Serious male doctor using mobile phone at work

Efficiencies increase health system revenues.

Major shifts in the healthcare landscape over the last 10 years have substantially impacted hospital system reimbursement. Placing an ever-greater importance on documentation accuracy and integrity, public payors such as Medicare and Medicaid have substantially reduced physician reimbursement. Similarly, private payors (Aetna, Cigna, UnitedHealthcare, etc.) have negotiated fee-for-service contracts, with treatment process and claims data submissions being screened with an unprecedented degree of scrutiny.

A fundamental problem plaguing the healthcare system lies in the ongoing struggle that healthcare providers face in their administrative dealings with insurers. Across the country, providers lose considerable time and money in their efforts to submit, dispute, and collect payment for their work. Each year, the provider revenue cycle strives to maximize the impact of $496 billion spent on billing and insurance-related costs (BIRCs), which includes a 50-percent administrative excess amounting to $250 billion annually (Crocker, 2006).

This article examines the evolution of clinical documentation integrity (CDI) processes and the importance of implementing a CAPD (computer-assisted physician documentation) solution in a health system to create workflow efficiencies and increase revenues. With payors demanding precise accuracy and flagging accounts for changed documentation, the considerable complexity and friction that complicates the process of getting paid has left physicians inundated with a surge of reimbursement-related documentation queries. These excess administrative responsibilities are outside of their patient-to-patient workflow, and result in a lack of timely and accurate compliance (January 2019). Reduction in revenues through more account denials and reduced reimbursement for inpatient and outpatient encounters for both professional and hospital billing can be countered effectively through direct and timely documentation advice to clinicians within their workflows. Since doctors spend most of their time in the electronic health record (EHR) conducting results review, documentation, and data entry, it is the ideal environment to capture their attention to notify them of a documentation query while they are still within the record of a patient (Sinsky et al., 2016).

Evolution in Resourcing CDI Processes

Around the turn of the century, health information management     (HIM) departments saw an opportunity to correct physician documentation to ensure that it was clinically congruent with insurance coding rules. HIM departments began hiring nurses and physicians to serve as clinical documentation improvement specialists (CDISs) to identify conditions that were evaluated, monitored, and treated over the duration of a hospital stay, but failed to be documented in a fashion that could be accurately coded (Towers, 2013) and  ultimately, reimbursed. Similar to how they operate today, these  CDISs combed through patient charts mostly in the inpatient setting, and generated paper queries and CDI forms to physicians to complete the chart documentation. The addition of this labor-intensive format initially brought along a great return on investment (ROI), demonstrating the enormous financial benefit of improved    documentation. However, with organizations growing, there was increased complexity in managing documentation, and a lack of   qualified individuals left the prospect of hiring additional human resources for the purpose of CDI unfeasible.

As the EHR and computerized decision support systems (CDSS) have evolved, with a greater capacity to manage medical records,   the paper-based approach of conducting chart review and generating queries by CDISs has migrated to an electronic medium. Commercial CDI applications are available to assist CDISs to perform a computerized chart review and enhance the efficiency of data abstraction. Usually, these products exist outside of the EHR, creating additional workflow problems for the CDI team, with double and triple data entry of the same data (severity of illness, or SOI, risk of mortality, or ROM, etc.), and necessitate untimely communication to the documenting doctor. Experiments with physician notifications through emails, text messages, and other communications back to documenting authors results in further delays in obtaining physician feedback to queries, since they are inundated with inbox messages, emails and phone messages.

The Arrival of Real-Time CAPD

The rapid uptake of the EHR over the past 10 years, along with advances in interface standards and clinical analytics techniques, has afforded the opportunity for a more integrated approach to CDI. The goal of maximizing clinician efficiency and optimizing revenue cycle is to apply CDI advice directly into the workflow process of patient care. Similar to how CDSS were integrated years ago with the decision-making processes of clinicians in mind, CDI advice can be seamlessly integrated into the workflow within the EHR prior to the completion of the clinical note.

Modern CDSS are available as computerized alerts and reminders, order sets, patient data reports, computerized guidelines, clinical workflow tools, and documentation templates (Sutton et al., 2020; Farion et al., 2013; Hoffman et al., 2011). The leading EHR software providers have opened their data and workflow interfaces and collaborated with vendors. The intent is to operationalize CDI as a CDSS that can improve compliance with clinical documentation requirements in a physician-centered fashion, learn from the clinician feedback to the automated queries, and catch the doctor before they sign their note so that corrected documentation is not denied by payors.

Vendors have collaborated with EHR systems to develop a “preferred workflow” approach, with initial efficacy described in the current article. CAPD is an EHR-integrated and physician-centric software solution that optimizes documentation in real time at the point of care, facilitating the construction of documentation that aligns with completeness expectations of third-party payors. Real-time suggestions in the EHR screen are delivered in conjunction with the note-writing and reviewing processes, along with unobtrusive task list and to-do reminders for a fluid preferred workflow.

EHR-Integrated CAPD Automation Query Notification

The preferred workflow consists of notifications for both direct and silent modes, which bring a physician to the automated query screen shown here. The screen has the following features, incorporating CAPD query feedback to the EHR once the note is written (saved):

  1. Query title is shown at the top of the left-hand panel, placed next to the note found deficient in documentation;
  2. Suggested text (up to six options), which is linked to underlying ICD-10 terms, can be added with a click wherever the physician places their cursor in the note on the right, which also prompts the physician to add that diagnosis onto the patient’s problem list, if warranted;
  3. Workflow options to agree, reject, state as clinically undetermined, ask me later, or defer the query;
  4. Evidence summarized and linked for the physician to evaluate the reason for the query, which allows streamlined review of the patient’s data;
  5. Info button, which opens an EHR window with highlighted abstracts of text and other data to reduce the need for lengthy chart review; and
  6. Items to address button under the note is highlighted to quickly show the physician how many queries are awaiting response.



Almost as important as the CAPD technology is the clinical reference knowledge used to create the automated query rules. THE CAPD vendors query library has content for more than 110 query types of various diagnoses and clinical subject areas. Healthcare providers start with the templates to create new query models, or use them as-is. In addition, a hierarchical relationship between query types allows generalized queries like presence of encephalopathy to be automatically replaced by more specific queries, such as metabolic encephalopathy, when certain evidence in the patient’s chart is present, mimicking the manual CDI query building process. The ability to notify providers to adjust clinical documentation in a timely and obvious way, prior to signing, also prevents denials from payors who flag claims submissions when they see altered or corrected documentation. The CDI notifications, as part of a preferred CDI workflow, are controlled by the EHR vendor, who can use their graphical user interface (GUI) to notify users with additions to their task list, sidebar menus, and on the bottom of each note.

As part of the real-time nature of the CAPD experience, clinicians (both CDISs and documenting physicians) become educated on the latest documentation requirements for both inpatient and outpatient clinical encounters. The quality of documentation is improved at the point of care, while the patient facts and physician decision-making are still in recent memory, reducing the administrative burden of wasted reorientation to the patient’s chart when a retrospective query is issued.

A requirement of participating in the analysis was the usage of CAPD integrated with the Epic EHR’s Advanced CDI License, including the NoteReader CDI module, from May 2019 to July 2021. One of the CAPD vendor’s clients, a healthcare system with multiple years of experience in reorganizing and optimizing their CDI processes, was proficient and interested in analyzing their outcomes related to the use of CAPD software. The system includes four community hospitals (labeled CH1-4 in the charts) and one academic medical center (AMC) with a total of 1,621 hospital beds, located on the Eastern Seaboard of the  United States. Prior to redesigning their CDI function with new processes and technology, they had a 40-percent gap in coverage of inpatient accounts. The selection of a technology vendor to automate their processes included the ability to cover 100 percent of patient encounters, with enough of the most frequently issued queries, and others that were less frequent, but still time-consuming to generate.

To improve accuracy and compliance with documentation queries, CAPD provides real-time queries within Epic as the physician writes their note. Notifications for CDI specialists and documenting physicians promote transparent communication between them in the clinical workflow system. One hundred percent of inpatient admissions and outpatient practice visits are run through the algorithms to provide a systematic review of all documentation. CAPD in Epic is a CAPD improvement tool that results in three times the number of queries and 3-5 times the number of responses from physicians.

It operates as an Epic-integrated, cloud-based NLP, with artificial intelligence (AI) and clinical workflow service that suggests improvements to clinical documentation, with the goals of increasing CDI coverage to the entire chart, improving physician documentation compliance, and reducing the number of manually-generated CDI queries. It uses a proprietary engine to locate medical terminology in clinical note text, along with relevant structured labs, vital signs, and medications, and applies algorithms from an advanced, configurable query knowledge base. CAPD identifies and presents documentation improvement suggestions at the point of service to physicians and CDI specialists through the Epic NoteReader CDI workflow, and prioritizes queries to either the CDI specialist or the physician for review. Clinicians access the suggestions and associated evidence as they write their inpatient or outpatient notes, and then respond to CAPD suggestions with a click, or enter new text in their note to complete accurate and compliant documentation.

Active query types used in the automated queries during the course of this analysis included the following direct mode queries (turned on Nov. 20, 2019)  and silent mode queries (turned on May 9, 2019).

The most frequently triggered queries during the two-year course of usage were as follows:

●       Malnutrition (Direct);

●      Obesity (Direct);

●       Heart Failure, Acute and Chronic;

●      Acute Respiratory Failure; and

●       Sepsis.

While not all the queries proved to be DRG impactors, there were other reasons to include the queries that didn’t impact DRGs: promoting an appropriate terminology to the documentation to comply further with reimbursement and quality guidelines, improving accuracy and calculation of SOI/ROM.

As shown by the response rates to CAPD queries, each query results in documentation addition, which results in new or more specific claims codes to the patient’s encounter. The implementation of CAPD resulted in a greater number of valid and completed queries, compared to the manual CDS query process. In its first six months of usage in silent mode, a range of 800-1500 queries per month were generated by CAPD, and the case activation frequency was high.

The newly introduced preferred workflow screens in Epic facilitated queries from CAPD to be presented efficiently to physicians in their workflow, and showed a significant and sustained response for approximately 20 percent of all queries. Overall physician compliance to respond to the CAPD queries showed a 321-percent increase in direct mode compared to silent mode, where the manual process of CDISs reviewing CAPD queries delayed sending the queries to the physicians, even though the queries were sent in real-time to the CDISs.

The advantage of a physician seeing a query in direct mode is that they are engaged while they are still writing and reviewing their note. They are presented the fact that a query is available for their review, and they know that they must return to their documentation to complete it, either in that moment or at a later time.

Most CDI-related feedback to physicians, which is patient and encounter-specific, is best done at the point of care, when the patient facts are still fresh in their minds and the time to reorient themselves to the patient’s chart is minimized. CAPD queries are delivered within 1.2 seconds of a note being saved, amended, or signed in the EHR system.

The newest approach to CAPD includes CAPD queries while physicians are actively writing their note (prior to the note being saved or signed in the EHR), which creates an additional CDI advantage because payors do not have an indication of documentation changes, reducing the tendency of denials due to changed documentation. While most CDI practices are currently CDS-centric, the current analysis shows that physician-centric solutions like direct mode CAPD have the greater impact on CDI and revenues, compared with silent mode and manual processes. However, silent mode, which is real-time to the CDIS, affords additional advantages to CDI, compared with manual CDI processes, because it ensures complete coverage of CDI for more than 110 different diagnoses (also referred to as query types), and delivers queries in the preferred workflow screens, which bring the physician directly to the note requiring their response. The process also includes reminders such as task list/to-do, and CDISs following up tracking are presented efficiency methods within workflow, which avoids inefficient email and in-basket messages.

At this point, it would be apt to answer a few questions:

1. What will be the impact of automating queries if the end result is a greater volume of queries for physicians to respond to?

Physicians will respond in a more timely manner, and in the context of their EHR workflow, making the overall experience of receiving and responding to CDI queries easier.

2. What will be the role of the CDIS in the CDI process, if many queries can be automated?

CDIS responsibilities will shift to more of a review function, rather than a primary initiator of queries. CDISs are essential to the creation and review of the query logic, which is implemented as part of the automated queries, and continue to be required to create complex manual queries and clinical validation of documentation.

3. What will be the impact of delivering physician notifications for queries and the actual queries themselves in a timelier manner than is currently the case, for manually produced queries?

Because of the timely nature of the queries, physicians are more likely to respond in timely fashion, which impacts the DRG while the patients are still admitted. The DRG shift results in a significant improvement of revenues, on the order of a 3-percent increase for cases receiving direct mode queries.

4. How will the organization measure the impact of a CAPD query tool, since one has not existed thus far, and there has been little experience in the industry with such a capability?

Continuous measurement and tracking of financial and workflow metrics are important to assess the impact of automated technologies. Examination of response rates by physicians, overall revenues, and DRG shifts are paramount to continued improvement of the system.

While other technology-enabled prioritization and abstraction tools provide  a mechanism for data abstractors (e.g. CDISs) to do their work in a consistent and scalable manner, CAPD is meant to be a complement    to the challenges facing those creating documentation. In the short term, CAPD assists physicians to optimize their documentation in a timely manner, at the point of care, reducing the documentation-related challenges that will  be later encountered. In the long term, it enables physicians to more closely align with the expected documentation practices of the period,  ensuring that physicians remain educated on the latest requirements. Greater deficiency identification with a proactive, immediate feedback component delivered within workflow imparts a revenue cycle advantage to organizations that can embrace automated query enhancement to their existing manual prioritization and query-building workflows.

Using a newfound physician-centric approach, CAPD is meant to streamline the clinical workflow while improving the quality of documentation that is available, when compared to the manual approach. Furthermore, consistent and compliant standards and knowledge bases can be regularly applied using computerized methods, reducing the risk of non-compliant processes. CAPD in silent mode provides automated queries, first to CDISs and then to documenting authors – the automatically generated queries enable the CDISs to better identify areas for documentation improvement in a shorter time, and prioritize identification of the patient account, compared to a totally manual CDS process.

In the silent mode process, CDISs are responsible for reviewing and “activating” the query to documenting authors. This semi-automated process is more efficient and financially beneficial compared to the manual process, and reveals that the automation translates to greater discovery of opportunities to improve the quality of clinical documentation. The increase in the count of activated and completed responses illustrates that the physicians also become increasingly receptive to the queries over time. In an era when physicians find themselves increasingly vexed by technology, the increased count of responses by physicians reveals that the AI benefits of improving CDIS processes are real.

Additionally, it was found that when shifting from silent mode to direct mode, where the CDISs were removed from notification workflow, the count of physician responses to identified deficiencies in documentation increased. The fundamental difference between silent mode and direct mode is the timeliness of the delivery of the query to documenting physicians. The silent mode CAPD is an upgrade from the manual mode, and facilitates an automated delivery of queries to CDISs; in other words, documentation improvement gaps are recognized by the software and delivered directly to the CDISs, who can review the recommendation and forward it to the documenting author, freeing them from the responsibility of initially identifying deficiencies and enabling a focus on more complex documentation challenges.

More importantly, the increase in the count of responses when direct mode is activated serves to reveal that there is a direct relationship between the speed at which a query is delivered and the likelihood that an improvement in documentation will take place: the faster the delivery of the query to the physician, the greater the likelihood that the documentation will be improved as the physician responds to the query in a timely manner. It was found to be 20 percent of the time within the immediate workflow screen of NoteReader CDI, and 80 percent of the time within subsequent documentation workflow within the patient’s encounter visit.

This has major implications for the direction that clinical decision support systems (CDSS) and CAPD should move towards: it corroborates the notion that AI solutions focused on improving clinical documentation should be organized in a physician- centric fashion, should be centered around improving the timeliness of the delivery of queries to physicians, and should be placed within the EHR workflow screens to be noticed best by the documenting physicians. The result is tens of millions of dollars in increased revenues from the accounts where CAPD is involved.

These findings collectively suggest that the implementation of CAPD integrated with the EHR workflows, as well as other physician-centric solutions that automate the generation of queries and cut the time it takes for a query to reach its documenting author, will lead to improved clinical documentation. Investment in the technology for automated, real-time queries can help HIM and revenue cycle functions build a reliable and timely revenue cycle, which front-loads the ability to capture accurate primary diagnoses, secondary co-morbidities, and justification for care.

This responsiveness by physicians, who are notorious for lack of response to computerized alerting in general, corroborates the notion that AI solutions focused on improving clinical documentation should be organized in physician-centric fashion, and centered on improving the timeliness of the delivery of queries to physicians within the EHR workflow screens to be noticed best by the documenting physicians.

Finally, the investment in the technology and training for automated, real-time queries reduces denials through the avoidance of addendums and other changed documentation, and more completely captures the patient story.

Programming Note: Listen live when Dr. Petratos reports this story on Talk Ten Tuesdays, 10 Eastern.

Share This Article

Facebook
Twitter
LinkedIn
Email
Print