May 2014

What Sepsis Numbers Do and Don’t Tell Us
By Lisa A. Eramo
For The Record
Vol. 26 No. 5 P. 22

Sepsis data are talking, but should the health care industry be listening?

Big Data has become a big commodity in the health care industry. The government, payers, researchers, providers, and even patients are interested in what data are trying to tell us about the quality of care provided. EHRs have allowed the complicated manipulation of these data, yet there continues to be limitations on the conclusions that can be drawn.

Consider sepsis, one of many diagnoses that researchers and others are interested in from a data perspective. Sepsis is the 10th leading cause of death in the United States, and the majority of cases develop after admission, often as a result of postsurgical complications. Can the data gathered from various sources help shed light on this serious problem or do they merely serve to skew reality?

Defining Sepsis
The American College of Chest Physicians and the Society of Critical Care Medicine have established the following four levels of sepsis:

• Systemic inflammatory response syndrome (SIRS): hypothermia (temperature lower than 36˚C/97˚F) or fever (temperature higher than 38˚C/100˚F), tachycardia (heart rate of more than 100 beats per minute), tachypnea (more than 20 breaths per minute) or hypocapnia (arterial CO2 below 32 mm Hg), and leukopenia or leukocytosis (white blood cell count that either is too high or too low). SIRS does not include a confirmed infectious process.

• Sepsis: SIRS in response to a confirmed infectious process.

• Severe sepsis: sepsis plus organ dysfunction, hypotension (low blood pressure), or hypoperfusion (insufficient blood flow) to one or more organs.

• Septic shock: sepsis with persisting arterial hypotension or hypoperfusion despite adequate fluid resuscitation.

However, many variations of these clinical criteria exist, with some more detailed than others.

The Surviving Sepsis campaign defines sepsis as “the presence [probable or documented] of infection together with systemic manifestations of infection.” Severe sepsis is defined as “sepsis plus sepsis-induced organ dysfunction or tissue hypoperfusion.” The systemic manifestations of infection can vary and are specific to the clinical scenario, making the definition difficult to apply in a universal fashion. The campaign has published specific clinical criteria as part of its International Guidelines for Management of Severe Sepsis and Septic Shock.

To complicate matters, diagnostic criteria for pediatric patients are different. According to the Surviving Sepsis campaign, criteria for pediatric patients include “signs and symptoms of inflammation plus infection with hyper- or hypothermia (rectal temperature > 38.5˚C or < 35˚C), tachycardia (which may be absent in hypothermic patients), and at least one of the following indications of altered organ function: altered mental status, hypoxemia, increased serum lactate level, or bounding pulses.”

“Sepsis is a clinical diagnosis based on a physician’s clinical judgment and medical decision making,” says Glenn Krauss, BBA, RHIA, CCS, CCS-P, CPUR, C-CDI, CCDS, revenue systems manager at Accretive Health. “I’ve seen patients meet two of the four criteria, but the context of the patient’s presentation doesn’t support sepsis.”

Ultimately, experts say a sepsis diagnosis is subjective and thus inconsistently applied. To add to the confusion, sepsis is an example of a clinical term that is frequently lost in translation when being coded. Physicians often use terms such as “bacteremia,” “urosepsis,” and “SIRS” that don’t equate to sepsis in coding terms.

Nevertheless, the reality is that research and payment are based on the words that physicians use and the codes that coders assign. Therefore, any research is only as accurate and thorough as the data on which it’s based.

Documentation Dilemma
Because clinical care comes first, documentation isn’t always a priority for physicians, says Steve Claypool, MD, vice president of clinical development and informatics at Wolters Kluwer Health. “Physicians, by tradition, will only identify sepsis in the chart when a patient is getting so ill that he or she is starting to die,” he says. “That’s usually the first time that you’ll see documentation of sepsis or septic shock.”

As a result, the health care community is missing out on potential lifesaving information. “If we’re able to identify patients who are starting to have those abnormalities in the data suggesting they’re becoming more ill, we can watch them more closely and get them on some screening tests and plans,” says Claypool, who is working with others at Wolters Kluwer Health to develop software that can accomplish this task by analyzing real-time data in the EHR.

“Sepsis is a big problem because it can progress very rapidly and lead to death or prolonged hospitalization, cost, and discomfort,” he continues. “When patients develop more severe sepsis, even the delay [of treatment] of an hour can start to affect mortality rates. Prompt antibiotics and IV fluids have a big impact on survival.”

Why Data Matter
Aside from the potential clinical ramifications, documentation also affects research. Consider a recent sepsis-related study performed by the University of Pennsylvania’s Perelman School of Medicine. The study, which used US county death data from the 2010 Multiple Cause of Death data files, compiled by the National Center for Health Statistics (NCHS), along with 2010 Area Resource File demographic data, identified hotspots for infection and severe sepsis-related deaths. These hotspots were defined as regions where the infection death rate was significantly higher than the national mean and surrounding counties.

The study found notable hotspot clusters located in the Midwest, mid-Atlantic region, and the South while also identifying 157 counties across the Southwest and Mountain states deemed to be “coolspots.” Unlike previous research, the University of Pennsylvania’s study identified trends on a national rather than state level.

“It’s really just preliminary work,” says David Gaieski, MD, an associate professor of emergency medicine at the university and a study author. “The incidence of sepsis should be really uniform across the United States once you adjust for the elderly population. If it isn’t, what’s going on? We don’t know what’s driving this.”

Coded data are, in all likelihood, a big driver in the analyses, Gaieski says. “Most of the big statistical studies that have talked about incidences of sepsis have all used coding data because there’s very little other way to get at this information,” he says.

But coded data are complicated because sepsis requires coders to report multiple codes in order to fully capture the diagnosis, says Gail I. Smith, MA, RHIA, CCS-P, president of Gail Smith Consulting. For sepsis due to an infection, coders must report one code for the underlying systemic infection or causal organism. For severe sepsis, they must report one code for the underlying infection, one code for severe sepsis (with or without septic shock), and another code for the specific organ dysfunction.

Not all coders consistently query for sepsis, which can skew the data and compromise their integrity, says Allison Errickson, CPC-H, director of coding compliance at ProVation Medical, part of Wolters Kluwer Health. “Coders don’t want to question a physician’s clinical judgment,” she says.

The University of Pennsylvania study relied on mortality codes representing the underlying cause of death. Experts say even these codes are complicated. If a physician doesn’t identify sepsis throughout the record, chances are he or she won’t consider it as a potential cause of death either, even when it may be. Even if sepsis is documented throughout the record, some physicians may not feel that the condition is the actual underlying cause of a patient’s death.

Consider this example: The victim in a traumatic accident is rushed to the emergency department. After a period of time, he starts to develop acute respiratory distress syndrome and sepsis, and subsequently dies. Some physicians will identify the cause of death as trauma, others as acute respiratory distress syndrome, and others as sepsis, Claypool says. The death certificate could reasonably reflect any one of these diagnoses.

The World Health Organization defines the underlying cause of death as “the disease or injury which initiated the train of morbid events leading directly to death, or the circumstances of the accident or violence which produced the fatal injury.”

However, even this definition is subjective. Physicians choose—sometimes at random—what condition caused the patient’s ultimate demise. They document this information on the patient’s death certificate, and it’s subsequently coded by the NCHS using an automated software program. When researchers use these mortality data for research purposes, they may not be entirely accurate or paint the full picture. In fact, sepsis deaths actually may be underestimated.

Gaieski says researchers at the University of Pennsylvania tried to compensate for these variables by defining sepsis deaths more broadly. Unlike coders who are bound by coding rules and guidelines, researchers have more flexibility to interpret physician documentation. For example, the Penn researchers defined sepsis-related deaths to include those that were attributed to an infection. These infections included septicemia, respiratory (eg, pneumonia), abdominal and gastrointestinal (eg, appendicitis and diverticulitis), cardiac (eg, endocarditis), kidney and genitourinary (eg, pyelonephritis and pelvic inflammatory disease), and neurologic (eg, meningitis). They also included deaths due to influenza, acute bronchitis and bronchiolitis, and pneumonitis due to solids and liquids.

Data Influencers
HIM experts say the validity of wide-range sepsis results are affected by other factors, including the following:

• Hospital size: Severely septic patients may be more likely to be transferred to larger facilities that can provide more comprehensive care. If these patients expire, the data will reflect poorly on these facilities.

• Clinical documentation improvement (CDI) programs: Some hospitals may have a more robust CDI program. As a result, coders and CDI specialists may be more diligent in capturing sepsis, which could skew the data unfavorably and not necessarily reflect on the clinical care.

• Complication or comorbidity (CC) or major CC (MCC): A discussion of sepsis data would be incomplete without examining the overall population health, including CC and MCC conditions. Patients with a higher risk of morbidity and mortality, particularly the elderly, will develop sepsis more frequently. This could account for the higher volume of diagnoses in certain regions. Regional differences in health behaviors, diet, socioeconomic status, genetics, or environmental exposures also may potentially alter the risk of sepsis.

• Execution of sepsis treatment protocols: Treatment for sepsis includes the administration of IV fluids, antibiotics, and vasopressors. Claypool says regional variations in how these treatments are rendered could account for statistical differences.

• Auditors: Because of their higher reimbursement, sepsis diagnosis-related groups (DRGs) have been targeted by recovery auditor contractors. They’re specifically looking for clinical indicators of sepsis, Krauss says, adding that some organizations may code and document more conservatively simply to avoid becoming the target of an audit.

• The limitations of the NCHS’ software: The software that provides automated coding of the underlying cause of death isn’t sophisticated enough to compare the cause of death listed on the death certificate with the actual medical record documentation—something that must be done to validate the data, says Minnette Terlep, BS, RHIT, vice president of business development at Amphion Medical Solutions.

For example, a physician may document urosepsis as the cause of death, but the record indicates that sepsis technically is the cause of death. “I’m not sure whether the NCHS software would equate urosepsis to sepsis,” she says. “My gut feeling is that it does not. So in areas where physicians are using that term, it could explain some variations in the data.”

In ICD-10, the term “urosepsis” is nonspecific and not considered synonymous with sepsis, says Anne M. Pavlik, RHIT, an education consultant at Amphion Medical Solutions, adding that the condition has no default code in the ICD-10 Alphabetic Index, and coders should query for clarification if used by a physician. Is it a localized urinary tract infection (UTI)? A UTI with bacteremia (ie, a positive blood culture without clinical evidence of sepsis)? Generalized sepsis that originates from a UTI?

Where to Go From Here
Each of these variables makes it difficult to draw conclusions, Claypool says. “You probably can’t rely on the coded data across institutions that have different documentation patterns,” he notes. “But any individual organization can use the coded data to compare year to year.”

Claypool cautions organizations from drawing quick conclusions based on the University of Pennsylvania’s research. “What I gleaned from it is that it’s interesting and that someone needs to take a closer look,” he says. “Someone needs to dig in and see if there’s a real reason that the coding might be done differently or not. Is there a difference in care or is there a difference in coding?”

In the past, the Office of Inspector General has found as much as a 21.2% error rate associated with DRG 416 (septicemia), which, according to Terlep, suggests that inaccurate coding does play a role.

Experts say the bottom line is that studies such as the one published by the University of Pennsylvania can significantly impact hospitals. Not only can the research draw the attention of recovery auditors and others, but it also can affect perceptions about the quality of care being provided.

“If I was a hospital administrator and I saw this, I’d be concerned,” says Renée Brown, RHIT, CCS, corporate manager of coding operations at HRS, adding that organizations must be aware of the data they’re producing regardless of whether coders generate the data or the NCHS automatically generates the data based on physician documentation. This is the only way that organizations can survive—and thrive—in a data-driven health care environment, she says.

“If you can’t answer the questions behind the numbers, then the numbers don’t really have much meaning,” Smith says. The good news is that HIM can at least help explain variations in the data. To receive greater clarity from the data, HIM professionals should consider the following questions when reviewing sepsis data:

• Does the CDI program improve the quality of sepsis documentation? CDI professionals may overquery for sepsis, resulting in the condition being listed throughout the record and ultimately on the death certificate when it’s not justified, effectively skewing data and research efforts, Krauss says. They should ensure that physicians document clinical indicators as well as their rationale for a sepsis diagnosis.

• Do coders ensure clinical coding accuracy when capturing sepsis? Coders increasingly are tasked with reading between the lines to infer the true clinical picture. Do clinical indicators point to a sepsis diagnosis even though it may not be documented? Krauss says this is the question every coder should be asking to ensure correct data reporting.

Brown says coders also should be mindful of instances in which physicians use the term “sepsis” loosely. “You want to show exactly what’s going on with the patient. We really have to make sure that we have all of the clinical indicators to support the diagnosis, even if the physician is saying it,” she says, while recommending coders double-check each other’s work, particularly for tricky sepsis cases with unclear documentation.

Errickson agrees: “You shouldn’t be coding this diagnosis unless that causative relationship is documented clearly.”

• Do physicians understand the importance of their own documentation? Physicians must understand “there are no decisions made in this country without looking at data,” Smith says. Death certificates particularly are important. “Do they see [the certificate] as a bureaucratic step or important clinical piece?” she adds, pointing out that although the document is basic, researchers and others are drawing conclusions from its contents.

Physicians slowly are realizing that their documentation matters, Terlep says. “As our health care delivery system moves forward in new ways, physician profiling of cost-effective and high-quality provision of care is going to really increase,” she notes. “I think the improvement in documentation is going to be physician driven instead of coder or CDI specialist driven because physicians will realize they have a major stake in this as well.”

— Lisa A. Eramo is a freelance writer and editor in Cranston, Rhode Island, who specializes in HIM, medical coding, and health care regulatory topics.