Tackling Diagnostic Ambiguity
The CDC stats are sobering: Out of the 37 million adults in the United States with chronic kidney disease, only 10% know that they have it. Further, about 2 in 5 adults with severe CKD remain unaware they have CKD at all.
Why don’t 90% percent of the people in the US affected by CKD know they have it? Simply put, the diagnostic power of the American healthcare system has failed them. When it comes to diagnosing patients, the American healthcare system aims for quantity over quality.
A recent article in the Journal of the American Medical Association (JAMA) gets to the heart of the matter: “The U.S. fee-for-service health system pays for tests and treatments, not for diagnostic reasoning or accuracy, and does not make a distinction about whether payment is for the appropriate diagnostic tests or whether treatment selections are based on a correct diagnosis. … Fee-for-service reimbursement rewards empirical treatment; even if the initial treatment is incorrect, the next treatment is reimbursable, too. … A fee-for-service system does not pay more when decision-support tools are used and does not consider timeliness of diagnosis and initiation of efficacious treatment in how it pays.”
Lab Results and Labels
Most diagnoses rely primarily on lab results and labels. Consider a care manager who sees a patient with a hemoglobin A1C of seven. More than likely, that patient would be diagnosed with diabetes and placed on a long-term treatment path that shapes the rest of their healthcare - and life - trajectory. But what if there is something else affecting that measurement? A curious clinician might say: “That’s a one-time reading. Let’s look a bit more to see if that is really the right diagnosis before we label that patient a diabetic forever.”
When diagnosing a patient, physicians are taught to look for standard patterns: “When you hear hooves, think horses not zebras.” Oftentimes, the easiest answer is the right one. But what happens to our zebras?
With CKD, it’s very easy for a primary care provider to place a non-standard CKD patient into a generic CKD bucket. Say the patient has an eFGR of X. The PCP automatically assumes diabetic kidney disease and sets the patient on their pre-set path. But an experienced nephrologist might see that the eFGR has fallen too fast and decide to dig deeper. They might do a urine test and discover the patient has myeloma, which is what caused the CKD progression. That kind of diagnostic digging doesn’t happen very often, because it doesn’t follow the prescribed path. Diseases, just like humans, often deviate from the expected path. When we treat all patients like horses, we risk failing to identify our zebras.
A study on DNA sequencing at Columbia University Irving Medical Center found 10% of CKD patients had another novel disease that would potentially change their CKD therapy. Providing the wrong clinical management and therapies for 10% of patients means nearly 4 million people may be mistreated. Correcting to the right course of action could stop the progression- avert the worsening of conditions- for those who wind up with end-stage renal disease (ESRD). This means substantially reducing the number of people who have to go into dialysis each year. This brings focus to where we need to and can improve. How can we support our clinicians - who are already racing against time and lacking resources- in identifying and treating these rarer cases?
A Helping Diagnostic Hand
“For the past 20 years, there have been claims that artificial intelligence (AI) will surpass physicians, particularly for diagnosis,” the JAMA article states. But what if, instead of assuming AI will surpass physicians, it could be used to help them make better, fully informed diagnoses?
“The advantage of AI and machine learning is we can give physicians a way of reviewing patient records with computation power,” says Jung Hoon Son, MD, pulseData’s director of informatics. This means much needed support not replacing physician review. It can mean that for the first time physicians have what they need to become more successful in one place, at the patient level to empower them to help more people in time to avert worsening - or worse- uncorrectable conditions.
The issue isn’t a lack of data - quite the contrary. With EHR data,claims records, ADT feeds, pharmacy data, clinical laboratory results, and complementary data sources all available, the problem becomes drawing meaningful insights from a mountain of data points.
“A physician can’t read through 50,000 patients (worth of records) and weed out 100 of them every day.” Dr. Son asks. “It’s not feasible. But AI platforms such as pulseData can do that. We have a clinical-level algorithm that automatically filters out just the highest risk patients. You don’t have to task your clinical team with trying to find them. We reduce that burden and help healthcare providers make better diagnoses — and better decisions.”
Diagnostics is both an art and a science. With AI supporting their processes, physicians are able to cut through the noise and dial in on some of the more uncommon elements that make diagnosing patients so very complex. A partnership between clinicians and algorithms means physicians have more real and relevant data at scale; on hand and in time to improve more lives, with better diagnoses.