Utility of routine chest radiographs in a medical–surgical intensive care unit: a quality assurance survey

It is not clear whether the performance of routine CXRs alters management in patients admitted to the ICU. Studies evaluating the use of routine CXRs have mainly been in the form of prospective observational studies, with contradictory results. Fong et al found that 48% of CXRs performed in a surgical ICU were routine studies, and only 17% had an impact on clinical management [1]. In a pediatric ICU, Price et al found that 37% of CXRs could be avoided by establishing specific indications, thereby resulting in significant cost savings [2]. In a prospective study, Hall et al compared bedside clinical diagnosis with the diagnosis made from the routine CXR [3]. Of 538 routine CXRs, 8% presented new 'major' findings; however, 58% of these were anticipated by the clinical examination, and only 3.4% of all routine CXRs presented findings not clinically anticipated. Conversely, several studies have concluded that routine CXRs are beneficial to patient care. Brainsky et al observed that 20% of routine CXRs performed in a medical ICU had 'major important' findings, and 8% prompted a change in management [4]. The majority of changes related to diuretic use, antibiotic coverage, initiation of a diagnostic test, or decisions regarding ventilator weaning. Similarly, Bekemeyer et al found that 27% of both routine and non-routine CXRs revealed clinically unsuspected abnormalities, but that non-routine films were more likely to change investigative or therapeutic management [5].

Although there may be benefits related to the performance of routine CXRs, there are also significant associated economic and clinical costs. Adverse consequences associated with patient repositioning for the performance of CXRs can include patient discomfort, hypotension, oxyhemoglobin desaturation, and displaced endotracheal tubes (ETTs), naso-gastric tubes (NGTs), or vascular catheters.

The financial costs, potential adverse clinical consequences, and the uncertainty surrounding the value of routine CXRs in previously published studies prompted us to prospectively evaluate their utility in our medical–surgical ICU as part of a quality assurance survey. The goals of this study were to determine the percentage of routine and non-routine radiographs that change management in our medical–surgical ICU, and to determine the specific resultant management changes.

All medical and surgical patients admitted to the ICU at Mount Sinai Hospital, a university-affiliated hospital, over a 6-month period were enrolled and prospectively evaluated. Because this was an observational study, no attempt was made to alter the performance of routine CXRs. Informed consent was not obtained from patients because this study was part of an ICU quality assurance program.

For each CXR performed (routine and non-routine), the clinical fellow completed a data sheet documenting the patient's ICU admission diagnosis, the indication for the CXR, and any resulting changes in management.

The ICU team, consisting of the attending physician, a clinical fellow, and a group of housestaff, interpreted the daily CXRs.

CXRs were defined as routine if they were performed first thing in the morning or at ICU admission. In our ICU, the on-call resident decides which patients should have routine CXRs. CXRs performed for a specific indication (e.g. desaturation, fever) were defined as non-routine.

Changes in patient management were categorized as ETT placement or change in position, central line placement or change in position, thoracostomy tube placement or change in position, ventilator setting change, antibiotics started, congestive heart failure (CHF) treated, lung or pleural biopsy, thoracentesis, or other.

Given that medical and surgical patients often have different complications and varying lengths of stay, the data for each were analyzed separately. Surgical patients were divided into two groups retrospectively by ICU length of stay ≥ 48 hours or <48 hours. Medical patients were defined as non-surgical patients admitted from a medical ward, the emergency department, or another hospital.

The hospital's computerized radiographic database (eFilm workstation 1.5.2, © 2000; eFilm Medical Inc. Toronto, Ont., Canada) was reviewed to determine whether there were additional radiographs not documented on a daily datasheet. Indications for these non-routine CXRs were not determined retrospectively. All data were entered into a computerized database (Excel 97; Microsoft Corp., Redmond, Washington, USA).

Over a 6-month period, 850 CXRs were performed in 198 patients: 645 CXRs in 97 medical patients and 205 CXRs in 101 surgical patients. Major admitting diagnoses for the medical and surgical patients are presented in Tables 1 and 2, respectively.

Table 3 presents the various indications for the CXRs in the medical and surgical patients. The two most common indications for non-routine CXRs were following a procedure to verify the position of a medical device and exclude complications, and for evaluation of a suspected new medical condition. Table 4 presents the management changes resulting from the CXRs in each of the patient groups.

In this quality assurance survey, we observed in our medical patients that 22% of all routine CXRs, and 40% of non-routine CXRs, led to a change in management. Similarly, in surgical patients with ICU stays longer than 48 hours, 26% of routine and 40% of non-routine films changed management. In surgical patients with ICU stays shorter than 48 hours, a smaller percentage of routine CXRs (17%) resulted in a change in management. In both the medical and surgical patients, the two most common changes resulting from the CXR were adjustment of a medical device, and the diagnosis and treatment of CHF. Furthermore, 46% of the medical patients and 26% of the surgical patients with an ICU stay ≥ 48 hours had one or more CXRs performed, in addition to the routine CXR, on a given day.

Our study probably overestimates the utility of routine CXRs owing to the introduction of selection bias, since the houses-taff decide which patients have morning CXRs. In contrast, the percentage of non-routine CXRs that alter therapy may have been underestimated, as 63-68% of these radiographs had no data sheets completed.

Our results are very similar to those of Fong et al, who observed that only 17% of routine CXRs prompted a change in clinical management in a surgical ICU [1]. Other studies have yielded varied results, most probably due to the heterogeneous patient population in the ICU setting, as well as large differences in study design and terminology [3,4,6,7]. For instance, Silverstein et al found that 27% of routine CXRs performed in a surgical ICU presented worse or new findings; however, only 1.4% of these required immediate action [6].

Our study evaluated the impact of routine CXRs without having recorded the information yielded by the bedside physical examination. Thus, given that no clinical correlation was made, the impact of CXRs on clinical management was most likely overestimated. This is supported by Hall et al, who reported that the incorporation of information from the clinical examination reduces the utility of routine CXRs, with only 3.4% leading to a change in management. The majority of the changes (78%) were related to repositioning of an ETT or a NGT [3]. Similarly, another prospective study reported that general physical examination had a sensitivity greater than 90% in predicting clinical change, which led to a 52% reduction in the number of CXRs performed [7].

Numerous studies have concluded that only selected patients should have routine CXRs performed [1,2,6,7,8,9,10]. Several investigators have evaluated the need for CXRs to check placement of a medical device. Palesty et al concluded that CXRs are not necessary following the placement of a central line over a guide wire, as they observed no complications in 380 such changes [10]. Gray et al found that clinicians were fairly accurate in determining the placement of subclavian or internal jugular (IJ) vein pulmonary artery (PA) catheter introducer sheaths, but the clinicians were not accurate for clinical determination of ETT or PA catheter position [9]. In contrast, Gladwin et al found that the sensitivity of a clinical decision protocol for detecting complications and malpositions of IJ catheter insertion was only 44%. They concluded that routine CXRs are necessary following IJ catheter insertion [11]. The major difference in these opposing studies is that Gray et al evaluated mostly IJ canulations with a PA catheter introducer sheath, whereas Gladwin et al inserted longer central venous catheters, which have a higher likelihood of being placed in the right atrium.

Daily CXRs are often performed in ICUs to assess the placement of medical devices. However, there are currently several ways to clinically judge the position of these devices. Once it has been established that the devices are in the correct position, clinical evaluation including ETT position at the lips could potentially eliminate a large number of CXRs, resulting in significant cost savings.

The authors conclude that although routine CXRs prove to have some value in the management of critically ill patients, they may not be warranted for all patients, specifically surgical patients admitted for post-operative monitoring. Moreover, the use of clinical decision protocols may reduce the number of CXRs performed following placement of a medical device.