QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Robert, R. C. Articles by Black, R. E. Search for Related Content PubMed PubMed Citation Articles by Robert, R. C. Articles by Black, R. E.

---

Community and International Nutrition

Process Evaluation Determines the Pathway of Success for a Health Center–Delivered, Nutrition Education Intervention for Infants in Trujillo, Peru¹

Rebecca C. Robert^*,2, Joel Gittelsohn^*, Hilary M. Creed-Kanashiro, Mary E. Penny, Laura E. Caulfield^*, M. Rocio Narro and Robert E. Black^*

^* Department of International Health, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD 21205 and Instituto de Investigación Nutricional, La Molina, Lima 12, Peru

² To whom correspondence should be addressed. Email: rrobert{at}jhsph.edu.

	ABSTRACT

TOP ABSTRACT SUBJECTS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Process evaluation was used to explain the success of a randomized, controlled trial of an educational intervention to improve the feeding behaviors of caregivers and the nutritional status of infants in Trujillo, Peru. Health personnel delivered a multicomponent intervention within the environment of usual care at government health centers. We created a model of the expected intervention pathway to successful outcomes. Process data were then collected on health center implementation of the intervention and caregiver reception to it. Using multivariate models, we found that variables of health center implementation, caregiver exposure, and caregiver message recall were all significant determinants in the pathway leading to improved feeding behaviors. These outcomes were consistent with our original intervention model. Further support for our model arose from the differences in caregiver reception between intervention and control centers. Process data allowed us to characterize the pathway through which an effective nutrition intervention operated. This study underscores the importance of including process evaluation, which will lead to the development and implementation of more effective nutrition interventions.

KEY WORDS: • process evaluation • implementation • infant and child nutrition • nutrition education • Peru

Childhood malnutrition remains a critical and preventable public health problem throughout the developing world (1,2). In Peru, specific problems include stunting (25% had length for age Z-score below –2SD) and iron deficiency anemia (50%) for children <5 y old (3,4). To improve childhood nutrition, public health education programs commonly target the feeding behaviors of the caregivers of young children. Although outcomes of these programs have been evaluated, limited attention has been given to their implementation and to describing how the interventions actually work to improve caregiver feeding behaviors. Understanding the pathway through which interventions work will lead to improved understanding, development, and practical application of public health nutrition programs (5,6).

Process evaluation involves the use of indicators that reflect how well interventions are delivered and received, and provides data on what, how, why, and for whom intervention programs work (7,8). Results of process evaluation can then provide essential information about how an intervention leads to (or does not lead to) successful behavior change. Although process evaluation research has grown in the last 2 decades, there is a dearth of such research in developing country settings (9,10). This paper uses process data to explain the success of a randomized, controlled trial of an educational intervention to improve the feeding behaviors of caregivers and the nutritional status of infants in Trujillo, Peru (11). In this trial, dietary outcomes were improved and stunting was reduced in a cohort of children followed from birth to 18 mo (n = 187 intervention, n = 190 control). To understand how the program succeeded, a model of the expected intervention pathway was created and process indicators were used to examine intervention implementation, caregiver reception to the intervention, and proximal outcomes.

The intervention

The 2-y cluster randomized, controlled trial (1999–2001) was a joint effort of the Instituto de Investigación Nutritional (IIN)³ and the regional Ministry of Health of La Libertad (MOH), in collaboration with the Johns Hopkins Bloomberg School of Public Health. The intervention trial took place in the large periurban, economically poor areas located around Trujillo, and involved 6 intervention health centers matched with 6 control centers. The overall study design, intervention components, and evaluation design are reported elsewhere (11). The process evaluation described here occurred from 1999 to 2000 and included ongoing measures as well as a community survey at mid-trial.

The intervention built on existing MOH nutrition education and was based on principles of social cognitive theory, formative research, and experience gained from other intervention programs in Peru; it concentrated, in particular, on complementary feeding of infants 6 mo old (12–14) [H. M. Creed-Kanashiro, M. Fukumoto, and M. E. Ugaz, unpublished report (Alimentación infantil: resultados de una campaña educative en la comunidad, 1995, IIN, Lima, Peru)]. The primary goals included raising the profile of nutrition in the health centers and integrating nutrition education more fully into all pediatric services. Essential intervention components to be implemented by health centers included: 1) delivery of age-appropriate messages in all pediatric services, aided by color photo flipcharts and follow-up checking questions to verify caregivers' understanding; 2) infant food preparation demonstrations in which caregivers prepare or observe other caregivers preparing a recipe, and all caregivers and infants taste-test the prepared food; 3) provision of infant food recipe fliers to caregivers attending either the growth and development monitoring or nutrition consultation services; 4) anthropometric assessment, including weight and length measurements of all infants attending the health center and explanation of the infant's growth; 5) growth and development monitoring in a group format in which, after anthropometric assessment, several caregiver-infant pairs with the same age infants actively participate in developmental and play activities and receive nutrition and health advice; and 6) multidisciplinary problem-solving sessions in which health personnel address nutrition concerns in their health center or community.

Caregiver exposure to the promoted feeding behaviors was expected from interactions with health personnel during routine health center visits (e.g., for illness, or monitoring of growth and development) and in the community (e.g., food preparation demonstrations). Knowledge, skills, the confidence to initiate the new behaviors and actual adoption of behaviors were expected to follow intervention exposure.

The message component was central to the intervention and was integrated with the other components. Of the various messages promoted, the following 3 were known as key messages and all health personnel were encouraged to deliver them: 1) "A thick purée satisfies and nourishes your baby, equivalent to 3 portions of soup; at each meal give the thick purée or main dish first"; 2) "Add a special food to your baby's meal: liver, egg, or fish"; 3) "Use love, patience, and good humor when teaching your child to eat."

The first of these 3 messages addressed a particularly strong cultural practice of giving soup to infants, either as the first food in the meal, or as the meal itself. The second message was developed to promote the use of locally available, inexpensive micronutrient-rich foods. The third message was directed at how to feed vs. what to feed in response to commonly expressed frustrations by caregivers about the actual feeding process [R. Villisante, personal communication and unpublished report (Impresiones y resultados preliminaries: proyecto alimentación infantil, 1997, IIN, Lima, Peru)].

Health centers received training and follow-up visits to support implementation of the intervention. A system of accreditation was introduced for health centers to receive public recognition and the title "Pioneer Centers in Infant Nutrition." Standards for implementation, written jointly by the IIN and regional MOH, were set out in a series of criteria and publicized. An outside team evaluated the centers to ascertain accreditation status.

This paper focuses on the following primary research question: through what pathway does the nutrition education intervention work to achieve program success? To address this, we answered the following specific questions, based on our model of the expected intervention pathway: 1) How does health center implementation influence caregiver exposure? 2) How does caregiver exposure influence key message recall? 3) How does key message recall influence initial feeding practices?

A secondary aim was to address the following: immediately after a health center visit, how did caregiver exposure to nutrition education and caregiver efficacy for a key feeding behavior compare between the intervention and control groups?

	SUBJECTS AND METHODS

TOP ABSTRACT SUBJECTS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Development of the process evaluation methodology began by creating a model of the expected intervention pathway leading to improvements in infant nutrition (Fig. 1). Process indicators were then linked to salient points of the model and included the following:

View larger version (15K): [in this window] [in a new window]   FIGURE 1  Model of the intervention pathway including process indicators (shown in italics).

Fidelity. The percentage of intervention activities completed by health personnel that adhered to intervention protocol; a measure related to the quality of the activity with acceptable standards being specified in the accreditation criteria (e.g., message delivery with use of educational material and checking question to verify understanding).

Exposure (dose received). The percentage recall of exposure to intervention activities by caregivers, including attendance at, awareness of, or engagement with any of the intervention components.

Message recall. The percentage recall of key message content by caregivers.

Proximal outcomes. The percentage of key feeding behaviors carried out by caregivers at mid-trial and mean efficacy scores for a key infant feeding behavior.

Two additional process indicators were included in the model: context, a qualitative measure of the environment in which the intervention was introduced and in which facilitating factors for implementation are identified; and barriers to implementation, potential impediments to intervention implementation, both of which are addressed elsewhere (10).

Data collection

Methods used to collect implementation data included record reviews and structured observations. A mid-trial community survey with 24-h dietary recall and exit interviews were used to assess caregiver reception to the intervention and proximal outcomes. A summary of the data collection methods, power calculations, and sample size, as well as the composite score variables created for data analysis from the specific data sources are provided in Table 1. Data collectors were trained and taught to use standardized methods through interactive classroom sessions and practice in a pilot health center/community; they met predesignated competencies and were not involved in delivery of the intervention. All training was conducted by one or more of the authors; several evaluation team members, all healthcare professionals, received additional training to carry out 24-h dietary recall assessments and structured observations. Approval to conduct the study was given by the Ethics Committee of the IIN, Lima, Peru.

    Structured observations of health personnel and caregivers in consultation. Structured observations of the interactions between health personnel and caregivers of infants 0–24 mo old were completed over 1 y at 3 separate time points in the following health services: 1) growth and development monitoring; 2) nutrition consultation; and 3) medical consultation. Observations were event based and each health worker was observed with 5 to 10 caregivers. A structured format consisting of dichotomous and multiple choice items was used to check off specific behaviors (e.g., delivery of recipe flier) and communications. All nutrition recommendations were recorded as well as the use of educational materials and/or checking questions to verify caregiver understanding. For this paper, the observation data used were from consultations with caregivers of infants 6 mo old.

Before each round of data collection, a randomization procedure was used to assign observers to health centers and services. Adjustments were made to ensure that each of the 4 data collectors observed health workers in both the intervention and control groups, as well as a mix of health services. An interrater reliability of at least 93% was achieved among the observers [(number of items agreed upon/total number of items) x 100].

    Mid-trial community survey. In the intervention group, a randomized, cross-sectional, community survey was completed at mid-trial in the homes of caregivers of infants 6–14 mo old. Sample size for the mid-trial community survey was based on census projections for the maximum number of infants available in the smallest community. This projection included at least 50 infants for each month of age (6–14 mo) from the 6 communities for a total of 450 surveys.

Fieldworkers read questions aloud to caregivers and recorded responses with predesignated codes, or written verbatim answers that were later coded according to a priori rules. Recall of messages was assessed through 3 types of questions: 1) spontaneous recall (e.g., "What, if any, nutrition recommendations did you receive?"); 2) prompted recall (e.g., "Have you heard anything about soups or purées?"); and 3) display recall in which a caregiver was asked to recall messages associated with educational materials. Several options of educational materials were shown including both study and nonstudy materials. The caregiver was asked to select which, if any, she had seen, followed by query of the message associated with it.

Twenty-four hour dietary recall assessments were conducted in the home with caregivers of 298 infants in the specific age groups of 7–9 mo and 12–14 mo, a mean of 4.1 ± 5.0 d after completing the mid-trial survey. To help caregivers recall the infant's intake of the previous day, caregivers were asked to show the amount consumed by the child in his/her usual utensils and where possible, for example with drinks, the estimated content was weighed. In addition, ceramic food models of different sizes, utensils, plates, and pictures of different food consistencies were used as prompts. Each food was categorized into 1 of 32 preparation forms (e.g., juice, thin soup, thick purée, or main dish). Using these data, 3 key feeding behaviors, consistent with those listed in the health center accreditation criteria, were assessed: 1) caregiver gave infant a thick consistency (energy dense) food first in the main meal of the day; 2) caregiver gave infant either liver, egg, or fish; and 3) caregiver gave infant at least 2 preparations of thick consistency (energy dense).

    Exit interviews with caregivers. Exit interviews were conducted with caregivers of infants 6–24 mo old in the intervention and control groups as they left the health center. Fieldworkers read out questions to caregivers and recorded responses using predesignated codes or written verbatim answers that were later coded. To measure caregivers' efficacy for infant feeding practices, several questions were asked using a 5-point Likert-type scale (strongly agree to strongly disagree). After each item, caregivers were asked to respond to why they had chosen the answer. Text and Likert scale answers were compared; those obviously discordant as well as caregiver responses of "don't know" were removed from the analysis (2.3% of items).

Sampling for both exit interviews and structured observations included all infants attending the health center on the day chosen for data collection until the required number was reached. Repeat visits were made to the center as necessary. Data were collected on a mix of days. No overlap occurred among exit interviews and structured observations on the same day in the same health center. Health workers were not advised of data collection dates in an attempt to preserve usual care.

Data analysis

Data were entered into Fox Pro 2.6a using purpose-designed programs with range and consistency checks. Double entry of 5–10% of the data was completed to check accuracy. Data were analyzed using SPSS (version 11.5) and STATA (version 8.2). For structured observation data, mean frequencies of observations were calculated for each health worker. Cumulative frequencies were generated for the variables from record reviews. Bivariate analysis included ² for categorical data, and t test or Mann Whitney U test statistics to compare means. Nonparametric correlations were analyzed using Spearmen's r. Factor analysis was used to analyze socioeconomic variables from the mid-trial survey. Two clusters of associated variables were identified, housing and household possessions. Each item in a cluster was given 0 or 1 point with a total score being summed. Reliability analysis calculated a coefficient of 0.64 for housing and 0.68 for household possessions.

Composite score variables were created to characterize each of the following: 1) health center intervention implementation (assessed from structured observations and record reviews); 2) caregiver exposure to the intervention at mid-trial (assessed from mid-trial community survey); 3) caregiver recall of key messages at mid-trial (assessed from mid-trial community survey); and 4) caregiver exposure to nutrition education immediately after a health center visit (assessed from exit interviews) (Table 1). For the first composite score entitled intervention implementation, items represented the main intervention components. Each health center received between 0 and 1 point for an item, reflecting the percentage frequency with which the item was implemented in that particular health center (an average of all observations recorded). Scores were summed for each health center and caregivers were assigned an implementation score according to the health center community in which they resided. An item was not included for health center problem-solving sessions because these did not involve direct patient care. The item for anthropometry/growth explanation was not included because we did not have an accurate measure of caregiver reception to it at mid-trial and wanted these 2 composite scores to be compatible. For the remaining 3 composite scores, measured at the caregiver level, items were scored dichotomously with either 0 or 1, (e.g., attended an infant food demonstration or not) and a total score summed. For the composite score of caregiver exposure after a health center visit, created from exit interview data, an item representing an explanation of children's growth was included. An item representing exposure to an infant food demonstration was not included because this was not a part of daily care (demonstrations were organized as a specific event).

Random effects multivariate logistic regression analyses, to allow for cluster randomization at the health center level, were used to build a series of parsimonious models to explain the following: 1) independent variables associated with caregiver exposure; 2) independent variables associated with caregiver key message recall; 3) independent variables associated with each of 3 key feeding practices. Following the conceptual model of the expected intervention pathway, independent variables of particular interest in the models were as follows: 1) the role of health center implementation in explaining caregiver exposure; 2) caregiver exposure in explaining key message recall; and 3) the role of caregivers' key message recall in explaining caregivers' practice of a promoted feeding behavior. For analyses, a P-value of <0.05 was considered significant.

	RESULTS

TOP ABSTRACT SUBJECTS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
    Mid-trial community survey demographics. The sample for the mid-trial community survey consisted of 476 caregivers in the 6 intervention communities. This represented 65% of eligible households from the census completed in these communities. For the reminder, 1% refused, 5% could not be contacted, and for 29%, a child of the correct age was present but the sample for the child's specific age group had been completed. Demographic data were collected on the infant's mother even if she was not the caregiver interviewed (Table 2).

    Caregiver recall and feeding behaviors. Results from the 24-h dietary recall data demonstrated that 32% of caregivers gave a thick consistency food first in the main meal of the day; 60% of caregivers gave liver, egg, or fish; and 22% caregiver gave at least 2 preparations of thick consistency for infants 7–9 mo. These 3 feeding practice outcomes, dichotomized as feeding practice done or not done, were examined via multivariate analyses. For the first of these outcomes, "gave thick consistency food first in the main meal of the day," adjusted OR were significant for household possessions (OR = 1.24, CI = 1.03, 1.48) and key message recall (OR = 2.16, CI = 1.18, 3.96) (Table 5).

    Caregiver exposure, message recall and efficacy—comparing the intervention and control groups. Characteristics of the caregivers with infants 6–24 mo old completing exit interviews did not differ between the intervention and control centers. This included caregiver age, number of children, level of education, age of infant, and service attended (P > 0.05). Because the intervention built on existent nutrition education in the health centers, caregiver exposure to nutrition education was compared between the 2 groups using the composite score created from exit interview items. The percentage of caregivers reporting any exposure to nutrition education was 77.8% in the intervention group and 49.1% in the control group (² = 19.2, df = 1, P < 0.001). Mean scores (possible 4 points) in the 2 groups were as follows: intervention group = 1.66, control group = 0.59, (t = 7.7 df = 214 P < 0.001). Comparison of caregiver recall (spontaneous and/or prompted) of a key message or message with similar content in the case of the control group, revealed that 47.2% of the intervention group recalled a message compared with 2.8% of the control group (² = 56.9, df = 1, P < 0.001). Correlation analysis between the degree of caregiver exposure and number of key messages recalled was significant (r_s = 0.63, P < 0.001).

Mean scores on 3 items measuring efficacy for a key feeding behavior for the 2 groups are shown in Table 6. Lower scores are equated with greater caregiver confidence. When leaving the health center, intervention group caregivers had significantly greater confidence (outcome efficacy) that thick consistency purées, and not soups, would help an infant to be well nourished. In this connection, the same caregivers tended to have greater confidence for feeding an infant thick consistency purées compared with those in the control group (P = 0.067). Correlation analyses between key message recall and each of the 3 efficacy items demonstrated significance as follows: 1) recall and self-efficacy for feeding thick consistency purées to an infant (r_s = –0.16, P = 0.020); 2) recall and outcome efficacy—feeding soup would not be helpful to nourish the infant (r_s = –0.22, P = 0.001); and 3) recall and outcome efficacy—feeding thick purées would be helpful to nourish the infant (r_s = –0.18, P = 0.011).

	DISCUSSION

TOP ABSTRACT SUBJECTS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Further support for our model was demonstrated in our exit interview data. Although nutrition education was part of existing MOH services, significantly greater caregiver exposure and message recall were found in health centers that had implemented the intervention. Differences between the intervention and control groups were also found in caregiver self-efficacy, her confidence in her ability to carry out a key feeding behavior, and in outcome efficacy, her confidence that the feeding behavior would result in a positive outcome for her child.

A goal of the trial was to determine whether health personnel, within the environment of usual care in the government health centers, could effectively deliver a multicomponent intervention to improve the complementary feeding behaviors of caregivers. To that end, we developed a health center intervention based on behavior change theory and previous experience, and diagrammed how the intervention was expected to work. We then measured important indicators along the pathway. Our study provides an important example of how to develop and apply process evaluation to interventions addressing behavior change to improve child health. Similar steps can be taken to include process evaluation as part of the evaluation strategy for future interventions. In the field of international health, it is particularly critical to understand how successful interventions work to allow the best utilization of limited resources. Process evaluation results provide this understanding.

Implementation was less than desired, as measured by dose (to what extent did the intervention take place) and fidelity (quality or adherence to protocol) markers. Nonetheless, health centers were successful in exposing the majority of caregivers to some aspect(s) of the educational intervention. Detailed examination via multivariate modeling demonstrated that the degree of health center implementation was important in explaining caregiver exposure. This supports the case for developing strategies to improve intervention implementation at the health center level. Examination of the individual components included in the implementation score points out areas in which improvements may begin. The obstacles to health center implementation, identified by health personnel, are reported in detail elsewhere; these should be addressed to improve implementation. Included were factors such as staff turnover and the difficulty in securing caregiver attendance to form groups for group growth and development monitoring. In that same paper, facilitators to health center implementation were discussed, such as use of key messages for all personnel and the positive response from caregivers attending group growth and development monitoring, which should be supported (10).

We chose to use a summary score based on each center's degree of implementation to reflect the overall culture or climate of nutrition in the center. Although with summarization we lost the individual contribution of components, our primary goal was to increase the profile of nutrition education and integrate it into all pediatric services. The use of a composite score at the health center level best reflected these goals.

The next process indicator on the pathway to changing caregiver feeding behaviors was caregiver key message recall. Although a substantial percentage of caregivers could recall one message (70%), a smaller percentage could recall 2 (38%) or 3 (17%). For any intervention, caregivers must be exposed to be able to recall promoted messages correctly; however, exposure to an intervention does not guarantee recall. In our study, we found exposure to be significantly influential in explaining recall. Thus, improving caregivers' exposure to more aspects of the intervention appears beneficial. Examination of the items comprising the composite score showed that a modest percentage of caregivers were exposed to infant food preparation demonstrations (9.2%) and the group format for growth and development monitoring (19.5%), both theoretically powerful methods that used experiential learning and observation to promote learning (including recall) and self-efficacy. In addition, caregiver education was influential in recalling messages, suggesting that caregivers with less education may require more time and reinforcement to promote message recall.

Completing our expected intervention pathway, key message recall was examined in relation to explaining initial caregiver feeding behaviors, considered proximal outcomes in our study. In multivariate analyses, recall helped explain 2 of the 3 feeding behaviors. For the behavior of giving a thick consistency food first at the main meal of the day, this association is impressive considering the strong cultural practice this challenges (giving soup before main dish), and underscores the importance of our formative research in the development of key messages.

Key message recall was also significant in explaining the feeding behavior "caregiver gave at least 2 preparations of thick consistency to infants 7–9 mo." This behavior stands in contrast to the widely accepted cultural belief that young children can choke or develop indigestion from eating foods of thick consistency. The prevalence of giving liver, egg, or fish (in particular, chicken liver) was quite high at baseline, which helps to explain why associations with message recall or other independent variables were not found.

In exit interviews, we were able to measure caregiver efficacy as well as exposure and message recall in both the intervention and control groups. This provided additional data for understanding how the intervention worked, for example, examining the gains in exposure and efficacy compared with that received under usual care in the health centers.

Correlation analyses demonstrated that the association of exposure and recall was highly significant; for recall and efficacy, the association was significant but considerably weaker. This is not surprising. First, efficacy is a much more difficult concept to measure and our items may not have fully captured it. Second, self-efficacy, the confidence to carry out a promoted behavior, is created after a caregiver has the knowledge and skills to carry out the behavior. It is particularly bolstered by experiential and observational learning (for example, during infant food preparation demonstrations in which caregivers participate or observe, or after trial of the behavior at home); therefore, our exit interviews may have captured only early formation of efficacy.

In the international literature, Santos et al. (15) reported significant weight gain for children 12 mo old at study entrance in a randomized, controlled trial of the nutrition component of the "Integrated Management of Childhood Illness" in Pelotas, Brazil. In a substudy, Pelto et al. (6) examined 2 steps in the pathway to help explain how the nutrition component was working to achieve success: physician behavior/implementation of training received, and caregiver retention of the nutrition counseling. Descriptive and comparative analyses of each step demonstrated positive and significant changes in the intervention group physicians and caregivers.

The study had several limitations. The mid-trial community survey did not include control group communities, which would have allowed us to characterize any exposure occurring in the control communities, and compare exposure and feeding practices between the groups. Nonetheless, the focus of interest at mid-trial was on understanding what was occurring in the intervention communities as a result of the intervention. A measure of caregiver feeding behaviors before the intervention was not included, which may have provided additional insight to our results at mid-trial. The study could not be conducted without knowledge of treatment group; however, data collectors were taught to use standardized methods and were periodically rotated to ensure that the same individuals were not always collecting data in the same treatment group. During structured observations, reactivity of health personnel and caregivers to the presence of an observer, measured with ordinal scales, was minimal; however, it is acknowledged that reactivity is not an easy construct to measure, and may have occurred (16).

The mid-trial survey did not contain an item that accurately represented caregiver's exposure to anthropometry/explanation of the child's growth; this would have made the composite score of caregiver exposure more complete. We also eliminated this item from the health center implementation score to retain compatibility between composite scores. Measurements of implementation included structured observations of caregivers of infants 6–24 mo old, whereas the mid-trial community survey included caregivers of infants 6–14 mo old. However, health personnel were trained to use the nutrition messages and materials for infants up to 24 mo; thus, we included these observations in calculating the mean frequency for each health worker.

Sample size was more limited for the dietary recall data; however, the 2 groups chosen (7, 8, and 9 mo and 12, 13, and 14 mo) were selected to represent developmental differences in the infants. Twenty-four hour dietary recall data were collected several hours to several weeks after survey data, which may have intentionally or unintentionally influenced the caregiver's recall of the infant's intake. However, ² analyses of the time between surveys (time categorized in quartiles: 0, 1–2, 3–6, and 7+ d) and each outcome did not differ between the categories, supporting our findings.

Finally, although efficacy measures were included in exit interviews, they were not included in the mid-trial community survey, which would have improved understanding of the link between message recall and caregiver feeding behaviors. Efficacy of the promoted infant feeding behaviors was not an easy construct to measure, and more work is required to improve the ability of its measure, particularly in different cultural contexts.

In conclusion, process evaluation provided valuable data for this effectiveness trial, which helped to explain how the intervention was working to improve caregiver feeding behaviors. Despite less than optimal levels of heath center implementation, caregiver exposure, and caregiver recall, all were significant variables in the pathway. Future studies can learn from our experience to design process evaluations that better characterize the pathways through which interventions operate to achieve outcomes, and lead to the development and implementation of effective programs, a global health priority (17).

	ACKNOWLEDGMENTS

 
We acknowledge and gratefully thank our team of data collectors (Yorli Gómez, Ana García, Betty Hernández, Lucy Hernández, Roxana Paredes, Liliana Quevedo, Monica Rodríguez, Alicia Diaz), Claudia Cuba and Victor Montoya for data management assistance, and the health staff and Trujillo MOH commission (Manuel Burgos, Sonia Castillo, Gladis Rodríguez, Lidia Moreno, Victoria Pinedo).

	FOOTNOTES

 
¹ Funded by the United States Agency for International Development through the Johns Hopkins Family Health and Child Survival Cooperative Agreement and a predoctoral research fellowship (National Research Service Award) awarded to R.C.R. by the National Institute of Nursing Research, National Institutes of Health.

³ Abbreviations used: IIN, Instituto de Investigación Nutritional; MOH, Ministry of Health of La Libertad; OR, odds ratio.

Manuscript received 4 May 2005. Initial review completed 27 July 2005. Revision accepted 4 December 2005.

	LITERATURE CITED

TOP ABSTRACT SUBJECTS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 

1. Fishman S, Caulfield LE, de Onis M. Childhood and maternal underweight. In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. Comparative quantification of health risks: global and regional burden of disease attributable to selected major risk factors. Cambridge, MA: Harvard University Press; 2004. p. 39–61.

2. World Health Organization. Complementary feeding of young children in developing countries: a review of current scientific knowledge. Geneva: World Health Organization; 1998. Report No. WHO/NUT/98.1.

3. National Institute of Statistics and Information. Demographic and health family survey, 2000. Lima, Peru: Instituto Nacional de Estadistica e Informatica; 2003.

4. Lutter CK, Rivera JA. Nutritional status of infants and young children and characteristics of their diets. J Nutr. 2003;133:2941S–9.[Abstract/Free Full Text]

5. Pelto GH, Freake HC. Social research in an integrated science of nutrition: future directions. J Nutr. 2003;133:1231–4.[Abstract/Free Full Text]

6. Pelto GH, Santos I, Goncalves H, Victoria C, Martines J, Habicht J. Nutrition counseling training changes physician behavior and improves caregiver knowledge aquisition. J Nutr. 2004;134:357–62.[Abstract/Free Full Text]

7. Baranowski T, Stables G. Process evaluations of the 5-a-day projects. Health Educ Behav. 2000;27:157–66.[Abstract/Free Full Text]

8. Steckler A, Linnan L. Process evaluation for public health interventions and research: an overview. In: Steckler A, Linnan L, editors. Process evaluation for public health interventions and research. 1st ed. San Francisco: Jossey-Bass; 2002. p. 1–23.

9. Dickey VC, Pachon H, Marsh DR, Lang TT, Claussenius DR, Dearden KA, Ha TT, Schroeder DG. Implementation of nutrition education and rehabilitation programs (NERPs) in Viet Nam. Food Nutr Bull. 2002;23:78–85.[Medline]

10. Robert RC. Process evaluation of a health center delivered, nutrition education intervention to improve infant feeding in Trujillo, Peru. PhD Thesis. Johns Hopkins University, Baltimore, MD; 2004.

11. Penny ME, Creed-Kanashiro HM, Robert RC, Narro R, Caulfield LE, Black RE. Effectiveness of an educational intervention to improve young child nutrition: a cluster-randomised controlled trial. Lancet. 2005;365:1863–72.[Medline]

12. Bandura A. Social foundations of thought and action: a social cognitive theory. Englewood Cliffs: Prentice Hall; 1986.

13. Bandura A. Exercise of personal and collective efficacy in changing societies. In: Bandura A, editor. Self-efficacy in changing societies. Cambridge: Cambridge University Press; 1995. p. 1–45.

14. Creed-Kanashiro H, Uribe T, Bartolini R, Fukumoto M, Lopez TT, Zavaleta N, Bentley ME. Improving dietary intake to prevent anemia in adolescent girls through community kitchens in a peri-urban population of Lima, Peru. J Nutr. 2000;130:459S–61.[Medline]

15. Santos I, Victoria CG, Martines J, Goncalves H, Gigante DP, Valle NJ, Pelto G. Nutrition counseling increases weight gain among Brazilian children. J Nutr. 2001;131:2866–73.[Abstract/Free Full Text]

16. Gittelsohn J, Shankar AV, Ram R, Gnywali T, West KP. Estimating reactivity and its effects in direct observation studies of health behavior. Hum Org. 1997;56:182–9.

17. Caulfield LE, Richard SA, Rivera Donmarco JA, Musgrove P, Black RE. Underweight and micronutrient deficiency disorders. In: Jameson DT, Breman J, Measham A, editors. Disease control priorities in developing countries II. World Bank/Fogarty International Center; 2005.

This article has been cited by other articles:

				R. C. Robert, J. Gittelsohn, H. M. Creed-Kanashiro, M. E. Penny, L. E. Caulfield, M. R. Narro, A. Steckler, and R. E. Black Implementation examined in a health center-delivered, educational intervention that improved infant growth in Trujillo, Peru: successes and challenges Health Educ. Res., June 1, 2007; 22(3): 318 - 331. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Robert, R. C. Articles by Black, R. E. Search for Related Content PubMed PubMed Citation Articles by Robert, R. C. Articles by Black, R. E.