Are mothers certain about their perceptions of recalled infant feeding history?

Abstract

Background:

Maternal recall of infant feeding practices is used frequently in epidemiologic studies to identify the relationship between the dose and exposure of mothers’ own milk (MOM) and subsequent health outcomes. Maternal recall has been identified as a potential source of measurement bias.

Objective:

To describe maternal recall certainty of the dose and exposure of MOM feedings at four sequential time periods post birth.

Materials and Methods:

In this secondary analysis of data from a larger study, mothers of children 4–36 months of age were asked to (1) describe infants’ MOM dose and exposure at four sequential time periods post birth and (2) rate the certainty of their recall.

Results:

Eighty mothers provided complete data. MOM was the first feeding for 78.5% of infants and was received by 83% during the first week, 85% during the first month, and 62% during the fourth month. Maternal ratings of recall certainty were >95% for each of the four time periods. Although reported recall certainty was significantly different for the four time periods (Χ² =9.67, p=0.02), no two time periods were significantly different in post hoc analyses. Maternal recall certainty was highest with the infants’ first feed, with no relationship for other time periods.

Conclusion:

Maternal recall certainty of infant feeding at four sequential time periods post birth was high regardless of elapsed time. The measurement of maternal recall certainty may be useful in clinical practice and subsequent studies linking MOM exposure to later health outcomes.

Introduction

Exposure to mother’s own milk (MOM; by breast, bottle or tube) is a commonly used independent and dependent variable, especially in epidemiologic studies that seek to link MOM feeding with long-term health outcomes in recipient infants and in the mothers themselves (Feltner et al., 2018). Although some of these studies are prospective (Abdel-Hady & El-Gilany, 2016; Agampodi, Fernando, Dharmaratne, & Agampodi, 2011; Amissah, Kancherla, Ko, & Li, 2017; Bland, Rollins, Solarsh, Van den Broeck, & Coovadia, 2003; Burnham et al., 2014; Cupul-Uicab, Gladen, Hernández-Avila, & Longnecker, 2009; Gillespie, d’Arcy, Schwartz, Bobo, & Foxman, 2006; Natland, Andersen, Nilsen, Forsmo, & Jacobsen, 2012; Promislow, Gladen, & Sandler, 2005; van Zyl, Maslin, Dean, Blaauw, & Venter, 2016), the majority rely upon maternal recall of early feeding practices and mothers are often asked about these practices weeks, months, or even years after MOM feeding has ended. For example, maternal recall has been used to study the relationship between MOM feeding and childhood infections (Downham, Scott, Sims, Webb, & Gardner, 1976; Pullan et al., 1980), allergies and asthma (Abreo, Gebretsadik, Stone, & Hartert, 2018; Lodge & Dharmage, 2016), neurocognitive development (Kramer et al., 2008; Victora et al., 2015), and obesity (Ortega-García et al., 2018), as well as long-term maternal outcomes including cardiac disease and breast cancer (Feltner et al., 2018). The findings from these studies have been used to inform global health policies focused on the promotion and protection of MOM feedings (Agostoni et al., 2009; American Academy of Pediatrics Section on Breastfeeding, 2012; World Health Organization, 2002). Thus, it is critically important to understand and reduce the bias of techniques that measure maternal recall.

The extensive use of maternal recall in epidemiologic studies was highlighted as a methodological weakness in the most recent Agency for Healthcare Research and Quality (AHRQ) report on the outcomes of MOM feeding (Feltner et al., 2018). Maternal recall of MOM exposure is multifaceted and involves not only remembering the infant’s feeding history, but also reporting it truthfully. For example, if mothers are aware of a purported benefit of MOM feeding, they may be inclined to report their own infant’s feeding history inaccurately by overstating the exposure to MOM during research participation and during clinician visits (Drews & Greeland, 1990). Similarly, the accuracy of maternal recall may be affected by socioeconomic status, social norms, precision of memory, or differential recall (Abdel-Hady & El-Gilany, 2016; Amissah et al., 2017; Cupul-Uicab et al., 2009; Drews & Greeland, 1990; Promislow et al., 2005). Researchers have attempted to reduce the bias in the techniques for measuring maternal recall by shortening the time between exposure to MOM and its measurement, as well as measuring MOM exposure prospectively and validating these responses later with instruments that use recall techniques (Abdel-Hady & El-Gilany, 2016; Agampodi et al., 2011; Amissah et al., 2017; Bland et al., 2003; Burnham et al., 2014; Cupul-Uicab et al., 2009; Gillespie et al., 2006; Natland et al., 2012; Promislow et al., 2005; van Zyl et al., 2016). However, retrospective recall of MOM exposure remains the primary technique in this line of research, and little is known about the certainty with which mothers perceive their responses to questions regarding early infant feeding practices. Asking mothers about certainty of their recall of infant feeding is an assessment of the memory in real-time. It is clinically important, as asking about certainty may lead to the provision of additional information about infant feeding.

As a component of a larger study, we measured maternal recall of MOM exposure in 80 mothers of children who had participated as a follow-up cohort for a previous randomized trial (RESTORE) (Curley & Watson, 2018; Curley et al., 2015). Using a structured survey instrument, we queried mothers about the child’s exposure to MOM at four time points post birth: the first feeding and over the first week, the first month, and the first 4 months of life. However, unlike previous surveys of this type, we also asked the mothers to rate the certainty of correctness of their recall for each time period. Thus, the purpose of this study was to: (1) describe the mothers’ reported infant exposure to MOM at these four time periods over the first 4 months post birth, (2) describe maternal recall certainty about infant MOM exposure at each time period; (3) compare time-dependent differences in maternal recall certainty over these time periods; and (4) explore the relationship between maternal recall certainty and infant age at the time when mothers completed the feeding survey.

Materials and Methods

Sample and Design

This was a secondary analysis of completed maternal recall surveys from RESTORE-Cognition (Curley & Watson, 2018), a follow-up study for an original 31-center cluster randomized controlled trial (RCT) evaluating a nurse-led sedation algorithm on the duration of mechanical ventilation in 2449 critically ill infants and children with acute respiratory failure (RESTORE) (Curley et al., 2015). To be eligible for the current study, mothers had to speak English, answer “yes” to the question of their ability to respond to questions about feeding during their infant’s first 4 months of life, and have children ages ≥4 months to <3 years at the time of enrollment into the original RCT. Although this sample included 157 completed maternal recall surveys, only 80 surveys contained information about all four selected post-birth time periods of interest (Figure 1). Institutional Review Broad approval was secured for Ann & Robert H. Lurie Children’s Hospital of Chicago, the University of Pennsylvania and Rush University Medical Center.

FIGURE 1.

CONSORT STATEMENT

Measures

A history of MOM feeding survey was developed to collect data about the amount and duration of breastfeeding and formula feeding. The original survey was designed to measure the dose and exposure period of MOM feedings as well as the certainty of maternal recall. This instrument incorporated a Likert scale for each question and the mother’s corresponding rating of her recall certainty for that question at five sequential time points (first feed, first week, and 1-month, 2- to 3-month, and 4- to 6-month feedings). After development, the survey was tested with 10 mothers of diverse racial and ethnic backgrounds and was reviewed by research experts in MOM feeding and pediatric critical care. Then, the instrument was modified so that the original Likert scale was replaced with a numeric percentage scoring system to measure dose of total MOM feeding and the certainty of maternal recall. The survey was retested with the same 10 mothers and reviewed again by the research experts, resulting in the removal of the 6-month time point measure and the summary questions.

The final instrument was an 18-item survey consisting of multiple choice questions asking mothers about child feeding at four time periods: the first feed immediately post-birth, over the first week of feeding, over the first month of feeding, and over the fourth month of feeding. For the very first feeding, mothers were asked to choose from breastmilk, formula, sugar water, water, or donor breastmilk. Mothers were then asked to rate their perception of the certainty of their recall using a percentage (0–100%). For the following three time periods (first week, first month, and fourth month), mothers were asked about infant feeding, with choices including MOM, formula, a variety of other liquids such as juices, tea, and water, and solids. If the mother reported that the infant received MOM over any of the three remaining time periods, she was asked to quantify the percentage of all feedings that consisted of MOM during the specific time period. Mothers were asked to rate their perception of their recall certainty for each of the three time periods, using a scale from 0% to 100%.

Other measures used in this study were collected from the RESTORE dataset (Curley et al., 2015) and included infant demographic characteristics, estimated family income using zip code as a proxy variable, infant health characteristics, diagnoses and intensive care unit (ICU) course, and length of stay in the ICU and the hospital. The age of the child (months) at the time of the survey was calculated by summing the infant age at the time of RESTORE enrollment to the number of months between RESTORE enrollment and survey completion. This number was then converted to child age in years by dividing the number of months by 12.

Study Procedures

During participation in RESTORE-Cognition (Curley & Watson, 2018), eligible mothers were invited by telephone to join this study. The survey was administered by trained research assistants during a phone interview with mothers or was completed on paper and returned to the data center by mothers themselves. Data were entered into a secure data management program and linked securely with data from the initial RESTORE dataset (Curley et al., 2015).

Statistical Analysis

Descriptive statistics were used to summarize demographic data, feeding type at each of the four time periods, and mothers’ certainty of their recall about feeding over the four time periods. The Friedman Non-Parametric Ranked Means test was used to compare recall certainty over the four time periods and infant age at time of survey completion. A post hoc analysis was done using the Wilcoxon-Signed Rank test. All analyses were done with SPSS® (SPSS).

Results

Eighty mothers completed the survey; 48 completed the survey during a phone interview and 32 completed it on paper and returned it to the data center. Characteristics of the 80 infants whose mothers completed the surveys are described in Table 1. At the time of enrollment into RESTORE (Curley et al., 2015), the infants ranged in age from 4 months to 3 years, were more likely to be female, non-Hispanic Caucasian, healthy at baseline, and free of previous medical illnesses. The mean age of the children at the time their mothers completed the survey was 4.9 years (2.3–7.9 years).

TABLE 1.

CHARACTERISTICS OF THE SAMPLE (n = 80)

Characteristic	Number (%)
Infant age years mean, (SD)	1.2 (1.7)
Male	33 (41.3)
Race/ethnic group
non-Hispanic white	51(63.8)
non-Hispanic black	11(13.8)
Hispanic	7 (8.8)
Other	11(13.8)
Baseline PCPCa=1	1 (100)
Baseline POPCa=1	77 (96.3)
Income category based on zip codeb
<$40,000	8 (10)
$40,000–79,999	50 (62.5)
>=$80,000	22 (27.5)
PRISM-III-12 score median (IQR)	5.5 (2.25–10)
Risk of Mortality median (IQR)	1.96 (0.76–8.04)
Primary Diagnosis
pneumonia	26 (32.5)
bronchiolitis	32(40)
acute respiratory failure relatad to sepsis	7 (8.8)
asthma or reactive airwy disease	5 (6.3)
aspiration pneumonia	2 (2.5)
otherc	8 (10)
Past Medical Hisory
prematurity	0 (0)
asthma	1 (1.3)
seizures	0 (0)
cancer	1 (1.3)
chromosomal abnormality	0 (0)
CPR	0 (0)
PICU LOS in days (IQR)	10.3 (5.5–10.9)
Hospital LOS in days (IQR)	14.9 (7–19)
Child age at survey completion years median, (IQR)	2.9 (3.5–5.9)

Abbreviations: CPR, cardiopulmonary resuscitation; IQR, interquartile range; OI, oxygenation index; OSI, oxygen saturation index; PARDS, pediatric acute respiratory distress syndrome; PCPC, Pediatric Cerebral Performance Category; PICU, pediatric intensive care unit; POPC, Pediatric Overall Performance Category; PRISM III-12, Pediatric Risk of Mortality III score from first 12 hours in the PICU.

^aPCPC and POPC range from 1 to 6, with higher categories indicating greater impairment.

^bMedian household income of zip code of residence in 2011.

^cOther primary diagnoses include laryngotracheobronchitis, pulmonary edema, and thoracic trauma

Types and amounts of feeding for the four time periods are summarized in Table 2. MOM was the first feeding for 78.5% of infants. At the following three time periods, 83% of infants during the first week, 85% during the first month, and 62% during the fourth month received some MOM. Receipt of some formula for the first feed, and during the first week, first month, and fourth month was reported as 21.3%, 43.8%, 58.8%, and 68.8% of infants, respectively. Although the first feeding was a mutually exclusive choice, each subsequent time period could have included more than one feeding choice. Over time, the percentage of total feeds consisting of MOM decreased from nearly 73% at 1 week to 49% at 4 months post birth.

TABLE 2.

AMOUNT AND FEEDING TYPE BY TIME PERIOD (n = 80)

Feeding Time Perioda	Feeding Type Number (%)
	MOM	Formula	Cereal	Baby Food	Water	Otherb	Unsure
First feed	62 (78.4)	17 (21.3)	0 (0)	0 (0)	0 (0)	0 (0)	1 (1.3)
1 week	69 (83.3)	35 (43.8)	0 (0)	0 (0)	1 (1.3)	1 (1.3)	0 (0)
1 month	68 (85)	47 (58.8)	0 (0)	0 (0)	1 (1.3)	0 (0)	0 (0)
4 months	50 (62.5)	55 (68.8)	18 (22.5)	12(15)	3 (3.8)	3 (3.9)	0 (0)

^aSubjects may have fed more than one feeding type at each time period

^bSugar water, tea, rice milk, cow’s milk

Maternal recall certainty is summarized in Table 3. Nearly 89% of mothers rated their recall of the infant’s first feed as 100% certainty, whereas one mother rated her recall as 0% certainty. The remainder of mothers rated their recall certainty for the first feed between 45% and 95%. Over the subsequent time periods, maternal recall certainty was scored as 100% for most mothers, with 81%, 77%, and 74% of mothers citing 100% recall at 1 week, 1 month, and 4 months, respectively. In addition to one mother who reported 0% recall certainty of the first feed, two additional mothers reported recall certainty of 0%: one at 1 week and the other at 1 month. However, all three mothers completed feeding information for these time periods.

TABLE 3.

MATERNAL CERTAINTY OF RECALL AT FOUR TIME PERIODS (n = 80)

	Mean (%)	Std. Deviation (%)	Minimum (%)	Maximum (%)	Median (%)	Mode (%)
Recall 1st feed	96.63	14.14	0	100	100	100
Recall 1 Week Feed	95.94	14.17	0	100	100	100
Recall 1 Month Feed	96.06	12.82	0	100	100	100
Recall 4 Months Feed	95.06	10.39	50	100	100	100

A Friedman Non-Parametric Ranked Means test revealed a statistically significant difference in maternal recall certainty among the four time periods (Χ² =9.67, p=0.02). However, a post-hoc analysis using Wilcoxon-Signed Rank test revealed no significant difference between any of the possible paired time periods. A Spearman’s Rho correlation revealed a moderate negative correlation between child age at the time of the survey and recall certainty at first feed (r_s= −.311, p=0.005), but no relationship between child age at the time of the survey and recall certainty was found at 1 week (r_s= −.089, p=0.43), 1 month (r_s= −.072, p=0.53), or 4 months (r_s=.072, p=0.52).

Discussion

To our knowledge, this is the first study to describe maternal perception of recall paired with certainty of correctness of infant feeding practices at four time points during the first 4 months post-birth. The results demonstrate that most infants received MOM as the first feed and continued to receive MOM over the remaining three time periods, with decreasing percentages of MOM and increasing percentages of formula between 1 and 4 months. By 4 months, mothers reported that nearly 40% of infants had begun receiving cereal and/or baby food as portion of the total diet. Whereas certainty of maternal recall was high for each of the four time periods, the three mothers who reported 0% recall certainty also answered survey questions about these feeding practices. Despite a statistically significant difference in maternal recall certainty among the four time periods, none of the six pairs of time points was significantly different in post hoc analyses. Maternal recall certainty of the first feeding was higher with younger infant age at the time of the survey; however, there was no difference in recall certainty over the following three time periods with respect to infant age at the time of survey completion.

To apply research on the protective effects of MOM feeding, clinicians must understand the process used to obtain and measure infant feeding data. It is also important for clinicians to understand that no universal standards for the measurement of maternal recall currently exist, and the most recent AHRQ summary of outcomes associated with MOM feeding cited the extensive use of maternal recall as a source of methodological bias in epidemiologic studies (Feltner et al., 2018). In the absence of standardized approaches, investigators have measured MOM exposure using maternal recall with different instruments, over different exposure periods of interest, at different time points between MOM exposure and data collection, and with different definitions for MOM feeding (Abdel-Hady & El-Gilany, 2016; Agampodi et al., 2011; Amissah et al., 2017; Bland et al., 2003; Burnham et al., 2014; Cupul-Uicab et al., 2009; Engebretsen et al., 2007; Fenta, Yirgu, Shikur, & Gebreyesus, 2017; Gillespie et al., 2006; Mulol & Coutsoudis, 2018; Natland et al., 2012; Promislow et al., 2005; Strippoli, Silverman, Michel, & Kuehni, 2007; van Zyl et al., 2016; Vossenaar, van Beusekom, Doak, & Solomons, 2014). Although most investigators, like this study, have used paper and/or interview questionnaires to acquire these data (Abdel-Hady & El-Gilany, 2016; Agampodi et al., 2011; Amissah et al., 2017; Bland et al., 2003; Burnham et al., 2014; Cupul-Uicab et al., 2009; Engebretsen et al., 2007; Fenta et al., 2017; Gillespie et al., 2006; Mulol & Coutsoudis, 2018; Natland et al., 2012; Promislow et al., 2005; Strippoli et al., 2007; van Zyl et al., 2016; Vossenaar et al., 2014), others have used medical record extraction (Burnham et al., 2014), maternal feeding diaries (Bland et al., 2003; Burnham et al., 2014), event calendars (Agampodi et al., 2011), and isotope measurement in infants (Mulol & Coutsoudis, 2018). Maternal recall has been used to measure MOM exposure for both exclusive MOM feeding (Abdel-Hady & El-Gilany, 2016; Agampodi et al., 2011; Bland et al., 2003; Burnham et al., 2014; Fenta et al., 2017; Mulol & Coutsoudis, 2018; Vossenaar et al., 2014) and for any MOM feeding (Amissah et al., 2017; Cupul-Uicab et al., 2009; Gillespie et al., 2006; Natland et al., 2012; Promislow et al., 2005; Strippoli et al., 2007; van Zyl et al., 2016), over various post-birth periods of time. Measures of the duration of MOM exposure have ranged from one study in which maternal recall for both exclusive and partial MOM feeding was measured at only one time point (Vossenaar et al., 2014), to another study in which exposure to any MOM feeding was measured by collecting serial recall data over a 13-month period (Natland et al., 2012).

Time points between MOM exposure and data collection, often referred to as elapsed time, have also varied widely among studies, yielding inconclusive results. Whilst most mothers have been asked to recall feeding information during the infant’s first year of life, the elapsed time between MOM exposure and data collection has ranged from 24 hours to 50 years (Abdel-Hady & El-Gilany, 2016; Agampodi et al., 2011; Amissah et al., 2017; Bland et al., 2003; Burnham et al., 2014; Cupul-Uicab et al., 2009; Engebretsen et al., 2007; Fenta et al., 2017; Gillespie et al., 2006; Natland et al., 2012; Promislow et al., 2005; van Zyl et al., 2016; Vossenaar et al., 2014). Although some investigators report that maternal recall of infant feeding is accurate over an elapsed time of 24 hours to 20 years (Fenta et al., 2017; Natland et al., 2012; van Zyl et al., 2016; Vossenaar et al., 2014), others have revealed that greater elapsed time is associated with decreased maternal recall accuracy (Abdel-Hady & El-Gilany, 2016; Amissah et al., 2017; Gillespie et al., 2006). One study focused on addressing the reliability of maternal recall by collecting prospectively measured MOM exposure and comparing it with a recall measure up to 50 years later in a cohort of 140 mothers (Promislow et al., 2005). Results revealed a moderate agreement between the two measures, with 94% of the sample accurately recalling “ever feeding MOM,” but only 54% accurately recalling the exposure period (Promislow et al., 2005). These data suggest that recall of “any” MOM feeding after a lengthy elapsed time may be relatively accurate, but the exposure period represents a source of recall bias. In our study, we found an inverse association between elapsed time and maternal recall certainty only for the first feed, suggesting that recall certainty for the first feed may have been affected by birth complications such as maternal anesthesia use or infant emergencies (Keenan et al., 2017).

Although the World Health Organization (WHO) (2002) has recommended the measurement of maternal recall using an entire prior 24-hour feeding period, referred to as 24-hour recall, bias still exists in this measurement. Several studies revealed that mothers overestimate MOM exposure during the prior 24-hour feeding period (Engebretsen et al., 2007; Fenta et al., 2017; Mulol & Coutsoudis, 2018). In contrast, other investigators report that both under- and overestimation of MOM exposure is greater when mothers are asked to recall “any” versus “exclusive” MOM feedings during any interval, including the 24-hour recall period (Engebretsen et al., 2007; Fenta et al., 2017; Vossenaar et al., 2014) recommended by WHO (2002). This difference in bias for “any” versus “exclusive” MOM feedings is present regardless of the elapsed time between birth and the recall measurement (Abdel-Hady & El-Gilany, 2016; Agampodi et al., 2011; Amissah et al., 2017; Bland et al., 2003; Burnham et al., 2014; Fenta et al., 2017; Gillespie et al., 2006; Natland et al., 2012; Promislow et al., 2005; Vossenaar et al., 2014). The primary strategy employed by investigators to reduce maternal recall bias has been the measurement of MOM exposure at a baseline time point, and then comparing these data to those collected by maternal recall at a later time point(s) (Abdel-Hady & El-Gilany, 2016; Agampodi et al., 2011; Amissah et al., 2017; Bland et al., 2003; Burnham et al., 2014; Fenta et al., 2017; Gillespie et al., 2006; Natland et al., 2012; Strippoli et al., 2007; van Zyl et al., 2016). Lastly, Mulol and Coutsoudis (2018) compared the accuracy of 24-hour recall with isotope-labeled water concentrations in infant and maternal saliva to evaluate exclusivity of MOM feeding. To complete this procedure, mothers drank a stable isotope that appeared in the saliva and MOM and in infant saliva after breastfeeding. The higher salivary concentrations in exclusively breastfed infants were compared to maternal 24-hour recall data. While this technique was determined to be an accurate measurement of exclusive MOM feeding, it is costly and labor-intensive. Furthermore, mothers still over-reported the exclusivity of MOM feeding, demonstrating maternal recall bias regardless of the accuracy of the measure.

Our findings indicate that mothers are willing to rate their certainty of infant feeding recall. Therefore, clinicians can ask about recall certainty during healthcare interactions, providing an additional tool for assessing infant feeding. More specifically, if mothers report infant feeding recall followed by a low percent of recall certainty, clinicians can engage in further discussion to assess infant feeding. Because infant feeding is critical to growth and development, particularly early in life, clinicians can use these discussions to provide anticipatory guidance and education.

This study has several strengths, including testing the survey instrument with mothers of diverse racial and ethnic backgrounds and the novel approach of asking mothers to rate their recall certainty at each time period of interest. Additionally, to our knowledge, this is the first study to collect these detailed data over four specific sequential time periods post-birth. However, the generalizability of our study is limited by the small sample size and the inclusion of primarily mothers with non-Hispanic white infants who had experienced a critical illness. Furthermore, we did not collect co-variates that may have affected the reliability of maternal recall, including maternal characteristics such as parity, smoking, and years of formal education (Abdel-Hady & El-Gilany, 2016; Amissah et al., 2017; Cupul-Uicab et al., 2009; Promislow et al., 2005).

Conclusions

In summary, we found that maternal perceptions of recall certainty of infant feeding over four sequential time periods during the first 4 months of life were very high for all time periods. Three mothers rated their recall certainty as 0%, but still provided estimates of MOM exposure. These findings fill an important gap in the literature by acquiring data about exclusive and any MOM feedings using precise definitions at sequential time points throughout the first 4 months of life and partnering these data with measurement of maternal recall certainty for each time point. We conclude that this approach can be used in subsequent studies of maternal recall of MOM feedings as well as clinical care.