*2.1. Stage 1*

The categorization of major types of diseases was made according to the 11th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-11) [67] of the World Health Organization (WHO). Out of 26 ICD-11 codes, fourteen types of diseases were included in the study as *Yn* regressands—those repeatedly reported by the WHO and many scholars among the most widespread health problems in both indigenous communities and urban settlements in the Arctic [68–70] (Table 2).



Note: for all *Yn*, the measure is the incidence rate per 1000 people. Source: authors' development.

An array of *Xn* regressors was established along with three types of variables, reflecting environmental (*X*1–6), nutritional (*X*7–12), and economic (*X*13–16) dimensions of health-related effects (Table 3).


**Table 3.** Regressors included in the model.

Source: authors' development.

Physical environment, including quality of the air, safe drinking water, and adequate sanitary facilities, is one of the critical parameters of public health in the Arctic [71,72]. Despite the large gaps and significant uncertainties, which exist around quantification of influence of Arctic air pollution on public health [6], emissions can be severe, negatively affecting public health [42], particularly around the Russian cities of Norilsk, Vorkuta, and Monchegorsk, the areas of highest air pollution in the Arctic [5,73]. Nilsson et al. [47], Parkinson and Butler [74], and Thomas et al. [75] reported waterborne infectious diseases among the people living in the circumpolar territories in many Nordic countries.

Nutritional effects on health are measured as per capita consumption of major food products, including meat, fish, dairy, vegetables, and bread [49,76]. A parameter of traditional food proportion in a diet was included in the array like the one relevant in circumpolar and, particularly, indigenous communities. Many authors consider traditional food systems as essential sources of nutrients, n-3 polyunsaturated fatty acids [77], and vitamins C, B2, and B12 [78]. Sheehy et al. [79] reported that more traditional foods in a diet translated into greater dietary adequacy for proteins and a number of vitamins and minerals, including vitamin A, several B-vitamins, iron, zinc, magnesium, potassium, sodium, and selenium. According to Wesche and Chan [80], traditional food reduces the intake of saturated fats, sucrose, and excess carbohydrates that often are found in marketed food. However, while most of the studies report health advantages of traditional food patterns, including a lower incidence of cardiovascular disease [81], stability of gut microbiome [82], sources of bioavailable iron [83], among others, there are alternative findings of adverse health effects of traditional food. For instance, Jeppesen et al. [84] concluded that traditional food was positively associated with type 2 diabetes mellitus, impaired fasting glucose, and fasting plasma glucose. Bjerregaard et al. [85] found that impaired fasting glucose increased among the Inuit in Greenland with the consumption of traditional marine food, which might result in impaired insulin secretion – a link revealed by Færch et al. [86] and Weyer et al. [87,88]. Jørgensen et al. [89] discovered a strong association between persistent organic pollutants in a traditional seafood and low insulin secretion, while Kuhnlein [90] found higher health risks of traditional food systems containing sea mammals due to environmental pollution and increased organochlorine consumption. Contamination of traditional food sources is one of the reasons for lower β-cell function, an important early stage in the development of type 2 diabetes mellitus.

Among economic variables, the real value of cash incomes is used as one of the parameters of the economic accessibility of adequate healthcare services and nutrition [91]. The proportion of the population living below a minimum subsistence income along with the proportion of food expenditures in total household expenditures is the measures of economic accessibility of a healthy diet, which are commonly used by the Food and Agriculture Organization of the United Nations (FAO) [92]. They were included in the array to reflect the ability of households to generate sufficient income, which, along with their own production, can be used to meet food needs. The selection was also based on the idea that within a monetary dimension, access to food required a steady income to ensure a consistent, year-round supply of high-quality goods in the stores and a ready supply of healthy wildlife to be harvested [93]. Indigenous people do not rely much on marketed food; their food expenditures are low. But they still have to deal with the high cost of many commodities, such as oil, fuel, and transportation, essential for hunting, fishing, or reindeer herding activities [33]. Since the primary means for obtaining and producing food in indigenous communities are provided by hunting, herding, fishing, and gathering activities, a presence of a hunter, a herder, or a fisherman in a family is used as one of the economic regressors.

For all *Yn* and *Xn*, the data were obtained from the Federal Service of State Statistics of the Russian Federation [41], as well as from the authors' calculations.

#### *2.2. Stage 2*

A critical issue in building multiple regression models is how to eliminate independent variables with strong correlations between each other, whether positive or negative. Identification of collinear variables involves several approaches, one of the most widely used being the variance inflationary factor (VIF) (Equation (1)). It has been successfully applied for measuring and reduction collinearity, for instance, by Zainodin et al. [94] in an ordinary least squares regression analysis, Bowerman and O'Connell [95] in expressing independent variables in regression models as the functions of the

remaining regressors, and Dan and Vallant [96] in the analysis of variances between independent variables in complex survey data.

$$VIF = \frac{1}{1 - R\_n^2} \tag{1}$$

where *VIF* = variance inflationary factor; *R*<sup>2</sup> *<sup>n</sup>* = coefficient of multiple determination for a regression model.

According to Snee [57], Kutner et al. [58], Montgomery et al. [59], and Levine et al. [61], collinearity between the variables is considered high when *VIF* exceeds 5. The approach used at Stage 2 was that if *VIF* for a particular set of *Xn* regressors was less than 5, these regressors were included in the model. In case it was not, the *Xn* variable was eliminated from a subset. The computation was made across eight subsets of *Xn* variables, one per territory included in the study (see Stage 4 for the list of territories).
