Missing portion sizes in FFQ: alternatives to use of standard portions

TY - JOUR

T1 - Missing portion sizes in FFQ

T2 - alternatives to use of standard portions

AU - Køster-Rasmussen, Rasmus

AU - Siersma, Volkert Dirk

AU - Halldorson, Thorhallur I.

AU - Olivarius, Niels de Fine

AU - Henriksen, Jan E.

AU - Heitmann, Berit

PY - 2015/8

Y1 - 2015/8

N2 - Objective: Standard portions or substitution of missing portion sizes with medians may generate bias when quantifying the dietary intake from FFQ. The present study compared four different methods to include portion sizes in FFQ.Design: We evaluated three stochastic methods for imputation of portion sizesbased on information about anthropometry, sex, physical activity and age. Energyintakes computed with standard portion sizes, defined as sex-specific medians(median), or with portion sizes estimated with multinomial logistic regression(MLR), ‘comparable categories’ (Coca) or k-nearest neighbours (KNN) werecompared with a reference based on self-reported portion sizes (quantified by aphotographic food atlas embedded in the FFQ).Setting: The Danish Health Examination Survey 2007–2008.Subjects: The study included 3728 adults with complete portion size data.Results: Compared with the reference, the root-mean-square errors of the meandaily total energy intake (in kJ) computed with portion sizes estimated by the fourmethods were (men; women): median (1118; 1061), MLR (1060; 1051), Coca(1230; 1146), KNN (1281; 1181). The equivalent biases (mean error) were (in kJ):median (579; 469), MLR (248; 178), Coca (234; 188), KNN (−340; 218).Conclusions: The methods MLR and Coca provided the best agreement with thereference. The stochastic methods allowed for estimation of meaningful portionsizes by conditioning on information about physiology and they were suitable formultiple imputation. We propose to use MLR or Coca to substitute missing portion size values or when portion sizes needs to be included in FFQ without portion size data.

AB - Objective: Standard portions or substitution of missing portion sizes with medians may generate bias when quantifying the dietary intake from FFQ. The present study compared four different methods to include portion sizes in FFQ.Design: We evaluated three stochastic methods for imputation of portion sizesbased on information about anthropometry, sex, physical activity and age. Energyintakes computed with standard portion sizes, defined as sex-specific medians(median), or with portion sizes estimated with multinomial logistic regression(MLR), ‘comparable categories’ (Coca) or k-nearest neighbours (KNN) werecompared with a reference based on self-reported portion sizes (quantified by aphotographic food atlas embedded in the FFQ).Setting: The Danish Health Examination Survey 2007–2008.Subjects: The study included 3728 adults with complete portion size data.Results: Compared with the reference, the root-mean-square errors of the meandaily total energy intake (in kJ) computed with portion sizes estimated by the fourmethods were (men; women): median (1118; 1061), MLR (1060; 1051), Coca(1230; 1146), KNN (1281; 1181). The equivalent biases (mean error) were (in kJ):median (579; 469), MLR (248; 178), Coca (234; 188), KNN (−340; 218).Conclusions: The methods MLR and Coca provided the best agreement with thereference. The stochastic methods allowed for estimation of meaningful portionsizes by conditioning on information about physiology and they were suitable formultiple imputation. We propose to use MLR or Coca to substitute missing portion size values or when portion sizes needs to be included in FFQ without portion size data.

KW - semrap-2014-3

U2 - 10.1017/s1368980014002389

DO - 10.1017/s1368980014002389

M3 - Journal article

C2 - 25382388

SN - 1368-9800

VL - 18

SP - 1914

EP - 1921

JO - Public Health Nutrition

JF - Public Health Nutrition

IS - 11

ER -