I have a reference in which the authors have applied a three-level multilevel model for binary response:

• Prenatal and Delivery Care and Childhood Immunization in Guatemala: Do Family and Community Matter? Demography, Vol. 33, No. 2., pp. 231-247. http://www.jstor.org/stable/2061874
• Teerenstra, S., Lu, B., Preisser, J.S., van Achterberg, T., & Borm, G.F. (2010). Sample size considerations for GEE analyses of three-level cluster randomized trials. Biometrics, 66(4), 1230-1237. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2896994/pdf/nihms-162479.pdf; http://www.bios.unc.edu/~jpreisse/personal/software.htm
  

• Pituch, K.A., Murphy, D.L., & Tate, R.L. (2010). Three-Level Models for Indirect Effects in School- and Class-Randomized Experiments in Education. The Journal of Experimental Education, 78(1), 60-95.
  

• Lutz, W., Leon, S.C., Martinovich, Z., Lyons, J.S., & Stiles, W.B. (2007). Therapist effects in outpatient psychotherapy: A three-level growth curve approach. Journal of Counseling Psychology, 54(1), 32-39.
• Tasca, G.A., Illing, V., Joyce, A.S., & Ogrodniczuk, J.S. (2009). Three-level multilevel growth models for nested change data: a guide for group treatment researchers. Psychotherapy Research, 19(4-5), 453-461.
• Ai, X. (2002). Gender differences in growth in mathematics achievement: three-Level longitudinal and multilevel analyses of individual, home, and school influences. Mathematical Thinking and Learning, 4(1), 1–22.
• Segawa, E. (2005). A Growth Model for Multilevel Ordinal Data. Journal of Educational and Behavioral Statistics, 30(4), 369-396.
• Raman, R., & Hedeker, D. (2005). A mixed-effects regression model for three-level ordinal response data. Statistics in Medicine, 24, 3331–3345.http://www.uic.edu/classes/bstt/bstt513/RamanStatMed.pdf
• de Jong, K., Moerbeek, M., & Van der Leeden, R. (2009). A priori power analysis in longitudinal three-level multilevel models: An example with therapist effects. Psychotherapy Research. http://www.informaworld.com/smpp/115353608-31391521/content~db=all~content=a917167003
• Preacher, K. J. (2011). Multilevel SEM strategies for evaluating mediation in three-level data. Multivariate Behavioral Research, 46, 691–731. http://quantpsy.org/pubs/preacher_2011.pdf
• http://quantpsy.org/pubs/preacher.2011.online.appendix.051211.pdf
  

We had a paper several years ago that involved patients nested within surgeons nested in hospitals that presented both hierarchical logistic and linear regression models:

Factors Associated with Pattern of Care before Surgery for Breast Cancer in Quebec between 1992 and 1997
Author(s): Ningyan Shen, Nancy E. Mayo, Susan C. Scott, James A. Hanley, Mark S. Goldberg, Michal Abrahamowicz and Robyn Tamblyn
Reviewed work(s):Source: Medical Care, Vol. 41, No. 12 (Dec., 2003), pp. 1353-1366

You might like to take a look at the Centre for Multilevel Modelling's free online multilevel modelling course

    http://www.bristol.ac.uk/cmm/learning/course.html

where we have recently added a 50 page 'concepts' report on multilevel modelling of continuous response three-level data. The illustrative application is a three-level students within school-cohorts within schools analyses.

You will also find there an additional 50 page 'practical' on how to conduct three-level analyses in MLwiN and Stata with other packages to follow. This practical replicates the analyses by Hedeker (1994) who fits a three-level students-within-classes-within-schools model. Data sets and syntax files are provided so that you can replicate all analyses.

thanks for your sharing, xie xie

Dear lisrelchen,
The multilevel structure of the data is not common and there is not a unique methode to nested structure. It depends on the specific question of your research and you cannot adopt a nested order (level 1 units in level two units in level three units) simply from other works. There are many references for your work and i can send you a chapter from an important book which treat this task. If you prefer to discuss of your idea in a private manner you can post your email and i can help you to structure your data.
Hope that helps.