[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
Purchase Options:
[Skip to Content Landing]
Figure 1.
Cohort Creation Diagram for the Study of Norepinephrine Shortage and Septic Shock
Cohort Creation Diagram for the Study of Norepinephrine Shortage and Septic Shock

IQR indicates interquartile range.

aPatients meeting the criteria by Angus and colleagues13 for severe sepsis and receiving 1 or more vasopressor during hospitalization.

bDefined as mean use over the first 8 quarters of the study period.

cHospitals meeting 3 criteria for norepinephrine use patterns consistent with shortage: (1) more than 20% relative decrease in norepinephrine use from baseline in at least 1 quarter of 2011; (2) return to norepinephrine use rates within 10% relative to the baseline rate by the second quarter of 2012; and (3) no more than 1 quarter of norepinephrine use more than 20% below baseline before or after 2011.

dHospitals with use of norepinephrine that did not decrease by more than 20% in any study quarter of 2011.

eHospitals that did not meet either set of inclusion criteria (ie, had decrease in norepinephrine use during ≥1 quarter of the shortage year, but also had decreases of >20% in norepinephrine use in ≥2 other quarters).

Figure 2.
Changes in Cohort Rates of Vasopressor Use Over Time Among Consistent-Use Hospitals and Shortage Hospitals
Changes in Cohort Rates of Vasopressor Use Over Time Among Consistent-Use Hospitals and Shortage Hospitals

Changes in cohort rates of vasopressor use (alone or in combination) over time among consistent-use hospitals (n = 120 759 patients; mean, 6038 patients/quarter; median [interquartile range], 5952 [5793-6239] patients/quarter) (A) and shortage hospitals (n = 27 835 patients; mean, 1392 patients/quarter; median [interquartile range], 1399 [1325-1448] patients/quarter) (B). Error bars indicate 95% confidence intervals.

Figure 3.
Trends in In-Hospital Mortality Among the Entire Cohort, Shortage Hospitals, and Consistent-Use Hospitals
Trends in In-Hospital Mortality Among the Entire Cohort, Shortage Hospitals, and Consistent-Use Hospitals

The entire cohort includes 168 304 patients (mean, 8415 patients/quarter; median [interquartile range], 8335 [8074-8777] patients/quarter), the shortage hospitals include 27 835 patients (mean, 1392 patients/quarter; median [interquartile range], 1399 [1325-1448] patients/quarter), and the consistent-use hospitals include 120 759 patients (mean, 6038 patients/quarter; median [interquartile range], 5952 [5793-6239] patients/quarter). Error bars indicate 95% confidence intervals.

Table 1.  
Characteristics of Shortage, Consistent-Use, and Unclassified Hospitals
Characteristics of Shortage, Consistent-Use, and Unclassified Hospitals
Table 2.  
Characteristics of Patients With Septic Shock From July 2008 Through June 2013, Stratified by Shortage Hospitals During Norepinephrine Shortage or Nonshortage Quarters vs Consistent-Use Hospitals
Characteristics of Patients With Septic Shock From July 2008 Through June 2013, Stratified by Shortage Hospitals During Norepinephrine Shortage or Nonshortage Quarters vs Consistent-Use Hospitals
Table 3.  
Association of Norepinephrine Shortage With In-Hospital Mortality for Patients With Septic Shock and Comparison With a Negative Control Cohort of Patients With Severe Sepsis Who Did Not Receive Vasopressors
Association of Norepinephrine Shortage With In-Hospital Mortality for Patients With Septic Shock and Comparison With a Negative Control Cohort of Patients With Severe Sepsis Who Did Not Receive Vasopressors
1.
Gatesman  ML, Smith  TJ.  The shortage of essential chemotherapy drugs in the United States.  N Engl J Med. 2011;365(18):1653-1655.PubMedGoogle ScholarCrossref
2.
Griffith  MM, Pentoney  Z, Scheetz  MH.  Antimicrobial drug shortages: a crisis amidst the epidemic and the need for antimicrobial stewardship efforts to lessen the effects.  Pharmacotherapy. 2012;32(8):665-667.PubMedGoogle ScholarCrossref
3.
Fox  ER, Sweet  BV, Jensen  V.  Drug shortages: a complex health care crisis.  Mayo Clin Proc. 2014;89(3):361-373.PubMedGoogle ScholarCrossref
4.
Roberts  R, Ruthazer  R, Chi  A,  et al.  Impact of a national propofol shortage on duration of mechanical ventilation at an academic medical center.  Crit Care Med. 2012;40(2):406-411.PubMedGoogle ScholarCrossref
5.
Thoma  BN, Li  J, McDaniel  CM, Wordell  CJ, Cavarocchi  N, Pizzi  LT.  Clinical and economic impact of substituting dexmedetomidine for propofol due to a US drug shortage: examination of coronary artery bypass graft patients at an urban medical centre.  Pharmacoeconomics. 2014;32(2):149-157.PubMedGoogle ScholarCrossref
6.
Kaakeh  R, Sweet  BV, Reilly  C,  et al.  Impact of drug shortages on U.S. health systems.  Am J Health Syst Pharm. 2011;68(19):1811-1819.PubMedGoogle ScholarCrossref
7.
US Food and Drug Administration. Current drug shortages. https://web.archive.org/web/20110202172802/http://www.fda.gov/Drugs/DrugSafety/DrugShortages/ucm050792.htm. 2011. Accessed February 6, 2016.
8.
Dellinger  RP, Levy  MM, Carlet  JM,  et al; International Surviving Sepsis Campaign Guidelines Committee; American Association of Critical-Care Nurses; American College of Chest Physicians; American College of Emergency Physicians; Canadian Critical Care Society; European Society of Clinical Microbiology and Infectious Diseases; European Society of Intensive Care Medicine; European Respiratory Society; International Sepsis Forum; Japanese Association for Acute Medicine; Japanese Society of Intensive Care Medicine; Society of Critical Care Medicine; Society of Hospital Medicine; Surgical Infection Society; World Federation of Societies of Intensive and Critical Care Medicine.  Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock: 2008.  Crit Care Med. 2008;36(1):296-327.PubMedGoogle ScholarCrossref
10.
Lindenauer  PK, Pekow  P, Wang  K, Mamidi  DK, Gutierrez  B, Benjamin  EM.  Perioperative beta-blocker therapy and mortality after major noncardiac surgery.  N Engl J Med. 2005;353(4):349-361.PubMedGoogle ScholarCrossref
11.
Rothberg  MB, Pekow  PS, Lahti  M, Brody  O, Skiest  DJ, Lindenauer  PK.  Antibiotic therapy and treatment failure in patients hospitalized for acute exacerbations of chronic obstructive pulmonary disease.  JAMA. 2010;303(20):2035-2042.PubMedGoogle ScholarCrossref
12.
Schneeweiss  S, Seeger  JD, Landon  J, Walker  AM.  Aprotinin during coronary-artery bypass grafting and risk of death.  N Engl J Med. 2008;358(8):771-783.PubMedGoogle ScholarCrossref
13.
Angus  DC, Linde-Zwirble  WT, Lidicker  J, Clermont  G, Carcillo  J, Pinsky  MR.  Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care.  Crit Care Med. 2001;29(7):1303-1310.PubMedGoogle ScholarCrossref
14.
Iwashyna  TJ, Odden  A, Rohde  J,  et al.  Identifying patients with severe sepsis using administrative claims: patient-level validation of the Angus implementation of the international consensus conference definition of severe sepsis.  Med Care. 2014;52(6):e39-e43.PubMedGoogle ScholarCrossref
15.
Elixhauser  A, Steiner  C, Harris  DR, Coffey  RM.  Comorbidity measures for use with administrative data.  Med Care. 1998;36(1):8-27.PubMedGoogle ScholarCrossref
16.
Barnato  AE, Alexander  SL, Linde-Zwirble  WT, Angus  DC.  Racial variation in the incidence, care, and outcomes of severe sepsis: analysis of population, patient, and hospital characteristics.  Am J Respir Crit Care Med. 2008;177(3):279-284.PubMedGoogle ScholarCrossref
17.
Agency for Healthcare Research and Quality.  Healthcare Cost and Utilization Project (HCUP) Procedure Classes. Rockville, MD: Agency for Healthcare Research & Quality; 2014.
18.
Begg  MD, Parides  MK.  Separation of individual-level and cluster-level covariate effects in regression analysis of correlated data.  Stat Med. 2003;22(16):2591-2602.PubMedGoogle ScholarCrossref
19.
Enders  CK, Tofighi  D.  Centering predictor variables in cross-sectional multilevel models: a new look at an old issue.  Psychol Methods. 2007;12(2):121-138.PubMedGoogle ScholarCrossref
20.
Dimick  JB, Ryan  AM.  Methods for evaluating changes in health care policy: the difference-in-differences approach.  JAMA. 2014;312(22):2401-2402.PubMedGoogle ScholarCrossref
21.
Centers for Disease Control and Prevention.  Interruptions in supplies of second-line antituberculosis drugs—United States, 2005-2012.  MMWR Morb Mortal Wkly Rep. 2013;62(2):23-26.PubMedGoogle Scholar
22.
Mayr  FB, Yende  S, Angus  DC.  Epidemiology of severe sepsis.  Virulence. 2014;5(1):4-11.PubMedGoogle ScholarCrossref
23.
Angus  DC, van der Poll  T.  Severe sepsis and septic shock.  N Engl J Med. 2013;369(9):840-851.PubMedGoogle ScholarCrossref
24.
Morelli  A, Lange  M, Ertmer  C,  et al.  Short-term effects of phenylephrine on systemic and regional hemodynamics in patients with septic shock: a crossover pilot study.  Shock. 2008;29(4):446-451.PubMedGoogle Scholar
25.
Morelli  A, Ertmer  C, Rehberg  S,  et al.  Phenylephrine versus norepinephrine for initial hemodynamic support of patients with septic shock: a randomized, controlled trial.  Crit Care. 2008;12(6):R143.PubMedGoogle ScholarCrossref
26.
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(3):486-552.PubMedGoogle ScholarCrossref
27.
Singer  M, Deutschman  CS, Seymour  CW,  et al.  The third international consensus definitions for sepsis and septic shock (Sepsis-3).  JAMA. 2016;315(8):801-810.PubMedGoogle ScholarCrossref
28.
Hall  WB, Willis  LE, Medvedev  S, Carson  SS.  The implications of long-term acute care hospital transfer practices for measures of in-hospital mortality and length of stay.  Am J Respir Crit Care Med. 2012;185(1):53-57.PubMedGoogle ScholarCrossref
Original Investigation
Caring for the Critically Ill Patient
April 11, 2017

Association Between US Norepinephrine Shortage and Mortality Among Patients With Septic Shock

Author Affiliations
  • 1Department of Anesthesiology, Columbia University, New York, New York
  • 2Department of Medicine, Albert Einstein College of Medicine, Montefiore Medical Center, Bronx, New York
  • 3Department of Neurology, Albert Einstein College of Medicine, Montefiore Medical Center, Bronx, New York
  • 4Department of Epidemiology, Columbia University Mailman School of Public Health, New York, New York
  • 5Pulmonary Center, Boston University School of Medicine, Boston, Massachusetts
  • 6Evans Center for Implementation and Improvement Sciences, Boston University School of Medicine, Boston, Massachusetts
  • 7Department of Medicine, University of Toronto, Toronto, Ontario, Canada
  • 8Interdepartmental Division of Critical Care Medicine, University of Toronto, Toronto, Ontario, Canada
  • 9Department of Critical Care Medicine, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada
  • 10Department of Anesthesia, University of Toronto, Toronto, Ontario, Canada
JAMA. 2017;317(14):1433-1442. doi:10.1001/jama.2017.2841
Key Points

Question  Was there an association between the 2011 US norepinephrine shortage and mortality among adults with septic shock?

Findings  In this cohort study of 27 835 patients with septic shock admitted to 26 hospitals, phenylephrine use significantly increased during 3-month periods of active norepinephrine shortage. Compared with hospital admission with septic shock during periods of normal use, admission during hospital periods of shortage was associated with an increased rate of in-hospital mortality (35.9% vs 39.6%, respectively).

Meaning  The US norepinephrine shortage was significantly associated with increased mortality among patients with septic shock.

Abstract

Importance  Drug shortages in the United States are common, but their effect on patient care and outcomes has rarely been reported.

Objective  To assess changes to patient care and outcomes associated with a 2011 national shortage of norepinephrine, the first-line vasopressor for septic shock.

Design, Setting, and Participants  Retrospective cohort study of 26 US hospitals in the Premier Healthcare Database with a baseline rate of norepinephrine use of at least 60% for patients with septic shock. The cohort included adults with septic shock admitted to study hospitals between July 1, 2008, and June 30, 2013 (n = 27 835).

Exposures  Hospital-level norepinephrine shortage was defined as any quarterly (3-month) interval in 2011 during which the hospital rate of norepinephrine use decreased by more than 20% from baseline.

Main Outcomes and Measures  Use of alternative vasopressors was assessed and a multilevel mixed-effects logistic regression model was used to evaluate the association between admission to a hospital during a norepinephrine shortage quarter and in-hospital mortality.

Results  Among 27 835 patients (median age, 69 years [interquartile range, 57-79 years]; 47.0% women) with septic shock in 26 hospitals that demonstrated at least 1 quarter of norepinephrine shortage in 2011, norepinephrine use among cohort patients declined from 77.0% (95% CI, 76.2%-77.8%) of patients before the shortage to a low of 55.7% (95% CI, 52.0%-58.4%) in the second quarter of 2011; phenylephrine was the most frequently used alternative vasopressor during this time (baseline, 36.2% [95% CI, 35.3%-37.1%]; maximum, 54.4% [95% CI, 51.8%-57.2%]). Compared with hospital admission with septic shock during quarters of normal use, hospital admission during quarters of shortage was associated with an increased rate of in-hospital mortality (9283 of 25 874 patients [35.9%] vs 777 of 1961 patients [39.6%], respectively; absolute risk increase = 3.7% [95% CI, 1.5%-6.0%]; adjusted odds ratio = 1.15 [95% CI, 1.01-1.30]; P = .03).

Conclusions and Relevance  Among patients with septic shock in US hospitals affected by the 2011 norepinephrine shortage, the most commonly administered alternative vasopressor was phenylephrine. Patients admitted to these hospitals during times of shortage had higher in-hospital mortality.

Introduction

Quiz Ref IDDrug shortages are an increasing problem with potential effects on patients and health care delivery costs.1,2 In the United States, the issue has received considerable attention from federal legislators, media, and physician and pharmacist advocacy groups. However, the effects of drug shortages on patient outcomes or on alternative prescribing practices are poorly described.36 In addition, there is no national reporting system that captures hospital-level data on drug shortages or their effect on patients.

Quiz Ref IDIn February 2011, the US Food and Drug Administration (FDA) announced a severe nationwide shortage of norepinephrine caused by production interruptions at 3 drug manufacturers that persisted until February 2012.7 Because norepinephrine is recommended as the first-line vasopressor for treatment of hypotension due to septic shock,8 the hypothesis was that admission to a hospital affected by the norepinephrine shortage would be associated with increased mortality in patients with septic shock. This question was addressed using a large, nationally representative database of hospitalized patients in the United States to assess the associations between the norepinephrine shortage, use of alternative vasopressors, and mortality among patients with septic shock.

Methods

The study protocol was approved by the institutional review boards of Columbia University Medical Center and Albert Einstein College of Medicine. Each granted waivers of informed consent. We conducted a retrospective analysis of administrative and pharmacy billing data captured in the Premier Healthcare Database between July 1, 2008, and June 30, 2013.911 Premier data contain standard claims data (eg, patient demographic characteristics, International Classification of Diseases codes) as well as detailed, date-indexed logs of all services billed to patients or insurance, including diagnostic tests, imaging, laboratory tests, and medications (including dose and route). Premier Inc performs an iterative data validation and audit process,12 returning missing or invalid data to source hospitals for correction prior to final data validation and release; missing data are minimal.

Definition of Septic Shock

Patients with septic shock were defined as adults (aged ≥18 years) admitted to US acute care hospitals and meeting criteria for severe sepsis using a previously validated algorithm that uses a combination of International Classification of Diseases, Ninth Revision (ICD-9) codes to identify a bacterial or fungal infectious process and a diagnosis of acute organ dysfunction,13,14 with the addition of use of any of 5 vasopressors (norepinephrine, phenylephrine, dopamine, epinephrine, and vasopressin) for 2 or more days during hospital admission. Vasopressor use was defined as any daily pharmacy charge for a given vasopressor. The cohort was restricted to patients who received vasopressor treatment for 2 or more days to exclude patients with infection and organ dysfunction who might have received brief infusions of vasopressors for indications other than sepsis-induced shock (eg, anesthesia-induced vasodilation). All hospitalizations that met these criteria were included, as we were unable to identify whether the same individual was hospitalized at different points in time using these data.

Definitions of Hospital Vasopressor Use

The study period was divided into quarter years (3-month periods), and the baseline period was defined as the first 8 quarters of data (July 2008 through June 2010) (eFigure 1 in the Supplement). Hospital-level vasopressor use was defined as the proportion of patients with septic shock in each study hospital who received the given vasopressor, alone or in combination with other vasopressor agents, in each study quarter. We calculated baseline and quarterly rates of vasopressor use and relative changes in quarterly use from baseline for each study hospital.

Hospital Exclusions

We excluded low-volume hospitals (defined as those with <500 septic shock admissions during the study period, or <5 septic shock admissions in any study quarter) to ensure that a small change in the number of patients receiving norepinephrine did not disproportionately affect observed hospital-level rates of norepinephrine use. We excluded hospitals with a baseline norepinephrine rate of less than 60% of patients with septic shock to both maximize the number of hospitals included and increase the likelihood that clinicians would experience a shortage of the first-line vasopressor within their hospital.

Primary Hospital Cohort

The possible norepinephrine shortage period in each hospital was defined as the 4 quarters of calendar year 2011 (as reported on the FDA’s drug shortage website).7 To evaluate potential factors and outcomes associated with the norepinephrine shortage, the primary analysis cohort focused on hospitals with consistent use of norepinephrine before and after the shortage, and with a reduction in norepinephrine use by more than 20% during at least 1 quarter of the shortage period. These characteristics were operationalized using 3 a priori criteria to define shortage hospitals: (1) a relative decrease by more than 20% in norepinephrine use from baseline in at least 1 quarter of 2011; (2) a return to norepinephrine use rates to within 10% of the baseline rate by the second quarter of 2012; and (3) no more than 1 quarter of norepinephrine use that was more than 20% below baseline before or after 2011.

Hospitals with norepinephrine use that did not decrease by more than 20% in any quarter of 2011 were designated as consistent-use hospitals and used for comparison with shortage hospitals in a secondary difference-in-differences (DID) analysis. Examples of temporal norepinephrine use patterns among study hospitals meeting shortage and consistent-use criteria are included in eFigure 2 in the Supplement, showing individual hospital data from 2 hospitals. We excluded hospitals not satisfying either set of inclusion criteria (unclassified hospitals), for example, those with intermittent quarters of decreased norepinephrine use throughout the study period (Figure 1).

Covariates

Covariates included patient-level variables such as demographic characteristics, comorbid disease (Elixhauser ICD-9, Clinical Modification classification system) and individual acute organ dysfunction during hospitalization as previously defined.15 Race was included as a variable, based on fixed categories available in Premier data, because of the known association between race and outcomes for patients with sepsis.16 We identified surgical admissions from ICD-9 procedure codes for major diagnostic or therapeutic surgery as classified by the Healthcare Cost and Utilization Project.17 Intensive care unit (ICU) admission was identified from hospitalization bed charges. Hospital-level variables included number of hospital beds, US geographic region, teaching status, and urban vs rural location.

Exposure and Outcomes

Individual patients were considered exposed to the norepinephrine shortage if they were admitted to a hospital in 2011 during a quarter in which the hospital demonstrated at least a 20% relative decrease in norepinephrine use compared with its baseline rate. The primary outcome was in-hospital mortality. The secondary outcome was the use of individual vasopressors.

Statistical Analysis

Characteristics of cohort hospitals are described for the shortage, consistent-use, and unclassified hospitals (eTable 1 in the Supplement). Patient characteristics were compared for patients who received care in shortage hospitals during identified shortage quarters vs patients in shortage hospitals during nonshortage quarters and vs patients in consistent-use hospitals, using means and standard deviations for parametric data and medians and interquartile ranges (IQRs) for nonparametric data. Unadjusted between-group comparisons were performed using t, Wilcoxon rank sum, Kruskal-Wallis, and χ2 tests as appropriate to compare patient characteristics for patients in shortage hospitals in shortage vs nonshortage quarters.

Vasopressor use rates (for each of 5 vasopressors) and in-hospital mortality rates were plotted by study quarter for the entire cohort, shortage hospitals, and consistent-use hospitals. We assessed for trends over time in the entire cohort using linear regression. To better assess use of other vasopressors during the norepinephrine shortage, we correlated the relative change from baseline in quarterly use of norepinephrine (in 2011) and each other vasopressor in shortage hospitals using unadjusted multilevel linear regression with hospital as a random effect to account for correlated observations.

Quiz Ref IDThe primary analysis evaluated the association between admission during quarters of 2011 with 20% or greater relative decreases in norepinephrine use (the primary exposure, as a dichotomous variable) and in-hospital mortality within shortage hospitals. Multivariable mixed-effects logistic regression analysis was performed with hospital of admission as a random intercept to account for clustering by hospital.18,19 Additional covariates included all individual patient and hospital-level factors and year of hospital discharge to adjust for temporal changes in norepinephrine use and in-hospital mortality. The coefficient generated by this analysis describes the adjusted odds of in-hospital mortality attributable to admission to a hospital with a quarterly lower-than-expected rate of norepinephrine use during the period of national shortage, while accounting for baseline differences in mortality rates between hospitals.

To better account for secular effects, in particular the possibility that the norepinephrine shortage period was associated with transiently higher mortality at all hospitals, a DID model was developed.20 This model included both shortage hospitals and consistent-use hospitals (as controls). It compared the difference in mortality in shortage hospitals between identified quarters of shortage in 2011 and all other study quarters (including quarters in 2011 not meeting shortage criteria) with the difference in mortality in consistent-use hospitals in quarters of 2011 vs other study years (independent of individual hospital rates of norepinephrine use). The DID estimator generated by the model represents the effect on in-hospital mortality of receiving care in a shortage hospital during a quarter of 2011 in which that hospital experienced a norepinephrine shortage, adjusted for secular trends in in-hospital mortality in patients with septic shock. As a negative control, the primary analysis was repeated but with patients in the primary cohort of shortage hospitals who met criteria for severe sepsis (ie, infection with organ dysfunction) but who did not receive vasopressors during the hospital stay.

To further examine the association between norepinephrine shortage and in-hospital mortality, we conducted several additional sensitivity analyses. Because of secular trends in mortality, we excluded data from after the shortage (2012-2013) and reran the primary and DID models using data from preshortage quarters and shortage quarters. To examine the stability of the results to alternative definitions of shortage hospitals, we recategorized unclassified hospitals as shortage hospitals if they met shortage hospital criteria using only the preshortage and shortage quarters (n = 15). The models were repeated excluding the patients who were transferred in from other acute care hospitals as they may have had other vasopressor exposure. The primary analysis was also repeated using the cohort of nonsurgical patients who received vasopressors for 1 or more days (vs ≥2 days in the primary analysis) and then categorizing the primary exposure variable into approximate tertiles of relative decreases in norepinephrine use (≤25%, 26%-35%, and >35% from baseline) to assess for an association with the severity of the shortage.

Given the size of the data set, results were evaluated for both clinical and statistical significance (defined as P < .05 without adjustment for multiple comparisons). All significance testing was 2-sided. Statistical analyses were performed with Stata/IC version 13.1 statistical software (StataCorp LP).

Results
Patient and Hospital Characteristics

The total cohort consisted of 168 304 patients with septic shock in 150 hospitals (Figure 1). Of these hospitals, 26 (17.3%) were classified as shortage hospitals and 102 (68.0%) were classified as consistent-use hospitals. The remaining 22 hospitals (14.7%) did not fit either norepinephrine use pattern. Among shortage hospitals, 8 (30.8%) were teaching hospitals, 24 (92.3%) were in urban locations, and 11 (42.3%) were located in the South region of the United States (Table 1). These hospitals were medium to large in size (19 of 26 hospitals [73.1%] with ≥300 beds).

A total of 27 835 patients (median age, 69 years [IQR, 57-79 years]; 47.0% women) were admitted to shortage hospitals. Among 1961 patients admitted to shortage hospitals during shortage quarters (7.0%), the median age was 68 years (IQR, 57-79 years) (Table 2). Nearly all of these patients (1869 of 1961 [95.3%]) were admitted to an ICU. The median length of the first ICU stay was 6 days (IQR, 3-11 days) and the median length of hospital stay was 13 days (IQR, 7-22 days). Characteristics were similar to the patients admitted in nonshortage quarters.

Vasopressor Trends

Overall, norepinephrine use increased during the 5 years in the entire cohort from 76.7% (95% CI, 75.7%-77.6%) in the third quarter of 2008 to 80.0% (95% CI, 79.2%-80.9%) in the second quarter of 2013 (P < .001) (eFigure 3 in the Supplement). At baseline, norepinephrine was used by 78.5% (95% CI, 78.2%-78.7%) of patients with septic shock, and there was a detectable decrease in norepinephrine use during the shortage period, with recovery in 2012. For consistent-use hospitals, the mean baseline use of norepinephrine was 79.3% (95% CI, 78.9%-79.6%) and was stable throughout 2011 (Figure 2A). For shortage hospitals, baseline norepinephrine use was 77.0% (95% CI, 76.2%-77.8%) (Figure 2B). During the shortage year (2011), quarterly norepinephrine use among shortage hospitals decreased to a nadir of 55.7% (95% CI, 52.0%-58.4%) during the second quarter. Mean quarterly phenylephrine use among the same hospitals increased simultaneously from a baseline rate of 36.2% (95% CI, 35.3%-37.1%) to a peak of 54.4% (95% CI, 51.8%-57.2%) in the second quarter of 2011. At the level of the individual shortage hospital in 2011, there was a moderate inverse correlation between the relative change in quarterly use of norepinephrine vs phenylephrine (β = −1.18 [95% CI, −1.55 to −0.81]; P < .001) and dopamine (β = −0.59 [95% CI, −0.84 to −0.34]; P < .001), and no significant correlation between changes in use of norepinephrine vs vasopressin (β = −0.61 [95% CI, −1.39 to 0.17]; P = .13) or epinephrine (β = −0.71 [95% CI, −1.46 to 0.04]; P = .06) (eFigure 4 in the Supplement).

Association Between Norepinephrine Shortage and In-Hospital Mortality

Overall, in-hospital mortality decreased in shortage hospitals during the 5 years assessed, from 34.9% (95% CI, 32.2%-37.5%) in the third quarter of 2008 to 31.9% (95% CI, 29.5%-34.3%) in the second quarter of 2013 (P < .001) (Figure 3). Compared with hospital admission with septic shock during quarters of normal norepinephrine use, admission to shortage hospitals during quarters of 2011 in which relative norepinephrine use decreased by 20% or more was associated with increased odds of in-hospital mortality (9283 of 25 874 patients [35.9%] vs 777 of 1961 patients [39.6%], respectively; absolute mortality difference = 3.7% [95% CI, 1.5%-6.0%]; adjusted odds ratio [AOR] = 1.15 [95% CI, 1.01-1.30]; P = .03) (Table 3 and eTable 2 in the Supplement).

In the DID analysis comparing in-hospital mortality rates between shortage hospitals and consistent-use hospitals, the association was similar (for in-hospital mortality in shortage hospitals during quarters of 2011 meeting shortage criteria, AOR = 1.17 [95% CI, 1.06-1.31]; P = .003), indicating that the observed association between the norepinephrine shortage and mortality was robust to additional adjustment for secular trends (Table 3 and eTable 3 in the Supplement). For the control cohort of patients with severe sepsis who did not require vasopressors, there was no significant association between admission during a shortage quarter and in-hospital mortality (Table 3 and eTable 4 in the Supplement).

Assessment of the primary cohort of shortage hospitals, excluding hospitalizations after the shortage period (when in-hospital mortality was the lowest) yielded similar estimates, as did the DID model (eTable 5 in the Supplement). Many of the unclassified hospitals (n = 15) were excluded because of continued low use of norepinephrine in 2012 (eFigure 5 in the Supplement). Regrouping of these unclassified hospitals as shortage hospitals also yielded a very similar estimate, as did exclusion of patients transferred in from other acute care hospitals (eTable 5 in the Supplement). Analysis of nonsurgical patients who received vasopressors for 1 or more days did not show a significant increase in mortality. When assessed as tertiles of relative decrease in norepinephrine use from baseline as a measure of severity of the shortage, adjusted odds of in-hospital mortality were significantly increased only for the group of patients admitted to shortage hospitals with rates of decreased norepinephrine use in quarters of 2011 in the middle tertile (representing relative decreases of 26%-35% in norepinephrine use) (AOR = 1.29 [95% CI, 1.07-1.56]; P = .007) (eTable 5 in the Supplement).

Discussion

Quiz Ref IDThis study demonstrates changes in patterns of use of vasopressors for patients with septic shock in selected hospitals during the period of norepinephrine shortage in the United States. The decreased use of norepinephrine during periods of shortage was associated with an increase in use of phenylephrine. Receipt of care in a hospital experiencing a norepinephrine shortage was associated with an increased rate of in-hospital mortality (3.7% increase, from 35.9% to 39.6%).

The use of national patient-level data is a strength of the study, as most existing evidence of the effect of shortages on patient care comes from clinician surveys, government reports, and single-center observational studies.4,5,21 The approach of a hospital-level exposure also reduces the possibility of confounding by indication; the results were robust to multiple sensitivity analyses, and the inclusion of a “negative” control (patients with severe sepsis but without a requirement for vasopressors) showing no significant association of care in hospitals experiencing the norepinephrine shortage and increased mortality also strengthens the findings. Placing these findings in the context of an estimated 150 000 deaths from sepsis in the United States each year,22,23 the norepinephrine shortage observed in 17% of hospitals may have been associated with hundreds of excess deaths. A sustained nationwide shortage of norepinephrine affecting all hospitals in the United States could result in thousands of additional deaths in a year. However, these estimates are based on assumptions that the point estimate represents the true effect size and that the entire difference is causal with no confounding, both of which are speculative.

Quiz Ref IDSeveral factors may explain the observed associations between norepinephrine shortage and increased patient mortality. First, other specific vasopressors selected to replace norepinephrine may result in worse outcomes for patients with septic shock. In this study, phenylephrine use increased the most during the norepinephrine shortage. There are fewer data comparing outcomes between patients treated with phenylephrine vs norepinephrine in septic shock.24,25 Concerns about increased ventricular afterload in the setting of myocardial dysfunction inform the Surviving Sepsis Campaign guidelines, which in 2008 (most relevant to the care of the patients in this study) did not recommend use of phenylephrine.8 Moreover, dopamine (the other vasopressor that increased most with the decrease in norepinephrine use) is described in the most recent guidelines as a vasopressor that should be used “as an alternative vasopressor agent to norepinephrine only in highly selected patients”26 based on conclusions of a meta-analysis assessing studies comparing norepinephrine vs dopamine in septic shock.

Another possible explanation for the observed association with mortality is that observable decreases in norepinephrine use in the setting of shortage may be a marker of related unmeasured factors that affected patient outcomes. These could include the absence of a dedicated shortage pharmacist to optimize distribution of limited supplies, delayed administration of vasopressors, or lack of clinician familiarity with dosing and titration of alternative vasopressor agents.

This study has a number of limitations. First, the specific criteria to identify shortage hospitals required specification of cutoffs to determine the presence or absence of a shortage. Some hospitals (22 of the 150 hospitals [14.7%]) could not be classified as either shortage hospitals or consistent-use hospitals. There was limited ability to determine why some hospitals appeared to experience a shortage, while others did not. Defining a shortage in terms of decreased number of patients and days of drug use likely underestimates the number of patients and hospitals affected by the 2011 norepinephrine shortage. These methods may not capture hospitals that experienced shortage but used alternative options to compensate (eg, capping doses in patients to extend available supplies, reducing use in the operating room to ensure availability for the ICUs); such events could result in underattribution of the norepinephrine shortage to outcomes. Although the 2011 norepinephrine shortage was unique among vasoactive drug shortages in its duration and severity, US drug manufacturers reported a brief concurrent shortage of phenylephrine (late April to June 2011) and reduced inventory of vasopressin due to interruption of production at some plants in 2011,7 although the data in the present study did not show a decrease in overall use during these periods.

There is the possibility of residual confounding from unidentified differences between hospitals with and without periods of decreased norepinephrine use associated with shortage, and between patients admitted who did or did not receive norepinephrine during the shortage period. In particular, the data did not include a physiology-based severity-of-illness score for patients, did not allow for identification of specific clinical decision making leading to selection of norepinephrine for individual patients, and did not include information on dosage or actual duration during a 24-hour period. The use of administrative data and the definition by Angus and colleagues13 for patients with severe sepsis is an ICD-9–based approach to defining sepsis and may or may not identify patients who would meet the criteria for sepsis based on the most recent clinical definition.27 The cohort of patients was also limited to those who received vasopressors for 2 or more days and excluded patients who died on the first day of vasopressor treatment. This study did not demonstrate a dose-response association between hospital quarters with greater relative reductions in norepinephrine use and increased mortality. This may be due to a threshold effect of vasopressor shortage on mortality or lack of power due to relatively few hospital quarters at the extreme levels of vasopressor shortage. Also, the limitation of in-hospital mortality means that the results may have underestimated mortality, particularly for hospitals that tend to transfer patients early to other skilled care facilities.28

Conclusions

Among patients with septic shock in US hospitals affected by the 2011 norepinephrine shortage, the most commonly administered alternative vasopressor was phenylephrine. Patients admitted to these hospitals during times of shortage had higher in-hospital mortality.

Section Editor: Derek C. Angus, MD, MPH, Associate Editor, JAMA (angusdc@upmc.edu).
Back to top
Article Information

Corresponding Author: Hannah Wunsch, MD, MSc, Department of Critical Care Medicine, Sunnybrook Hospital, 2075 Bayview Ave, Room D1.08, Toronto, ON M4N 3M5, Canada (hannah.wunsch@sunnybrook.ca).

Published Online: March 21, 2017. doi:10.1001/jama.2017.2841

Author Contributions: Dr Gershengorn had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Vail and Gershengorn are co–first authors.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Vail, Gershengorn, Hua, Walkey, Wunsch.

Drafting of the manuscript: Vail, Rubenfeld, Wunsch.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Vail, Gershengorn, Hua, Rubenfeld, Wunsch.

Supervision: Wunsch.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Walkey reported receiving a grant from the National Institutes of Health and personal fees from UpToDate. Dr Hua reported receiving grants from the National Institute on Aging and the American Federation for Aging Research. No other disclosures were reported.

Funding/Support: This study was supported by funds from the Herbert and Florence Irving Scholars Program at Columbia University (Dr Wunsch).

Role of the Funder/Sponsor: The funding agency had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Meeting Presentation: This article was presented at the 37th International Symposium on Intensive Care and Emergency Medicine; March 21, 2017; Brussels, Belgium; and was presented as an abstract at the 46th Critical Care Congress of the Society of Critical Care Medicine; January 23, 2017; Honolulu, Hawaii.

References
1.
Gatesman  ML, Smith  TJ.  The shortage of essential chemotherapy drugs in the United States.  N Engl J Med. 2011;365(18):1653-1655.PubMedGoogle ScholarCrossref
2.
Griffith  MM, Pentoney  Z, Scheetz  MH.  Antimicrobial drug shortages: a crisis amidst the epidemic and the need for antimicrobial stewardship efforts to lessen the effects.  Pharmacotherapy. 2012;32(8):665-667.PubMedGoogle ScholarCrossref
3.
Fox  ER, Sweet  BV, Jensen  V.  Drug shortages: a complex health care crisis.  Mayo Clin Proc. 2014;89(3):361-373.PubMedGoogle ScholarCrossref
4.
Roberts  R, Ruthazer  R, Chi  A,  et al.  Impact of a national propofol shortage on duration of mechanical ventilation at an academic medical center.  Crit Care Med. 2012;40(2):406-411.PubMedGoogle ScholarCrossref
5.
Thoma  BN, Li  J, McDaniel  CM, Wordell  CJ, Cavarocchi  N, Pizzi  LT.  Clinical and economic impact of substituting dexmedetomidine for propofol due to a US drug shortage: examination of coronary artery bypass graft patients at an urban medical centre.  Pharmacoeconomics. 2014;32(2):149-157.PubMedGoogle ScholarCrossref
6.
Kaakeh  R, Sweet  BV, Reilly  C,  et al.  Impact of drug shortages on U.S. health systems.  Am J Health Syst Pharm. 2011;68(19):1811-1819.PubMedGoogle ScholarCrossref
7.
US Food and Drug Administration. Current drug shortages. https://web.archive.org/web/20110202172802/http://www.fda.gov/Drugs/DrugSafety/DrugShortages/ucm050792.htm. 2011. Accessed February 6, 2016.
8.
Dellinger  RP, Levy  MM, Carlet  JM,  et al; International Surviving Sepsis Campaign Guidelines Committee; American Association of Critical-Care Nurses; American College of Chest Physicians; American College of Emergency Physicians; Canadian Critical Care Society; European Society of Clinical Microbiology and Infectious Diseases; European Society of Intensive Care Medicine; European Respiratory Society; International Sepsis Forum; Japanese Association for Acute Medicine; Japanese Society of Intensive Care Medicine; Society of Critical Care Medicine; Society of Hospital Medicine; Surgical Infection Society; World Federation of Societies of Intensive and Critical Care Medicine.  Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock: 2008.  Crit Care Med. 2008;36(1):296-327.PubMedGoogle ScholarCrossref
10.
Lindenauer  PK, Pekow  P, Wang  K, Mamidi  DK, Gutierrez  B, Benjamin  EM.  Perioperative beta-blocker therapy and mortality after major noncardiac surgery.  N Engl J Med. 2005;353(4):349-361.PubMedGoogle ScholarCrossref
11.
Rothberg  MB, Pekow  PS, Lahti  M, Brody  O, Skiest  DJ, Lindenauer  PK.  Antibiotic therapy and treatment failure in patients hospitalized for acute exacerbations of chronic obstructive pulmonary disease.  JAMA. 2010;303(20):2035-2042.PubMedGoogle ScholarCrossref
12.
Schneeweiss  S, Seeger  JD, Landon  J, Walker  AM.  Aprotinin during coronary-artery bypass grafting and risk of death.  N Engl J Med. 2008;358(8):771-783.PubMedGoogle ScholarCrossref
13.
Angus  DC, Linde-Zwirble  WT, Lidicker  J, Clermont  G, Carcillo  J, Pinsky  MR.  Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care.  Crit Care Med. 2001;29(7):1303-1310.PubMedGoogle ScholarCrossref
14.
Iwashyna  TJ, Odden  A, Rohde  J,  et al.  Identifying patients with severe sepsis using administrative claims: patient-level validation of the Angus implementation of the international consensus conference definition of severe sepsis.  Med Care. 2014;52(6):e39-e43.PubMedGoogle ScholarCrossref
15.
Elixhauser  A, Steiner  C, Harris  DR, Coffey  RM.  Comorbidity measures for use with administrative data.  Med Care. 1998;36(1):8-27.PubMedGoogle ScholarCrossref
16.
Barnato  AE, Alexander  SL, Linde-Zwirble  WT, Angus  DC.  Racial variation in the incidence, care, and outcomes of severe sepsis: analysis of population, patient, and hospital characteristics.  Am J Respir Crit Care Med. 2008;177(3):279-284.PubMedGoogle ScholarCrossref
17.
Agency for Healthcare Research and Quality.  Healthcare Cost and Utilization Project (HCUP) Procedure Classes. Rockville, MD: Agency for Healthcare Research & Quality; 2014.
18.
Begg  MD, Parides  MK.  Separation of individual-level and cluster-level covariate effects in regression analysis of correlated data.  Stat Med. 2003;22(16):2591-2602.PubMedGoogle ScholarCrossref
19.
Enders  CK, Tofighi  D.  Centering predictor variables in cross-sectional multilevel models: a new look at an old issue.  Psychol Methods. 2007;12(2):121-138.PubMedGoogle ScholarCrossref
20.
Dimick  JB, Ryan  AM.  Methods for evaluating changes in health care policy: the difference-in-differences approach.  JAMA. 2014;312(22):2401-2402.PubMedGoogle ScholarCrossref
21.
Centers for Disease Control and Prevention.  Interruptions in supplies of second-line antituberculosis drugs—United States, 2005-2012.  MMWR Morb Mortal Wkly Rep. 2013;62(2):23-26.PubMedGoogle Scholar
22.
Mayr  FB, Yende  S, Angus  DC.  Epidemiology of severe sepsis.  Virulence. 2014;5(1):4-11.PubMedGoogle ScholarCrossref
23.
Angus  DC, van der Poll  T.  Severe sepsis and septic shock.  N Engl J Med. 2013;369(9):840-851.PubMedGoogle ScholarCrossref
24.
Morelli  A, Lange  M, Ertmer  C,  et al.  Short-term effects of phenylephrine on systemic and regional hemodynamics in patients with septic shock: a crossover pilot study.  Shock. 2008;29(4):446-451.PubMedGoogle Scholar
25.
Morelli  A, Ertmer  C, Rehberg  S,  et al.  Phenylephrine versus norepinephrine for initial hemodynamic support of patients with septic shock: a randomized, controlled trial.  Crit Care. 2008;12(6):R143.PubMedGoogle ScholarCrossref
26.
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(3):486-552.PubMedGoogle ScholarCrossref
27.
Singer  M, Deutschman  CS, Seymour  CW,  et al.  The third international consensus definitions for sepsis and septic shock (Sepsis-3).  JAMA. 2016;315(8):801-810.PubMedGoogle ScholarCrossref
28.
Hall  WB, Willis  LE, Medvedev  S, Carson  SS.  The implications of long-term acute care hospital transfer practices for measures of in-hospital mortality and length of stay.  Am J Respir Crit Care Med. 2012;185(1):53-57.PubMedGoogle ScholarCrossref
×