Keywords
Pediatrics - Ambulatory care information systems - Electronic Health Records - Clinical
Care - Quality Improvement
1. Background and Significance
1. Background and Significance
Twenty-four hour cardiac rhythm monitors (“Holter monitors”) are ambulatory monitors
that record a continuous electrocardiographic tracing. The monitors are small, portable
devices (non-invasive electrodes taped to the patient’s chest and a small computer).
The surface electrograms are processed by a technician to generate summary data and
representative tracings. A cardiologist interprets the summary data and trends, then
issues a final report. The ordering physician receives the report for use in patient
care. In pediatrics, clinical indications for Holter monitors include rhythmsymptom
correlation, monitoring of medication efficacy, and identification of arrhythmias
[[2]–[6]]. Abnormal Holter results contain important information for patient care, including
suggesting an elevated risk for sudden cardiac death in some settings [[7], [8]]. The prompt delivery of the results to the ordering physician has benefits to the
physician, to the hospital and to the patient.
1.1 Local problem
In our center, return of Holter results was taking too long from the time a Holter
was returned by the patient until the time that the ordering physician received a
written report. Delays in returning results to physicians had the potential to lead
to poor patient outcomes.
Proprietary systems for electronic Holter interpretation exist, but they require a
major investment in information technologies infrastructure for implementation [[9]]. Our hospital uses the EPIC® electronic health record (EHR) system (Epic Systems
Corporation, Verona, WI). EPIC collects and stores clinical results, then distributes
report to ordering physicians. No reading or processing system for Holter reading
and interpretation was available in our EPIC build. However, EPIC has a mature software
product for evaluating and reporting radiology results (the Radiant Radiology Information
System).
2. Objectives
The aim of this project was to adapt a radiology EHR product for reporting of Holter
monitor results. Our primary goal was to decrease the duration between the date that
the patient returned his or her Holter monitor to our institution and the date that
his or her ordering physician received a final report. From a clinical standpoint,
our mean turnaround time at baseline was approximately 20 days. This generated concern
that clinical follow-up might be overly delayed and was a stressor on patient-physician
relationships [[10]]. From an administrative standpoint, the extensive paper-based system created inefficiencies
in storage, communication, and personnel time management.
We saw parallel workflows in cardiovascular Holter monitoring and the radiology department.
In both departments, a study is initiated by an electronic order. The study is performed
and additional technical processing is required prior to interpretation. A physician
who does not have direct clinical responsibility for the patient performs the interpretation
and the final report needs to be delivered quickly and efficiently to the ordering
physician. We generated the hypothesis that adapting a radiology EHR system to use
in cardiology would produce a shorter turn-around time, using fewer resources than
implementing a proprietary electronic Holter results system.
3. Methods
This quality improvement project was implemented in a tertiary children’s referral
hospital. Holter monitors are processed at our facility (“Main Processing”) and at
a subcontractor (“Outside processing”, MedNet Healthcare Technologies, Ewing, NJ).
The outside subcontractor performs the scanning function that our technician would
normally perform and does not otherwise alter the workflow. Our team consists of 4
electrophysiologists and 10 ECG technicians, plus an administrative hospital leader
who supervises the technicians. Our initial step was value stream mapping (VSM), followed
by establishing a reading template in serial Plan-Do-Check-Act (PDCA) cycles (see
section 3.2) and then a final digital implementation strategy. A back-up system was
in place to default back to the original reading method if necessary during the workflow
transition. We have an independent process to identify “urgent” or “emergent” clinical
findings on Holter. Urgent or emergent clinical findings trigger rapid notification
to an electrophysiologist and the ordering physician. This system remained in place
both before and after implementation and was not altered by the QI process described
here.
3.1 Existing Workflow
See ►[Figure 1A] for a diagram of our workflow procedures prior to the intervention, established
in our initial VSM sessions. In summary, a combination of electronic transmission,
paper-based interpretation, transcription, and manual log procedures were required
to successfully deliver a Holter monitor result. Holter tracings were interpreted
by a physician with the aid of a paper-based “Reading Template”, a semi-structured
reading tool with blanks left for numerical data and physician interpretation (Online
Supplement 1). The result was available in the medical record, but the ordering physician
did not receive notification that the result was available.
Fig.1 Workflow Summary: Contrasts the workflow before implantation (panel A) against the
workflow after implementation (panel B). The workflow summary in both A and B begins
once the technician scanning completed. Until technician scanning is completed, the
workflows are the same in two methods. EHR: Electronic Health Record.
We identified four primary inefficiencies in this system. First, the reliance on paper-based
summary reports and a paper template system required multiple person-to-person transfers
of the Holter monitor tracings. Second, the step of using a transcriptionist to type
the interpretation on a report was time-consuming, prone to introducing inadvertent
errors, and expensive. Third, the process of scanning a paper report to the EHR was
time-consuming. Fourth, ordering physicians did not receive notification of results
when a report was available.
3.2 New Workflow
Using the radiology reading template software, the paper template was converted to
a digital document and spaces for handwritten notes were converted to drop-down multiple
selection lists, plus space for unstructured interpretation. A core team consisting
of one ECG technician, one physician and one IT analyst created a digital template
based on the original paper template. The template underwent four PDCA cycles in a
test environment until we had obtained consensus on a digital template. Our information
management professionals implemented the final version of the digital template in
our EHR system, using RadiantTM software.
Once the new template was available, we implemented the final workflow shown in ►[Figure 1B]. Reports were signed electronically and the results in digital form were immediately
routed to the ordering physician. In addition, the logbook that was kept on paper
by technicians could be digitally generated after implementation of the new workflow.
3.3 Information Technologies Investment
The template required 50 person-hours encode and test. Custom software was not required;
all programming was done with the tools that were provided with EPIC. Coordination
with the technical support team from the Epic Systems Corporation was not required.
3.4 Analysis
Outcome measures were determined prior to study initiation. We pre-specified 100 Holters
for analysis at each time point, with a higher number of Holters 6 months after implementation
because we anticipated that this was the critical time point for sustainable results.
Holter monitors were excluded from analysis if there was significant loss of data
or other technical issues (6/533, 0.01%). In advance, we identified 6 data collection
points in the process for each Holter monitor: the date that the Holter monitor was
placed on the patient, the date it was returned to us by the patient, the date that
the technician delivered the Holter data packet to the physician, the date that the
physician finished the interpretation, the date the transcriptionist finished processing
the data and the date that the final report was available in our EHR. With this data,
four key outcome measures could be identified:
Total Time: The Total Time was the time elapsed in days between the date the patient
returned the Holter monitor and the date the ordering physician had access to a final
report. This was our primary outcome measure.
Physician Interpretation Time: The Physician Interpretation Time was the time in days
elapsed between the date the technician made the Holter data available to the physician
until the date the physician signed a final report.
Technician Scanning Time: The Technician Scanning Time was the time elapsed in days
between the date the patient returned the Holter monitor until the date the technician
made the data available to the physician for reading.
Patient Return Time: The Patient Return Time was the time elapsed in days between
the date the patient received a Holter and the time the family returned the Holter
to the hospital.
Differences in time intervals were analyzed by t-test and corrected for multiple comparisons
with the Bonferroni method. Stata 12.1 was used for data analysis and creation of
figures (StataCorp, College Station, TX).
4. Results
A retrospective review of 527 Holter reports was performed. We evaluated 100 Holters
prior to implementing our intervention. We evaluated the first 100 Holters immediately
after the intervention was implemented. We evaluated 227 Holters 6 months after intervention.
Finally, we evaluated 100 Holters 1 year after intervention.
The mean age of patients at the time of Holter placement was 8.7 years (range 1 day
of age – 42 years of age). Forty-nine percent of patients were female.
Before our intervention, the Total Time between the date the patient returned the
Holter monitor until the date the ordering physician received a final report was 19.8
days (SD 42.3). Immediately after the intervention, the Total Time decreased to 1.5
days (SD 1.5, p<0.001). This result was durable over the next 12 months of follow-up
(►[Figure 2]).
Fig.2 Total Time Elapsed: Total Time is the number of days between the date the patient
returned the Holter monitor and the date the referring physician received a final
report. (*) p-values are < 0.001 for all three comparisons. (ɸ) p-values are non-significant
(>1). A Bonferroni correction was used for all six pairwise comparisons.
There was a decrease in the Physician Interpretation Time with the new system (2.1
days versus 0.6 days, p=0.01, [Figure 3]). In addition, we eliminated the time that was required for the transcriptionist
to type the interpretation based on the physician’s handwritten notes (1.9 days, SD
2.6). The largest improvement in the workflow was eliminating the time between the
completion of transcription and the final report being scanned into our EHR system.
Prior to implementation of the new system, this scanning step required a mean of 14.0
days (SD 41.6). This workflow was entirely eliminated by the new system. We did not
find any differences in Physician Interpretation Time between the 4 electrophysiologists
involved in the project.
Fig.3 Physician Interpretation Time Elapsed: Physician Interpretation Time is the number
of days between the date the technician made the Holter data available to the physician
and the date the physician signed a final report. (*) p-value is 0.01. (¥) p-values
are < 0.001. (¥) p-values are non-significant (>1) for all pairwise comparisons. A
Bonferroni correction was used for all six pairwise comparisons.
Technician Scanning Time did not change with the new system (1.8 days before implementation
vs. 0.9 days after implementation, p=0.55). The Technician Scanning Time remained
similar throughout the 12-month follow-up (►[Table 1]). Patient Return Time did not change with the new system (4.1 vs. 3.8 days, p>1)
or during follow-up (►[Table 1]).
Table 1
Sustained improvement after intervention. See Methods for description of each time
interval (e.g. “Total Time”). Each cell shows the mean elapsed time in days, followed
by the standard deviation in parenthesis. All p-values reflect a Bonferroni correction
for multiple comparisons.
|
A
|
B
|
C
|
D
|
|
|
|
Elapsed time, days
|
Pre-inter-vention
|
Immediately after intervention
|
6 months after intervention
|
1 year after intervention
|
p-value, A vs. B
|
p-values, B-D[*]
|
|
Total Time
|
19.8 (42.3)
|
1.5 (1.5)
|
1.7 (2.0)
|
1.3 (1.5)
|
< 0.001
|
> 1
|
|
Physician Interpretation Time
|
2.1 (5.0)
|
0.6 (1.0)
|
0.1 (4.0)
|
0.1 (0.4)
|
0.01
|
> 1
|
|
Patient Return Time
|
4.1 (3.8)
|
3.8 (2.7)
|
4.7 (7.8)
|
3.2 (2.2)
|
> 1
|
≥ 0.7
|
|
Technician Scanning Time
|
1.8 (4.7)
|
0.9 (1.5)
|
1.6 (4.3)
|
1.1 (1.5)
|
0.55
|
≥ 0.15
|
(*) These p-values represent the pairwise comparisons for B vs. C, B vs. D and C vs.
D. The lowest p-value for any comparison is shown.
We reviewed additional Holters at the 6-month time point to allow internal comparisons.
There was no difference in Total Time or Technician Scanning Time between the “Main
Processing” Holters that were scanned internally our hospital and “Outside Processing”
Holters that were scanned using an outside subcontractor (1.8 days vs. 1.6 days, p=0.46
for Total Time; 1.5 days vs. 1.8 days, p= 0.62 for Technician Scanning Time). “Outside
Processing” Holters were associated with slightly longer Physician Interpretation
Time durations than “Main Processing” Holters (0.9 days vs. 0.3 days, p=0.001).
5. Discussion
The primary finding of our quality improvement initiative is that a small investment
in time can produce a substantial improvement in Holter monitor reporting efficiency
using existing EHR technologies. Specialized software and expertise are not always
required to improve physician workflow in EHR systems used for pediatric cardiology.
We improved our primary outcome by implementing changes to our technology configuration
and by implementing changes in our personnel workflow, emphasizing the overlapping
importance of both aspects of clinical data interpretation.
From a service perspective, non-invasive cardiology laboratories primarily care about
our primary outcome measure, Total Time. Our intervention improved the Total Time
by 18 days. This large improvement is a reflection of three smaller gains. First,
we eliminated a time-consuming step of having our medical records department scan
final reports into the EHR. By eliminating this single step, we removed an average
of 14 days from our process – nearly the entire time savings. An important implication
of the delay from the medical records department is the observation that cardiologists
may not have operational control over all delays in Holter reporting. This study suggests
that cardiologists may be able to implement internal cardiology systems that decreases
the need to rely on other hospital systems.
Secondly, we eliminated the role of a transcriptionist in our workflow. This allowed
the person who was previously employed in this role to be reassigned to other work,
at a substantial ongoing cost savings for the hospital. Removing the transcription
step accounted for another 1.9 days in time savings.
Physician Interpretation Time decreased by 1.5 days with implementation of the new
system. Prior to implementation, physicians had been concerned that eliminating transcription
and creating a direct-entry workflow would increase the workload on physicians and
increase Physician Interpretation Time. Instead, we discovered that the automated
worklist and the structured electronic reporting tool substantially improved physician
workflow. Physician Interpretation Time reflects that amount of time that Holters
spent in the physician work queue, as opposed to the time the physician spent interpreting
each Holter. However, all 4 electrophysiologists felt subjectively that the modifications
substantially improved their personal workflow for reading Holters. The time each
physician spent with each Holter is an unmeasured variable.
The results of this intervention were durable. A common pattern for novel interventions
is to demonstrate a burst of effectiveness at the time of implementation, followed
by a regression to the mean as practitioners withdraw their intense focus from the
new intervention. We found that our improvements were durable. We attribute this to
the value stream mapping sessions during the planning phase, followed by four PDCA
cycles of revision in the digital template prior to implementation, with buy-in from
all stakeholders at each cycle. This allowed us to create a system that improved the
workflow for each member of the team, even after the focus on this project dissipated
and it became the routine method of Holter interpretation.
We found Patient Return Time and Technician Scanning Time did not change before and
after the intervention. Neither of these processes were altered by the new workflow.
We included these intervals as an internal control against an enhanced vigilance effect
that might improve efficiency, even when workflows were unchanged. Only workflows
affected by the new system improved. This suggests that the durable improvements are
genuinely due to improved workflow and not due to a vigilance effect.
6. Limitation
This was a single institution study with only 4 electrophysiologists and 10 technicians
using a single EMR system. Others may not achieve the same gains that we documented
in our institution. However, the lessons of our experience may be relevant to many
physicians who are searching for efficiencies in their practice.
7. Conclusions
EHR systems are not typically designed for pediatric cardiology workflow. However,
aspects of these EHR systems can be adapted for cardiology with minimal investment
of time and resources. These adapted systems can improve patient care. In specialties
like pediatric cardiology, there is insufficient volume nationally to drive development
of dedicated commercial systems. This study demonstrates the general principle that
creative adaptation of EHR systems can improve result reporting in pediatric sub-specialties.
8. Clinical Relevance Statement
8. Clinical Relevance Statement
It is not cost-effective or time-effective to design a custom EHR software product
for every aspect of medical care. However, in many cases, EHR systems have designed
other software solutions that can successfully adapted to improve results reporting
for patients and clinicians.
Multiple Choice Questions
Multiple Choice Questions
Which of the following correctly summarizes the reasons that an EHR system improved
patient care in this study?
-
Decrease in total time between the patient returning the Holter and the final interpretation
being available for review.
-
Increase in total time between the patient returning the Holter and the final interpretation
being available for review.
-
Decrease in the technician scanning time required to process the Holter monitor for
physician interpretation.
-
Decrease in billing errors due to improved ability to track the ordering physician
and the reading physician’s workflow.
Correct Answer: A. Patients returned their Holter monitor to the hospital and a final
read was available to the ordering physician within 2 days. This improvement is a
marked decrease in time from 19.8 days, which was the average total time before the
intervention. There was no increase in time associated with our intervention (choice
B). Technician scanning time did not change in our intervention (choice C). This is
a reassuring finding since our automation did not alter the technician workflow, thus
we demonstrate that the overall time changes are not due just to enhanced vigilance.
Finally, we did not evaluate the rate of billing errors in this study (choice D).
What was the magnitude of change in the average total time saved per patient by instituting
an EHR-based reporting system?
-
18 hours.
-
1 day.
-
18 days.
-
180 days.
Correct Answer: C. The average total time per patient decreased from 19.8 days to
1.5 days (a mean improvement of 18.3 days). While the actual improvement in each medical
center may vary, the magnitude of the improvement in our center illustrates the value
of moving away from paper-based reporting systems into online systems that can be
accessed from any location.
