11478 McKenzie Parkway Carmel, IN 46032 317 7060321

1
ADVANCED METHODS CAPABILITIES PRESENTATION
11478 McKenzie Parkway
Carmel, IN 46032
[phone] www.abr-usa.com All materials
presented remain exclusively the property of
Advanced Business Research (ABR) and may not be
distributed (paper or electronic) to anyone else
without prior written approval from ABR. If these
materials are shown to a third party, ABR needs
to be identified as the source and the author.
2
Pharmaceutical Segmentation Applications

• Patient
  Physician
• Segmentation
  Segmentation
• ? Segment patient-market through the eyes
  ? Segment physicians based on
  their Rx Choice behavior,
• of the physician.
  
  treatment needs and preferences.
• ? Segments reflect homogeneous Patient
  ? Incorporate
  attitudes/perceptions influencing Choice of Rx
• types that physicians need to treat.
  via Structural
  Equations.

Descriptive Predictive
Behavioral
Behavioral/Media/Demo

• Symptoms
• Concomitant Cond
• Diagnoses
• Drugs
• Demographics
• Approaches
• Mixture Model
• Normative Segmentation

Group patients with similar treatment
patterns and therapies. Modeling
Approach Restricted Choice Set Model of Drug
Competition Gupta Sarkar (2003)
Group physicians with similar treatment
preferences. ? What they prescribe? ? Why they
prescribe? ? Who they prescribe? ? When they
prescribe? Modeling Approach Structural
Equation Latent Class Discrete Choice Model
Optimize physician targets for Product Launch
based on ? Treatment Needs/Preferences ?
Media Consumption ? Demographics ? Sales Force
Constraints Modeling Approach Mathematical
Optimization with Structural Equation Latent
Class Discrete Choice Model
3
DESCRIPTIVE PATIENT SEGMENTATION (Normative
Segmentation)
AIM Find patient-types
in Pain-with-Depression based on Symptoms,
Diagnoses, Treatment and Demographics. MODEL
INPUTS Patient charts filled out by
physicians. DATA SOURCES Patient
records. METHODOLOGY Mixture Model
Segment 9
Target Segment 1 Current Rx profile
Vicodin, Vioxx, Celebrex, Zoloft,
Paxil Consider Brand X 40 Consider Brand Y 33
Arthritic/Joint Pain
Classic Migraine
I
Target Segment 2 Easy to treat Consider Brand X
9 Consider Brand Y 10
Segment 6
II
V
IV
Target Segment 4 Current Rx profile Vioxx,
Celebrex, Neurontin, Vicodin Celexa, Flexiril,
Xanax Consider Brand X 35 Consider Brand Y 46
Mild Chronic Pain
Target Segment 3 Current Rx profile Vioxx,
Motrin, Paxil, Flexiril, Effexor Consider Brand
X 75 Consider Brand Y 60
OUTPUT Nine distinct patient types in
Pain-Depression market as seen through the eyes
of the physician four targets identified.
4
PREDICTIVE PATIENT SEGMENTATION Restricted Choice
Set Model of Drug Competition Gupta and Sarkar
(2003)
AIM Segment patients
based on similarities in patterns of treatment of
Dyslipidemia. MODEL INPUTS Experimental
Design Discrete Choice. DATA SOURCES
Physician Survey METHODOLOGY Restricted Choice
Set Model of Drug Competition
Patient Types
Brands
OUTPUT Six patient types with distinct
treatment patterns. Probability that a certain
brand is considered as a treatment option for a
patient type is shown in the PI-matrix.
5
BEHAVIORAL PHYSICIAN SEGMENTATION (ONCOLOGISTS)
AIM Segment physicians
based on differences in patterns of treatment of
Colorectal Cancer. MODEL INPUTS Experimental
Design Discrete Choice Attitudes towards
Treatment. DATA SOURCES Physician
Survey METHODOLOGY Latent Class Discrete Choice
Analysis
OUTPUT FOUR behaviorally distinct physician
segments with messaging implications for each.
6
BEHAVIORAL/MEDIA/DEMOGRAPHIC PHYSICIAN
SEGMENTATION (Targeting at Product Launch)
AIM Optimize physician
targets based on Treatment preferences, Media
consumption and Demographics. MODEL INPUTS
Experimental Design Discrete Choice Attitudes
towards Treatment Promotion preferences. DATA
SOURCES Physician Survey METHODOLOGY
Mathematical optimization with Structural
Equation Latent Class Discrete Choice Analysis
Primary Care Physicians 96 Proactive Treatment
Attitudes Early Adopters Low Prescribers of
X Prefer activities with large crowds
Urologists 95 Proactive Treatment
Attitudes Early Adopters, Feel in Control High
Prescribers of X Prefer activities with large
crowds
I
II
IV
III
Primary Care Physicians 98 Passive Treatment
Attitudes Late Adopters Low Prescribers of
X Prefer family-oriented activities
Urologists 67 Primary Care 33 Passive Treatment
Attitudes Feel less in Control Mid-High
Prescribers of X Prefer family-oriented activities
OUTPUT FOUR behaviorally distinct physician
targets with distinct promotion messaging
implications for each.
7
Segmentation Methods
8
Validity of Segmentation Solutions
STATISTICAL VALIDITY
MANAGERIAL VALIDITY
DISCRIMINANT VALIDITY

• Segment definition
• Actionability
• Identifiability
• Accessibility
• Segment Potential
• Responsiveness

• MANOVA, ANOVA
• Discriminant Analysis
• Logistic Regression
• Perceptual Mapping

9
Compare Segmentation Techniques
10
Custom Research FORECASTING HIV PATIENT
POPULATION
AIM Forecast the
available pool of HIV patients over the next 15
years. MODEL INPUTS Survival Rate, Death
Rate, New Patient Arrival Rate. DATA SOURCES
CDC Deaths, Incidence, Demographics Patient
Audit Charts METHODOLOGY Survival Analysis,
Time Series Analysis, Mathematical Model.
Survival Patient
Transition Analysis
Matrix (Conditional probability of transition
from one segment to another)
Mathematical Model output Death Rates in
Patient Segments
New Patient Arrival Rate Time Series Model
OUTPUT Forecast of HIV population over the
next 10 years with 95 precision via bootstrap
simulation.
11
NEW DRUG DEVELOPMENT MARKET RESEARCH NEEDS