Keywords geographical information systems - clinical documentation - billing - emergency
Background and Significance
Background and Significance
Accurate and detailed physician documentation is required for key stakeholders to
characterize patient acuity and ensure appropriate reimbursement.[1 ]
[2 ] All services provided by physicians during an emergency department (ED) visit, including
procedures and cognitive work, are described by common procedural terminology (CPT)
codes. These codes must be translated by trained medical coders into professional
charges.[3 ] However, physician descriptions of patient care vary widely and may omit essential
verbiage for appropriate coding and billing, leading to inaccurate reimbursement for
services and incomplete severity of illness metrics.[4 ]
For patients who require critical care services, physician documentation must reflect
the aggregate time spent on such care, in addition to medical necessity. CPT coding
depends on multiple factors, including time increments as documented by physicians
(e.g., less than 30 and 30–74 minutes).[5 ] This allows the coder to decide whether an emergency department CPT code (e.g.,
99285) or a critical care code (99291) is appropriate. Such codes alter the relative
value unit (RVU) assigned to the provided care and to financial reimbursement.
However, estimating the passage of time is difficult, especially in stressful situations
with varying stimuli.[6 ]
[7 ] This difficulty accurately perceiving the passage of time compounds the potential
for the inadvertent omission of key components of critical care documentation, namely
time-based care estimates. Prior work has demonstrated discrepancies between documented
elements of clinical care and those which were observed by third parties, highlighting
an opportunity to improve documentation by accurately tracking physician time at the
bedside during critical care encounters.[8 ] Real-time location systems (RTLS) have become common in clinical environments for
asset tracking. While RTLS applications to augment clinical quality improvement methods
have been described less commonly, efforts to apply situational analytics to health
care information technology implementation have resulted in insights into context-specific
activities and related outcomes.[9 ] As part of an ongoing initiative at our institution to explore the value of RTLS
in various operational and academic applications, we piloted this project to provide
insight into how co-locating patients and staff could enhance audit and feedback of
clinician documentation.
Our objective was to augment traditional feedback and audit quality improvement methodology
with the strategic integration of RTLS geolocation data. We sought to integrate RTLS
capabilities with QI methodology to precisely identify encounters in which critical
care services were provided and offer clinicians personalized, timely and specific
documentation support. This QI project was evaluated by Cincinnati Children's Hospital
institutional review board and did not meet the definition of human subjects research.
Methods
Setting and Context
This project was conducted at a pediatric tertiary care emergency department with
a volume of approximately 80,000 patients. This institution houses the regions only
pediatric intensive care unit and admits 85 to 90% of pediatric patients from a catchment
area of approximately 2,000,000 people. We formed an improvement team to strategically
integrate RTLS, evaluate baseline documentation patterns, form a framework of improvement,
and develop interventions based on key drivers to improve critical care documentation
for patients treated in our pediatric ED and admitted to an intensive care unit (ICU).
Our goal was to increase the proportion of physician documentation with accurate critical
care attestations for critically ill children from a baseline of 76% to a goal of
90% over a 6-week period and to gain insight into how co-location data could be incorporated
into documentation auditing. We incorporated multiple iterative process revisions
into the primary plan-do-study-act cycle of the project to refine our data collection
methods and intervention development.
Emergency Department Workflow
The ED is staffed by pediatric emergency medicine (PEM) faculty and fellows, staff
pediatricians, nurse practitioners, and resident physicians. All patients who are
determined to be critically ill, likely to become critically ill, or require intensive
resource use are evaluated and treated in the Shock Trauma Suite (STS), a designated
critical care area adjacent to regular rooms. To make this determination, triage nurses
apply criteria to patients who self-present, arrive via emergency medical services
or via transport from outside facilities. These criteria incorporate vital sign abnormalities,
physical exam findings, past medical history, and mechanism of injury to estimate
the risk of critical illness and need for high-resource utilization. Additionally,
patients whose medical status declines during the ED stay are moved from their ED
room to the STS for the duration of their need for critical care. Therefore, nearly
all critically ill patients are treated in this area for the duration of their ED
visit. Additionally, physicians at our site typically perform most critical care tasks
(consultant phone calls, chart review, and leading the care team) in the STS, allowing
us to capture these components of critical care time.
Internal analysis of our ED workflow prior to this project has shown that patients
who receive treatment in this area spend an average of 68 minutes (standard deviation
[SD]: 32 minutes) in the STS, while physicians spend an average of 35 minutes (SD:
18 minutes). Therefore, simply using patient-based location time metrics would result
in a large denominator upon which to base our interventions and an unacceptably large
number of “false positives.” We sought to utilize RTLS-derived data to inform our
improvement project to increase the precision of our interventions.
Critical Care Documentation
The Centers for Medicare & Medicaid Services (CMS) define critical care as a physician's
direct delivery of medical care for a critically ill or injured patient. A critical
illness acutely impairs one or more vital organ systems such that there is a high
probability of imminent or life-threatening deterioration. Critical care is time based
and involves complex decision-making, and must be reasonable and medically necessary.[5 ] Our institution embedded a “smart phrase” in the electronic medical record (Epic
Systems, Madison, Wisconsin, United States) to facilitate the documentation of required
elements of critical care, both to aid coding and to improve communication among providers.
This specifies critical care time that does not include separately reported billable
procedures. Addition of this smart phrase to the medical record requires the documenter
to take multiple steps and is not required for chart completion.
Real-Time Location System
RTLS refers to various technologies that automatically track the geolocation of objects,
equipment, or people within a defined physical space in real time.[10 ] Various communication methods, such as radiofrequency identification (RFID) and
infrared (IR), are used to locate transmitters on the objects of interest in relation
to installed sensors, allowing precise locations to be displayed and analyzed over
time. The virtual environment can be mapped to the physical environment and standard
workflow patterns to allow for robust analysis of movement data.[10 ] There are myriad applications of RTLS in healthcare, including operational, clinical,
safety and quality improvement initiatives.[11 ]
[12 ]
[13 ]
[14 ] In anticipation of a large scale RTLS installation in a new critical care tower,
our institution implemented an RTLS pilot program in the ED to troubleshoot installation,
streamline implementation, and test its operational functionality. An RFID-based RTLS
(CenTrak, Newtown, Pennsylvania, United States) was installed in the ED in October
2018 and all staff badging was operational by February 2019. The RTLS-derived data
were internally validated with accuracy to within 2 feet with a 3-second delay (range = 1–9 seconds
delay). Data transfer to internal servers was also validated and reliable. An operational
test was conducted from February to April 2020 in which all patients who arrived in
our ED had an RTLS badge immediately attached and were tracked throughout their encounter.
This test was designed to refine and solidify the operational choreography required
to quickly attach RTLS badges to patients upon arrival and remove the badges upon
discharge or admission. We conducted our project in conjunction with this test to
take advance of the unique opportunity to precisely co-locate patients with specific
physicians throughout their time in the ED. Consent for RTLS tracking was obtained
from patients as part of the general consent to treat process.
Our improvement team included three PEM physicians, a coding specialist, a patient
registration manager, a nurse manager, and an office administrator. We constructed
a key driver diagram to make explicit our framework of improvement ([Fig. 1 ]). Each of these drivers was identified as a component of the care process which
may reduce critical care documentation. We used multiple iterative process revisions
to design and refine our definitions, data collection, and processes to address these
drivers, ultimately resulting in our final email intervention. We collected data during
the operational test during which patient geolocation was available to precisely characterize
patient encounters, which likely included critical care services and to evaluate our
interventions.
Fig. 1 Key driver diagram of inaccurate critical care documentation. ICU, intensive care
unit; PEM, pediatric emergency medicine; RTLS, real-time location system.
We posited that, for patients who were ultimately admitted to an intensive care unit
(pediatric ICU, cardiac ICU, or neonatal ICU), physician time spent in close physical
proximity to a patient located in the critical care area of our ED would likely represent
the minimum amount of time providing CMS-defined critical care. Therefore, we sought
to determine the baseline rate of encounters in which physicians spent more than 25 minutes
in proximity to critically ill patients in the STS and included a critical care attestation
in the EMR. From February 2019 to February 2020, our RTLS capabilities included physician
and staff locations but not exact patient location. Therefore, we used the location
of the physician in the specific STS room where the patient was located based on EMR
and audited charts to determine rates of appropriate documentation. This baseline
determination occurred prior to implementation of interventions for this project.
Determination of Critical Care Encounters
Physicians in our division are educated on CMS documentation requirements upon hiring.
Reinforcement occurs during routine divisional staff meetings, but review of documentation
for critical care encounters was not routinely performed prior to this project. We
utilized the expertise of each team member to strategically integrate RTLS capabilities
to enhance the process to identify patients who were most likely to have received
critical care services. We used iterative process revisions to determine the criteria
for time and location, which would capture all the patients who were most likely to
receive critical care services while limiting “false positives.” We then identified
the physicians who were primarily involved in the care provided in the STS, and characterized
their RTLS-derived locations were compared with these patients' locations. Most of
the critical care for these patients was provided at the bedside but that portions
may occur at a separate location (phone calls with consultants, chart review, etc.).
Therefore, a 25-minute cutoff captured a portion of encounters with up to 5 minutes
of care occurring elsewhere in the ED, thus fulfilling CMS critical care criteria.
This strategy under-represented all critical care services provided (patients who
expired were not admitted to the ICU, etc). However, given the specific workflow patterns
in our ED, the strategy allowed for the most efficient and targeted testing of this
RTLS-based project.
Report Generation and Streamlining
The team created an operational process to streamline data collection from multiple
sources, including RTLS, scheduling software, and medical records. This process was
iteratively refined until we were able to generate a comprehensive daily report outlining
all the patients who met the aforementioned criteria and the physicians who likely
provided their critical care services for the previous 24 hours.
Email Intervention
For 6 weeks, an administrator reviewed the daily location report and sent a single
email to each physician who met the project criteria for critical care services for
specified patients. The email was sent only to the identified physicians and included
the date of care, names, and medical record numbers of the critical patients, and
the duration of time spent in the STS with those patients from the RTLS. The text
of the message reminded the physicians to consider if their care constituted critical
care as defined by CMS and, if applicable, to ensure their note was documented accurately.
To minimize “false-positive” documentation, the email was sent within 1 day of the
encounter and explicitly reminded physicians that, ultimately, documentation was at
their discretion in accordance with CMS regulations for separately billable procedures
(i.e., intubation).
Measures
Our operational definition for our primary process measure is the proportion of patient
encounters in which a critical care attestation was appropriately documented by the
attending physician. The denominator for this measure was all encounters of patients
were treated in the STS and admitted to an ICU and in which the attending physician
was co-located with the patient for >25 minutes. The numerator for this measure was
encounters in which a specific critical care attestation was documented, which included
critical care time spent on patient care, specific functions performed, and patient
complexity and risk of morbidity.
Analysis
A P-chart was constructed to analyze the proportion of eligible patient encounters
in which critical care was accurately documented. The identified process measure was
measured over time on a statistical process control chart to evaluate the impact of
the defined intervention. The rules for interpretation of a Shewhart chart were applied
to the P-chart to identify special cause variation.[15 ] We identified characteristics of encounters in which a critical care attestation
was likely warranted but not documented.
Results
We evaluated a total of 92 patient encounters during an operational phase in which
all patients had their location tracked via RTLS between February 3, 2020 and April
7, 2020. The median proportions of patient encounters meeting inclusion criteria with
critical care attestations changed from 76.0 to 84.9% in the postimplementation period,
but a change in centerline was not demonstrated ([Fig. 2 ]). In total, 22% of these patients spent more than 30 minutes in the STS, but physicians
did not spend more than 25 minutes co-located with them and did not document a critical
care attestation.
Fig. 2 Proportion of encounters of patients treated in the Shock-Trauma Suite and admitted
to a pediatric intensive care unit who had a critical care attestation documented
in the medical record (February 2020–April 2020).
Plan-Do-Check-Act Cycle and Iterative Process Revisions
Throughout the project, we performed multiple iterative process revisions to refine
RTLS and data management processes ([Fig. 3 ]). The initial revisions focused on configuring RTLS settings to accurately determine
the co-location times of critically ill patients and physicians. To efficiently use
resources, we initially placed RTLS badges only on patients who received care in the
STS in September 30, 2019 to November 7, 2019. This became problematic for registration
staff workflow as some patients arrived by car, some arrived by EMS, and some did
not immediately require critical care services, creating situations with inconsistent
badging. We ultimately utilized a comprehensive workflow for registration staff to
place badges on all patients in the ED. Coupled with a standard process to reclaim
badges upon patient departure, we were able to increase badges available for cleaning
and reuse.
Fig. 3 Flow chart of iterative process revisions.
Subsequent revisions were used to refine the criteria to identify patient encounters
that likely included critical care services and to efficiently produce the daily report
and email intervention. A cutoff time of 25 minutes of identified co-location was
used after multiple revisions to broadly capture relevant patient encounters accounting
for some critical care time outside of the STS without too many “false positives.”
Similarly, the email intervention was refined to effectively target relevant physicians
while reducing administrator workload. We initially created individual emails to each
physician for each patient encounter. However, due to the relative few daily patient
encounters, the email was consolidated into one reminder to each specific physician
which included all their relevant patients.
A Pareto chart was constructed to display the types of patient encounters in which
documentation was most commonly recorded inaccurately. Patients with primary respiratory
complaints accounted for the most common type of encounter with inaccurate critical
care documentation ([Fig. 4 ]).
Fig. 4 Pareto chart of patient category for inaccurate critical care documentation (February
2020–April 2020, n = 15).
Discussion
Summary and Interpretation
We integrated RTLS with improvement methodology to design, refine, and evaluate the
effect of interventions on critical care documentation. The inadvertent omission of
key components of time-based documentation for critical care services in the ED may
lead to inaccurate CPT coding and reduced reimbursement. Maximizing appropriate reimbursement
for patient care is paramount to the fiscal sustainability of health care facilities
nationwide, especially in economically challenging times. While we did not show a
change in the centerline of the proportion of encounters with critical care documentation,
we successfully developed, defined, and integrated an RTLS-based system to efficiently
identify applicable encounters and remind physicians to evaluate that encounter for
critical care services performed. During the postintervention period, the COVID-19
pandemic dramatically decreased patient volumes, changed ED workflow, and ultimately
ended our project. While we did not show special cause variation in our primary outcome,
we are encouraged by the operational success of the project, positive feedback from
clinicians, and the trajectory of our results. In a short-time period, we demonstrated
the potential for effective interventions and the implications for similar RTLS applications
in various healthcare environments.
In addition to a growing volume of critically ill patients, EM physicians are also
providing more critical care services in the ED.[16 ]
[17 ] There is increasing emphasis on appropriate documentation to capture these services
and accurately describe mortality and morbidity characteristics of these populations.
Prior clinical documentation improvement programs have focused on standardizing documentation,
physician education, and auditing.[18 ] Elkbuli et al found improved accuracy in mortality, case mix index, and severity
of illness after program implementation in a trauma service.[19 ] Kittinger et al showed desirable results after a similar improvement effort in a
plastic surgery practice. Other improvement projects focused on intensive documentation
education and concurrent auditing for trauma and critical care physicians.[20 ]
[21 ]
While these projects successfully achieved their improvement goals, they required
expensive and time-consuming educational sessions and retrospective team auditing.
We were able to perform a focused intervention utilizing a novel technology to augment
usual coding education and practices.
Improvement projects focusing on the electronic medical record have improved documentation
efficiency, coding completion, and improved reimbursement.[22 ]
[23 ] Utilizing the benefits of automation through electronic data capture can remove
some of the cognitive burden from physicians.[24 ]
[25 ] King et al demonstrated the utility of an RTLS-equipped EMR to improve performance
in locating patients and increase physician efficiency interacting with the EMR.[26 ] Our project demonstrated an effective method to automate components of critical
care documentation, which usually falls to the individual physician: the perception
of time and proactively documenting that time in the chart.
We used a Pareto chart to display the patient categories in which the proportion of
inaccurate documentation was most common. While not directly related to our stated
outcomes, these data demonstrate that patients treated for critical abnormalities
of the respiratory system were most commonly inadequately documented. This is expected,
as disorders of the respiratory system are a frequent reason to treatment in the pediatric
ED. Our findings may inform further targeted efforts and highlight the need to help
physicians accurately identify the critical care time they spend on services for common
conditions with a wide range of severity. Additionally, the RTLS-based processes developed
in this project have the potential to augment traditional process related analyses.
While this study was not powered for more detailed analysis, future investigation
of the patient–physician dyad, predictors of documentation errors and patient experience
with RTLS-based processes may be warranted.
Strengths
In this focused quality improvement initiative, we utilized a novel application of
RTLS to provide targeted, timely, and specific documentation support to EM physicians
in a tertiary care pediatric ED. Multiple iterations of intervention refinement resulted
in an automated report, requiring minimal staff efforts to create daily emails to
physicians. We reduced “false positive” emails by 22% utilizing patient–physician
co-location metrics compared with patient-only metrics. Ultimately, we postulate that
intervention provided specific feedback to individual physicians regarding patients
they had recently treated, allowing them to easily adjust their documentation as appropriate.
While our results did not result in a centerline shift to demonstrate special cause
variation, this pilot work demonstrates the feasibility of RTLS-based interventions.
With additional data collection after workflow “normalization” postpandemic, we expect
centerline shift.
This work has significant financial implications. For critical care services provided,
the targeted intervention described would increase the coded work relative value unit
from 3.80 to 4.50 (99285–99291). For emergency departments providing 5 to 8% of patients
with critical care services with annual volumes of 100,000 patients per year, this
intervention would result in $120 000 to $202 000 of additional professional billable
services.[27 ] These cost savings must be balanced against the significant financial investment
in RTLS installation and maintenance. The costs to install and maintain widespread
RTLS in clinical setting vary widely. These costs can be significant and include setup
(hardwired equipment costs, and installation work) and maintenance (ongoing RFID tag
purchasing, software licensing, and system updates, personnel management) costs. To
provide adequate coverage, reliability, and precision in our ED of 42 patient rooms,
four critical care bays, and three staff work areas, RTLS installation costs totaled
over $270,000. Hardware installation and wiring cost $114,000, software setup cost
$70,000 and badges, parts and equipment cost $84,000. While these systems can also
assist in asset management, inventory management, and supply chain logistics, operational
leaders must incorporate large initial investment with expected long-term cost savings.
This project demonstrates potential return on investment when thoughtfully applying
RTLS to clinical operations.
While this project was conducted in a pediatric ED, there are numerous lessons learned
that are applicable to broader contexts. We confirmed that documentation of time-based
care is difficult, and there is a need for objective time-based measurements and feedback.
Specific, timely and targeted feedback is effective and well received when coupled
with a demonstrable outcome measure. In a dynamic clinical environment, accurate measurement
of co-location of patients and physicians via RTLS depends on precise tracking of
both parties. We found that thoughtful planning was needed to strategically apply
RTLS capabilities to clinical workflow, thereby maximizing benefit and adapting to
system limitations.
This study provides proof of concept for the value of RTLS to objectively measure
time-based clinical activities for quality improvement. Future work is needed to further
refine these processes and create fully automated RTLS-based documentation support
to accurately capture and communicate critical care services.
Limitations and Lessons Learned
Our work has several limitations. It was performed at a large tertiary pediatric ED
with an operational RTLS that may limit generalizability. Physicians do not need to
be physically proximate to patients to perform critical care services. Other tasks
such as consultation with experts may take place away from the bedside. While this
certainly has the potential to impact our results, aspects of our work blunt this
limitation. We strategically chose a threshold of 25 minutes to reasonably include
all patient who most likely had critical care services performed, based on local practice
patterns and workflow. This threshold effectively serves as a minimum requirement
for a patient encounter and thus would be included in our email intervention. Physicians
could then decide after receiving feedback whether they performed critical care services
and could adjust their documentation as needed. This threshold could easily be adjusted
to account for various workflow patterns and optimize the sensitivity of the intervention.
We did not consider patients treated outside of the critical care area of the ED,
and it is possible that some patients received critical care services in regular ED
rooms. Thus, as our measurement of location serves as a proxy for our primary outcome,
the scalability and generalizability of our work is limited. We did not follow up
directly with physicians to determine their reasons for documentation decisions, as
our objectives were primarily to test the effectiveness of a partially automated intervention
with minimal staff efforts. While it is possible that our results could be skewed
by a small group of physicians contributing to documentation inaccuracies due to their
own style, our analysis showed only one instance of a single physician with two encounters
with inaccurate documentation. We were unable to evaluate the long-term effects of
our intervention due to drastic changes in the ED workflow due to the COVID-19 pandemic.
However, our work has produced an RLTS-based process of identification of critical
care services provided in an ED to be used in future QI work.
Key lessons learned may be applied to RTLS applications in other zone-based care environments,
including EDs and ICUs. Intervention design requires intimate knowledge and consideration
of local workflow patterns and physician practices. The completeness of RTLS linkage
to services provided in various areas may be improved by considering clinical resources
used, orders, and patient complaints and acuity. While documentation decisions are
ultimately at the discretion of clinicians, thoughtful nonintrusive co-locating feedback
can be helpful when provided in an efficient and timely manner. Finally, the costs
associated with RTLS installation and operation can be significant. While the cost-effectiveness
of any single RTLS application may not be sustainable, when developed in a concerted
effort to complement an overall RTLS implementation plan, these applications may provide
value for patients, clinicians, and hospital systems.
Conclusion
Implementation of a quality improvement initiative utilizing RTLS created timely,
specific and targeted physician feedback for critical care services provided for pediatric
ED patients who were admitted to an ICU.
Clinical Relevance Statement
Clinical Relevance Statement
This study describes a quality improvement project to improve critical care documentation
in an emergency department by integrating quality improvement methodology with geolocation
data from a real-time location system. This strategic integration improved the precision
of an audit and feedback intervention to efficiently target physician reminders and
minimize unnecessary interruptions. When thoughtfully applied, RTLS-derived data can
augment quality improvement methods to successfully improve health care delivery.
Multiple Choice Questions
Multiple Choice Questions
Real-time location systems can augment quality improvement methodology by:
Providing objective geolocation information to inform specific targets for interventions
and selected outcome measurement
Replacing improvement team planning and plan-do-check-act (PDSA) cycles
Reducing time requirements for data analysis
Circumventing staff acceptance of time-based measurements
Correct Answer: The correct answer is option a. RTLS-derived data can enhance quality improvement
initiatives after carefully planning and application of relevant geolocation data
to clinical workflows and human interactions. Successful application depends on this
careful application, which requires an improvement team and PDSA cycles to optimize.
Analysis of RTLS data can be time intensive and requires staff acceptance to smoothly
integrate into operations and improvement work.
Barriers to accurate clinical care documentation include:
Ambiguous definitions of critical care as outlined by CMS
Perception of time spent accomplishing a task in a stressful environment
Extensive physician knowledge of billing codes associated with critical care
Extended time passage between service provision and completion of documentation
Correct Answer: The correct answer is option b. Critical care documentation is dependent upon providers
understanding the CMS definitions of critical care, recognizing the need for documentation
of their services, and remembering to document the time spent providing services retrospectively.
The CMS definitions of critical care are explicitly outlined within published billing
and coding, and providers must be familiar with the definitions of “critical care”
services to appropriately document. Though some institutions require providers to
complete their own coding and billing, this is not a universal requirement, and most
institutions provide coding and billing specialists to complete this task or consult
if there are questions. Critical care is documented retrospectively in all cases,
and evidence shows that stressful environments can alter a person's perception of
the passage of time. Documentation of services must be completed “during, or as soon
as practicable after it is provided in order to maintain an accurate medical record,”
per the 2019 update to CMS Medicare Claims Processing Manual Chapter 12 (physicians/nonphysician
practitioners).
