An Epidemic of Septicemia?

While prevention efforts have largely been focused on chronic disease, spurred in part by the Centers for Disease Control and Prevention’s (CDC) insistence that 86% of healthcare dollars are spent on patients with chronic illness, it turns out that the largest, fastest-growing and possibly most preventable diagnosis is arguably the most acute diagnosis of all: septicemia.

Septicemia is any persistent systemic blood-borne bacterial infection, generally caused by contamination from an invasion of the bloodstream by an outside pathogen, and generally not easily addressed with antibiotics. Its many complications include death, 13% of the time. (There is some confusion about the various definitions of and distinctions among sepsis, septicemia and septic shock.)

The Statistics

The statistics below are drawn from the Agency for Health Research and Quality’s Healthcare Cost and Utilization Project (HCUP) database, and as such are easily replicated by entering ICD-9 “038.9” as “principal diagnosis.” The most recent year available, 2013, is the source of the screenshots.

According to the most recent statistic available, septicemia is not only #1 in total spending, but it’s also almost twice as costly as the next-highest ICD9.

Screen Shot 2016-06-16 at 11.02.42 PM

Taking the long view yields an even more concerning picture.  This is the same analysis, run for 1999 — the year the Institute of Medicine published To Err is Human, bringing errors and infections into the public consciousness.

Screen Shot 2016-06-16 at 11.07.59 PM

In 1999, septicemia was #20.  So septicemia is both the largest and the fastest-growing ICD9.  As it grew in incidence, it drew a little attention from AHRQ when it became #6 in 2009…but AHRQ has not mentioned anything since then. 

And yet, besides being #1, growth in septicemia has actually been accelerating. For contrast, we are graphing septicemia diagnosis rates against heart attacks (ICD9 410xx) and diabetes (ICD9 250xx), the current emphasis of most government and private sector (workplace wellness) prevention efforts. The dotted line represents an index of all non-birth admissions over the same period.

Screen Shot 2016-06-16 at 11.12.54 PM

The Magnitude of the Problem in Perspective

The number of suicides in the US has increased from 29,000 to 42,000 since 1999, a 24% increase adjusted for population.  The fastest-growing segment, middle-aged men, has suffered a 50% increase in the rate over that period.

Not to underestimate the implications of or be insensitive to that problem (which included one of my relatives), but the septicemia rate is 20 times higher and has jumped sixfold. So however much of a national priority it is to understand what is driving the suicide rate, it should be a much greater priority to understand the dramatic increase in septicemia.

Or put another way, while you are reading this article, 5-10 more people have been diagnosed and probably one person died.

What’s causing this dramatic increase?

Many septicemia cases — complications of burns, needle contamination, AIDS — originate outside the hospital. However, those cases are flat-to-declining. Burn admissions have remained unchanged at 32,000/year this century.  Non-sterilized needle-sharing is presumably also on the decline, while AIDS has likewise responded to improved treatments this century. Further, if cases originated outside the hospital one would expect discharges from the ER to be increasing at the same rate as admissions to the hospital from the ER, but treat-and-release ER visits have increased far more slowly.

Another hypothesis could be that due to the progress made in prolonging life for people with chronic disease, patients are dying of this rather than directly due to their chronic disease.  If that were the case, inpatient deaths due to septicemia would be high and increasing. Instead, inpatient septicemia deaths have declined, according to HCUP, from 22% of all septicemia diagnoses to 13%.

Yet another hypothesis could be that these diagnoses are some kind of Medicare revenue-maximization coding scheme. That is unlikely. Starting around 2001-2002, CMS became pretty good at shutting those down, while this one has been going on for years.

Failure to seek care for suspected infections in a timely way due to insurance status is another hypothesis worth exploring. It’s possible that people are waiting longer to seek care if they don’t have insurance or have high-deductible plans.  Using lack of insurance as a proxy for lack of coverage, a suggestive statistic is that roughly 3.7% of all non-maternal admissions are septicemia for uninsured people, while the comparable statistic for insured people is 3.4%. A noticeable difference to be sure, but far from a definitive explanation. (It is not possible to track high-deductible plans in the HCUP database.)

The only remaining hypothesis is that hospitals have increasingly harbored increasingly hardy bacteria that increasingly find their way into patients’ blood, at a rate more than six times greater than in 1999. (The Centers for Disease Control and Prevention writes that hospital-acquired infections have been declining, but didn’t reconcile that finding with the definitive HCUP Nationwide Inpatient Sample, which reports only roughly a 2% standard error. In research, it’s one thing to explain why a conclusion conflicts with the “hard data.” It’s another thing to ignore that conflicting hard data altogether.)

This hypothesis is a bit different from “superbugs,” which the CDC is addressing. True superbugs are still rare. Septicemia is very common.

Response to date and proposed solutions

A specific set of infections, such as catheter line infections and post-surgical infections for certain surgeries, fall into the category of “serious, reportable events” (formerly and still colloquially referred to as “never-events”) that Medicare won’t pay for.  (They still pay for the cases, but not for the extra resources needed to control the infection.)   However, these events don’t cover remotely the entirety of septicemia.

Septicemia is the second-biggest diagnosis (behind knee replacements) for the employed population, and arguably the most preventable. While some major companies and regional business coalitions have been working closely with Leapfrog Group and are addressing patient safety issues and infections in many ways, including using Hospital Safety Scores to guide network and benefits design, even Leapfrog can’t track septicemia yet. 

Instead, most large employers are focused instead on wellness programs, widely believed to be ineffective.  Employers should refocus their efforts away from these programs and instead start demanding safety improvements from their providers. It might be something as simple as not paying for certain medical errors (and paying a bit more for other cases, to compensate), as Medicare does.

Medicare, in turn, should expand its never-event categories to include more infections. This would put more risk back on the hospitals, and should motivate them to emphasize safety.

Next, caregiver-to-patient transmission of bacteria must be curtailed.  Hand-washing simply isn’t taking place as often as it should.  Notably, the increase in infections has coincided with the increase in the use of hand sanitizers. Perhaps sanitizers are being used as a substitute for hand-washing instead of a complement…and don’t work as well. The FDA is raising this issue too. Apparently the alleged safety and effectiveness of these dispensers are based on 1994 standards.

Finally, everyone could name hospitals known for cancer or transplants. But could anyone name hospitals known for safety?  One of the safest hospitals, according to Leapfrog, is Virginia Mason Medical Center.  Being a leader in cancer treatment or transplants creates high visibility. The large, high-technology investments are often displayed in advertisements. But being the leader in safety?  That means doing a million little things right that only the cognoscenti would ever notice. If somehow more attention were paid to “safest” rather than “highest-tech,” the basis for hospital preferences could change. Yet Congress has consistently blocked more transparent hospital reporting requirements.

Following the well-publicized and completely preventable death of a 12-year-old boy from sepsis, New York State implemented Rory’s Regulations, which require hospitals to implement a set of protocols to predict, identify, communicate and treat bacterial infections. Other states should follow suit, without waiting for a similarly well-publicized tragedy to spur them to action. (In 2013, the first nearly complete year of Rory’s Regulations, New York septicemia cases still rose 15%–but would have risen 17% had they simply tracked the country as a whole. So the law has already made a difference, though not a major one.) 

Unfortunately, since no patient expects to get an infection, no one selects a hospital whose expertise is avoiding them. Hence, few hospitals attempt to “compete” with VMMC and a handful of others for this distinction nationally.  Locally, though, business leaders could highlight the safest hospitals in their regions, and perhaps steer more employees towards them.

A better working hypothesis, and more detailed solutions, are for others to propose and implement. The focus of this column is to identify this as an unappreciated but presumably addressable epidemic. It should be treated as such.

Al Lewis is the CEO of Quizzify.com

Categories: Uncategorized

10 replies »

  1. Yours is the hypothesis we all worry about. We could be doing “everything right” and still have more sepsis. As to the CMS question, I can think of several reasons. They may have decided to push hospital to actively pursue sepsis as a policy goal, so the have little interest in tamping down this surge in codes. Alternatively, I would note that it just takes them a while to respond. Last, maybe they think this is real. Would have to ask them.


  2. I probably shouldn’t be posting this because it sounds like heresy, but there was a recent article in Science or Nature that suggested that the latest thinking on antibiotic resistance is that we shouldn’t be relying on the drug so much to do the bacterial killing. We should be trying to kickstart the immune mechanism into finishing the killing by using slightly less than the old recommended doses of antibiotics. Viz. we should not give the full dose for seven or ten days. We should want the bacterial antigen to stick around a bit and to enhance the immune system to do the final killing and healing. If we use multiple antibiotics and give the full dosages, we are getting rid of the bacterial antigens too soon before the patient’s own immune system can really be kicked in to full response against the bacterium.

    I’m still trying to find this citation, and I may have some of the above incorrect.
    Incidentally, bacteria can re-aquire their sensitivity to antibiotics also; over time this has been seen.

    I wish we could beef up our research on phages.

  3. Coding isn’t a science, it’s very fluid, “let’s pick the code that best fits”. Presumably this data was gleaned from the ICD9 days, where a very vague constellation of findings would be coded as sepsis. Also, this isn’t taking into account that many more immunocompromised people are walking around now for the past several decades–and they succumb to sepsis simply because they don’t have the T cells to fight. No amount of hand washing will prevent this. Bear in mind the explosion in transplants, chemotherapeutics, immune modulators for rheumatoid arthritis, psoriasis, IBD, and the millions with diabetes….
    Glen Frey died of sepsis, yes, but it came from the immunomodulators for his RH weakening the immune system and producing sepsis. So don’t lose sight of the forest.

  4. But the death rate is also way up (though not as fast as the dx rate, so some portion might be coding). And assuming you’re right, why hasn’t CMS noticed this? Why are hospitals still allowed to just plop “septicemia” in as the Primary Dx and get more money?

    Just a hypothesis–we are doing more things to prevent infection but the bugs themselves are fighting back — fewer get run the safety protocol gauntlet but the ones that do are hardier.

  5. Like Brad, I think this is probably mostly coding. It is also kind of odd that sepsis would be increasing now. While hand washing is not perfect now, it was pretty uncommon in 1999. We have central line protocols now. We wipe valves before injecting now. We do so much more than we used to to try to prevent infections. If it not coding, then maybe all of this stuff we are doing to try to stop infections is actually making things worse.


  6. Thanks for the comment. Diagnoses have increased more than sixfold while deaths have increased fourfold–hence the % of people with septicemia dying is down by about 40%. Therefore there could be overdiagnosis but as mentioned in my other comment doesn’t remotely explain all of it, not even close.

    And why is CMS making it so easy to get paid more money for infections? That doesn’t seem like it would make any sense at all.

  7. I am seeing over diagnosis secondary to the new criteria and the ease of getting admission reimbursed under that diagnostic group. Has death from sepsis been on the rise? If it is falling, certainly could be over diagnosis.

  8. Hi Brad. Thanks for your comment. I only looked at PRINCIPAL diagnosis, not all listed diagnoses, to avoid exactly what you are describing. And maybe you’re right and I’m giving CMS too much credit for vigilance. They read this blog. They should defend themselves here

    I would have a hard time believing they are completely asleep on this one. If they are this should wake them up.

    One piece of data suggesting you are partially right (but it only explains a portion of the variance ) is the decline in deaths as a % of diagnoses That would explain some but far from all of the increase in diagnoses as “easier” cases being primary-coded.

    As I mentioned, I am not finding the answer. Just raising the question

  9. Al
    What makes you think CMS has shut coding issues down? How hospitals and providers characterize and identify sepsis has gotten more attention in the last five to ten years than I have ever witnessed. I think you simplify the problem: it’s coding–both of the good and bad sort, as well as the changing spectrum of illness in the hospital.

    While the definition of sepsis changed this year–and not accepted by all, the prior definitions of sepsis, severe sepsis, and septic shock are highly sensitive (half the folks in the hospital can be diagnosed with it based on vitals alone), it’s still and imprecise science. Coders call all the time and ask, “Doctor, is this sepsis?” Number one question.

    Sepsis pays more. Hospitals have increased vigilance to identify, again, with more focus on that ICD–for myriad reasons. You cite a lot of meaningful stuff above, but give too little weight to the codes themselves.