Contributors: Dr Sarah Gebauer, Magda Dubois PhD, Jasmine Balloch, Dr Michael Chuang, Dr Mishaal Ali, Dr Ellie Asgari, Dr Dominic Pimenta, Sal Khalil
In this white paper, we discuss quality reporting metrics and processes in the United States and provide physician insight from a recent summit focused on identifying challenges which could become opportunities for leveraging AI in quality reporting.
The summit brought together clinicians across from the United States with insights from TORTUS’s Data Science team.
We cover three key themes:
- Documentation burden of quality metrics
- Collation and visualisation of quality data
- Physician and specialty-specific reporting
Introduction
Quality reporting is essential for healthcare providers to enhance patient care, pinpoint areas for improvement, and ensure compliance with accreditation standards. Nonetheless, quality reporting in healthcare faces considerable challenges, which hinder its effectiveness. AI-driven quality reporting can also benefit individual patients. It allows for risk assessment and personalised care recommendations, taking into account patients’ unique medical histories, preferences, and needs. For instance, AI can assess a patient’s medical history, lifestyle, and genetic factors to recommend personalised treatment plans and lifestyle modifications to improve overall health. AI-driven quality reporting systems have the potential to directly impact patient care by improving the accuracy of diagnoses and treatments, enhancing patient safety, and reducing the occurrence of medical errors. For example, AI can analyse large volumes of patient data to identify trends and anomalies, allowing healthcare providers to make more informed treatment decisions and reduce adverse events.
Levels of quality reporting
Quality reporting occurs at three main levels: local, health system, and national or governmental. Local reporting may focus on departmental or hospital-level initiatives, as well as include health system and national-level data. Common examples of local reporting include presentations for departmental or service-line meetings, physician or group-level specialty metrics, or hospital initiatives like patient fall prevention. Health system reporting may encompass issues like mortality reviews or infection-control initiatives. Data is sent to CMS directly or via third party vendors, or to national quality databases like NSQIP or vendors like Vizient.
Some of this data comes from billing data, and some from EHRs. Much of the data sent to the databases then comes back to the hospital and health system with benchmarking, which is then presented at the various levels.
Documentation Burden of Quality Metrics
Problem
Physicians were united in their desire to decrease the documentation burden of quality reporting. One physician stated that she had to click the same 8 boxes for every patient, even though that information was already being captured in the chart. Another physician stated that he spends much of his time documenting these metrics rather than caring for patients. The manual entry of quality metrics can be time-consuming and administratively burdensome for healthcare providers. Physicians noted that nurses also have a very heavy quality documentation burden. For example, they must indicate if any patient has a pressure ulcer prior to admission in order to avoid a CMS penalty if one is noted at discharge.
Additionally, much of the abstraction to clinical databases and CMS is still done manually, often by specialised nurses. This represents a significant cost and leads to a delay in receiving analytics about the quality data.
Physicians also debated whether the metrics being pre-filled would actually decrease cognitive burden since they would have to read through them regardless, and whether they would feel comfortable with the AI completing all the quality fields without reviewing them first.
Opportunity
AI-powered voice recognition and natural language processing in the patient chart could capture when a quality metric is met and automatically complete that part of the documentation. For example, when a physician or nurse asks a patient if they smoked prior to surgery, an AI scribe could ‘hear’ that information, document it appropriately in the chart, and also check that box in the quality documentation. Additionally, much of the abstraction can likely be transitioned to NLP-based measures, though physicians noted challenges with copy and paste portions of the medical chart and inconsistencies of data within the chart.
Collation and visualisation of quality data
Problem
Physicians noted that much of the presentation of quality reporting requires nurses and doctors to discover, create, or collate information from disparate sources including local quality improvement projects, health system initiatives, and national benchmarking data. No tool currently exists to find this data, suggest insights and action plans, and create slides for presentation. One physician estimated that each presentation takes at least 6-8 hours per month to create. Additionally, physicians noted a long period to receive quality data from a hospital’s IT department, noting time from placing a ticket to receiving data to be 4-7 months, by which point a quality initiative may already have been completed. They noted that Epic’s Slicer Dicer and similar tools have the ability to provide some of this information but that most physicians don’t have the time or training to use them.
Opportunity
AI can streamline this effort by automating the discovery and aggregation of data from disparate sources. Machine learning algorithms can analyse local quality improvement projects, health system initiatives, and national benchmarking data to extract relevant information efficiently. AI-driven tools have the potential to not only find this data but also suggest insights and action plans based on the analysis of these vast datasets. Moreover, AI can generate reports and presentation materials, significantly reducing the time and effort required for manual data collection and report creation.
Physician and Specialty – Specific Reporting
Problem
Various medical specialties demand unique quality metrics that are usually developed by specialty societies or are hospital metrics that apply primarily to one specialty, such as time from arrival in the ED to appropriate stroke management. However, many physicians feel that completing these metrics doesn’t actually contribute to patient care or improve outcomes. Additionally, continuous evaluation of physician performance is required for Joint Commission certification. However, the OPPE/FPPE process can be cumbersome and inefficient and lacks metrics that could help physicians improve. For instance, conducting regular evaluations using traditional manual methods can consume substantial administrative resources and may not provide timely insights into a physician’s performance.
Opportunity
Identifying areas for professional growth and improvement is a continuous endeavour for healthcare practitioners. AI has the potential to support this ongoing learning process by analysing a physician’s performance data and offering personalised recommendations for improvement. Real-time assessment of physician performance – only with physician consent, of course – could identify opportunities for improvement that could enhance patient care.
Recommendations
To harness the potential of AI in quality reporting, healthcare institutions should consider adopting AI-driven solutions that cater to their specific needs. Furthermore, ongoing research and development efforts should focus on improving the capabilities of AI in healthcare quality reporting, ensuring that it continues to evolve to meet the ever-changing demands of the healthcare industry.
An ideal AI-based quality reporting system would encompass various essential components, including streamlined documentation processes, real-time feedback on performance, seamless integration with existing hospital systems, customisation options for individual healthcare providers, and the ability to deliver personalised care to patients. For example, the system could integrate with Electronic Health Records (EHRs) to provide real-time data, offer customised performance dashboards for physicians, and provide patients with tailored care plans based on their unique needs and preferences.
