Integration of a Heart Rate Prediction Model into a Personal Health Record to Support the Telerehabilitation Training of Cardiopulmonary Patients - Biomedical Engineering Systems and Technologies - page 308

2.4 Personal Health Record System

To provide a data source that is both highly available for the patient and always under
his full control, our approach physically locates the PHR in the user's home environ-
ment where it can be installed on normal PCs as well as on selected set top boxes. It is
important to allow the user to disable any access to the PHR from outside (e.g. by
appropriately configuring the system or by simply keeping it off the Internet connec-
tion) to achieve higher acceptance, than a fully Internet-based system with remote
data storage at an “unknown”' location outside the control of the user would.
Because the medical models such as the HR model requiring interoperability with
different systems, it is important to choose appropriate standards for data storage and
communication. The “Integrating the Healthcare Enterprise” (IHE) initiative works on
improving the way computer systems in healthcare share information [18]. IHE does
not define new standards but uses and combines existing ones in support of specific
use cases. IHE is mainly involved with the professional side, but has defined an inte-
gration profile based on the Clinical Document Architecture (CDA) that describes the
information exchange between EPRs and PHRs, named Exchange of Personal Health
Record Content (XPHR) [19].
XPHR documents can be bound to one of IHE's transport profiles describing how a
standards-based communication between EPR and PHR is implemented. There are
three such IHE profiles for the exchange of documents: Cross-enterprise Document
Sharing (XDS) uses a central registry and repository infrastructure, which is not
available yet in most countries. Cross-enterprise Document Reliable Interchange
(XDR) describes a point-to-point document exchange using secure Web Services and
E-Mail. Cross-Enterprise Document Media Interchange (XDM) allows users to ex-
change data through media such as USB sticks or CD-Rs (see [20]).
We have implemented XPHR as content profile and XDM as transport profile for
the interoperable document exchange. This strengthens the approach to give the user
the physical control about the own health data. Relevant predictors have been prede-
termined in the model creation process and can be either transmitted through standar-
dized medical documents or entered directly in the PHR by physicians and patients.

3 Results

We modeled the four stages of a training session (one for each training phase) for the
five different scenarios and determined the weighted RMSE to quantify the error of
each model (see Fig. 2). As a proof of concept we implemented a training schedule
definition in the PHR that uses the S1 and S2 models to predict the patient's HR dur-
ing the creation of the schedule (see Fig. 3).
Table 1 shows the contribution of different predictors to the model and their effect
on the RMSE. Because of their naturally high correlation (also known as multicolli-
nearity) it is no surprise that four of the first five predictors have a high impact on the
model. Important other predictors are age and load (Overall>1.4). The only predictor
from the weather data is the air pressure with just a very small influence of ≈0.08%.
Most of the different blood pressure values and the Borg value have no impact on the
model.