Memorandum

JAMA on high prevalence of insulin-related hypoglycemia emergency department visits and hospitalizations – March 11, 2014

Executive Highlights

  • According to JAMA Internal Medicine, an estimated 97,648 emergency department (ED) visits for insulin-related hypoglycemia and errors (IHEs) occur annually in the US.
  • Notably, 29% of these events resulted in hospitalization. Insulin-treated patients >80 years were 2.5 times as likely to visit the ED and nearly five times as likely to be subsequently hospitalized for IHEs than those 45-64 years old.

In yesterday’s JAMA Internal Medicine, Geller et al. provided national estimates of insulin-related hypoglycemia and errors (IHEs) leading to emergency department (ED) visits and hospitalizations. Data was collected from insulin-treated patients seeking ED care and a national household survey of insulin use (January 2007-December 2011). Based on 8,100 cases, an estimated 97,648 ED visits for IHEs occurred annually, and 29% (28,317 events) resulted in hospitalization. Notably, severe neurologic sequelae (e.g., brain damage) were documented in 61% of ED visits for IHEs. Insulin-treated patients >80 years were 2.5 times as likely to visit the ED and nearly five times as likely to be subsequently hospitalized for IHEs than those 45-64 years old. The most commonly identified IHE precipitants were reduced food intake and administration of the wrong insulin product. The latter was particularly concerning to hear, since it seems somewhat avoidable with better labeling and user-centered design.

While the dangers of taking insulin are certainly widely acknowledged, this study serves as a critical reminder that severe events related to hypoglycemia are far too common and too costly – and many should be easier to avoid. There have been clear advances in insulin therapy, diabetes technology, and monitoring, but it’s still abundantly clear that reducing severe hypoglycemia is a major unmet need. This is a win-win for all concerned, especially the healthcare system – using the data above (97,648 visits and a 29% hospitalization rate) and a couple assumptions ($1,387 per ED visit and $17,654 per hospitalization; Quilliam et al., AJMC 2011), the total costs come out to $640 million. This is likely a conservative estimate – an oral presentation (279-OR) at ADA 2013 showed that hypoglycemia in type 1 diabetes in the US was associated with 20,839 hospitalizations and 284 deaths in 2009, with a total cost of about $1 billion (the average cost per hospitalization sounded very high to us at $46,039 and we’re not sure what contributed to this figure).

We hope payers and especially regulators increasingly recognize the true dangers and overall costs of using insulin. Speakers often mention the lack of incentives for payers to invest in next-gen drugs and devices, as the benefits don’t accrue until years later. Severe hypoglycemia is a major exception – at tens of thousands of dollars per hospital admission, the payback period is fast and the short- and long-term benefit to patients is enormous.

We are optimistic that newer drugs and next-gen devices could help cut severe hypoglycemia, particularly flatter basal insulins (e.g., Novo Nordisk’s Tresiba, Sanofi’s U300 glargine), more ultra-rapid-acting insulins (e.g., MannKind’s Afrezza, Novo Nordisk’s ultra-fast aspart, Biodel’s BIOD-123), the emergence of easier glucagon presentations (e.g., Xeris, Biodel, Latitude); increasingly automated insulin delivery (e.g., Animas, Medtronic, Roche, Tandem, plus scores of more automated systems still in development); more reliable and connected CGM (e.g., Dexcom Share and Gen 5, Medtronic Connected Care); and improved remote monitoring/detection algorithms (e.g., Telcare, WellDoc’s BlueStar, Glooko’s Population Tracker and Virtual Joslin).

Close Concerns Questions

Q: Do payers and regulators recognize the true dangers of taking insulin? What key data is missing to make the case clear?

Q: What is most critical for reducing severe hypoglycemia – better education, improved drugs, new devices, better monitoring, or something else?

--by Adam Brown and Kelly Close