It Pays to be Accurate: Improving Critical Care Documentation in a Pediatric Emergency Department Free

Our hospital is a 305 bed quaternary care pediatric referral center and has over 16 000 inpatient admissions and 53 000 ED visits annually. The emergency department (ED) provider team comprises of 27 PEM-trained physicians, 6 general emergency medicine physicians, 3 pediatric advance practice registered nurses, and 3 general pediatricians. Fellows and residents in the PED are not eligible for documenting CCB. All professional and facility charge billing from the PED is performed by professional service coders based upon documentation.

We evaluated the documentation of children who were evaluated with ultimate diagnoses of status asthmaticus, diabetic ketoacidosis, and anaphylaxis. These diagnoses were identified as the most common in the PED that may qualify for CCB and were clearly identifiable in chart review. Status asthmaticus was defined as albuterol-refractory respiratory distress requiring multiple treatments of albuterol in the PED and/or the use of intravenous adjuncts. Diabetic ketoacidosis was defined by the presence of acidosis with a venous pH <7.3, ketonuria, and hyperglycemia requiring insulin infusion and hourly neurologic assessments. Anaphylaxis was defined as symptoms involving 2 or more systems requiring the administration of intramuscular epinephrine while in the PED. While billing for critical care services only begins after 30 minutes of critical care, critical care documentation was counted if there were any documentation of critical care and was not restricted by the amount of critical care time documented. We chose this metric because we were most interested in the appropriate documentation of critical care services and not the amount of time spent on critical care. We did not want to indicate that 30 minutes of critical care services was a desired goal, only that accurate documentation of actual critical care time was the goal. Patients who received the majority of their emergency care at a referring hospital and did not require an escalation of their care in the PED were excluded from the analysis as these encounters do not warrant CCB.

A team comprised of PEM physicians and a pediatrician with an expertise in QI methodology was formed to standardize critical care documentation in January 2020. PEM faculty were surveyed to determine awareness of critical care documentation and potential barriers (Supplemental Fig 6). From survey results, 3 key drivers were identified to meet our stated aim (Fig 1). Key drivers focused on the knowledge of CCB, awareness of billing practices, and an electronic medical record that supports appropriate attestations.

In September 2020, the first interventions were implemented and included lower- reliability interventions focused on critical care documentation education. Simultaneously, a critical care attestation template (Supplementary Fig 7) was introduced in the electronic medical record, which standardized and simplified critical care attestations. This was felt to be a higher reliability intervention since it fit into providers existing workflows in the PED.

In October 2020, documentation quick reference guides were posted at physician workstations for documentation requirements and how to use the critical care template. In November 2020, monthly physician feedback was provided anonymously using funnel plots. In January 2021, weekly deficient chart notifications were sent to physicians for further documentation reinforcement and for greater system learning for documentation barriers.

Baseline data were obtained from January 2019 to the implementation of the first intervention in September 2020. Ongoing project data were collected weekly from September 2020 to December 2021 through the electronic medical record. Statistical process control charts were used to study the inclusion of critical care documentation related to the timing of our interventions.

Primary outcome was the percentage of encounters with status asthmaticus, diabetic ketoacidosis, and anaphylaxis that included any documentation of critical care. At the conclusion of the study, charge data analysis was performed on our population of interest.

Retrospectively, we also reviewed charts of all patients (excluding our diagnoses of interest) admitted from the PED to the ICU to determine the rates of critical care documentation.

Process measure was the percentage of critical care documentation utilizing the newly-created critical care template. For the process measure, we only included encounters where the project’s exact template was used.

Refusal of payment for critical care level of services by insurers was monitored as a balancing measure.

Critical care documentation for the diagnoses of interest was monitored by using annotated statistical process control charts using QI Charts software (Performance Improvement Products, Austin, TX). Patients with any of the 3 diagnoses of interest were organized in groups of 20 consecutive encounters with percentage of critical care attestations plotted on the p-chart. X-bar charts were used to evaluate average encounter professional and facility charges in groups of 20 consecutive encounters. This grouping of 20 consecutive encounters provided rapid data analysis opportunities as well as a meaningful lower control limit (>0) when improvement work began. Annotations of interventions were added to the p-chart to highlight the relationship of changes to outcomes. Rules of interpretation for special cause variation for p-charts and X-bar charts were used.

Our study was approved by our institutional review board as nonresearch QI.

During the study period, there were 1448 encounters for the diagnoses of interest. 709 baseline encounters were reviewed to determine a baseline rate of critical care attestation inclusion.

Given the significant degree of knowledge variability regarding CCB and attestations, our efforts focused first on creating a concise presentation for PEM faculty. To encourage the involvement in this project by faculty, maintenance of certification (MOC) credit was offered with 1 of the requirements being the completion of an educational session. Over the course of 3 months, all faculty completed this initial step with the majority completing the education within the first month. During the educational campaign period, the team also developed and released the standardized critical care attestation template within the electronic medical record. Following clinician education and release of the critical care template, special cause variation was detected, and the centerline shifted to 88.1% (Fig 2). A second centerline shift was again detected in December 2020 from 88.1% to 94.4%, with a similar temporal shift noted with the process measure of template usage from 34.4% to 51.1% (Fig 3).

To help maintain our documentation improvement practices and in response to feedback from PEM faculty, documentation guidelines were posted in the PED and emailed to faculty for use when completing billing outside emergency department. Data from the project were reviewed at monthly division meetings. Additionally, starting in November 2020, providers were given anonymous report cards with feedback utilizing funnel plots with comparisons to peers. These reports cards also reviewed how to use the critical care template in the electronic medical record. Monthly feedback continued for the next 10 months before transitioning to quarterly feedback. After the process improved in December 2020 with only a few documentation deficiencies appreciated, weekly chart deficiency inquiries were sent to the respective providers for feedback and identification of additional barriers.

Feedback from professional billing coders were periodically received for further refinement of physician documentation practices.

Baseline data indicated that 51% of encounters of diabetic ketoacidosis, anaphylaxis, and status asthmaticus included critical care documentation (Fig 2). Following the educational campaign and release of the critical care template, special cause variation was detected with a new process mean of 88.1% (Fig 2). Special cause variation was again achieved in December 2020 to a new mean of 94.4% and was sustained for 12 months (Fig 2).

Regarding financial data, our baseline X-bar chart revealed an encounter average professional and facility charge of $593 and $3934, respectively (Fig 4). Following the initiation of the QI project, special cause variation was detected with an increase in professional and facility charges to $648 and $4579, respectively and again detected in July 2021 with an increase in professional and facility charges to $661 and $4723, respectively. This performance has been sustained for 5 months.

Subsequently, we evaluated the percent of all patients (regardless of diagnosis) admitted to the PICU from our PED to determine how frequently the PED was billing for critical care services provided in the PED before PICU admission. Encounters were measured weekly with baseline data showing that 39.3% received CCB (Fig 5). At the end of the QI project, we had noted 2 upward centerline shifts indicating that, overall, a higher percentage of patients admitted to the PICU from the PED had included the critical care attestation than at baseline. The first upward shift had a temporal relationship with the onset of the QI project to 61.8% with the second occurring with the special cause variation in critical care template utilization 69.9%. This performance has been sustained for 8 months.

When contrasting financial charges before and after the initiation of the QI project, the average of charges for the diagnoses of interest increased $61 for professional charges and $700 more for facility charges. When aggregating these differences for 739 encounters after initiating the QI project, this equated to an increase of $44 753 in professional charges and $517 582 in facility charges.

In addition, for admissions from the PED to the PICU across all other diagnoses who ultimately were admitted to the PICU, we increased average professional charges by $39 and facility charges by $427. When totaled, the financial differences for these admissions led to an increase in charges of $43 301 in professional charges and $478 054 in facility charges.

Combining the diagnoses of interest and others admitted from the ED to the PICU, we appreciated an increase in billing charges by $1.1 million dollars over 15 months.

Critical care template use was initially measured at 34.4% (Fig 3). In December 2020, special cause variation was detected with a new mean of 51.1%. This centerline shift had a temporal relationship with the special cause variation of the primary outcome measure. After the first 6 months of the project, several faculty began customizing the standardized critical care attestation templates, an available feature of the electronic medical record. Because of this customization, there was a decrease in use of the official template created for the project and yet, not a decrease in the overall success of the project since critical care attestations remained high.

Financial and billing services reported no reported changes in our rate of denials by payers.

Our use of QI methodology led to sustained improvement in critical care documentation in the diagnoses of interest and an increase in charges of more than $1.1 million in 15 months. Provider education was essential in this project given variability in knowledge. Template utilization, a higher level of reliability intervention, led to additional improvement in critical care documentation. Provider feedback did not lead to further improvement but likely contributed to sustained process performance.

Although using education as an intervention has variable results in impacting outcomes compared with workflow changes,^7,8  this component was identified as a major barrier for accurate documentation in our study. Education does require voluntary involvement, which at times can be a barrier within itself, but incentivizing education through MOC credit was helpful in gaining interest in this effort. Additionally, to obtain MOC credit for this project, participants were also required to respond to periodic surveys and to be present at multiple meetings. This requirement helped maintain provider engagement.

Our findings are consistent with a systematic review for improving documentation that templates are effective in improving documentation.⁵  The template introduced at the beginning served as an educational tool for how to appropriately document for critical care and allowed for others to develop personal documentation tools for critical care. Similar efforts in the same setting for documentation and billing accuracy for point-of-care ultrasound have also found workflow interventions that simplify and standardize documentation led to significant process improvement, whereas education alone does not.⁹ 

After the educational campaign and release of the template, posted and distributed quick reference guides for critical care documentation were circulated. These printed educational materials did not lead to a significant change in our documentation performance and is consistent with the literature that these materials show little improvement in a process.¹⁰  Although likely not globally used by providers, this intervention did appear to help 1 physician who initially lagged behind in performance compared with peers. Continued system learning and tailoring interventions to underperformers may help with overall performance. Studies to improve physician documentation with multiple interventions showed success with interventions including chart audit, pocket reminders, and education.^11–13  Given the variability of knowledge in documentation, inquiring about barriers from providers as we did with a survey is important to thoughtfully plan PDSA cycles.

At the conclusion of the study, we noted a significant increase in CCB for all other children admitted from the PED to the PICU. The process control chart (Fig 5) for CCB from the PED to the PICU did not account for children within the QI project who did have critical care documentation but did not have at least 30 minutes of critical care to warrant CCB. Based on these findings, there was a spillover effect to general practice in critical care documentation.

While provider feedback and chart deficiency auditing did not make additional improvements to CCB documentation, we believe that consistent feedback helped maintain our performance. There are multiple studies that demonstrate improvement in providing feedback to providers.¹⁴  The use of funnel plots to compare individual feedback to others appeared to help underperformers ultimately improve their performance and is an established method within QI.¹⁵  Additionally, individualized performance also helped the QI team target the needs of those who needed additional education or resources.

In comparison with recent QI work in a similar setting for CCB,⁶  focusing on critical care documentation on a few diagnoses rather than all CCB ultimately led to similar improvements. We shared similar strategies in surveying physicians, simplifying documentation practices using templates, and providing education. While we did provide MOC as an incentive for participation, there were no direct financial incentives to participants as faculty physician salary is not based on relative value unit productivity at our institution; however, increased relative value unit production may support annual productivity bonuses. Feasibility for improving documentation through QI methodology may be more achievable by focusing on common presentations with anticipation that an improved process will have a larger effect. This logic follows QI teaching that starting small before making systemwide changes is the best way to enact change within a process.

Our improvement project was completed at a single institution. Our findings may not be generalizable to all institutions, particularly to those with different billing practices. Patient population was determined by billing codes solely which is subject to inaccuracies and may not be inclusive of all cases if final diagnosis was limited or inappropriately assigned.

Improving documentation practices by intently focusing on a few diagnoses may result in a significant improvement in capturing appropriate billing charges for care provided to the critically ill in the PED. Lower reliability interventions such as education should be partnered with higher reliability interventions that focus on workflow for greater improvement. Our project focused on provider education, documentation standardization, and ongoing provider feedback and subsequently realized significant financial benefits for our institution.