A multi-region assessment of population rates of cardiac catheterization and yield of high-risk coronary artery disease

We used the Alberta Provincial Project for Outcomes Assessment in Coronary Heart (APPROACH) disease database, which is a geographically defined, population-based registry that captures all patients undergoing cardiac catheterization in Alberta [10]. Corresponding cardiac catheterization data compiled through the partner cardiac registry initiatives in British Columbia and Nova Scotia were similarly obtained. The databases from each of the three Canadian provinces contain detailed clinical information, including demographic characteristics, co-morbidities and therapeutic interventions. Information on coronary anatomy is stored using a cardiac reconstruction software program (Heartview, Siemens Medical Systems). For the purposes of this analysis, high-risk CAD is defined as greater than 50% stenosis of the left-main coronary artery, or similar stenotic disease involving 3-vessels, or 2-vessels disease with involvement of the proximal left anterior descending artery [7]. Ethics approval was obtained from each respective academic institution ethics review board.

Alberta APPROACH data from Jan 1, 1995 to Dec 31, 2006 were used. Repeat procedures were excluded so that individual patients would only be included once in the analysis. The previous work reported data from 1995 to 2001 ("time period 1") [7]. Due to a health region boundary change in April, 2002, the data for time period 1 (1995-2001) were re-analyzed for this study using the Alberta lay-out of 9 health regions that existed in 2006 to allow for seamless comparison in yield of high-risk disease across the two time periods. Only 2 of Alberta's 9 health regions have cardiac catheterization and revascularization facilities. These regions provide invasive investigation and revascularization for the entire province.

The Nova Scotia and British Columbia cardiac registry initiatives provided corresponding data from 2002-2004 and 2000-2005, respectively. This represents the entire dataset within each province for which Heartview and catheterization data are available. The dataset for each province includes all health authorities within each province (i.e. complete geographically inclusive capture). Nova Scotia has 9 District Health Authorities with cardiac catheterization provided by 1 centre. British Columbia has 5 Health Authorities, with cardiac catheterization services offered in three urban centres.

Analyses were completed for males and females separately as the prevalence of coronary artery disease differs between males and females and because the optimal catheterization rate may differ between sexes. Patients were categorized into health regions based on their residential postal code. All catheterization procedures were included. For each region, the population rates of catheterization and high-risk detection were calculated. Cardiac catheterization rates were derived as the total number of catheterization procedures divided by the total population over the age of 20. Rates of high-risk CAD detected were calculated as the number of patients with high-risk CAD detected divided by the total population over the age of 20. Subsequently, using direct standardization, age-adjusted rates were calculated. Data were categorized into 5 age categories; 20-34, 35-49, 50-64, 65-74 and over 75. The number of cardiac catheterizations within each age group was divided by the population in that category. Rates of high-risk CAD were calculated in the same manner. Weighting the age-specific rates with the 1996 Canadian population, we thus obtained age-adjusted cardiac catheterization and high-risk detection rates per 100,000 population. The 1996 census was used as the population reference standard for age adjustment (by direct standardization) because this standard has been used in other notable procedure rate work [11], and comparability of rates across studies was desired.

Subsequent analyses used these adjusted catheterization rates and high-risk detection rates as data points. Initially, a scatter plot of the catheterization rate versus the high-risk detection rate was constructed, with each data point representing an individual year of data for a particular region. For the Alberta data, we compared more recent years of data with earlier years, in order to determine whether relationships persisted over time. For the other two provinces, the entire time period available was reported without distinction between years.

Given that the unit of analysis for this study is the health care region, rather than the individual patient, we employed a least squares linear regression analysis to assess the relationship between catheterization rate and high-risk CAD detection. As the observations from each region are not independent, we performed hierarchical modeling using a mixed effects linear model with a random effect for each region. In this analysis, we modeled a shared fixed intercept (fixed at zero) and random slopes. The results of this model were then used to plot a single weighted line to reflect the linear relationship between catheterization rate and high-risk CAD detection rate for all regions. The line of "best fit" with 95% confidence lines was then overlaid on the scatter plot of observed data. To explore if the relationship might be different for patients with acute coronary syndrome (ACS) or for those with stable coronary disease, we repeated the above analysis, stratified by indication.

Lastly, we tested quadratic terms in the modeling process for the data from each of the provinces studied. A statistically significant quadratic term would indicate evidence of a plateau in the yield of high-risk CAD associated with increased catheterization rates. This was performed separately for both time periods studied in the Alberta data, and repeated for Nova Scotia and British Columbia data.

Recognizing the complexity of the mixed effects modeling described above, we performed a sensitivity analysis using simpler ordinary least squares regression. This latter analysis, while simpler, overlooks the non-independence of some of the data points analyzed. The findings of this sensitivity analysis are not presented here but were similar to those of our hierarchical mixed effects analysis.

Table 1 presents the patient characteristics, intervention rates and outcomes by regional tertile of utilization in time period 1 and time period 2. The clinical characteristics and utilization are generally stable across tertiles and time periods. Table 2 shows the average population of males and females over 20 years of age, along with the crude and adjusted cardiac catheterization rates per 100,000 for each health region in Alberta in time periods 1 and 2. For males, the adjusted catheterization rates ranged from 379.7 to 538.4 per 100,000 in time period 1. In time period 2, the range is higher with slightly less variation -- i.e. population rates ranging from 438.9 to 587.8 per 100,000. There is an increase between time periods for both sexes, though females continued to have consistently lower rates than males.

The scatter plot of catheterization rates vs. high-risk rates and the results of the hierarchical modeling are presented in Figure 1. Each individual point represents one year of data for one health region. The results for 9 health regions analyzed in time period 1 (Panel A) confirm our previously reported findings for that period [7]. Time period 2 (Panel B), meanwhile, continues to show a positive association between cases of high-risk CAD detected and catheterization rates.

Describing the relationship quantitatively, for males in time periods 1 and 2, for every 2.3 and 2.6 catheterization procedures performed, respectively, an additional high-risk CAD patient was detected. For females, the detection rate is lower than males with an additional high-risk patient detected in every 3.7 and 4.3 catheterization procedures performed in time periods 1 and 2, respectively.

Table 3 presents the population, crude catheterization rate and adjusted catheterization rate for British Columbia and Nova Scotia. In general, the rates in both British Columbia and Nova Scotia are higher than those seen in Alberta, and as in Alberta, males have consistently higher rates than females.

The linear relationship between catheterization rate and high-risk yield is similarly demonstrated in both provinces, with findings that closely resemble those from Alberta (Figure 2). In British Columbia, every 2.4 catheterizations results in an additional high-risk case for males whereas for females, every 3.8 catheterizations results in an additional high-risk case. The detection rate is slightly higher in Nova Scotia with an additional high-risk case in every 2.1 catheterizations for males and every 3.5 procedures for females.

We formally assessed whether a plateau was evident in the high-risk CAD detection rate as catheterization rates increased. When using the Alberta data across both time points, the coefficient of the quadratic term entered into the regression model was not significant (p-value = 0.64). Similarly, there was no evidence of a significant quadratic term in either British Columbia (p-value = 0.67) or Nova Scotia (p-value = 0.61). This finding indicates that there is no statistical evidence of a plateau in yield of high-risk disease, in a range of data points extending to the highest catheterization rates seen in Nova Scotia of 850 procedures per 100,000 population.

A similar linear relationship was found for both ACS and non-ACS subgroups, and on formal testing for evidence of a plateau, there was again no evidence of a statistically significant plateau in the relationship between population rates of catheterization and yield of high risk disease (test for quadratic term indicating plateau: p-value = 0.43 for ACS and p-value = 0.167 for non-ACS) (Figure 3). As observed in the overall analysis, the yield of high-risk cases is lower in females than males for both subgroups and both time periods.