Overview of the study design and data collection decision tree.
Schmidt M, Lindenauer PK, Fitzgerald JL, Benjamin EM. Forecasting the Impact of a Clinical Practice Guideline for Perioperative β-Blockers to Reduce Cardiovascular Morbidity and Mortality. Arch Intern Med. 2002;162(1):63–69. doi:10.1001/archinte.162.1.63
β- Blockers reduce morbidity and mortality when administered to high-risk patients undergoing major noncardiac surgery, yet little is known about how often they are being prescribed. Clinical practice guidelines are tools that can be used to speed the translation of research into practice and may be one method to improve the use of β-blockers. Before implementing any guideline, it is important to forecast its potential clinical and financial impact.
We conducted a retrospective cohort study, using administrative and medical record review data, of all adult patients undergoing major noncardiac surgery at Baystate Medical Center, Springfield, Mass, during a 1-month period in 1999. Patients with 2 or more cardiac risk factors or with documented coronary artery disease were classified as high risk and were considered eligible for treatment with a β-blocker if they had no obvious contraindications to its use. We estimated the potential clinical benefit of treating eligible patients with a β-blocker by extrapolating the treatment effect observed in a previously reported randomized clinical trial.
Of 158 patients undergoing major noncardiac surgery, 67 (42.4%) seemed to be ideal candidates for treatment with perioperative β-blockers. Of these 67 patients, 25 (37%) received a β-blocker at some time perioperatively. During the course of a year, we estimate that between 560 and 801 patients who do not receive β-blockers might benefit from treatment with these medications. Full use of β-blockers among eligible patients at our institution could result in 62 to 89 fewer deaths each year at an overall cost of $33 661 to $40 210.
There seems to be a large opportunity to improve the quality of care of patients undergoing major noncardiac surgery by increasing the use of β-blockers in the perioperative period. A clinical practice guideline may be one method to achieve these goals at little cost.
PERIOPERATIVE myocardial ischemia and myocardial infarction are important causes of morbidity and mortality among patients undergoing major noncardiac surgery.1- 4 At highest risk are the estimated 7 to 8 million patients with multiple cardiac risk factors or with established coronary artery disease (CAD).1,3 By slowing the heart rate, decreasing blood pressure, and moderating hemodynamic stress responses, β-blockers reduce the incidence of perioperative myocardial ischemia.5- 7 In a recent Veterans Affairs hospital randomized controlled trial,8 the perioperative administration of the β-blocker atenolol to high-risk patients resulted in significant reductions in mortality and cardiovascular complications beginning at 6 months and extending to 2 years following surgery. Following this report, the American College of Physicians9 recommended that β-blockers be considered for perioperative use in all high-risk patients undergoing major noncardiac surgery. More recently, bisoprolol was shown to reduce mortality when administered perioperatively to high-risk patients undergoing vascular surgery.10 Although we are not aware of any studies documenting how often perioperative β-blockers are used in routine practice, if practice patterns in patients with myocardial infarction can serve as a guide, it is quite possible that these medications are being underused.11,12
Clinical practice guidelines are "systematically developed statements to assist practitioner decisions about appropriate health care for specific clinical circumstances."13 They are one of several tools intended to speed the translation of research findings into routine clinical practice.14- 19 While individual studies20- 23 of the effectiveness of clinical practice guidelines have demonstrated mixed results, a systematic review24 of the guidelines' literature supports their continuing role in quality improvement initiatives. We developed a clinical practice guideline, based on the eligibility criteria and treatment protocol used in a recent randomized controlled trial,8 to improve the use of β-blockers in patients undergoing major noncardiac surgery at our institution. Before implementing this guideline, we sought to forecast its clinical and financial impact.
Our main objective was to estimate the number of patients undergoing major noncardiac surgery at our institution who might benefit from perioperative β-blockers but who do not receive these medications. We used this information to estimate the clinical and financial impact if our guideline were reasonably successful in meeting its goals.
The study was performed at Baystate Medical Center, Springfield, Mass, a 550-bed community-based teaching hospital that serves as the western campus for Tufts University School of Medicine, Boston. The hospital is a regional referral center for a population of almost 1 million residents of western New England.
Relying on information contained in prior clinical trials and incorporating the input of a multidepartmental task force, we developed a practice guideline for perioperative β-blocker use. The guideline consists of a series of treatment recommendations that include specific eligibility criteria, dosing and monitoring recommendations, and a detailed treatment algorithm. Plans for dissemination and implementation of the guideline included a mailing to all anesthesiologists, surgeons, and internists on the medical staff; presentations at appropriate grand rounds and noon conferences; and the development of an online version of the recommended order set using the hospital's electronic physician order entry system. Implementation plans called for providing partial salary support to a faculty physician to serve as a "clinical champion" for the project. This physician was expected to maintain awareness of the guideline through lectures and informal discussions, through individualized academic detailing, and by being available to discuss questions or problems related to the guideline. As an additional implementation measure, we planned to have the hospital case manager for surgery play an active role in identifying eligible patients and reminding physicians about the availability of the guidelines.
The study was a retrospective cohort analysis using administrative and medical record review data of patients undergoing major noncardiac surgery at our institution.
All patients undergoing surgery in the institution's main operating suites between January 2, 1999, and February 2, 1999, were identified using the hospital's computerized database to determine potential eligibility. Surgery that took place outside this arena was limited to "day-stay" procedures. Patients who were 18 years and older and who underwent major noncardiac surgery were considered potentially eligible. Surgical procedures considered major for this study were determined by a group of physicians and nurses representing the departments of surgery, anesthesiology, and internal medicine. These procedures include the following: orthopedic procedures, including total knee/hip replacement, bipolar hip replacement, open reduction internal fixation (hip or pelvis), and limb amputation; vascular procedures, including aortofemoral bypass, abdominal aortic aneurysm repairs, and any other surgery on the aorta; intra-abdominal procedures, including colectomy, splenectomy, pancreatoduodenectomy (Whipple procedure), radical prostatotomy, radical hysterectomy, liver resection, nephrectomy, and total abdominal hysterectomy/bilateral salpingo-oophorectomy; and intrathoracic procedures, including all invasive noncardiac procedures. We excluded patients undergoing day-stay surgery and those who underwent a second surgical procedure during a single hospitalization. Demographic data, including age, sex, race, type of surgery, length of hospital stay, and prior use of β-blockers, were collected for all potentially eligible patients.
A 2-step process was used to determine patient eligibility for treatment with perioperative β-blockers. The first step identified all patients meeting the guideline definition of high perioperative risk. The eligibility criterion was established CAD (including any one of the following: prior myocardial infarction, classic angina, atypical angina with a positive stress test result, history of coronary revascularization, or cardiac catheterization showing significant stenosis) or being at high risk for CAD (including at least 2 of the following: 65 years or older, having hypertension, currently smoking, having a total serum cholesterol level ≥240 mg/dL [≥6.21 mmol/L], or having diabetes). The second step consisted of screening all high-risk patients for conditions that might be a contraindication to β-blocker use (including poorly controlled reactive airway disease [asthma or chronic obstructive pulmonary disease], decompensated congestive heart failure, bradyarrhythmia [heart rate <55 beats/min without a pacemaker], sick sinus syndrome, second- or third-degree heart block, a systolic blood pressure <100 mm Hg, or carotid hypersensitivity). High-risk patients who had no obvious contraindications to β-blocker use were considered to be ideal candidates for treatment with perioperative β-blockers according to our guideline.
For those patients who seemed to be ideal candidates for treatment with a β-blocker, we recorded the following: whether a β-blocker was used at any time during the hospital course, the development of potential contraindications to β-blocker use postoperatively, the ability to take oral medications following discharge from the postanesthesia critical care unit, the prescription of a β-blocker at discharge, the incidence of postoperative tachycardia (defined as a recorded heart rate ≥100 beats/min at any time during the postoperative course), and in-hospital death.
In the clinical trial by Mangano et al,8 patients who received β-blocker therapy had higher overall and event-free survival at 6 months, 1 year, and 2 years compared with those receiving a placebo. The 1-year mortality of 14% in the placebo group and 3% in the atenolol-treated group results in a number needed to treat to prevent 1 death of 9. The 1-year cardiovascular complication rate of 22% in the placebo group and 8% in the atenolol-treated group results in a number needed to treat to prevent 1 cardiovascular complication of 7. We extrapolated the results from that study to estimate the potential impact the implementation of this guideline might have on morbidity and mortality at our institution. Recognizing the limited effectiveness of clinical practice guidelines, even when they are aggressively implemented using proved methods, we estimated that only 50% of the patients who appeared to be ideal candidates, and yet were not receiving β-blockers, would likely be administered a perioperative β-blocker as recommended by our guideline. In other words, we assumed that the guideline would be successful in 50% of the cases.
To provide an estimate of the costs of developing and implementing this guideline, we considered the medical and the administrative costs that would likely be incurred. We also estimated the potential savings associated with a forecasted reduction in cardiovascular events during the course of 1 year.
Estimates of medical costs were based on the marginal drug and telemetric monitoring costs for the additional patients who we predicted might receive β-blocker treatment. Drug costs reflect pharmacy acquisition costs for generic oral and intravenous atenolol at the study institution's pharmacy in February 2000. Our guideline recommends intravenous atenolol for patients who are unable to take oral medications postoperatively. Moreover, it is the clinical policy at our institution that a patient undergo telemetry when intravenous β-blockers are used. We estimated the additional cost of telemetric monitoring by multiplying the expected number of patient-days in which postoperative telemetry would be required by the daily cost for telemetric monitoring. Costs were obtained directly from the hospital's cost-accounting system (Health Management Systems, El Segundo, Calif).
Estimates of the administrative costs incurred in developing and implementing the guideline were based on several factors. These included the following: the costs of supporting 10% of the annual salary of the physician who served as the clinical champion during the program's first year and 5% during subsequent years, the costs of supporting 20% of the salary of the surgical case manager each year, the computer programming costs associated with developing online versions of the recommended order sets, and the printing and mailing costs to distribute the guideline to the medical staff.
We used the average cost of caring for a patient with an acute myocardial infarction (International Classification of Diseases, Ninth Revision, Clinical Modification code 410) at the study institution in 1999 to estimate the potential cost savings associated with a reduction in cardiovascular events that might be expected each year. The cost for caring for patients within this International Classification of Diseases, Ninth Revision, Clinical Modification code was obtained from the hospital cost-accounting system. We multiplied this figure by the forecasted number of cardiovascular events averted each year to estimate the potential savings resulting from the guideline.
During the 1-month study period, 1132 patients were admitted to the institution's main operating room. Of these patients, 212 met the age and type of surgery inclusion criteria. Medical records were readily available for 158 (75%) of these patients. Of the 158 study subjects, 81 (51%) were found to have either established CAD or 2 or more major risk factors. Of these 81 patients, 14 (17%) had a contraindication to β-blocker use. The remaining 67 patients (42% of the total study group) seemed to be ideal candidates for treatment with β-blockers according to our guideline (Figure 1).
Demographic information on the subgroup of patients who seemed to be ideal candidates for β-blocker treatment and on the patients not eligible for treatment is presented in Table 1. There is a significantly higher proportion of men among ideal patients compared with those not eligible. The distribution of type of surgery also differs significantly between the groups, with higher rates of vascular and orthopedic surgery among the ideal patients and lower rates of abdominal and thoracic surgery in this group.
As expected, based on eligibility criteria, ideal patients tended to be older, with a higher proportion having established CAD and having more CAD risk factors present. The one exception we found was smoking, which was more common among the ineligible patients. Thirty-three patients overall (21%) were taking β-blockers before hospital admission.
Of the 67 ideal patients, 25 (37%) who we predicted might benefit from β-blocker therapy had received a β-blocker at some time perioperatively. If we include use before admission, this increases to 31 (46%) of the ideal patients. Of the 42 patients who were not taking β-blockers before admission, 6 (14%) had them initiated during hospitalization. Another 6 patients (6 [24%] of the 25 taking β-blockers before admission) did not continue taking them during the hospital stay. Almost all patients who were admitted to the hospital while taking β-blockers (30 [91%] of 33 patients) were prescribed them at discharge.
Of the 67 ideal patients, 21 (31%) were unable to take medications by mouth for an average of 3 days following surgery. Twenty (30%) of the ideal candidates developed a new contraindication to β-blocker use at some time postoperatively. Of these contraindications, hypotension (≥1 reading of a systolic blood pressure of <100 mm Hg) was the most common, and was found in 17 (25%) of the eligible patients. Four patients had β-blockers discontinued secondary to a contraindication. Postoperative tachycardia was observed in 25 (50%) of the 50 ideal patients who did not receive a β-blocker and in 12 (50%) of the 24 ideal patients who did receive β-blocker treatment. Of 42 eligible patients who did not receive a β-blocker perioperatively, 5 (12%) died before hospital discharge compared with 0 of the 25 patients who received treatment.
A 95% confidence interval on the observed proportion (42%) of patients eligible indicates that we can expect between 35% and 50% of patients undergoing major noncardiac surgery to be eligible for treatment with perioperative β-blockers. In light of the annual number of these operations carried out at our institution, we expect between 890 and 1272 patients per year to be eligible for treatment. Based on our observations, we anticipate that only slightly more than one third (25 [37%] of 67 patients) are already receiving a perioperative β-blocker. Therefore, each year, some 560 to 801 patients who seem to be ideal candidates are not receiving treatment. If our guideline is 50% successful, then some 280 to 400 ideal candidates who are not receiving this therapy would begin receiving β-blocker therapy perioperatively each year. Twenty (30%) of the 67 patients developed at least one potential contraindication to β-blockers postoperatively. Taking this into account, we estimate that approximately 196 to 280 patients would be able to tolerate treatment with β-blockers for the duration of their hospitalization.
Extrapolating the benefit observed in the Veterans Affairs' study, in which the number of patients needed to treat to prevent a single death at 1 year was 9, we estimate that as many as 62 to 89 additional patients might be alive at 1 year if all ideal candidates at our institution were to receive this treatment. Taking into account the limited effectiveness of guidelines in changing physician behavior, and the likelihood that some patients will not tolerate β-blocker therapy, we estimate 22 to 31 lives saved each year. Similarly, if all ideal candidates were to receive β-blocker treatment, we would anticipate 80 to 114 fewer cardiovascular events each year (number needed to treat, 7). Discounting the effectiveness of the guideline and the tolerability of β-blocker therapy, we predict 28 to 40 fewer cardiovascular events during the course of each year as a result of our intervention.
Additional drug cost estimates for oral and intravenous atenolol are $3087 to $4393 per year, and the cost for additional monitored beds is from $15 399 to $21 948. Thus, the total medical cost of guideline implementation at the study institution was estimated to be $18 486 to $26 348 per year.
Administrative costs included $12 000 in the first year for physician salary support and $6000 per year in subsequent years, $10 000 per year for nurse case management, and 1-time programming and printing and mailing costs of $4185 and $1125, respectively. Thus, the total administrative costs were projected to be $27 310 in the first year and $16 000 in subsequent years. During the first 5 years, the average administrative cost of developing and implementing this guideline is projected to be $18 262 per year. Taking into account medical and administrative factors, we estimate $33 661 to $40 210 in annual costs attributable to the perioperative β-blocker guideline or $1297 to $1530 per life saved.
Given predictions of 28 to 40 fewer annual cardiovascular events and an average cost per myocardial infarction at the study institution of $12 588, we estimate annual savings of $352 464 to $503 520. Balancing these savings against the costs of developing and implementing this guideline results in overall savings of $318 803 to $463 310.
In a recent review, the American College of Physicians9 recommended that all high-risk patients undergoing major noncardiac surgery be considered for treatment with a β-blocker in the perioperative period. We sought to forecast the impact of a clinical practice guideline designed to assist clinicians in following these recommendations. We estimated the proportion of patients undergoing major noncardiac surgery who were likely to be classified as high risk using the criteria from our guideline and identified practice patterns surrounding perioperative β-blockade. We found that 51.3% of the patients undergoing major noncardiac surgery had either established CAD or at least 2 major risk factors for CAD—and, thus, were classified as being at high risk for major postoperative complications. After excluding patients with known contraindications to β-blocker therapy, we found that 63% (n = 42) of this population did not receive a perioperative β-blocker. If all ideal candidates at our institution were to receive this treatment, we estimate that an additional 62 to 89 patients might be alive at 1 year and that 80 to 114 patients might survive 1 year without a major cardiovascular complication. Taking into account the challenges of achieving full compliance with any practice guideline and the tolerability of β-blockers, we anticipate 28 to 40 fewer cardiovascular events and 22 to 31 fewer deaths each year. The administrative and direct medical expenses required to fully implement this guideline were estimated to be $33 661 to $40 210 per year. Considering the potential savings associated with anticipated reductions in cardiovascular complications such as acute myocardial infarction, use of this guideline in the perioperative period would be anticipated to save between $318 000 and $463 000 annually.
There are several limitations to the present study. First, patient eligibility for treatment with β-blockers was determined retrospectively via medical record review. Retrospective classification of eligibility status may differ from eligibility classification performed during real clinical time because of, for example, incomplete patient histories in the medical record. However, this is unlikely to be a major source of error in the present study, as most of the CAD diagnosis and risk factor data came from preoperative anesthesia records, which we believe reflect a reasonably accurate assessment of cardiovascular risk. Second, the present study population includes only patients undergoing major noncardiac surgery. We developed this definition after obtaining the input of an internist, an anesthesiologist, a surgeon, and a nurse, but may have excluded some procedures that could arguably be considered major. As the definition of what constitutes major surgery changes, the proportion of high-risk patients would change as well. For example, if total abdominal hysterectomy and bilateral salpingo-oophorectomy—which are predominantly performed on younger women without a history of CAD—were not considered major surgery, the overall proportion of patients undergoing major surgery who are eligible for the protocol would increase.
Our method for estimating the clinical impact of the proposed guideline also has several significant limitations. First, evidence supporting the efficacy of perioperative β-blockers in reducing morbidity and mortality is not extensive,5,8,10 and only 2 clinical trials8,10 have examined the effects of β-blocker therapy administered during the entire hospital stay. Our forecast of clinical impact assumes that the effects observed by Mangano et al8 are generalizable. However, differences between the 2 study populations, such as in case mix or severity, may result in differences in the impact of β-blocker therapy on mortality and morbidity. We used eligibility criteria to select patients who were nearly identical to those used in the Veterans Affairs study. This should have the effect of adjusting for some of the differences between a veterans hospital's population and the population at an urban community-based teaching hospital. This is supported by a similar distribution of CAD diagnosis and CAD risk factors between the 2 studies. One notable difference between the 2 study populations is the use of an all-male population in the veterans hospital clinical trial. To the extent that sex modifies the beneficial effect of β-blocker therapy or significantly alters the risk of postoperative mortality and morbidity, error in the present study's estimates would result. We recognize that clinical practice guidelines are only one of several techniques used to improve quality of care and that the effectiveness of such interventions is often limited. We estimated that we would be successful in achieving 50% compliance with our guideline by incorporating several proved implementation techniques. Although this estimate may be either overly optimistic or alternatively too pessimistic, it falls within the range of reported compliance in other guideline trials25,26 that used similar implementation techniques.
Our methods used to quantify the expected number of patients who will develop a contraindication to β-blockers have several limitations. Most notably, our assessment of contraindications did not fully take into account actual β-blocker use (only 25 [37%] of 67 patients took a β-blocker at least once). Those patients who did take a β-blocker in the present study were more likely to have been taking a β-blocker before hospital admission and, thus, less likely to develop a new contraindication because of use. By including these patients in our calculation, we may have significantly underestimated the proportion of patients likely to develop a contraindication after beginning treatment with a β-blocker. Alternatively, our estimates of new-onset contraindications may overestimate the number of patients who would have to terminate treatment. For example, any study patient with a systolic blood pressure lower than 100 mm Hg on a single measurement was considered to have a new contraindication. However, it is unlikely that a single episode of mild hypotension would always be viewed as clinically meaningful and result in termination or suspension of treatment. Indeed, we found that many patients who were taking a β-blocker postoperatively continued to receive these medications even after we noted a potential contraindication, such as a low systolic blood pressure.
Owing to the retrospective nature of this study, we were unable to detect subtle, yet important, clinical outcomes such as myocardial ischemia or subclinical myocardial infarction. We found similar rates of tachycardia in the β-blocker–treated and the nontreated groups; however, we only measured the overall frequency of postoperative tachycardia and were not able to quantify the time patients spent tachycardic or the mean heart rate postoperatively. Although it is possible that patients who received β-blockers were suboptimally treated, our methods of measuring the effect of β-blockade were not sensitive enough to detect clinically important differences between the 2 groups. Furthermore, because treatment assignment was not random, we were not confident in attempting to compare the clinical outcomes of the treated with those of the nontreated patients.
A major contributor to our estimate of the cost of fully implementing this guideline was the cost of telemetric monitoring for patients who receive nothing by mouth following surgery—because these patients may require treatment with intravenous β-blockers. Although our hospital requires that telemetric monitoring be used for patients taking intravenous β-blockers, this may not be the case at other institutions. Moreover, we assumed that telemetric monitoring would be required for the entire time that patients were to receive nothing by mouth, yet it is conceivable that this monitoring could be limited to the first several doses of medication. In addition to calculating drug and monitoring costs, we estimated the administrative costs required to develop and implement this guideline. These included the costs of a physician champion, a nurse case manager to provide real-time reminders to clinicians, an electronic version of the guideline and order set programmed into the hospital electronic physician order entry system, and printing and mailing. While these techniques increase the likelihood that a guideline will be adopted by clinicians,27 not all institutions are likely to commit this level of support. Finally, we estimated potential savings resulting from a reduced incidence of cardiovascular complications during the course of 1 year using the cost of caring for a single episode of myocardial infarction at our own institution. Although we considered these savings taking the perspective of society, we recognize that many hospitals would not reap any financial benefit from reducing the incidence of complications 6 months following discharge.
If β-blocker therapy is a beneficial treatment for high-risk patients in the perioperative period, enhancing β-blocker use should be a priority. The use of a clinical practice guideline is one possible technique for improving the use of perioperative β-blocker therapy in high-risk patients undergoing major noncardiac surgery. Nevertheless, it is often difficult to foresee the applicability and the likely costs and benefits of a practice guideline before its implementation. For this reason, some researchers28,29 have suggested that retrospective analyses be used to forecast the impact of clinical guidelines before their implementation. We believe that this study provides a reasonable estimate of the clinical and financial impact of a perioperative β-blocker guideline by retrospectively applying guideline criteria to a population of consecutive patients undergoing major noncardiac surgery. We found that many of these patients (51%) are at high risk for perioperative complications and that most do not receive perioperative β-blockers. If practice at our hospital reflects care occurring at similar institutions, then there is a sizable opportunity to improve the care of patients undergoing major noncardiac surgery nationwide. Based on this analysis, the use of a clinical practice guideline or other measures intended to improve the use of perioperative β-blockers in high-risk patients should be strongly considered.
Accepted for publication April 19, 2001.
We thank Penelope Pekow, PhD, for her advice and assistance, and Alan Kulig, MD, for his leadership in developing and implementing the perioperative β-blocker guideline.
Corresponding author and reprints: Peter K. Lindenauer, MD, MSc, Division of Healthcare Quality, Baystate Medical Center, 759 Chestnut St, P-5928, Springfield, MA 01199 (e-mail: Peter.Lindenauer@bhs.org).