Thursday, January 3, 2019

Pittsburgh Authors Publish Review of National CMS SEP-1 Data

I have previously published two articles on CMS SEP-1 performance data, one in August when 3-quarter 2017 data was first released (here) and one in November when full year 2017 data was released (here).  The second article focuses on the poor performance of top US academic medical centers on SEP-1.

SEP-1 has been criticized elsewhere; see references in the above blogs and for two simple entry points here and here.  For more detail see Faust & Weingart here.

Barbash, David, and Kahn of University of Pittsburgh School of Medicine have published a late December 2018 article which is a peer-reviewed assessment of the CMS raw data.  (For an earlier report see Venkatesh, here.)

I quote the full abstract below.  In my November blog, Pittsburgh was in the top 20 academic hospitals for which I pulled SEP-1 data; Pittsburgh scored 42, below the national average but about the median of the top 20 academic centers.  The authors believe their data shows that the overall value of SEP-1 is "supported by providing additional construct validity," and quote other work suggesting that early identification saves lives.  If this is the case, it should be more upsetting that many hospitals score so poorly and in the open public record.  However, the authors conclude that overall associations between SEP-1 scores and other quality scores was weak, consistent with other weak or negative data for SEP-1 (here, here, here).




Crit Care Med. 2018 Dec 21.   [Epub ahead of print]
National Performance on the Medicare SEP-1 Sepsis Quality Measure.
Barbash IJ1,2, Davis B2,3, Kahn JM1,2,3.
   https://insights.ovid.com/pubmed?pmid=30585827       https://journals.lww.com/ccmjournal/Abstract/onlinefirst/National_Performance_on_the_Medicare_SEP_1_Sepsis.96060.aspx

OBJECTIVES:
The Centers for Medicare and Medicaid Services requires hospitals to report compliance with a sepsis treatment bundle as part of its Inpatient Quality Reporting Program. We used recently released data from this program to characterize national performance on the sepsis measure, known as SEP-1.

DESIGN:
Cross-sectional study of United States hospitals participating in the Centers for Medicare and Medicaid Services Hospital Inpatient Quality Reporting Program linked to Centers for Medicare and Medicaid Services' Healthcare Cost Reporting Information System.

SETTING:
General, short-stay, acute-care hospitals in the United States.

MEASUREMENTS AND MAIN RESULTS:
We examined the hospital factors associated with reporting SEP-1 data, the hospital factors associated with performance on the SEP-1 measure, and the relationship between SEP-1 performance and performance on other quality measures related to time-sensitive medical conditions. A total of 3,283 hospitals were eligible for the analysis, of which 2,851 (86.8%) reported SEP-1 performance data. SEP-1 reporting was more common in larger, nonprofit hospitals. The most common reason for nonreporting was an inadequate case volume.

Among hospitals reporting SEP-1 performance data, overall bundle compliance was generally low, but it varied widely across hospitals (mean and SD: 48.9% ± 19.4%). Compared with hospitals with worse SEP-1 performance, hospitals with better SEP-1 performance tended to be smaller, for-profit, nonteaching, and with intermediate-sized ICUs. Better hospital performance on SEP-1 was associated with higher rates of timely head CT interpretation for stroke patients (rho = 0.16; p < 0.001), more frequent aspirin administration for patients with chest pain or heart attacks (rho = 0.24; p < 0.001) and shorter median time to electrocardiogram for patients with chest pain (rho = -0.12; p < 0.001).

CONCLUSIONS:
The majority of eligible hospitals reported SEP-1 data, and overall bundle compliance was highly variable. SEP-1 performance was associated with structural hospital characteristics and performance on other measures of hospital quality, providing preliminary support for SEP-1 performance as a marker of timely hospital sepsis care.

PMID: 30585827