A self‐applied valid scale for rapid tracking of household food insecurity among pregnant women in Sri Lanka

Abstract Rapid household food insecurity (HFI) tracking has been identified as a priority in the context of the COVID‐19 pandemic and its aftermath. We report the validation of the Latin American and Caribbean Food Security Scale (Escala Latinoamericana y Caribena de Seguridad Alimentaria [ELCSA]) among pregnant women in Sri Lanka. The eight‐item adult version of the ELCSA was translated from English to Sinhala and Tamil. Cognitive testing (on 10 pregnant women and five local experts) and psychometric validation of the self‐administered HFI tool were conducted among pregnant women (n = 269) attending the special clinics of the Rajarata Pregnancy Cohort (RaPCo) in Anuradhapura in February 2020. We assessed the psychometric properties and fit using a one parameter logistic model (Rasch model analysis) using STATA Version 14 and WINSTEP software Version 4.3.4. Concurrent validity was tested using psychological distress. The scale was internally consistent (Cronbach's alpha = 0.79) and had a good model fit (Rasch items infit statistic range: 0.85 to 1.07). Item 8 (‘did not eat for the whole day’) was removed from the model fit analysis, as it was not affirmed by respondent. Item severity scores ranged from −2.15 for ‘not eating a diverse diet’ to 4.43 for ‘not eating during the whole day’. Concurrent validity between HFI and psychological distress was confirmed (r = 0.15, p < 0.05). The self‐applied version of ELCSA‐pregnancy in Sri Lanka (ELCSA‐P‐SL) is a valid and feasible valid tool. We recommend it to track HFI among pregnant women in lower income countries during the COVID‐19 pandemic.


eight-item adult version of the ELCSA was translated from English to Sinhala and
Tamil. Cognitive testing (on 10 pregnant women and five local experts) and psychometric validation of the self-administered HFI tool were conducted among pregnant women (n = 269) attending the special clinics of the Rajarata Pregnancy Cohort (RaPCo) in Anuradhapura in February 2020. We assessed the psychometric properties and fit using a one parameter logistic model (Rasch model analysis) using STATA Version 14 and WINSTEP software Version 4.3.4. Concurrent validity was tested using psychological distress. The scale was internally consistent (Cronbach's alpha = 0.79) and had a good model fit (Rasch items infit statistic range: 0.85 to 1.07).
Item 8 ('did not eat for the whole day') was removed from the model fit analysis, as it was not affirmed by respondent. Item severity scores ranged from −2.15 for 'not eating a diverse diet' to 4.43 for 'not eating during the whole day'. Concurrent validity between HFI and psychological distress was confirmed (r = 0.15, p < 0.05). The selfapplied version of ELCSA-pregnancy in Sri Lanka (ELCSA-P-SL) is a valid and feasible valid tool. We recommend it to track HFI among pregnant women in lower income countries during the COVID-19 pandemic. authorities have advised countries to be alert and prepared for a COVID-19-undernutrition syndemic strongly rooted in social determinants of health (SDH) inequities (Cornia et al., 2020). In 2005, the World Health Organization Commission for Social Determinants of Health was established to understand and act on inequities in health (Marmot, 2005). One of the three recommendations of the commission to close the gap in a generation was 'to measure and understand the problem and assess the impact of action' on inequities in SDH (Report of the Commission on Social Determinants of Health & The World Health Organization, 2008). Given that 'closing the gap in a generation' would be a challenge with the current COVID-19 pandemic, counties across the globe need to track HFI frequently as it is a crucial SDH. This information is necessary for equitable resource allocations through effective food and nutrition security actions.
HFI has been measured indirectly through indicators such as anthropometry, dietary intake surveys and household food expenditure. Experience-based HFI scales directly capture the experience of a household food situation, by selecting a respondent who is cognizant of the food situation in the family. The lack of access to a healthy and nutritious diet as a result of poverty, social deprivation or situations such as natural or man-made disasters is well captured through experience-based HFI scales (Ballard et al., 2013). Research in Mexico has shown that ELCSA can be applied by an interviewer over the phone to reliably and efficiently track HFI during the COVID-19 pandemic (Gaitán-Rossi et al., 2021). However, as far as we know, experience-based HFI scales have not been validated in a self-administered form in LMICs, which is a major gap in the context of pandemics or other public health emergencies, such as COVID-19, requiring social distancing measures.
The objective of this study was to test the validity of a selfapplied culturally adapted eight-item version of ELCSA in a large population-based cohort study of pregnant women in Sri Lanka (Agampodi et al., 2020) in the context of the COVID-19 pandemic.

| METHODS
We conducted cognitive and psychometric validation of the selfadministered ELCSA for pregnant women in Sri Lanka (ELCSA-P-SL).
The study population included pregnant women in their third trimester (32-36 weeks of gestation) of pregnancy enrolled in the Rajarata Pregnancy cohort (RaPCo) in Anuradhapura district, Sri Lanka. Ethical

Key messages
• Rapid household food insecurity (HFI) tracking has been identified as a priority in the context of the COVID-19 pandemic and its aftermath.
• Cognitive testing and psychometric validation of the selfadministered HFI experience-based ELCSA scale were conducted among pregnant women in Sri Lanka.
• The self-administered ELCSA pregnancy was valid and feasible to efficiently track HFI during pregnancy in Sri Lanka.
• We recommend ELCSA pregnancy to track HFI among pregnant women in lower income countries. clearance was obtained from the Ethics Review Committee of Faculty of Medicine and Allied Sciences, Rajarata University of Sri Lanka. The study was conceived in December 2019 and was carried out in February 2020, just before Sri Lanka started to be affected by the COVID-19 lockdown. Following the initial assessment, the pregnant women reporting severe HFI were recruited for a financial assistance donor programme voluntarily mediated by the institution. data showed that antenatal care coverage through the public health system is 100%. Of the females in the district, 90% have at least entered secondary level education. The ongoing prospective cohort study in Sri Lanka (RaPCo study) provided a unique opportunity to validate a self-reported measure of HFI due to its relatively high maternal literacy rates, universal maternal and child health service coverage and focus in equity for women.

| Cognitive validation
The 8 adult items of the English version of ELCSA was culturally adapted to the Sri Lankan context and translated into Sinhala and Tamil following a modified version of Sumathipala and Murray's method (Sumathipala & Murray, 2000). A panel comprised of eight panelists, who are fluent in both languages, translated and then assessed the adapted questionnaire. Each panellist was provided with the original questionnaire. The concept of food security including its dimensions, and the objectives of the study were explained by an investigator. In the first round, panellists recorded their own translations. Each of the original English version and its corresponding eight translations were discussed one at a time until consensus was reached. The items were further refined in two subsequent rounds.
Cognitive validation was performed through expert opinion with five local experts with diverse experiences in the fields of nutrition, scale development methodology, maternal health, SDH and public health. Necessary changes were made to incorporate the comments of the experts with the consensus of the investigators. Target group interviews were held with 10 pregnant women using the methods suggested by Bowden, Fox-Rushby, Nyandieka, and Wanjau (2002) for cognitive validation. A three-step process was followed. First, the investigators observed whether the participant could easily understand and provide an answer for each item. Then, the cognitive process of the participant was examined to find out whether the participants thinking matched the intended meaning of each scale's item. Then, the participants were asked whether they had alternative wording suggestions for each item. We ensured that the original meaning was preserved and that the wording was clear, simple and self-explanatory to facilitate its self-administration.

| Psychometric validation
The cognitively validated ELCSA-P-SL was self-administered by a con-

| Changes made in cognitive validation
Experts and pregnant women agreed that the questions were clear to them and that the intended meaning of each item was preserved.
Based on the consensus from the experts, the introduction of the questionnaire was adjusted to include lay language, which could be easily understood by rural women. It was clearly mentioned that the lack of access to food was being asked in the context of lack of socioeconomic resources and that it did not refer to food restrictions due to dieting, loss of appetite or pregnancy-related symptoms such as nausea, vomiting or heart burn. To maintain the respondents' focus on the time period, the phrase 'during the past three months' was added at the beginning of each question. To ensure that the exact intended reasons of food insecurity were taken into account, the phrases such as 'due to the above mentioned reasons' or 'due to less availability of food in your house' were used in each and every scale question.
Using these phrases repeatedly was designed to enable participants to remain focused on accurately answering the ELCSA-P-SL selfadministered questions.

| Characteristics of the study sample
Of the 328 pregnant women attended the RaPCo special clinics during this period, 269 responded to the questionnaire; hence, the response rate was 82%. The sample characteristics with regard to ethnicity and the level of education were representative of the study population (Table 1).

| Frequency distribution of responses
Inability to have a diversity of food due to lack of money or other resources was the item more frequently affirmed. There were no pregnant women in this sample who affirmed Item 8; that is, not eating throughout the day ( Table 2). The percentage of missing data ranged from 0.4% to 3.3% in each item. The missing data were replaced by the answer 'no', as it did not change the order of severity by doing so and it was predictable based on the pattern of responses in rest of items. As an example, a mother who responded 'no' to Item 7 (experience of hunger for one meal) would invariably not left with hunger during the whole day which indicates a 'no' answer for Item 8. Further assessment of model fit without replacing the missing data indicated no significant change in severity scores and infit values.

| Psychometric properties
The tool was internally consistent (Cronbach's alpha = 0.79). The one parameter logistic model indicated that dietary diversity was the first factor to be compromised and that experience of hunger represented the most severe stage of HFI ( Figure 1). The item infit statistics indicated good model fit (range: 0.85 to 1.07). Item 8 (did not eat for the whole day) was removed from the model fit analysis as none of the pregnant women affirmed this item. Overall, the item severity scores ranged from −2.15 for diversity food to 4.43 for hunger during the whole day. The item severity in Rasch model is based on the assumption that the log-odds of a household affirming an item is proportional to the difference between the "true" T A B L E 2 Frequency distribution of responses to the ElCSA-Prgenancy Sri Lank scale (ELCSA-P_SL, N = 269)

Item
Answered "yes" Percentage "yes" Answered "no" Percentage "no" Missing data severity of food security in the household and the "true" severity level represented by the item (Nord, 2014). The range of item severity was 6.28, and the ordering of severity (Figure 2) was as expected with the food diversity being the least severe and the item about not eating during the whole day the most severe. We observed a week positive (r = 0.15, p < 0.05) correlation between HFI and distress scores (Allen et al., 2018;Becerra et al., 2015) confirming the concurrent validity of the tool (Figure 3).

| DISCUSSION
This study reports the validation of the eight-item self-administered ELCSA-P-SL (Data S1 to the best of our knowledge, this is the first time that ELCSA was validated as a self-administered questionnaire among pregnant adult women in a LMIC.
The severity scores of ELCSA-P-SL behaved as expected indicating that dietary diversity is compromised first and that food intake reduction or hunger is the most severe manifestation of HFI. Poor diet diversity has indeed been identified as being prevalent in Sri Lanka (Aguayo, 2017;Sirasa et al., 2020). In contrast to other contexts, being worried or having fear about running out of food is the item that has been more frequently affirmed (Pérez-Escamilla, 2017; Pérez-Escamila et al., 2017). Identification of possible underlying reasons for this observed inter-country difference in items' ranking order requires further research.
This study is likely to have minimized social desirability bias as the HFI questionnaire was self-administered (Food and Agricultural Organization, 2016). Likewise, pregnant women who may have felt shy or uncomfortable reporting deficiencies in household food availability in a face-to-face interview might have reported more accurate details in this self-administered scale.
One important advantage of ELCSA is that it not only captures the access to food but also reflects the underlying pain and anxiety F I G U R E 2 Item characteristic and information functions of ELCSA-P-SLa (N = 269 pregnant women in Sri Lanka) F I G U R E 1 Scatter plot indicating a positive correlation between HFI and psychological distress (N = 269 pregnant women in Sri Lanka) experienced by the participants due to food insecurity and hunger.