PURPOSE: Field substitution of nonrespondents can be used to maintain the planned sample size and structure in surveys but may introduce additional bias. Sample weighting is suggested as the preferable alternative; however, limited empirical evidence exists comparing the two methods. We wanted to assess the impact of substitution on surveillance results using data from Progressi delle Aziende Sanitarie per la Salute in Italia-Progress by Local Health Units towards a Healthier Italy (PASSI). METHODS: PASSI is conducted by Local Health Units (LHUs) through telephone interviews of stratified random samples of residents. Nonrespondents are replaced with substitutes randomly preselected in the same LHU stratum. We compared the weighted estimates obtained in the original PASSI sample (used as a reference) and in the substitutes' sample. The differences were evaluated using a Wald test. RESULTS: In 2011, 50,697 units were selected: 37,252 were from the original sample and 13,445 were substitutes; 37,162 persons were interviewed. The initially planned size and demographic composition were restored. No significant differences in the estimates between the original and the substitutes' sample were found. CONCLUSIONS: In our experience, field substitution is an acceptable method for dealing with nonresponse, maintaining the characteristics of the original sample without affecting the results. This evidence can support appropriate decisions about planning and implementing a surveillance system.
PURPOSE: Field substitution of nonrespondents can be used to maintain the planned sample size and structure in surveys but may introduce additional bias. Sample weighting is suggested as the preferable alternative; however, limited empirical evidence exists comparing the two methods. We wanted to assess the impact of substitution on surveillance results using data from Progressi delle Aziende Sanitarie per la Salute in Italia-Progress by Local Health Units towards a Healthier Italy (PASSI). METHODS: PASSI is conducted by Local Health Units (LHUs) through telephone interviews of stratified random samples of residents. Nonrespondents are replaced with substitutes randomly preselected in the same LHU stratum. We compared the weighted estimates obtained in the original PASSI sample (used as a reference) and in the substitutes' sample. The differences were evaluated using a Wald test. RESULTS: In 2011, 50,697 units were selected: 37,252 were from the original sample and 13,445 were substitutes; 37,162 persons were interviewed. The initially planned size and demographic composition were restored. No significant differences in the estimates between the original and the substitutes' sample were found. CONCLUSIONS: In our experience, field substitution is an acceptable method for dealing with nonresponse, maintaining the characteristics of the original sample without affecting the results. This evidence can support appropriate decisions about planning and implementing a surveillance system.
Authors: Stefaan Demarest; Geert Molenberghs; Johan Van der Heyden; Lydia Gisle; Herman Van Oyen; Sandrine de Waleffe; Guido Van Hal Journal: Int J Public Health Date: 2017-05-03 Impact factor: 3.380
Authors: Patrizio Zanobini; Chiara Lorini; Vieri Lastrucci; Valentina Minardi; Valentina Possenti; Maria Masocco; Giorgio Garofalo; Giovanna Mereu; Guglielmo Bonaccorsi Journal: Int J Environ Res Public Health Date: 2021-11-26 Impact factor: 3.390