Background & Aims
Person centered approach in symptom science enables identification of patient subgroups who experience multiple coexisting symptoms differently. Information on these subgroups is important to inform clinical decisions on targeted assessment and intervention strategies. Two analytical methods for identification of patient subgroups are hierarchical cluster analysis (HCA) and latent profile analysis (LPA). In HCA, individuals with the most similar score patterns are grouped into one cluster. However, in LPA, a probability model-based approach is used for clustering, by determining probabilities that certain individuals are members of certain latent classes/subgroups based on the model. No comparison of HCA and LPA have been done thus far to determine which method is better and more informative in identifying clinically relevant patient subgroups. The aim is to compare patient subgroups generated by HCA and LPA among oral cancer survivors experiencing concurrent symptoms including pain.
Methods
Symptoms were assessed using MDASI-H&N in 300 oral cancer survivors. The top 11 severe symptoms that also occurred in ? 30% of the patients were considered for analyses. Agglomerative HCA was performed with squared Euclidean distances as the dissimilarity measure and weighted average linkage as the clustering method. Two, three, four, and five cluster solutions were obtained on the symptom data. A 2-cluster solution was chosen based on dendrograms, Calinski and Harabasz pseudo-F stopping rule index and the Duda and Hart Je(2)/Je(1) index. Next, LPA was performed, where models with various classes were built across a full range of within-class variance–covariance structures. A 5-profile solution was selected as the best-fit model using BIC, entropy, profile discrimination, model parsimony, and profile size. Data was managed using REDCap and analyzed using STATA 16.0. Concordance between the HCA and LPA solutions was assessed using chi-square and examination of subgroups obtained.
Results
Both HCA and LPA solutions had a significant relationship (?2(4) = 113.04, p = .000). While HCA identified a common Dysphagic subgroup (patients with severe dysphagia), LPA identified three distinct subgroups based on severe dysphagia, where dysphagia co-occurred differently with significantly increased mucus or dry mouth or psychoneurological symptoms. Pain scores differed significantly among the 5-profiles [F = 28.25 (4,295); p = .000].
Conclusions
The number and composition of patient subgroups differ depending on the statistical method used.
LPA enabled better objective evaluation of model fit with a balance of model parsimony and fit and hence, offers more confidence in the patient subgroups identified. LPA provided more nuanced information on the dysphagic group by differentiating it on the basis of dysphagia co-occurring with other symptoms in different profiles.
References
1. Lanza, S. T., & Rhoades, B. L. (2013). Latent class analysis: an alternative perspective on subgroup analysis in prevention and treatment. Prevention science : the official journal of the Society for Prevention Research, 14(2), 157-168. https://doi.org/10.1007/s11121-011-0201-1
2. Miaskowski, C., Barsevick, A., Berger, A., Casagrande, R., Grady, P. A., Jacobsen, P., Kutner, J., Patrick, D., Zimmerman, L., & Xiao, C. (2017). Advancing symptom science through symptom cluster research: Expert panel proceedings and recommendations. JNCI: Journal of the National Cancer Institute, 109(4).
3. Nylund-Gibson, K., & Choi, A. Y. (2018). Ten frequently asked questions about latent class analysis. Translational Issues in Psychological Science, 4(4), 440.
4. Rosenthal, D. I., Mendoza, T. R., Chambers, M. S., Asper, J. A., Gning, I., Kies, M. S., Weber, R. S., Lewin, J. S., Garden, A. S., & Ang, K. K. (2007). Measuring head and neck cancer symptom burden: the development and validation of the MD Anderson symptom inventory, head and neck module. Head & Neck: Journal for the Sciences and Specialties of the Head and Neck, 29(10), 923-931.
5. Spurk, D., Hirschi, A., Wang, M., Valero, D., & Kauffeld, S. (2020). Latent profile analysis: A review and “how to” guide of its application within vocational behavior research. Journal of Vocational Behavior, 120, 103445
6. Tein, J.-Y., Coxe, S., & Cham, H. (2013). Statistical power to detect the correct number of classes in latent profile analysis. Structural equation modeling: a multidisciplinary journal, 20(4), 640-657.
7. Williams, G. A., & Kibowski, F. (2016). Latent class analysis and latent profile analysis. Handbook of methodological approaches to community-based research: Qualitative, quantitative, and mixed methods, 143-151.
Presenting Author
Asha Mathew
Poster Authors
Asha Mathew
PhD
Christian Medical College, Vellore
Lead Author
Mark B. Lockwood
University of Illinois Chicago
Lead Author
Alana Steffen
University of Illinois Chicago
Lead Author
Amit Jiwan Tirkey
Christian Medical College Vellore
Lead Author
Crystal L. Patil
University of Illinois Chicago
Lead Author
Ardith Z. Doorenbos
University of Illinois Chicago
Lead Author
Topics
- Assessment and Diagnosis