Saturday, August 4, 2018

CMS Releases SEP-1 Data - A Sepsis Management Performance Measure for Hospitals

  • Sepsis is a major public health issue, and international guidelines urge rapid interventions.  
  • After several years, CMS has just released data on how well hospitals perform to its sepsis rapid intervention metric.   
  • About half of US hospitals failed Medicare's sepsis management measure for over half of their patients.
    • Academic centers ranged from 23% compliance (Yale) to 71% (Stanford).
    • Even Republican health leadership doesn't help - Indiana is near the bottom of national rankings.
    • Henry Ford Hospital - the "owner" of the national NQF500 Sepsis Metric - scores below the median of US hospitals at only 43%.
    • Update.  Data in this essay is 7/2018 date for Q1-Q3-2017.  In 10/2018, CMS released full CY2017 data.  I've put the CY17 raw data here.   This would only cause very minor changes in the essay below.   
      • In CY2017, 461,523 CMS patients were logged into the urgent care SEP-1 protocols of which just 229,093 completed the protocol successfully.

Sepsis as a Public Health Crisis
In the past decade, sepsis has increasingly been recognized as a major public health problem.   According to Sepsis Alliance, the cost of sepsis is over $24B and steadily growing, here.  See also a table and links here.   Accordingly, the CDC has a major public health effort on sepsis (home page here).

Surviving Sepsis Guidelines
One aspect of sepsis management is early recognition and intervention.   A global sepsis alliance regularly updates an overall guide to sepsis management, with special focus on activities in the first 1-3 hours of sepsis management, such as rapid fluid resuscitation and rapid blood draw for organism culture.    See links for the Surviving Sepsis Campaign Guidelines here.  There is an increasing focus on actions right in the first hour, here.   There are also concerns that sepsis guidelines (including CMS SEP-1) lead to overuse of antibiotics, with attendant patient harms (e.g. C. diff) and public harms (resistance), here.

NQF Sepsis Measure - NQF-500
As often occurs, guidelines designed by clinical experts can also be reflected in administrative quality measures and metrics.   NQF, the main U.S. body for endorsing measures, manages NQF-500, a 3-hour sepsis measure.  See here and here.  For a 2018 article on the measure, see Pepper et al., here.

State Mandates for Sepsis Interventions
Some states have mandates for quality tracking and reporting on early sepsis intervention, of which the first was New York State.  For entry points see here and here.

CMS Quality Measure "SEP-1"
For its part, CMS adopted the NQF Sepsis Measure 0500 in 2015, but modifies it with additional instructions for CMS hospital reporting.   While closely based on NQF 0500, CMS calls its measure, "SEP-1."  One issue is that while the intervention actions can be described in 50 words or less, providing exhaustive instructions, rules, and exceptions analysis takes >100 pages (for one entry point, here).

See over twenty papers on SEP-1 at PubMed, with a range of pro & con viewpoints, here.    For stakeholder group comments, here.  For, "Why IDSA Didn't Endorse SEP-1," here; for two blogs here, here.  For a recent hospital trade journal essay, see here.  For a June 2018 publication analyzing early actual SEP-1 data, see Rhee et al., 2018, here.  For ins-and-outs of administratively defining sepsis, see Rhee et al., 2017, here.

In August 2018, for Inpatient FY2019 final rulemaking, CMS determined it would drop chart-abstracted (and labor-intensive) metrics except for SEP-1.  In the final rule, CMS made a point to remark that it remained interested in SEP-1 and would continue to track SEP-1.


CMS Just Released The First Hospital-Level and State-Level SEP-1 Performance Data

In late July 2018, CMS released hospital-level and state-level SEP-1 data for the first time.  (See an article at Boston STAT, here.) There are two different routes to this data.   For consumers, CMS provides a brief summary of what SEP-1 is (here) - one of several hospital indexes for "timely and effective care."  You can search for hospitals by name or zip code and among other measures, you'll see their SEP-1 performance (here).

The professional method gives you Excel spreadsheets of hospital performance data.  Go to this professional's page (here) and click on Download Flat Files.  You'll get about 50 Excel files totalling 250MB of unzipped data.  See the ones titled "Timely and Effective Care" for data from the hospital, state, and national levels.

I've Done the Database Work For You (As .XLS)

I've created a 1 mb file of about 8 tabs of SEP1 data - hospitals, hospitals sorted by score, hospitals sorted by number of cases per year.  I've also included tabs for state level data and national data.  It's in the cloud HERE.  (Click the down-arrow to download as XLS.)

A Snapshot of SEP-1 Data

Of about 4800 hospitals listed, about 1700 don't report SEP-1 data. 

Of those that do, about 6 hospitals had a perfect 100% SEP-1 compliance score - but these hospitals had just 15-90 cases in the year.   A handful of hospitals had >700 SEP-1 category cases in the year.   About 190 hospitals had >80% SEP-1 score.

About half of hospitals had <50% score.   500 hospitals had only 30% compliance or worse with SEP-1.  The median hospital scored about 80 relevant patients per year, or about 1.5 per week.

State Level Data
In state data, Hawaii ranked highest at 68%, Puerto Rico lowest at 11% and DC second-lowest at 35%.  Indiana ranked in the lowest ten-percentile, in the bottom 4 of 50 states, at 43%.  (Indiana is the home state of Secretary of Health Azar, CMS Administrator Verma, and Vice President Pence).

SEP-1 Rules Are Complex
Note that the complex SEP-1 rules allow many patients to be excluded from counting, such as patients with sepsis but with DNR or other palliative-care-only status for whom rapid invasive interventions would be inappropriate.  Since SEP-1 is calculated manually through detailed chart review, some hospitals may be better or worse at applying the complicated appropriate or allowable exclusions.

Hospital Touts SEP-1 Score
I've seen one hospital already issue a press release on its high SEP-1 score (here).

Famous Academic Centers As Low as 23% Score
Academic centers had high variance.  Samples include Stanford (71), Johns Hopkins (53), Brigham Women's (42), and Yale (23).

click to enlarge


The chart below showed state or territory average scores.  Hawaii has a 68, at top left.  Puerto Rico has 11, at bottom right.  Most states fall between 40 and 60.

click to enlarge

In the next chart, below, 3000 hospitals are distributed across the possible scores from 0 (left) to 100 (right).   About 100 U.S. hospitals report SEP-1 scores between 0 and 15 (the three bars to the far left).  About 60 hospitals have scores about 90 (the three bars on the far right.)   
Hospitals are pretty broadly distributed for scores between 25 and 75 (about the middle twelve bars). 
click to enlarge

About 80% of hospitals fall between 25 and 75, and about 10% each are either above 75 or below 25.


Deep Dives Possible
Hawaii had the highest average score, at 68.  Of 23 hospitals in Hawaii, only half (12) reported, the others did not.  Hawaii hospital scores ranged from 41 to 82.  Scores were (41, 44, 51, 52, 53, 62, 68, 70, 71, 78, 80, 82).   Hawaii reporting is based on 2,778 total patients, about 230 average per hospital (range, 58-652, from Kona Community Hospital in Kealakekua [Big Island] to Straub Hospital in Honolulu, respectively).

Puerto Rico had the lowest regional score.  In Puerto Rico, of 51 hospitals, 29 did not report.  Of reporting hospitals, 11 reported between 1% and 88% compliance, with six below 10%.  I'm uncertain if 9 additional PR hospitals reported with a dash as their score, represent "0".

I was stumped that a category with $24B costs could be the largest health care driver for hospitals.  See data here.   20 categories, with 1%-6% each, tallied just 47% of costs at $181M (2013).   The link just provided is a quite interesting view into hospital cost categories for all payers, Medicare, Medicaid, etc.

For an August 2018 article on use of "big data" to map patient physiology to sepsis, here.  For sepsis in claims vs EHR data, here.