Science of Patient Experience
By: Doug Engfer: President & CEO, invivodata
invivodata integrates science and technology
to improve the process of measuring patient experience,
resulting in more timely, more valid clinical data. invivodata
delivers a new total solution for capturing patient experience-in
real time, in the real world-one that combines the advantages
of handheld electronic diary technology with the principles
of clinical and behavioral science.
Patient experience - self-observations of symptoms, relief, and
quality of life following treatment administration - is an increasing
focus of pharmaceutical science. Obtaining reliable, valid measures
of patient experience requires a deep understanding of the processes
through which patients observe and report about themselves and their
behavior. A valid scientific approach to measuring patient experience,
known as Ecological Momentary Assessment (EMA), is rooted in scientific
principles involving cognitive psychology, statistical sampling,
psychometrics, study design, and human memory. There are many considerations
in deploying a technology-based EMA solution; this article discusses
those considerations in light of new and emerging technologies.
It All Starts with the Subject
In trying to understand how to deploy a technology-based EMA-based
solution, it's useful to examine the roots of EMA research. When
Dr. Saul Shiffman, professor of psychology at the University of
Pittsburgh and co-founder of invivodata, inc, began his work on
nicotine addiction and smoking cessation in the 1970s, he approached
the work from a subject-centered perspective. Dr. Shiffman was interested
in taking research out of the laboratory and understanding the psychology
of smoking in everyday life. He wanted to know why people return
to smoking after quitting, and, more importantly, why people smoke
when they do-what triggers these smoking "episodes" and
how can we understand these episodes within the overall fabric of
the smokers' lives?
Like so many researchers using written diaries at the time, Dr.
Shiffman would frequently find subjects completing a week's worth
of diaries in the waiting room before a research visit. What was
needed was a way of capturing, in a scientifically valid way, what
people were doing and feeling in the real world at the time when
they were actually experiencing something. In his efforts to understand
how best to circumvent the biases of existing paper- based methods
of data collection, he established the study of Ecological Momentary
Assessment (EMA). The outcome of that work can be boiled down to
a fairly straightforward guideline: in order to capture and measure
valid and reliable patient experience data, one must collect the
data in the real world, in real time.
That simple statement has deep implications, however. Proper application
of EMA methodologies involves creating a rich set of samples of
patient experience, against which episodic events can be measured
and compared in statistically valid ways. From their work Dr. Shiffman
and his peers and associates derived some fundamental rules for
devices to be used in any EMA-based study:
1. The most appropriate data collection device must be portable,
because assessments are made throughout the day and must not interfere
with the subjects' normal life.
2. The user interface must be extraordinarily easy and fast to
use, because assessments are made repeatedly and in real time.
3. The means of user interaction must be natural, so that a wide
range of possible subjects can use the device.
4. The collection device must accommodate the subject's lifestyle,
so that the subject remains inclined to carry the device with
them at all times.
In clinical studies, researchers deal with a cross-section of the
population, covering many demographics. The devices we develop to
conduct this research must be usable by this broad demographic-whether
young or old, technically savvy or naïve, in the U. S. or worldwide.
Even with the explosion of technology in the U.S., the fact remains
that here, as well as throughout the rest of the world, a majority
of people gets through each day without directly, consciously interacting
with a computer. While companies have developed a number of simple,
keypad-based interfaces (ATMs, phones, etc.), there's nothing quite
so natural for people as simply pointing at something. A particular
beauty of today's handheld computing platforms is that they support
this very simple and literal "point-to-select" interface
modality. These easy- to-use, intuitive modes of interaction engage
the subject and help drive high levels of protocol compliance.
Programmability may seem to be a given, but it's worth exploring,
if only to look at some emerging technologies. EMA-based studies
rely on protocol designs that engage and interact with the subject.
In many cases, these interactions change based on when, where, and
how the subjects respond to the interview questions. For example,
consider a clinical trial designed to measure time to relief. In
such a study, when subjects record that they have taken a medication,
the device should repeatedly sample the subjects' perceived relief
during the critical time window to ensure a sensitive test of when
they first experience symptom relief. This type of dynamic sampling
dramatically increases the sensitivity of measurement (e.g., Shiffman,
Elash, et al., 2000). Only a programmable device allows researchers
to change sampling schedules so that certain data types and time
windows for data collection are built into the diary protocol.
Some interesting platforms to consider for the future are the intelligent
phone and the wireless PDA. Especially as devices similar to Kyocera's
SmartPhone become widely available (merging handheld computing with
wireless telephony), having a programmable platform that also supports
wireless communication will be very useful for EMA-based studies.
Similarly, wireless PDAs similar to the Palm VII provide a larger
display and back-end wireless communications over networks equivalent
to those provided by the paging industry.
This final technological factor is less technically oriented and
more a market consideration. Given the nature of clinical research-long
lead times and study durations, broad constituencies, and the need
for peer and regulatory approval-it's only natural that the industry
be concerned about the long-term viability of any computing platform
they use. It's important for anyone pursuing EMA-based research
to consider this and to choose a computing platform that is going
to be around not just when the study is run, but also when the results
are reviewed by FDA.
Most fundamentally, successful EMA-based research must stress ease-of-use
in the interface, almost literally to the point of instant usability.
Subjects are really the key, the linchpin, to clinical research
and they must feel respected, valued, and supported throughout the
study process because they carry a heavy burden of responsibility.
EMA techniques, properly applied in the user interface, can actually
encourage, support, and engage the subject, driving higher subject
compliance with the study protocol than is possible with any other
methodology. Conversely, complicated user interfaces that require
text entry or extended cognitive effort during assessments discourage
participation and undermine compliance.
As a consequence, in an EMA user interface, simple is better. Use
a small number of well-known interface widgets with good real-world
analogues (sliding scales, check-boxes, number spinners, and so
on) and sound psychometric properties. This approach also drives
consistency in the interface. Because the interface consists of
a small number of widgets, the subject quickly learns how each works
and can apply that knowledge to new and different assessments. The
result: higher subject performance, better subject compliance.
We touched earlier on the subjects' capacity to learn the assessment
regime. As subjects become familiar with the various questions in
the interviews, their response times drop dramatically. Subjects
quickly get to the point where they recognize questions ("Oh,
the mood question") and respond almost instantly to the prompts.
As such, there is little value in posing long-winded, wordy prompts.
Better to use brief, succinct prompts and support the assessment
with a bit of launch-time training. Wordy approaches continuously
burden the subject, eroding compliance.
Finally, the interface must be reasonable and respectful at all
times. The data collection device must fit into the subjects' daily
lives, both in terms of its form factor and with respect to user
interface. Well-designed EMA-based interfaces include many "livability"
features-suspend, bedtime, alarm, etc.-that are designed to accommodate
the varied patterns of subjects' daily lives. Let's discuss suspend
as an example. There are times during almost every day when a subject
simply can't be interrupted to respond to a random prompt. EMA devices
need to respect that, and provide a mechanism for the subjects to
remain compliant while suspending their participation for limited
amounts of time.
One of the most disrespectful things that a clinical study device
can do is create subject burden. In a user interface, few things
are as annoying as system failures-crashes, hangs, involved procedures,
reboots, whatever. Any EMA data collection device should be rock-solid.
While in the Windows® world we've come to accept our daily reboot
(if that infrequent!), the world of clinical research simply cannot
tolerate such a burden on the subject, not to speak of the risk
these errors pose to data reliability. A well-crafted EMA-based
study will generate, on average, as few as .025 Help Desk calls
per subject-week (one call per week with 40 patients in the field).
The next step in the process, after having assessed the subject
over some period of time, is to collect and analyze the subject's
data. Here, again, all of the basic principles of an EMA platform
come into play. The approach must balance the needs of the sponsor
with the needs of the subject, providing a secure, reliable means
to collect data while not burdening the subject unnecessarily. Some
handheld platforms provide a fairly natural means to collect data
that does not burden the patient: data synchronization. Palm's HotSync®
technology is probably the best-known, most-reliable, most-popular
example. Such tools provide a flexible, extensible, reliable means
to move study data from the collection device to the study database.
When accomplished at regular site visits, this approach adds no
burden to the subject, and delivers timely data to the sponsor for
For study designs that demand either more-frequent data upload,
or where site visits are infrequent, sponsors should consider using
wire-line modems to collect data from the subjects. In deploying
a modem-based solution, it's important to consider and mitigate
the potential subject burden associated with data-upload that may
occur as often as daily. Well-designed software eliminates any need
for the patient to be concerned about re-dial attempts, busy signals,
etc. Well-designed equipment packaging makes it straightforward
for even technologically naïve subjects to set up their data-transfer
kit. The net result is a system of data collection that allows study
sponsors ready access to recent data.
Finally, for studies that need data even more frequently than once
per day, or that include real-time subject feedback, wireless systems
are ideal. Such systems can allow the patient to send and receive
data effortlessly in the normal course of daily living. Cost and
coverage are both important considerations when deploying wireless
studies. Also, given the limitations of wireless coverage, study
software must not rely on continuous access to the network. A thick-client
approach, taking occasional advantage of the wireless network, provides
the best compromise between ready data access for sponsors and continual
application availability to study subjects.
Personal Area Networking
One truly fascinating area of current research and development concerns
short-distance or "personal area" networking. These wireless
protocols, whether infrared- or radio-based, offer the promise of
integrating various sensors that can augment the patient experience
data stream. For example, imagine being able to integrate the EMA
data collection device with a device that reads heart rate and blood
pressure. At a minimum, one could devise a protocol that reads that
physiological data, automatically, as a part of the EMA assessment
interview. Further, though, one could imagine triggering an EMA
assessment based on a given heart rate or BP threshold value being
achieved. Or, conversely, based on the specific results of a diary
assessment, one could cause the device to take a reading. In all
cases, those physiological data would be logged as part of the EMA
data stream. To a limited extent, in this case the future is now.
Studies have linked Palm technology to ambulatory blood pressure
monitors to collect simultaneous streams of physiological and psychological
data (see Kamarck et al., 1998).
As work with Bluetooth, 802.11b, and other short-distance wireless
protocols continues to advance, expect that this kind of integration
will proliferate, not only in clinical research, but also in patient
and chronic-disease management applications.
Doug Engfer is president, CEO and co-founder of invivodata, inc.
invivodata is headquartered in Scotts Valley, CA and the Clinical
Operations Center is located in Pittsburgh, PA.