Hospitals increasingly use process models for structuring their care processes. Activities performed to patients are logged to a database or a log. These data can be used for managing and improving the efficiency of care processes and quality of care. In this article, we propose the method for estimation of deviation of care processes from a care plan. Care plan defines the steps of a patient treatment for a certain disease in a specific hospital. Care plan is built on the base of care process model. A care process model is built on the base of exemplars of care processes, stored in a database or a log. The Dynamic Time Warping (DTW) algorithm was used for estimation of deviation. The DTW algorithm measures a distance-like quantity between two given traces containing information about execution or not execution of actions defined by care plan.
Explore of methods for management of care process (CP) as a flow of therapeutic and diagnostic activities allows us to analyze situations with patients and recommend to decision-makers (DM) appropriate action for the treatment of patients. In this connection, occurs increased interest in an automation of hospitals units (accounting, reception, offices, warehouse and so on) and an analysis and a formal representation of care processes in order to support doctors work.
The first area is deep enough elaborated - powerful medical information systems for supporting of therapeutic and diagnostic processes are developed. Flows of patients can be optimized by simulation in order to identify bottlenecks.
The second area requires additional researches to
estimate stages of a care process and to develop
recommendations for decision-makers. Currently tools
for analysis and medical diagnostics, using different
classifiers with high accuracy and precision are
developed [
The aim of our research is to develop algorithms and tools that assist to a doctor by making proposals (recommendations) on the organization of care process in accordance with an actual patient’s state and care plan. One of the objectives is to assess the quality of the rendered medical services by comparing a patient’s care process and care plan. The care plan defines the actions of the doctor and is based on the care process model, discovered from precedents.
The multi-dimensional distance based on Dynamic Time Warping (DTW) algorithm is used to differences between care process and care plan. Experiments show that DTW algorithm is effective to compare sequences of actions in relation to care processes. We consider the use of DTW algorithm to detect deviations between real care process of a patient and care plan. 2 Formalization and optimization of care processes
Some methods for formalization and application of
care processes are described in [
Automatic discovery of care process model is
difficult and actual problem. Some methods and
algorithms to solve this problem are described in articles
[
Identification Radiologist 3 Identification of possible deviations of a care process from the care plan
Qualitatively expert assessment of a trace as a way on
the graph with temporary marks will allow to reveal and
estimate various deviations from the course of medical
and diagnostic process due to both objective and
subjective reasons and to eliminate them in the
subsequent realizations of care process.
Nonperformance of action which has to be surely executed in
care process, non-compliance of actions sequence,
performance of actions not provided by care process, etc.
[
The method for detection and visualization of the
deviations associated with performance of actions not
included in the model of care process and
nonperformance of actions that need to be executed is
described in the article [
1 F 1 2 ik1 n i mi 1 ik1 n i ci 2 1
k i1 n i ri ik1 n i pi
In case of deviation detection in the course of treatment of a specific patient, n and k are equal to 1 since there is one trace and one instance. Function f will be as follows:
1 f 1 2 m
c 1 1 2 r p
If f is equal to 1, then the trace is completely
consistent with the care process model. Otherwise, there
are deviations. In [
Let’s consider a method for detection of the
deviations associated with the performance of actions not
provided by the care plan and non-performance of
actions provided by care process. The DTW method was
applied calculate the deviation or distance. The method
allows to find closeness between two measurement
sequences for a certain period of time. Generally, the
length of sequences can be different, and measurements
can be made with different rates [
DTW algorithm calculates the optimal sequence of
transformation (deformation) between two time series [
We consider the example of DTW algorithm application for comparison of processes of bronchial asthma treatment that is presented in Table 1 (data are provided by the Medical center of the Central bank of the Russian Federation). Table contains care plan that includes operations mandatory to perform and reports of real care processes of three patients.
To formalize care process reporting we will define the following variables: «executed» → α = 1, «not executed» → β = -1, «not required» → σ = 0.The distance between «executed» and «not executed» operations we will define as |α - β| = 2, and the distance between «executed» («not executed») and «not required», respectively |α - σ| = |β - σ| = 1.
Obtained care process parameters are included in Table 2 in the form of sequences
The sequences can now be compared. We will apply the DTW method to calculate care process deviations of care processes from the care plan. In Fig. 2 the process of comparison of the Patient 3 (P3) course of treatment with the plan (P) is presented as comparison of two sequences by the DTW method using the dynamic programming scheme. In the table the way determining the minimum value of deviation R is highlighted in color. In this case R =2.257.
We will apply the DTW method to calculate deviations of Patients 1 and 3 (respectively P1 and P3) care processes. All necessary data for comparison of processes are shown in Fig. 3.
Table 3 demonstrates possibility of distances calculation between precedents. The more the distance, the more precedents are differed from each other. The table shows that Precedent 1 (care trace of the Patient 1) is close to the plan.
Transfer from the reception area to the ward with beds /ICU through 2 hours or less External respiration function or Peak expiratory flow rate Pulse oximetry
Inhaled glucocorticosteroi ds Inhaled ß2agonists
Systemic glucocorticosteroi ds Inhaled glucocorticosteroi ds Inhaled ß2agonists
d required Execute Execute d d 1 day in ICU / ward with beds Not Not required required Iaoncrhtfiaonlrgemdßo2sth-eaorgortol-nists Exedcute 2-7 days in ward with beds edCxooecnrtsocuirsletatthioenraopfyan Exedcute Consultation of a physiotherapist
8-21 days in ward with beds Execute Not
d required
Not required
Not required
Not required Not execute d
d
Not execute d
Not execute d
The distance according to the established scheme is R = 1.935. Results of pair comparison of all patients care processes are given in Table 3.
Medical processes describing care of patients are useful in daily work of physicians, especially in difficult situations. Certain step towards the creation of tools to support physician’s work in the course of care process is taken in the article. At the current stage of research, methods for the analysis and evaluation of deviations associated with performance and non-performance of actions for elementary care processes are chosen and studied. The problem is solved on the existing generalized care process model and reduced to evaluation of deviations of care precedent from available traces. The proposed method allows verifying compliance of treatment with the care plan and procedures established by care standard. In addition it helps the decision-maker with the choice of a rational way of the treatment carried out with the use of strategies and rules. In reality, to make a choice of a rational way of treatment it is necessary to sort and estimate rather large number of admissible trajectories taking into account strict binding of operations at the time-point that certainly complicates process of comparison and determination of distances. Authors intend further to research similar processes.
The reported study was funded by RFBR according to the re-search project No 16-37-00034 «Research and development of methods for analysis of deviations of executed medical processes from their model»