We first examine graphs of the measures of pharmacy accessibility in each province, presented in Figs. 1 and 2. For completeness, the data are also reported in Tables 1 and 2. The top panel in Fig. 1 shows the provincial variation in pharmacy density. This density varies from 2.0 per 10,000 population in Newfoundland to 3.3 per 10,000 population in Saskatchewan. Accessibility, however, depends not only on the number of pharmacies but also the hours each pharmacy is open. The bottom panel of Fig. 1 shows the median and the 90th percentile of the distribution of pharmacy level weekly operating hours. Again, there is marked variation across provinces in the median operating hours per week – from a low of 53 hours in Manitoba to a high of 75 in Quebec. While the median hours per pharmacy is lowest in Manitoba, the 90th percentile is among the highest. This indicates that there is a contingent of Manitoba pharmacies that are open for as long as pharmacies operating in other provinces.
Figure 2 focuses on pharmacy accessibility during weekends. The top panel shows the median and the 90th percentile of the distribution of pharmacy level weekend operating hours. Again, there is marked interprovincial variation in median hours, with the lowest values observed in Manitoba and Ontario and the highest values observed in Quebec. There is less variation in the 90th percentile of pharmacy weekend operating hours. The bottom panel of Fig. 2 shows the fraction of pharmacies open on Saturdays and on Sundays. The fraction of pharmacies open on these days is the highest in Quebec. This could reflect the fact that the density of pharmacies in Quebec is relatively low so that it makes business sense to remain open on weekends.
We next turn to the analyses of pharmacy accessibility at the FSA level. We begin with Fig. 3, which shows two plots. The left-hand side is a plot of the number of pharmacies per 10,000 population and median total household income in the non-rural residential FSA regions across the 10 provinces of the FSA. The right-hand side of the Figure plots the total pharmacy operating hours and income in each non-rural FSA. A lowess curve (a local average of the y-axis variable evaluated at each income value) is superimposed in each of the plots. The figures reveal that pharmacy density is higher, the lower is the regional median household income. There is however substantial variation in pharmacy density and hours for a given level of household income.
In our dataset, the median FSA-level household income ranged from $23,317 to $216,260. The average was $75,378 and the median was $70,997. The age 65 + share of the FSA population varied from 2–44% and had an average value of 18%. The FSA land area varied from 8 km2 to 483,227 km2; the median was 92 km2.
Next, we report estimates of the linear regression models for each of the three measures of pharmacy availability in the FSA, namely the number of pharmacies, the total weekly operating hours and the number of weekend operating hours, all expressed per 10,000 population. We model how each of these outcome variables depends on median household income, the percentage of the population 65 and older, a variable that indicates whether the FSA is rural or non-rural, and indicators for each of the provinces. Ontario is the reference group. To allow for non-linearities in the relationship between income and community pharmacy availability, we used indicators for each the 10 deciles of median household income. The first decile (the 10% of FSAs with the lowest median household income) formed the reference group.
The parameter estimates (and associated 95% confidence intervals) for the model of the number of pharmacies per 10,000 population in the residential FSA regions are illustrated in Fig. 4. (Numerical parameter estimates for this and the other linear regression models appear in the Appendix.) The estimates reveal the following: First, holding constant other factors, pharmacy density is higher, the lower is the regional median household income. Indeed, there are about 2.5 fewer pharmacies per 10,000 population in the most affluent regions compared to the least affluent regions. This corroborates the results of Fig. 3. Second, there are more pharmacies per capita in regions with a larger elderly population share. Each percentage increase in the elderly share of the FSA population results in a 0.03 increase in the number of pharmacies per 10,000 population. Third, pharmacy density is lower in rural regions (about 1 fewer pharmacy per 10,000 population). Finally, pharmacy density also varies across the provinces, even after controlling for household income, demographics and rurality. There are fewer pharmacies per capita in Newfoundland, New Brunswick and Quebec, than in Ontario and more in Alberta. Results are qualitatively similar if the 2019 taxfiler reported income data are used instead of the 2015 household income data reported in the Census. These results are available from the author by request.
Figure 5 reports the parameter estimates of the model of total FSA level pharmacy weekly operating hours per 10,000 population. The same patterns hold as for the models of pharmacy density, but the estimated effects are now larger. For example, there are about 150 fewer operating hours per week per 10,000 population in the most affluent regions compared to the least affluent regions. The interprovincial variation in hours is also substantial, with a 150 hour difference between New Brunswick, the province with the fewest hours per 10,000 population, and Alberta, the province with the most. The same patterns are also observed in Fig. 6, which reports parameter estimates of the model of total FSA level pharmacy weekend operating hours per 10,000 population.