Associations between breastfeeding intention, breastfeeding practices and post‐natal depression during the COVID‐19 pandemic: A multi‐country cross‐sectional study

Abstract Associations between breastfeeding intention, duration and post‐natal depression (PND) have been shown in pre‐COVID‐19 studies. However, studies during COVID‐19 have not examined the associations between breastfeeding intention, breastfeeding practices, and PND in an international sample of post‐natal women, taking into consideration COVID‐19 related factors. This is the first study to address this gap as both PND and breastfeeding may be affected by COVID‐19, and have important long‐term effects on women's and infant's health. A cross‐sectional internet‐based survey was conducted with 3253 post‐natal women from five countries: Brazil, South Korea, Taiwan, Thailand, and the United Kingdom from July to November 2021. The results showed that women who intended to breastfeed during pregnancy had lower odds of having PND than women who did not intend to. Women who had no breastfeeding intention but actually breastfed had greater odds (AOR 1.75) of having PND than women who intended to breastfeed and actually breastfed. While there was no statistical significance in expressed breast milk feeding in multivariable logistic regression models, women who had shorter duration of breastfeeding directly on breast than they planned had greater odds (AOR 1.58) of having PND than those who breastfed longer than they planned even after adjusting for covariates including COVID‐19‐related variables. These findings suggested the importance of working with women on their breastfeeding intention. Tailored support is required to ensure women's breastfeeding needs are met and at the same time care for maternal mental health during and beyond the pandemic.

With the advent of the COVID-19 pandemic, infection preventative measures such as self-isolation, social distancing, facemask wearing and restricted hospital visits from partners or relatives have been put in place in hospitals and 'hotspot' areas to control the transmission of the virus in many countries. Studies undertaken during the first year of COVID-19 pandemic have reported high rates of PND.
For example, a survey conducted with 614 post-natal women between April 2020 and May 2020 in the UK found that 43% of women had an Edinburgh post-natal Depression Scale (EPDS) score ≥13, indicating a major post-natal depressive disorder (Fallon et al., 2021). About 35% of 162 post-natal women in London between May 2020 and June 2020 were assessed to have EPDS ≥ 13 (Myers & Emmott, 2021). A survey of 184 post-natal women from two hospitals in Brazil between 8th June 2020 and 23rd December 2020 reported a 38.8% PND rate (EPDS ≥ 12; Galletta et al., 2022). An Italian study presented that 23.03% of 152 post-natal women in a hospital who filled in the EPDS questionnaire on the second post-natal day at hospital discharge between 22nd February 2020 and 18th May 2020 had an EPDS score ≥ 12 compared to 11.56% of 147 women from the nonconcurrent control group in 2019 (p < 0.001; Zanardo et al., 2021).
These findings are concerning due to the negative and long-term impact of PND on the woman, her infants and her family (Myers & Johns, 2018;Slomian et al., 2019;Tammentie et al., 2004).
The short-and long-term benefits of breastfeeding to the health of women and infants have been well documented (Victora et al., 2016).
Breastfeeding has also been found to provide protection against infectious respiratory diseases such as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) which caused COVID-19 (Didikoglu et al., 2021;Verd et al., 2021). For women diagnosed or suspected of COVID-19, breastfeeding remains the recommended infant feeding practice but with precautions, such as wearing a facemask during feeding and good hand hygiene (WHO, 2020). A narrative review that included 12 studies between January 2020 and January 2021 reported that COVID-19 had both positive and negative impacts on women's breastfeeding plans (Pacheco et al., 2021). The positive impact included that some women were able to spend more time at home with their infants and increased breastfeeding duration. The negative impact included reduced breastfeeding duration and frequency, and earlier breastfeeding cessation due to increased childcare responsibilities at home and perceived lack of support from family and professionals (Pacheco et al., 2021).
Breastfeeding could be a protective factor for PND and reduce post-natal depressive symptoms (Figueiredo et al., 2014). However, unmet breastfeeding expectations may increase the risk of PND in some women (Gregory et al., 2015). Borra et al. (2015) reported that among women who were not depressed before childbirth, those who had intended to breastfeed and actually breastfed had the lowest risk of PND, while those who had intended to breastfeed but did not breastfeed had the highest risk of PND. On the other hand, PND has been indicative of early breastfeeding cessation (Brown et al., 2016;Dias & Figueiredo, 2015).
Limited studies conducted during the COVID-19 pandemic have investigated the relationships between PND and breastfeeding. Zanardo et al. (2021) study during the COVID-19 lockdown in an Italian hospital showed that women who breastfed their infants exclusively at hospital discharge on the second post-natal day had significantly lower EPDS scores compared to women who practised formula feeding and complementary feeding. In a study amongst Malaysian post-natal women with premature infants at the beginning of the pandemic, Yahya et al. (2021) found women who were of a high risk of PND had less positive attitudes towards breastfeeding compared to those who had a low risk of PND.
However, none of the studies during COVID-19 explored the relationships between (un)changed breastfeeding plans and PND, and on an international level. Our study is the first that addresses this gap.

Key messages
• This study identified independent associations of breastfeeding intention and actual breastfeeding practices with PND after adjustment for COVID-19-related covariates.
• Although women who intended to breastfeed during pregnancy were less likely to have PND, those who did not intend to breastfeed but did breastfeed were more likely to have PND than those who intended to breastfeed and breastfed.
• Women with shorter (vs. longer) than planned duration of breastfeeding on breast were more likely to have PND.
• Health care providers and policymakers should ensure tailored support is available for women's infant feeding plans and mental health.
The aim of this article was to report the associations between breastfeeding intention, breastfeeding practices, and PND considering COVID-19-related factors among post-natal women in five countries. This article formed part of a larger multi-country project examining various aspects (including infant feeding, PND, social support, maternity care, COVID-19 infection and COVID-19 vaccination acceptance) of post-natal women's experiences of having a baby during the COVID-19 pandemic.

| Study design and participants
A cross-sectional online survey was conducted in five countries: Brazil, South Korea, Taiwan, Thailand, and the UK from July 2021 to November 2021. The choice of countries was a convenience sample of countries that had similar PND rates pre-COVID (around 21%), except Thailand (12.52%; Wang et al., 2021). A convenience sampling technique was used to recruit participants. Survey participants' inclusion criteria were post-natal women who were: (a) up to 6 months postpartum (b) living in one of the participating countries, (c) aged 18-49 years old (except in Taiwan 20-49 years old), and (d) literate in the residential country's official language. The recruitment information and the survey web link were advertised online and/or through hard copies of posters or flyers, with a quick response (QR) code where used, as planned by the study lead of each country. The researchers from each country distributed the survey information in the official language of the country (i.e., Portuguese, South Korean, Mandarin with traditional Chinese characters, Thai, and English) via various channels such as emails, social media (e.g., Twitter, Facebook, WhatsApp groups, Line groups, etc.), parenting online forums, personal networks, relevant health care services, and notfor-profit organisations, including services supporting breastfeeding, women, children and families.

| Data collection
Data were collected anonymously via an online Google Form in each country's official language. Except for the outcome variable mentioned below, the survey questions were initially developed in English and subsequently translated into the official language of the participating country. Back translations were also undertaken. Online informed consent was obtained from all participants before they started the survey. All data were anonymised. Ethical approval was granted from each country's relevant ethical approval body (detailed in Section 2.4).

| Outcome variable
The outcome variable in this study is depressive symptoms. The EPDS, a 10-item self-report scale, was used to assess post-natal women's mental health in the last 7 days, as stated in the EPDS, using Likert scales (scoring 0-3, with a total score ranging between 0 and 30; Cox et al., 1987;Khalifa et al., 2016). An EPDS cutoff point of 13 or above was used to classify those with depression. Validated EPDS versions in each country's official language were used.

| Independent variables
The independent variables were: (1) intention to breastfeed during pregnancy, (2) breastfeeding intention during pregnancy and actual breastfeeding practices during postpartum, (3) impact of COVID-19 on baby fed directly from breast, and (4) Impact of COVID-19 on feeding expressed breast milk.
1. Intention to breastfeed during pregnancy was asked with response option ('yes' or 'no' or 'don't know') and was categorised into ('yes' or 'no/don't know') in the analyses. 3. Impact of COVID-19 on baby fed directly on breast was asked using the question 'Does COVID-19 affect your infant feeding behaviour?' with response options (i) not intend to feed, (ii) shorter than intended, (iii) the same duration as intended, and (iv) longer than intended.
4. Impact of COVID-19 on baby fed expressed breast milk was asked using the question 'Does COVID-19 affect your infant feeding behaviour?' with response options (i) not intend to feed, (ii) shorter than intended, (iii) the same duration as intended, and (iv) longer than intended.

| Covariates
Socio-demographic, obstetric, health and support characteristics Socio-demographic variables of women included country, mother's age, education level, work status, education level, residence and marital status. Obstetric variables were pregnancy intention, mode of childbirth, parity, birthweight and preterm birth. Health and support variables were (a) health problem of mother during pregnancy, (d) social support post-birth (scores)measured using a six-item self-report scale-Maternity Social Support Scale (MSSS;Webster et al., 2000).

COVID-19 knowledge, attitudes and practices (KAP), and beliefs of breastfeeding in relation to COVID-19
Questions regarding KAP on infection prevention and control measures against COVID-19 were drawn and modified from previous studies (Hussain et al., 2020;Islam et al., 2020). Nine questions were included to assess knowledge of COVID-19 by answering if a statement (e.g., 'COVID-19 can NOT spread through respiratory droplets of infected individuals') was 'true', 'false' or 'do not know'.
A correct answer was awarded a score indicating 'adequate answer'.
A total score ranged from 0 to 9, with a higher score indicating better knowledge. Attitudes toward the severity and prevention of COVID-19 had seven questions (e.g., 'Social distancing is important to prevent COVID-19') with 5-point Likert scales ('Strongly disagree', 'disagree', 'undecided', 'agree' and 'strongly agree'. A total score ranged from 7 to 35. A higher score indicated a more positive attitude. Precaution practices of COVID-19 contain six questions including questions such as 'During the last 7 days, did you avoid touching eyes, nose and mouth with unwashed hand?' with 4-point Likert scales ('never', 'occasionally', 'sometimes' and 'always'). A total score ranged from 6 to 24, with higher scores indicating a more adequate practice.
Six questions (e.g., If the mother is confirmed or suspected to have COVID-19, the mother should not breastfeed) were developed for breastfeeding beliefs in relation to infection prevention and control measures for COVID-19 based on WHO's (2020) breastfeeding Q&A. The answer options were 'disagree', 'uncertain' and 'agree'.
A total score ranges from 0 to 12, with higher score showing a more positive breastfeeding belief.

COVID-19 impact variables
Impact of COVID-19 on food security was assessed before and during COVID-19 using two questions: 'Did you ever run out of food before the end of the month or cut down on the amount you ate to feed others in 2019 BEFORE COVID-19?' and 'Did you ever run out of food before the end of the month or cut down on the amount you ate to feed others DURING COVID-19 in 2020-2021?' The two variables were combined and categorised into (i) no change: insecure to insecure, (ii) worse: secure to insecure, (iii) better: insecure to secure and (iv) no change: secure to secure. Variables regarding COVID-19 positive diagnosis (yes or no), and COVID-19 vaccination uptake (yes or no) were also included.

| Statistical analysis
The data analyses were conducted using SAS statistical software package version 9.3 (SAS Institute Inc.). Descriptive statistics were used to analyse categorical variables (frequencies and percentages) and continuous variables (mean and standard deviation [SD]). To test the association between PND (outcome variable) and the four independent variables: (1) intention to breastfeeding during pregnancy, (2) breastfeeding intention during pregnancy and actual breastfeeding practices, 3) impact of COVID-19 on baby fed directly from breast, and (4) Impact of COVID-19 on feeding expressed breast milk, bivariate associations were assessed using chi-square test and simple logistic regression. Subsequently, to assess the impact of each independent variable on PND, adjusted for other variables, three multivariable logistic regression models were used: (i) model I adjusted for socio-demographic, obstetric, health and support characteristics; (ii) model II adjusted for the covariates in model I and COVID-19 related KAP and belief towards breastfeeding during COVID-19; and (iii) model III adjusted for the covariates in model II and COVID-19 impact variables: food security status before and during COVID-19, positive COVID-19 diagnose, COVID-19 vaccination uptake. Adjusted odds ratio (AOR) with a 95% confidence interval (CI) was used to assess the strength of these associations.

| Ethical statement
Online informed consent was obtained from all participants before they started the survey. All data were anonymised. Ethical

| RESULTS
A total of 3253 eligible responses were received from post-natal women living in five countries: Brazil, South Korea, Taiwan, Thailand and the UK. Table 1 shows the characteristics of the participants. The majority of all women were between 30 and 39 years old (61.6%), had a university or higher degree (75.8%), lived in urban areas (72.6%), were married (95.5%), were on either paid or unpaid maternity leave (59.3%), had vaginal birth (61.0%), were primiparous (57.0%), and had received at least one dose of COVID-19 vaccine (72.2%). About 74% (73.5%) of all women fed their baby directly on breast, 40.6% fed their baby with infant formula, and 38.3% with expressed breast milk in the past 24 h of survey completion (Table 1).  did not intend to and actually breastfed, 2.2% intended but did not breastfeed and 1.7% did not intend and did not breastfeed. The chi-square test revealed that there was a significant association with PND in the pooled data (p < 0.0001). At the country level, a statistical significance was only seen in Thailand (p < 0.0001).
Women were asked 'Does COVID-19 affect your infant feeding behaviour?' with two types of breastfeeding: directly on breast and expressed breast milk. About 54% (53.7%) of all women responded to breastfeeding directly on breast for the same duration as planned, while 18.6% shorter than planned, and 12.7% longer than planned.
The bivariate analysis showed a significant association between the impact of COVID-19 on breastfeeding directly on breast and PND in the pooled model (p < 0.001) and by country in Taiwan, Thailand and the UK (p < 0.01). In terms of expressed breast milk feeding, 40.8% fed their baby expressed breast milk the same duration as planned, 20.1% shorter than planned and 11.5% longer than planned. The bivariate analysis revealed a significant association with PND in the pooled model (p < 0.0001) and in almost all countries (p < 0.05), except South Korea. Longer than intended 1.00 -1.00 -1.00 -1.00 -Note: Model: 1 included covariates of country (Brazil, South Korea, Taiwan, Thailand or the UK), maternal age (18-29, 30-39 or 40-49), intended pregnancy (yes or no), mode of childbirth (vaginal or caesarean section), health problem of mother during pregnancy, delivery or postpartum (yes or no), work status (employed, on paid maternity leave, on unpaid maternity leave or housewife/unemployed), residence (urban or rural), marital status (married or others), parity (1 or 2+), birthweight (<2.5, 2.5-3.5 or >3.5 kg), preterm birth (yes or no), breastfeeding directly on breast in the last 24 h (yes or no), expressed breast milk in the last 24 h (yes or no), infant formula in the last 24 h (yes or no), solid, semi-solid or soft foods in the last 24 h (yes or no), social support (score), number of post-natal care (never, 1-2, 3 or 4+) Model 2: covariates in Model 1+ COVID-19 knowledge (score), attitudes (score), and practices (score) and breastfeeding belief (score) Model 3: covariates in Model 2+ changes in food insecurity before and during covid-19 (no change (insecure), worse, better, or no change (secure)), ever diagnosed as COVID-19 positive (yes or no),  and belief towards breastfeeding in relation to COVID-19 in model 2 and COVID-19 impact variables in model 3, the associations became a bit greater (1.59, 1.15-2.19 and 1.58, 1.15-2.18, respectively).

| DISCUSSION
To the best of our knowledge, this is the first international study investigating the associations between breastfeeding intention, breastfeeding practices and PND during the COVID-19 pandemic.
We focus the discussion on the associations of the outcome variable and independent variables stated in this article. Discussions specifically on (a) breastfeeding and (b) PND were presented in separate articles as separate topics Coca et al., 2022). Some key pooled results from the five countries in this study were that those who intended to breastfeed during pregnancy had lower odds of having PND (p < 0.0001) while those who had no breastfeeding intention but actually breastfed (p < 0.0001), and ceased breastfeeding directly on breast earlier than planned had higher odds of having PND (p < 0.0001), and to cease breastfeeding (directly on breast and expressed breast milk) earlier than they planned (p < 0.0001) compared to those with no PND.
Similar findings were shown in some pre-pandemic findings. For example, Gregory et al. (2015) found that women who met their breastfeeding expectation had lower odds of post-natal depressive symptoms compared to those who did not. However, our study did not establish causal relationships. For example, we do not know if women already had depressive symptoms before childbirth and then developed PND, which may affect their breastfeeding decisions; or women's decision to stop breastfeeding earlier than planned due to the impact of COVID-19 may trigger their PND, although previous pre-pandemic studies have reported bidirectional relationships between breastfeeding and PND (Dias & Figueiredo, 2015;Pope & Mazmanian, 2016). Nevertheless, our findings demonstrate the importance of supporting women to achieve their breastfeeding plans as well as when their infant feeding plans change while at the same time caring for their mental health.
We further analysed the association between the type of changes in breastfeeding plans and PND. We showed in pooled models statistically significant results before and after adjusting for confounders, including COVID-19-related factors, that women with no intention to breastfeed but actually breastfed had greater odds of having PND than women who intended to breastfeed and breastfed.
This showed the importance of understanding women's breastfeeding intention, working with women on their breastfeeding intention, and their subsequent breastfeeding practices which may be different from their intention. Borra et al. (2015) reported a similar finding that among women who did not have depressive symptoms during pregnancy, breastfeeding increased the risk of PND in women who had not intended to breastfeed and decreased the risk of PND in women who had intended to breastfeed. Other statistically significant results in our pooled models before and after adjusting for confounders were that women who breastfed directly on breast for shorter duration than they planned had greater odds of having PND than those who breastfed longer than they planned. Many prepandemic studies have shown that women who had PND were more likely to have a shorter breastfeeding duration than those who did not have PND (Butler et al., 2021;Dias & Figueiredo, 2015;Pope & Mazmanian, 2016). On the other hand, Costantini et al. (2021) conducted an online survey in the UK during the COVID-19 lockdown with women whose children were aged 0-3 years old and found no statistically significant difference in post-natal depressive scores as measured by Patient Health Questionnaire (PHQ-9) between women who breastfed more than 6 months and those who breastfed less than 6 months. However, Costantini et al. (2021) did not investigate if women changed their breastfeeding plans.
As mentioned above, the statistically significant associations PND. An online survey of UK post-natal women over 4 weeks in May 2020 and June 2020 found that women ceased breastfeeding due to COVID-19-related concerns, such as lockdown and lack of support (Brown & Shenker, 2021). Piankusol et al. (2021) survey conducted between 17th July 2020 and 17th October 2020 reported that lack of family support with infant feeding was the risk factor associated with changing breastfeeding practices (e.g., reduced breastfeeding fre- conducted to understand the specific aspects of COVID-19 that impacted on women's breastfeeding plans and mental health outcomes, especially when 'living with COVID' becomes a norm. The importance of receiving support from health care professionals, partners, family members, and friends for breastfeeding and PND has consistently been shown in studies before and during COVID-19 (da Silva Tanganhito et al., 2020;Myers & Emmott, 2021;Pacheco et al., 2021). The statistically significant associations found in our study between breastfeeding intention, breastfeeding practices and PND further illustrate the importance of providing effective breastfeeding support from health care professionals, partners and families to tailor infant feeding support to women's infant feeding needs and decisions, and to minimise the risk of PND. A Canadian prospective pre-COVID study reported that women who experienced breastfeeding challenges scored lower EPDS scores when they did not report a negative experience with breastfeeding support (Chaput et al., 2016). This highlighted the need of positive and high-quality breastfeeding support for post-natal women to reduce the risk of PND (Chaput et al., 2016). There are few interventions to support women at risk of PND with breastfeeding. Reach Out, Stand Strong, Essentials for new mothers (ROSE) study which utilised a group interpersonal therapy to promote social support and self-care in lowincome pregnant women at risk of PND has indicated a positive outcome of increased breastfeeding duration among women receiving the therapy, although further evidence is needed (Kao et al., 2015). need support to be able to support women (Chang et al., 2021).
Breastfeeding peer supporters were found to be beneficial in not only providing practical support for breastfeeding but also emotional/ psychological support, as well as decreasing social isolation , which was reported as a challenge for post-natal women during the pandemic due to restrictive measures and lack of support (Ipsos MoRI, 2021). Some peer support, including face-to-face in-person interactions, continued to be provided during the COVID-19 lockdown (Hann et al., 2021). Evaluating and learning from these support services may help inform the development of improved breastfeeding and mental health support with integration into health services for women and their families.

| Limitations
This study has several limitations. Due to the use of convenience sampling, generalisability may not be appropriate to all post-natal women in the participating countries and other countries. Other limitations of this study include the nature of assessing breastfeeding and PND as we used self-report assessments with retrospective data, which increase the risk of eliciting social desirability bias and may lead to recall bias. We used the EPDS scale with a cutoff point of 13, in which previous research has shown that identifying PND using a tool, such as EPDS, may be insufficient to recognise women who need support (Fellmeth et al., 2019). Further, we did not ask if women had antenatal depression or previous mental illness or received support for previous mental illness. Despite the limitations of this study, this study has certain strengths. Our study is the first that addresses the relationships between (un)changed breastfeeding plans and PND.
Furthermore, being an international study across five countries on this important topic during COVID-19 is a strength and the results can also inform practices and policies for future pandemics.

| CONCLUSIONS
Our study highlighted that breastfeeding intention at pregnancy and change of breastfeeding plans were associated with PND during the COVID-19 pandemic. Further investigation is needed to identify effective breastfeeding interventions in preventing PND and reducing post-natal depressive symptoms, combining 'living with COVID' desired preventative measures. Working with women on their breastfeeding intention during pregnancy is important. post-natal care should include supporting women's breastfeeding decisionmaking and identify breastfeeding/infant feeding support needs for women more likely to be at risk of PND. It is also essential that policymakers and health care providers provide guidance and take actions on mitigating the long-term effects of unmet breastfeeding plans and PND, and preventing/reducing occurrences of PND and post-natal depressive symptoms and negative breastfeeding outcomes for women during the pandemic and beyond.