Diet and food insecurity among mothers, infants, and young children in Peru before and during COVID-19: A panel survey

The COVID-19 pandemic may impact diet and nutrition through increased household food insecurity, lack of access to health services, and poorer quality diets. The primary aim of this study is to assess the impact of the pandemic on dietary outcomes of mothers and their infants and young children (IYC) in low-income urban areas of Peru. We conducted a panel study, with one survey prepandemic (n = 244) and one survey 9 months after the onset of COVID-19 (n = 254). We assessed breastfeeding and complementary feeding indicators and maternal dietary diversity in both surveys. During COVID-19, we assessed household food insecurity experience and economic impacts of the pandemic on livelihoods; receipt of financial or food assistance, and uptake of health services. Almost all respondents (98.0%) reported adverse economic impacts due to the pandemic and 46.9% of households were at risk of moderate or severe household food insecurity. The proportion of households receiving government food assistance nearly doubled between the two surveys (36.5%–59.5%). Dietary indicators, however, did not worsen in mothers or IYC. Positive changes included an increase in exclusive breastfeeding <6 months (24.2%–39.0%, p < 0.008) and a decrease in sweet food consumption by IYC (33.1%–18.1%, p = 0.001) and mothers (34.0%–14.6%, p < 0.001). The prevalence of sugar-sweetened beverage consumption remained high in both mothers (97%) and IYC (78%). In sum, we found dietary indicators had not significantly worsened 9 months into the COVID-19 pandemic. However, several indicators remain suboptimal and should be targeted in future interventions. This was an unbalanced panel study conducted amongst low‐income urban households in Peru. We collected data via two surveys: one conducted from December 2019 to March 2020 before the COVID‐19 pandemic (PERUSANO survey) and one conducted in December 2020 (STAMINA survey), 9 months after the onset of the pandemic. The PERUSANO survey was part of a wider interdisciplinary project to address multiple forms of malnutrition, particularly stunting, anaemia, and risk of overweight/obesity, in urban Peru while the STAMINA survey was designed to examine the nutritional risks of mothers and IYC in the same communities during COVID‐19. The surveys took place in Manchay, Pachacamac District, Lima, and the city of Huánuco, Huánuco District in the Andean Highlands. In each study area, we purposively selected the principal health centre and one subsidiary health centre. Periurban (low‐income urban) communities within the jurisdiction of these health centres were selected to participate. In the PERUSANO survey, we recruited participants by systematic random sampling using quotas via house‐to‐house visits. Before recruitment, enumerators mapped each block of houses within each “sector” (local planning administrative unit for urban areas) of the community within the health centre catchment, and the block was used as the sampling unit. We chose a block at random as the starting point from the mapped sector, then proceeded to knock on the first house, and every third house thereafter until completing the sector. A random starting point was chosen for the next sector, and recruitment continued. The target sample size for PERUSANO was 360 mother–infant dyads based on the original aims before the emergence of COVID‐19. The sample size rationale was to capture the diversity of characteristics relevant to IYC diets, food patterns, and practices via quota sampling by age (6–11, 12–17, and 18–23 months), maternal employment (employed/nonemployed) and study area (Lima/Huánuco). Recruitment stopped in March 2020 because of the pandemic, at which point 244 mother–infant dyads had been recruited. For the STAMINA survey, the sample size was set to match the pre‐COVID sample (n = 250) with the same quota sampling by age and study site. Sampling took place through follow‐up of the pre‐COVID‐19 survey participants (PERUSANO) for those still eligible (aged 6–23 months) (24% of the sample). We recruited the remaining participants using systematic random sampling from the local authority child registration records for the participating health centres. From the total number of eligible IYC records within an age group, we selected every nth record to attain the required number of participants. At least two attempts were made to contact participants via telephone before noting them as unavailable. If one participant declined or was unavailable, the next birth record after that case was selected. In both surveys, a screening questionnaire was used to check the eligibility of mothers and IYC with the following inclusion criteria: i. singleton infants aged ≥6 months and <24 months on the day of interview; ii. no congenital malformations affecting nutrition or growth; and iii. primary residence of mother/primary caregiver in the study site for the previous 6 months. For both surveys, data collection took place using structured questionnaires with precoded responses, administered by trained enumerators with >10 years’ experience in conducting community‐based health, demographic, and nutrition surveys. Interviews were conducted face‐to‐face via household visits (pre‐COVID‐19) and by telephone (during COVID‐19). The data collection team underwent a 2 weeks’ training programme before each of the two surveys. For both surveys, the questionnaire was produced in English and translated to Spanish, and rechecked in both languages by a team member fluent in both languages. For PERUSANO, questionnaires were piloted in Lima (n = 20 interviews) using mothers of IYC living in another similar community (Canto Grande) to the target communities and changes were made accordingly. In the pre‐COVID survey, two enumerators worked as a pair on household visits in Manchay, Lima, and a further two in Huánuco. Two experienced supervisors, one in Manchay and one in Huánuco, checked the completed questionnaires at the end of each week for quality control and to ensure there were no missing data. For STAMINA, new questions were piloted via telephone with 15 caregivers. The telephone interviews during COVID‐19 were conducted by the same team of four enumerators with one additional enumerator. One supervisor from the original pre‐COVID‐19 team checked the completed questionnaires. Hence, the same team was used for both surveys and had close knowledge of the communities from which participants were selected. For the PERUSANO survey, paper questionnaires were completed, and data were entered in Microsoft Access. Automated consistency checks were run and double data entry for a random sample of 10% of questionnaires was used to calculate the error rate. A predefined threshold was set to determine an acceptable level of variation in responses, with the error rate set at 1%. For the STAMINA survey, questionnaires were completed using tablets (Samsung Galaxy Tab‐A) with an electronic data capture (CsPRO) during telephone interviews. Data quality was enhanced in CSPrO through programming of skips, controls, and other logic of the questionnaire. Data collected in both surveys included: household sociodemographic characteristics; infant and young child feeding (IYCF) practices using standardised and validated questions (WHO, 2008; WHO and UNICEF, 2021), and child and maternal qualitative 24‐h dietary recalls using the standardised open recall method (WHO, 2008). These questionnaire modules were the same in both surveys. New questions were added in the STAMINA questionnaire to assess the impact of COVID‐19 on households including changes in employment status, adaptations to finance, sources of financial support, household food insecurity experience as well as access to, and uptake of, well‐child clinics and vaccination health services. We developed questions through engagement with policy‐makers, stakeholders, and expert discussion forums that highlighted gaps in knowledge and concerns surrounding the potential impacts of the pandemic in Peru. We also reviewed emerging literature on the consequences of COVID‐19 on maternal and infant nutrition in LMICs. We measured the experience of household food insecurity using the validated food insecurity experience scale (FIES) survey module (Cafiero et al., 2018). For both surveys, we used the most recent guide to generate IYCF practices for infants aged 0–23 months (WHO and UNICEF, 2021). Breastfeeding indicators (i.e., ever breastfed; early initiation of breastfeeding; exclusive breastfeeding under 6 months; continued breastfeeding 12–23 months) were generated using the retrospective recall of the mother/caregiver. Given that our sample of IYC only included those aged 6–23 months, we were not able to generate the exclusive breastfeeding indicator using 24‐h recall as advised in the IYCF guide. Instead, we asked a series of questions to establish retrospectively whether mothers practised exclusive breastfeeding in the first 6 months or whether foods and beverages were introduced early. Complementary feeding indicators (i.e., introduction of solid, semisolid or soft foods; dietary diversity score; minimum dietary diversity [MDD]; minimum meal frequency [MMF]; minimum acceptable diet [MAD]; egg and/or flesh food consumption; sweet beverage consumption; unhealthy food consumption; and zero vegetable or fruit consumption) were generated using mother/caregiver reported intakes of foods and beverages during the past 24 h. For the mother/caregiver, the reported food and beverage consumption during the past 24 h was used to calculate dietary diversity scores (FAO and FHI 360, 2016). Women who had consumed at least 5 out of the 10 predefined food groups were classified as meeting the adequate MDD for women. We also derived standardised maternal indicators of: egg and/or flesh food consumption, zero fruit and vegetable consumption, sweet beverage consumption, and unhealthy food consumption to match with the IYC indicators. Full definitions of maternal and IYC dietary indicators are provided in Supporting Information Appendix 1. The FIES is an eight‐item experience‐based scale of food insecurity severity. Respondents answered (yes/no) questions on household food‐related behaviours and experiences of limited access to food due to lack of resources in the past 30 days. Statistical techniques borrowed from the toolkit of item response theory (Rasch models) allowed the generation of two prevalence rates comparable across countries: i. moderate or severe food insecurity and ii. severe food insecurity only (Cafiero et al., 2018). The internal reliability of the instrument was good as the modified Rasch reliability test was estimated at 0.80 (Agarwal et al., 2009). We created a common household wealth index for both surveys using factor analysis (i.e., multiple correspondence analysis) applied to proxy indicators of the household environment (ownership of consumer durables; source of drinking water and type of toilet facilities; the number of household members per room used for sleeping; type of materials used for the floors, roof, and walls; and livestock ownership). Analysis of proxy indicators of the household environment revealed no major differences between Lima and Huánuco; hence, we ran the factor analysis with the two settings combined. Variables with low variability (i.e., either less than 5% or more than 95% ownership) were not included in the factor analysis. We split the continuous score of the household wealth index into tertiles of socioeconomic status (SES), with the first tertile representing the relatively poorest households. The first component retained explained 85% of the overall variance. We assessed internal validity by tabulating ownership of durable assets and other housing characteristics by SES tertile. We categorised mothers’ self‐report of completed educational level as less than secondary; secondary/technical, and university level. Other sociodemographic characteristics included maternal working status (yes/no); marital status (married/living together vs. not); child’s age and sex; maternal age; and place of residence (Lima vs. Huánuco). From the survey during COVID‐19, we collated responses to questions on financial impacts of the pandemic on households, adaptations to financial impacts, sources of financial support, as well as access to and uptake of nutrition‐related health services reported by the caregiver. Questions regarding sources of food assistance were asked in both surveys. We generated descriptive statistics (mean, standard deviation or number [n], and percent) for sociodemographic and maternal and infant nutrition‐related factors for each survey. For the STAMINA survey, we described the additional variables on the impact of the pandemic on households and livelihoods, that is, economic impact; coping mechanisms; household food insecurity; and access and uptake of nutrition‐related health services. To examine whether the pandemic influenced dietary outcomes, we performed univariate and multivariable logistic regressions for binary outcomes and linear regressions for continuous outcomes, with the survey type pre‐COVID‐19 (PERUSANO) versus during COVID‐19 (STAMINA) as the main exposure variable. To account for repeated measures (24% of the sample) across the two surveys, we used a mixed‐effects model with a random effect (intercept). Models were adjusted for competing exposures (i.e., those not related to the exposure but the outcomes only) as adjusting for these increases precision in estimates of exposure–outcome associations. These competing exposures included maternal education; maternal working status; wealth index; and place of residence. Models for maternal and child outcomes controlled for maternal age or child age, respectively. We used Stata SE version 16 for statistical analyses. The type I error risk was set at 0.05. Ethical approval for the PERUSANO project was obtained from the Ethical Review Committee of the Instituto de Investigación Nutricional (IIN), Peru (Reference 388‐2019/CIEI‐IIN) and Loughborough University (C19‐87). Written informed consent was provided by all participants after receiving written and verbal information about the study. For the STAMINA project, ethical approval was obtained from IIN (Reference 270‐2020/PROY.367) and Loughborough University (Reference 1926). Verbal consent was provided by mothers or other caregivers following guidelines for research under COVID‐19 of the Ethical Research Committee of IIN. The project was also registered with the National Institute of Health under the requirement for all studies related to COVID‐19 (PRISA Reference EI00000001577). In both surveys, participants were informed of the right to withdraw from participation.