Clinical Toxicology (2009) 47, 425–429

Copyright © Informa UK, Ltd.

ISSN: 1556-3650 print / 1556-9519 online
DOI: 10.1080/15563650902953586

ARTICLE
LCLT

The use of an automated interactive voice response system

to manage medication identification calls to a poison center

EDWARD P. KRENZELOK1,2 and RITA MRVOS1

IVR to manage medication ID requests

1
Pittsburgh Poison Center, University of Pittsburgh Medical Center, Pittsburgh, PA, USA
2
Schools of Pharmacy and Medicine, University of Pittsburgh, Pittsburgh, PA, USA

Introduction. In 2007, medication identification requests (MIRs) accounted for 26.2% of all calls to U.S. poison centers. MIRs are
documented with minimal information, but they still require an inordinate amount of work by specialists in poison information (SPI).
An analysis was undertaken to identify options to reduce the impact of MIRs on both human and financial resources. Methods. All
MIRs (2003–2007) to a certified regional poison information center were analyzed to determine call patterns and staffing. The data
were used to justify an efficient and cost-effective solution. Results. MIRs represented 42.3% of the 2007 call volume. Optimal
staffing would require hiring an additional four full-time equivalent SPI. An interactive voice response (IVR) system was developed
to respond to the MIRs. Discussion. The IVR was used to develop the Medication Identification System® that allowed the
diversion of up to 50% of the MIRs, enhancing surge capacity and allowing specialists to address the more emergent poison exposure
calls. This technology is an entirely voice-activated response call management system that collects zip code, age, gender and drug
data and stores all responses as .csv files for reporting purposes. The query bank includes the 200 most common MIRs, and the
system features text-to-voice synthesis that allows easy modification of the drug identification menu. Callers always have the option
of engaging a SPI at any time during the IVR call flow. Conclusions. The IVR is an efficient and effective alternative that creates
better staff utilization.

Keywords IVR; Medication; Identification; Poison center

Introduction Methods

In 2003, the American Association of Poison Control Centers’
(AAPCC) National Poison Data System (NPDS) (formerly
designated as the Toxic Exposure Surveillance System)
reported 617,414 medication identification requests (MIRs) –
17.3% of all inquiries directed to poison information centers.
In 2007, there were 1,070,537 MIRs – 26.2% of U.S. poison
center incoming call volume.1,2 The Pittsburgh Poison Center
(PPC) experienced considerable growth in the volume of
MIRs as well. The impact of the added MIRs call volume has
the potential to be significant for centers that provide this
service. To determine the ramifications of MIRs on the PPC
service capacity, an analysis was undertaken to evaluate the
effect of the service on call volume and staffing and to
identify a solution to reduce specialist in poison information
workload related to MIRs.
Received 26 January 2009; accepted 6 April 2009.
Address correspondence to Edward P. Krenzelok, Pittsburgh Poison
Center, University of Pittsburgh Medical Center, 3705 Fifth Avenue,
Pittsburgh, PA 15213, USA. E-mail: krenzelokep@upmc.edu
The PPC is an AAPCC-certified regional poison information
center that serves approximately 5.9 million residents of 44
Pennsylvania counties. The PPC Visual Dotlab (Visual Dotlab
Enterprise, version 4.3.6, 2008; WBM Software, Fresno, CA,
USA) electronic medical record system was queried for the
period of 2003–2007 using Crystal Reports (Version
11.0.0895; Business Objects, San Jose, CA, USA) to identify
all MIRs. Additionally, the PPC annual reports for the years
2003–2007 were queried to profile the total number of calls
for human exposures. The operational interactive voice
response (IVR) database of 200 prescription medications was
based on the top 50% of 2007 PPC MIRs. The data related to
the MIR call volume by time, date, and specific pharmaceutical
product were exported into Excel for analysis. An IVR system
was developed in partnership with Computer Instruments, Inc.
(Overland Park, KS, USA) to automate responses to MIRs
and placed into service on July 22, 2008. IVR data from
5 months (August 1, 2008–December 31, 2008) were ana-
lyzed to determine the number of MIRs that were completed
successfully. MIR documentation did not include unique
identifiable patient information; therefore, approval by the
investigational review board was unnecessary. Descriptive
statistics were used to characterize the data.

426
Results
The PPC partnered with Computer Instruments, Inc., an
established innovator in the development of self-service IVR
applications, to develop an all-in-one integrated voice service
application to manage MIRs – the PPC Medication Identifi-
cation System®. The application-based system was designed
to streamline poison center MIR activities through automated
speech recognition and text-to-speech technologies. Although
the IVR has the capability of managing and triaging all calls
electronically, PPC callers are first screened by a specialist in
poison information to identify the emergent poisoning calls
and to determine whether the nonemergent MIR calls should
be managed by a specialist (e.g., calls that originate from a
health-care professional, law enforcement personnel, and
elderly person seeking assistance with their medications) or
transferred to the IVR. The appropriate callers with an MIR
were transferred to the automated self-service application,
thereby increasing poison center efficiency by freeing
specialists to manage exposure calls.
Through this application, the PPC is able to collect caller
demographic information. In a sequential manner, the system
collects, via speech recognition, the caller’s zip code, gender,
age, and the medication identification code and then identi-
fies the medication for the caller (Fig. 1). In the event that the
caller is transferred to the IVR inappropriately, the caller is
given the option of being transferred to a specialist in poison
information if a poisoning has occurred. Prior to querying the
caller, a disclaimer regarding the limitations of the Medica-
tion Identification System® is provided. The IVR system then
asks the caller to “speak” the medication identification code
on the tablet or capsule. Based on the caller’s input, the sys-
tem queries the database. The medication identification that
was entered is confirmed by the text-to-speech technology to
insure that the correct medication is being identified and
continues to speak the medication description and use – the
system is set at a 100% confidence/sensitivity level to insure
that the caller confirms their MIR before a response is pro-
vided. For example, if the caller is requesting information
about a medication with the identification code of 512, the
text-to-voice synthesis informs the caller that 512 “contains
acetaminophen and oxycodone, is a white, round tablet, and
is used most commonly as a pain reliever.” Upon completion
of the MIR information, a menu is played to the caller allowing
them to repeat the medication description, look up another
medication, return to the main menu, complete a survey, or
terminate the call. Based on these options, the IVR applica-
tions will execute the feature chosen by the caller. All IVR
caller activity is e-mailed to the PPC daily as a .csv file for
review and analysis. The IVR can also be queried for current
MIR data at any time.
If the medication is not found, an application that is
referred to as a “form filler” requests the caller to participate
in a quality assurance survey. The form filler enables the cen-
ter to record the caller speaking the medication identification
Clinical Toxicology vol. 47 no. 5 2009
E.P. Krenzelok and R. Mrvos
codes as they would speak it into the system using speech
recognition. This feature was built into the product to allow
the PPC to capture identification codes of MIRs that were not
completed successfully, either because it was not in the pre-
existing database of 200 medications or the speech recogni-
tion software was unable to recognize the caller’s input.
These recordings are e-mailed as .wav files to the PPC on a
daily basis for review and allow contemporary updating of
the system.
The system was built to protect both the caller and the
PPC during the automated medication identification pro-
cess. Disclaimers are played throughout the identification
sequence informing the caller that “the system is intended
solely for the general information of the caller and is accurate
and reliable to the best of our [the Pittsburgh Poison Center]
knowledge, but is not guaranteed to be so. The information
given is not intended to take the place of professional advice
of a health-care professional. If this is a poisoning emer-
gency, you may press zero now to be transferred to a poison
information specialist.”
The IVR has optional features that allow it to control all
incoming calls when the center’s surge capacity may be over-
whelmed. All calls can be routed through the automated sys-
tem and triaged accordingly. For example, during a terrorism
event, callers with poisoning emergencies can be directed to a
specialist in poison information, whereas those who merely
need information about “white powder” (e.g., ricin) can be
directed to an informative message on the IVR through voice
recognition technology. Poison center call volumes often
surge in response to media stories about counterfeit tooth-
paste, lead in toys, food poisoning, and so on. The IVR allows
the center to create text messages rapidly that are synthesized
to speech and uploaded easily to the IVR, allowing specialists
to triage callers to the message rather than reiterating it repeat-
edly. These features can be turned on and off through a simple
Web interface provided with the system and accessed
remotely.
The PPC Medication Identification System® was a new
application for Computer Instruments, Inc. and required approx-
imately 2 months of development time and an additional month
of refinements to build, test, and implement. The direct cost
associated with software development, hardware, installation,
and training was $123,412. The indirect costs encountered by
the PPC staff involved approximately 175 h of time to iden-
tify the top 200 MIRs; verify all information regarding each
medication code, color, and shape; entering the text into the
IVR template for synthesis to text; listening to each entry to
insure accuracy; refine text-to-speech pronunciation during
the beta-test period; and training. The annual maintenance
cost is $16,075.
MIRs were responsible for 24,643 (18.5% of total call
volume) requests in 2003 and 55,473 (42.3% of total call
volume) in 2007 – a 225% increase (Fig. 2). In 2007, the
MIRs occurred in a pattern that was similar to the human
exposure call volume (Fig. 3) and exceeded human
Clinical Toxicology vol. 47 no. 5 2009
IVR to manage medication ID requests
Fig. 1. Self-service IVR application.
exposure call volume between the hours 14:00 and 06:00.
During 2007, the mean documented MIR daily call volume
was 153 and the mean human exposure volume was 134.
Each of the 153 MIRs was documented as a unique med-
ical record. After implementation of the IVR Medication
Identification System® in 2008, the number of MIRs
dropped from an average of 134 to 86.5 requests per day.
However, the number of MIRs that required electronic
documentation decreased to a mean of 6.2/day. Figure 4
427
shows the period from January 1, 2008 to December 31,
2008, and reflects the monthly MIR call volume and the
subsequent decrease in the number of MIR calls that
required medical record documentation. This represents a
95.4% decrease in the daily mean number of unique poi-
son center records that were completed for the purpose of
documenting MIRs. The mean percentage of MIRs that
were matched successfully was 20.1%/day (19.3–21.9%).
In 2007, the PPC received MIRs for 3,165 medications
 Clinical Toxicology vol. 47 no. 5 2009

428 E.P. Krenzelok and R. Mrvos

60,000 Discussion
50,000
The MIRs present a significant call volume burden for poi-
Number of requests

40,000 son centers. As illustrated, the PPC MIR call volume

increased by 225% over a 5-year period. During the same
30,000
period of time, the number of full-time equivalent SPI did
20,000 not increase. Therefore, the individual workload related to
responding to and documenting the MIRs increased
10,000 dramatically. As an initial response to the increasing MIR
0
volume, only minimal information was required to be doc-
2003 2004 2005 2006 2007 umented in the electronic medical record – date and time
Year
(autoentry), caller zip code, and the seven-digit Poisindex®
Fig. 2. Number of MIRs by year 2003–2007. code of the medication that was identified. This reduced
the amount of time that professional staff spent document-
ing MIRs but still insured that adequate data about the
trends in the potential abuse of prescription medications
were being collected. However, the sheer volume of MIRs
competed with the core responsibility of a specialist in poi-
son information – managing exposure calls. Secondary fall-
out included staff job dissatisfaction, the inability to work
on special projects, low participation in continuing educa-
tion, and an insufficient number of work breaks.
The increased MIR volume was responsible for a substan-
tial amount of overtime compensation and off-shift differen-
tial pay for additional staffing during the evening shift,
which correlated with both the peak human exposure and
MIR call volumes. Additional overtime expenses were also
incurred as specialists could not complete all the medical
record documentation related to actual exposures during
their assigned shifts. To compensate for the increased MIR
Fig. 3. Annual MIRs versus human exposure call volume by hour of call volume, it was estimated that an additional four full-
the day. time equivalent SPI would be necessary at an annual cost of
$277,453 (2008 U.S. dollars) to maintain optimal response
to poisoning exposure calls. Other than opting to eliminate
MIR the medication identification service, the only other alternat-
4,500
ive would have been to employ poison information providers
4,000
(individuals who are not qualified to work as SPI) at a
3,500
reduced salary. Because the agencies (e.g., law enforcement
3,000
and public health) that benefited from the medication identi-
Call 2,500 fication service data did not provide the funds to support hir-
volume 2,000
ing additional staff members, a pilot project conducted at
1,500
1,000
another poison center provided the idea to automate the
500
medication identification service through the use of IVR
0
technology.3 Now that the system is functional, the annual
Jan Mar May Jul Sep Nov cost of $16,075 is a modest expense to maintain the service,
Month
collect a significant amount of data regarding prescription
Fig. 4. 2008 monthly MIRs that necessitated documentation. medications that may signal regional prescription medica-
tion abuse trends, reduce staffing expenses, improve effi-
ciency and the quality of care, and enhance staff job
satisfaction.
with unique medication identification codes. Two hundred Measuring the success of the PPC Medication Identifica-
of the medications were identified as being responsible for tion System® is a matter of interpretation and individual
50% of the MIRs (medications from 200 to 3,165 repre- perception. Subjectively, the SPI appear to have improved
sented 30 or fewer individual requests per year) and were job satisfaction because their workload has decreased sig-
included as the unique production identification codes in nificantly and they no longer have as many ethical concerns
the initial IVR software development process. about the information that they provide through this service.
Clinical Toxicology vol. 47 no. 5 2009
IVR to manage medication ID requests
Although all calls, including MIRs, are answered by the staff,
appropriate MIRs are transferred with a single touch keypad
stroke to the PPC Medication Identification System®. Prior to
the implementation of the system, all calls were docu-
mented as unique electronic medical records. With the
advent of the IVR, only 4.6% of MIRs required documenta-
tion (e.g., medical professional and law enforcement inquir-
ies). Therefore, staff workload, even with the minimal
amount of information that was being collected prior to
implementation of the Medication Identification System®,
has been reduced by approximately 128 fewer calls/day that
necessitate documentation.
An evident limitation is that these data are no longer auto-
uploaded to the AAPCC NPDS (the loss of approximately
51,000 records annually to the NPDS), and this creates a data
void regarding medication identification inquiries from the PPC
service region. At the present time, Visual Dotlab does not pro-
vide the option of converting the IVR data automatically into an
AAPCC NPDS-compatible format. Positively, the PPC Medica-
tion Identification System® has reduced workload during the
hours of peak activity, thus allowing staff to manage exposure
calls more effectively and to have sufficient time to conduct peer
review activities. Additionally, overtime compensation has been
reduced, as there is now adequate time to complete medical
record documentation during the work shift.
Another limitation is the disparity between the 20.1%
successful completion rate and the anticipated 50% suc-
cess rate because of the nature of the database being
restricted to only 200 available medications. Another fac-
tor that has accounted for the apparent low success rate is
the lack of compliance by the caller, resulting in dropped
calls – users who drop the call prior to the completion of
their inquiry. As the “culture” of the caller acclimates to
the new system, it is anticipated that the successful com-
pletion rates will improve. An average of only two users
per day participated in the survey (form fill application) at
the end of an unsuccessful MIR and indicated that the
medication that they sought to identify was unavailable.
The form fill application helps to identify newly emerging
prescription pharmaceutical MIRs and provides direction
with regard to what new medications should be added to
429
the PPC Medication Identification System®, and this may
possibly contribute to a higher successful completion rate.
The PPC Medication Identification System® database con-
tains the identification codes of 200 medications that were
responsible for 50% of the MIRs. The base unit included a
module with the capability of being able to store 250 different
medication identification scenarios. Increasing the options
would undoubtedly enhance the successful completion rate;
however, there is a significant cost associated with the instal-
lation of each module that is capable of storing 250 additional
medication identification profiles. More modules will be
added as funds become available.
Conclusion
The PPC Medication Identification System®, utilizing IVR
technology, is an accurate and effective means of responding
to MIRs. The primary mission of the PPC is to reduce mor-
bidity and mortality because of poisonings by providing 24/7
emergency telephone consultation; medication identification
is a secondary service. The PPC Medication Identification
System® was developed to reduce the growing demands on
the staff and to still offer a medication identification service for
those who do not have the ability to consult with a pharmacist
or use one of the many alternative medication identification
services.
References
1. Watson WA, Litovitz TL, Klein-Schwartz W, Rodgers GC, Youniss J,
Reid N, Rouse WG, Rembert RS, Borys D. Annual Report of the
American Association of Poison Control Centers Toxic Exposure
Surveillance System. Am J Emerg Med 2004; 22:335–404.
2. Bronstein AC, Spyker DA, Cantilena LR, Green JL, Rumack BH, Heard
SE. Annual Report of the American Association of Poison Control Cen-
ters’ National Poison Data System (NPDS): 25th annual report. Clin
Toxicol 2008; 46:927–1057.
3. Bronstein AC, Creavin MD, Spyker DA, Seroka AM, Wruk KM.
Automation of drug identification using interactive voice response
technology. Clin Toxicol 2007; 45:607 (Abstract).