Paul Kallukaran, IMS America [Division of Cognizant], Plymouth Meeting, PA USA Jerry Kagan, IMS America [Division of Cognizant], Plymouth Meeting, PA USA ABSTRACT IMS America, a division of Cognizant Corporation, is the principal source of informationused in marketing and sales management by health care organizations throughout the UnitedStates. Of the various potential applications of neural networks, pattern recognition isconsidered one of major importance. This paper presents the results of using neuralnetworks to classify time-series data into several trend pattern classifications [e.g.,Increasing Trend, Decreasing Trend, Shift Up, Shift Down, Spike Up, Spike Down, and NoPattern], and the information generated from the classifier is used to detect variousmarketing related phenomena [e.g., Brand Switching, Brand Loyalty, and Product Trends].The data used for the test consisted of prescription data for 12 months, for 600,000prescribers writing four drugs in the Anti-ulcer market. Using the neural network classifierand brand switching algorithm, the system was able to detect 2500 prescribers who werechanging their prescribing behavior. The model is a promising formula for analyzing times-series information from extremely large databases, and presenting the user with onlyinformation relevant for decision making. This data-mining system, designed as part of theIMS Xplorer product, uses SAS System components for data retrieval, datapreparation, graphical user interface, and data visualization. The IMS Xplorer productis a sales and marketing decision support system based on commercially-available, client-server technology created for pharmaceutical companies in their effort to fully utilize themission critical information found in the Xplorer data warehouse.
KEY WORDS: Time-Series Data, Neural Networks, and Data Mining
Introduction
cluster and discriminant analysistechniques.
During the past decade there has been asignificant increase in the use of artificial
In fact, there is research to indicate that
control, predictive modeling, data analysis,
pattern recognition and signal processing
traditional statistical methods in the areas
(Nibset, Mclaughlin, and Mulgrew 1991).
of forecasting (Sharda and Patil 1992).
well established in applied fields, where
neural networks are recognized as flexible
and powerful tools for solving prediction
alternatives to the more traditional time-series analysis (de Groot and Wuertz
1991), linear and nonlinear model fitting
into various groups [e.g., IncreasingTrend, Decreasing Trend, Level Shift Up,
Up]. The results from this neural network
products and are also used at IMS to detect
prescribers that have changed theirprescribing behaviors. Other statistical
The Xponent database, in the last couple
etc.] can then use the classifications from
becomes extremely difficult to identifyphysicians changing their prescribing
Database Background
behaviors. The neural-network modelprovides a method of analyzing times-
series data and identifying physicians that
are growing at an alarming rate. As these
over time. The method provides a tool for
effectiveness, data-mining techniques gain
each prescriber, yields an annual gain of$52 million in sales. So, if you’re not
175,000 sites across the United States.
physicians, pharmacists, veterinarians,drugstores, hospitals, distributors, and
Applications of the Data Mining Technique for Targeting
including computer tape, microfilm,purchase invoices and surveys.
Xponent, the first true physician level
impossible for human analysts to visually
database for the health-care industry, is
examine all of the time-series data and to
classify time-series data into the various
prescribing activity by using a customized
projection factor for each prescriber.
collected over successive time intervals. The data can be product volumes for a
Xplorer is a customized decision support
particular market or prescription data for
model is applicable regardless of the data
research analysts, economists, behavioral
allow clients to access the data, conduct a
scientists, security analysts, and others,
broad range of analyses, and produce easy-
study time series to gain an understanding
of general market trends and to takesubsequent action based on the trends.
developed to classify time-series data into
various trend patterns and to detect various
performance. For our data-mining test, themodel was used to detect physicians that
Implementation of a Data- Mining Solution Using the SAS System
product B, C, or D over a 12 month period. The data used for the test consists of
600,000 prescribers, writing four drugs in
the Anti-ulcer market. The purpose of the
tool had to have the ability to operate in a
client/server environment, to have access
Table One: SAS System Components Used in the Project SAS COMPONENT FUNCTION 1. SAS/AF ®
All the graphical user screens , to select type of analysis,data elements and viewing the final results. 2. SAS/CONNECT ®
Connectivity to the UNIX server and interface to theOracle Database from Client. [Windows 95]
3. SAS/ BASE ®
All data preparation, transformation and statistics. 4. SAS/GRAPH ®
All the graphical reports for data visualization.
switching brands. The results of the test
transformation and statistical functions,and to furnish the user with several
were changing their prescribing behavior.
physician that was previously loyal to the
drug A, and in the last 5 months switched
to prescribing drug B. This reportprovides a useful tool for thepharmaceutical company’s sales-force touse when targeting the right prescribers for
Table Two: List of Information Provided by the Client to Run the Model CRITERIA SETTING Time Periods Doctor Specialty Payment Type Distribution Channel Geography Products
For a typical analysis, the user may specify
5.0 Conclusion
the examination and analysis of 600,000time series would take weeks of work. By
After the criteria are selected for the data-
person can use this model to target doctors
analysis are shielded from the end user.
First, the analysis uses SAS to build SQL
selling and to devise a specific message to
database residing on the server, andtransforms and manipulates the data set to
provide the correct input formats for the
preprocessing capabilities, and offering the
the prescribing of one medication over the
visualization. SAS provides all the tools
prescribing of another. The final results
required to implement such a system.
from the analysis are saved in a SAS dataset. The complete analysis is performed
To help the user to understand the factors
various statistical techniques [Multinomial
producing a final SAS data set about 2,500
classifications from the neural-networkmodel as a response variable. In the
future, IMS plans to implement additional
support system, to help the business user
reports. The entire process is seamlessly
within the mountains of captured data.
integrated with the SAS system, giving theuser complete flexibility in running theprocess without knowing any SAS. Biography
Lapedes, A., and Farber, R. (1987),“Nonlinear Signal Processing Using
Operations Research from Illinois Institute
of Technology in Chicago. Sincegraduation, he has worked in the
expertise in statistics, operations research,
and artificial intelligence. Currently, he is
Application,” Proceedings of the Eight
Nelson, M., and Illingworth, W. T. (1991),
A practical Guide to Neural Nets, New
York, NY: Addison Wesley PublishingCompany, Inc.
Jerry Kagan is a Senior Programmer in theAdvanced Analytics Group at IMS
and Neural Works Explorer, Pittsburgh
SAS system for the past nine years,working primarily in the pharmaceutical
development. Jerry has been an invitedand contributing speaker at SUGI and
(1992), “Neural Networks As anAlternative to Statistical Modeling of
Contacts
Radio Wave Propagation”, AmericanStatistical Association Winter Conference,Acknowledgments References
Institute Inc. in the USA and othercountries. ® indicates USA registration
deGroot, C., and Wuertz, D. (1991),“Analysis of Univariate Time series with
registered trademarks of their respective
Classical Examples,” Neurocomputing, 3,4
Figure One: Physician Targeting Report Doctor Targeting Report Company: Lily Market: Anti-Depressant September 1994 - August 1995 Product: Prozac Doctor: John Smith DEA NUMBER: AA0382856 Market Share Rx Volume PROZAC ZOLOFT PROZAC ZOLOFT Speciality: OBSTETRICS Number of Samples: Number of Details: Product Switched: PLANS ASSOCIATED: US HEALTHCARE Reasons for Switch:
Less Drug Interactions, More Cost effective
Market Share By Product/Doctor Rx Volume By Product/Doctor Rx Volume Appendix A TREND ANALYZER/SAS INTERFACE TOOL PROTOTYPE USING SAS/FRAME PROCESS FLOW DIAGRAM
BIJLAGE A: Formulier voor aanvraag tot terugbetaling van de specialiteit SEEBRI (§ 6690000 van hoofdstuk IV van het K.B. van 21 december 2001) I – Identificatie van de rechthebbende (naam, voornaam, inschrijvingsnummer bij de V.I.): II – Elementen te bevestigen door de geneesheer: Ik ondergetekende, dokter in de geneeskunde, attesteer dat de hierboven vermelde patiënt aan COPD van
Short Communications and Brief Case NotesSHORT COMMUNICATIONS AND BRIEF CASE NOTESthat the determination of specifi c IgE, and especially IgG, to Fatal Intraoperative Anaphylaxis After aprotinin should be evaluated in patients with prior contact Aprotinin Administration with this peptide. Patients with high antibody titers should be considered at risk, whereas the absence of aprotinin-s