Dental labour force collections attempt complete enumeration but response rates vary from as low as 45% to over 90%. In these collections labour force participation (LFP) of non-respondents must be imputed. If respondents to the survey differ in their LFP from non-respondents, biased estimates may be obtained. The aims of this study were: 1) to investigate the extent of non-response bias in the collection; and 2) to determine the impact on estimates of LFP of an imputation method that adjusts for non-response bias. Methods: Data from the dental labour force collections in SA, WA and Vic for the years 2001 and 2002 were examined (2001: n=4420, response rate 73.9% and 2002: n=4529, response rate 88.5%). For Aim 1, we evaluated whether non-response to the 2001 survey was associated with differences in LFP by the same dentists who subsequently responded to a 2002 survey. For Aim 2, we computed ratios of response bias for each of 18 strata defined by sex age group (5 year categories). Stratum specific ratios were used to impute LFP of non-respondents to the 2001 survey. Results: 4008 dentists responded to the 2002 survey, of which 19.7% (790) had been non-respondents to the 2001 survey. Dentists who were non-respondents in 2001 but who responded in 2002 reported a lower LFP (84.1% in 2002) than those dentists who had responded in both surveys (90.5% in 2002). Within the 18 sex and age group strata, LFP rates differed from 0.3% to 17.5% (median = 7.8%) between 2001 respondents and 2001 non-respondents and were significantly different (P<0.05) in only 4 strata. Using stratum specific response bias ratios to impute for 2001 non-response, the estimated LFP in 2001 was 87.5%, which was only marginally lower than the default (average-imputed) LFP of 89.1% (95% CI=88.0% to 90.2%). Conclusions: While non-response bias cannot conclusively be ruled out, alternative methods of imputing practising rates for non-respondents did not result in substantial differences from standard methods that assume no bias.
DN Teusner*, AJ Spencer, GD Slade
Presented at the 43rd Annual Meeting of the IADR (ANZ Division), 21 September – 1 October 2003, Melbourne, Australia
Note: * indicates presenter