Queuing for Coronary Angiography During Severe Supply-Demand Mismatch in a US Public Hospital: Title and subTitle BreakAnalysis of a Waiting List Registry FREE

In recent years a debate has been waged over the problems of escalating costs and increasingly limited access to medical care in the United States.^{1 - 3} As Americans consider health care system reform and restricting access to health care, experiences from countries with universal access policies and national health care systems have come under scrutiny. An integral component of most national health care systems is the use of waiting lists for nonurgent medical care and procedures.^{1 - 3} In this setting, the care of patients with cardiovascular diseases is of particular concern because adverse cardiac events and excess deaths have been increasingly reported during waits for either percutaneous transluminal coronary angioplasty (PTCA) or coronary artery bypass grafting.^{4 - 10}

Diagnostic procedures are equally as important as therapeutic interventions, raising the question of whether access to coronary angiography (CA) can be safely limited. The use of CA has undergone extraordinary expansion during the last decade, stimulated by the explosive growth of percutaneous revascularization techniques.¹¹ Although the existence of waiting lists for CA in different non-US health care delivery and/or financing systems has been reported,^{12 - 14} it is generally accepted that US hospitals provide the shortest waits for invasive cardiovascular procedures.¹⁵ Importantly, the most cited study on this subject¹⁵ reported data generated by the response of physicians to hypothetical case scenarios and did not consider practical issues such as access to health care. As the number of Americans lacking health insurance increases—to more than 41 million in recent surveys¹⁶ — it is likely that a sizable proportion of the US population may experience difficulty in accessing health care and delays in receiving medical services.^{17 - 24} This problem has been exacerbated as declining reimbursement rates and market forces have led to an increasing inability of private hospitals to cross-subsidize charity care for the underinsured and uninsured.^{1 - 3} More so than ever before in the United States, large numbers of patients who were previously cared for in private hospitals are being diverted to public hospitals. These have become increasingly crowded, ^{21 - 25} leading to a de facto need to ration care.

The University of Texas Medical Branch (UTMB) at Galveston is a US public teaching hospital with 1236 licensed beds that serves as a tertiary referral center for patients from 231 of Texas' 254 counties. Uninsured patients represent 23% of UTMB's admissions and 43% of outpatient visits (UTMB Healthcare Financial Management Office, unpublished data, fiscal year 1997). The high volume of patients seeking care at UTMB has taxed resources available to provide health services. In the years 1993-1994, an unprecedented influx of patients referred to UTMB for CA overwhelmed the capability of the cardiac catheterization laboratory and a waiting list for nonurgent procedures had to be established. The present study describes the occurrence of death, nonfatal myocardial infarction (MI), and cardiac-related hospitalizations among these patients. The clinical importance of this study is that objective data on the risk of adverse events in patients awaiting elective CA are limited, and none are available from a US population.

The observational cohort was composed of 381 adult outpatients placed on a waiting list for CA. Patient queuing was based on referral date. Patients not eligible for the waiting list were those with (1) recent MI (<30 days from symptom onset), (2) potentially life-threatening diseases of the ascending aorta (aneurysm or dissection), and (3) modified Canadian Cardiovascular Society (CCS) angina classes IVb or IVc.²⁶ To accurately describe the medical consequences of prolonged waits on cardiovascular morbidity and mortality, we excluded patients with a short life expectancy due to other illnesses such as cancer or severe pulmonary, hepatic, or renal disease. The study was approved by the UTMB institutional review board.

At the time of referral, all patients were entered into a registry consisting of a modular database, which included demographic, clinical, and insurance data.

Noninvasive stress tests included treadmill exercise-electrocardiogram (ECG) and exercise or pharmacologic stress imaging (thallium scintigraphy or echocardiography). A positive exercise-ECG result was defined as a 0.1 mV or greater horizontal or downsloping ST-segment depression in a lead with a normal baseline ST segment or typical angina. A strongly positive result was defined as an early (<3 minute) or striking (≥2 mV) ST depression, a drop in systolic blood pressure of 20 mm Hg or more during exercise, or the persistence of ST depression longer than 6 minutes after exercise. A positive nuclear imaging study result was defined as a reversible thallium perfusion defect in 2 or more contiguous zones or abnormal distribution associated with increased lung uptake in the absence of severely depressed resting left ventricular function. A positive stress echocardiography result was defined as the development of stress-induced, new, or worsening regional wall motion abnormalities in 2 or more contiguous segments.

Clinical data, symptom status, and stress test results were evaluated in combination to determine whether a patient met appropriate indications for CA. Appropriateness was rated according to the classification proposed by the American College of Cardiology and American Heart Association (ACC/AHA).²⁷

Significant coronary artery disease (CAD) was defined as at least 50% lumen diameter narrowing of a major epicardial coronary artery or its major branches. In classifying the number of diseased vessels, a left main coronary artery stenosis was regarded as equivalent to stenosis in both the left anterior descending and the left circumflex arteries.

Patients were scheduled for CA according to their position on the waiting list. A common method for filling unexpected cancellations in the catheterization laboratory was to call a waiting list patient for the next day. If the patient was unable to come in, the available opening was offered to the next patient on the list. We used chart abstraction to obtain information concerning the frequency of outpatient visits, prescription changes, and hospitalizations, including diagnosis and length of stay. This evaluation included not only the UTMB hospitals but all hospitals to which patients were admitted during follow-up.

For study purposes, follow-up was considered complete at the time of CA or at the time of an adverse event, if such event occurred before performance of CA at the UTMB or another hospital. Adverse events were as follows: (1) cardiac death (defined clinically as fatal MI, documented arrhythmic death, death related to heart failure, or sudden death without a noncardiac cause); (2) acute MI (defined as chest pain ≥20 minutes in duration and a rise in serum cardiac enzyme levels to at least twice the upper limit of normal, or the appearance of new pathologic Q waves); and (3) need for emergency or urgent hospital admission due to unstable angina (defined as accelerating or new-onset angina at rest or on exertion in patients who either have new ST segment–T wave changes, or in patients without clear ECG changes but with a history of CAD documented by angiography or noninvasive stress tests) or worsening heart failure (signs or symptoms of congestion or low cardiac output believed to be secondary to cardiac dysfunction).

Data are expressed as mean (SD) or percentages, where appropriate. Waiting times, because of their highly skewed distribution, are presented as medians with 25th and 75th quartiles. To compare differences between groups for continuous and categorical data, t tests and χ² tests with Yates correction, respectively, were used. Waiting times were compared using the Mann-Whitney U test. The probabilities of events were estimated with the Kaplan-Meier method. Individual and joint associations of demographic and clinical variables to adverse events were estimated using Cox proportional hazard regression analysis. The proportional hazards assumption was verified by excluding a crossover between survival curves for each binary variable. Conformity to the assumption of a linear gradient for continuous or ranked variables was graphically examined by plotting the observed and predicted values for the outcome over the range of each predictor. As a measure of the strength of the association between each predictor and the outcome variable, univariate crude odds ratios (ORs) and multivariate adjusted ORs were computed, with respective 95% confidence intervals (CIs). Given the limited number of events, only the univariate correlates with associated P≤.10 values were entered into multivariate analysis to minimize the risk of model overfitting. The likelihood ratio χ ² value was used as an indicator of the overall fit of the multivariate model. Internal stability of risk estimates was validated via bootstrap procedures. Missing data were analyzed using the pairwise deletion method. For variables with more than 10% of data unavailable, the potential for a systematic bias in risk estimates resulting from a "hidden" nonrandom distribution of the missing data was controlled by verifying their pattern across the correlation matrix of parameter estimates iterated using the Newton-Raphson method. For all tests, a 2-sided P<.05 was regarded significant.

Follow-up data were available for all 381 patients on the waiting list, including 66 (17%) who were dropped from the list for the following reasons: CA canceled by the referring physician (n=22), declined the study (n=6), or underwent CA in another hospital (n=38).

During a mean (SD) follow-up of 8.4 (6.5) months, 36 patients (9.4%) experienced adverse events, including 6 deaths (1.6%), 4 MIs (1.0%), and 26 unplanned hospitalizations (6.8%) for heart failure (n=5) or unstable angina (n=21). The occurrence of unstable angina in these 21 patients accounted for 58% of all adverse events. Of the total 36 adverse events, 4 (2 deaths, 1 MI, and 1 hospitalization for heart failure) occurred in the 66 patients dropped from the waiting list.

The temporal distribution of the 36 adverse events is shown as Kaplan-Meier event-free survival curves (Figure 1). The probability of events was minimal in the first 2 weeks and increased steadily between weeks 3 and 13. At 3 months, event-free rates for death, MI, and hospitalization were 98%, 99%, and 92%, respectively. Freedom from all adverse events at 3 months was 89%.

Table 1 shows the demographic and clinical characteristics for all patients and for patients grouped according to the occurrence of adverse events. Patients experiencing adverse events had a significantly higher frequency of cigarette smoking, more than 2 cardiac risk factors, and prior PTCA. A combination of 2 to 3 anti-ischemic drugs was given to 67% of the patients with events vs 39% of those without events (P=.001). There were no significant differences in the frequency of outpatient visits and prescription changes during follow-up between patients with and without adverse events.

Diagnostic categories, presenting symptoms, stress test results, and ACC/AHA Class of indications for CA are shown in Table 2. Compared with the event-free group, patients with adverse events more frequently had a history of known CAD, CCS angina class III or IVa, and positive stress test results. Furthermore, the percentage of strongly positive exercise-ECG or positive stress imaging results was significantly higher in patients with events than in those without (P =.001). Patients with and without events differed significantly in the rate of ACC/AHA class I (P=.01) and class III (P=.03) indications for CA.

Cox regression analysis of the explanatory variables listed in Table 1 and Table 2 identified 7 univariate correlates of adverse events with P≤.10 (Figure 2). A multivariate model that included all these univariate correlates provided significant ability to predict adverse events (χ² likelihood ratio=17.76, P=.001). However, the only individually significant predictors of events in this model were a strongly positive exercise-ECG or positive stress imaging result at referral (adjusted OR, 2.32; 95% CI, 1.22-4.16; P=.01) and use of 2 to 3 anti-ischemic medications (adjusted OR, 1.98; 95% CI, 1.19-3.96; P=.04).

Of 381 patients placed on the waiting list, 311 underwent CA at UTMB (4 died before the scheduled CA and 66 were removed from the waiting list). Of the 311 patients, 283 (91%) had uneventful waits, whereas 28 (9%) experienced adverse events before the scheduled CA. Overall, 64% had significant CAD. Patients with events had a significantly higher prevalence of CAD ( P<.001) and more frequently required revascularization ( P<.001), predominantly coronary artery bypass grafts ( Table 3).

Overall, 8% of the patients underwent CA within 2 weeks of referral, 44% waited 2 to 6 weeks, 32% waited 6 weeks to 3 months, and 16% had waiting times longer than 3 months. Waiting time and length of hospital stay are reported in Table 4. Patients who experienced adverse events had a significantly shorter waiting time vs patients without adverse events (P=.01). However, the length of hospital stay increased significantly for patients with adverse events ( P=.001). To determine whether differences in demographics and clinical characteristics influenced the length of wait for CA and the duration of hospital stay, we stratified patients according to selected variables ( Table 4). Age, sex, race, and symptom status were not different for patients with longer waiting times. Longer delays were observed among Medicaid-insured and uninsured patients compared with those privately insured (P<.001), but length of hospital stay was similar across insurance groups.

To clarify reasons for longer waiting times for the Medicaid-insured and uninsured patients, we calculated the percentages of patients who rescheduled CA or refused a next-day opening and of those who lived farther than 50 miles from the hospital. Only 11% of the privately insured patients were unable to make the appointments vs 38% of the Medicaid-insured and uninsured patients (P<.001). In addition, 40% of the Medicaid and uninsured patients lived farther than 50 miles from the hospital, whereas none of the privately insured patients did (P<.001).

The present study is the first to provide objective data on morbidity and mortality in a US population awaiting elective CA. Our experience suggests that delaying CA places patients at some risk of adverse cardiac events, longer hospital stays, and, potentially, a worse prognosis.

An important finding is that a 1- to 2-week wait for CA confers minimal risk, but that there is a progressive increase in the probability of adverse events between 3 and 13 weeks. Thus, waits longer than 2 weeks should be avoided, particularly for patients with strong indications for CA (ACC/AHA class I, representing 69% of those with adverse events in our population). In particular, the need for emergent or urgent hospitalization accounted for most of the adverse events, whereas the rates of death (1.6%) and MI (1.0%) were low. However, these rates are approximately 10 times those associated with cardiac catheterization (mortality, 0.11%; MI rate, 0.06%²⁸ ), emphasizing the importance of potentially preventable deaths or MIs among patients waiting for CA.

The risks outlined here are similar to other reports in the literature. In a study from Manitoba,¹⁴ after a mean wait for CA of 4.2 weeks, cardiac arrest, acute MI, and death occurred in 0.5%, 0.9%, and 0.4% of the patients, respectively, with a 3.7% rate of emergency admissions. In a second-opinion trial among patients recommended for CA for whom the procedure was deferred, annual rates of cardiac death and MI were 1.1% and 2.7%, respectively, during a nearly 4-year follow-up.²⁹ In addition, 21% of the patients had a cardiac hospitalization and 32% had aggravation of angina. The rates of adverse events while waiting for CA in our study and others are strikingly similar to those observed in patients awaiting coronary revascularization. Mortality rates of 0.4% to 2.2% have been reported in coronary artery bypass graft candidates after mean waiting times of 17 to 98 days,^{4 - 9} and adverse coronary events may occur in 23% of patients awaiting routine PTCA.¹⁰

It is significant that development of unstable angina accounted for 58% of the adverse events in our population because this diagnosis has a negative prognosis even if revascularization is performed.^{30 - 31} In fact, although the technical success of percutaneous interventions in unstable angina is similar to that for stable angina, the incidence of periprocedural complications as well as restenosis is higher.³⁰ In addition, emergency surgical intervention in refractory unstable angina is accompanied by increased operative mortality rates.³¹

We investigated whether characteristics predictive for long-term prognosis in patients without overt CAD or with stable ischemic heart disease could also be used to identify patients at risk among those awaiting CA.^{32 - 33} That patients taking multiple anti-ischemic medications had an increased risk of short-term adverse events is particularly interesting because it contradicts the belief held by many investigators that medical therapy and revascularization are equally safe in patients with stable angina and nonlimiting symptoms.^{34 - 35} Furthermore, close medical follow-up did not prevent adverse events: there was no significant difference in the intensity of outpatient follow-up between the 2 groups. In our population, the need for multiple drugs most likely identified patients with more severe or diffuse underlying CAD. Indeed, rapid progression of preexisting stenoses leading to acute coronary events may occur despite combined antianginal treatment, as demonstrated by a study of medically stable patients placed on a waiting list for routine PTCA.³⁶

Our findings that a strongly positive exercise-ECG or positive stress imaging result constitutes an independent risk of early cardiac events suggests that a strategy of objectively documenting the degree of inducible ischemia is not only useful for posing an appropriate indication for CA but also for identifying patients in whom CA can be safely deferred and those requiring prompt invasive evaluation.

To examine reasons for long waits in our registry, we analyzed several variables that have been shown to be relevant to the delivery, quality, and outcome of health care.^{17 - 20} Our finding of a disparity in waiting times between Medicaid-insured and uninsured patients and those privately insured raises important issues of access to health care. The high percentage of uninsured patients in our study sample is representative of Texas demographics as a whole, including large numbers of African American and Hispanic individuals, who have high rates of noncoverage.^{16 ,18} Additionally, Texas has relatively restrictive Medicaid programs, so that low-income, working individuals are less likely to qualify for federally funded medical coverage than they are in other states.¹⁶ Since health care services at UTMB are offered regardless of ability to pay, it is unlikely that cost-containment policies encouraging lower levels of resource utilization for uninsured patients adversely affected the waiting time for CA. Because Medicaid-insured and uninsured patients had to be called more often than privately insured patients to schedule CA and were more likely to live farther than 50 miles from the hospital, we speculate that factors such as conflicts with work schedules, child care responsibilities, and limited ability to afford transportation to access distant medical care may have contributed to the longer delays for this group.

Some limitations of our study merit discussion. The interval between referral and performance of CA was not constant for all patients, but varied depending on the dynamics of the waiting list, which were, in turn, determined by the occurrence of clinical events, bed availability, and other practical issues. Therefore, some caution is recommended in interpreting the impact of the length of time on the waiting list because the selection process itself may have influenced outcome.

While our study included some patients who underwent CA for other cardiac disease, our population was predominantly composed of patients evaluated for known or suspected CAD. Therefore, our findings probably can be generalized to the vast majority of patients referred for elective CA. Whether the results from this single tertiary public hospital can be extrapolated to the US population in general is unknown.

Physicians would prefer not to delay needed medical and surgical services. However, our experience underscores the problem of limited access and delays in the delivery of health services for a population largely composed of uninsured patients in the setting of a US public hospital. Apart from the effects on patients of persisting symptoms and the anxiety associated with waiting,^{12 ,37} our results suggest that destabilization of patients' clinical conditions while awaiting CA may lead to unplanned hospital admissions and longer hospital stays (with their own costs to patients and society), expedited procedures (bumping other patients down the waiting list), and increased risks in the context of acute conditions such as MI, unstable angina, or worsening heart failure. The issue of consecutive delays for both CA and revascularization also should be considered. The benefits of PTCA and coronary artery bypass graft in reducing symptoms, improving quality of life, and extending the life expectancy of defined patient groups are well known,^{33 ,38 - 39} and a long waiting time before receiving definitive treatment may result in a reduced probability of overall benefit.

Identifying patients at high and low risk by using clinical predictors, along with monitoring for excessive waits and other inequities, could help in devising fair and practical solutions for allowing the health care system to provide flexibility for accommodating periodic surges in caseload, ultimately minimizing patient jeopardy. In conformity with the recommendation to collect outcome data on a regular basis, endorsed by the 23rd Bethesda Conference on access to cardiovascular care,¹ our experience suggests that in any health care system, there should be an ongoing triage of patients on waiting lists, with those at greatest risk being moved to the head of the line.