Peripartum depression and infant care, sleep and growth

Peripartum depression is a common, serious complication in mothers. To assess the influence of infant care, sleep and growth on the risk of peripartum depression, 1,271 mothers of healthy one-month-old infants completed comprehensive questionnaires including the Edinburgh Postnatal Depression Scale. Relationships between high depression scores and variables related to infants’ care, sleep and growth were assessed adjusting for other variables. High depression scores were found in 233 mothers, which were associated with variables related to infants’ care (poor satisfaction with infant care support, p < 0.001; mothers’ passive response to night-time fussing, p = 0.001), sleep (dim bedroom lighting, p < 0.05; short sleep duration, p < 0.05) and growth (poor weight gain, p < 0.05), as well as maternal variables (primiparas, p < 0.001; poor income satisfaction, p < 0.001; poor sleep satisfaction, p < 0.001; daytime sleepiness, p = 0.001). Considering the observed association between high depression scores and infants’ care, sleep and growth, a multidisciplinary approach accounting for infant care would be required to prevent peripartum depression.

whereas very high EPDS scores (≥12) were found in 115 (9.0%) mothers. The infants were 39.5 ± 1.2 weeks gestation and 3091 ± 374 g at birth, whose body weight gain was 44.1 ± 10.4 g/day. The night-time and daily sleep durations of the infants were 8.3 ± 1.8 h and 15.8 ± 2.8 h, respectively.
Findings from univariate analysis: variables associated with infant care, sleep and growth ( Table 2). Elevated EPDS scores were associated with smaller body weight gain of infants (p < 0.001), dim lighting of infant's bedroom (p = 0.031 for 'dim' compared with 'bright'), irregular light-off time (p = 0.002 for 'irregular' compared with 'regularly early'), shorter daytime sleep duration (p = 0.009 for 'short' and 0.048 for 'medium'; compared with 'long'), shorter night-time sleep duration (p < 0.001 for 'short' and p = 0.041 for 'medium'; compared with 'long'), and shorter daily sleep duration (p < 0.001 for both 'short' and 'medium' compared with 'long'). For variables associated with infant care support and child-rearing style, the presence of an additional caregivers (p = 0.016 for 'one' compared with 'no' extra caregiver), poor satisfaction with social infant care support (p = 0.022 for 'generally satisfied' and p < 0.001 for 'not satisfied'; compared with 'no support'), formula-milk-based feeding (p = 0.004 for 'mainly formula milk' compared with 'exclusive breastfeeding'), passive response of mothers to night-time fussing (p < 0.001 for 'observe' without dandling compared with 'rock briefly'), and poor confidence with child rearing (p < 0.001 for mothers of both 'anxious' and 'generally confident'; compared with 'confident') were associated with high EPDS scores.
Findings from univariate analysis: maternal variables. Higher maternal age (p = 0.032 for ≥35 years old compared with 29≤ and <35 years old) and dissatisfaction with family income (p < 0.001 for 'not satisfied' and p = 0.003 for 'generally satisfied'; compared with 'fully satisfied'), primipara status (p < 0.001 for primiparas compared with both parity 2 and ≥3) and vaginal delivery (p = 0.030 for vaginal delivery compared with elective caesarean delivery) were associated with high EPDS scores (Table 3). Of variables associated with maternal www.nature.com/scientificreports www.nature.com/scientificreports/ sleep status during the last 7 days of the baby check, irregular bedtime (p < 0.001 for 'irregular' and p = 0.027 for 'regularly late'; compared with 'regularly early'), shorter night-time sleep duration (p < 0.001 for 'short' and p = 0.043 for 'medium'; compared with 'long'), more frequent night-time wakefulness (p < 0.001), poor satisfaction with sleep (p < 0.001 for both 'dissatisfied' and 'mildly satisfied'; compared with 'satisfied'), and greater Epworth Sleepiness Scale (ESS) scores (p < 0.001) were associated with high EPDS scores (See Supplementary  Table 1 for the variables in relation to maternal sleep during the last month of pregnancy).

Discussion
Using a comprehensive dataset from a survey of mothers of 1-month-old infants, infant care, sleep and growth were identified as potential independent variables of PD even after adjustment for the influence of known independent variables. These findings suggest that, for the prediction, prevention, early diagnosis and efficient treatment of PD, multifactorial approach would be required focusing on both maternal and non-maternal variables, especially on those related to the care, sleep and growth of infants.
peripartum depression and sleep of mothers and infants. Infants' sleep duration start aggregating in the night-time approximately one month after birth; however, the timing and progress vary between infants 11 . A delay in this process may increase the anxiety of mothers, rendering them susceptible to psychiatric conditions. Our data suggest the dependence of maternal PD on the schedule, quantity, and environment of their infants' sleep, as well as their own sleep quality and quantity. Using the same dataset, we have previously demonstrated that maternal night-time sleep duration is tightly coupled with that of their infants 10 . These findings suggest the importance of providing good night-time sleep to infants in order to secure good maternal sleep and ultimately reduce the risk of maternal PD. We have previously shown that punctual light-off time is associated with longer night-time sleep duration in infants 10 . Univariate analysis in the current study suggested that punctual light-off time for infants and punctual bedtime for mothers are associated with a lower risk of maternal PD. Regular, and possibly early, light-off time for infants might promote a swift acquisition of mature sleep patterns, potentially leading to better maternal sleep and lower incidence of PD. peripartum depression and infant care. Pathological clinical conditions of newborn infants, such as preterm delivery and hospitalisation of infants at neonatal intensive care centres, have been recognised as prominent risk variables of PD 9 . Our study suggested that, even after delivery of a healthy infant, relatively poor infant  www.nature.com/scientificreports www.nature.com/scientificreports/ growth might be the risk of PD. Because infant weight is one of the few visible markers of 'successful infant care' for mothers, poor weight gain might easily enhance mothers' anxiety. Univariate analysis in our study suggested that, in addition to the infant growth, mothers' confidence and anxiety regarding infant care was associated with high EPDS scores. These findings highlight the importance of attending to mothers' psychological characteristics to predict and prevent PD. To ameliorate the excessive pressure on mothers, it is important to provide appropriate social infant care support 8 . In the current study, the risk of PD was paradoxically dependent on the type and quality of social infant care support. While dissatisfaction with social infant care support was associated with high EPDS scores, mothers who were caring for their infants on their own showed the lowest risk of PD, suggesting that satisfaction with social infant care support is not necessarily dose-dependent, but is affected by mothers' psychological characteristics.
With respect to the type of feeding, our findings from the univariate analysis were consistent with previous studies showing the benefits of breastfeeding in reducing the risk of PD 12,13 . Breastfeeding mothers are exempted from additional night-time tasks required to prepare the formula milk, allowing them longer night-time sleep duration. Increased serum oxytocin levels of breastfeeding mothers may also influence the threshold for developing PD 14 . Although our data suggest the benefit of exclusive breastfeeding, the supplemental use of formula milk did not increase the risk of PD. Given that failure to establish breastfeeding is associated with a significant increase in the risk of PD 15 , encouragement of breastfeeding should be accompanied by sufficient support to minimise maternal psychological pressure.
other independent variables of peripartum depression. Consistent to previous studies, which demonstrated the association between maternal socio-economic status and the development of PD 5,6 , we confirmed that poor satisfaction of mothers with their family income was associated with high EPDS scores. Dependence of EPDS scores on delivery mode and parity was also suggested. However, unlike a previous study, which observed an elevated risk of PD following caesarean delivery 16 , our data in the univariate analysis suggested the association between elective caesarean delivery and lower risk of PD. Background characteristics associated with the indication of caesarean delivery might be associated with the paradoxical impact of caesarean delivery to the risk of PD. In our study, the majority of women who underwent caesarean delivery were multipara mothers, however, no further data were available regarding the indication of caesarean delivery. There currently is no  www.nature.com/scientificreports www.nature.com/scientificreports/ large-scale survey, which assessed the indication of caesarean delivery in Japanese women. In Japan, the rate of elective caesarean delivery is 11.0%, which is lower than in other developed countries 17 . Considering that this delivery mode is covered by the national insurance, it is unlikely that socio-economic status affected the decision of the delivery mode in our study cohort. Future studies need to address the impact of caesarean delivery on the risk of PD.

Strength and limitation of the study. The development of PD is dependent on factors rooted in various
aspects of maternal life, such as socio-economic status, pregnancy plan, problems with family and spouse, history of mental illness in the family, and social support 5,6 , suggesting that maternal background is insufficient to fully explain the mechanism of PD. By incorporating various background variables associated with maternal life during and after delivery, we have demonstrated that variables associated with infant care, sleep and growth are key factors to understand the mechanism of maternal PD. However, considering that clinical variables of mothers, infants, and other family members are often inter-correlated with each other (such as lifestyle and sleep schedule), it remains uncertain whether variables other than those related to maternal background are directly associated with PD, or merely via the inter-correlation between the variables. As previously described, our study was conducted by recruiting typical Japanese mothers in terms of age, academic background, and other socio-economic variables 18 . Second, we used a relatively low cut-off value of EPDS ≥ 9 to represent increased risks of PD [19][20][21] . Based on the high-enough sensitivity and specificity of this threshold to predict PD in Japanese women (82% and 95%, respectively), it was speculated that Japanese women are too reluctant to express their feelings and emotions when they are depressed 21 . Therefore, when generalising our findings to mothers in other parts of the world, careful consideration is required for cultural and traditional factors.

Conclusions
In addition to established maternal factors associated with PD, our survey with mothers of healthy 1-month-old infants highlighted that independent variables rooted in infant care, growth and sleep, were associated with the risk of maternal PD even when adjusted for known independent variables. A comprehensive, multidisciplinary approach is required to predict, prevent, identify, and treat maternal PD. Future studies need to address whether a strategy, which incorporate infant care, growth and sleep, improve prediction and prevention of PD, and thus help minimise the adverse influence of PD on the mother-child relationship. Study population. Data collection was performed with the participation of two tertiary perinatal centres (Kurume University Hospital and Kyushu University Hospital) and three private obstetrics hospitals (Izumi Ladies' Clinic, Fukuoka Birth Clinic, and Sato Women's Hospital) in Fukuoka, Japan. Preliminary data on the photoperiod and sleep patterns of infants and their mothers were reported previously, and data collection methods were described in detail 10 . Briefly, mothers of newborn infants were asked to complete a questionnaire at the time of the 1-month baby check between December 2014 and November 2015.

Methods
Questionnaires. Information was collected using the originally developed Sleep Questionnaire for Mothers of Newborn Infants 10 , EPDS [20][21][22] and ESS 23,24 . Our original sleep questionnaire comprised 45 questions in relation to mothers and their family, birth of infants, social infant care support, infants' sleep pattern and environment, and mothers' sleep condition 10 . For the current study, 30 questions were carefully chosen to represent the variables of interest; relatively a wide range of variables (birth information, child-rearing style, growth and sleep of infants, and maternal sleep status before and after delivery, and ESS scores after delivery) were considered to assess the potential dependence of high EPDS scores on infant care, sleep and growth with adjustment for known independent factors of PD and other factors associated with various aspects of maternal life. The ESS assesses sleepiness while engaged in eight different activities (score 0-24), with higher scores indicating higher daytime sleepiness 23 . The EPDS is a 10-item self-report scale specifically designed to screen for postpartum depressive symptoms (score 0-30), with higher scores indicated higher levels of depressive symptomatology. For the Japanese version of EPDS, a relatively low cut-off score of 8/9 is recommended for the screening of PD 19 . Data analysis. Questionnaires with more than 20% missing data without relevant reasons were excluded from the analysis 25 . To minimise biases derived from missing data, multiple imputation of the variable was performed (n = 5 imputations) based on the correlation between variables with missing values and other participant characteristics using SPSS ver. 21.0 (IBM, Armonk, NY, U.S.A.). A univariate logistic regression model was used to evaluate the association of potential independent variables with EPDS scores ≥9. Continuous variables with normal distribution were categorised into 2-3 levels using quartiles or clinically relevant thresholds when appropriate. P-values for the univariate analysis were not corrected for multiple comparisons because of the exploratory nature of the analysis. However, to avoid type-1 errors, p-values between 0.01 and 0.05 were regarded as chance level. A multivariate logistic regression analysis was performed to identify the final model to explain EPDS scores ≥9. For the multivariate model, mandatory variables (e.g. satisfaction with infant care support, body weight gain, daily sleep duration of infants, maternal satisfaction with sleep, and maternal satisfaction with family income) were chosen based on our hypothesis driven by previous studies, whereas additional variables were chosen based on clinical relevance, collinearity and the univariate analysis results.