Agreement between self and partner reports of paternal drinking and smoking. The ALSPAC Study Team. Avon Longitudinal Study of Pregnancy and Childhood.

BACKGROUND: We examined agreement between self and proxy reports of paternal drinking and smoking behaviour using data collected as part of the prospective, population-based Avon (England) Longitudinal Study of Pregnancy and Childhood.
METHODS: Information on the smoking and drinking habits of pregnant women's male partners was obtained through self-administered questionnaires completed by pregnant participants and by their partners. For dichotomous indicators (e.g. smoker versus non-smoker), we evaluated self/proxy agreement by calculating Kappa coefficients and per cent agreement. For ordinal measures of smoking and drinking amounts, we calculated per cent perfect agreement, per cent agreement within one category, and Spearman correlation coefficients. Data from 8414 respondent pairs were included in the analyses.
RESULTS: Men's and women's reports of paternal smoking and drinking status were in nearly complete agreement (95% and 98%, respectively). For analyses of smoking and drinking amounts, agreement within one category remained high (90% and 98% for smoking and drinking, respectively), but perfect agreement on amount was somewhat lower (81% and 71%, respectively). Per cent perfect agreement on smoking amount was especially low (50%) when non-smokers were excluded. When couples' reports were not in perfect agreement, women tended to report lower amounts of smoking and drinking for their partners compared to the men's self reports.
CONCLUSIONS: Our results suggest that women's proxy reports of their partners' drinking and smoking status can be used with considerable confidence in reproductive epidemiological studies when the enrollment of both women and men as participants is infeasible for financial or logistical reasons. Caution is warranted, however, when proxy reports are used for more detailed information on smoking and drinking amounts.