Intensive care unit utilization after implementation of minor severe pneumonia criteria in real-time electronic clinical decision support

Abstract

Objective:

To determine if the implementation of automated clinical decision support (CDS) with embedded minor severe community acquired pneumonia (sCAP) criteria was associated with improved Intensive Care Unit (ICU) utilization among Emergency Department (ED) patients with pneumonia who did not require vasopressors or positive pressure ventilation on admission.

Design:

Planned secondary analysis of a stepped-wedge, cluster-controlled CDS implementation trial.

Setting:

16 hospitals in 6 geographic clusters from Intermountain Health; a large, integrated, nonprofit health system in Utah and Idaho.

Patients:

Adults admitted to the hospital from the ED with pneumonia identified by: 1) discharge ICD-10 codes for pneumonia or sepsis/respiratory failure and 2) ED chest imaging consistent with pneumonia, who did not require vasopressors or positive pressure ventilation on admission.

Intervention:

After implementation, patients were exposed to automated, open-loop, comprehensive CDS that aided disposition decision (ward versus ICU), based on objective severity scores (sCAP).

Measurements and Main Results:

The analysis included 2747 patients, 1814 before and 933 after implementation. The median age was 71, median Elixhauser index was 17, 48% were female and 95% were Caucasian. A mixed effects regression model with cluster as the random effect estimated that implementation of CDS utilizing sCAP increased 30-day ICU-free days by 1.04 days (95% confidence interval [CI] 0.48, 1.59; p<0.001). Among secondary outcomes, the odds of being admitted to the ward, transferring to the ICU within 72 hours and receiving a critical therapy decreased by 57% (odds ratio [OR] 0.43 95% CI 0.26, 0.68; p<0.001) post-implementation; mortality within 72 hours of admission was unchanged (OR 1.08 95% CI 0.56, 2.01; p=0.82) while 30-day all-cause mortality was lower post-implementation (OR 0.71, 95% CI 0.52, 0.96; p=0.03).

Conclusion:

Implementation of electronic CDS using minor sCAP criteria to guide disposition of patients with pneumonia from the ED was associated with safe reduction in ICU utilization.

Introduction

More than 6 million adult cases of pneumonia occur annually in the United States, leading many patients to be evaluated in the emergency department (ED) and admitted to a hospital for further treatment.^1,2 European and American pneumonia guidelines recommend objective severity assessments to guide treatment location (home, general ward, or intensive care unit [ICU])^3,4 for patients because their use is associated with improved patient outcomes and processes of care.^5–8

Whereas Clinical Decision Support (CDS) utilizing these assessments safely improves the number of patients treated at home and can be automated and disseminated,^7–9 little beyond validation of risk scores has been done for determining admission to the ward versus the ICU for those being admitted.^10–13 The consequences of inappropriate disposition for inpatients are significant. Delayed ICU admission (first going to the floor and then transferring to the ICU within 72 hours) is associated with increased mortality and length of stay^14–16 while misallocation of ICU resources to low-risk patients increases costs and risks associated with unnecessary interventions.

One recommended objective severity assessment is the Infectious Disease Society of America/American Thoracic Society (IDSA/ATS) Severe Community Acquired Pneumonia (sCAP) “minor criteria” (“major” criteria are use of vasopressors or mechanical ventilation).^3,4 Utilizing data universally available in EDs and validated in multiple retrospective cohorts, the sCAP minor criteria predict need for ICU admission within 72 hours of ED arrival with sensitivity of 57% and specificity of 91%.^10,12–14

Beginning in 2017, Intermountain Health (a large non-profit health system in the western United States) implemented automated CDS for pneumonia (ePNa) that included the sCAP minor criteria to guide ICU disposition into 16 community hospitals in a pragmatic stepped-wedge clinical trial across 6 hospital clusters. A previously published manuscript from this data showed that 30-day all-cause mortality was reduced 38% and outpatient disposition increased 61% post ePNa implementation. How ICU utilization was impacted was not fully explored. sCAP logic is embedded within ePNa and prompts providers to consider ICU admission at appropriate thresholds for every patient. Thus, ePNa implementation provided the first opportunity to evaluate minor sCAP criteria guided disposition for patients being admitted across a large health system. We hypothesized that implementation of ePNa was associated with a safe reduction in ICU utilization among patients for whom clinical judgement alone was previously used to determine disposition.

Materials and Methods

Study design:

Planned secondary analysis of a cluster-controlled, stepped-wedge implementation trial (www.clinicaltrials.gov identifier NCT03358342) to assess the impact of ePNa implementation on ICU utilization among patients without an sCAP major criterion on admission.

Setting:

16 Intermountain Health hospital-based EDs in Utah and Idaho.

Implementation:

ePNa was implemented in a stepped-wedge process into 6 geographic clusters which were selected to encompass ED, hospitalist and critical care providers working at more than one hospital. Order of cluster implementation was dictated by a prior commitment to prioritize deployment at “usual care” hospitals from a previous study.⁸ Details of the development and implementation of ePNa have been published previously.^{5,8,9,17–19} Additional details of hospital clusters are presented in appendix table 1.

Timeline:

Data was collected retrospectively for patients presenting between December 2016 and June 2019 (before the onset of the COVID-19 pandemic). For each cluster, patients presenting in the 12 months prior to deployment at that cluster served as baseline data. Starting on the first day of local deployment we enforced a 2-month wash-out period in which no patient data were included in the analysis to allow for provider education and uptake. Following the end of the washout period, patient data were included from each cluster until the end of study in June 2019. Implementation at the first cluster began December 1^st, 2017 and implementation in the final cluster began November 1^st, 2018. Specific dates for each cluster are shown in appendix table 1 and appendix figure 1.

Inclusion Criteria:

We included all patients ≥ 18 years old presenting to study EDs during study dates identified by International Classification of Diseases, 10^th Revision (ICD-10) discharge codes for pneumonia or sepsis with respiratory failure per previously published methodology (appendix table 2).^8,20 Pneumonia was confirmed based on chest imaging obtained in the ED by hierarchy of available imaging (computed tomography of the chest, 2 view x-ray of the chest, 1 view x-ray of the chest). Patient imaging was categorized for presence of pneumonia (likely, possible, or unlikely with patients categorized as “unlikely” excluded from analysis), unilateral versus bilateral infiltrates, and presence of an effusion. For those with chest x-rays, images were categorized for these criteria by the CheXED artificial intelligence platform.²¹ For those with CT scans, the authors reviewed radiologist reports (blinded to all other data). We included only the first encounter for pneumonia within a 12-month period.

Exclusion Criteria:

We excluded patients discharged home from the ED, who died in the ED, who were admitted from the ED to a non-Intermountain hospital, patients with an sCAP major criterion on admission (invasive mechanical ventilation [IMV], non-invasive positive pressure ventilation [NIPPV], or vasopressors) because these are absolute indications for ICU admission in most hospitals, patients admitted directly to inpatient hospice, and patients admitted to intermediate care units (present in 2 of 16 hospitals) because neither ePNa nor relevant guidelines offer logic for considering these locations and there were no standardized criteria for what constituted intermediate care.

Patients transferred to a different Intermountain hospital were attributed to the ED and cluster where they initially presented. Data were gathered through queries of the Intermountain enterprise data warehouse with mortality also identified from state departments of health. Missing data were identified by manual chart review (most commonly oxygen saturation and mental status, missing data was <1% after chart review).

Clinical Decision Support:

ePNa automates the collection and tabulation of data and provides recommendations via the user interface but providers can choose to accept or reject any recommendation (“open loop” decision support). Providers are encouraged to utilize recommendations unless their clinical judgement dictates deviation.

Patients with pneumonia presenting to the ED are assessed and, if meeting criteria for hospital admission, the minor sCAP score is tabulated by ePNa (Table 1). Patients having scores of 0–2 are recommended for general ward, scores greater than 3 are recommended for ICU, and a score equal to 3 is left to provider discretion but the display notes a likelihood of ICU admission requiring critical therapies of 54%.^10,11 To generate explicit recommendations in real time from the electronic health record (EHR), we adapted the minor sCAP criteria as noted in table 1.^17,22

Table 1:

sCAP Minor Criteria

IDSA/ATS Guideline Criteria	Adaptation for electronic calculation
Respiratory rate ≥ 30	First recorded value in ED
PaO2:FiO2 ≤ 250*	Hierarchy of measured PaO2 from arterial blood, then estimated from SpO2 using the Ellis equation, adjusted for mean altitude (1400 meters) of study hospitals using 213.
Multilobar infiltrate	Natural language processing of radiology report using a hierarchy of ED performed CT scan, 2-view chest series or 1-view portable radiograph when more than one study available.
Confusion/Disorientation	Glasgow Coma Scale <15 (standard nursing documentation) or documentation of altered mental status or disorientation
Blood Urea Nitrogen ≥ 20 mg/dl	First recorded value in ED
White Blood cell count <4,000 cells/μl	First recorded value in ED
Platelet count <100,000/μl	First recorded value in ED
Core temperature <36° C	First recorded value in ED
Hypotension requiring aggressive fluid resuscitation†	First recorded systolic blood pressure, <90 mmHg used as cutoff

Measurements:

Our primary outcome was 30-day ICU free days. We define ICU-free days as 30 minus the number of days in the ICU (range, 0–30 days) and assigned 0 ICU-free days for death within 30 days. If the patient was dispositioned to the floor and never admitted to the ICU they were assigned 30 ICU-free days. Because of the intense skew in the primary outcome, we conducted a post hoc reanalysis calculating ICU days/100 inpatient days, defined as ICU length of stay divided by the hospital length of stay. Patients never admitted to the ICU had a value of zero.

Secondary outcomes were 1) odds of being admitted to the floor, subsequently transferred to the ICU within 72 hours, and receiving a critical therapy. A critical therapy was defined as receiving positive pressure ventilation (either IMV or NIPPV), receiving >60% FiO2, placement of an arterial or central-venous catheter, hemodialysis, infusion of a vasopressor or inotrope, or fluid resuscitation totaling 4 liters of crystalloid within 2 hours, consistent with prior published work.¹⁴ The critical therapy had to be administered within 24 hours of ICU admission. 2) odds of being admitted to the ICU and not receiving a critical therapy, 3) odds of being admitted to the ICU, staying less than 24 hours, and not receiving a critical therapy, 4) ICU length of stay (LOS), 5) hospital LOS, 6) Total cost of hospital admission (in $USD) and 7) 30-day all-cause mortality. Patients who died during hospital admission were excluded from all LOS calculations.

By standing rules, ≥60% FiO2 required ICU admission at study hospitals. “Receipt of a critical therapy” identifies ICU stays “requiring” ICU care as opposed to those admitted for operational reasons such as staffing or close observation.

Mortality Review:

The lead author conducted a manual review of patients before and after implementation who were admitted to the floor and died within 72 hours without transfer to the ICU. This was done to explore whether implementation of ePNa influenced provider behavior such that high-risk patients who would have benefitted from ICU-level care were instead admitted to the floor. Each patient’s ED and hospital course were reviewed including labs and imaging. Specific items of interest included the minor sCAP score on admission, cardiac arrest team activations, and documented limitations of care on admission or transitions to hospice.

Statistical analysis.

Mixed effects regression models with cluster as the random effect were used to predict both 30-day ICU-free days and ICU days/100 inpatient days. Variables included in the model to account for differences are Influenza season (November 1-May 31), age (years), sex, eCURB score,¹⁸ PaO2:FiO2 ratio, Health Care Associated Pneumonia Criteria (HCAP; present or absent),²³ Elixhauser comorbidity index, and pleural effusion²⁴ (present or absent). There were insufficient data in each cluster to model a random effect for the secondary outcomes, so we used linear or logistic regression using the same patient characteristics as the primary analysis. Prior to conducting this analysis, we were aware, based on prior publication, that there were insufficient data to utilize an interrupted time series model.⁹

All analyses were done by implementation period (pre vs post). At the time of data collection, no reliable indicator of ePNa use by clinicians was available in the EHR, ruling out an analysis that considered actual use of the tool during a patient encounter. We have previously published that the estimated utilization of ePNa in the post-implementation period was 67%.⁹ Differences in outcomes between pre-and-post-implementation were tested for statistical significance using Fisher’s exact test in categorical variables and T-test for continuous variables. R version 4.1.0 was used for these analyses.¹⁸

Ethical considerations:

The study, data collection, and analysis were in accordance with the requirements of the Institutional Review Board of Intermountain Health who approved the original study (titled: “Pneumonia ED Decision Support Screening Tool Performance and Outcomes”) as IRB #1024929 on January 31^st, 2014 with a waiver of informed consent. All study procedures were in accordance with the Helsinki Declaration of 1975, as most recently amended.

Results

Between December 2016 and June 2019, 3091 patients met initial inclusion criteria. After exclusions the final data set included 2747 patients, 1814 before implementation of ePNa and 933 after implementation (Figure 1). Median age was 71 (IQR 58–81), 48% were female and 95% white. Baseline patient characteristics were notably different in the pre- versus post-implementation groups for median eCURB predicted mortality (4% vs 3%), presence of pleural effusion (4% versus 8%), 30-day mortality (10% vs 7%), confusion (13% vs 8%), leukopenia (4% vs 2%), hypotension (4% vs 2%), and baseline median sCAP score (2 vs 1). Complete details of the pre and post implementation cohorts are presented in Table 2.

Figure 1:

ICD-10: International Classification of Diseases, 10^th revision

Table 2:

Characteristics of included patients

Variable			Pre		Post
	n		1814		933
Age	Median (IQR)		72 (59–81)		70 (56–81)
Female	% (n)		49% (882)		48% (449)
Race	% (n)
	American Indian or Alaska Native		1% (15)		1% (8)
	Asian		1% (15)		0% (3)
	Black or African American		0% (9)		1% (6)
	Native Hawaiian or Pacific Islander		1% (19)		1% (13)
	White		95% (1731)		95% (888)
	Multiple		0% (0)		0% (1)
	Unavailable		1% (25)		2% (14)
Ethnicity
	Not Hispanic, Latino, or Spanish Origin		93% (1695)		95% (883)
	Hispanic, Latino, or Spanish Origin		6% (106)		5% (42)
	Unavailable		1% (13)		1% (8)
eCURB (%)	Median (IQR)		4 (2–9)		3 (1–6)
PaO2:FiO2	Median (IQR)		297 (243–362)		304 (250–362)
HCAP	% (n)		24% (435)		26% (245)
Pleural effusion	% (n)		4% (68)		8% (77)
Flu Season (1-Nov to 1-Jun)	% (n)		68% (1231)		65% (607)
Elixhauser (VW)	% (n)		18 (8–28)		16 (6–26)
30-day Mortality	% (n)		10% (179)		7% (66)
sCAP Components
Variable			Pre	Post
		N	1814	933
RR >= 30 breaths/min		% (n)	10% (178)	10% (93)
P:F ratio <= 250*			28% (514)	25% (234)
Multilobar infiltrates			49% (884)	50% (462)
Confusion			13% (231)	8% (77)
Uremia (BUN >= 20 mg/dL)			52% (941)	50% (465)
Leukopenia (WBC count < 4000 cells/mm3)			4% (74)	2% (22)
Thrombocytopenia (platelet count <100,000 cells/mm3)			5% (83)	4% (33)
Hypothermia (core temperature < 36º C)			7% (130)	6% (56)
Hypotension (Systolic BP <90 mmHg)			4% (68)	2% (23)
sCAP score		Median (IQR)	2.00 (1.00–2.00)	1.00 (1.00–2.00)

Primary outcome:

A mixed effects regression model with cluster as the random effect showed that ePNa implementation was associated with an increase in 30-Day ICU-free days of 1.04 days, (95% CI 0.48–1.59; p<0.001), median 30.00 (IQR 29.00–30.00) before vs 30.00 (IQR 30.00–30.00) after deployment. In the post hoc reanalysis, a mixed effects regression model with cluster as the random effect showed that ePNa implementation was associated a decrease in ICU days/100 inpatient days by 4.21 days (95% CI −6.49 to −1.93; p<0.001), mean 17 days (SD 31) before vs. 13 (SD 28) after deployment. Disposition of ED patients by their sCAP score before and after implementation is shown in Table 3 (unadjusted for severity) demonstrating an overall shift, particularly among patients with sCAP score <3, away from the ICU in accordance with ePNa recommendations.

Table 3:

Disposition by sCAP score

Variable	Pre	Post
N	1814	933
sCAP score – Actual Disposition
sCAP<3 – Floor	65% (1177)	74% (693)
sCAP<3 – ICU	13% (230)	7% (65)
sCAP=3 – Floor	11% (191)	10% (94)
sCAP=3 – ICU	4% (78)	4% (38)
sCAP>3 – Floor	3% (63)	3% (24)
sCAP>3 – ICU	4% (75)	2% (19)

Secondary outcomes:

Details of secondary outcomes are in Table 4. The distribution of critical therapies received by patients admitted to the ICU is shown in Table 5. The odds of being admitted to the floor, transferred to the ICU within 72 hours, and receiving a critical therapy decreased by 57% post implementation (5.1% vs 2.3%, OR 0.43, 95% CI 0.26–0.68; p<.001). The odds of being admitted to the ICU, staying less than 24 hours, and not receiving a critical therapy decreased by 56% post-implementation (2.6% vs 1.2%, OR 0.44, 95% CI 0.21–0.82; p=0.015). ICU LOS did not significantly change between pre- and post-implementation (2.4 vs 2.2 days, estimate 0.01, 95% CI −0.94 to 0.95; p=0.99). Hospital LOS decreased by 1.08 days post-implementation (3.1 vs 2.4, 95% CI −1.47 to −0.69; p<.001). Median total cost per patient pre-implementation was $8,033 vs $6181 post-implementation (estimate -$3,222, 95% CI −4,708 to −1,736; p<.001). Mortality within 72 hours did not change significantly post-implementation (1.8% vs 1.7%, OR 1.08, 95% CI 0.56–2.0; p=0.82), but the odds of 30-day mortality decreased by 29% (9.9% vs 7.1% OR 0.71, 95% CI 0.52–0.96; p=0.03).

Table 4:

Clinical Outcomes

Variable		Pre	Post		P-value
N		1814	933
Primary Outcome				Adjusted Difference (95% CI)
30-day ICU-free days	Mean (IQR)	30 (29.00–30.00)	30.00 (30.00–30.00)	1.04 (0.48, 1.59)	<0.001
Exploratory Outcome
ICU days/100 inpatient days	Mean (SD)	17 (31)	13 (28)	−4.21 (−6.49, −1.93)	<.001
Secondary Outcomes				Estimate (95% CI)
Admitted to the floor, transferred to the ICU within 72 hours, and received critical therapy	%(N)	5.1% (92)	2.3% (21)	0.43 (0.26, 0.68)*	<.001
Admitted to the ICU and did not receive critical therapy	%(N)	5.4% (98)	4.1% (38)	0.73 (0.49, 1.07)*	0.11
Admitted to the ICU and stayed less than 24 hours and did not receive critical therapy	%(N)	2.6% (47)	1.2% (11)	0.44 (0.21, 0.82)*	0.015
ICU LOS on index admission	Median (IQR)	2.4 (1.1–4.9)	2.2 (1.4–4.9)	0.01 (−0.94, 0.95)	0.99
Hospital LOS on index admission	Median (IQR)	3.1 (2.0–5.0)	2.4 (1.7–4.0)	−1.08 (−1.47, −0.69)	<.001
Total costs of hospital admission (USD)	Median (IQR)	8033 (5339–14273)	6181 (4285–10910)	−3222 (−4708, −1736)	<.001
Death within 72 hours	%(N)	1.8% (33)	1.7% (16)	1.08 (0.56, 2.01)*	0.82
30-day all-cause mortality	%(N)	9.9% (179)	7.1% (66)	0.71 (0.52, 0.96)*	0.03

Table 5.

Critical therapies administered to patients admitted to the ICU

Critical Therapy	Pre	Post
N	1814	933
Any critical therapy	22.3% (404)	12.5% (117)
Mechanical ventilation	3.9% (70)	2.9% (27)
NIPPV	9.8% (178)	5.4% (50)
FiO2 >= 60%	13.7% (248)	8.6% (80)
Arterial line placement	5.5% (105)	1.9% (18)
Central line placement	11.1% (201)	4.6% (43)
Emergent dialysis	0.6% (10)	0.6% (6)
Vasopressor/inotrope administration	7.2% (131)	3.3% (31)
4L crystalloid fluid in a 2-hour period	0% (0)	0% (0)

We attempted to explore how secular change in critical therapy use may have impacted. our secondary outcomes by analyzing the odds of being admitted to the ICU and not receiving a critical therapy. This did not change significantly between pre- and post-implementation periods (5.4% vs 4.1%, OR 0.73, 95% CI 0.49–1.07; p=0.11).

Mortality review.

40 patients were admitted to a hospital ward and died within 72 hours without ICU transfer; 26 (1%) pre-implementation and 14 (1%) post-implementation. Median Age (IQR) was 79 (71–90) pre- versus 83 (75–90) post-implementation. The median Elixhauser comorbidity index (IQR) was 26 (16–37) pre- versus 32 (14–45) post-implementation. Pre- versus post-implementation, the percentage of patients with an sCAP score of 4 (indicating likely benefit from ICU admission) was 15% (4,) vs 29% (4); with an sCAP of 3 (indicating possible benefit from ICU admission) 42% (11) vs 21% (3); with an sCAP score of 2 (unlikely to require ICU admission) 27% (7) vs 21% (3); with an sCAP score of 1, 15% (4) vs 29% (4). Twenty seven percent of patients (7) pre and 19% (3) of patients post-implementation were residents of a long-term care facility. Nineteen percent (5) of patients pre and 7% (1) patient post implementation had cardiac arrest team activations immediately prior to their deaths. Pre implementation, 81% (21) versus 93% (13) of patients post implementation had documented limitations of care or were transitioned to hospice during their hospital stay.

Discussion

Implementation of electronic CDS for pneumonia using minor sCAP criteria in a pragmatic, stepped-wedge clinical trial was associated with increased 30-day ICU-free days and fewer delayed ICU transfers, shorter hospital LOS, and reduced cost. While prior work had validated the performance of minor sCAP criteria to predict ICU admission,^10–12,14 low strength of guideline recommendations reflects these studies’ limitations and may have discouraged broader use. This is the first large-scale, prospective use of sCAP minor criteria to guide ICU admission. We believe our results should increase the weight of guideline recommendations to use objective severity assessments that guide ICU admission decisions.

While a prior manuscript from the same pragmatic trial reported improvements in mortality and appropriate outpatient disposition, it also noted a possible change in ICU utilization that could not be fully explored. To conduct this analysis property, we restricted the population to those in whom, prior to ePNa, judgement alone may have been used to determine inpatient disposition (i.e. excluding sCAP major criteria, which are absolute indications for ICU admission in most hospitals and included patients on the floor at risk for transfer to the ICU). Here, we focused on ICU related processes of care (ICU-free days) and use of critical therapies as a surrogate for “ICU need.”

We hypothesize two main reasons sCAP implementation may have influenced outcomes. First, presenting sCAP scores in CDS may increase provider confidence in identifying low-risk patients. These patients then avoid the ICU altogether and have shorter hospital stays. Second, the sCAP score may prompt reconsideration of high impact clinical data. In high-risk (sCAP >=3) patients, this may lead to recognition and encourage action and communication about high-risk status. High-risk patients who go to the ICU early may receive more aggressive resuscitation, be stabilized earlier, and recover more quickly. This may explain the effect observed on ICU utilization in one prior study that employed sCAP scores to identify high risk patients in the ED but did not use sCAP scores to recommend disposition.²⁵

Importantly, fewer high-risk patients were admitted to the floor and delayed ICU transfers were reduced. This is supported by our manual review of patients with early mortality which showed that, both before and after implementation, these patients expressed significant limitations for their care and appropriately avoided the ICU.

One notable source of potential bias in our analysis is that clinical or administrative detection of pneumonia may have changed pre versus post implementation. This would impact our captured population even though our case identification method was the same across all study periods. We are unaware of specific administrative efforts which would have impact case identification through ICD-10 codes. However, we hypothesize that ePNa improves outcomes, in part, by aiding real-time diagnosis and this could impact case identification. If this is true, ePNa still has important impacts on care by reliably identifying low-risk patients and improving therapeutic guidance (established in prior work)^9,26 or possibly by improving diagnostic accuracy (under further investigation).

Limitations:

First, the post-deployment group is generally younger and less ill that the pre-deployment group. Despite consistent results across analyses, this may leave residual confounding. Second, our data come from a single large, integrated health system which, while geographically diverse, shares unified leadership. How other shared system-wide initiatives may have impacted ED provider behavior is unknown. In addition, the relative homogeneity of patient populations and processes of care limit the generalizability of our findings. Third, the impact of providing sCAP scores might differ in health systems with different baseline utilization of ICU care for pneumonia patients. Finally, secular changes in critical care or pneumonia management may impact interpretation of our results.

Conclusions:

A large scale implementation of CDS which provided sCAP minor criteria to support bedside decision-making for ICU admission in various settings (urban and rural, large and small) was associated with reduced ICU utilization without evidence of harm. These results suggest that CDS is an effective way to implement objective severity assessment that guides ICU admission decisions. Our results support guideline statements favoring incorporation of sCAP in pneumonia care.

Key Points:

Question:

Does implementation of clinical decision support (CDS) utilizing the Infectious Disease Society of America/American Thoracic Society (IDSA/ATS) Severe Community Pneumonia (sCAP) minor criteria improve intensive care unit (ICU) utilization among patients presenting to the Emergency Department (ED) with community acquired pneumonia (CAP) who do not require mechanical ventilation or vasopressors on admission?

Findings:

In a cluster-controlled, stepped-wedge implementation of automated, open-loop, comprehensive CDS using the IDSA/ATS sCAP minor criteria to aid clinician judgement, 30-day ICU-free days increased 1.04 days after deployment (p <0.001).

Meaning:

ICU utilization can be safely reduced among patients with CAP by using the IDSA/ATS sCAP minor criteria as part of comprehensive CDS.