The Use of Sucking Time to Establish Models to Estimate Breastmilk Intake of One Breastfeeding Session for 1-12-Month Infants: A Cross-sectional Study

Background: Observation method of assessing breastmilk intake (BI) by observing 20 medical indices has less interference in breastfeeding than test - weighing method, but for the 21 calculation certain influential predictor variables had not been recognized and needed to be 22 further studied. 23 Method: Infants between 1 and 12 months were enrolled in a hospital of Shenzhen city 24 in northern China. A cross - sectional survey was conducted in a sample of mother - infant 25 dyads. A breastfeeding session of each participating mother subject was observed in the clinic. 26 150 mother - infant dyads were adopted for data analysis. BI was measured by test - weighing 27 method. Sucking time (ST), Information of maternal and infant health, maternal breast, and 28 infant feeding practice were collected. 29 Results: There was a relationship between sucking time and breastmilk intake per 30 breastfeeding session (r=0.57, p<0.05). In the 4 multilevel models (r>0.6, p<0.05) established 31 to estimate breastmilk intake the best fit multilevel model was to estimate breastmilk intake 32 per kg (infant weight) for 1 - 4 - month infants ( R 2 =0.53, p<0.05). Sucking time as the key 33 variable and other factors including infant birth weight/ birth weight, breast side, maternal 34 BMI, maternal vocation significantly had effect on breastmilk intake and those factors that 35 remained significant varied according to different age groups. 36 Conclusions: Sucking time had a significant association with breastmilk intake in a 37 breastfeeding session. Establishing multilevel models based on sucking time to estimate 38 breastmilk intake per infant weight for specific age group of infants greatly improved the effectiveness. But the accuracy of the estimation needed to be improved for further application in telemedicine.

further studied.

23
Method: Infants between 1 and 12 months were enrolled in a hospital of Shenzhen city 24 in northern China. A cross-sectional survey was conducted in a sample of mother-infant 25 dyads. A breastfeeding session of each participating mother subject was observed in the clinic. 26 150 mother-infant dyads were adopted for data analysis. BI was measured by test-weighing 27 method. Sucking time (ST), Information of maternal and infant health, maternal breast, and 28 infant feeding practice were collected.

29
Results: There was a relationship between sucking time and breastmilk intake per 30 breastfeeding session (r=0.57, p<0.05). In the 4 multilevel models (r>0.6, p<0.05) established 31 to estimate breastmilk intake the best fit multilevel model was to estimate breastmilk intake 32 per kg (infant weight) for 1-4-month infants (R 2 =0.53, p<0.05). Sucking time as the key 33 variable and other factors including infant birth weight/ birth weight, breast side, maternal 34 BMI, maternal vocation significantly had effect on breastmilk intake and those factors that 35 remained significant varied according to different age groups.

45
Breastmilk is the optimal source for infants during the first 6 months of life, and 46 breastfeeding has short-term and long-term beneficial effects on infant health and development. WHO Global Strategy recommended that infants should be exclusively 48 breastfed for the first 6 months of life, then continued breastfeeding to 2 years old or above [1] . 49 Monitoring breastfeeding is an important part to increase the rate of breastfeeding, especially 50 exclusively breastfeeding [2] . However, breastfeeding surveys were largely supported by 51 maternal recall [3] . Thomas et al concerned about the data quality from those recalling method 52 for measuring breastmilk intake for population-based studies and suggested that available 53 analytical approaches should provide consistent results [4] .

54
Since the 1950s, researchers from various specialties have focused on measuring 55 breastmilk intake via various methods, such as test weighing, Deuterium Dilution 56 Dose-to-mother Method (DTM), observation, flowmeter, milk expression, and milk 57 production calculation [5] . Test weighing, and DTM are more accurate than the other methods 58 and regarded as gold standards for assessing breastmilk intake. However, test weighing 59 methods can often interfere with the normal breastfeeding process and DTM method can be 60 timely and costly [6][7] .

61
Observation method is often more convenient because they usually use medical 62 indices to estimate breastmilk intake and require less operating process on mothers [8] , so the 63 method can be easily used for assessment of infant feeding in telemedicine. But the accuracy 64 of this method needs to be improved.

65
There were several studies on observation method to evaluate breastmilk intake.

66
Audible swallowing was considered as a valid observational indicator to determine milk 67 intake per individual feed [6] [7] [9] . However, there was great variation among observers, and 68 observers should be trained to recognize feeding patterns [7] . What's more, each swallow 69 should be recognized and counted from the beginning to the end during one breastfeeding 70 session; meanwhile swallowing behavioral cues were not explicitly [7] . So without assistant 71 tools, swallow counting is timely and needs great effort.

72
Timekeeping is easier to observe rather than counting swallows. Scott et al used 73 breastfeeding time to estimate the amount of breastmilk consumed, and milk intake was 74 calculated as 10g/min [10] . However, this was likely an over-simplified approach, and the 75 definition of breastfeeding time was not strictly defined. Thomas et al reviewed that in 76 Dewey's study estimated 24 h breastmilk intake equaled 100*infant weight (3 months, 77 kg)+0.49*nursing time (min), but for the calculation certain influential predictor variables 78 should be considered, such as infant age and sex, infant feeding data, to validate and modify 79 the volume calculations [4] . However, measuring 24 h breastmilk intake was not easy to 80 perform, and breastfeeding frequency should be another contributor, and that would make 81 measuring work complex and time-consuming.

82
As it is presented above, we aimed to use breastfeeding time to explore a simple 83 observational method to estimate breastmilk intake of one breastfeeding episode and try to 84 find out other influence factors to improve the accuracy of the method by establishing 85 multilevel models in order to further apply in telemedicine in future.

87
Design 88 This study was a cross-sectional survey conducted in a sample of mother-infant dyads.

89
Usually, lactation is established by 1 month., so infants aged from 1 month to 12 months were 90 enrolled due to the stability and sustainability of breastmilk expression [11] . 30 dyads were 91 planned to be recruited in each for four groups (1-2, 3-4, 5-6, 9 -12 months) by stratified   Test-weighing was used to measure milk profile per observed breastfeeding session by weighing infants on an accurate digital scale (Sophisticated Baby Weight Scale, Betterren Inc, 107 Shanghai, China, resolution 2g) immediately before and after feeding with one breast. And 108 weights before breastfeeding were adopted as infants present weights. Infants did not have 109 their clothing removed for the weighing to minimize disruption.  Questionnaires were administered to determine mother-infant dyad health, maternal 123 breast information during the breastfeeding session, and infant feeding information.  Numerical variables examined for non-normal distribution, outliers, and missing data, were 127 presented as the mean (X), standard deviation (SD), and minimum and maximum.

138
General information 139 We enrolled 181 mother-infant dyads, among which 31 were excluded, due to 140 breastfeeding time recorded on bilateral breasts (24 mothers), birth weights less than 2500g 141 (5 infants) and ineligible breastfeeding data (2 infants). Considering breast side might 142 contribute to the variances of breastmilk intake, our study collected the data of one episode, 143 and separated the right and the left breast.

144
In total 150 mothers who breastfed their infants aged 1-12 months with one breast were 145 included in the analysis.   Figure 1 shows the scatter plot of BI distribution by ST, and BI was positively  Table 2 indicates that, average BI from the right breast was higher than the left 162 (61.14+32.14 vs. 49.88+29.35g, p<0.05); average VB from the right breast was greater than 163 the left (8.12+4.48 vs. 6.66+3.38 g/min, p<0.05); average VB in housewives was higher than 164 working mothers (8.02+4.22 vs. 6.60+3.65 g/min, p<0.05); and VB was higher in infants of 165 5-12 months compared to those of 1-4 months (8.15+4.25 vs. 6.67+3.70 g/min, p<0.05 ).  evaluation one breastfeeding session was observed [6] . swallowing will affect milk intake [14] . Therefore, when models were established to estimate 210 breastmilk intake of unilateral breast, factors influencing the short-term ability of breastmilk 211 production and intake should be considered. Here we divided factors into three groups: 212 factors from infants, maternal factors and feeding practice, which may have short-term effect. 24-hour volume of breastfeeding [4] . Our data indicated that gender did not play a significant 224 role in the multilevel models to estimate breastmilk intake of one feed.

226
In the present study, whether working played significantly in one of our multilevel 227 models. Gilany et al reported that housewife was one of the independent predictors of 228 exclusive breastfeeding [15] , which supported that housewives tended to express more 229 breastmilk. Maternal postpartum BMI was negatively associated with breastmilk intake. 230 Diana et al got the similar finding that maternal BMI was not a protective factor of milk 231 production for breastfed infants [17] .

233
Our multilevel models showed breast side was a key contributor of breastmilk intake 234 and the right side produced more breastmilk. This may be due to the fact that mostly the right .Also, using bilateral breasts separately to assess milk production was reported by Kent et 239 al [20] . There was a hypothesis that right breasts produced more milk in right-handed mothers 240 due to more blood flow of the right side [11]. Therefore, it's necessary to distinguish breast 241 sides in estimating BI of one episode, and information of maternal handedness which was not 242 available in the study should be also recorded. 243 Thomas suggested approaches to classify feeding style, including exclusively 244 breastfeeding and combination feeding [4] , but our study used breastfeeding frequency for 245 replacement and found that feeding frequency was not a significant factor in the multilevel 246 models. Although breastfeeding frequency can be an indicator of assessing breastfeeding, 247 further research still needs to separate exclusively breastfeeding and combination feeding [18] .

248
Better fit models 249 We found birth weight impacted breastmilk intake in younger infants while present   Figure 1 Distribution of BI by ST in one breastfeeding session. Figure Legends: Scatter plot of BI distribution by suckling time, and breastmilk intake was positively related to ST (r=0.57, p<0.05). Linear regression was used, and there was a linear relationship between suckling time and breastmilk intake (b=5.37, p<0.05).