To describe the selection and modification of an appropriate diet score to assess diet quality in early old age.

Cross-sectional analyses of the Boyd Orr cohort – a long-term follow-up of men and women whose families took part in a survey of diet and health in pre-war Britain. Dietary data were obtained from a 113-item food-frequency questionnaire. A nine-item Healthy Diet Indicator (HDI) developed by Huijbregts and colleagues was identified from the literature and modified because some dietary variables were unavailable and to accord more closely with recommendations of the UK Committee on Medical Aspects of Food Policy.

In total, 1475 traced, surviving cohort members aged 60 years and over.

Modification resulted in a 12-item Healthy Diet Score (HDS). We found that about half the variation in the HDS was explained by variation in the HDI (r = 0.71). There was, however, little misclassification of subjects (<10%) into extreme thirds of the distribution by the HDS compared with the HDI. Items of the score most strongly correlated with overall score were saturated fat (r = −0.57), red meat (r = −0.46), dietary fibre (r = 0.58), fruit and vegetables (r = 0.54) and percentage energy from carbohydrates (r = 0.51). Modifying existing items had greater impact on agreement between HDI and HDS than the addition of new items.

The selection and modification of diet scores is more complicated than often assumed. Furthermore, modest changes to an existing score can produce a score that is different from the original, and although it was not possible to test this issue, it may no longer predict subsequent health experience.