What Insight Does Big Data Offer Into Adrenal Insufficiency?

Richard M. Plotzker, MD


April 08, 2022

Although adrenal insufficiency is uncommon, it is not so infrequent that we never encounter patients who have it. The sometimes-tricky presentations, along with the glorious responses to therapy, provide many endocrinologists with a blend of fascination and professional gratification. We think of the condition as a possibility more often than we find the real thing, doing cosyntropin tests that yield normal results for a variety of presentations, from critical illness to hyponatremia. We inherit people with panhypopituitarism and figure it out ourselves when an adult survivor of cancer who received central nervous system radiation years ago begins to gradually underperform.

Despite our fascination with the condition and the frequent challenge of figuring out a clinical puzzle, few of us see more than a few dozen of these individuals, each unique in glucocorticoid and mineralocorticoid replacement requirements, and often without reliable laboratory tracking of our therapeutic judgments. These people have a way of generating calls to our offices, either from their own setbacks or emergency department (ED) doctors and surgeons who are skittish about elective outpatient surgery. But despite the central role of the endocrinologist in ongoing care, we have surprisingly little data on how these patients fare over time as they navigate the increasingly baffling maze of medical care of chronic diseases. Until now.

A group in Alberta, Canada, performed a review of medical encounters among patients with primary or secondary adrenal insufficiency over a 5-year interval. An accumulation of data from a few central locations combined with computerized sorting capabilities has enabled better exploration of relatively rare conditions like adrenal insufficiency, which, when pooled, exposes common paths this condition takes. We sense from our phone calls that these people need emergency care, but the frequency of setbacks varies from never to a lot. Thus, large clinical databases have made health planning less laborious and more reliable but, as this study reveals, not exactly glitch-free.

In Alberta, each person is assigned a personal identification number (PIN) that remains constant for all physician and pharmacy encounters. This allowed the research group to match diagnostic codes for primary and secondary adrenal insufficiency, and to capture individuals receiving multiple refills of glucocorticoids and mineralocorticoids. Because there were three active databases, however, some encounters appeared in multiple places, which the authors corrected to avoid overcounting.

They found 2637 unique individuals seeking outpatient or ED care who had a coding diagnosis of primary or secondary adrenal insufficiency. An estimated prevalence of 839 per million Alberta residents exceeded the frequency of other national or regional estimates.

The limitations of teasing out the realities of medical care from administrative codes may be best illustrated by the researchers' attempts to sort out how much medical care these people received and for what purpose. On average, the claims and encounter data show 2.2 ED visits and 17.8 outpatient visits per person with adrenal insufficiency. However, the diagnostic codes show that only a small fraction of these have an adrenal or pituitary diagnostic code for most episodes of medical care. The inclusion of physical therapy and radiologic procedures as outpatient visits makes it difficult to assess how the adrenal insufficiency affected other elements of health.

The ED visits may be more revealing. Adrenal or pituitary codes rarely appeared as the primary diagnosis; however, among the top 10 were abdominal pain, volume depletion, electrolyte disorders, chest pain, and abscesses, often the sequelae of either adrenal insufficiency or its treatment. The authors determined who the patients were by their PINs and got some sense of what happened to them by tabulating the complications that resulted in emergency care. Unfortunately, the top 10 primary ED codes comprised only a small fraction of the 2699 different diagnostic codes submitted, so it remains unclear how much ED attention was directly or indirectly a consequence of hypoadrenalism, only that these patients seek emergency care considerably more frequently than others in Alberta.

This analysis of administrative data confirms what most clinicians providing longitudinal care to these patients already suspected: they need more than average amounts of physician attention and they sometimes need rescue from crises unique to adrenal impairment. However, as we depend more and more on large amounts of aggregated data to isolate specific conditions and assess the treatment history, this study also exposes some of the limitations of how we code, classify, sort, and retrieve these patient encounters. Even data that should be binary, such as you went to the doctor or you didn't, have a way of slipping through the cracks, as not all medical visits in Alberta needed to be centrally reported.

As much as physicians, myself among them, dislike our secondary roles as coding clerks, the defaults permitted, including "follow-up visit," often give no medical information yet are sufficient for payment. Having 2699 unique ED visit diagnoses impedes sorting them into the manageable categories we depend on as clinicians to understand diseases and as planners to assess public needs. Even things as basic as why the adrenals or pituitary failed, whether by immune, surgical, trauma, or medication (including opiate use) causes, are not adequately captured for later assessment.

Big Data and the ability to access it has expanded our capabilities but, as this study demonstrates, more can be done to make it reveal to clinicians and planners what we most need to know. How many people with other uncommon but threatening diseases still drift in and out of our exam rooms and EDs, controlled by the appropriate specialist but shared with many others? It looks like we can figure out from administrative data more than we knew before but less than we would seek to find out.

Follow Medscape on Facebook, Twitter, Instagram, and YouTube


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.