Outcomes of patients with hypertensive emergency managed in intensive versus intermediate care settings: A Multi-hospital Retrospective Cohort Study

Abstract

Background:

Hypertensive emergency is characterized by end-organ dysfunction due to markedly elevated blood pressure. Though guidelines recommend rapid blood pressure lowering with intravenous anti-hypertensives and continuous monitoring in an intensive care unit (ICU), many patients are managed in intermediate care (IMC) settings. In this study, we compare outcomes of hypertensive emergency patients admitted to ICU versus IMC settings.

Methods:

Adult patients in the emergency departments of three hospitals were included if they had ≥2 systolic blood pressure measurements >180 mmHg within 12-hours of one another, were treated with continuous intravenous anti-hypertensive medications, and were subsequently admitted to an ICU or IMC setting. We excluded patients receiving mechanical ventilation in the ED and those presenting with acute coronary syndrome, aortic dissection, or acute cerebrovascular accident. The primary outcome was hospital length of stay (LOS) penalized for death and adjusted for baseline patient features. Secondary outcomes included ICU and IMC LOS, and hospital mortality. Process outcomes included arterial line usage, frequency of blood pressure measurements in the first 24 hours, and escalations of care. Lastly, we report the proportion of patients reaching blood pressure targets and/or episodes of hypotension in each setting.

Results:

ICU patients (n=649) were younger with a male predominance, and more frequently received non-invasive respiratory support or were dialysis dependent than IMC patients (n=629). In an adjusted analysis there was no difference in time to hospital discharge between patients admitted to ICU versus IMC settings (absolute difference: +0.29 days [95%CI: −0.07, 0.70]). Similarly, there was no difference in ICU versus IMC LOS, hospital readmission rate, or hospital mortality. Blood pressures were more frequently measured in the ICU, but there was no difference in time to target blood pressure goal. Any hypotensive episode in the first 24 hours was more common in the ICU.

Conclusions:

In this retrospective observational study of patients with hypertensive emergency, we did not observe significantly different LOS outcomes between patients admitted to ICU versus IMC settings, nor did we detect a signal of harm among those admitted to IMC. IMC settings may be a reasonable alternative to ICU admission for select patients with hypertensive emergency.

Introduction

Hypertensive emergency is prevalent and characterized by end-organ dysfunction due to markedly elevated blood pressure.^1,2 The American College of Cardiology and the American Heart Association give a Class I recommendation for admission of such patients to an intensive care unit (ICU) for controlled, rapid lowering of blood pressure with close blood pressure monitoring.³ In contrast, the British and Irish Hypertension Society recommends admission to either an ICU or intermediate care (IMC) setting.⁴ Both groups advocate for the use of intravenous anti-hypertensive agents and invasive blood pressure monitoring (i.e., arterial line) for rapid and controlled lowering of blood pressure.^3,4 However, in observational studies in the United States, 38–52% of patients with hypertensive emergency are cared for outside the ICU,^5,6 and fewer than half receive invasive blood pressure monitoring.⁶ Data comparing outcomes of these patients treated in different care settings is lacking.

To preserve ICU beds for more severely ill patients, most hospitals in the United States and Europe maintain IMC settings.^7,8 IMC is intended for patients who do not need all of the human and technical resources of an ICU, but whose care needs exceed what is possible on a general ward.⁹ As an alternative to intensive care for some patients, IMC may be appropriate for many patients with hypertensive emergency because initiation of intravenous anti-hypertensive agents should place them on a predictable clinical trajectory of improvement.⁹ IMC may even be superior to intensive care for some patients in light of an ICU culture that tends towards intervention, invasive monitoring, and testing, even when potentially unnecessary.¹⁰ Moreover, admission to IMC may decrease hospital length of stay (LOS), as many ICUs are not staffed and organized for efficient patient discharge.^11,12 However, the lower nurse-to-patient staffing ratios of IMC, and less invasive and less frequent monitoring, could also result in delays in blood pressure control, unintended overcorrection, or both.

To characterize outcomes of patients with hypertensive emergency initially treated in ICU versus IMC settings, we conducted a retrospective observational study of ED patients meeting hypertensive emergency criteria without other clear indications for ICU admission. We hypothesized that after adjustment for illness and patient characteristics, patients with hypertensive emergency initially admitted to IMC settings would be discharged alive more quickly. If this could be accomplished safely, it would have important implications for ICU utilization.

Methods

The Johns Hopkins University School of Medicine Institutional Review Board acknowledged this analysis as exempt from human subject review and granted a waiver of consent for secondary use of data (IRB00453058).

Study Design and Setting

We included patients admitted to one of three hospitals within the Johns Hopkins Health System (JHHS) where intravenous anti-hypertensive agents and arterial lines were permitted in both ICU and IMC settings. These included one academic hospital (1038 beds) with two contributing IMC settings that house medical patients (12 and 21 beds), and two community hospitals (226 beds each), both with IMC settings (17 and 20 beds) that house both medical and surgical patients. All IMC settings maintained a nurse-to-patient ratio of 1:3 and provided all patients with continuous pulse-oximetry and cardiac telemetry. While each IMC has admission guidelines, these do not direct triage decision beyond detailing which services and monitoring are permitted (Table 1). ICUs contributing patients to the study included the non-surgical cardiac critical care unit (12 beds) and medical ICU (24 beds) of the academic hospital, and three mixed specialty ICUs (16, 12, and 12 beds) in the two community hospitals.

Table 1:

Organizational features of intermediate care settings

IMC Feature	IMC1a	IMC2	IMC3	IMC4b
Unit Structure/Process
Hospital type	Academic	Academic	Community	Community
Number of IMC beds	21	12	17	20
Stand-alone unit	Yesc	Yesd	Yese	Yesf
Unit specialty	Med/non-cardiac	Med/cardiac	Mixed med/surg	Mixed med/surg
Intensivist approves beds	No	No	Yes	Yes
Staffing
Provider staffing	Closed	Closed	Open	Open
Attending specialty	Internist	Cardiologist	Intensivist	Mixed
Intensivist consult req	No	No	No	No
In-house 24-attg coverage	No	No	Yes	Yes
Nurse:patient ratio	1:3	1:3	1:3	1:3
Ancillary staffing
Dedicated pharmacist	Yes	No	No	Yes
On-site respiratory therapist	Yes	No	No	Yes
Nursing technician/shift	2	≤ 1	2	2
Multi-disciplanry rounds/week	3	5	7	7
Monitoring
Continuous 12-lead telemetry	Yes	Yes	Yes	Yes
Continuous pulse oximetry	Yes	Yes	Yes	Yes
Frequency of vital signs	Every 4 hours	Every 4 hours	Every hour	Every hour
Arterial line monitoring	Yes	Yes	Yes	Yes
Pulmonary artery catheter	No	Yes	No	No
Frequency of intake/output assessments	Every 4 hours	Every 4 hours	Every 2 hours	Every hour
Frequency of glucose check	Every hour	Every hour	Every hour	Every hour
Frequency of lab draws	Every 4 hours	Every 4 hours	Every hour	Every hour
Frequency of neuro checks	Every 4 hours	Every 4 hours	Every 2 hours	Every 2 hour
Interventions/Treatments
Frequency of nebulizers	Every 2 hours	Every 2 hours	Every 3 hours	Every 2 hours
Continuous nebulizer treatment	No	No	No	No
High flow nasal oxygen	Yes	Yes	Yes	Yes
New continuous positive airway pressure	Yes	Yes	Yes	Yes
New bilelelvel positive airway pressure	Yes	Yes	Yes	Yes
Invasive ventilation via endotracheal tube	No	No	No	No
Invasive ventilation via tracheostomy	Yes	Yes	Yes	Yes
Bedside bronchoscopy	Yes	No	No	Yes
Continuous anti-hypertensive infusion	Yes	Yes	Yes	Yes
Vasopressors	No	No	Yes	Yes
Inotrope infusion	Yes	Yes	Yes	Yes
Continuous rate controlling infusion	Yes	Yes	Yes	Yes
Left ventricular assist device	No	Yes	No	No
Pericardial drain	No	Yes	No	Yes
Intermittent hemodialysis at bedside	Yes	Yes	Yes	Yes
Continuous renal replacement therapy	No	No	No	No
Thrombolytics for peripheral thrombosis	Yes	Yes	Yes	Yes
Thrombolytics for central thrombosis	No	Yes	No	No
Thrombolytics for peripheral arterial thrombosis	Yes	Yes	No	No
Plasmapheresis	Yes	Yes	Yes	Yes
Plasma exchange	Yes	Yes	Yes	Yes
Continuous benzodiazepine infusion	Yes	No	No	No
Continuous naloxone infusion	Yes	No	Yes	Yes
Other
Arterial sheaths	No	Yes	Yes	Yes
Large bore venous resuscitation line	No	Yes	Yes	Yes
Bladder pressure monitor	No	Yes	No	Yes
Procedural sedation	No	No	No	Yes

IMC = intermediate care

Academic = has a residency/fellow training program

Med/surg = medical and surgical patients are cared for in this location

Closed staffing model = there is an on-site provider team responsible for patients on the unit who do not have care responsibilities elsewhere in the hospital

Open staffing model = multiple different provider teams may have patients admitted to the IMC setting.

Mixed = intensivists and non-intensivists (e.g., hospitalists) both staff patients on this unit

^aOpen staffing model before March 2020. Closed staffing model thereafter.

^bOnly contribute data beginning in 2021 when arterial lines and intravenous anti-hypertensives became permissible.

^cIMC1 is 5 floors down from the medical ICU in an adjacent building connected by a bridge. It is on the same floor as the non-surgical cardiac ICU in an adjacent building connected by a bridge.

^dIMC2 is 5 floors down from the medical ICU in an adjacent building connected by corridors on several levels. It is on the same floor as the non-surgical cardiac ICU in an adjacent building connected by corridors on several levels.

^eIMC3 is in the same building down the hall from the ICU

^fIMC4 is in the same building, one floor down from the ICUs.

Population

Inclusion criteria were: 1) age ≥18 years, 2) presentation to the emergency department (ED) of one of the three contributing hospitals, 3) ≥2 systolic blood pressure (SBP) readings >180 mmHg in the ED and within 12 hours of each other, 4) treatment with a continuous infusion of either nicardipine, nitroglycerin, or nitroprusside initiated in the ED, and 5) admission from the ED directly to an ICU or IMC setting. Patients were excluded if a condition suggesting a definitive need for ICU care was present in the ED (i.e., receipt of mechanical ventilation, acute coronary syndrome operationally defined as a cardiac catheterization within 24 hours of presentation, or aortic dissection or acute cerebrovascular accident defined by International Classification of Diseases version 10 [ICD-10] code present on admission). The unit of observation was the hospital admission, and for patients with repeat admissions meeting study inclusion criteria, each of these hospitalizations was included.

Data Source and Covariates

We used electronic medical record (EMR) data from the Johns Hopkins Critical Illness and Recovery Center of Excellence registry, which contains comprehensive EMR data for JHHS patients admitted to ICU and/or IMC settings.¹³ For this analysis, we included data from July 1^st, 2017 through December 31^st, 2024. We extracted patient features including age, gender, race, Elixhauser comorbidity index,¹⁴ COVID-19 hospitalization status (by ICD-10 codes),¹⁵ dialysis dependence,¹⁶ dialysis in the first 24 hours of presentation, documented vital signs, use of advance respiratory support in the ED or after admission (non-invasive ventilation [NIV], high-flow nasal oxygen [HFNO], invasive mechanical ventilation [IMV]), and maximal sequential organ failure score (SOFA) measured in the ED.¹⁷ We also extracted administrative data (e.g., ICU, IMC, and hospital length of stay[LOS]) and duration of continuous infusions of anti-hypertensive agents to characterize patient outcomes. Further details on covariate definitions and missingness are included in Table E1. For missing data on SOFA scores we used imputation of normal values if missing sub-scores. After this imputation only 1 other variable, mean arterial blood pressure in the ED, had missing values (n=13, 1%), and all variables for outcomes and adjustment were complete.

Outcomes

The primary outcome of this study was days to hospital discharge alive, with in-hospital death placed at the 99^th percentile of the distribution to account for potential bias from early death.¹⁸ Secondary outcomes included days to first transfer out of the ICU or IMC to a ward or hospital discharge directly from the ICU or IMC, and hours of anti-hypertensive infusion (end time indicated by beginning of first 8 hour interval off anti-hypertensive infusion), both penalized for death. Other secondary outcomes were hospital discharge location (ICU, IMC, or acute care ward), in-hospital death or discharge to hospice, and 30-day readmission rate to a JHHS hospital. Unless otherwise stated, time zero for all time-based outcomes was initiation of the continuous anti-hypertensive infusion. Process outcomes included use of an arterial line for blood pressure monitoring, number of blood pressure measurements documented in the EMR during the first 24 hours following admission and escalation of advanced respiratory support. Lastly, we report blood pressure control effectiveness and safety outcomes. We characterized blood pressure control effectiveness with three measures: 1) the proportion of patients who reached a target reduction in SBP of 25% in the first hour of intravenous anti-hypertensive therapy, 2) goal SBP of < 160 mmHg reached between hours 2 and 6 after initiation of infusion, and 3) maintenance of SBP < 160 from 6 to 24 hours (documented within each time frame).² Adverse safety events were defined by a composite measure of SBP < 90 mmHg, reduction in SBP of > 40% by hour 6, or mean arterial pressure < 65 mmHg in the first 24 hours after start of the anti-hypertensive infusion.⁵ We also report the proportion of IMC patients transferred to the ICU within 48 hours of admission and at any time during the admission.

Statistical Analysis

Cohort and process outcomes by initial care setting following ED encounter (ICU versus IMC) were summarized using counts and proportions for categorical variables, and medians and interquartile ranges (IQRs) for continuous variables. They were compared statistically with Wilcoxon rank-sum and chi-square tests for continuous and categorical variables, respectively. All standard errors were calculated accounting for clustering at the patient level.

For our primary and key secondary outcome analyses, we used a log transformation to approximate a normal distribution for continuous time-based outcomes. We then used univariable and multivariable linear regression models to compare outcomes for patients admitted to ICU versus IMC settings. We initially prespecified covariates including age, race, gender, COVID-19 status, maximum SBP in the ED, use of advanced respiratory support in the ED (except IMV), Elixhauser Comorbidity Index, dialysis dependence, SOFA Score, and hospital. After review, we excluded race and gender from the primary analysis to avoid adjusting away differences that could represent disparity.¹⁹ They are retained in a sensitivity analysis. In these models, exponentiated coefficients are interpreted as the relative percent difference in outcome duration. We additionally calculated the effects on the absolute day scale to provide a more clinically interpretable estimate.²⁰ We used counterfactual predictions using a standardization approach and estimated the average difference in time to discharge. To estimate confidence intervals, we employed clustered bootstrapping with the percentile method using 1000 replicates.²¹ For 30-day readmission and in-hospital death outcomes, we constructed univariate and multivariate logistic regression models adjusted for the same covariates noted above. All blood pressure control effectiveness and safety outcomes were compared using unadjusted Wilcoxon rank-sum and chi-square tests accounting for clustering by patient.

Finally, in sensitivity analyses we used the same methods to evaluate primary and secondary outcomes (1) when race and gender were retained as covariates, (2) with study quarter as a covariate to detect changes in outcomes over time, and (3) among patients chronically dependent on dialysis. Lastly, we assessed outcomes related to blood pressure control among patients without arterial lines. Study reporting follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidance (Table E2).²² Analyses were done in R (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria).

Results

There were 4,550 patient encounters in which systolic blood pressure was >180 mmHg on ≥2 assessments during the first 12 hours in the ED. After excluding 2859 encounters during which intravenous anti-hypertensive infusions were not used, 412 for whom a clear ICU indication was present, and 1 who was still admitted at the time of database lock, there were 1278 encounters among 1008 individual patients in the final cohort (Figure 1 and Figure E1).

Figure 1: Hypertensive emergency managed in intensive versus intermediate care settings cohort:

Flow diagram for cohort creation for the hypertensive emergency in ICU versus IMC settings

Patients admitted to the ICU (n=649) were younger with a male predominance, more often presented with dialysis dependence and received dialysis in the first 24 hours of admission, had higher median blood pressures (SBP, diastolic blood pressure, and mean arterial pressure) and SOFA Scores, and more often received advance respiratory support (NIV or HFNO) than those admitted to IMC settings (n=629; Table 2). Comorbidities did not differ between groups. After log transformation, LOS and duration of continuous anti-hypertensive infusion data were normally distributied (Figure E2).

Table 2:

Characteristics of Patients with Hypertensive Emergency Admitted to the Intensive Care Versus Intermediate Care Setting

Characteristics	ICU n=649 Encounters n=530 Patients	IMC n=629 Encounters n=551 Patients	p-value
Demographics/Comorbidities
Age, years, median [IQR]	57 [47, 67]	61 [52, 71]	<0.001
Female, n (%)	284 (44)	314 (50)	0.04
Race, n (%)
American Indian, Alaska Native	0 (0)	2 (0.3)	0.14
Asian	18 (3)	23 (4)
Black or African American	523 (81)	474 (75)
Other	33 (5)	47 (7)
White	75 (12)	83 (13)
Hispanic Ethnicity, n (%)	18 (3)	27 (4)	0.35
Elixhauser Comorbidities, median [IQR]	10 [7, 14]	10 [7, 13]	0.32
ESRD Present on Admission, n (%)	268 (41)	207 (33)	<0.01
Current Illness Characteristics
Highest SBP in ED, median [IQR]	231 [216, 248]	228 [213, 244]	<0.01
Highest DBP in ED, median [IQR]	132 [117, 150]	125 [111, 142]	<0.001
Highest MAP in ED, median [IQR]	166 [149, 184]	162 [144, 175]	<0.001
SOFA Score, median [IQR]	4 [2, 5]	2 [1, 4]	<0.001
COVID-19, n (%)	25 (4)	21 (3)	0.61
NIV or HFNO in ED, n (%)	184 (28)	132 (21)	<0.01
Dialysis in 1st 24 Hours, n (%)	216 (33)	155 (25)	<0.01
Hospital, n (%)
Academic Hospital	498 (77)	474 (75)	0.01
Community Hospital #1	115 (18)	93 (15)
Community Hospital #2	36 (6)	62 (10)

Definition of Abbreviations: ESRD=End-stage renal disease; ED=Emergency Department; HFNO=High-flow Nasal Oxygen; ICU=Intensive Care Unit; IMC=Intermediate Care; IQR=Interquartile range; MAP=Mean arterial pressure; NIV=Non-invasive ventilation; SBP=Systolic Blood Pressure; SOFA=Sequential Organ Failure Assessment Score

There was no difference in the unadjusted primary outcome of hospital LOS, which was 4.8 days (IQR:3.0-8.4) for ICU patients and 4.3 days (IQR:2.6-7.4) for those admitted to an IMC setting (Table 3). In the adjusted analysis, this indicated a nonsignificant increase in hours in the hospital by a factor of 1.06 (95% CI:0.96-1.16, p=0.27) and a risk difference of 0.29 more hospital days (95% CI:-0.07-0.70) among those admitted to an ICU versus IMC setting. Similarly, no difference was appreciated between groups when comparing LOS in ICU versus IMC settings, odds of readmission after hospital discharge, or in-hospital death (including hospice). However, ICU patients remained on intravenous anti-hypertensives approximately three hours longer (13.2 vs. 10.2 hours [unadjusted comparison]). In an adjusted analysis, this translated to a 1.3 times longer (95% CI:1.13-1.50, p<0.001) use of intravenous anti-hypertensives, or an adjusted marginal effect of 2.72 hours more (95% CI:1.69-3.93, p<0.001). Sensitivity analyses of primary and secondary outcomes in models including race and gender, study quarter, and in patients chronically dependent on dialysis showed nearly identical finding (Tables E3-E5).

Table 3:

ICU versus Intermediate Care As Initial Location for Hypertensive Emergencyand Patient Outcomes

Outcome	ICU n=649 Encounters n=530 Patients	IMC n=629 Encounters n=551 Patients	Adjusted Analysesa,b Estimate [95% CI], %(ICU vs IMC)	P Valuec
Primary Outcome	Median [IQR]	Median [IQR]	Adjusted % Difference and Adjusted Absolute Difference
Days to Discharge Alive (Death or DC to Hospice Ranked at 99th Percentile), Days	4.8 [3.0, 8.4]	4.3 [2.6, 7.4]	1.06 [0.96, 1.16] 0.29 days [−0.07, 0.70]	0.27
Secondary Outcomes
Days at ICU/IMC Level of Care (Death or DC to Hospice Ranked at the 99th Percentile), Days	2.0 [1.2, 3.2]	1.9 [1.2, 3.3]	0.94 [0.86, 1.02] −0.14 days [−0.28, 0.01]	0.13
Hours on Anti-Hypertension Medication Infusion, (Death or DC to Hospice Ranked at the 99th Percentile), Hours	13.2 [5.6, 27.3]	10.2 [4.4, 19.6]	1.30 [1.13, 1.50] 2.72 hours [1.69, 3.93]	<0.001
	n (%)	n (%)	Adjusted Odds Ratio
30-day Readmission to Health System Hospital, n (%)	85 (13)	82 (13)	1.00 [0.70, 1.42]	1.0
In-hospital Death or Discharge to Hospice, n (%)	26 (4)	10 (2)	2.05 [0.84, 5.23]	0.14

Data and Abbreviations: ICU=Intensive Care Unit; IMC=Intermediate Care; DC=Discharge; CI=Confidence Interval

^aDays to outcomes all calculated with linear regression with log-transformed outcome and association measures reported on the relative difference scale and absolute difference scale. Logistic regression is used readmission and in-hospital death or hospice discharge outcomes.

^bModels adjusted for: Age, Elixhauser mortality index, maximum sequential organ failure assessment (SOFA) score in the emergency department, SARS-Cov2 Status, dialysis dependent on admission, maximum SBP in the emergency department, use of advanced respiratory support (non-invasive positive pressure ventilation [NIPPV]/high flow nasal oxygen[HFNO]) in the emergency department, and hospital where the patient was admitted.

^cP-value obtained from statistical testing of regression coefficient

There were notable differences in process outcomes between groups (Table 4). Patients admitted to the ICU had blood pressure documented more frequently in the first 24 hours of admission (45 measurements; IQR:34, 60 vs. 20 measurements; IQR:12, 31; p<0.001), more often had an arterial line (24% vs. 5%; p<0.001), and more often progressed to require IMV (12% vs. 0.4%; p<0.001) or NIV (8% vs 5%; p=0.02). Lastly, hospital discharge alive directly from the ICU was less common than discharge from an IMC setting (17% vs. 34%; p<0.001).

Table 4:

Process and Blood Pressure Control Outcomes for Patients Admitted to the ICU or Intermediate Care with Hypertensive Emergency

Outcome	ICU n=649 Encounters n=530 Patients	IMC n=629 Encounters n=551 Patients	P-value
Process Outcomes
BPs Documented in 1st 24 Hours of ICU/IMC Stay, median (IQR)	45 [34, 60]	20 [12, 31]	<0.001
Presence of Arterial Line, n (%)	156 (24)	30 (5)	<0.001
Intubated within 1st 48 Hours, n (%)	80 (12)	3 (0.4)	<0.001
NIV Started within 1st 48 Hours, n (%)	50 (8)	28 (5)	0.02
IMC to ICU transfer within 1st 48 Hours, n (%)	NA	19 (3)	NA
IMC to ICU transfer ever, n (%)	NA	27 (4)	NA
Discharge Alive Directly from ICU/IMC, n (%)	113 (17)	215 (34)	<0.001
Blood Pressure Control Outcomes
SBP reduced by 25% in 1st hour on Drip, n (%)	164 (25)	149 (24)	0.52
SBP lowered to ≤ 160 mmHg in 1st 6 hours on Drip, n (%)	463 (71)	456 (73)	0.66
SBP Maintained at ≤ 160 mmHg in hours 6-24, n (%)	55 (9)	86 (14)	0.003
% of SBP Documentations at ≤160 mmHg in hours 6-24, median patient percent [IQR]	52 [22, 81]	55 [25, 83]	0.45
Hypotension within 1st 24 Hours of Starting Drip, n (%)a	246 (38)	148 (24)	<0.001
% of Hypotension Documentations in 1st 24 hours, median patient percent [IQR]	0 [0, 4]	0 [0, 0]	<0.001

Definition of Abbreviations: BP=Blood Pressures; ICU=Intensive Care Unit; IMC=Intermediate Care; IQR=Interquartile range; NA=Not applicable; SBP=Systolic Blood Pressure

^aDefined as SBP≤90 mmHg or mean arterial pressure ≤65 during the first 24 hours of continuous anti-hypertensive infusion, or SBP≤60% of baseline in the first 6 hours after starting the continuous anti-hypertensive infusion.

From the perspective of blood pressure control effectiveness, there was no difference in the proportion of ICU versus IMC patients that achieved target SBP in the first hour, or first six-hour interval, after starting treatment with continuous intravenous anti-hypertensives (Table 4). Only a minority of patients maintained SBP below goal between hours 6-24, and this was less likely to be achieved in ICU versus IMC settings (9% vs. 14%; p=0.003). However, from the perspective of the percent of measurements at target between hours 6-24, there was no difference (median patient percentage of measured blood pressure ≤160 mmHg, 52% vs. 55%, p=0.45). From a safety perspective, any incidence of hypotension was more common among ICU versus IMC (38% vs. 24%; p<0.001). Lastly, 3% of IMC patients were transferred to the ICU in the first 48 hours of admission and this only increased to 4% for the duration of the hospitalization. Sensitivity analyses of blood pressure control outcomes among patients without arterial lines were similar (Table E6).

Discussion

In this study, we compared outcomes of patients with hypertensive emergency admitted from the ED to ICU or IMC settings. We observed no difference in the primary outcome of hospital LOS, or in secondary outcomes of LOS in the ICU or IMC settings, odds of hospital readmission, or odds of death. While a statistically significant difference in time on intravenous anti-hypertensive agents was noted (longer in the ICU), this is unlikely to affect clinical or operational outcomes in an important way. Several process outcomes also differed between groups, which appear to correlate with more intensive monitoring of a sicker population. Lastly, while time to target blood pressure was similar between groups, blood pressures above goal between 6- and 24- hours were common in both groups, whereas episodes of hypotension were more common in the ICU setting.

We hypothesized that discharge of hypertensive emergency patients to home after admission to an IMC setting would be more expedient than after admission to an ICU, as ICUs are often not well resourced for this activity.^11,12 However, despite adjustment for severity of illness and other illness feature metrics, we did not observe a significantly shorter hospital LOS. While residual confounding is certainly a possibility, it is also possible that IMC settings, like ICUs, are organized more around acute management than discharge operations. Alternatively, comorbidities, which were similar between groups, may have a greater effect on hospital LOS than the relatively short duration of acute critical care management (continuous anti-hypertensive infusion) in this population. Regardless, nearly half of our population with hypertensive emergency was admitted to an IMC setting and achieved outcomes that were similar to their ICU counterparts. Policies that mandate admission of these patients to an ICU could limit the availability of critical care to more severely ill, contemporaneous patients.²³ Moreover, assuming the daily cost of operating an ICU bed is greater than that of an IMC bed, admitting appropriate patients to the IMC rather than the ICU could reduce patient care fees. However, the net effects on societal costs depend in complex ways on where the resources for the IMC come from (e.g., closing ICU beds, repurposing ward beds, or adding entirely new beds) and how the system adapts to these total capacity changes and their cascading effects. When navigating these complexities, a main theme to keep in focus is improved ICU utilization.²⁴

Indeed, the interest in creating IMC arose from a recognition that the life-saving resources of ICUs are both expensive and limited in supply,²⁵ and patients who qualify for intensive care are more likely to die if this resource is not available.^26,27 For these reasons, IMC was developed to unload ICUs of patients who simply needed close monitoring rather than active interventions.^28-30 In some studies, IMC has had the desired effect; sicker patients have better access to the ICU, while those admitted to IMC do well.^23,31-33 However, other studies have reported poor outcomes among patients erroneously identified as being appropriate for IMC.³⁴ While guidelines for IMC admission have been generated, they are largely based on expert opinion rather than rigorous investigation,^35-37 and ICU versus IMC triage decisions rely heavily on clinician judgement.³⁸

In our study, we add to an evolving literature addressing which patients are appropriate for IMC.^39-42 We chose to study hypertensive emergency (with a few easily identifiable exclusions: use of IMV, stroke, aortic dissection, and acute myocardial infarct), because it is a common emergency department presentation,^43-45 and typically stabilizes following initiation of therapy. Though we cannot interpret our results causally, we found no association between IMC admission and worse outcomes. Patients in both ICU and IMC settings achieved target blood pressures on a similar timeline, and overcorrection of blood pressure and hypotension was less common among patients in the IMC setting. This is significant because overcorrection of blood pressure in the setting of hypertensive emergency is strongly associated with adverse cerebrovascular and cardiovascular events.^46-48 Moreover, the potential for overcorrection has been cited as the primary rational for ICU admission and use of intra-arterial blood pressure monitoring for this syndrome.^2,49 Though we cannot comment on how providers made decisions to admit patients to ICU or IMC settings, or use arterial lines, we presume our exclusion criteria selected for patients that were more likely to do well in an IMC with less invasive monitoring.

Syndromes that may have a similar trajectory of improvement and could also be appropriate for IMC include select patients with diabetic ketoacidosis, bronchodilator responsive COPD exacerbations requiring non-invasive respiratory support, and those with sepsis who do not require vasopressors.⁹ While these conditions require more frequent and intensive monitoring than is feasible on an acute care ward, many patients with these conditions are unlikely to deteriorate and require the advanced technical and human resources of an ICU. In our population, only 4% of IMC patients required transfer to the ICU (3% in the first 48 hours), which is low compared to the ~12% IMC failure rate previously reported for general populations of emergency department patients admitted to IMC.^50,51

While randomized trials evaluating different levels of critical care for patients like these would be optimal, they would be methodologically and operationally challenging. Future studies using rigorous causal inference methods may be a good alternate analytical strategy for comparisons of different levels of critical care. Evidence of this sort would inform policy makers and clinicians who manage these patients.^52,53

Limitations

There are several limitations to this study. First, the data are reflective of care processes (e.g., triage practices, IMC services and monitoring, IMC admission guidelines) and outcomes in a subset of hospitals in a single health system, which may limit generalizability. While there was good overlap between features of our population and others (e.g., similar age, gender, race, qualifying SBP, dialysis dependence, and hospital LOS),^5,6,43 ours was treated with intravenous anti-hypertensives for longer (10.2 to 13.2 vs. 6.5 hours),⁶ and more patients experienced hypotensive events (24% to 38% versus 4% to 14%).^5,6 While our mortality was lower than two other studies (1% vs 4.8% and 6.9%),^6,43 those did not exclude patients with clear indications for ICU admission. Second, our definition of hypertensive emergency is pragmatic. We did not review patient records directly for evidence of acute organ dysfunction related to hypertension at the time of admission. It is therefore possible some patients in our population may have better fit a definition of hypertensive urgency, which does not require aggressive intervention and perhaps could be managed as an outpatient.^54,55 However, our definition incorporated clinician judgement as the use of an antihypertensive infusion in the emergency department is generally reserved for cases presumed to represent hypertensive emergency. Third, we have presented a retrospective, observational, cohort study. While we did not detect adverse associations with IMC admission, our study is descriptive and not designed to discriminate between which patients are most appropriate for ICU and IMC settings; we cannot comment on how patients admitted to the ICU would have done if admitted to the IMC setting. The higher SOFA scores in the emergency department, and more frequent progression to IMV after admission, suggests clinicians distributed sicker patients to the ICU. While our analysis adjusted for severity of illness and comorbidity, there is likely residual confounding by indication. Fourth, our assessments of effectiveness and safety are limited to blood pressures that were documented in the EMR.⁹ Fewer blood pressure measurements in IMC patients could have missed out of range blood pressures. Lastly, readmission rates at 30 days are dependent on readmission to a health system hospital. We would not have been able to detect readmissions to other hospitals.

Conclusions

In a retrospective observational study of patients with hypertensive emergency, we did not detect significant differences in LOS outcomes between patients admitted to ICU versus IMC settings, nor did we detect any signals indicating worse clinical outcomes among those admitted to IMC. Further study is indicated to more clearly define which patients with hypertensive emergency may be safely managed in an IMC setting with the objective of preserving ICU beds for those with more severe illnesses and less predictable clinical trajectories.