With COVID-19 still pressuring hospital revenues, providers are looking for ways to improve financial health. One area in which providers may find substantial savings is unwarranted clinical variation, which results in more than $760 billion of avoidable healthcare costs, a number that’s still growing. Variations in clinical practice occur because of delayed implementation of or updates to evidence-based best practices, misaligned incentives in fee-for-service environments, patient demand for tests and treatments and entrenched workflows.
Those billions of dollars include waste from unnecessary tests and treatments as well as variation in clinical supplies ordered. Clinical variation results in patients unhappy with longer lengths of stay and higher bills that don’t necessarily equate to better outcomes. Further, providers miss financial and quality goals — essentially leaving money on the table.
Hospitals have much of the data they need to implement clinical variation management (CVM) in their own financial and electronic health records. Machine learning (ML) and artificial intelligence (AI) tools can help them mine that data at scale and pair it with industry standards and evidence-based practices to identify evidence-based pathways that will help eliminate unnecessary tests and procedures while improving outcomes. These benefits will help persuade clinicians to support CVM efforts. Further, providers monitoring the efficacy of their COVID-19 treatments are laying the groundwork to expand data sources for CVM beyond hospital walls. The timing is optimal to uncover and address variation and wasteful spending.
Uncovering unwarranted clinical variations
The first step toward CVM is automating the grouping of patient cohorts with similar diagnostic codes, treatment plans or chronic conditions to identify clinical pathways. That process uses an organization’s own data so that the analysis reflects the provider’s current practices and financial and patient outcomes. It results in a quantitative comparison of sources and evaluation of the impact of all clinical variations. Typically, we see that about 80% of patients can be cared for through a standard evidence-based practice.
A physician prescribing a second-line antibiotic because the patient’s allergies, lab results and medical history contradict first-line choices is an acceptable variation; prescribing that antibiotic due to preference without supporting clinical data or evidence is unwarranted variation. Ordering daily post-operative chest X-rays because the patient is intubated or exhibits symptoms requiring monitoring is acceptable variation. Failing to discontinue the X-rays after the patient is extubated because the X-rays were part of the original order set is an unwarranted variation.
Such insights typically reside in healthcare system data stores, which have grown at a rate of 878% since 2016. Yet much data within a provider enterprise goes unused for advanced analytics. In addition, physicians may find it overwhelming to keep up with today’s medical advances. In 1950, researchers estimated that the volume of medical knowledge would double every fifty years. Now the volume of medical knowledge is expected to double every seventy-three days.
The growth in medical knowledge combined with the sheer volume of hospitalizations, surgeries, office visits and more that occur daily across a large health system clearly makes CVM a task for machine intelligence. Machine learning (ML) tools rapidly analyze electronic medical record (EMR) and financial data that represent thousands of patient procedures and millions of individual events. ML analysis may also incorporate benchmarks from other institutions to help distinguish evidence-based practices leading to acceptable variations from entrenched legacy workflows that create unwarranted variations.
Integrating those findings into EMRs and modern data ecosystems facilitates rapid deployment of insights about variations to physicians, nurses, pharmacists, administrators and executives. Systematically created timely and detailed reporting provides an unbiased mechanism for tracking adoption of and adherence to standardized clinical pathways. Intuitive dashboards updating in real-time make it possible for clinicians to understand their individual performance compared to peers and change practices to help eliminate unwarranted variation. Drilling down into data allows physicians to see precisely where they are not in compliance with a care pathway, such as being slow to write discharge orders or consistently ordering tests or treatments that do not align with pathway evidence.