One-week recall period gives a more accurate estimate of exclusive breastfeeding practice than 24-h recall among infants younger than six months of age

Exclusive breastfeeding (EBF) is an important indicator of infant feeding behavior [1]. World Health Organization (WHO) defines EBF as feeding an infant only with breast milk excluding other liquids, semisolid or solid foods, except medicines, ORS, vitamin and minerals [2]. The WHO stress the importance of breastfeeding and proposed a set of criteria and indicators to measure infant feeding practices [3]. These indicators are used to identify high risk groups for malnutrition, compare the adequacy of infant feeding practice nationally and internationally, monitor and evaluate nutritional interventions and programs and for research purposes [4]. Infant and young child feeding (IYCF) indicators is a group name given to a set of indicators measuring breastfeeding and other young child feeding practices, and “the proportion of infants less than six months of age who are exclusively breastfed in the past 24 hours” is a key indicator measuring breastfeeding practices [5]. It draws on data collected using the 24-h recall method, where mothers are asked to report their child’s feeding practice of the previous day [6].

There are practical justifications to use a one-day recall as a proxy to determine breastfeeding practice. Resource implications of taking repeated measurements, ease of the indicator to make cross country comparisons, and reduced risk of recall bias are among the reasons [7]. Nonetheless, the low accuracy of using a 24-h recall to describe typical dietary pattern of specially a smaller sample is evident. Failing to account for feeding practices beyond the last 24-h could lead to misclassifying infants as exclusively breastfed [8].

Exclusive breastfeeding prevalence estimates obtained using the 24-h recall method often overestimated the true prevalence [9‐12]. Studies indicated a 9.2 and 47.4% overestimation when breastfeeding practice data collected using single 24-h recall was compared with prospectively collected data [13, 14]. In our former publication, we have identified a 23.5% overestimation of the prevalence of EBF practice when estimates from single 24-h recall was compared with the cumulative score of seven repeated 24-h recalls [12]. Though seven repeated 24-h recall is more accurate, it is not practical for routine use.

Having a more accurate indicator is important to show the true picture of the practice thereby informing policy makers to take appropriate actions towards improving child feeding practices. In this study we hypothesized that estimates of breastfeeding practice employing a one week recall period has a superior accuracy than the estimates obtained using single 24-h recall. The validation involved comparing the estimates from a recall of breastfeeding practice over a period of one week against the reference standard (cumulative of the estimates from multiple 24-h recall of 14 consecutive days) among infants younger than six months of age.

From the cohort of 408 mother-infant pairs recruited at the beginning of the follow-up, 391 (95.8%) participated throughout the follow-up period. Sex of the study participants was comparable where 181 (46.3%) infants were male and 210 (53.7%) were females. The mean age of infants was three months (± 1.5SD), 36.5% of the participants cannot read or write and the majority, 315 (80.6%) were housewives (Table 1).

Table 2 shows exclusive breastfeeding prevalence estimated by using single 24-h recall and multiple 24-h recall. This multiple 24-h recall is the cumulative score of EBF practice derived from more than one day (two and above consecutive days) interview. The highest prevalence was found from single 24-h recall, 71.4% (95% CI 66.9, 75.9) and the second highest estimate was from two consecutive 24 h recall 62.7% (95% CI 57.7, 67.3). The results showed that as the number of repeated 24-h recall increased the prevalence of EBF decreased. The fourteen repeated 24-h recalls resulted in the lowest EBF prevalence, 47.1% (95% CI 42.1, 52.1).

Based on the data from the short period recalls, the prevalence of exclusive breastfeeding decreased as the duration of recall increased until a recall of the past seven days. After the 7th day recall the estimate in the prevalence of EBF started to increase. A recall of the feeding practice over the past two days yielded an estimated 64.2% (95% CI 59.4, 68.9) prevalence of EBF, but it decreased to 54.2% (95% CI 49.3, 59.2) when the recall period covered the previous seven days (one-week) (Table 3).

A significant difference in exclusive breastfeeding prevalence was observed between the data from short recalls and the reference standard. The highest overestimation, 24.3% (95% CI 19.8, 28.8) was observed when single 24-h recall was employed (Table 3). The overestimation grew smaller when shorter recalls were considered. A recall within a period of one week bear the lowest overestimation, 7.1% (95% CI 3.8, 10.5), but when the recall period exceeded seven days period the over estimation increased.

Table 4 shows change in the patterns of EBF prevalence. The difference in exclusive breastfeeding prevalence was examined among the different age groups. The proportion of infants who were exclusively breastfed decreased as the age of infants increased regardless of the method used. Using short period recalls the prevalence of EBF ranged from 74.4–88.0% among infants aged 0–1 months, from 60.6–81.9% among infants aged 2–3 months and from 29.5–46.7% among infants aged 4–5 months.

A significant overestimation was observed when short period recall was compared with the reference standard among infants of age 0–1 month, 2–3 months and 4–5 months. Using seven days recall the lowest overestimation of EBF was observed among the age category of 2–3 months and 4–5 months (Table 4).

The sensitivity and specificity of short recall methods was determined using the 14 repeated 24-h recall as a reference. The specificity of short period recalls ranged from 54.4 - 83.5%. Table 5 shows comparison of short period recalls with 14 repeated days 24-h recall as a reference in determining exclusive breastfeeding. The lowest specificity (54.4%) and positive predictive value (66.3%) was observed when single 24-h recall was compared against the reference standard. The highest specificity (83.5%) and positive predictive value (84.0%) was found when seven days recall was applied. The ability of the short period recalls to exclude those who were not exclusively breastfed (specificity) improved up to the recall over seven days period (Fig. 1).

Sensitivity and specificity of short period recalls was also examined across the different age groups. The specificity in all age groups except the first group increased as the number of recall days increased up until the 7th day and decreased afterwards. The specificity increased from 37.5% obtained by 24-h recall to 77.5% obtained by six days recall among infants of age 0–1 month. Similarly, it increased from 39.0% obtained by 24-h recall to 81.4% obtained by seven days recall among infants age 2–3 months. This improvement in specificity was also observed among infants age 4–5 months.

The specificity, sensitivity and positive predictive value (PPV) of seven-day recall was examined against different sociodemographic variables. The specificity of seven days’ recall was higher (88.9%) among infants of age of 4–5 month and lower (72.5%) among 0–1 month old. Higher specificity of seven days recall was observed among mothers who attended school (86.5%). The variability was also observed among housewife and working mothers with specificity of 90 and 81.9% respectively.

This study aimed to assess an alternative indicator for EBF prevalence that addresses the trade-off between accuracy and feasibility of the measurement. We tested the accuracy of EBF prevalence estimated using recall over a one-week period. In doing so we tested the accuracy of estimates from varying recall periods as well. The exclusive breastfeeding prevalence estimated using the single 24-h recall was 71.4% (95% CI 66.9, 75.9) while it was 47.1% (95% CI 42.1, 52.1) when estimated by the reference method resulting in a 24.3% overestimation. A week recall (a recall over seven days’ period), resulted in the smallest discrepancy in EBF prevalence estimate (7.1%) as compared to the reference method.

Our findings indicated that 24-h recall over estimates exclusive breastfeeding prevalence. Such overestimations from 24-h recall have also been demonstrated in other studies [18, 19]. Similarly, former prospective studies identified an over estimation ranging 8.4 to 14.6% when estimates of 24-h recall method was compared with recall since birth [18‐20]. Misclassification is common when single 24-h recall is used and this is misleading and undermines the effort to further improve breastfeeding practices.

Short periods recall methods showed a significant improvement in performance of estimating EBF practice. Comparing with the reference method, short period recalls showed improvement in specificity and PPV. For instance, there was improvement in specificity from 54.4% obtained from 24-h recall to 67.1% from a recall over two days period. Even though two days (48 h) recall had a better specificity and PPV, there were children who were still misclassified as being exclusively breastfed. This might be due to children who were fed additional food and drinks after the second day. This shows even though two days recall is better than 24-h recall in estimating breastfeeding practice, it still doesn’t capture the true change in feeding patterns. Several previous studies have also showed disagreement between different methods of recall. A study conducted in South Africa comparing 48 h recall with data collected prospectively at three data points 48 h recall methods similarly found low specificity (65–89%) and positive predictive value (31–48%) of 48 h recall method [21].

In our result, we have seen that EBF prevalence estimates decreased as the number of days the recall period spans progressively increased from one day to seven days. In addition, the ability of the short period recall methods in representing the feeding pattern improved up to a recall period that spanned for seven days period (a week recall). Among the multiple short periods of recall, a week recall had optimal specificity (83.5%) and positive predictive value (84.0%) promising a better ability to capture typical child feeding behavior. A study done to validate different methods of data collection on duration of exclusive breastfeeding has found similar results from a recall period spanning seven days [21]. Comparing with three times a week recall, a study has found seven day recall to accurately reflect EBF practices with 96% sensitivity and 94% specificity [21]. The increase in the number of days involved in the recall period, beyond a mere 24-h, might have captured more variability in child feeding practices since infant feeding practices vary widely within short periods of time [22].

In addition, week recall gives a more accurate picture of the usual infants feeding practice by capturing the day-to-day variation since it involves all days of the week. For example, in our study area there is a culture of giving infants an herb called “Anita” twice a week believing it would help the infant to gain weight. This plant is given on Wednesdays and Saturdays, which are market days. Such practices that have an impact on infant feeding practice cannot be captured by using a single day measurement, but by involving a longer recall period. Using a week recall method would help us capture such variations in feeding practices. In addition, since the recall period/duration is relatively short, mothers tend to have a better recall of what they have given thereby minimizing the possibility of recall bias.

Even though we have observed an improvement in the EBF estimate with short period recalls, we should be careful when we extend the recall period beyond seven days. We have observed a persistent increment in EBF over estimation when we asked a mother to recall a feeding practice beyond a week. In addition, the specificity and PPV also drops as the recall period extends from 8 to 14 days ranging (80.2–64.7%) and (81.2–70.8%) respectively. The presence of more cases of exclusive breastfeeding in the recall days exceeding 7th day, might be due to the presence of more false positive cases of EBF since the positive predictive value drops as the recall duration increases. As the number of days of recall increased beyond a week period, it is likely that women forget their feeding practices of the earlier days, and participants could have been misclassified as exclusively breastfed though that was not the case [6]. Ethiopia saw a rapid primary healthcare expansion in the last few decades. This promoted access of rural and peri urban communities to quality health information including optimal child feeding practices [23]. We presume most of our study participants have the information. This could have caused a desirability bias into denying introducing additional food in the first six months, a practice that they were taught was wrong, increasing the EBF prevalence.

We found out that 24-h recall does not capture the actual breastfeeding practices, at least in a small sample. It is intelligible that prospective studies are likely to be more accurate since it helps us to see infants moving in and out of the feeding categories [24]. A study done comparing single 24 h recalls with seven observation days has found discrepancy on the feeding pattern stating the possibility of capturing the day to day variation increased as the observation of days increase [12]. The consecutive days data help us to see the real picture of infants feeding history in order to categorize exclusive breastfeeding status. However, as longitudinal studies are resource intensive it may not meet the requirements of programmatic evaluations.

Infant feeding patterns are complex as an infant can be exclusively breastfed for a period, receive other food due to a change in circumstances, and then return to exclusive breastfeeding again. This complexity can only be captured either by repeated measurements or by employing a longer recall period. In order to overcome the challenge with overestimation from 24-h recall studies suggest using of recall since birth in addition to 24-h recall [18, 24]. A study exploring methods of measuring exclusive breastfeeding suggested the best approach would be to report indicators based on both point-in-time and life-long data [24]. On the other hand a longitudinal prospective study proposed adding EBF since birth as an indicator since indicators based on 24-h recall period could be inadequate and misleading for many reasons [13]. There are also different studies suggesting the use of recall since birth as an alternative method of assessment for EBF practice [18, 24]. Even though recall since birth determines the feeding practice throughout the infant’s life, there is high possibility of recall bias. However, we can certainly minimize the recall bias if we used a shorter recall period [24]. A study done in South Africa recommended the use of recall no longer than seven days to assess the prevalence EBF practice [21]. Our finding also indicated short recall periods can be an alternative method. However, these recall periods should not be greater than seven days (week recall) since the recall bias increased as we extended the recall period beyond the first seven days. Therefore, by taking improved specificity and feasibility into consideration we recommend a recall over a one-week period as it gives a more accurate estimate of the actual practice.

The major limitation of this study was the high possibility of desirability bias secondary to conducting repeated 24-h recalls. Though we could not completely avoid its effect we tried to minimize it by explaining the objective of the study and changing data collectors each day in order to give a fresh start for interviews every day. Another limitation of this study is not identifying potential confounding factors that might have influence on the practice. The strength of this study is taking repeated measure for consecutive days and itemized check list to aid the short period recalls method and minimize recall bias.

A significant difference in exclusive breastfeeding estimate was observed among the different recall periods used. A week recall gave a comparable estimate to the reference method with optimal specificity and sensitivity. The recall period that is longer than 24-h helps to better capture variability in feeding practices and this point in time measurement optimizes the use of resources. We recommend for future efforts to report EBF using longer recall of 7 days. In addition, researches could also investigate factors that influence exclusive breastfeeding practice using longer recall.