Computer order entry template.
Comparison of coronary care unit (CCU) and medical intensive care unit (MICU) testing trends. P value corresponds to comparison of CCU and MICU trends.
Wang TJ, Mort EA, Nordberg P, Chang Y, Cadigan ME, Mylott L, Ananian LV, Thompson BT, Fessler M, Warren W, Wheeler A, Jordan M, Fifer MA. A Utilization Management Intervention to Reduce Unnecessary Testing in the Coronary Care Unit. Arch Intern Med. 2002;162(16):1885–1890. doi:10.1001/archinte.162.16.1885
There have been no studies of interventions to reduce test utilization in the coronary care unit.
To determine whether a 3-part intervention in a coronary care unit could decrease utilization without affecting clinical outcomes.
Practice guidelines for routine laboratory and chest radiographic testing were developed by a multidisciplinary team, using evidence-based recommendations when possible and expert opinion otherwise. These guidelines were incorporated into the computer admission orders for the coronary care unit at a large teaching hospital, and educational efforts were targeted at the house staff and nurses. Utilization during the 3-month intervention period was compared with utilization during the same 3 months in the prior year. The hospital's medical intensive care unit, which did not receive the specific intervention, provided control data.
During the intervention period, there were significant reductions in utilization of all chemistry tests (from 7% to 40%). Reductions in ordering of complete blood counts, arterial blood gas tests, and chest radiographs were not statistically significant. After controlling for trends in the control intensive care unit, however, the reductions in arterial blood gas tests (P = .04) and chest radiographs (P<.001) became significant. The reductions in potassium, glucose, calcium, magnesium, and phosphorus testing, but not other chemistries, remained significant. The estimated reduction in expenditures for "routine" blood tests and chest radiographs was 17% (P<.001). There were no significant changes in length of stay, readmission to intensive care, hospital mortality, or ventilator days.
The utilization management intervention was associated with significant reductions in test ordering without a measurable change in clinical outcomes.
THE ROUTINE USE of laboratory and radiologic testing has come under increased scrutiny in recent years. One reason for this relates to the finding that substantial variation in testing practice exists, which suggests that a large fraction of tests may be unnecessary.1 Numerous studies have examined interventions to reduce test ordering.1,2 Few of these have been in the intensive care unit (ICU) setting, despite the fact that ICU expenditures occupy a disproportionate fraction of overall health care costs.3 Furthermore, ICU patients may be more likely to receive a large panel of "routine" diagnostic tests under the presumption that critically ill patients require increased laboratory surveillance. To our knowledge, there have been no studies of interventions to change practice in a coronary care unit (CCU). This omission is notable, given that critically ill cardiac patients are relatively homogeneous compared with patients in a general medical ICU. This population would naturally lend itself to the use of evidence-based or guideline-based care.
An example of a common, but unsubstantiated practice in the CCU is the routine determination of "extended" chemistries, such as calcium, magnesium, and phosphorus. A standard justification for checking these parameters in cardiac patients is prophylaxis of arrhythmias. Calcium and phosphorus are not important risk factors for arrhythmias, however. Whether hypomagnesemia predisposes to arrhythmias is controversial, and studies are often confounded by the associated presence of hypokalemia.4 Further, magnesium infusion as a routine therapy for acute myocardial infarction is not beneficial.5 Arterial blood gas (ABG) measurements and daily portable chest radiographs (CRs) are additional examples of tests that are often ordered routinely rather than in response to a change in a patient's clinical status.6- 12
We undertook this study to determine whether a 3-part utilization management intervention, consisting of guideline development, computerized order template design, and educational efforts, could decrease routine test utilization without affecting clinical outcomes in a large academic hospital CCU.
The Massachusetts General Hospital (MGH), Boston, is an 855-bed teaching hospital that serves as both a tertiary referral center and a community hospital for the local population. The 15-bed CCU accounted for 618 admissions in 1999, most of which were related to coronary artery disease. The medical ICU (MICU) at MGH, which was used as a control unit in our analysis, has 18 beds and had 490 admissions during the same period.
Testing guidelines were developed by a multidisciplinary team consisting of 2 cardiologists, an internist, a pulmonary/critical care specialist, and 3 CCU nurses. A MEDLINE literature search was conducted to identify studies and guidelines pertaining to laboratory testing practices, the use of ABG analysis, and indications for daily portable CRs. The following guideline addressed routine testing:
Laboratories. Order daily tests for electrolytes (sodium, potassium, chloride, and carbon dioxide), serum urea nitrogen, creatinine, and complete blood cell count (CBC). A routine measurement of calcium, magnesium, and phosphorus is not indicated.
Portable CR. Order daily CRs in patients (1) receiving ventilator support, (2) with a pulmonary artery line, or (3) with a chest tube. For stable patients with a tracheostomy, a daily CR is not required.
Order ABG measurements every 12 hours in ventilated patients. No need for a routine ABG measurement in patients not receiving ventilator support. No need for a routine ABG measurement after every ventilator change. Consider additional ABG measurements if there is a (1) significant change in minute ventilation, (2) significant fall in oxygen saturation (oxygen saturation as measured by pulse oximetry <88%), or (3) significant change in clinical condition, such as tachypnea (respiratory rate >30/min), labored breathing, or hemodynamic instability.
House staff were allowed to order additional tests as clinically indicated. The intervention was approved by the Clinical Care Management Unit at MGH.
While no data exist to support recommendations for routine laboratory testing in the ICU, it was the consensus of the guideline development group that daily CBCs, and electrolytes, serum urea nitrogen, and creatinine measurements were reasonable tests for critically ill patients. On the other hand, routine determinations of "extended" chemistries, such as calcium, magnesium, and phosphorus, were not recommended. Our ABG guidelines were similar to those proposed by Roberts et al6 and emphasized that blood gases did not have to be assessed with every change in ventilator settings.
The practice of ordering daily portable CRs has been extensively studied.7- 12 Unfortunately, no consensus exists, although the practice appears to have a higher diagnostic yield in medical than in surgical patients.7,9 No study has specifically looked at the utility of radiographs in the CCU. We recommended ordering a daily CR in patients receiving ventilator support or with an indwelling pulmonary artery catheter.
At the beginning of the intervention period, 2 of the investigators (the CCU director and a cardiology fellow) held sessions with the house staff team and nursing staff to explain the guidelines. Sessions with the house staff were repeated every month when the house staff teams changed. The guidelines were also posted in the unit. Admission templates on the computer order entry system were revised to reflect the recommendations for routine testing; a sample page is reproduced in Figure 1. House staff were encouraged, but not required, to use these templates; it was possible to use the computerized order entry system without the predesigned order templates. All orders in the CCU are written by house staff.
Using historical data on ordering CBCs and sodium, potassium, and magnesium measurements, we estimated the sample size needed to detect a 20% reduction in test ordering with 80% power and a type I error of .05. The required sample size ranged from 57 patients (for magnesium) to 163 patients (for potassium). A 3-month intervention period from April through June 1999 was therefore chosen.
We analyzed the effect of the intervention on utilization of diagnostic tests in the CCU. The primary outcome measure was the number of tests ordered per patient-day in the CCU. We focused on tests that were specified in the routine testing guidelines: CBCs, chemistry and ABG measurements, and portable CRs. A secondary measure was aggregate utilization, which was estimated by multiplying the number of each test ordered by its corresponding unit cost and summing over the panel of tests. Unit costs were based on an allocated assignment of the total cost of performing each test and not based on hospital charges. Costs were held constant at 1998 levels to facilitate comparisons between years.
We compared utilization during the intervention period (April through June 1999) with utilization in the CCU prior to the intervention. We chose the same months in the preceding year (April through June 1998) to minimize the effect of seasonal variation. Noncardiac patients may be admitted to the CCU when other ICUs are full; given our intended focus on cardiac patients, we performed analyses on both the entire sample and after excluding patients with noncardiac diagnoses. We were also concerned that hospital-wide changes between the 2 years might affect this comparison, so we analyzed utilization in the MICU during the same 2 periods. During the study period, the hospital was engaged in a variety of utilization management efforts that focused on reducing unnecessary laboratory tests.
To assess whether the intervention had adverse effects on patient outcomes, we looked at mortality, length of stay, readmission to the ICU, readmission to the hospital, and time on the ventilator in both the intervention group and the control groups. We also examined an index of illness severity known as the All Payer Refined Diagnosis Related Group (APR-DRG) morbidity class. This index places patients into 1 of 4 categories based on International Classification of Diseases, Ninth Revision diagnosis and procedure codes and other billing data.
t Tests were used to compare the difference in utilization between the 2 periods within each ICU. The difference in change over time between the 2 ICUs was compared by testing the ICU and time interaction term in linear regression models. All tests with P values less than .05 were considered statistically significant.
During the intervention period, 225 patients were hospitalized in the CCU. There were 246 patients in the CCU during the corresponding months in 1998. Neither demographic characteristics of the patients nor APR-DRG morbidity class differed significantly between the 2 years (Table 1). There were more patients in 1999 with a noncardiac principal diagnosis (30.5% vs 18.5%; P = .004) (Table 2). No significant differences existed with respect to comorbid diagnoses.
Utilization of diagnostic tests in the 2 years is given in Table 3. Statistically significant reductions were noted in the ordering of all chemistry tests, with the largest changes seen for calcium, magnesium, and phosphorus (40%, 31%, and 40% reductions, respectively). Reductions in the use of other chemistry tests ranged from 7% (chloride) to 23% (potassium). Nonsignificant reductions were noted in the ordering of CBCs (P = .34), ABG measurements (P = .07), and CRs (P =
.10). In aggregate, the estimated reduction in expenditures for routine blood tests and CRs was 17%, from $84.20 to $69.98 (in 1998 dollars) per patient-day (P<.001). Analyses restricted to those patients with cardiac diagnoses yielded similar results. Because of the decrease in heart failure and arrhythmia admissions between 1998 and 1999, we performed additional analyses excluding patients with these diagnoses and observed similar reductions in utilization (data not shown).
Corresponding data for the MICU are also given in Table 3. In the MICU, there was no change between the 2 years in either demographic characteristics or APR-DRG morbidity class (data not shown). Ordering of diagnostic tests was not significantly changed, except for ionized calcium and radiographs, which were ordered more often. We compared the change over time between the CCU and MICU to account for potential hospital-wide changes in test-ordering practice (Figure 2). Observed trends in CCU testing were significantly different from observed trends in the MICU for the following tests: potassium, glucose, serum calcium, ionic calcium, magnesium, and phosphorus. In addition, the reductions in ordering CRs (P<.001) and ABG measurements (P = .04) were also significant when compared with MICU trends.
Data for patient outcomes in the CCU are given in Table 4. There were no significant differences between 1998 and 1999 in length of stay, hospital mortality, readmission to the ICU, readmission to the hospital, or days of ventilator support. In contrast, readmission to the hospital increased in the MICU population (data not shown).
We demonstrated a modest reduction in routine testing in the CCU after an intervention consisting of guideline development, order template design, and education. This change in practice was more marked when changes in a control ICU were considered. These reductions were not associated with any change in measured clinical outcomes, although our power to detect such a change was limited. Our results build on prior work showing that it is possible to reduce testing through guidelines, administrative interventions, and educational efforts.1,13- 15
A wide variety of strategies have been tried to improve physician test ordering, targeting both selected tests and broad panels of laboratory and radiographic studies.1,2 Solomon et al1 have drawn on a behavioral model to classify these strategies into those that target attitudes or knowledge (predisposing factors), those that facilitate behaviors or reduce barriers to them (enabling factors), and those that reward or penalize certain behaviors (reinforcing factors). They reviewed 49 studies and found that interventions targeting multiple behavioral factors appeared more effective than those that focused on a single factor. Educational efforts (to address predisposing factors) coupled with interventions to target enabling factors were particularly effective.
Our intervention attempted to address both predisposing and enabling factors. A major focus was to educate residents and nursing staff on "appropriate" ordering of routine laboratory and radiographic tests. We derived our guidelines from a review of the literature and by the consensus of a group of clinicians and nurses experienced in the care of critically ill patients. In addition, we attempted to facilitate appropriate test ordering by incorporating these guidelines into an admission computer order template for CCU patients. However, our intervention did not include explicit feedback on testing behaviors to ordering clinicians (ie, we did not target reinforcing factors).
An additional feature of the intervention was that it involved nurses as well as physicians. Unlike the house staff, the CCU nursing staff does not change substantially from month to month. As a consequence, enlisting the support of the nurses and ensuring their understanding of the guidelines were important elements of the intervention.
Few of the large number of studies examining test-ordering behavior have been set in ICUs, and, to our knowledge, none have focused on the CCU in particular. Interventions that are successful in other settings may not be relevant in the ICU setting because of the nature of diseases encountered in ICUs. In addition, specific perceptions may be attached to CCU patients—for instance, magnesium should be checked daily because cardiac patients are vulnerable to life-threatening arrhythmias.
Our study demonstrates the feasibility of a simple intervention in the CCU to change test-ordering practice. The most marked change was observed in ordering of additional chemistry measurements, such as calcium, magnesium, and phosphorus. Only modest reductions were observed in the ordering of other tests, including the standard electrolyte measurements, CBCs, and CRs, perhaps because our guidelines supported the use of these tests on a daily basis in most patients.
Similarly, the ordering of ABG measurements was only modestly affected. After the intervention, an average of 4.24 ABG measurements per day were still being ordered for ventilated patients. Measurements of ABGs are the most frequently ordered tests in ICUs.6 Prior studies suggest that twice-daily or even once-daily blood gas measurements may be sufficient in ventilated patients, without prolonging ventilator weans.6,16 Nonetheless, we encountered a widespread (and difficult to modify) perception that every change in ventilator settings should be accompanied by an ABG determination. Because house staff teams changed monthly, the practice of ordering frequent ABG measurements outside of the CCU may have hindered efforts to change practice within the CCU.
Despite the intervention, the absolute number of laboratory tests remained higher than would be predicted on the basis of the guidelines. For instance, while the guidelines recommended once-daily electrolyte measurement and CBC, actual ordering was approximately 1.3 tests and 1.5 tests per patient-day, respectively. While some of this excess may have been clinically indicated, these figures suggest scope for further reductions in utilization.
Several limitations of our study deserve comment. We assessed utilization only during the intervention period. It is important to know whether these reductions can be maintained over longer periods. While we found no measurable differences in clinical outcomes between the intervention and control groups, our study was not powered to assess these differences. In addition, our measures of severity of illness were crude.
We also acknowledge the limitations inherent in our comparison groups. The use of historical controls may not account for secular trends in the severity of illness or hospital practices. For this reason, we used a second control group (ie, MICU patients hospitalized during the same period) and reached similar conclusions. Randomly assigning patients to guideline-based or nonguideline-based care would not have been possible given the nature of our intervention.
Our experience suggests that interventions to improve the use of diagnostic testing in the CCU are feasible, but also highlights which testing practices may be more difficult to modify. In particular, the dependence on ABG measurements in ventilated patients was only modestly affected by our intervention. Broader dissemination of guidelines and greater educational efforts may have helped. Whether alternate methods, such as feedback or barriers to ordering, would have promoted larger reductions warrants further investigation.
Accepted for publication January 17, 2002.