Digital consumer behavior and medical tourism: A regional analysis in Mexico

Medical tourism has increasingly become an important alternative to receive healthcare services given medical systems’ limitations such as: treatment availability, access


Introduction
Medical systems, which include patient care, administration, and infrastructure, amongst others (Nolan, 1998), and their policies are different in every country. Since such systems do not always fit the population's needs and/or match their affordability, medical tourism has increasingly become a popular means to obtain medical attention. Thus, medical tourism is not a new or modern practice; in fact, Connell (2013) has posited that the practice has been carried out for over 2000 years. Throughout such time, the industry has certainly evolved in terms of physician care, services, and consumer behavior; however, its growth has importantly been impacted by globalization and technological advancements and, particularly, with the increasing availability of internet services and digital platforms. Medical tourism can be analyzed from two different perspectives, that is, from the standpoint of travelers seeking medical attention, and physicians traveling to provide medical services (Reed, 2008); this study focuses on the consumer behavior of the former.
Medical tourism has been defined as the practice of individuals travelling to another country to obtain health care (Benowitz & Gaines, 2020), treatments, and health services. Therefore, countries' conditions may drive consumers to engage in medical tourism, in turn, impacting both the country of origin and destination. The economy of destination countries is not only impacted by the medical aspects of the travelers' visit, but also because many take the opportunity to make a trip of it, take a tour, go sightseeing, and take in the local culture. While a significant number of people would like to indulge in medical tourism, the reality is that the vast majority of the population cannot; thus, although growing, it continues to be a relatively small industry that is relatively unexplored.

243
There are many countries that are medical tourism destinations and Mexico is one of the most important due to affordable prices, availability of treatments, and types of destinations (Orozco Núñez et al., 2014). However, further research is required to comprehend the current consumer behavior dynamics, particularly regarding such emerging market. For instance, medical tourists' use of digital platforms and preferred type of destination remained unclear. Therefore, the general objective of this study was to analyze three consumer profiles related to the use of digital platforms and the preferred type of destination for medical tourism. In order to do so, a cluster segmentation analysis was performed using k-medoids, followed by a correlation analysis as well as a regional analysis to identify the relationship among the study's variables and groups of traveler profiles. The paper is sectioned as follows: section two includes the study's theoretical framework, section three encompasses the methodology, section four includes the study's results, and section five consists of the concluding remarks, limitations, and future research directions.

Medical tourism
One of the most common ways to analyze medical tourism is from the perspective of travelers seeking medical attention (Reed, 2008). From such standpoint, according to the Center for Disease Control and Prevention (CDC), medical tourism occurs when people seek medical attention out of the country in which they reside (Traveler's Health, 2021). The Organisation for Economic Co-operation and Development (OECD) has stated that medical tourism is described as people travelling to other countries, typically from developed to developing countries, to obtain healthcare services (Lunt et al., 2011). And Gupta & Das (2012) posit that medical tourism is an activity that occurs when healthcare treatments and/or services are unavailable and/or too expensive; in other words, it occurs when patients cannot access the treatments and services they need in their country of residence.
Medical tourism is, therefore, defined as a practice that occurs when a person travels out of the country of residence to receive health treatments and/or services  whether they are programmed, emergent, biological, dental, or cosmetic. Consumers commonly travel to developing countries and emerging markets (Reed, 2008;Horowitz et al., 2007), for treatments and services, because affordability is a primary factor associated with medical tourism. However, although medical tourism definitions emphasize that consumers travel abroad to access health treatments and services, Connell (2013) has argued that the practice is currently also "short distance, cross border and diasporic". Therefore, consumers, patients, may be traveling both long and short distances to receive healthcare.

Medical tourism in Mexico
There are various countries around the world that are popular destinations for medical tourism, including but not limited to: Brazil, Costa Rica, Cuba, Dubai, India, Malaysia, Mexico, Singapore, South Africa, Thailand, as well as various countries in Europe (Lunt et al., 2011). Mexico is not only a popular destination for medical tourism, but it is also the largest destination in North America (Angeles Health International, 2022), and, according to Mexico's National Bank of Foreign Trade (Bancomext), it is the second most important destination for medical tourism worldwide (Bancomext, 2019).
Medical tourism, therefore, has an important impact on the country's economy. The latter occurs because medical tourism includes a series of services in addition to treatments such as, travel agencies, hospitals, laboratories, restaurants (Solis, 2019), as well as transportation, and digital platforms, amongst Mexico receives approximately 1.2 million visitors per year and common countries of origin include China, France, Germany, Japan, and United States (Bancomext, 2019). Reasons include: types of destinations, availability and access to treatments, and prices, amongst others (Johnston et al., 2010).
The prices for medical tourism treatments in Mexico are in the low-end and mid-range among some of the most visited destinations. Table 1

includes a comparison between Mexico and the United
States in relation to the prices of the twenty-one most sought out procedures in medical tourism.
Furthermore, medical tourism in Mexico enables consumers to save between 35 and 85 percent in treatments, as well as significantly reduce waiting times for their procedures (Bancomext, 2019).
Additionally, it is a popular destination due to diverse cultural attractions, landmarks, archeological treasures, culinary activities, amongst others.

Medical tourism and digital platforms
Digital platforms have significantly impacted just about every industry and sector. As such, the internet is a key element in the growth of medical tourism because it provides fundamental information (Lunt et al., 2011) regarding treatments, medical professionals, destinations, hospitals, etcetera; furthermore, it enables communication amongst consumers and service providers worldwide (Reed & Dawood, 2008).
Certain digital platforms facilitate consumer electronic word-of-mouth (Connell, 2013) (eWOM), through which current and potential consumers can exchange experiences, opinions and recommendations. For instance, social media, such as Facebook, has been associated with a positive effect on consumer purchase decision making related to medical tourism (Alghizzawi et al., 2020).
Regarding technology, digital consumers display a certain degree of readiness, meaning, an inclination towards the use of new technologies (Parasuraman, 2000). As such, said consumers tend to invest more in new technology and devices (Kierzkowski et al., 1996). The reality is that not all consumers 246 are so inclined to use digital platforms; that being said, critical events, such as COVID-19, have somewhat forced late majorities and laggards to the fold.

Medical tourism consumer behavior
Each country's health care system's characteristics are well associated with consumer behavior related to medical tourism (Béland & Zarzeczny, 2018). Meaning that, medical tourist consumer behavior is influenced by various factors emanating from health institutions and their policies; in such way that the availability, quality, access, and price of treatments are some factors that drive consumers to partake in medical tourism. The destination itself also plays an important role in consumers' decision-making when they are inclined to take advantage of the trip to vacation.
As consumer loyalty is well associated with their perception of brand value (Vera & Trujillo, 2017), consumers' opinions of their experiences are significant for potential medical tourism consumers.
Meaning that individuals and groups can influence consumer behavior, such as reference groups (Piotrowska, 2017); while this is true for in-person word-of-mouth, considering that there may be greater impact and influence on decision-making when information is retrieved online from acquaintances and other trusted sources, is central to understanding consumer behavior during the fourth industrial revolution.
For instance, according to Benowitz & Gaines (2020), in addition to lower prices, some factors influencing medical tourisms include a suggestion from family or friends, the opportunity to combine medical care with a vacation destination, a preference for care from providers who share the traveler's culture, or to receive a procedure or therapy not available in their country of residence. Those consumers that can be part of medical tourism are those that have a certain degree of disposable funds. While many consumers seek treatment and services in other countries because they are significantly less expensive (Çapar & Aslan, 2020), the practice is still prohibitive for most of the population globally.
Medical tourists may be classified in various manners depending on the approach of the analysis.
For instance, according to Awadzi & Panda (2006), there are three main categories of medical tourism consumers, these include: consumers that do not have insurance, those that are underinsured, and consumers that cannot access insurance.
In terms of destinations, a study performed in Turkey found that the factors influencing consumer behavior in medical tourism included "accessibility of health care service, level of security and safety, quality of health care service, level of hygiene, potential of savings low cost, and tourism opportunities" (Çapar & Aslan, 2020, p.80). a previous study developed in Mexico found three main consumer profiles, these being: The "Frequent/Air/Modern Traveler, the Saver/Land/Modern Traveler, 247 and the Occasional/Air/Traditional Traveler" (Arrioja-Castrejón & López-Fernández, 2021), which will be used for this study; Table 2 includes the classification of medical tourism consumers and their main characteristics.  (2021) The consumers' characteristics abovementioned depict frequency, willingness to spend, means to plan travels, and overall priorities, medical treatment and/or procedure notwithstanding; however, the proclivity for the types of destinations remains undefined. Hence, the following hypotheses have been framed: H1: The greater the frequent digital traveler consumer behavior, the greater the preference for a city destination for medical tourism services.
H2: The greater the thrifty digital traveler consumer behavior, the greater the preference for a beach destination for medical tourism services.
H3: The greater the occasional traditional traveler consumer behavior, the greater the preference for a city destination for medical tourism services.

Procedure
The general objective of this study was to analyze three consumer profiles related to the use of digital platforms and the preferred type of destination for medical tourism, in order to do so, this study used a clustering segmentation procedure, that refers to a broad set of techniques for finding groups, or 248 subgroups, within a data set. This grouping is carried out to divide the observations so that those belonging to the same group are fairly like each other, while those belonging to other groups are different. This study's model began with a correlation analysis to determine if the variables' relationships were significant. Subsequently, a cluster segmentation analysis was performed using k-medoids to identify the relationship among the study's variables and groups of traveler profiles, considering the study's sample and the effect of those profiles' reservation tool selection and preference for type of destination.

K-medoids
K-medoids is a partition clustering technique where each cluster is represented by an element of the cluster (Kaufman & Rousseeuw, 1987), these are points known as the medoids; the latter refer to an object in a cluster where its average dissimilarity is minimal in relation to other members in the cluster (Loy-García et al., 2021). K-medoids is less sensitive to noise and outliers since, instead of centroids (used in k-means), it uses medoids as cluster centers. Medoids for each cluster are calculated by finding object i within the cluster that minimizes the equation (1), where Ci is the cluster containing object i and d (i, j) is the distance between objects i and j (Reynolds et al., 2004).
(1) Furthermore, this algorithm is an effective alternative to K-means to segment a dataset into groups. In this method, groups are represented by a particular object in the cluster; said objects, medoids, correspond to the cluster's most central points. The Partitioning Around Medoids (PAM) algorithm is the most common grouping method (Park & Jun, 2009). This method requires that the researcher know the data and select the appropriate data and number of clusters to be produced.

Dunn index
The Dunn index, shown in the equation (2) is an internal cluster validation measure that enables the calculation of the distance between each group's objects and those in other groups (Dunn, 1973). This index calculates the minimum of this pairwise distance as the separation between clusters and uses the maximum distance within the cluster as intracluster compactness.

Silhouette coefficient
The Silhouette coefficient, shown in the equation (3) below, was used as an optimization criteria to define the number of partitions in a cluster using a measure of the cluster classification's quality (Gentle et al., 1991).

Definition of variables
This research first retrieved data from several sources of information including, Patients Beyond Borders, Medical Tourism Association, Medical & Dental Departures, and Health Tourism, considering the main medical treatments that are programmed under the medical tourism industry worldwide. Then, a structured survey was administered to a sample of 100 medical tourism tourists and consumers incorporating variables defined in a previous study carried out in Malaysia (Musa et al., 2012), in order to identify travel preferences to better establish the traveler profile.
The instrument included 8 questions that assessed the following variables: type of trip (business, pleasure, or both), gender (male or female), age, average expenses per trip in Mexican pesos, number of trips per year (1,2,3,4,5,6,7 or more), preferred place of travel (between beach and city), preferred type of transportation (air, land, or sea), and preferred reservation method (internet/mobile app, social media, or travel agency/phone). Data was collected over a period of three months with a hundred percent response rate; the sample included 49 women and 51 men.
The variables' definitions and descriptions are included in Table 3. The variables in the dataset were codified by assigning values 0 and 1 to convert them into dichotomic and trichotomic variables and determine which factors should be considered when designing a digital tool based on user preferences.

Correlation analysis
A correlation analysis was performed to study the relationship between the set of variables associated with the traveler profile and the variables linked to the type of reservation and quantify the number of existing independent dimensions. The selected variables included three associated with the reservation medium, three variables associated with the type of transportation, four related to the traveler profile, and one variable associated with the destination.
The variables associated with the traveler's profile were Expenses, Age, Travel frequency (TRF), and Gender. The variables linked to the reservation type were Internet_App (ROB), Social_Network (RSN), and Agency_Phone (RAP). Table 4 includes the descriptive analysis of the studied variables, where we can see that we have 3 quantitative (EXP, TRF, AGE) and 7 qualitative (GDR, ATT, TTT, MTT, ROB, RSN, RAP) variables.  Table 5 includes the variables considered in this study for analysis, i.e. the reservation group (reservation type, composed of the variables: ROB, RSN, and RAP), and traveler information (traveler's profile, consisting of the variables: EXP, TRF, AGE, and GDR), to be able to determine if there is a correlation among these variables, and see if the use of digital tools such as the Internet, mobile applications, or social networks have a significant relationship with the type of user profile.
As can be seen in Table 5, there is a clear correlation between the age and the type of reservation; this correlation is positive within older age groups in relation to the telephone or travel agency reservation method, and negative in relation to the Internet and social networks. The latter is consistent with the notion that younger consumers are more comfortable with the use of digital platforms. There is also a positive correlation between average expenses and travel frequency.

Cluster with mixed data using K-medoids
In order for an algorithm to group observations, a notion of (dis)similarity between observations must be defined first. There are a variety of metrics in cluster analysis to help choose the number of clusters to extract. The silhouette width, which is an internal validation metric that determines an observation's similarity to a cluster in comparison to that of the nearest neighbor cluster, was used for the analysis. The metric can range from -1 to 1, where the highest values are better. After calculating the Silhouette width for clusters ranging from 2 to 10 for the PAM algorithm (which uses a partition around medoids) for the analyzed data, three clusters show the first change in the initial value (see Figure 1).

Cluster validation using Silhouette width
The Silhouette width coefficient was used to validate the quality of the cluster and determine how well the observations were classified (see Table 6); higher values imply a good classification, negative values mean poorly classified observations, and values of 0 mean that there is a possible overlap. After running the algorithm and selecting three clusters, we can interpret the clusters by running a summary on each one (see Table 2), based on the following results: Tourism and Medical Tourism Association. This map was created using the GeoDa tool, considering a cluster segmentation analysis for mixed data using K-medoids. In this study, the Z standardization was used with 150 repetitions and a random seed, with a maximum of 1000 iterations, using the Manhattan distance function. (i) The states in dark blue do not have a defined cluster for the medical tourism market; however, they have tourist and medical services that are isolated or not registered by any of the associations previously mentioned.
(ii) The states in light blue do have medical tourism services, but not a cluster formed under this concept.

(iii)
The states in green have well-defined clusters with well-formed communication, certifications, and supply chains, which allow them to generate a more competitive market for all its stakeholders.
Medical tourism travelers' preferences regarding destinations were evaluated after running the clustering segmentation (see Table 7) by counting the number of travelers in each group and reviewing the data obtained in the survey related to their preferences to visit a city or a beach destination. The results confirm that user profiles have some relationship with the consumer preferences for the destination for consuming medical tourism services in Mexico. As such, the three hypotheses proposed in this research, according to the data analyzed, were accepted; however, the data shown in Table 7 is not clear-cut in two of the profiles, these being: Frequent digital traveler and Thrifty digital traveler.

Conclusions
The medical tourism industry is small yet fast growing as the number of people travelling to other countries to receive medical attention continues to increase. Tourism and medical services in Mexico represent the main source of direct jobs and social welfare in the country (Secretaria de Turismo, 2018); therefore, it is necessary to strengthen both industries particularly in an uncertain environment where the fear of the contagion of COVID-19 and other diseases prevails in the minds of consumers (Garcia-Muñoz et al., 2020). Understanding consumer behavior in relation to digital technology is key to making sound strategic decisions to foster growth within the industry.
A key factor today to increase the competitiveness of medical tourism providers is taking advantage of the increase in the use of digital platforms to purchase products and services, which has been further enhanced by the COVID-19 pandemic (Esposti et al., 2021). The use of social networks as a means of reference and knowledge of recommendations based on the experience of other users, as well as the use of mobile applications, have contributed to exponentially exploding e-commerce and making it increasingly profitable for companies all around the world (Gasser et al., 2020). Therefore, it is crucial for medical tourism service providers to take advantage of these digital tools, particularly for younger 256 consumers, and to identify user profiles clearly to reach them and offer more accurate services adapted to the needs of specific types of consumers.
Based on the results obtained from the cluster segmentation analysis, we conclude that consumers of medical tourism services in Mexico, previously categorized into three different profiles based on their preferences and consumption habits, also have an effect on the consumer preferences related to the destination; however, the differences between one type of preference and another are limited.
Likewise, these results confirm that having clearly identified profiles of medical tourists will help service providers identify and offer a better user experience to improve the consumption of medical tourism services in Mexico, which confirms the results reached by other researchers (i.e. Montiel, 2015;Christensen et al., 2016;Garcia-Muñoz et al., 2020).
The study's main limitation was the sample; that is, given the size, characteristics and context, the results may not be generalizable. In such way that future research could extend this study with a larger sample, a broader consumer base, by collecting data from various locations within the medical tourism network; further, future research could also study the impact of different types of transportation on medical tourism consumers' preferences. Finally, future research could develop a cross-cultural study to assess whether medical tourism consumer profiles as well as their behavior significantly varies in different countries, as well as the effect on destination preferences.