, toilet facility) factors that were identified through the bivariate analysis and the review of the literature. An HIV-PND interaction term was included in the model to test for the interaction between HIV status and PND; the reference cell was HIV-N with no depressive symptoms. Variables were removed from the model in a manual backward stepwise procedure until only significant variables remained. All analyses were completed with SAS version 9.13 for Windows (SAS Institute, Cary, NC, USA) and significant results were reported at α < 0.05 unless otherwise indicated.

Ethics

Ethical approval was obtained from the institutional review boards of McGill University, University of Ghana, Iowa State University, and University of Connecticut. Signed informed consent was obtained from each woman for herself and her infant.

Results

Study Sample

Six hundred and ninety-two pregnant women were informed about the study, out of which 552 women were enrolled. Out of 505 recorded live births, morbidity data were available for 492 infants.

Baseline Characteristics

Maternal age ranged from 18 to 48 years with a mean of 28.5 ± 0.3 years. Study subjects had 0-8 prior live births, with a median of 1. The women reported illness symptoms (respiratory, gastrointestinal, skin, and gynecological) on about 18% of the observed days, with no difference by HIV status. The majority of women (69%, n = 339) were of the Ga-Adangme ethnic group. More than two-thirds of the women were not married and were either living with a partner (n = 245) or had no partner (n = 95). Most women (n = 440) had some form of formal education although one-third (n = 162) had not reached secondary school. The public tap served as the main water source for the majority of households (67%, n = 328). Less than one-sixth (16%, n = 77) of the women had flush toilets. Wood was used by a minority of households (10%, n = 50) as the main cooking fuel. On the whole, HIV-P women were poorer than other women, as indicated by less secondary or higher education, lower likelihood to be married and less access to flush toilet, in-house tap and charcoal, gas or electric stove (Table 1).

Table 1 Demographic and socioeconomic characteristics of Ghanaian study population at baseline, by maternal HIV status
Full size table

HIV and Diarrhea

The total number of days ill with diarrhea was 2.2 per 100 days observed (8.0 days ill/year). There was an average of 0.6 new episodes of diarrhea per 100 days at risk (2.3 episodes/year). Over the 12 months, although the diarrheal incidence was similar among the three groups of infants, infants of HIV-P mothers tended to have a lower number of days ill (P = 0.06; Table 2).

Table 2 Infant diarrheal days ill (per 100 days of observation) and incidence of diarrhea (episodes per 100 days at risk), by maternal HIV status
Full size table

Maternal Postnatal Depression and Diarrhea

About 10% of women reported symptoms of PND shortly after birth. HIV-P women tended to be more likely to report symptoms at birth than HIV-N and HIV-U women (14% vs. 7% and 10%, respectively; χ2 = 4.76; P = 0.09). The analysis of diarrheal morbidity by maternal PND status showed that having symptoms of PND shortly after birth was associated with more episodes and days ill with diarrhea (Table 3). In the first 3 months of life, infants of mothers reporting PND symptoms had almost twice the number of diarrheal episodes (χ2 = 12.24; P = 0.0005) and more than twice as many days ill with diarrhea (χ2 = 13.81; P = 0.0002) compared to infants whose mothers reported no PND symptoms. No significant association was seen after three months. At 6 months, 9.3% of women reported PND symptoms; these reports were more common among HIV-P women than HIV-N or HIV-U women (17% vs. 3% vs. 10%, χ2 = 14.83; P = 0.0006). PND symptoms reported at 6 months were not significantly associated with diarrheal morbidity between 6 and 12 months (data not shown).

Table 3 Infant diarrheal days ill (per 100 days of observation) and incidence (episodes per 100 days at risk), by symptoms of maternal postnatal depression reported by mothers shortly after birth
Full size table

Other Maternal, Child, and Household Factors and Diarrhea

Increased risk of diarrhea in the first three months of life was associated with parity, PND symptoms, single marital status, and use of wood as cooking fuel and decreased risk with low birth weight (Table 4). These factors, with the exception of parity, remained significant in the adjusted model (Table 5). Maternal stress was a significant predictor of risk of diarrhea only in the adjusted model. Sanitation indicators (water source and latrine) were not significant in either the unadjusted or adjusted model. Exclusive breastfeeding was common in this population: 87, 80 and 63% infants were EBF at 1, 3, and 5 months, respectively. Exclusive breastfeeding during the first 3 months was not a significant factor in either model.

Table 4 Unadjusted odds ratios for explanatory variables of infant risk of having diarrhea in the first 3 months of life
Full size table
Table 5 Factors associated with infant risk of having diarrhea in the first 3 months after birth
Full size table

Maternal HIV Status-PND Interaction

To examine the synergistic effect of maternal HIV infection and symptoms of PND on risk of diarrhea, the analysis focused on the first three months postpartum when a strong association between risk of infant diarrhea and maternal PND assessed soon after birth was noted. Multiple logistic regression results showed a significant interaction between maternal HIV status and PND symptoms soon after birth on infant risk of having diarrhea in the first three months of life (Table 5). The association between PND symptoms and infant diarrhea differed according to maternal HIV status. Compared to infants of HIV-N women who reported no PND symptoms, infants of HIV-N women with symptoms had no increased risk whereas infants of HIV-P and HIV-U women with symptoms had a twofold and threefold increased risk of diarrhea, respectively. Infants of HIV-P women who showed no symptom of PND however had a reduced risk of having diarrhea compared to infants of HIV-N women who showed no symptom of PND.

Finally, the difference in diarrheal risk that was associated with PND symptoms was compared within HIV groups, adjusting for perceived prenatal stress, marital status, low birth weight, and cooking fuel. Whereas there was no difference in infant risk of diarrhea by reported PND symptoms among HIV-N women (OR = 0.99, 95% CI: 0.28–3.43), reporting PND symptoms was associated with a large increase in diarrheal risk for infants of both HIV-P and HIV-U mothers (Fig. 1). Infants of HIV-P and HIV-U women had a fivefold (OR = 5.39, 95% CI: 2.48–11.68) and fourfold (OR = 4.53, 95% CI: 2.47–8.32) increased risk of having diarrhea, respectively, if their mother reported PND symptoms compared to if their mother did not have PND symptoms.

Fig. 1
figure 1

Infants’ risk of having diarrhea on any day, 0 – 3 months of age, associated with symptoms of postnatal depression (PND), by maternal HIV status. Symbols indicate adjusted Odds Ratios (aOR; adjusted for perceived prenatal stress, marital status, low birth weight, and cooking fuel); vertical lines indicate 95% CI. The comparisons shown use absence of PND symptoms as the reference group within each HIV category: HIV-P (aOR = 5.39, 95% CI: 2.48-11.68); HIV-U (aOR = 4.53, 95% CI: 2.47-8.32); HIV-N (aOR = 0.99, 95% CI: 0.28-3.43). Dotted line represents OR = 1. HIV-P, HIV positive; HIV-N, HIV negative; HIV-U, unknown HIV status

Full size image

Discussion

To the best of our knowledge, this is the first study to show a synergistic relationship between maternal HIV status and symptoms of PND on an indicator of infant health (risk of having diarrhea). One explanation of the increased risk may be that the presence of PND contributes to HIV-P mothers performing fewer preventive child health behaviors; however, the evidence on PND is inconsistent in the literature. Grupp-Phelan and colleagues [16] reported that mothers with symptoms of depression perceived that they were less able to “take care of their children in the way they would like to.” In a study by Minkovitz et al. [17], mothers with PND symptoms were about 20% less likely to carry out preventive health care behaviors (well-child visits, vaccinations) as women without symptoms. In contrast, breastfeeding and timely well-baby visits did not vary by presence of symptoms in a study by Chung et al. [18]. These inconsistent findings may reflect different social environments, the severity of other stressors that women experienced, and the quality of health services received. Watson and Kemper [19], for example, reported that maternal depression did not explain frequency of delayed immunizations in a population highly stressed by poverty and drug abuse. There may be two reasons for their results: (i) the effect of PND was not detectable because other factors such as drug use were much stronger determinants of child health, or (ii) the sample selected (a population already receiving health services) was one in which clinic support successfully mitigated the effect of PND. A close examination of the overall environment and services as well as the co-existence of PND with other stressors (such as HIV in this study) will help elucidate the role that PND plays in child health in diverse settings.

Increased infant diarrheal morbidity may also have occurred because HIV-P women with PND symptoms were clinically sicker and therefore unable to adequately care for their infants. Antelman and colleagues reported an increase in risk of disease progression (HR = 1.61, 95% CI: 1.28–2.03) and all-cause mortality (HR = 2.65; 95% CI: 1.89–3.71) among Tanzanian HIV-infected women who were depressed [20]. This explanation, however, is unlikely in our study as most women were in early stages of the disease and maternal illness rates were similar across HIV status.

The only other study that has examined the relationship between maternal depression and infant diarrhea was conducted in a non-HIV population in Pakistan and demonstrated similar results as our study. The researchers reported a three-fold (OR = 3.1; 95% CI: 1.8–5.6) increase in risk of frequent diarrheal illness (defined as ≥5 diarrheal episodes/year) among infants of women who were depressed [9]. Underweight (<−2 weight-for-age) at 6 months in the Pakistani study was common (18% overall) and was associated with more than twice the risk. We chose to not include a similar mid-point anthropometric predictor in our model for various reasons. First, only predictors that occurred prior to the outcome of interest were included to clarify temporal relationships. Second, differences in anthropometric measurements were pronounced from birth and retained thereafter (data not shown). Finally, birth weight is highly correlated with weight during the first year of life and therefore an additional weight indicator may be redundant in the model.

The estimate of diarrheal incidence (2.3 episodes/year) in this study is similar to that reported in a WHO global review (2.6 episodes/year) [21] and a study of non-HIV affected children in Turkey (2.8 episodes/year) [22]. This estimate contrasts the reports of 7.7 –18.6 episodes per year in South Africa, Kenya and Zimbabwe [23]. The comparatively lower rates recorded in Ghana could be due to high breastfeeding and exclusive breastfeeding (EBF) rates; only nine infants of HIV-P mothers (2.7%) did not breastfeed at birth and only 20% of all mothers were not EBF at 3 months. Exclusive breastfeeding is protective against diarrhea [24, 25]. Among HIV-uninfected infants of HIV-P mothers, non-exclusive breastfeeding was associated with a 2.5 fold (95% CI: 1.7–3.8) increased risk of diarrhea [26]. The high EBF rate seen in our study is consistent with another Ghanaian study [27], and may have reduced incidence rates to a level at which we were unable to detect significant differences in diarrhea across HIV groups. Regardless, the HIV group difference was in the opposite direction of what was expected. Infants of HIV-P women tended to report less, not more, diarrheal illness for their infants. This unexpected relationship may reflect the health education in the HIV clinics. The VCT nurses provided HIV-P women with one-on-one counseling that included infant feeding and care. This education may also explain the lower risk of diarrhea associated with low birth weight as the majority of low birth weight babies were in the HIV-P group.

The lack of a significant direct association between maternal HIV and infant diarrhea in other settings may also reflect the environment in which the study is conducted. For example, maternal HIV was also not associated with increased diarrhea morbidity among 6-month-old infants in Botswana even though overall infant mortality was higher [28]. In contrast to our study, the 5 month EBF rates were low (17.5% vs. 9.5% for HIV-P vs. HIV-N, respectively) and infants were weaned early (median was 5.8 and 9 months among HIV-P and HIV-N women, respectively). About one-third of 6-month-old infants had at least one episode of diarrhea regardless of whether their mother was HIV-P or HIV-N, suggesting that other factors, such as poor hygiene and non-EBF, were the primary determinants of diarrhea.

Maternal marital status was significantly associated with diarrhea in this study. Being unmarried may indicate poor socioeconomic status [29] and low socioeconomic status may affect a mother’s ability to access adequate resources for the care of her child. Mothers who were not married had infants with increased risk of having diarrhea; this association is consistent with the increased risk seen with another indicator of poverty, the use of wood as the main source of cooking fuel. Poverty was positively associated with diarrhea morbidity in a previous study in Ghana [3].

Twice weekly data collection reduced errors from poor maternal recall and allowed for confirmation of health status by the field staff when the illness coincided with the visit. Field workers were masked to mothers’ HIV status, thereby reducing differential bias. Despite these methodological strengths, this study had limitations. Firstly, maternal HIV status was tested only at enrollment. Some mothers may have become infected during the study; new infections would have diminished our ability to observe group differences. Secondly, our analysis did not include the infant’s HIV status. At the time of the study, the health ministry protocol included testing of infants only at 18 months of age or with presentation of symptoms. Although the project offered free testing for all infants, few mothers agreed to have their child tested. Infant HIV infection is associated with increased morbidity; however, given the decreased risk of HIV transmission with EBF compared to mixed feeding, the majority of infants may have remained uninfected during early infancy.

In summary, infants of HIV-P and HIV-U mothers who reported symptoms of PND had an increased risk of having diarrhea in the first three months of life. Health care programs in HIV-affected communities need to train their personnel to identify and address the multiple stressors that contribute to poor maternal mental and physical health and affect infant health. Multisector prevention and curative programs that address the complex situations in which HIV-affected households live are needed to improve maternal well-being and decrease infant morbidity and mortality.