Comparison of the Use of Medical Resources and Outcomes in the Treatment of Aneurysmal Subarachnoid Hemorrhage Between Canada and the United States
Objective—Using data from a randomized trial of tirilazad mesylate, we assessed the differences between Canada and the United States in the use of medical resources and outcomes in the treatment of aneurysmal subarachnoid hemorrhage during the first 90 days after admission to the hospital.
Methods—Of the 877 patients for whom economic data were available, 194 were enrolled in Canada and 683 were enrolled in the United States. The differences between the countries in patient characteristics, use of medical resources, and outcomes were analyzed by comparing means and the 95% confidence intervals (CIs) around the differences in means. These differences also were predicted with use of multivariable regression analysis.
Results—The average hospital stay was 4.2 days longer (95% CI, 1.3 to 7.1 days) in Canada, but most of the extra stay was among patients admitted to the study in poor neurological condition. In general, however, hospital stays in Canada were substantially less intensive. Patients treated in Canada spent 3.7 fewer days (95% CI, 1.2 to 6.1 days) in nursing homes and rehabilitation centers than did patients in the United States. No statistically significant differences were seen for Glasgow Outcome Scale score, death, and occurrence of vasospasm.
Conclusions—For patients admitted to the study in good neurological condition, the apparent difference in length of stay between Canada and the United States was caused by a shift in the sites of formal care rather than to the length of this care. For those admitted in poor neurological condition, both the length and sites of care differed between the two countries. No significant difference in outcomes appeared to justify these differences in the use of medical resources.
Differences between the health care systems of Canada and the United States have received growing attention because Canada spends a smaller percentage of gross domestic product on health care and exerts greater government control over expenditure than does the United States.1 2 3 4 5 6 7 8 9 10 11 Much research has focused solely on variations in practice patterns, without simultaneous assessment of the outcomes of this variation.1 2 4 6 8 9 11 Recently, however, investigators have used data from two multinational clinical trials of drug therapies after acute myocardial infarction to assess treatment differences for specific clinical populations and to assess differences in the processes and outcomes of care.3 5 They found that hospitalization in the United States was longer and more intensive than that in Canada. In one study, these differences did not translate into differences in mortality or the rate of a second infarction,5 whereas in the other, after 1 year patients treated in the United States had a significantly higher survival and substantially greater improvement in functional status.3 The latter finding, however, was questioned in a series of letters to the editor.12 13 14 15
We used data from a randomized trial of tirilazad mesylate16 to assess differences between Canada and the United States in the use of medical resources and in clinical outcome in the treatment of aneurysmal subarachnoid hemorrhage (SAH) during the first 90 days after randomization. Studying aneurysmal SAH allowed us to determine whether the cross-national differences found in the studies of myocardial infarction can be generalized to another disease. Moreover, by including postdischarge use of medical resources in our cross-national comparisons, we gained a better understanding of the size and scope of the differences in treatment between the two countries over an extended period.
Subjects and Methods
The clinical trial was conducted from 1992 to 1994 in 54 centers located in the United States and Canada.16 Patients were randomly assigned to one of three treatment groups: vehicle (n=300) or tirilazad at 2.0 or 6.0 mg/kg of body weight per day for 8 to 10 days (n=298 and n=299, respectively). Treatment began within 48 hours of the development of SAH and ended 10 days after the hemorrhage occurred. The primary clinical outcome was mortality at day 76 after the first dose of study medication. Secondary outcomes included diagnosis of clinical vasospasm in the 14 days after randomization and functional status at 90 days (defined by the Glasgow Outcome Scale17 ). Both clinical data and data on use of medical resources for the first 90 days after admission to the hospital were collected prospectively during the trial.
The clinical results of this study showed that there was no difference in mortality or in symptomatic vasospasm between the treatment groups. When patients were stratified by gender and neurological grade on admission to the study, men with grades IV and V who received tirilazad at 6 mg/kg per day had an 85% reduction in the risk of death (5% versus 33% among patients who received placebo; P=.03). Full results have been reported elsewhere.16
Data on use of medical resources were available for 877 of the 897 patients who were enrolled in the trial and received at least one dose of treatment or vehicle. Of these 877 patients,194 were enrolled in Canada and 683 in the United States. All 20 patients who were excluded from the analysis because of insufficient data on use of medical resources were enrolled in the United States.
Data on the patients’ clinical and demographic characteristics were collected before randomization. These data included age, characteristics of the aneurysm, blood pressure, pulse, days between onset of SAH and randomization, sex, race, neurological grade (range, I through V, with V the most severe), and distribution of vasospasm. Data were also collected on the size of the hospital (measured by number of beds) in which the patient was treated.
For the initial hospitalization, use of a selected set of resources was recorded daily for the first 21 days and weekly for all days after the first 3 weeks. Data elements included total length of stay subdivided by days in intensive care, step-down care, routine care, and other units; number of surgical procedures by type of procedure; drugs; days on which the patients received hypertension, hypervolemia, and hemodilution therapy; use of albumin and volume expanders; days on which a Swan-Ganz catheter was in place; number of CT scans, transcranial Doppler ultrasonograms, and radiographs; and number of chemistry panels, ECGs, and lumbar punctures.
Many surgical procedures were performed during this trial. We analyzed aneurysm clipping and shunt placement separately because these procedures were the most common (performed in 842 and 148 patients, respectively). We also summarized total use of surgery in the two countries by converting the number of surgical procedures to a common relative value unit scale by use of the workload and practice expense relative value units from the Medicare Fee Schedule.18
Data were also collected on days spent in nonhospital facilities (nursing homes and rehabilitation centers), days spent in the hospital because of readmission, and days spent at home.
The clinical outcomes used in this analysis were the occurrence of vasospasm by day 14 after randomization, death during the initial hospitalization, death during the first 90 days after randomization, and the Glasgow Outcome Scale score 90 days after randomization. Among the 877 patients included in the study, 9 (2 treated in Canada and 7 treated in the United States) had missing values for the Glasgow Outcome Scale score at 90 days; these patients were excluded from the evaluation of this outcome.
Patient Characteristics and Use of Medical Resources
We report means and standard deviations (continuous variables) and percentages (categorical variables) for a selected set of patient characteristics by country. We also report the difference in means and percentages between the two countries and, where applicable, the 95% confidence intervals (CIs) around the difference.
We report percentages for the four clinical outcome variables by country. We also report differences between the two countries and, where applicable, 95% CIs around the differences.
Accounting for Severity of Illness
Although data used in our study were derived from a randomized controlled trial, randomization was by treatment and not by where patients received their care. Thus, some of the differences that were observed between the two countries may have been due to differences in the patients’ underlying severity of illness.12 15 To limit the potential for such bias, we used multivariate regression analysis to reevaluate differences between Canadian and US use of a selected set of resources (for the initial hospitalization, total length of stay, stays in the intensive care, step-down care, and routine care units, and surgery relative value units; after discharge from the initial hospitalization, days in nursing homes or rehabilitation centers). We also reevaluated differences between Canada and the United States in the occurrence of vasospasm by day 14 and death during the first 90 days after randomization. These analyses controlled for potential differences in severity of illness among the patients.
We used ordinary least squares regression for the prediction of total length of stay, intensive care and routine care unit stays, and for surgery relative value units. Due to the lower frequency of their use, we used a two-stage procedure to predict step-down care unit and nursing home and rehabilitation center stays, first predicting the probability of any stay and then predicting the length of stay among those who used these services. For the prediction of vasospasm and death, we used logistic regression.
The explanatory variables for these regressions included the country in which the patient was treated, age, a diagnosis of hypertension prior to SAH, pulse rate at randomization, characteristics of the SAH as diagnosed from the admission CT scan (the presence of left to right or right to left shifts of the midline structures, the presence of localized or generalized brain swelling, and evidence of enlarged lateral ventricles), a borderline or abnormal finding on the admission ECG, prior history of vasospasm, neurological grade on admission to the study, location of the aneurysm, and the order in which the patient was enrolled at the center where he/she received care (a proxy for experience). These variables were selected because they were highly correlated with the measures of use of resources and outcome.
Variables that were considered for the analysis but which were not used because they had insufficient independent predictive ability included treatment group, gender, race, handedness, smoking behavior, history of prior stroke, the number of abnormal items on the medical history, the number of aneurysms, abnormalities on the admission chest roentgenogram, employment status, blood pressure, and hospital size.
In addition to evaluating whether our results were due in part to differences in the severity of illness of patients treated in the two countries, we evaluated whether the relative differences in the use of medical resources and outcome observed between Canada and the United States were similar for patients admitted to the study in good neurological condition (grades I through III) and patients admitted in poor neurological condition (grades IV through V). Neurological status on admission to the study was assessed with a modified version of the World Federation of Neurological Surgeons grading scale.16
Differences between the countries were considered to be statistically significant if the 95% CIs around the differences excluded zero. For the univariate comparisons, we tested whether these statistical results were sensitive to this parametric test of significance by use of the Wilcoxon rank sum test for continuous variables and the Fisher exact test for categorical variables. For the ordinary least squares regressions, we tested the sensitivity of the parametric specification of the error terms by using least absolute value models that predicted the median rather than the mean for the dependent variables.19 We note cases in which the detection of significance differs between the parametric and nonparametric methods.
For the evaluation of differences by neurological condition on entry to the study, we compared the coefficients for the differences between Canada and the United States with an F test.
For the ordinary least squares regressions, we report R2 values as the measure of the predictive ability of the models; for logistic regressions, we report the concordance index (which equals the area under the receiver operating characteristic curve20 ). We did not adjust our tests of significance to account for the multiplicity of comparisons (ie, we were more likely to detect differences between Canada and the United States in use of medical resources and outcomes than we would have been had we made this adjustment).
Table 1⇓ shows patient characteristics by country measured when the patients were assigned to one of the three treatment groups. Patients in Canada were more likely to be white (absolute difference, 17.8%) and on average had 0.4 fewer aneurysms; a lower diastolic blood pressure (−3.3 mm Hg); a slower pulse rate (5 fewer beats per minute); 0.1 more days between onset of SAH and randomization; and treatment in hospitals that had 171 fewer beds. The neurological grade did not significantly differ between patients treated in Canada and those treated in the United States, nor were there differences in the clot thickness as measured on the admission CT scan.
Table 2⇓ shows use of medical resources after randomization during the initial hospitalization in Canada and the United States. The average length of stay in Canada was 4.8 days longer (95% CI, 1.3 to 8.3 days) than the average length of stay in the United States. In general, however, hospital stays in Canada were substantially less intensive. On average, patients treated in Canada spent 6.6 fewer days in intensive care units (95% CI, −8.0 to −5.2 days), had a lesser amount of surgery (as indicated by fewer relative value units per patient [−6.9; 95% CI, −12.7 to −1.1]), received 4.4 fewer types of drugs(95% CI, −5.7 to −3.0), and had lower rates of diagnostic testing. In all but one case (total surgery relative value units), the results of the parametric statistical tests were confirmed by the Wilcoxon rank sum test.
Table 3⇓ reports the average number of days of care by site in the first 90 days after SAH. Patients treated in Canada spent 4.1 fewer days (95% CI, 1.7 to 6.5 days) in nursing homes and rehabilitation centers than did patients treated in the United States. This reduction almost completely offset the increased length of stay for the initial hospitalization observed among patients treated in Canada.
Table 4⇓ reports outcomes among patients treated in Canada and those treated in the United States. None of the four outcomes statistically significantly differed between the two countries.
Accounting for Severity of Illness
Table 5⇓ reports the differences in the use of medical resources and absolute differences in the probability of the clinical outcomes between Canada and the United States estimated from the regressions that predicted lengths of stay; the number of relative value units for surgery; death at 90 days; and the presence of vasospasm within 14 days after randomization.
Although a number of the characteristics of the patients treated in Canada and those treated in the United States differed, they had little effect on the observed differences in the use of medical resources or clinical outcomes. After adjustment using multivariable analysis, differences in length of stay in the intensive care, step-down, and routine care units as well as in nursing home and rehabilitation centers were similar to the differences identified in the univariate analyses. The difference in the number of surgical relative value units was also similar. In addition, the small nonsignificant difference for death at 90 days that was observed in the univariate analysis became even smaller (absolute difference, −0.5%, 95% CI, −5.0% to 5.3%). A slightly larger difference was seen for the occurrence of vasospasm at day 14 (absolute difference, −2.5%, 95% CI, –12.9% to 6.6%).
As with the univariate analyses, in all cases except for surgery relative value units the results of the median regression were similar to those for the ordinary least squares regression predicting the mean values.
Table 5⇑ also reports these same analyses repeated for patients who were admitted into the study in good neurological condition (grades I through III) and those who were admitted in poor neurological condition (grades IV through V). These subanalyses indicated that differences between Canada and the United States in the use of intensive care units, surgery, and nursing home and rehabilitation centers were similar for patients admitted to the study in good versus poor neurological condition (P=.27, P=.62, and P=.36 respectively; probability values not shown in the table).
The relative use of step-down and routine care units in Canada and the United States, on the other hand, differed between these two patient groups. Patients admitted to the study in good neurological condition spent 2.7 more days in step-down care units and 6.5 more days in routine care units when treated in Canada compared with when they were treated in the United States. Patients admitted to the study in poor neurological condition spent 2.7 fewer days in step-down care units and 15.9 extra days in routine care units when treated in Canada compared with when they were treated in the United States (P=.003 and P=.001, respectively; probability values not shown in the table).
Overall stay during the initial hospitalization tended to be slightly longer for patients admitted in good neurological condition in Canada compared with those admitted in the United States (P=.06), but the length of the episode of care (including days in nursing homes and rehabilitation centers) was similar in the two countries (P=.46; data not shown). Stay during the initial hospital was substantially longer for patients admitted in poor neurological condition in Canada compared with those in the United States (P=.01). The decreased use of nursing homes and rehabilitation centers reduced but did not fully offset the differences in the length of the treatment episode.
While our subanalysis of the clinical outcomes had little power, we were unable to detect relative differences between Canada and the United States in the clinical outcomes of patients admitted to the study in good versus poor neurological condition (P=.36 for being alive at 90 days and P=.91 for vasospasm by day 14; probability values not shown in the table).
We found that use of medical resources in the treatment of patients with aneurysmal SAH differed substantially between Canada and the United States. On average, the initial hospitalization was 4.7 days shorter for patients treated in the United States compared with those treated in Canada but involved 6.6 more days of intensive care. After discharge for the initial hospitalization, patients treated in the United States spent 3.7 more days in nursing homes and rehabilitation centers than did those treated in Canada. The absolute difference in the severity-adjusted probability of survival at 3 months was −0.5% (Canada−United States) and was not statistically significant.
We also found that the differences in the use of intensive care units observed between Canada and the United States were similar for patients admitted to the study in good as well as poor neurological condition, but the differences in the use of step-down and routine care units observed between Canada and the United States were not the same for these two patient populations. Patients in good neurological condition in the two countries had similar overall lengths of stay in the hospital, but Canadian physicians substituted days in step-down and routine care units for days in intensive care unit. Patients in poor neurological condition who were treated in Canada had substantially longer stays in the hospital than those treated in the United States, and substantially more of their stays were in routine care units.
Our findings generally confirm and expand on the findings from previous comparisons of use and outcome of health care between Canada and the United States (these comparisons focused on care after myocardial infarction). As did Rouleau and colleagues,5 we found that patients treated in Canada had longer hospital stays. Like Rouleau and colleagues, we found no significant difference in outcomes to justify the additional use of medical resources in the United States.
While the studies of treatment after myocardial infarction3 5 did not comment on follow-up institutionalization, we found that more patients in the United States received follow-up care in nursing homes and rehabilitation centers than those in Canada. For patients admitted to the study in good neurological condition, the length of the episode of care was similar in the two countries, but differed in the sites of care. In Canada more days were spent in step-down care and routine care units during the initial hospital stay; in the United States more days were spent in intensive care units during the initial hospital stay and in nursing homes and in rehabilitation centers after the discharge from the hospital. For patients admitted to the study in poor neurological condition, the episode of care for those treated in Canada was substantially longer than for those treated in the United States, but was marked by fewer days in intensive care units. These findings appear to confirm prior observations that the US hospital payment system, which is based on diagnosis related groups, may provide incentives for earlier discharge from the hospital and more use of nursing homes and rehabilitation centers after discharge.21 22 23
Hospital bills were available for most of the patients treated in the United States but were unavailable for all patients treated in Canada. Had we predicted the latter bills using multivariable analysis that associated the resource use observed in the trial with the total bills, and had we also adjusted for differences between US costs and charges and between Canadian and US unit costs, we would have found that on average, patients treated in Canada cost $15 300 (95% CI, $9800 to $20 900) less than those treated in the United States. The difference in costs would have been greater among patients admitted in good neurological condition ($17 300, 95% CI, $12 200 to $22 300) and smaller among patients admitted in poor neurological condition ($9800, 95% CI, $3300 to $16 200).
Our study is one of the few multinational comparison studies that controlled for observed differences in severity of illness among the patients treated in the different countries. Our univariate analysis of patient characteristics indicated that study investigators in the two countries may have enrolled different types of patients in the trial (based on race, number of aneurysms, etc). Nevertheless, our multivariable analyses showed that adjustment for measured disease severity did not substantially affect our results.
Our study has several limitations. First, our analysis (along with the other two that have compared use of medical resources and outcomes3 5 ) was based on data from a large clinical trial rather than on data from observation of the usual care of patients. Although the trial itself did not impose major restrictions on the usual care of these patients, it may have resulted in more monitoring and longer stays than would have occurred had the patients not been enrolled in a randomized trial. On the other hand, the investigators in the clinical study were carefully selected and may have been more efficient than other providers in the community.
Second, although our results may accurately reflect differences in use of medical resources among larger hospitals with substantial neurosurgical capabilities in Canada and the United States, it is not clear that the magnitude of differences we observed accurately reflects the magnitude that might exist between community hospitals in the two countries. For example, Taylor et al,24 in a recent study of care after SAH in the United States, reported that among older patients with SAH only 18.9% were treated surgically (compared with over 90% of patients in our study who were surgically treated). This difference in the rates of surgery was unlikely to have resulted from differences in the ages of the populations in the two studies (as with the overall sample, among the 84 patients in our study who were aged 70 and above, over 90% were treated surgically). It may instead have been due to the fact that the enrolling physicians in our study generally were neurosurgeons.
In conclusion, our study highlights the potential, based on practices observed in Canada and the United States, for physicians in both countries to maintain their standard of care while potentially reducing the use of medical resources. Patients and physicians in both countries stand to benefit from comparative studies such as ours that identify use and outcomes for relatively homogeneous groups of patients with well-defined clinical characteristics.
Supported in part by a grant from Pharmacia and Upjohn Inc, Kalamazoo, Mich. The authors thank Sankey Williams, MD, who read a prior version of the manuscript.
Reprint requests to Henry Glick, University of Pennsylvania, Division of General Internal Medicine, Room 312A, 3615 Chestnut St, Philadelphia, PA 19104.
Richard Willke is an employee of Pharmacia and Upjohn.
- Received September 17, 1997.
- Revision received November 18, 1997.
- Accepted November 18, 1997.
- Copyright © 1998 by American Heart Association
Mark DB, Naylor CD, Hlatky MA, Califf RM, Topol EJ, Granger CB, Knight JD, Nelson CL, Lee KL, Clapp-Channing NE, Sutherland W, Pilote L, Armstrong PW. Use of medical resources and quality of life after acute myocardial infarction in Canada and the United States. N Engl J Med. 1994;331:1130–1135.
Redelmeier DA, Fuchs VR. Hospital expenditures in the United States and Canada. N Engl J Med. 1993;338:772–778.
Rouleau JL, Moye LA, Pfeffer MA, Arnold JM, Bernstein V, Cuddy TE, Dagenais GR, Geltman EM, Goldman S, Gordon D, Hamm P, Klein M, Lamas GA, McCans J, McEwan P, Menapage FJ, Parker JO, Sestier F, Sussex B, Braunwald E, for the SAVE Investigators. A comparison of management patterns after acute myocardial infarction in Canada and the United States. N Engl J Med. 1993;328:779–784.
Zwanziger J, Anderson GM, Haber SG, Thorpe KE, Newhouse JP. Comparison of hospital costs in California, New York, and Canada. Health Aff (Milwood). 1993;12:130–139.
Kessler RC, Frank RG, Edlund M, Katz SJ, Lin E, Leaf P. Differences in the use of psychiatric outpatient services between the United States and Ontario. 1997;336:551–557.
Deckelbalm LI, Quality of life after myocardial infarction: Canada versus the United States. N Engl J Med. 1995;332:470–471. Letter.
Vajo Z. Quality of life after myocardial infarction: Canada versus the United States. N Engl J Med. 1995;332:470. Letter.
Greenfield S, Kaplan S. Quality of life after myocardial infarction: Canada versus the United States. N Engl J Med. 1995;332:471. Letter.
US Department of Health and Human Services, Health Care Financing Administration. Medicare program: fee schedule for physicians’ services: final rule. Federal Register (Addendum B). November 25, 1991;56:59502–59811.
Stata Corporation. Stata Reference Manual, Release 5. Vol 3. College Station, Tex: Stata Press; 1997:94–104.