Weighting Methods and Their Application in the Dental Practice Panel

The Dental Practice Panel is the cost structure survey of the National Association of Statutory Health Insurance Dentists. Each year, around 38,000 practices are invited to respond. Participation is voluntary, and the response rate had most recently fallen to around nine per cent. Because participating and non-participating practices may differ systematically, the results are weighted. On behalf of the Central Research Institute of Ambulatory Health Care, we examined whether the established procedure still met the requirements in the face of declining participation. A non-response analysis showed that, apart from regional differences, no meaningful self-selection could be detected. We then compared several weighting methods and tested their accuracy against the actual turnover figures known for the population. The two post-stratification methods performed markedly better than unweighted values, whereas weighting by estimated participation probabilities offered no advantage. A simulation also showed that estimates become more precise the more weighting cells can be formed. From this we derived two recommendations: to lower the minimum cell size from six to three practices, and to use different methods at the national and regional level.