Chest X-ray Overuse in Cardiothoracic Intensive Care Unit per American College of Radiology Criteria

Abstract

Background

A routine chest radiograph (CXR) is frequently incorporated into postoperative clinical pathways. Whereas prior studies evaluated CXR use for specific indications, overall CXR "stewardship" within cardiothoracic surgery remains undescribed. We characterized CXR use within a cardiothoracic intensive care unit (ICU) and quantified X-ray–driven changes in patient management.

Methods

All CXRs performed during 3 months in the cardiothoracic ICU of an academic 900-bed tertiary care hospital were retrospectively reviewed. Attending radiologist CXR reports and clinical history from ordering providers were manually reviewed to classify CXRs as indicated or not indicated by American College of Radiology (ACR) Appropriateness Criteria. Multiple resident-level providers further reviewed CXRs to determine whether the image demonstrated any actionable or unexpected findings.

Results

Between October 1, 2021, and January 3, 2022, 1273 CXRs were performed on adult cardiac and general thoracic surgery patients; the most common indication was routine daily (55.9%), followed by support device monitoring (12.4%) and new ICU transfer (10.4%). Only 34.3% of performed CXRs met ACR Appropriateness Criteria; 65.7% did not (P = .002). Only 3.8% of not indicated CXRs had actionable or significant findings vs 13.0% of indicated CXRs (P = .02). Of the 32 not indicated CXRs with actionable findings, 37.5% recommended endotracheal tube repositioning and 28.1% recommended other line or tube repositioning.

Conclusions

CXRs are overused in the cardiothoracic ICU according to ACR Appropriateness Criteria. CXRs performed outside these guidelines have significantly fewer actionable or meaningful findings. Optimizing CXR use can improve resource utilization and efficiency in care of cardiothoracic surgical patients.

In Short.

• ▪Chest radiographs are overused in cardiothoracic intensive care units compared with American College of Radiology Appropriateness Criteria.
• ▪Chest radiographs that do not align with American College of Radiology indications have significantly fewer actionable or unexpected findings, supporting more judicious X-ray use in the intensive care unit.

Chest radiographs (CXRs) are the most frequently ordered imaging study in the intensive care unit (ICU).¹ The American College of Radiology (ACR) released updated Appropriateness Criteria in 2021 to encourage more responsible use of CXRs in the ICU.² Whereas many well-regarded reports of ICU X-ray use have been published from nonsurgical ICUs or even noncardiac surgical ICUs, there remains a paucity of data on CXR use specific to the cardiothoracic ICU (CTICU), in which invasion of the thoracic and mediastinal cavities may make providers think increased routine surveillance of these structures is necessary.³ We sought to characterize current CXR stewardship and ACR guideline compliance in a CTICU and to quantify X-ray–driven changes in patient management.

Patients and Methods

All CXRs performed within the CTICU of a 900-bed academic tertiary care center within a 3-month period were retrospectively extracted from the electronic medical record using an institutional review board–approved protocol. The ICU included, on average, 58% post–cardiac surgery patients and 42% post–thoracic surgery patients. All data were deidentified, and a waiver of informed consent was granted (Northwestern University IRB: STU00214410). Imaging procedures were selected from the institutional data warehouse by searching procedure orders related to “chest X-ray” and other closely related Boolean text searches while the patient’s disposition was recorded as the CTICU. The procedures were tied to radiology reports by a procedure order key that exactly matched the procedure order key of the original order. One CXR order did not include an accompanying radiology report and was not included in the data set. To fully anonymize the data, each patient was assigned a randomized unique identifier; each report had all imaging, diagnosis, note, and death dates shifted within a 30-day window, and patients aged >90 years were coded as 90 years old. Care was used to eliminate studies and reports that did not match these criteria and those with identifiable patient data.

Radiographs were then grouped by indications as determined through a manual review of fellowship-trained attending chest radiologist reports and coinciding clinical history provided by ordering providers. ACR Appropriateness Criteria were used to classify radiographs as indicated or not indicated on the basis of this manual review.² Indicated radiographs were performed for clinical changes, line placement or adjustment, postoperative ICU transfer, invasive circulatory support device monitoring, or surveillance of known clinical conditions. Common findings, including pleural effusions, consolidations, opacities, and pneumothorax, among others, were manually searched within the radiograph imaging report.

Each radiograph report was further reviewed by multiple resident-level providers to determine whether the radiograph demonstrated any actionable or unexpected findings; any actionable findings were recorded during review. Actionable findings were those that were unexpected or that prompted unanticipated, tangible changes in patient management. These included findings such as new pneumothorax, new pleural effusion, increasing pulmonary edema, and concern for pneumonia. Findings that would not prompt management changes, such mild air space opacities or resolving pulmonary edema, were not classified as actionable.

Results

Between October 1, 2021, and January 3, 2022, 1273 CXRs were performed on 186 adult cardiac and general thoracic surgery patients in a CTICU. Patients were, on average, 58.8 ± 14.1 years old, 67.2% (125/186) male, and 32.8% (61/186) female and had an average body mass index of 28.2 ± 6.5 kg/m² (Table 1). Patients received an average of 1.25 X-rays per day while in the ICU. The most common indication for chest radiography overall was routine daily (55.9% [711/1273]), followed by support device monitoring (12.4% [158/1273]) and new ICU transfer (10.4% [133/1273]; Table 2).

Table 1.

Baseline Demographics of Patients (N = 186)

Demographic Characteristic	No. (%) or Mean ± SD
Age, y	58.8 ± 14.1
Male	125 (67.2)
Female	61 (32.8)
Body mass index, kg/m2	28.2 ± 6.5
White	109 (58.6)
Black	32 (17.2)
Asian	5 (2.7)
Other	40 (21.5)
Length of stay, d	4.9 ± 5.0
X-rays per patient per day	1.25

Table 2.

Alignment of Chest X-ray Utilization in the Cardiothoracic ICU With ACR Appropriateness Criteria

Indication	No. of Chest X-rays	Percentage of All Chest X-rays	No. of Chest X-rays With Actionable Findings	Percentage of X-rays With Actionable Findings
Indicated per ACR Appropriateness Criteria
Postoperative ICU transfer	134	10.5	27	20.1
Support device monitoring	158	12.4	4	2.5
Line placement/adjustment	92	7.2	16	17.4
Clinical change	46	3.6	8	17.4
Clinical condition surveillance	7	0.5	2	28.6
Subtotal	437	34.3	57	13.0
Not indicated per ACR Appropriateness Criteria
Routine daily	711	55.9	28	3.9
Chest tube removal	86	6.8	3	3.5
Post-bronchoscopy	24	1.9	1	4.2
Establish baseline	6	0.5	0	0.0
Line removal	3	0.2	0	0.0
Other	6	0.5	0	0.0
Subtotal	836	65.7	32	3.8
Total	1273	100.0	89	7.0

ACR, American College of Radiology; ICU, intensive care unit.

Only 34.3% (437/1273) of performed radiographs were indicated per ACR Appropriateness Criteria; 65.7% (836/1273) were not indicated (P = .002). The most common X-ray indications that did not align with the ACR Appropriateness Criteria included routine daily, chest tube removal, post-bronchoscopy, establish baseline, and line removal.

Only 3.8% (32/836) of not indicated CXRs had actionable or significant findings vs 13.0% (57/437) of indicated CXRs (P = .02; Table 3). Of the 32 not indicated CXRs with actionable findings, 37.5% (12/32) recommended endotracheal tube repositioning and 28.1% (9/32) recommended line (pulmonary artery catheter or nasogastric tube) repositioning. There was no suggestion of any major patient events that occurred secondary to endotracheal tube or other line malpositioning. Other actionable findings in this group involved imaging changes and cardiopulmonary findings (pneumonia, worsening edema, small pneumothoraces, and a pulmonary nodule flagged for follow-up). One patient had a large pneumothorax; this CXR report mentioned that a chest tube had already been placed before CXR read dictation.

Table 3.

Characterization of Actionable Findings Observed in X-rays

Findings	Indicated X-rays (%)	Not Indicated X-rays (%)	P Value
Cardiopulmonary findings	17 (29.8)	11 (34.3)	.830
Endotracheal tube malposition	24 (42.1)	12 (37.5)	.739
Other line/tube malposition	16 (28.1)	9 (28.1)	.780
Total	57 (100)	32 (100)

Comment

Inappropriate CXR use in the ICU carries implications for both patient safety and cost of care. Our report demonstrated a rate of 1.25 CXRs per patient per ICU day, and 66% were not appropriate per ACR. For reference, Vanderbilt University Medical Center, another academic tertiary care center, reports an average of 1.17 CXRs per patient per day in the ICU.⁴

Further analysis revealed that 3.8% of CXRs not meeting ACR guidelines had actionable findings. This aligns with previously published rates of intervention from routine CXR use; rates of intervention are documented in 3.9% of combined medical-surgical ICUs and 4.5% of post–thoracic surgery patients.^5,6 Despite reported diagnostic efficacy of CXR, the therapeutic value of routine CXR seems low with a number needed to treat of 27 patients based on a 3.8% rate of intervention. This may stem from abnormalities on CXR being expected findings, rather than new discoveries, in the postoperative setting.

In our work, most CXRs that did not meet ACR guidelines were performed as routine daily images. Numerous studies, including 1 meta-analysis of more than 9600 patients and 36,000 X-rays, have highlighted no changes in ICU mortality, hospital mortality, ICU length of stay, hospital length of stay, or duration of mechanical ventilation after employing more restrictive approaches to CXR use.⁷ Furthermore, the RARE study presented a cluster-randomized crossover trial across 21 ICUs in 18 hospitals and demonstrated a 32% reduction in CXR use when imaging was performed on-demand rather than daily. This report highlighted no changes in duration of mechanical ventilation, ICU length of stay, or ICU mortality after elimination of routine daily CXRs.⁸

Chest tube removal was another area of notable CXR use in our series. There is general disagreement on the necessity of CXRs after chest tube removal; however, several reports have recommended selective X-ray use only for symptomatic patients after chest tube removal.² A clinical trial of >1000 postoperative cardiac surgery patients demonstrated the safety of omission of routine postoperative chest tube removal X-rays, with only 1.5% of patients requiring therapeutic intervention after chest tube removal.⁹

Of the not indicated CXRs in our series that revealed actionable findings, most were related to malpositioning of an endotracheal tube, nasogastric tube, or Swan-Ganz catheter. Various alternatives have been proposed in the literature and successfully tested to safely reduce the use of CXRs for such monitoring. Such interventions include documentation of endotracheal or nasogastric tube position in the electronic medical record with reduction of scheduled CXRs to once or twice weekly or even use of chest ultrasound.¹⁰

The costs of routine chest radiography are not insignificant. Based on an estimate of $671 charged by the studied hospital per X-ray, our work shows a direct cost of $560,956 for not indicated CXRs performed during the studied 3-month period. Beyond direct costs, routine morning CXRs involve patient discomfort from being woken up at odd hours, dedicated X-ray technologists who are then unavailable for urgent imaging when needed, and hours of attending radiologist time spent reading unnecessary imaging studies. Anecdotal experience from our institution highlights discontent from radiologists who spend about up to 10 hours each day reading routine CXRs and dictating reports that few providers read.

Our study is not without limitations. First, our patient population included a mix of adult cardiac and lung transplant patients. Because an imaging database was used to source data, patients could not be stratified on the basis of procedure while maintaining deidentifiable data. Patients and radiography findings were also unable to be stratified by postoperative day. These may be useful analyses for future reports. We also acknowledge that the guiding body of evidence for the ACR Appropriateness Criteria references all ICUs, possibly limiting its generalizability to a purely cardiothoracic surgical population. In addition, ACR criteria do not draw firm conclusions on certain indications. Within our study, we classified indicated CXRs as those with a “usually appropriate” or greater level of recommendation. Criteria for which the panel “did not agree” on recommending radiography were characterized as not indicated. Finally, our report is unable to analyze clinical modifications made on the basis of imaging that were not commented on by radiologists, such as diuresis management and drain decisions. However, we hold that numerous other clinical parameters can be used to guide these decisions that incorporate judicious rather than daily CXR use.

Routine CXR practices after cardiothoracic surgical procedures currently diverge from ACR Appropriateness Criteria, contributing to unnecessary health care costs and provider burden. Whereas there is undoubted utility for chest radiography in the ICU, we recommend the development of modern, thoughtful postoperative protocols and perhaps incorporation of artificial intelligence algorithms to drive responsible use of CXRs within the ICU.