Conceptual considerations and methodological recommendations for the use of the mortality‐to‐incidence ratio in time‐lagged, ecological‐level analysis for public health systems‐oriented cancer research

Sunkara and Hébert1 analyzed the correlation between the mortality-to-incidence ratio (MIR) for colorectal cancer and the World Health Organization health systems ranking for all 34 countries of the Organization for Economic Co-operation and Development (OECD). Using scatterplots and linear regression models, they concluded that MIR is useful for identifying disparities in cancer screening and treatment in high-income countries. Although the actual quantitative results are not questionable, herein we critically revise some conceptual considerations with regard to the MIR in this type of studies and provide some methodological recommendations. The MIR indicator is a proxy measurement for survival rates of specific cancer sites. It is a ratio of the death and incidence figures (either actual absolute case numbers or population rates) registered in the same period of time for a specific cancer.2 Therefore, it depends on cancer-specific screening and prevention, treatment, and survival dynamics, along with the availability of high-quality data. Conceptually, in the long run, a given health system will have an impact on the MIR. However, understanding the actual mediating mechanisms for such an impact and the latency with which such effects appear are essential to define the appropriateness of the study measurements and design a study aimed at capturing such complex interrelationships. In the article by Sunkara and Hébert, MIR data were obtained from the 2012 GLOBOCAN database. In settings with high-quality data, a criterion that countries in the OECD surely fulfill, the MIR performs better in predicting the 5-year survival rate with distinctions according to cancer site.3 Therefore, the actual usefulness of MIR estimates lays in their future predictor capacity. In the article by Sunkara and Hébert, that means before 2017 because the 2012 GLOBOCAN data do not contain exclusively data from 2012 but the latest available data up to that year for those 34 OECD countries. This time span is very important to take into account to make the right interpretation of the MIR. Conversely, the WHO health systems ranking indicator encompasses dimensions not exclusive to cancer but for health care achievements in general.4 Sunkara and Hébert described the specific situation of cancer guidelines for some countries found to be either above or below their calculated expected MIR. However, no absolute quantitative data or evidence are presented to establish that the ranking is a good proxy for the actual country-level performance of cancer screening and care. An indicator that is more related to cancer is more appropriate in studies such as this. An example is the total number of external beam radiotherapy devices per capita as used before by Batouli et al for another study similar to that by Sunkara and Hébert but expanded to also include lower-income and middle-income countries.5 Moreover, the WHO ranking dates back to 2000 and also used the latest data available up to that year. Consequently, there is an approximate time gap of close to 20 years between the ranks and the survival predicted by MIR. This should lead us to think about the actual time required for each country, or health system, to change its cancer epidemiology and the appropriateness of linking these 2 measurements. Finding a statistical correlation is a great exercise but it is just the first step in the larger task of establishing prediction. The MIR depends on cancer epidemiology and health care provision over time, and health systems are also very likely to change and evolve over time as a result of factors such as population demographics, health needs assessments, priorities in resource allocations, disasters, health politics, and even government-level agendas. With so many mediating mechanisms, we should be cautious when interpreting the results of time point, country-level aggregated measurements or their correlations such as those presented by Sunkara and Hébert because MIR and health systems are not static or fixed entities. Finally, we suggest the following steps for a much more comprehensive analytical exercise to better elucidate the original research question: 1) to use data more closely related in time (eg, the 2000 WHO health rankings and GLOBOCAN 2002, 2008, and 2012); 2) to assess time trends of MIR estimates; 3) to compare time trends between countries and classify them; 4) to assess how these MIR time trends (not just the single-point estimates) correlate with a given baseline health system ranking; 5) ideally, to also measure in a longitudinal fashion the health systems ranking; 6) use of random effects linear mixed models, multilevel models, or generalized equation models to assess how changes in MIR and health systems within and across countries and over time are related; and 7) keep in mind these ecological analyses using time-lagged, country-level indicators are prone to ecological fallacy obscuring the mediating mechanisms and the subnational reality within each country. We salute and welcome all timely efforts to understand how health systems performance translates into actual population health status, but we also have to consider many other epidemiological aspects beyond just the numbers. No specific funding was disclosed. The authors made no disclosures. Alejandra Cordero-Morales Maria Jose Savitzk Karoline Stenning-Persivale Eddy R. Segura, MD School of Medicine Universidad Peruana de Ciencias Aplicadas. (Peruvian University of Applied Science)