Helping a healthcare brand assess the market using Latent Class Segmentation
Scientists around the world have made advances targeted towards improving the body’s immune system for the treatment of various types of cancer. These advancements have led to new therapies. This survey we executed focused on the treatment of cancers, across different tumour types. The purpose of this research was to understand how the patients with the listed conditions were treated.
The specific objectives of the study were:
- To identify the drivers of prescribing behavior and how they may differ among different segments.
- To identify patient’s differential needs, symptoms, psychographics, and treatments, for each segment.
- To profile the segments on demographics, behaviors, attitudes, and future intent to prescribe drug.
- To understand HCP attitudes and behaviors, and how they differ.
- To understand the inherent characteristics of the market to capture and grow share for the client’s drug.
A quantitative survey was conducted for physicians treating cancers in US. The study covered patients’ demographics, current treatment, attitudes towards Immunotherapy, future treatment, sources of learning and practice characteristics, thoughts of others (e.g. peers or specialists), external controls (for example, guidelines), and finally, the therapy of reasoned action and planned behavior (TPB).
The objective was to optimise drug targeting and messaging strategy given its expected entry into a crowded market of immuno-oncology agents by identifying key segments in the market. The challenge was to apply an analytical method that could handle different types of variables measured on different scales.
We opted for a model-based Latent Class Segmentation, which is a finite mixed probabilistic model. The technique uses probability modelling to maximise the overall fit of the model to the data. The model identified patterns across multiple variables (for example, attitudes and needs) and quantified correlation of variables with related variables (for example, buying behaviours). It handles different types of variables, unlike the traditional segmentation where the input variables are of continuous or scale. Moreover, as the study involved mixed variables, traditional approach fails as these depends on distance measure. This scenario is best explained by latent class technique, as it is based on probability.
For each survey respondent, the analysis delivers the probability of belonging to each cluster (segment). Respondents are assigned to the cluster to which they have the highest probability of belonging.1 We iterated over couple of iterations before finalizing the model. Based on the statistics, the optimal number of clusters were selected.
The findings of the survey helped the client to understand the profiles of HCPs who are most likely to try and adopt the drug by assessing their prescribing behaviour, practice characteristics, demographics, among other factors. It highlighted the drivers that had an impact on the prescribing behaviours and were instrumental in developing the promotional material and strategies for marketing the drug to specific segments.