Processes and Outcomes of Care for Diabetic Acute Myocardial Infarction Patients in Ontario: Do physicians undertreat?

We obtained information from the Ontario Myocardial Infarction Database (OMID), which links together a variety of administrative databases. The accuracy of AMI coding in OMID as well as complete details of the rationale and inclusion-exclusion criteria have been reported previously (11). The cohort consisted of all patients admitted to a hospital with a diagnosis of AMI, code 410 of the ICD-9, between 1 April 1992 and 31 December 1993 (inclusive). The time period was specifically intended for examination of processes of care and their impact on long-term outcomes after AMI. Each patient was tracked for a minimum of 5 years. This cohort recently has been used to demonstrate the long-term process-outcome relationship across sociodemographic and hospital subgroups throughout Ontario (9,12).

Diabetic patients were identified using the Ontario Diabetes Database (ODD) (1). The ODD is an administrative data-derived ongoing registry of all individuals in the province with diagnosed diabetes. Cases are identified on the basis of diagnosis codes recorded in hospital discharge abstracts and physicians’ service claims. Cases of gestational diabetes (diabetes-defining administrative data records occurring within 4 months of an obstetrical event) were removed. Persons who enter the ODD remain until death or migration out of the province, as recorded in the provincial registry, even if they have no diabetes-related claims in subsequent years. Validation against data abstracted from primary care charts has confirmed a sensitivity of 86% and specificity of >97% for identifying patients with diabetes using ODD (1). We further categorized patients with diabetes into two groups: those with complications and those without. Diabetes with complications was identified using ICD-9 codes 250.1–250.9 (inclusive). These codes included metabolic complications (ketoacidosis, hyperosmolarity syndromes) and microvascular and peripheral vascular disease but did not specifically include cardiac complications.

Baseline patient characteristics including age, sex, socioeconomic status, and disease severity were obtained from the hospital discharge abstracts of the index AMI admission. Socioeconomic status was derived using ecological income measures (i.e., median neighborhood household income) from 1991 official Canadian census data. Measures of illness severity incorporated variables included in the Ontario AMI mortality prediction rule for 30-day and 1-year mortality (11,13). In addition to diabetes (with and without complications), clinical variables included the following: shock, congestive heart failure, pulmonary edema, cardiac arrhythmia, stroke, malignancy, acute renal failure, and chronic renal failure. The clinical prediction rule has been validated against several population-based AMI cohorts illustrating good predictive accuracy (areas under the receiver operating curve of 0.77 in validation datasets) (13).

Attending physician specialty was identified using hospital discharge abstracts (9,12). Patients were categorized according to the type of facilities available at the admitting institution (on-site versus no on-site angiography and revascularization facilities), regardless of whether they were subsequently transferred to another hospital. Patients at six hospitals that were on-site catheterization-only facilities were excluded from the analysis because of the small sample size (3.5% of the study cohort) (12).

Previous work by our group and others has demonstrated a positive relationship between resource intensity and long-term AMI outcomes, for selected treatments when applied within the first several months that follow the AMI presentation (12). Accordingly, we concentrated on care provided to patients within the first several months after discharge from the AMI admission during the early postinfarct time intervals. Processes of care variables examined included the use of coronary angiography and myocardial revascularization within the first 6 months and 1 year after AMI, respectively (to allow for the appropriate interval of time for patient risk stratification and waiting lists), the provision of follow-up care, and the prescribing of evidence-based pharmacotherapies within the first 90 days after AMI discharge. The importance of such time intervals in relation to long-term outcomes in Ontario has been previously demonstrated (12). Information pertaining to invasive cardiac procedure use was obtained using both hospital discharge abstracts from the Canadian Institute of Health Information (CIHI) and physician claims data from the Ontario Health Insurance Plan (OHIP) databases; information pertaining to physician prescribing patterns was obtained from the Ontario Drug benefits database (available only for those patients ages 65 years and older), whereas data on outpatient physician visits was obtained from OHIP.

Outcomes included mortality and recurrent emergent cardiac hospitalization rates at 5 years after AMI: i.e., recurrent AMI (ICD-9 code 410), congestive heart failure (ICD-9 code 428), unstable angina (ICD-9 codes 411, 413). Emergency visits were identified using physician claims data. In total, 96.2% of readmissions were accompanied by an emergency visit. Accordingly, recurrent cardiac hospitalizations reflected urgent rather than elective indications for the vast majority of cases (12).

We compared the unadjusted baseline characteristics, processes of care, and outcomes between diabetic and nondiabetic patients using χ² test of proportions and Student’s t test where appropriate. Multivariable analysis adjusted for age, sex, socioeconomic status, illness severity (predicted probability of 30-day mortality), and attending physician and hospital characteristics using Cox proportional hazards model. All multivariable models were constructed in a similar fashion, incorporating backward stepwise regression techniques and then comparing the −2 log likelihoods of the Cox proportional hazards model. Death was used as the main censoring variable when assessing nonfatal outcomes at 5 years. In our statistical models, congestive heart failure was examined using a four-level categorical variable: those without congestive heart failure, those with congestive heart failure (but without pulmonary edema or shock), those with pulmonary edema (but without shock), and those with shock. Similarly, statistical models incorporated diabetes as a three-level variable: those without diabetes, those with diabetes (but without complications), and those with diabetes with complications. We also reanalyzed each model by combining the two subgroups of diabetes into one category (diabetes with or without complications) to examine its overall association with outcomes.

We undertook a secondary analysis to examine the extent to which outcome differences between diabetic and nondiabetic elderly patients could be explained by intergroup variations in the intensity and follow-up care. Specifically, we determined the magnitude of association between diabetes and outcomes, before and after adjustments for evidence-based therapies, outpatient follow-up care, and revascularization procedures. This secondary analysis was confined to the elderly population because information on the use of evidence-based pharmacotherapies was unavailable for patients <65 years of age. Statistical significance was defined as P < 0.05 in all analyses. SAS statistical software package (version 8.2; SAS Institute, Cary, NC) was used.

The baseline patient, attending physician, and hospital characteristics are shown in Table 1. The study cohort consisted of 25,697 patients, 6,052 (23.6%) of whom were diabetic. Diabetic patients were more likely to be older, women, less affluent, and sicker than nondiabetic patients. There were no significant interphysician or interhospital differences between patients with and without diabetes.

The relationship between diabetes and processes of care after AMI are shown in Table 2. Despite their higher risk profile, diabetic patients were 13.3% less likely to be followed by a cardiologist (22.2 vs. 25.6%, P < 0.001), 17.8% less likely to receive coronary angiography (20.3 vs. 24.7%, P < 0.001), and 10.2% less likely to receive myocardial revascularization (12.6 vs. 14.9%, P < 0.001) than were nondiabetic patients. Once coronary angiography had been undertaken, patients with diabetes were 21.4% more likely to receive bypass surgery (36.9 vs. 30.4%, P < 0.001), but 25.9% less likely to receive angioplasty (19.5 vs. 26.3%, P < 0.001) compared with patients without diabetes. With regard to medical therapy, the use of evidence-based therapies was extremely low in both diabetic and nondiabetic patients. However, elderly diabetic patients were 18.2% less likely to receive a statin (3.6 vs. 4.4%, P = 0.06), 22.3% less likely to receive β-blockers (34.2 vs. 44.0%, P < 0.001), 6.0% less likely to receive aspirin (59.7 vs. 63.5%, P < 0.001), but 24.4% more likely to receive ACE inhibitors (44.8 vs. 36.0%, P < 0.001) than nondiabetic elderly patients.

Although there were no significant differences in waiting times to invasive coronary interventions or cardiology outpatient visits between the two groups after AMI, patients with diabetes generally waited 1 day less for internal medicine or primary care outpatient visits.

Table 3 demonstrates that diabetes was associated with significantly higher rates of fatal and nonfatal adverse outcomes compared with nondiabetes. Nonfatal outcome differences were driven predominantly by higher rates of recurrent AMI and congestive heart failure admissions. Intergroup differences in cardiac readmission rates resulted in an average increase of 4.1 days spent in hospital per diabetic patients over the 5-year follow-up (27.6 vs. 23.5 days, P < 0.001).

After adjusting for baseline sociodemographic, clinical, attending physician, and hospital factors, diabetes was a significant independent predictor of long-term morbidity and mortality. Moreover, the magnitudes of association between diabetes (with or without complications) and long-term adverse outcomes were large, ranging from as low as 26% for emergency consultation (i.e., adjusted hazard ratio 1.26, 95% CI 1.22–1.31) to as high as 109% for congestive heart failure hospitalization (2.09, 1.95–2.24). Diabetes (with or without complications) was associated with a 52% higher risk of any cardiac readmission (1.52, 1.45–1.59) and a 54% higher risk of mortality at 5 years (1.57, 1.50–1.63) (Table 4). Apart from age and congestive heart failure (with or without pulmonary edema), diabetes was the strongest predictor of long-term mortality and morbidity after AMI (Table 5).

We examined whether variations in resource intensity, cardiac procedure use, and evidence-based therapies explain outcome differences between patients with and without diabetes. Figure 1 illustrates that variations in resource intensity, cardiac procedure use, and evidence-based therapies exerted only modest impact on explaining the association between diabetes and nonfatal outcomes in elderly patients. Variations in available process measures could not account for long-term AMI mortality differences between diabetic and nondiabetic elderly patients. Furthermore, among those patients receiving coronary interventions post MI, there was no significant interaction between diabetic outcomes and revascularization modality. Specifically, diabetic patients undergoing percutaneous transluminal coronary angioplasty had similar long-term outcomes as those undergoing coronary artery bypass grafting (20.7 vs. 22.5, percutaneous transluminal coronary angioplasty versus coronary artery bypass grafting, respectively; P = 0.60).

Our study provides a comprehensive assessment of health services and outcome experiences of persons with diabetes who have had an AMI in Ontario. As compared with nondiabetic patients, individuals with diabetes were more likely to be significantly older, women, less affluent, and sicker at presentation. Not surprisingly, diabetes was an important independent predictor of long-term mortality and morbidity after AMI. Despite their higher baseline risk, patients with diabetes were generally managed less aggressively than patients without diabetes. Nonetheless, intergroup process differences only marginally accounted for variations in outcomes between patients with and without diabetes.

Our findings are consistent with the large body of evidence already demonstrating the important independent deleterious impact of diabetes on outcomes for patients presenting with AMI (5–7,10). For example, among elderly Medicare patients in Michigan, 30-day mortality rates were 21 and 17% for patients with and without diabetes, respectively (10). In an international study consisting of 8,013 patients with acute coronary syndrome enrolled into the Organization to Assess Strategies for Ischemic Syndromes (OASIS) Registry, diabetes was associated with a 1.57-fold risk of long-term mortality after adjusting for baseline clinical and treatment factors, an effect similar in magnitude across many countries (Australia, Brazil, Canada, the U.S., Hungary, and Poland) and one that closely resembles the effect size observed in our own study (5).

Although mortality has served as the focus for many of the diabetes outcome studies, relatively few studies have compared the long-term morbidity of patients with and without diabetes after AMI. We demonstrated that diabetic patients experienced significantly higher recurrent cardiac readmission rates, driven predominantly by congestive heart failure and AMI hospitalizations. The prevalence of heart failure and reinfarction among diabetic patients also supports one widely held hypothesis that attributes excess mortality to a higher susceptibility of myocardial pump failure and reinfarction among patients with diabetes (4). Nonfatal outcome differences translated into 410 additional days in hospital per 100 AMI patients, thus highlighting the significant morbidity and economic burden associated with diabetes in Ontario.

The diminished intensity of treatment provided to diabetic patients observed in our study is also consistent with the results of studies published elsewhere (5,7,10). With two notable exceptions (the higher utilization rates of ACE inhibitors and the preferential use of bypass surgery over angioplasty among those patients receiving angiography), diabetic patients were managed less aggressively in relation to specialty cardiac follow-up care, evidence-based pharmacotherapies, and therapeutic interventions after AMI than were their nondiabetic counterparts. The higher use of ACE inhibitors among diabetic patients likely appropriately reflected their higher prevalence of congestive heart failure and diabetic nephropathy (14). Similarly, the predilection toward bypass surgery over angioplasty (among patients referred for angiography) may have appropriately reflected their higher prevalence of underlying multivessel coronary artery disease (7).

If we are to assume that absolute benefits will always be greatest for those patients at highest baseline risk (8), risk-benefit tradeoffs should have resulted theoretically in more aggressive management for patients with diabetes, as compared with patients without diabetes. However, we observed otherwise. The inverse relationship between physician treatment intensity and baseline patient risk, while seemingly counterintuitive, is exemplified frequently in cardiovascular health services research. For example, available evidence has consistently demonstrated the presence of socioeconomic and age-related treatment disparities, whereby physician aggressiveness in the management of cardiovascular disease is disproportionately targeted toward younger, more affluent, and healthier patients (9,15,16).

Although the precise reasons for the paradoxical relationship between baseline risk and physician intensity in the management of patients with diabetes are unknown, several hypotheses exist. First, available evidence suggests that physicians perceive differences in the efficacy or responsiveness of evidence-based therapies and interventions when the number of chronic comorbid conditions increases (17). Uncertainties arising over the expectations of prognostic benefits incurred for patients with chronic disease, coupled with a clinician’s aversion toward doing harm may sway physician decision-making processes in favor of more conservative approaches when managing patients with diabetes (18,19). Second, treatment decisions may be heavily influenced by the intrinsic attitudes or biases of physicians or their patients. For example, in a survey eliciting practice patterns and attitudes of physicians who cared for patients with renal disease, most physicians acknowledged that dialysis services in Ontario were being implicitly rationed based on such factors as short life expectancy, poor quality of life, patient preferences, and age (20). Third, available evidence suggests that clinicians who are caring for patients with one chronic disease are less attentive to the treatment necessities required for managing other concurrent diseases (21).

Admittedly, our study demonstrated that variations in treatment intensity only marginally accounted for nonfatal outcome differences between diabetic and nondiabetic elderly patients; however, available evidence has demonstrated that the relative prognostic benefits associated with evidence-based therapies when applied to patients with diabetes are similar to, if not greater than, those experienced by patients without diabetes (22). Accordingly, one might reasonably hypothesize that the magnitude of outcome differences between diabetic and nondiabetic patients observed in our study might have been further attenuated if physicians had been more aggressive at managing diabetic patients post MI. In summary, our study strongly suggests that physicians undertreat diabetic patients in Ontario after AMI.

There are several limitations to our study. First, we incorporated administrative data, which lack clinical details (e.g., heart rate, blood pressure, infarct location, prior coronary revascularization, risk factors) required to both characterize the initial presentation of the myocardial infarction and to predict outcome. Furthermore, information pertaining to contraindications was unavailable. Second, only a few therapies with proven lifesaving interventions were examined. Third, our cohort is 10 years old, which likely contributed to the overall low utilization rates of evidence-based therapies observed in this study. Contemporary management patterns for diabetic and nondiabetic care in Ontario may have changed. Indeed, available evidence has suggested that the use of evidence-based pharmacotherapies and cardiac interventions have increased over time among patients with and without diabetes (23). Nonetheless, evidence originating from other populations conducted more recently (and in regions other than Ontario) has similarly demonstrated that patients with diabetes are managed less aggressively with regard to evidence-based cardiovascular therapies and specialized cardiovascular interventions than patients without diabetes (10). Furthermore, these limitations must also be balanced against the comprehensiveness of our sample, which is highly representative of the Canadian population.

In conclusion, our findings suggest that diabetic AMI patients in Ontario have more extensive comorbidity, receive fewer evidence-based therapies, and fare far worse than nondiabetic AMI patients. More aggressive cardiovascular treatment initiatives may significantly mitigate the poor outcomes associated with diabetes. Given the high prevalence of diabetes, the wider implementation of aggressive management strategies specifically focused on cardiovascular risk modification applied to patients with diabetes may significantly translate into demonstrable outcome differences for the entire AMI population.

This project was supported by an operating grant from the Canadian Institutes of Health Research. The Institute for Clinical Evaluative Sciences is supported, in part, by a grant from the Ontario Ministry of Health. D.A.A. is a new investigator of the Canadian Institutes of Health Research and the Heart and Stroke Foundation of Canada. J.E.H. is a career scientist of the Ontario Ministry of Health and Long Term Care. J.V.T. is supported by a Canada Research Chair in Health Services Research.