Concept

As it has been pointed out, although it is frequently used, defining the concept of quality is not a simple task (^{Parasuraman et al., 1985}; ^{Alén, 2006}; ^{Colmenares & Saavedra, 2007}). For the American National Standards Institute (ANSI) and the American Society for Quality (ASQ), cited by ^{Evans and Lindsay (2005, p. 16)}, quality represents “the totality of traits and characteristics of a product or service that support its ability to satisfy certain needs.” According to the previous, regardless of its conceptual complexity, quality represents a series of attributes that consumers or clients of a product or service will evaluate as positive or not based on the capacity to satisfy their requirements and expectations. In addition, as ^{Yu and Hyun (2019)} state, goods and services are different. ^{Zeithaml et al. (1993)} add that service quality assessment is usually more complex than that for products due to three particular features: 1. They are intangible (they represent experiences, not objects) 2. They are heterogeneous (performance varies from one organization to another, from one consumer to another and from one day to another). 3. In general, their production and consumption are inseparable (the quality of services materializes during their delivery, not controlled in a manufacturing plant, as in the case of products).¹ These features make it more complex to define specifications for devising and thus standardizing the quality of services.

On the other hand, clients rate the quality of service by considering the final result and the process of receiving the service (^{Zeithaml et al., 1993}).² It should also be emphasized that the only valid criteria for evaluating the quality of a service are those established by its users. In this sense, and following ^{Cabello and Chirinos (2012)}, the central aspect in defining and assessing the quality of services, particularly in the health area, is the satisfaction of external users. Thus, the perception of the quality of the service is determined by how the provider organization executes the service, which is contrasted with the consumer’s expectations before receiving the service (^{Grönroos, 1984}). Coincidentally, research results in various service sectors led ^{Zeithaml et al. (1993, p. 21)} to indicate that service quality can be defined as “the extent of the discrepancy between clients’ expectations or desires and their perceptions.”

The Service Quality Model (SERVQUAL)

In this model, quality measurement is determined by the difference (gap) between the perceptions (assessment) and expectations (what is expected) expressed by the users of the services. ^{Parasuraman et al. (1985}, ¹⁹⁸⁸⁾ developed the SERVQUAL model after asking the following questions: how does the client evaluate the quality of the service, do they make a global evaluation or do they first evaluate the different stages of the service, and if so, which stages do they use to evaluate the service? For them, key variables that define consumer expectations are word-of-mouth communication, personal needs, previous experiences and external communications from the provider organization.³ They also identified ten dimensions that represent the criteria for evaluating service quality on the part of clients: tangibles, reliability, responsiveness, professionalism, courtesy, credibility, security, accessibility, communication, and client comprehension, which form the basis of the SERVQUAL. Since they observed a strong statistical correlation between several dimensions, they integrated some into others, refining the model, as seen in Figure 1, into five criteria: 1. Tangible elements, 2. Reliability, 3. Responsiveness, 4. Safety (comprising professionalism, courtesy, credibility, and security) and 5. Empathy (consisting of accessibility, communication, and client comprehension).⁴

Source: created by the authors based on ^{Parasuraman et al. (1988)}, ^{Zeithaml et al. (1993)}

Figure 1 SERVQUAL model criteria 

^{Colmenares and Saavedra (2007)} affirm that SERVQUAL is an obligatory reference in the study of service quality. Following their methodology, several investigations have been conducted around the world and in Ibero-America in various settings, including university canteens (^{Cevallos, 2015}); the postal service (^{Roslan et al., 2015}); public transportation (^{Valenzo et al., 2019}); and fitness centers (^{Peitzika et al., 2020}). Works on public health include, among others, those of ^{Cabello and Chirinos (2012)}, ^{Pedraza et al. (2014)}, ^{Zapata (2014)}, ^{Pedraza et al. (2015)}, ^{Henao et al. (2018)}, and ^{Teshnizi et al. (2018)}.

The Service Performance Model (SERVPERF)

The SERVPERF model was created by ^{Cronin and Taylor (1992}, ¹⁹⁹⁴⁾, who argue that although the quality dimensions proposed by ^{Parasuraman et al. (1985}, ¹⁹⁸⁸⁾ are adequate, the SERVQUAL paradigm and methodology are not the best for assessing the quality of services. They propose that this can be measured solely based on the analysis of the perceptions of the client receiving the service regarding the performance of the provider organization and no longer based on the differences between expectations and performance. According to these researchers, the SERVPERF model is a more functional method since, based on the same instrument used in SERVQUAL, it reduces by 50% the number of items to be measured because it does not require the study of expectations (^{Cronin & Taylor, 1992}).

Likewise, authors such as ^{Cabello and Chirinos (2012)} and ^{Le and Fitzgerald (2014)} have criticized the methodology and complexity surrounding the identification of client expectations and interpreting results of the SERVQUAL model. In this regard, ^{Salomi et al. (2005)} and ^{Ibarra and Casas (2015)} recognize the advantage of SERVPERF in terms of its use since not having the clients reflect on their expectations makes it a more agile instrument concerning the time required for the application of the questionnaire and in terms of less difficulty in the task of analysis and interpretation, since it is based solely on perceptions.

Regardless of the methodology of each model and the approaches of their creators, it should be pointed out that both SERVPERF and SERVQUAL are reliable and valid instruments for estimating the quality of services, which has been confirmed in works such as those of ^{Jain and Gupta (2004)}, ^{Salomi et al. (2005)}, ^{García and Díaz (2008)}, and ^{Bayraktaroglu and Atrek (2010)}. Similarly to SERVQUAL, numerous researches have been executed based on the SERVPERF model in activities such as hospitality (^{Ho et al., 2014}), banking (^{Torres & Luna, 2017}), and the aeronautics industry (^{Yu & Hyun, 2019}). Additionally, studies such as those by ^{Le and Fitzgerald (2014)}, ^{Basantes et al. (2016)}, ^{Ampah and Ali (2019)}, ^{Castellano et al. (2019)}, ^{Pedraja et al. (2019)}, and ^{Subiyakto et al. (2020)}, state that SERVPERF is suitable for measuring the quality of healthcare services offered by public institutions. The above, in addition to its functionality, motivated the use of this model in the present study.

Quality in government services

As mentioned above, the questioning to which the public administration was subjected at the international level regarding the effectiveness and efficiency of its operations led to the emergence of a new form of administration, known as new public management. Based on identifying the citizen’s needs and expectations as the core of public action (^{Vargas & Rodríguez, 2011}), quality management applied to the services provided by the government constituted one of the strategic axes of its new administrative paradigm. ^{Torres, Vásquez, and Luna (2011)} add that the need to offer services that meet national and international regulations, together with the demands of the public, requires the incorporation of mechanisms for their continuous evaluation to improve their quality.⁵

Although researchers such as ^{Vargas and Rodríguez (2011)} have found organizational resistance to change through the adoption of quality systems in government, other scientists, such as ^{Sáenz and Serrano (2012)}, indicate that they serve as change management models in organizations. In this context, ^{Ruiz (2012)} adds that the quality approach made possible a cultural transformation in the public sphere: the transition from administration to management. Simultaneously, it became necessary because of its ability to provide tools to governmental organizations that needed to relegitimize themselves by providing higher-level services. Although there are irregularities, the progress of quality management in Latin America is highly significant, which is confirmed by the review of the many successful experiences in this area (^{Torres et al., 2011}) and by the fact that it is referred to in various legal systems of the region’s governments (^{Ruiz, 2012}).

A milestone in this area was the publication of the Ibero-American Charter for Quality in Public Management (^{Latin American Center of Administration for Development, CLAD, 2008}), approved within the framework of the 10th Ibero-American Conference of Ministers of Public Administration and State Reform. The following elements inherent to quality public management stand out: inspiring principles, citizens’ rights and duties, and specific measures and instruments for designing policies and strategies. Coinciding with ^{Vargas and Rodríguez (2011)} that quality management should be conceived as a mechanism at the service of citizens and not as an end in itself to obtain certificates and recognition, the charter states that quality in public management is a transforming culture that drives the Public Administration to continuously improve to fully satisfy the needs and expectations of the citizens with justice, equity, objectivity, and efficiency in the use of public resources. (…) Quality in public management makes sense if its ultimate goal is to achieve a society of well-being, with justice and equity, guaranteeing the development of the individual and respect for their dignity, social cohesion, and the guarantee of the effective and comprehensive fulfillment of human rights. In particular, it ensures everyone has timely and universal access to information, benefits, and quality public services (^{CLAD, 2008, pp. 6-7}).

The links between the document mentioned above and the SERVPERF model are obvious if one considers that ongoing evaluation and continuous improvement, centered on the citizen, in addition to conducting opinion surveys to detect the level of satisfaction of the population with services, are fundamental principles for promoting quality public management (^{CLAD, 2008}). Tools such as SERVPERF can be very useful for this purpose.

General characteristics of the sample

A total of 794 surveys were administered, of which 196 corresponded to users of the CS, 213 of the HG, 326 of the IMSS, and 59 of the ISSSTE. Although a priori it was established that at least 175 questionnaires should be conducted per health unit, only 59 were conducted at the ISSSTE due to the low number of beneficiaries; the other units complied with the stipulations. It should be noted that when a minor or a user is unable to seek health services on their own due to temporary or permanent incapacity received medical care, their family member or guardian was surveyed.

Regarding the occupation of those entitled to health services, it is noteworthy that housewives were the ones who requested them most (40.9%), a result highly correlated with the fact that two-thirds of the respondents were women (Tables 3a and 3d). These statistics were expected because in the Mexican population, three-quarters of the people who provide care within households, including health care, are women (^{Instituto Nacional de las Mujeres, 2013}). In this context, women not only come to receive health services but are also (in general) responsible for taking their minor children or accompanying the elderly to medical appointments.

Another important aspect to note is that one out of every four ISSSTE users is a pensioner, consistent with their age, since in this health unit, more than half of the respondents stated that they were 50 years old or older (Table 3b).

Regarding the educational level of the beneficiaries, the majority of those with no education were served at the CS or the HG; 66% and 63% of the respondents, respectively, stated that they had no education or only basic education. In contrast, the population with higher education was attended to mainly by the IMSS or ISSSTE (Table 3c).

Reliability of the instrument

Cronbach’s alpha calculated with the 15 items of the instrument yielded a value of 0.962, indicating that the questionnaire has high reliability for collecting information on the quality of health services. On the other hand, the coefficient for reliability was 0.91, which also shows high reliability. The coefficients for responsiveness, security, empathy, and tangible items were 0.88, 0.89, 0.81, and 0.88, respectively. These results indicate that in these dimensions, the instrument has acceptable reliability (^{Hernández et al., 2010}). The decrease in the value of Cronbach’s alpha in each dimension, compared to the overall value, is due to one of the properties of the coefficient: if the number of items in the instrument increases, the value of the index increases (^{Cortina, 1993}).

Confirmatory factor analysis

The matrix of sample correlations and the standard deviations (SD) for estimating the factor model in Figure 3 are shown in Table 4. It should be noted that latent factors require a measurement metric, achieved by fixing the factor loading of an observable variable on each factor. The loads set with a value of 1.0 were λ3,1,λ3,2,λ4,3,λ2,4 and λ3,5 The model estimated with the LISREL software is shown in Figure 3.

Source: created by the authors

Figure 3 Unstandardized estimation of the five-factor SERVPERF model to explain the perception of health services in Ciudad Ixtepec, Oaxaca 

One way to evaluate the goodness of fit of the model is to observe the discrepancies between the observed variances and covariances matrix and the one reproduced by the model: if there is a good fit, then the matrix reproduced by the model will be similar to the one calculated with the observed variables. As can be seen in Table 5, both matrices are similar, although this similarity is relative, so it is necessary to resort to a formal statistic procedure.

Taking the RMSEA goodness of fit index as a reference, the results show a value of 0.079, indicating that the model reasonably fits the data (^{Holgado et al., 2019}). Another index for evaluating model fit is the comparative fit index (CFI). A model with a good fit will produce a CFI coefficient greater than or equal to 0.95. This work obtained a CFI of 0.988, indicating that the fitted model is better than one that assumes zero population covariances among all items.

To test the significance of each parameter, i.e., that it is different from zero, the Wald test was used. This procedure divides the estimated coefficient in the standardized solution by its standard error. At a significance level of 0.05, the null hypothesis that the parameter is equal to zero is rejected if the ratio, in absolute value, is greater than 1.96. The respective results are shown in Figure 4a. It should be noted that all parameters are statistically significant, which leads to the conclusion that each model parameter is important. All factor loadings show high values, indicating the relational relevance of each item to its theoretical construct.

Source: created by the authors

Figure 4 a) Significance tests for each model parameter and b) Estimated standardized model. 

In the CFA, the factor loading represents the correlation between the item and the latent factor. In addition, for a factor to explain at least 50% of the variance of the item, it must have a factor loading greater than 0.70, a condition that all factor loadings meet (Figure 4b). This situation ensures that in the standardized solution, the square of the factor loading is the proportion of the variance of the item explained by the model and the respective latent variable. In this context, the Reliability factor is the one that most explains the variability of the respective items, with at least 74% in each (Table 6). The latent variables Responsiveness and Tangible elements also explain a high percentage of the variability of their respective indicators. Although the item “The language of the person who assisted was clear” presented the lowest factor loadings, its latent variable, Safety, explains 62.2% of its variability.

On the other hand, it should be noted that the sample size corroborated the stability of the factorial model shown in Figure 3, which, following the recommendation of ^{Johnson and Wichern (2007)}, was verified by randomly dividing the original sample into two subsamples and estimating the confirmatory factorial model stipulated in Figure 3 in each one. The results always showed stability.

Status of indicators

The results were presented above, showing the reliability and validity of the proposed instrument to capture the perception of health services, specifically the relevance of the indicators in each of the five dimensions. Nevertheless, given the malleability of the indicators, an analysis considering each item of the model is presented below. Analyzing the average estimates of the items, it is possible to highlight four points. First, most averages are between eight and nine, meaning health services are rated “good.” It is worth noting that this convenient appreciation of quality levels by users resembles that reported in similar works, such as those of ^{Hamui et al. (2013)}, ^{Zapata (2014)}, ^{Pedraza et al. (2015)}, ^{Basantes (2016)}, ^{Pedraja et al. (2019)}, and ^{Ampah and Ali (2019)}.

Second, the indicators that show the highest ratings correspond to Tangible elements, specifically those related to the presentation of health personnel, in addition to the pleasantness and cleanliness of the facilities, which coincides with the findings of ^{Pedraja et al. (2019)} and ^{Monroy and Urcádiz (2019)}. Third, the indicators corresponding to Responsiveness present the lowest ratings, representing a key opportunity to improve the quality of services, specifically in establishing measures to respect pre-established service schedules and especially the reduction of waiting time to receive the service. On this aspect, there is convergence with ^{Le and Fitzgerald (2014)}, ^{Pedraza et al. (2015)}, and ^{Basantes et al. (2016)}. Finally, the overall rating of the care received and the fulfillment of expectations not only presented statistically the same average rating but also had a high positive correlation (r = 0.88), i.e., if the fulfillment of expectations increases, so will the overall rating (Figure 5).

Source: created by the authors

Figure 5 Intervals with 95% confidence for the average estimation of health services quality indicators 

Status of indicators by health institution

Figure 6 shows the average estimates for each item; nevertheless, it is necessary to question how they behave according to the healthcare institution. Accordingly, and based on the Kruskal-Wallis test for comparing k independent samples, the null hypothesis of equality of distributions by health institution is rejected for each item (p-value of 0.000 for each item). The breakdown of the averages of the ratings for each indicator by health institution shows that the best perception was found at IMSS in the Tangible elements factor, while the lowest scores in the five dimensions were found at ISSSTE. In addition, in HG, Responsiveness obtained average scores as low as in ISSSTE (Table 7a).

Regardless of the areas of opportunity detected, although at first glance it could be considered that poor management procedures underlie the ISSSTE in comparison with other organizations, this is not necessarily true if one considers that the average overall rating, 7.8, places it very close to the “good” parameter. The CS, HG, and IMSS received average scores of 8.6, 8.2, and 8.8 on the same item, which was clearly favorable but not very far from that of ISSSTE. It seems that this result reflects the fact that the relative majority of users of this institution (as opposed to the others) stated that they have higher education, which makes them more demanding when expressing their evaluations, in addition to the fact that it receives a markedly lower percentage of housewives, who, together with unemployed personnel, give the highest evaluations.

The passing grades observed in the study could be somewhat explained by the approach of ^{Padma et al. (2010)}, in that hospital patients consider the interpersonal aspect of care the most important. In Ciudad Ixtepec, the relations between health personnel and clients tend to be very warm and close.

Status of indicators according to gender, occupation, age, and education

The Mann-Whitney procedure used to test the null hypothesis of equal distributions of indicators between men and women showed that statistically both are the same for each item (p-values greater than 0.10). That is, there are no significant differences between men’s and women’s perceptions of each criterion. In contrast, occupation is a significant variable in explaining the variability of each item. In this context, employees and students in the Reliability dimension provided the lowest average perceptions. Also, in the Responsiveness factor, students were the most rigorous in issuing ratings. In contrast, users who said they were not working gave the best ratings in each dimension (Table 7b). These results are similar to those of ^{Ferraces et al. (2000)}, who, when analyzing the quality of bus service, found that pensioners report a higher degree of satisfaction than students and professionals.

According to the Kuskal-Wallis procedure, age and level of education are useful variables in explaining the variability of the scores on each item (p-values of 0.000 for each item and both variables). Thus, taking the age of the beneficiaries as a reference, it is possible to distinguish three characteristics in each quality dimension: a) those between 20 and 30 years old registered the lowest averages; b) the highest ratings were provided by users aged 60 years and older; and c) those aged 30 years and older, but younger than 60 years old, presented the same average ratings (Table 7c). The above is consistent with the study of the ^{University of Almeria (2015)}, where it is noted that older people award higher ratings than younger people. Similar behavior was observed with the level of education, where in each dimension it was observed that: a) users with elementary and high school education registered the same average ratings, statistically equal to the general averages in each item; b) beneficiaries with no education issued, on average, the highest ratings; and c) the distributions with the lowest location parameter, the mean, refer to users with university studies (Table 7d).

It should be noted that the results achieved are consistent with those recorded in other research on service quality evaluation in the region based on the SERVPERF model (^{Torres & Luna, 2017}). In both studies, the most highly rated quality dimension was Tangible elements, and the one with the lowest rating was Responsiveness, and there were certain differences in scores by institution. Likewise, the two studies showed no differences in the perception of quality by gender. Nevertheless, there were differences in occupation, education, and age since students and professionals, people with higher education, and younger people tended to give lower ratings. Housewives and the unemployed, citizens with basic education or no education, and people over 60 years of age give higher marks to the services. Although there are differences in the overall evaluation of the quality dimensions: 8.3 on average in this study and 9.1 in the one mentioned above, both are positive.