Readmissions metrics, procalcitonin testing in CAP, and more

Summaries from ACP Hospitalist Weekly.


CMS metric for 30-day readmissions may not be linked with improved care quality, outcomes after MI

Excess readmission ratio (ERR), the performance measure used by CMS as part of its Hospital Readmissions Reduction Program, is not associated with improved care quality or long-term clinical outcomes after acute myocardial infarction (MI), a recent study suggested.

Photo by Thinkstock
Photo by Thinkstock.

To examine the association between MI-ERR and inpatient care quality, researchers used data from consecutive patients admitted with MI (n=229,252) to 519 hospitals from July 1, 2008, to June 30, 2011 (the time period used to calculate readmission penalties for the first cycle of the CMS program). Process-of-care outcomes included adherence to MI performance measures (acute and discharge), as well as a composite defect-free care metric.

To examine readmissions and mortality, researchers performed a one-year outcome analysis in 51,453 patients from 377 sites who were discharged alive from the index hospitalization and had follow-up data available. Clinical outcomes included days from discharge to the composite of all-cause mortality and all-cause readmission within one year of discharge.

Results were published online on April 26 by JAMA Cardiology and appeared in the July print issue.

Researchers grouped hospitals into four categories based on their ERR for MI for fiscal year 2013 (one group with MI-ERR ≤1 and three groups with MI-ERR >1). There were minimal differences in age and proportion of women across groups, but the proportion of black patients was significantly higher among groups with MI-ERR greater than 1 than the group with MI-ERR of 1 or less (7.6% vs. 4.5%; P=0.01).

Overall adherence to MI process-of-care measures was high, and researchers found no significant differences between groups in the unadjusted adherence to acute MI measures. Centers with higher MI-ERR showed a statistically significantly yet modest reduction in use of aspirin and beta-blockers at discharge (P=0.03 and P=0.04, respectively) compared to those with lower MI-ERR. Defect-free care was not significantly associated with continuous measures of MI-ERR in unadjusted or adjusted analyses.

For long-term outcomes, the overall adjusted analysis (1 to 365 days after discharge) showed that the risk for the composite outcome of mortality or all-cause readmission within one year was higher with increasing MI-ERR (9% higher risk per 0.1-unit increase in MI-ERR up to 1; 6% higher risk per 0.1-unit increase thereafter). However, in the adjusted landmark analysis (31 to 365 days after discharge), MI-ERR was not significantly associated with long-term risk for composite outcomes.

The study authors noted limitations to their work, such as how hospitals were National Cardiovascular Data Registry/Acute Coronary Treatment and Intervention Outcomes Network Registry–Get With the Guidelines centers that participated in the first cycle of the Readmissions Reduction Program, so results may not be generalizable to all hospitals. They added that they did not consider ED visits or observation status admissions during follow-up and that the results may not be applicable to the time period after the study (2012 and later).

They emphasized that the number of black patients and those with higher clinical severity (e.g., more prevalent heart failure symptoms, lower ejection fraction, more bleeding events) were significantly higher at centers with higher MI-ERR. They added that the CMS measure adjusts for many hospital-level differences in patient characteristics (e.g., age, comorbidities, patient case-mix) but does not account for race/ethnicity or any measures of symptoms or ejection fraction.

“Because MI-ERR does not associate with indices of quality of care or 1-year mortality or readmissions between 31 and 365 days, our findings raise questions of whether CMS readmissions penalties are equitably and justly applied for hospitals with a high prevalence of socially and/or medically complex patients,” they wrote.

Procalcitonin can help differentiate bacterial from viral CAP, but no specific cutoff found

Procalcitonin values demonstrated accuracy in differentiating viral from bacterial community-acquired pneumonia (CAP) in a recent study, but not at clear-cut thresholds.

Researchers used data from 1,735 adults hospitalized with CAP (median age, 57 years) who were enrolled in the CDC Etiology of Pneumonia in the Community study from January 2010 through June 2012. Participants had procalcitonin measurements and underwent systematic bacterial and viral pathogen testing.

Based on microbiology test results, researchers categorized patients into five pathogen groups: typical bacteria (169 patients [10%]), atypical bacteria (67 [4%]), virus (409 [24%]), mycobacteria/fungus (15 [1%]), and no pathogen detected (1,075 [62%]). They conducted three analyses to assess the accuracy of procalcitonin in identifying bacterial CAP, comparing 1) patients with any bacterial pathogen to those with a viral pathogen, 2) patients with typical bacteria to those with atypical bacteria or viruses, and 3) patients with any bacterial pathogen to those who did not have bacteria detected.

Results were published online on April 12 by Clinical Infectious Diseases and appeared in the July 15 print issue.

Researchers used the following procalcitonin values, which have been described in the literature as thresholds for identifying bacterial CAP: <0.1 ng/mL, 0.1 to 0.24 ng/mL, 0.25 to 0.49 ng/mL, and ≥0.5 ng/mL.

Median procalcitonin was lower in the viral group compared to the atypical bacterial group (0.09 ng/mL vs. 0.20 ng/mL, P=0.05) and typical bacterial group (2.5 ng/mL, P<0.01). Typical bacteria were more prevalent in patients with higher procalcitonin concentrations (3% among those with <0.1 ng/mL vs. 21% among those with ≥0.5 ng/mL). Among patients with typical bacteria, 23.1% had procalcitonin <0.25 ng/mL, and 12.4% had procalcitonin <0.1 ng/mL.

For distinguishing any bacterial CAP from viral CAP, a procalcitonin threshold of ≥0.1 ng/mL produced a sensitivity of 80.9% (95% CI, 75.3% to 85.7%) and a specificity of 51.6% (95% CI, 46.6% to 56.5%). For distinguishing typical bacterial CAP from viral and atypical CAP, the same threshold produced a sensitivity of 87.6% (95% CI, 79.6% to 90.7%) and a specificity of 49.3% (95% CI, 44.8% to 54.0%). For bacterial CAP versus nonbacterial CAP, sensitivity was 80.9% (95% CI, 75.3% to 85.7%) and specificity was 46.2% (95% CI, 43.7% to 48.8%).

The study authors noted limitations, such as how almost a quarter of patients from the CDC study were excluded because procalcitonin measurements were not available and that, for those with values, the biomarker was only measured at the time of admission. They wrote that procalcitonin concentrations “could be a useful adjunct in the etiologic assessment of patients hospitalized with CAP,” but added that “clinicians cannot rely on procalcitonin alone to guide antibiotic decisions.”

Procalcitonin appears to correlate with viral or bacterial etiologies of acute respiratory tract infections, “but not terribly well,” according to an accompanying editorial. However, the biomarker can likely be used to help guide clinical management, the editorialists stated. “The challenge going forward is learning how to reconcile host and pathogen-based diagnostics to gain a comprehensive understanding of the patient's disease,” they wrote.

Rivaroxaban reduced recurrence of unprovoked VTE compared to warfarin

For patients with an unprovoked venous thromboembolism (VTE), rivaroxaban was associated with lower risk of recurrence and similar rates of bleeding as warfarin, according to a recent study.

The propensity-matched cohort study used Danish health registries to identify patients with a first hospital diagnosis of unprovoked VTE from Dec. 9, 2011 to Feb. 28, 2016. None had taken oral anticoagulation before and none had other indications for anticoagulation. The study included 1,734 patients prescribed rivaroxaban matched with 2,945 patients prescribed warfarin. Results were published by The Lancet Haemotology on April 11 and appeared in the May print issue.

At six-month follow-up, the rate of recurrent VTE was 9.9 per 100 person-years in the rivaroxaban group, compared to 13.1 per 100 person-years in the warfarin group, yielding a hazard ratio (HR) of 0.74 (95% CI, 0.56 to 0.96) for rivaroxaban. During the same follow-up period, the rate of major bleeding was 2.4 per 100 person-years in rivaroxaban users compared to 2.0 in warfarin users (HR, 1.19; 95% CI, 0.66 to 2.13).

The results support and expand on the findings of previous clinical trials and observational studies of rivaroxaban, according to the study authors. The current study looked at routine care, rather than patients recruited for a trial, and included patients with pulmonary embolism but excluded patients with provoked VTE to reduce confounding. They noted that the current study also found shorter length of stay with rivaroxaban, which is likely driven by the time required in the hospital to transition patients from heparin to warfarin.

Overall, the study indicates that “rivaroxaban is an effective and safe alternative to warfarin in oral anticoagulation-naïve patients with incident unprovoked [deep venous thrombosis] and pulmonary embolism,” according to the authors.

ICU schedule simulation finds increased continuity, time off with shared service

A computer simulation of intensivist scheduling found that a “shared service” schedule increased continuity for patients and weekends off for physicians compared to a seven-on schedule.

Researchers created a simulation of an ICU with two daytime teams each covered by a different attending and both covered by one overnight on-call attending and looked at the effects of different scheduling models over a year. The model included simulated patients randomly admitted on different service days to assess continuity of care. Results were published online by Critical Care Medicine on March 30 and appear in the July print issue.

Photo by Thinkstock
Photo by Thinkstock.

Under the seven-on schedule, the same attending covered days for seven consecutive days, and off-service attendings cross-covered each night. Under the shared schedule, four attendings shared the majority of day and night service for the two teams over two weeks. Over the two-week block, each attending had one 24-hour shift in the hospital. The shared model resulted in a 9% increase in continuity of care, as measured by the Continuity of Attending Physician Index, the study found. A scheduling system that increased continuity of care could maximize the benefits of having intensivist attendings on overnight, the study authors suggested.

The shared service was also predicted to give the attendings 3.4 more full weekends off per year and 1.3 more weekends in which they had either Saturday or Sunday off. They also gained four additional full weeks without clinical obligations, and the mean time between clinical obligations increased by 5.8 days. The additional blocks of time off could improve physician productivity by allowing uninterrupted time for research, teaching, and administration, as well as reduce burnout, the study authors said.

They noted that different scheduling models “have been an area of intense debate among hospitalists,” but that “there have not been other quantitative studies of the impact of staffing models on physician outcomes and on continuity of care.”

The scheduling model proposed by this study would need to be validated in real-word prospective studies, the authors said. Other limitations and uncertainties include whether physicians would consider the combined day and night coverage associated with the shared service model to be a burden.