Segmenting Patients and Physicians Using Preferences from Discrete Choice Experiments

  • Ken DealEmail author
Practical Application


People often form groups or segments that have similar interests and needs and seek similar benefits from health providers. Health organizations need to understand whether the same health treatments, prevention programs, services, and products should be applied to everyone in the relevant population or whether different treatments need to be provided to each of several segments that are relatively homogeneous internally but heterogeneous among segments. Our objective was to explain the purposes, benefits, and methods of segmentation for health organizations, and to illustrate the process of segmenting health populations based on preference coefficients from a discrete choice conjoint experiment (DCE) using an example study of prevention of cyberbullying among university students. We followed a two-level procedure for investigating segmentation incorporating several methods for forming segments in Level 1 using DCE preference coefficients and testing their quality, reproducibility, and usability by health decision makers. Covariates (demographic, behavioral, lifestyle, and health state variables) were included in Level 2 to further evaluate quality and to support the scoring of large databases and developing typing tools for assigning those in the relevant population, but not in the sample, to the segments. Several segmentation solution candidates were found during the Level 1 analysis, and the relationship of the preference coefficients to the segments was investigated using predictive methods. Those segmentations were tested for their quality and reproducibility and three were found to be very close in quality. While one seemed better than others in the Level 1 analysis, another was very similar in quality and proved ultimately better in predicting segment membership using covariates in Level 2. The two segments in the final solution were profiled for attributes that would support the development and acceptance of cyberbullying prevention programs among university students. Those segments were very different—where one wanted substantial penalties against cyberbullies and were willing to devote time to a prevention program, while the other felt no need to be involved in prevention and wanted only minor penalties. Segmentation recognizes key differences in why patients and physicians prefer different health programs and treatments. A viable segmentation solution may lead to adapting prevention programs and treatments for each targeted segment and/or to educating and communicating to better inform those in each segment of the program/treatment benefits. Segment members’ revealed preferences showing behavioral changes provide the ultimate basis for evaluating the segmentation benefits to the health organization.


Random Forest Latent Class Analysis Adjusted Rand Index Segmentation Strategy Holdout Sample 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.



The author greatly appreciates the generous access provided by Charles E. Cunningham, McMaster University, to the cyberbullying research data. The cyberbullying research (Cunningham et al., personal communication, 2013) was supported by a Community-University Research Alliance grant from the Social Sciences and Humanities Research Council of Canada, the Canadian Institutes of Health Research, the Jack Laidlaw Chair in Patient-Centred Health Care held by Dr. Charles E. Cunningham, and a Canada Research Chair from the Canadian Institutes of Health Research held by Dr. Tracy Vaillancourt. There were no conflicts of interest.


  1. 1.
    Bensing J. Bridging the gap, the separate worlds of evidence-based medicine and patient-centered medicine. Patient Educ Couns. 2000;39:17–15.Google Scholar
  2. 2.
    Lerer L. Pharmaceutical marketing segmentation in the age of the internet. Int J Med Mark 2. 2002;2(2):159–66.CrossRefGoogle Scholar
  3. 3.
    Bassi F. Latent class factor models for market segmentation: an application to pharmaceuticals. Stat Methods Appl. 2007;16:270–87.CrossRefGoogle Scholar
  4. 4.
    Vaughn S, Sarianne S. Examining physician segments. Pharm Represent. 2009;39(4):12–5.Google Scholar
  5. 5.
    American Marketing Association. The American Marketing Association releases new definition for marketing. Accessed 13 Nov 2013.
  6. 6.
    Andreasen AR. Redesigning the marketing universe. Keynote address, World Marketing Summit, Dhaka, 2 Mar 2012.Google Scholar
  7. 7.
    Levitt T. The marketing imagination. New York: The Free Press; 1983.Google Scholar
  8. 8.
    Smith W. Product differentiation and market segmentation as alternative marketing strategies. J Mark. 1956;21:3–8.CrossRefGoogle Scholar
  9. 9.
    Greengrove K. Needs-based segmentation: principles and practice. Int J Mark Res. 2002;44(4):405–21.Google Scholar
  10. 10.
    Ferrandiz J. The impact of generic goods in the pharmaceutical industry. Health Econ. 1999;8(7):599–612.PubMedCrossRefGoogle Scholar
  11. 11.
    Cunningham C, Deal K, Chen Y. Adaptive choice-based conjoint analysis: a new patient-centered approach to the assessment of health service preferences. Patient. 2010;3(4):257–73.PubMedCentralPubMedCrossRefGoogle Scholar
  12. 12.
    Cunningham C, Deal K, Neville A, Miller H, Lohfeld L. Modeling the problem-based learning preferences of McMaster University undergraduate medical students using a discrete choice conjoint experiment. Adv Health Sci Educ Theory Pract. 2006;3(2):245–66.CrossRefGoogle Scholar
  13. 13.
    Cunningham C, Vaillancourt T, Rimas H, Deal K, Cunningham L, Short K, Chen Y. Modeling the bullying prevention program preferences of educators: a discrete choice conjoint experiment. J Abnorm Child Psychol. 2009;37(7):929–43. doi: 10.1007/s10802-009-9324-2.PubMedCrossRefGoogle Scholar
  14. 14.
    Yin Y, Zhang X, Williams R, et al. LOGISMOS—Layered optimal graph image segmentation of multiple objects and surfaces: cartilage segmentation in the knee joint. IEEE Trans Med Imag. 2010;29(12):2023–37.CrossRefGoogle Scholar
  15. 15.
    Schaap M, van Walsum T, Neefjes L, et al. Robust shape regression for supervised vessel segmentation and its application to coronary segmentation in CTA. IEEE Trans Med Imag. 2010;30(11):1974–86.CrossRefGoogle Scholar
  16. 16.
    Van Gerven MA, Jurgelenaite R, Taal BG, et al. Predicting carcinoid heart disease with noisy-threshold classifier. Artif Intell Med. 2007;40(1):45–55.PubMedCrossRefGoogle Scholar
  17. 17.
    Giuly RJ, Martone M, Ellisman M. Method: automatic segmentation of mitochondria utilizing patch classification, contour pair classification, and automatically seeded level sets. BMC Bioinform. 2012;13:29.CrossRefGoogle Scholar
  18. 18.
    Dolnicar S, Lazarevski K. Methodological reasons for the theory/practice divide in market segmentation. J Mark Manag. 2009;25(3–4):357–73.CrossRefGoogle Scholar
  19. 19.
    Aldenderfer M, Blashfield R. Cluster analysis. Newbury Park: Sage Publications; 1984.Google Scholar
  20. 20.
    Everitt BS. Unresolved problems in cluster analysis. Biometrics. 1979;35:169–82.CrossRefGoogle Scholar
  21. 21.
    Aaker D. Developing business strategies. 5th ed. New York: Wiley; 1998. p. 47.Google Scholar
  22. 22.
    Vermunt J. Latent class modeling with covariates: two improved three-step approaches. Polit Anal. 2010;18:450–69.CrossRefGoogle Scholar
  23. 23.
    Dolnicar S, Leisch F. Evaluation of structure and reproducibility of cluster solutions using the bootstrap. Market Lett. 2010;21:83–101.CrossRefGoogle Scholar
  24. 24.
    Retzer J, Shan M. Cluster ensemble analysis and graphical depiction of cluster partitions. Proceedings of the 2007 Sawtooth Software Conference, Sequim (WA); 2007.Google Scholar
  25. 25.
    Williams G. Data mining with Rattle and R: the art of excavating data for knowledge discovery. New York: Springer Science+Business Media; 2011.CrossRefGoogle Scholar
  26. 26.
    Orme B. Getting started with conjoint analysis: strategies for product design and pricing research. Madison: Research Publishers LLC: p. 65.Google Scholar
  27. 27.
    Breiman L. Random forests. Mach Learn. 2001;45(5–3):2.Google Scholar
  28. 28.
    Arabie P, Hubert L. Cluster analysis in marketing research. In: Bagozzi R, editor. Advanced methods of marketing research, Cambridge: Blackwell; 1994. p. 160–189.Google Scholar
  29. 29.
    Rousseeuw P. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Comput Appl Math. 1986;20:53–65. doi: 10.1016/0377-0427(87)90125-7.CrossRefGoogle Scholar
  30. 30.
    Calinski T, Habarasz J. A dendrite method for cluster analysis. Commun Stat. 1974;3:1–17.CrossRefGoogle Scholar
  31. 31.
    Schwartz G. Estimating the dimension of a model. Ann Stat. 1978;6:461–4.CrossRefGoogle Scholar
  32. 32.
    Akaike, H. Information theory as an extension of the maximum likelihood principle. In: Petrov BN, Csaki F, editors. Second international symposium on information theory. Budapest: Akademiai Kiado; 1973. p. 267–8.Google Scholar
  33. 33.
    Bozdogan H. Model selection and Akaike’s information criterion (AIC): the general theory and its analytical extensions. Psychometrika. 1987;52:345–70.CrossRefGoogle Scholar
  34. 34.
    Sugiura N. Further analysis of the data by Akaike’s information criterion and the finite corrections. Commun Stat Theory Methods. 1978;A7:13–26.CrossRefGoogle Scholar
  35. 35.
    Banfield JD, Raftery AE. Model-based gaussian and non-gaussian clustering. Biometrics. 1993;49:803–21.CrossRefGoogle Scholar
  36. 36.
    Rand WM. Objective criteria for the evaluation of clustering methods. J Am Stat Assoc. 1971;66:846–50.CrossRefGoogle Scholar
  37. 37.
    Hubert L, Arabie P. Comparing partitions. J Class. 1985;2:193–218.CrossRefGoogle Scholar
  38. 38.
    Morey LC, Agresti A. An adjustment to the rand statistic for chance agreement. Classif Soc Bull. 1981;5:9–10.Google Scholar
  39. 39.
    Fowlkes EB, Mallows CL. A method for comparing two hierarchical clusterings. J Am Stat Assoc. 1983;78(383):553–69.CrossRefGoogle Scholar
  40. 40.
    Hultsch L. Untersuchung zur Besiedlung einer Sprengfläche im Pockautal durch die Tiergruppen Heteroptera (Wanzen) und Auchenorrhyncha (Zikaden).Google Scholar
  41. 41.
    Krieger AM, Green PE. A generalized Rand-index method for consensus clustering of separate partitions of the same data base. J Classif. 1999;16:63–89.CrossRefGoogle Scholar
  42. 42.
    Zweig M, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem. 1993;39(4):561–77.PubMedGoogle Scholar
  43. 43.
    Vuk M, Curk T. ROC curve, lift chart and calibration plot. Metodoloski zvezki. 2006;3(1):89–108.Google Scholar
  44. 44.
    Goodman LA. The analysis of systems of qualitative variables when some of the variables are unobservable. Part I: a modified latent structure approach. Am J Sociol. 1974;79:1179–259.CrossRefGoogle Scholar
  45. 45.
    Magidson J, Vermunt JK. Latent class factor and cluster models, bi-plots and related graphical displays. Sociol Methodol. 2001;31:223–64.CrossRefGoogle Scholar
  46. 46.
    The CBC Latent Class Technical Paper. Version 3. Sawtooth Software Technical Paper Series, 2004.Google Scholar
  47. 47.
    Latent Class v4.5, Sawtooth Software Inc., 26 Sep 2012.Google Scholar
  48. 48.
    Vermunt JK, Magidson J. Latent Gold Choice 4.0 user’s guide. Statistical Innovations; 2005.Google Scholar
  49. 49.
    Allenby G, Arora N, Ginter J. On the heterogeneity of demand. J Market Res. 1998;35:384–9.CrossRefGoogle Scholar
  50. 50.
    Rossi P, Allenby G, McCullough R. Bayesian statistics and marketing. New York: Wiley; 2005.CrossRefGoogle Scholar
  51. 51.
    Revelt D, Train K. Mixed logit with repeated choices: households’ choices of appliance efficiency level. Rev Econ Stat. 1998;30(4):647–57.CrossRefGoogle Scholar
  52. 52.
    Johnson FR, Mansfield C. Survey design and analytical strategies for better healthcare stated-choice studies. The Patient. 2008;1(4):299–307.PubMedCrossRefGoogle Scholar
  53. 53.
    MacQueen JB. Some methods for classification and analysis of multivariate observations. In: Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability. Vol. 1. University of California Press; 1967: p. 281–297.Google Scholar
  54. 54.
    Witek E. Comparison of model-based clustering with heuristic clustering methods. Folia Oeconomica. 2011;255:191–7.Google Scholar
  55. 55.
    Wang X, Qiu W, Zamar RH. CLUES: a non-parametric clustering method based on shrinking. Comput Stat Data Anal. 2007;52(1):286–98.CrossRefGoogle Scholar
  56. 56.
    Chang F, Qiu W, Zamar RH, Lazarus R, Wang X. Clues: An R package for nonparametric clustering based on local shrinking. J Stat Softw. 2010;33:4.Google Scholar
  57. 57.
    Kaufman L, Rousseeuw PJ. Finding groups in data. New York: Wiley; 2005.Google Scholar
  58. 58.
    R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2013.
  59. 59.
    Magidson, J. SI-CHAID 4.0 user’s guide. Statistical Innovations; 2005.Google Scholar
  60. 60.
    Retzer J, Shan M. Cluster ensemble analysis and graphical depiction of cluster partitions. In: Proceedings of the 2007 Sawtooth Software Conference, Sequim (WA); 2007.Google Scholar
  61. 61.
    Strehl A, Ghosh J. Cluster ensembles: a knowledge reuse framework for combining multiple partitions. J Mach Learn Res. 2002;3:583–617.Google Scholar
  62. 62.
    Orme B, Johnson R. Improving K-means cluster analysis: ensemble analysis instead of highest reproducibility replicates. Sawtooth Software Research Paper Series; 2008.Google Scholar
  63. 63.
    Arseneault L, Walsh E, Trzesniewski K, Newcombe R, Caspi A, Moffitt TE. Bullying victimization uniquely contributes to adjustment problems in young children: a nationally representative cohort study. Pediatrics. 2006;118(1):130–8. doi: 10.1542/peds.2005-2388.PubMedCrossRefGoogle Scholar
  64. 64.
    Arseneault L, Bowes L, Shakoor S. Bullying victimization in youths and mental health problems: ‘Much ado about nothing’? Psychol Med. 2010;40:717–29.PubMedCrossRefGoogle Scholar
  65. 65.
    Kim YS, Leventhal BL, Koh YJ, Hubbard A, Boyce WT. School bullying and youth violence: causes or consequences of psychopathologic behavior? Arch Gen Psychiatry. 2006;63(9):1035–41. doi: 10.1001/archpsyc.63.9.1035.PubMedCrossRefGoogle Scholar
  66. 66.
  67. 67.
    Bridges JFP, Hauber AB, Marshall D, et al. Conjoint analysis applications in health—a checklist: a report of the ISPOR good research practices for conjoint analysis task force. Value Health. 2011;14:403–13.PubMedCrossRefGoogle Scholar
  68. 68.
    Chen C, Liaw A, Breiman L. Using random forest to learn imbalanced data. UC Berkeley: Department of Statistics; 2004.Google Scholar
  69. 69.
    Svetnik V, Liaw A, Tong C, et al. Random forest: a classification and regression tool for compound classification and QSAR modeling. J Chem Inf Comput Sci. 2003;43:1947–58.PubMedCrossRefGoogle Scholar
  70. 70.
    Liaw A, Wiener M. Classification and regression by randomForest. R News. 2002;2(3):18–22.Google Scholar
  71. 71.
    Haley R. Benefit segmentation: a decision-oriented research tool. J Mark. 1968;32(3):30–5.CrossRefGoogle Scholar
  72. 72.
    Zapert K, Spears D. Reengineering a US-based diabetes patient segmentation for Japan: lost in translation. Presented at 2011 Annual National Conference of the Pharmaceutical Marketing Research group; 2011.Google Scholar
  73. 73.
    Bogle A, Simpson SL, Mills TM. Segmentations that work. First Annual Meeting of the Pharmaceutical Marketing Research Group; 2007.Google Scholar
  74. 74.
    Ross C, Steward CA, Sinacore JM. The importance of patient preferences in the measurement of health care satisfaction. Med Care. 1993;31(12):1138–49.PubMedCrossRefGoogle Scholar
  75. 75.
    Magidson J, Eagle T, Vermunt JK. New developments in latent class choice models. In: Sawtooth Software Conference Proceedings; 2003: p. 89–112.Google Scholar
  76. 76.
    The CBC Latent Class Technical Paper. Version 3. Sawtooth Software Technical Paper Series; 2004.Google Scholar
  77. 77.
    McCullough PR. Comparing hierarchical Bayes and latent class choice: practical issues for sparse data sets. In: 2009 Sawtooth Software Conference Proceedings, Delray Beach (FL); Mar 2009.Google Scholar

Copyright information

© Springer International Publishing Switzerland 2013

Authors and Affiliations

  1. 1.McMaster UniversityHamiltonCanada

Personalised recommendations