Validation of a measurement instrument for parental child feeding in a low and middle-income country

Background: Parental child feeding practices (PCFP) are a key factor influencing children's dietary intake, especially in the preschool years when eating behavior is being established. Instruments to measure PCFP have been developed and validated in high-income countries with a high prevalence of childhood obesity. The aim of this study was to test the appropriateness, content, and construct validity of selected measures of PCFP in a low and middle-income country (LMIC) in which there is both undernutrition and obesity in children.

Methods: An expert panel selected subscales and items from measures of PCFP that have been well-tested in high-income countries to measure both "coercive" and "structural" behaviors. Two sequential cross-sectional studies (Study 1, n = 154; Study 2, n = 238) were conducted in two provinces in Indonesia. Findings of the first study were used to refine subscales used in Study 2. An additional qualitative study tested content validity from the perspective of mothers (the intended respondents). Factorial validation and reliability were also tested. Convergent validity was tested with child nutritional status.

Results: In Study 1, a confirmatory factor analysis (CFA) model with 11 factors provided good fit (RMSEA = 0.045; CFI = 0.95 and TLI = 0.95) after two subscales were removed. Reliability was good among seven of the subscales. Following a decision to take out an additional subscale, the instrument was tested for factorial validity (Study 2). A CFA model with 10 subscales provided good fit (RMSEA = 0.03; CFI = 0.92 and TLI = 0.90). The reliability of subscales was lower than in Study 1. Convergent validity with nutrition status was found with two subscales.

Conclusions: The two studies provide evidence of acceptable psychometric properties for 10 subscales from tested instruments to measure PCFP in Indonesia. This provides the first evidence of the validity of these measures in a LMIC setting. Some shortcomings, such in the reliability of some subscales and further tests of predictive validity, require further investigation.