Journal Logo

Late Breaker Articles

Sepsis Among Medicare Beneficiaries: 1. The Burdens of Sepsis, 2012–2018*

Buchman, Timothy G. PhD, MD1,2; Simpson, Steven Q. MD1,3; Sciarretta, Kimberly L. PhD1; Finne, Kristen P. BA4; Sowers, Nicole MPP5; Collier, Michael BA5; Chavan, Saurabh MBBS, MPH5; Oke, Ibijoke MPA5; Pennini, Meghan E. PhD1; Santhosh, Aathira MA5; Wax, Marie MBA1; Woodbury, Robyn PhD1; Chu, Steve JD6; Merkeley, Tyler G. MS, MBA1; Disbrow, Gary L. PhD1; Bright, Rick A. PhD1; MaCurdy, Thomas E. PhD5,7,8,9; Kelman, Jeffrey A. MD, MMSc6

Author Information
doi: 10.1097/CCM.0000000000004224


Sepsis is a syndromic illness most generally recognized as one or more adverse host responses to infection. Anyone can develop sepsis. Children younger than 1 year, adults 65 years old and older, persons with weakened immune systems, and persons with chronic medical conditions are at increased risk (1–3). According to recent (2016) Centers for Disease Control and Prevention estimates, 1.7 million adult Americans become septic each year; of those, nearly 270,000 Americans die, and one in three patients who die in a hospital have sepsis (4). Septicemia, which is the detection of a pathogen in the bloodstream in the context of adverse host responses, was recently recognized to be the most costly hospital inpatient condition and accounted for $23.663 billion in costs in 2013 (5). Early survivors are often too ill to return to their homes and require ongoing progressive healthcare in venues such as long-term acute care hospitals (LTCHs) and skilled nursing facilities (SNFs).

Medicare is the U.S. federal health insurance program authorized under the Title XVIII of the Social Security Act for persons who are 65 years old or older, for certain younger persons with disabilities, and for persons with permanent end-stage renal disease requiring long-term dialysis. Medicare Program spending grew to $706 billion in 2017: those costs reflect 20% of the national health expenditure (6). The number of beneficiaries is growing as Americans are aging into Medicare and they are living longer. Summaries of the Medicare program and its beneficiaries are publicly available (7,8).

With rising numbers of beneficiaries come rising numbers of sepsis diagnoses. At the same time, clinical scientists have created and deployed new diagnostics and treatments. The aggregate effects of these demographic, diagnostic, and treatment changes are poorly understood. We therefore sought to clarify the contemporary burdens of sepsis among Medicare beneficiaries, focusing on those who elect fee-for-service (FFS) under Medicare parts A and B, and comparing wherever possible those who have elected Medicare Advantage (MA) under Medicare part C (9–11).

We sought to assess the recent and current burdens of sepsis borne by Medicare beneficiaries, their families, and the nation. Specifically, we wished to 1) count the numbers and calculate the percentages of inpatient admissions linked to sepsis; 2) report the payments for inpatient admissions and subsequent SNF admissions for those patients; and 3) describe the sepsis-associated admission mortality during the inpatient admission and 3 years subsequent to discharge from the inpatient hospital.


We used claims from the Centers for Medicare & Medicaid Services (CMS) DataLink Project. Under the DataLink contract, Acumen, LLC (, Burlingame, CA) produces reports with funding from the CMS and the Assistant Secretary of Preparedness and Response. These reports are compiled using data matching strategies across multiple databases containing claims, payments, and outcomes data. This DataLink project relies initially on pre-adjudicated administrative claims data to conduct near real-time monitoring and research and quality improvement analyses (12–14).

We include a glossary of terms (Table 1).

Glossary and Description of Terminology Used in Medicare Claims

For all three reports in this set, we restricted analysis to include only those claims having their final action as paid and to omit claims either denied or awaiting final action. For this study, the study interval included all claims for services rendered January 1, 2012, to December 31, 2018 (21,22).

We identified: 1) all FFS Medicare beneficiaries with an inpatient hospital admission claim associated with one or more explicit sepsis codes (vide infra); 2) the subset of that cohort with a SNF admission claim within 90 days immediately subsequent to the inpatient hospital claim; and 3) encounter data submitted by MA program insurance plans on behalf of enrollees who required inpatient hospital admission associated with one or more sepsis codes. In cost analysis estimates, we assign costs through encounter data claims made on behalf of MA beneficiaries equivalent to claims paid on behalf of FFS beneficiaries using prevailing FFS rates.

Each inpatient admission is defined by paid (FFS) and encounter data (MA) claims that provide beneficiary, admission date, and provider-level information. The discharge date, diagnoses, and procedure information for the admission were taken from the most recent claim, whereas the admission date information was taken from the earliest claim associated with the particular inpatient admission. Sepsis admissions were identified by the presence of any sepsis diagnosis code listed on the last claim in the inpatient admission. Thus sepsis admissions include both sepsis present on admission (POA) and sepsis not POA (NPOA) meaning that the condition was acquired during the inpatient hospital admission.

During this study interval, Medicare and other payers transitioned their coding bases from International Classification of Disease (ICD), 9th Edition (ICD-9) to 10th Edition (ICD-10) (23). We used the common standard of general equivalence mappings (“GEMs”) to crosswalk the explicit sepsis diagnosis codes (DGNs) of ICD-9 into ICD-10 (24). In addition to these explicit codes, we also evaluated claims reflecting ICD-10 codes used to denominate the sepsis CMS quality metric (SEP-1) (25). Thus, three specific code sets were initially identified:

  • “ICD-9”—Explicit sepsis using ICD-9 codes 038, 995.91, 995.92, and 785.52.
  • “ICD-10 crosswalk”—Explicit sepsis using ICD-10 codes obtained using the GEMs for the ICD-9 codes above: A409, A412, A4101, A4102, A411, A403, A414, A4150, A413, A4151, A4152, A4153, A4159, A4189, A419, A419, R6520, R6521.
  • “ICD-10 SEP-1 metric”—Sepsis defined using ICD-10 codes for the SEP-1 metric: A021, A227, A267, A327, A400, A401, A403, A408, A409, A4101, A4102, A411, A412, A413, A414, A4150, A4151, A4152, A4153, A4159, A4181, A4189, A419, A427, A5486, B377, R6520, R6521.

Note that the ICD-10 crosswalk and ICD-10 SEP-1 metric code sets are similar but not identical: the SEP-1 code set is a superset (contains more codes) compared with the ICD-10 set obtained from the GEMs-enabled crosswalk. In order to capture the maximum number of sepsis-associated inpatient admissions, we generally used the SEP-1 denominator code set to identify, administratively, sepsis following the transition from ICD-9 to ICD-10 that occurred in October 2015. For details of the code sets, see Supplement 1 (Supplemental Digital Content 1,

Conventionally, sepsis is classified according to level of severity, and that severity reflects (1) the extent of the physiologic derangement (2), the individual patient’s ability to respond to that derangement, and (3) the intensity of the corrective treatments. We stratified the severities into tiers based on ICD-9→10 codes as follows:

  1. Septic Shock is specified by ICD-9 code 785.52 or ICD-10 code R6521.
  2. Severe Sepsis without Shock is specified by ICD-9 code 995.92 or ICD-10 code R6520, and excepting septic shock codes.
  3. Non-Severe Sepsis, Organism-Specific is specified by organism-specific sepsis codes (ICD-9: 0380, 03810, 03811, 03812, 03819, 0382, 03840, 03841, 03842, 03843, 03844, 03849, 0388 and ICD-10: A409, A412, A4101, A4102, A411, A403, A414, A4150, A413, A4151, A4152, A4153, A4159, A4189, A021, A227, A267, A327, A400, A401, A408, A4181, A427, A5486, B377) and excepting severe sepsis or septic shock codes.
  4. Non-Severe Sepsis, Unspecified, specified by ICD-9 codes 0389 or 995.91 or ICD-10 code A419, and excepting severe sepsis, septic shock, or organism-specific sepsis codes.

We classified sepsis as being POA when at least one sepsis diagnostic code on the last claim in the admission reported a POA indicator on that claim as Y (yes). Conversely, we classified sepsis as NPOA when all sepsis diagnostic codes on the last claim in the admission had POA indicators as equals to N (no). Some admissions were therefore indeterminate (designated as unknown presence on admission) because there was no sepsis diagnostic code on the last claim in the admission having POA indicator equals to Y and yet not all sepsis codes “during” the admission have POA indicator equals to N. This situation meant that at least one sepsis code had unknown status on admission. (The fraction of sepsis admissions classified as unknown is reported below and was always < 0.4%. These indeterminate admissions were omitted from analyses aimed to compare outcomes of POA and NPOA admissions but were otherwise included, e.g., in aggregate cost reporting.)

Total payment was computed by adding the inpatient stay payment and when applicable, the payment from the most recent Emergency Department outpatient claim with a through date 3 days/72 hours prior to the beneficiary’s inpatient sepsis admission.

Regarding MA inpatient admissions, they are considered MA inpatient admissions if they are a) not FFS inpatient admissions and b) if their associated claims contain a related condition code equal to “04” (signifying an “information only” claim and used by all MA systems) (26). We counted both paid and encounter data inpatient claims for MA beneficiaries that capture services rendered to MA beneficiaries.

MA costs for inpatient sepsis admissions were assigned hierarchically as follows:

  1. Based on the CMS diagnosis-related group (DRG), admission month, and 1-week mortality status (either death within 1 wk of discharge or survival) for the MA inpatient sepsis admission, we assigned a payment equal to the average FFS payment for that DRG, admission month, and 1-week mortality group.
  2. If data for the method in the prior paragraph were not available, we assigned a payment equal to the average FFS payment for the same DRG and 1-week mortality group.
  3. If data for the two prior paragraphs were not available, we assigned a payment equal to the average FFS payment for the same 1-week mortality group.

MA payments for subsequent SNF stays were assigned by assuming that the proportion of SNF stays is the same as that for the FFS group and also that the average payment for MA SNF stays are the same as the FFS SNF stays.

The analysis included dual-eligible beneficiaries. Dual-eligible beneficiaries (Medicare dual eligible [“duals”]) are patients who qualify for and are enrolled in the federal Medicare program and state-operated Medicaid programs. They are considered an at-risk population due to their qualifying disability, chronic illness, and/or socioeconomic status and commonly have higher occurrence of hospitalizations due to their poorer heath status and making proportionately greater claims (27).

Medicaid-only beneficiaries were not analyzed or reported in this study.

The interval to death is reported in reference to the date of discharge from the inpatient hospital reported in the index hospital claim. More precisely, mortality percentages are reported referencing an admission associated with the death of the beneficiary at intervals following each admission and reported herein stratified by the inferred severity. Mortality percentages are computed for 6-month, 1-year, and 3-year “look-forward” periods starting from the beneficiaries’ inpatient hospital discharge dates. (This introduces a slight bias in the data, in that admissions represented earlier in the cohort have completed a 3-yr look-forward, whereas more recent admissions are overrepresented in the shorter look-forward intervals.)

The analyses for this report were generated using SAS software (Version 9.4, SAS System for Windows; SAS Institute, Cary, NC). Herein, we report only descriptive (counts, rates, and costs) statistics.

This analysis and publication is exempt from institutional review board oversight. It was performed as a healthcare quality improvement analysis. CMS is a covered entity. Deidentification methods were implemented in accordance with CMS policy, Privacy Act of 1974 (5 U.S.C. § 552a) and HIPAA (45 Code of Federal Regulations Part 160 and Subparts A and E of Part 164) requirements.

Supplement 2 (Supplemental Digital Content 2, and Supplement 3 (Supplemental Digital Content 3, include the source data used to create figures and also unabridged tables.


Dynamic interactive visualizations accompany some of the results reported below and can be viewed at

Medicare FFS sepsis claims and costs steadily increased during the study interval. There is seasonal variation in claims and costs, with increases during those winter months generally associated with excess respiratory infections including but not limited to seasonal influenza (28). Despite the transition from ICD-9 to ICD-10 codes, there was not visible discontinuity in the growth of costs or of inpatient admission rates owing to the GEMs crossover (Fig. 1, dashed line vs dotted line, respectively). Inpatient hospital admissions that included SEP-1 explicit sepsis codes tracked and slightly exceeded the ICD-10 explicit sepsis codes (Fig. 1, dotted line vs solid line). (The slight excess is a result of the SEP-1 code set being a superset of the sepsis code set obtained from cross-walking ICD-9 to ICD-10 as mentioned in Methods and shown in detail in Supplement 1 [Supplemental Digital Content 1,].)

Figure 1.
Figure 1.:
Analysis of the transition between editions of the coding system (n = 6,731,828 inpatient [IP] admissions to acute care hospitals of Medicare Part A/B beneficiaries). A, Transition from International Classification of Diseases, 9th Edition (ICD-9) to International Classification of Diseases, 10th Edition (ICD-10), IP admission counts. Counts of Medicare fee-for-service–only IP admissions with a sepsis code, by month. B, Transition from ICD-9 to ICD-10 sepsis IP admission rates out of enrolled beneficiaries. Sepsis IP admission rates as a fraction of enrolled beneficiaries, by month. Dashed line, ICD-9; dotted line, ICD-10 crosswalk; solid line, ICD-10 Centers for Medicare & Medicaid Services core measure (SEP-1) metric denominator code set. Note that the SEP-1 metric denominator is a superset of the ICD-10 crosswalk from ICD-9 (the filled squares are slightly higher valued than the open squares).

The number and proportion of inpatient admissions that included a sepsis diagnostic code increased steadily during the interval at all levels of severity (Fig. 2, A and B). A sepsis diagnostic code was included among an average of about 17.8 diagnostic codes in 2012, rising to about 19.5 diagnostic codes in 2018. Although the largest increases occurred in the least severe sepsis tier group admissions, year-over-year increases in both admission counts and proportion of total admissions within the beneficiary community rose steadily (albeit with seasonal variation) through the 7-year period at all severity tiers. The proportions of admissions coded as septic shock, severe sepsis, and sepsis of lower severity varied little (averaging about 2.3% between code types) over the 7-year study interval; however, identification of a specific causative organism declined among admissions with lower severity sepsis (Fig. 2C).

Figure 2.
Figure 2.:
Analysis of sepsis admissions stratified by severity by counts, rates, and proportions (n = 6,731,828 inpatient [IP] admissions to acute care hospitals of Medicare Part A/B beneficiaries). A, IP sepsis admissions, by severity. Sepsis by IP hospital admission counts. Filled circles: septic shock; filled squares: severe sepsis; open circles, nonsevere sepsis (organism unspecified); open triangles, nonsevere sepsis (organism-specified). B, Sepsis admission rates versus all IP admissions, by severity. Percentage of IP admissions featuring a sepsis code (rate). Even for severe sepsis and septic shock, both the count and the rate of sepsis admissions are rising. The impact of seasonal infections on sepsis rates during the winter months is apparent. Note to reviewers, “count” plots have abnormal right “tails” because claims through December 2018 are not yet complete. These tails will disappear when the data and plots are updated in January 2020 prior to publication. C, Fractional severity tiers, by month. Despite the increase in counts and in the fraction of total admissions requiring a sepsis code, the fraction of admissions coded as septic shock and as severe sepsis has remained stable. The fraction of less severe sepsis has also remained stable; however, the identification of specific organisms has declined among the less severe sepsis IP admissions.

The mortality of sepsis diagnoses was different across the severity tiers with greatest 1-week, 6-month, 1-year, and 3-year mortality among patients initially diagnosed with septic shock. Although the least severe cases had less initial mortality, that sepsis tier continued to be associated with increased risk for death 3 years following the index inpatient admission (Fig. 3). Under the Sepsis-3 rubric, these less severe cases are not classified as sepsis, yet despite the absence of organ failure or shock during the index inpatient admission, all-cause mortality among those initial survivors continues to accumulate to total 60% for 3 yr following that stay. For comparison, nonsepsis inpatient admissions, all-cause mortality at 3 years is approximately 40%. See Supplement 2 (Supplemental Digital Content 2, for data. There was a steady reduction in the mortality rate over the study period at all levels of sepsis severity.

Figure 3.
Figure 3.:
Medicare fee-for-service mortality stratified by sepsis severity within 1 wk (A), 6 mo (B), 1 yr (C), and 3 yr (D) of hospital discharge. Even the least severe sepsis admissions serve to mark substantial risk of late mortality. Filled circles, septic shock; filled squares, severe sepsis; open circles, nonsevere sepsis (organism unspecified); open triangles, nonsevere sepsis (organism-specified).

Dual beneficiaries were found to have a higher proportion of their inpatient admissions attributable to or at least accompanied by sepsis (Fig. 4).

Figure 4.
Figure 4.:
Percentage of Medicare fee-for-service inpatient (IP) admissions associated with a sepsis code. Dual beneficiaries have a sepsis code assigned to an IP admission more than twice as frequently as non–dual beneficiaries. Furthermore, the likelihood of a sepsis code assignment is rising faster in the dual beneficiary population. Dual beneficiaries: open circles, International Classification of Diseases, 9th Edition (ICD-9); open squares, International Classification of Diseases, 10th Edition (ICD-10); filled squares, Centers for Medicare & Medicaid Services core measure (SEP-1). Non–dual beneficiaries: open triangles, ICD-9; open diamonds, ICD-10; filled diamonds, SEP-1.

Sepsis POA admissions have been rising generally, whereas sepsis acquired during the hospitalization (NPOA) have modestly declined, albeit with minor seasonal variation (Fig. 5).

Figure 5.
Figure 5.:
Counts of Medicare fee-for-service sepsis present on admission (POA), not POA (NPOA), and unknown. During the 7-yr study period, the counts of sepsis POA have risen steadily, whereas the counts of sepsis NPOA (i.e., acquired during the inpatient [IP] stay) have declined slightly. The fraction of admissions where the sepsis status at admission could not be determined was typically 0.2–0.3% of the total. POA: open circles, International Classification of Diseases, 9th Edition (ICD-9); open squares, International Classification of Diseases, 10th Edition (ICD-10); filled squares, SEP-1. NPOA: open triangles, ICD-9; open wedge, ICD-10; filled wedge, Centers for Medicare & Medicaid Services core measure (SEP-1). Unknown: circle-dot, ICD-9; open hexagon, ICD-10; filled hexagon, SEP-1.

Medicare FFS inpatient sepsis payments totaled $ ≈22.4 billion for CY2018 (Fig. 6). Although many beneficiaries’ cost responsibility is reduced or eliminated through supplemental insurance plan coverage (30% employer-sponsored insurance, 29% Medigap insurance, 22% Medicaid Program), nearly one in five Medicare beneficiaries in FFS Medicare did not have a supplemental coverage in 2016, placing them at financial risk (29). On average, Medicare covered around 95% of the total care payment for the average FFS inpatient admission that included a sepsis code (Table 2).

Total Payments and Beneficiary Responsibilities, Inpatient, and Part B, Fee-for-Service, Acute Care Hospitals, Psychiatric Hospitals, Rehabilitation Hospitals, and Long-Term Care Hospitals
Figure 6.
Figure 6.:
Total monthly payments for all inpatient sepsis admission by severity, Medicare fee-for-service only (n = 6,998,888 inpatient admissions [acute care hospitals, psychiatric hospitals, rehabilitation hospitals, and long-term care hospitals]). Solid line, total payment; dashed line, septic shock; dash and single dot, sepsis with organism unspecified; dotted line, severe sepsis; dash and double dot, sepsis with organism specified.

The cost of inpatient sepsis care declined on a per-admission basis at all levels of sepsis severity (Fig. 7).

Figure 7.
Figure 7.:
Average payment for inpatient hospital admission by sepsis severity, Medicare fee-for-service only. Data are given in current dollars, not constant dollars. Solid line, overall average payment; dashed line, septic shock; dash and single dot, sepsis with organism unspecified; dotted line, severe sepsis; dash and double dot, sepsis with organism specified. There has been a reduction in payment across all sepsis severities and in the average payment. The payment for organism-specific treatment is now less than that caused by an unspecified organism.

A substantial proportion of FFS beneficiaries who survived their inpatient hospital admission was transferred to SNFs due to their qualifying medical condition. Their trajectories are discussed in greater detail in the second report of this set (21). Those costs for a SNF admission following a sepsis inpatient admission appear to be approaching the costs of that prior inpatient admission owing to the decline in the latter (Fig. 8).

Figure 8.
Figure 8.:
Average payments for sepsis survivors, Medicare fee-for-service (FFS) only. Dashed line, average inpatient (IP) payment for beneficiaries who survive and will go on to a skilled nursing facility (SNF) stay; solid line, average IP payment for all beneficiaries who survive the IP admission; dotted line, average 90 d payment to SNF for beneficiaries who survive sepsis and require SNF care. Although average payments for IP care are declining, payments for SNF care are steady. As a consequence, the average payments for IP care and for subsequent SNF care appear to be converging.


To our knowledge, this is the largest set of Medicare sepsis data ever reported. We observe that the large number of beneficiaries and claims analyzed for this report (100% of the Medicare population over a 7-yr interval) can suggest statistical differences even when the actual differences between the groups being compared are trivial. Therefore, we have relied on illustrations, counts, percentages, and mortality to make key points.

Despite improvements in survival and reductions in costs per case, Medicare and its beneficiaries face significant challenges around their long-term survival and quality of life. Furthermore, the rapid expansion of the Medicare population owing to the post-war “baby boom” reaching the age of eligibility has increased case volume in excess of cost-per-case efficiencies: at the beginning of this cohort (i.e., end 2011), Medicare person-years totaled 48,892,758, whereas at the end of the reporting period (end 2018), Medicare person-years totaled 59,950,214, an increase of 22.6%. During that same interval, the overall U.S. population (including these beneficiaries) grew from 312.8 million to 328.1 million, only 4.9%. In other words, the beneficiary population is growing more than 4.5 times faster than the nation as a whole (30). Improvements in care of other chronic (e.g., cancer) and acute conditions (e.g., myocardial infarction) have enabled Medicare beneficiaries to live longer and more productive lives, paradoxically increasing the lifetime probability that they may experience an inpatient admission caused by or complicated by sepsis. Summing the known costs of inpatient and SNF care for FFS beneficiaries with the inferred costs based on encounter data from MA inpatient admissions and (based on the Long-Term Care Minimum Data Set assessment that nursing facilities must report to CMS) a similar rate of transfer of MA sepsis survivors to SNFs, the annual costs of sepsis inpatient admissions and subsequent SNF care to Medicare exceed $41.5 billion (Table 3).

Total Payments for Sepsis Inpatient and Subsequent Skilled Nursing Facility Care

The method for identification of sepsis cases often is debated and sometimes disputed (1,31). There are no fewer than six current definitions of sepsis, and analysis of each is likely to yield a unique cost (and mortality) estimate (Table 4) (42). We aimed to describe the current costs to Medicare, and therefore adopted the current CMS definitions for this report.

Contemporary Definitions of Sepsis

The CMS definition relies on the expertise of professional coders trained to evaluate medical records and assign codes using specific guidance (43). This report further focuses on a code set containing only four ICD-9 codes that collectively are widely accepted as explicit sepsis diagnostic codes beginning in 2002 (34).

We were concerned initially that the change in coding bases over the study interval could affect findings independent of patient condition and care rendered. We observed continuity of costs and of mortality during the ICD-9→ICD-10 transition at all tiers of sepsis severity. We also observed that the SEP-1 denominator is a superset of the ICD-10 codes identified by the GEMs crosswalk. An advantage of choosing the explicit code approach is that it is fully described and therefore can be readily migrated to diverse datasets using a detailed method in a way that simplifies comparisons across patient populations (22).

Although the rise in the use of explicit diagnosis codes might reflect increasing patient and provider awareness of sepsis and possibly financial incentives to use those codes, the observed steady rise in the severest sepsis diagnoses with objective findings (such as septic shock) suggests that sepsis in fact may be becoming more common as opposed to more commonly coded. The fact that the proportions of patients with septic shock, severe sepsis, and less severe sepsis were little changed (varying only about 2.3%) over the 7-year study interval, even while the rate of sepsis admissions increased approximately 50% suggests that changes in coding behavior are not contributing substantially to the increased number of admissions. Rather, there are more beneficiaries and a rising incidence of POA sepsis. We observed that the number of sepsis-related claims and their aggregate dollar cost is rising, even while the cost per inpatient admission and mortality by severity are declining.

Plausible alternatives may also contribute to the rise in counts and percentages of sepsis inpatient admissions. Those alternatives note that a) although the shock aspect of septic shock is arguably objective, the sepsis part of septic shock is certainly not given that up to half of septic shock is culture negative; b) the estimated sensitivity of coding for septic shock relative to clinical markers of shock (i.e., vasopressors) has been reported by others to be only 66% and therefore codes are an imperfect proxy for true disease incidence; c) other investigators have reported substantial differences in the trajectories of septic shock incidence and mortality when using claims data compared with data in the electronic health record; and d) more sensitive and complete coding could be occurring at all levels of sepsis severity (1,44–46). We further note that the accuracy of POA coding has been reported as only moderate; however, this assessment is based on data now a decade and half old (47).

In 2016, Torio and Moore (5) reported that septicemia was the most costly inpatient diagnosis in the United States, totaling $23.7 billion in 2013. Their analysis leveraged the Health Care Utilization Project, which itself was based on the national inpatient sample covering all U.S. acute care hospitals with the exception of federal hospitals operated by the Department of Veterans Affairs and the Defense Health Agency (48). The cost estimated by Torio and Moore (5) is contrasted to the costs reported herein as follows. Our study population was restricted to Medicare beneficiaries, who accounted for 61.5% of the total cost in the Torio and Moore (5) study, or about $15 billion in 2013 dollars. Different from Torio and Moore (5), a) we did not restrict the analysis to a single ICD-9 code (038) but rather used an expanded and widely accepted set of codes (including ICD-9 code 038) and b) we further assessed the costs of SNF inpatient admissions subsequent to a sepsis inpatient hospital admission claim.

Similar to Torio and Moore (5), no attempt was made to attribute any portion of the cost directly to sepsis diagnosis and treatment. No attempt was made to estimate the secondary costs of sepsis including (but not limited to) costs related to prehospital care, transport to the hospital, loss of productivity or employment of the beneficiary, loss of employment by family workers as they assume caregiver roles, costs associated with temporary relocation in order to accompany the beneficiary to inpatient hospital and SNF settings, or costs of outpatient care such as home health, provider visits, and other supplier supports such as home oxygen. The costs reported herein represent only payments for facility care rendered in inpatient and SNF settings. As such the summed costs reported herein inform only a lower bound for the actual costs for Medicare beneficiaries only and must not be interpreted as an estimate of national costs. (Such an estimate is considered in the third report of this series [22].) We observe that, for Medicare FFS beneficiaries who are discharged to SNF care, the average costs of their inpatient care and the average costs of subsequent SNF care are converging. Fortunately, the percentage of sepsis patients requiring SNF care declined steadily over CY2012–CY2018 from 37% to 30%.

The data illuminate differences in costs and mortality associated with sepsis POA to the inpatient hospital compared with costs and mortality associated with sepsis that is NPOA and is acquired as a complication during an inpatient hospital admission. Although NPOA (i.e., hospital-acquired) sepsis constitutes a declining minority of cases (approximately 13% at the beginning of the study interval, declining to about 7.5% at the end of the study interval), such hospital-acquired sepsis continues to portend disproportionate mortality (≈60% for NPOA vs ≈40% for POA sepsis at 6 mo) and costs (generally, more than double the costs of sepsis POA) (49,50). Similarly, dual beneficiaries are at risk for later mortality and accumulated expense (22). Such analyses draw attention to uniquely vulnerable populations that might otherwise be lost by use of a single estimate of cost or mortality risk.

Although the data are comprehensive (100% of Medicare beneficiaries for the interval 2012–2018), the analysis strategy creates its own limitations. First, there is an accounting anomaly. We studied all admissions with a sepsis diagnostic code and reported both costs and mortality relative to each admission and not relative to each beneficiary. Such admission-based data may provide additional perspective for discussing outcomes with patients and families based on the most recent inpatient admission and reflecting the beneficiary’s most current response to infection. (For a companion, beneficiary-focused exploration of the impact of serial admissions, we studied beneficiaries who have not experienced any inpatient admission for a year prior to an index admission [21].) Second, there is a labeling uncertainty. Although sepsis is defined as organ dysfunction consequent to infection, it is widely understood that approximately half of patients “thought to be septic” never yield a positive culture (51). Specifics aside, every sepsis definition generally depends on a clinical impression that infection is sufficiently likely that cultures are obtained and antibiotics are initiated. We do not assert that the administrative codes that identify the patients we term “septic” reliably collects all patients with adverse systemic responses attributable to microbial pathogens. We only assert that there is a consistent method attempting to identify such patients, and the clinical care of patients so identified (diagnostic, treatment, rehabilitation, progressive care) is associated with claims. There are no generally accepted criteria for reliably classifying patients as infected or uninfected, reliably distinguishing physiologic from pathologic responses to infection, or reliably attributing a cost to sepsis or some other illness.

This report offers insight into mortality associated with sepsis administrative codes and costs associated with sepsis administrative codes. Sepsis may accompany other conditions that are also lethal and costly.


The human and economic burdens of sepsis experienced by Medicare beneficiaries continue to grow. Although there are improvements in mortality and in cost-per-case throughout a pragmatic hierarchy of sepsis severity, the year-over-year growth of the beneficiary population, the year-over-year increase in the total number of sepsis deaths, and the year-over-year increase in the total cost of sepsis care highlight the need to understand how beneficiaries become septic, their clinical courses once septic, and how sepsis survivors fare following discharge from the acute care hospital. Such insights are necessary to prevent and to more rapidly detect sepsis earlier, to mitigate the effects of sepsis, and to improve the lives of surviving beneficiaries and their families after they leave the hospital. Analysis of these trajectories of sepsis is the focus of the second report in this series (21). Finally, these recent data might be used to create models of the sepsis population to predict future circumstances including the number of cases and associated costs. Methodology, models, and forecasts are reported in the third article in this series (22).


1. Rhee C, Dantes R, Epstein L, et al.; CDC Prevention Epicenter Program: Incidence and trends of sepsis in US hospitals using clinical vs claims data, 2009-2014. JAMA 2017; 318:1241–1249
2. Rhee C, Wang R, Zhang Z, et al.; CDC Prevention Epicenters Program: Epidemiology of hospital-onset versus community-onset sepsis in U.S. Hospitals and association with mortality: A retrospective analysis using electronic clinical data. Crit Care Med 2019; 47:1169–1176
3. Tillmann B, Wunsch H. Epidemiology and outcomes. Crit Care Clin 2018; 34:15–27
4. National Center for Emerging and Zoonotic Infectious Diseases (NCEZID), Centers for Disease Control and Prevention, Division of Healthcare Quality Promotion (DHQP) CDC Data and Reports, 2016. Cited August 23, 2016. Available at: August 23, 2016
5. Torio CM, Moore BJ. National Inpatient Hospital Costs: The Most Expensive Conditions by Payer, 2013: Statistical Brief #204. Healthcare Cost and Utilization Project (HCUP) Statistical Briefs. 2006Rockville, MD, Agency for Healthcare Research and Quality.
6. United States Center for Medicare and Medicaid Services: NHE Fact Sheet, 2019. Available at: Accessed April 26, 2019
7. United States Center for Medicare and Medicaid Services: Medicare Beneficiaries at a Glance, 2017. Available at: Accessed December 1, 2019
8. Medicare Payment Advisory Commission: Medicare Beneficiary Demographics in A Data Book: Health care spending and the Medicare program, 2019. Available at: Accessed December 1, 2019
9. United States Center for Medicare and Medicaid Services: What’s Medicare? 2019. Available at: Accessed December 1, 2019
10. United States Center for Medicare and Medicaid Services: What Medicare Covers, 2019. Available at: Accessed December 1, 2019
11. United States Center for Medicare and Medicaid Services: MLN Booklet: Medicare and Medicaid Basics, 2018. Available at: Accessed December 1, 2019
12. DeSalvo K, Lurie N, Finne K, et al. Using Medicare data to identify individuals who are electricity dependent to improve disaster preparedness and response. Am J Public Health 2014; 104:1160–1164
13. Kelman J, Finne K, Bogdanov A, et al. Dialysis care and death following hurricane sandy. Am J Kidney Dis 2015; 65:109–115
14. Lurie N, Finne K, Worrall C, et al. Early dialysis and adverse outcomes after hurricane sandy. Am J Kidney Dis 2015; 66:507–512
15. United States Center for Medicare and Medicaid Services: CMS Research-General Information, 2018. Available at: Accessed December 1, 2019.
    16. United States Center for Medicare and Medicaid Services: Long Term Care Hospitals. Available at: Accessed June 1, 2019
      17. United States Center for Medicare and Medicaid Services: Inpatient Rehabilitation Hospitals. Available at: Accessed March 5, 2012
        18. United States Center for Medicare and Medicaid Services: Psychiatric Hospitals, 2019. Available at: Accessed September 7, 2017
          19. United States Center for Medicare and Medicaid Services: Skilled Nursing Facility, 2019. Available at: Accessed December 1, 2019
            20. United States Center for Medicare and Medicaid Services: Nursing Home Care, 2019. Available at: Accessed December 1, 2019
              21. Buchman TG, Simpson SQ, Sciarretta KL, et al. Sepsis Among Medicare Beneficiaries: 2. The Trajectories of Sepsis, 2012–2018. Crit Care Med 2020:289–301
              22. Buchman TG, Simpson SQ, Sciarretta KL, et al. Sepsis Among Medicare Beneficiaries: 3. The Methods, Models, and Forecasts of Sepsis, 2012–2018. Crit Care Med 2020; 48:302–318
              23. United States Center for Medicare and Medicaid Services: ICD-10 2015. Available at: Accessed May 17, 2018
              24. United States Center for Medicare and Medicaid Services: General Equivalence Mappings: ICD-9-CM to and from ICD-10-CM and ICD-10-PCS, 2009. Available at: Accessed December 1, 2019
              25. The Joint Commission: Specifications Manual for National Hospital Inpatient Quality Measures, 2019. Available at: Accessed December 1, 2019
              26. United States Center for Medicare and Medicaid Services: Condition Code 04, 2010. Available at: Accessed December 1, 2019
              27. United States Center for Medicare and Medicaid Services: Medicare Dual Beneficiaries at a Glance, 2018. Available at: Accessed December 1, 2019
              28. United States Centers for Disease Control and Prevention, National Center for Immunization and Respiratory Diseases (NCIRD): Influenza (flu), 2019. Available at: Accessed June 11, 2019
              29. Cubanski JD, Damico A, Neuman T, et al. Sources of Supplemental Coverage Among Medicare Beneficiaries in 2016, 2019. Available at: Accessed December 1, 2019
              30. United States Census Bureau: United States Census, 2019. Available at: Accessed December 1, 2019
              31. Gaieski DF, Edwards JM, Kallan MJ, et al. Benchmarking the incidence and mortality of severe sepsis in the United States. Crit Care Med 2013; 41:1167–1174
              32. American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 1992;20:864–874
              33. Bone RC, Sprung CL, Sibbald WJ. Definitions for sepsis and organ failure. Crit Care Med 1992; 20:724–726
              34. Levy MM, Fink MP, Marshall JC, et al.; SCCM/ESICM/ACCP/ATS/SIS: 2001 SCCM/ESICM/ACCP/ATS/SIS International sepsis definitions conference. Crit Care Med 2003; 31:1250–1256
              35. Whitemyer B. Sepsis Coding Decision Tree in ICD-9 and ICD-10 2015. Available at: Accessed December 1, 2019
                36. Rhodes A, Evans LE, Alhazzani W, et al. Surviving sepsis campaign: International guidelines for management of sepsis and septic shock: 2016. Crit Care Med 2017; 45:486–552
                37. Levy MM, Evans LE, Rhodes A. The surviving sepsis campaign bundle: 2018 update. Crit Care Med 2018; 46:997–1000
                38. Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of clinical criteria for sepsis: For the third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA 2016; 315:762–774
                39. Shankar-Hari M, Phillips GS, Levy ML, et al.; Sepsis Definitions Task Force: Developing a new definition and assessing new clinical criteria for septic shock: For the third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA 2016; 315:775–787
                40. Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA 2016; 315:801–810
                41. Rhee C, Zhang Z, Kadri SS, et al.; CDC Prevention Epicenters Program: Sepsis surveillance using adult sepsis events simplified eSOFA criteria versus sepsis-3 sequential organ failure assessment criteria. Crit Care Med 2019; 47:307–314
                42. Saria S, Henry KE. Too many definitions of sepsis: Can machine learning leverage the electronic health record to increase accuracy and bring consensus? Crit Care Med 2020; 48:137–141
                43. American Association of Professional Coders: American Association of Professional Coders, 2019. Available at: Accessed December 1, 2019
                44. Rhee C, Jentzsch MS, Kadri SS, et al.; Centers for Disease Control and Prevention (CDC) Prevention Epicenters Program: Variation in identifying sepsis and organ dysfunction using administrative versus electronic clinical data and impact on hospital outcome comparisons. Crit Care Med 2019; 47:493–500
                45. Kadri SS, Rhee C, Strich JR, et al. Estimating ten-year trends in septic shock incidence and mortality in United States academic medical centers using clinical data. Chest 2017; 151:278–285
                46. Rhee C, Murphy MV, Li L, et al.; Centers for Disease Control and Prevention Epicenters Program: Improving documentation and coding for acute organ dysfunction biases estimates of changing sepsis severity and burden: A retrospective study. Crit Care 2015; 19:338
                47. Goldman LE, Chu PW, Osmond D, et al. The accuracy of present-on-admission reporting in administrative data. Health Serv Res 2011; 46:1946–1962
                48. HCUP Home: Healthcare Cost and Utilization Project (HCUP). October 2019. Agency for Healthcare Research and Quality. Healthcare Cost and Utilization Project (HCUP) 2019. Available at: Accessed December 1, 2019
                49. Leisman DE, Angel C, Schneider SM, et al. Sepsis presenting in hospitals versus emergency departments: demographic, resuscitation, and outcome patterns in a multicenter retrospective cohort. J Hosp Med 2019; 14:340–348
                50. Rothman M, Levy M, Dellinger RP, et al. Sepsis as 2 problems: Identifying sepsis at admission and predicting onset in the hospital using an electronic medical record-based acuity score. J Crit Care 2017; 38:237–244
                51. Nannan Panday RS, Lammers EMJ, Alam N, et al. An overview of positive cultures and clinical outcomes in septic patients: A sub-analysis of the Prehospital Antibiotics Against Sepsis (PHANTASi) trial. Crit Care 2019; 23:182

                cost; Medicare; mortality; sepsis

                Supplemental Digital Content