Factors associated with the evaluation of Primary Health care from the user’s perspective: results of the telephone survey

This paper aims to evaluate the performance of PHC from the perspective of users and its association with sociodemographic characteristics, self-reported health conditions, and behavioral risk factors for Chronic Noncommunicable Diseases. This is a population-based cross-sectional study with data from the 2015 VIGITEL Telephone Survey. The Primary Care Assessment Tool short version was adopted. The study population covers adults over 18 years of age who used PHC services in Belo Horizonte in the last 12 months (n = 872). The multiple logistic regression model was performed to estimate the odds ratio. We observed that adults without a health insurance plan are 3.21 (95% CI 2.08-4.96) more likely than those with a health insurance plan to evaluate PHC with a high score (≥ 6.6), and adults with low schooling (95% CI 1.48-5.32), people with diabetes (95% CI 1.05-3.24), obese (95% CI 1.20-3.24), and older adults (95% CI 1.00-1.41) were 2.81, 1.84, 1.97, and 1.19 more likely to report a high score for PHC quality than the others, respectively. The use of the PCATool short version in a telephone survey showed a new possibility for PHC performance assessment and can become useful in managing health services.

introduction Primary Health Care (PHC) is the guiding axis of the Health Care Network (RAS) in the Brazilian Unified Health System (SUS). It is responsible for ensuring universal and equal access to available health actions and services 1 and reducing hospitalizations for conditions sensitive to primary care. A strong and resolute PHC contributes to curbing health system costs and upholding SUS 2 principles.
According to Starfield and Shi 3 , PHC should be considered the gateway to the health system and offer access to prevention, cure, and rehabilitation services. It must also rationalize all available resources for health promotion and maintenance and integrate the health system's points of care to ensure the timely provision of care appropriate to the user's needs 3 .
A strengthened and well-structured PHC must include four structural or essential elements: a) first contact; b) longitudinality; c) comprehensiveness; and d) coordination. It should also include two derivative elements: family approach and community orientation 4 . Thus, one of the benchmarks for assessing PHC services is the assessment of these attributes.
Even with the advances in the last decades in health with the consolidation of the SUS and the implementation of the Family Health Strategy (ESF) 5 , it is essential to ensure quality care that meets the users' needs. Qualifying the services requires evaluation processes with approaches that show the perspectives of the various health care stakeholders, such as managers, professionals, and users. The assessment also contributes to the identification of barriers and weaknesses of PHC services 6,7 .
The evaluation of health services must be understood as a management tool in all health actions. It can direct or redirect health policies and programs, promoting and qualifying health care, and strengthening SUS principles 5 . It also contributes to social control when the results are shared with the population, favoring participation in the decision-making process of managers [6][7][8] .
The Primary Health Care Secretariat (SAPS) was created thirty years into the SUS establishment, thus emphasizing the PHC's relevance as a priority for the SUS. Among the SAPS objectives are the strengthening of PHC's essential and derived attributes, training, professional staffing, care support strategies, and development of information and care technologies 9 .
Some tools used in several countries were developed considering the PHC assessment. In a review and meta-synthesis carried out between 1979 and 2013, Fracolli et al. 10 identified the leading national and international PHC assessment tools. They also stated that the Primary Care Assessment Tool (PCATool) 11 is the most widely used instrument in Brazil 10 . In another bibliographic study of scientific production between 2007 and 2017 on the assessment of PHC in the Brazilian context, Ribeiro and Scatena 12 also noted that PCATool 11 was the most widely used instrument in studies published in this period. This instrument is very relevant, considering that it has already been validated and used in several countries and different Brazilian regions, thus allowing comparing outcomes in this research with other studies 12 .
Another important issue concerns the profile of health services users. The study by Malta et al. 13 confirmed the recurrent use of these services by people with NCDs, which can be explained by the greater demand for routine visits or complications, more significant associated comorbidities, and the need for continuous monitoring 14,15 . Chronic conditions are severe public health problem 16 and entail high costs for the health system. They also significantly impact the population's quality of life, which shows us that PHC has a fundamental role in representing the link in the health system responsible for monitoring these cases, which often require more complex and coordinated care between different services.
Several risk factors are related to NCDs, such as inadequate diet, excessive salt intake, alcohol abuse, physical inactivity, overweight, tobacco use, and glucose and lipid metabolism disorders 17 . These risk factors are the target of interventions in health policies, mainly within PHC. In this context, this study is relevant considering the scarcity of PHC performance assessment works from the user's perspective and studies with an analysis relating PHC performance assessment to clinical outcomes. Also, PHC assessment using a national population database such as the Surveillance System for Risk and Protective Factors for Chronic Diseases by Telephone Survey (VIGITEL) is of great importance 18 . It was used for the first time for this purpose, which is an innovative and low-cost possibility.
Considering the above, this study carried out in Belo Horizonte using PCATool aims to assess PHC performance from the users' perspective and its association with sociodemographic features, self-reported health conditions, and behavioral risk factors for NCDs.

Methods
This is a cross-sectional population-based study. Data from the Belo Horizonte sample of VIGI-TEL 2015 18  VIGITEL 2015 interviewed the adult population (≥18 years old) living in households with at least one landline 18 through a structured questionnaire. The telephone interview starts with using a VIGITEL 2015 questionnaire with questions addressing the demographic and socioeconomic characteristics of individuals, behavioral risk factors for NCDs, and self-reported health conditions 18 . After applying this questionnaire, respondents answered questions to identify those who used any health service in the last 12 months 20,21 , as follows: -"When you are sick or in need of treatment to take care of your health, which health service do you usually look for?" (If public or private, whether PHC, hospital, or emergency department); -"In the last 12 months, did you seek care at a PHC Unit (UBS) (whether a health post, health center, or family health unit) to take care of your health? ("Yes" or "no"). If so, how many times?".
Thus, for this study, the adult interviewed who answered that he had sought some PHC health service at least once in the last 12 months and that mentioned the name or location of the UBS sought in the city of Belo Horizonte was considered 20,21 . These respondents were then invited to answer the VIGITEL evaluation module to assess the performance of the municipality's PHC services 22 .
In this study, we used only the part of the VIGITEL evaluation module made up of the PCA-Tool-Adult-Brazil short version for PHC services users, translated into Portuguese and validated in Brazil 23 . This instrument consists of 23 items arranged in blocks of questions that correspond to the PHC attributes' performance evaluation (access, longitudinality, comprehensiveness, coordination, family orientation, and community orientation) 23 .
PCATool is a PHC assessment tool developed in Baltimore, Maryland (USA), by Starfield 24 . It was built from the health service quality assessment model proposed by Donabedian 25 , whose evaluation is based on the measurement of aspects of health services' structure, process, and results. The PCATool 11 proposes to measure the presence and extent of PHC attributes according to structure and process aspects. Empowered by statistical methods, the PCATool enables the association with the effectiveness of the actions and services provided and establishing associations with other clinical outcomes 26,27 .
Responses to PCATool items use the Likerttype scale where the respondent specifies his level of agreement with the item, ranging from 1 to 4 for the analysis of each attribute (1 = certainly not; 2 = probably not; 3 = probably; 4 = certainly), with the addition of option 9 (I don't know/I don't remember) 11 . The values are transformed on a continuous scale, ranging from zero to ten (Chart 1) 11 after consolidating each attribute's data. The essential, derived, and general scores are calculated along with the score by attribute. We also calculated the standardized general score representing the cutoff point, considering the general score found (Chart 1) 11 to carry out the statistical analyses. A general score ≥ 6.6 shows a strong PHC orientation, equivalent to a value of 3 on the Likert scale (probably) and, consequently, a good quality of care (Chart 1) 11 . It is worth mentioning that the degree of affiliation aims to identify the professional or service that serves as a benchmark for the respondent and, therefore, is not considered a PHC attribute but is included in the calculation of essential and general scores 11 .
A total of 2,125 interviews were conducted in the VIGITEL Belo Horizonte 2015 sample of the 3,800 telephone lines used (equivalent to 19 replicates of 200 telephone numbers each), in which 2,006 respondents reported having sought some health service when they needed care. Of these, 795 users answered the VIGITEL evaluation module ( Figure 1). The study population consisted of adult PHC users who agreed to answer the VIG-ITEL evaluation module. The sample size was defined as 1,000 adults, obtained by the expression: , where p=50%, z=1.96, and error margin of 3.1.
The sample obtained with the VIGITEL evaluation module was 795 interviews, and it was necessary to add five replicas with 200 phone numbers each, totaling 1,000 phone numbers, to reach the minimum size defined by the sample calculation. Of these, another 118 adults were interviewed, who answered the short version of the questionnaire of VIGITEL and the VIGITEL evaluation module, thus totaling 913 interviews. Forty-one interviews were excluded due to the impossibility of locating the address of the PHC Unit (UBS) that the respondent said he used ( Figure 1).
Thus, the population of this study consists of adults over 18 living in households served by at least one landline in Belo Horizonte, who used the PHC services in the city of Belo Horizonte in the last 12 months before the interview with identified UBS address and who agreed to answer the VIGITEL evaluation module (n = 872) (Figure 1).
Post-stratification procedures calculated using the rake method to expand the sample to the total population were applied to reduce the sample selection bias of the VIGITEL Belo Horizonte 2015 that interviews adults with a landline. Details on the sample design of the VIGITEL survey and post-stratification process have been described in other publications 18,28 .
New post-stratification weights were calculated to adjust PHC users' distribution by age, gender, and schooling. These weights were calculated using the Data Analysis and Statistical Software (STATA) version 14.0 using the SURVWGT package and adopting the rake method and estimating the PHC user population obtained from the VIGITEL evaluation module as a reference population 29,30 .
A descriptive analysis of the variables was performed using absolute and relative frequencies to characterize Belo Horizonte PHC service users. Then, Pearson's χ 2 test was used to identify associations, with a significance level of 5%. chart 1. Description of PCATool score calculations. escore cálculo Descrição Standardized general score (score-minimum score)*10 maximum score-minimum score (1) if standardized general score ≥ 6.6 (0) if standardized general score < 6.6 General score A+B+C+D+E+F+G+H+I 9 Sum of the degree of affiliation plus mean score of the components of the essential and derived attributes, divided by the total number of components Essential score A+B+C+D+E+F+G 7 Sum of the degree of affiliation plus mean score between the components of the attributes first contact (B), longitudinality (C), coordination (D and E) and comprehensiveness available (F), of the services provided (G), added to the degree of affiliation (A) Derived score H+I 2 Sum of the mean of the attributes family approach (H) and community orientation (I) Score by attribute After consolidating the relative data of each attribute, the values are transformed on a continuous scale, ranging from zero (0) to ten (10) given by the expression [score obtained -1 (minimum value)] X 10/4 (maximum value) -1 (minimum value)

comprehensiveness provided (H) family approach (I) community orientation
Note: The degree of affiliation aims to identify the service or health professional (doctor/nurse) that serves as a reference for care, which is not considered an attribute of PHC but is used in the calculation of essential and general scores.
The outcome variable of this study (extracted from the VIGITEL evaluation module) was the standardized general score (if ≥ 6.6 or < 6.6). The explanatory variables (extracted from the VIG-ITEL questionnaire) can be described in three groups. The first one is the sociodemographic  In assessing the presence and extent of PHC attributes, according to the general score obtained, 19.61% (n = 171) of users evaluated with a score ≥ 6.6, and 80.39% (n = 701) gave a score < 6.6 ( Table 1). Table 1 describes the profile of users of PHC services, according to the general assessment score. We observed that most of those who best evaluated PHC (score ≥ 6.6) have low schooling, i.e., ≤ 8 years of study ( Table 2 shows PHC service users' assessment, according to behavioral risk factors for NCDs and self-reported health conditions. We observed that hypertensive (26.59%; 95% CI 21.33-32.60), diabetic (32.94%; 95% CI 23.11-44.52) and obese (31.23%; 95% CI 23.16-40.64) users are among those who best evaluated PHC (score ≥ 6.6). Table 3 shows the result found in the application of the multiple logistic regression model. In the crude model, we can see that users without health insurance are 3.26 more likely (95% CI 2.11-5.03) to report a high score (≥ 6.6) for PHC quality than the others. Less educated, that is, with less than eight years of study (95% CI 1.66-5.79), obese (95% CI 1.28-3.57), diabetic (95% CI 1.14-3.57), and hypertensive users (95% CI 1.14-2.53) are 3.10, 2.04, 2.08, and 1.70 more likely to report a high score, respectively. = In the model adjusted for confounding variables (age and gender), users without health insurance are 3.21 more likely (95% CI 2.08-4.96) to report a high score (≥ 6.6) for PHC quality than adults with health insurance, while less educated users (0-8 years of study) are 2.81 more likely (95% CI 1.48-5.32) to report high scores (Table  3). Considering self-reported health conditions, people with diabetes (95% CI 1.05-3.24) and obese individuals (95% CI 1.20-3, 24) are 1.84 and 1.97 more likely to report a high score, respectively. Regarding the age group, older adults (over 60 years old) are 1.19 more likely (95% CI 1.00-1.41) to report a high score for PHC quality than adults in other age groups. The outcome "arterial hypertension" lost statistical significance (p = 0.095) and did not show any difference after applying the adjusted model (Table 3).

Discussion
The population-based study built on telephone interviews presents the evaluation of PHC service performance from the perspective of users in Belo Horizonte, using the PCATool-Brasil short version 23 .
The study innovates by applying the PCATool to a population sample in Belo Horizonte by telephone interviews to assess PHC performance and its association with sociodemographic characteristics, self-reported health conditions, and behavioral risk factors for NCDs, which differs from most published studies. It identifies the score of evaluation of the attributes from the users' viewpoint and knowing the PHC service use profile and factors associated with use 20,31 . It is worth mentioning that studies that apply the PCATool and analyze the score obtained with the users' lifestyles and morbidity are still scarce in the country.
The analysis using the multiple logistic regression model showed that the general score was better evaluated by PHC service users and associated with elderly users (aged 60 and over), with low schooling, without a health insurance plan, and with behavioral risk factors for NCDs or self-reported diseases, such as diabetes and obesity.
Considering the instrument chosen in this study to assess the performance of PHC services in Belo Horizonte, in a systematic global review,    Prates et al. 32 searched for studies published from 2007 to 2015 on using the PCATool instrument from the user's perspective for the evaluation of PHC performance. They found that several countries used PCATool, such as Canada, Spain, Korea, and China. However, studies evaluating PHC from the perspective of users in Brazil are still scarce 32,33 .
The results indicate a predominance of older adults who best evaluated PHC (score ≥ 6.6). Evidence points out that older adults have more multimorbidity and consequently use health services more, especially PHC, for individual or group care, or even for the purchase of medications, thus creating a bond with the service and the teams, facilitating better care assessment 17,20,34,35 .
In a household survey to analyze the pattern of use of health services by older adults in public services in Guarapuava, state of Paraná, Pilger et al. 36 concluded that this population is a large user of health services. Dotto et al. 37 evaluated the orientation of PHC services and compared the quality of PHC between UBS and Family Health Units (FHU), according to older adults' use ex-perience, by employing the PCATool, in two districts of Porto Alegre, Rio Grande do Sul. They identified that most older adults (77.9%) used the UBS services and, regarding the quality of the services, they observed that 22.9% of older adults evaluated PHC with a high-quality score 34 .
Users with low education and without a health insurance plan also evaluated PHC services better, corroborating with other studies that indicate that less-educated people and without health insurance plan use PHC services more, as these are mostly dependent on the SUS 20 . PNS data showed that ESF coverage is higher among people with low schooling. It is worth mentioning that the results found show the potential of the PHC services' contribution to reducing health inequalities, promoting greater access to health care 15,31,38,39 . However, the study by Perillo et al. 20 records that 45.22% of users with health insurance also used PHC services, which reinforces the scope of these services.
The study by Augusto et al. 40 shows that older adults without a private health insurance plan living in the Metropolitan Region of Belo Horizon- te showed a better evaluation in the attributes of care coordination, first contact access, and com-prehensiveness, and a worse evaluation in community orientation. It also observed that very old adults, women, and higher education rated the service better. Those who reported greater use of the service and chronic conditions had a worse assessment of PHC. The authors concluded that worse health conditions and greater use of services are associated with a more negative perception of PHC attributes among older adults 40 . Araújo et al. 41 evaluated the quality of PHC care provided to older adults according to their perspective in a municipality in the metropolitan region of Natal (RN) and identified that the sociodemographic factors linked to vulnerability (lower-income, rural area, and older age) were positively associated with different attributes of PHC.
The positive evaluation of users with NCDs, such as diabetes and obesity, shows that PHC services play a fundamental role in NCD surveillance and monitoring risk factors since they seek to develop activities to prevent these diseases, promote health, and implement harm reduction. These users require continuous monitoring, should address complications with specialists, and obtain supplies. PHC plays an important role in articulating the points of care of the RAS, ensuring the principles of comprehensiveness and care coordination.
A study by Sala et al. 42 that evaluates the comprehensiveness attribute in PHC services from the perspective of users of health units in São Paulo showed a very favorable evaluation in the issue of the gateway, list of services, and coordination.
Nevertheless, it is worth emphasizing the importance of investments to strengthen PHC to reduce NCDs 7,43 effectively.
Studies indicate that the best assessment of diabetic users of PHC services may be related to these users' low individual demand. Also, a feeling of gratitude could prevent users from evaluating the services received more critically for fear of weakening the bond with the health team and limiting access to the care received or the purchase of supplies 44 .
A limitation of this study refers to a possible selection bias from the use of the landline telephone register, minimized with the use of weighting and post-stratification weights, adjusting the sample composition to the demographic features of the municipality's population. Choosing the PCATool-adult-Brazil short version from the user's perspective has known limitations. The first would be to use only the experience of the actors involved (in the case of this study, users) in care as an evaluation criterion, not incorporating, for example, the technical evaluation of the service provided. However, considering that the opinion of users of PHC services is important in the evaluation process of the service and that the telephone survey can be useful in collecting data and is a low-cost process, further studies can be carried out with the evaluation of healthcare professionals to complement the assessment. The second points out a limitation regarding the fact that this instrument was not developed for the analysis of scores by attribute and measured the presence and extent of the essential and derived attributes of PHC through the general score. However, the feasibility of using the full version or analyzing the short version for use in telephone surveys must be considered. Evaluation becomes an important instrument for decision-making by professionals, managers, and academics [45][46][47][48] , and should be incorporated into management, especially the local one.
This study presents a new possibility of using the evaluation of health services, especially PHC, through the telephone survey, which can be a proper monitoring strategy, capturing users' perspective, at a lower cost and faster. Also, the PCATool-Brazil instrument is important in evaluating the quality of PHC services, considering the structural and process aspects in health services and facilitating associations with clinical outcomes.

conclusion
The study innovates by using PCATool in a telephone survey and was useful in assessing the performance of PHC in Belo Horizonte from the perspective of users and its association with sociodemographic features, self-reported conditions, and behavioral risk factors for NCDs. Further studies are required to assess PHC performance from the user's perspective and present an analysis relating PHC performance assessment to clinical outcomes. The study proved that this is an innovative type of assessment that can be replicated nationwide and contribute to the management of services. It also has a negligible cost and can be rapidly applied. It allows comparability of the findings as it is a tool used worldwide with different versions validated for local contexts. RD Perillo and DC Malta worked on the study design, the writing of the paper, the analysis and interpretation of data, the literature review, and the final review of the text. RTI Bernal and KC Poças participated in the design of the study, in the analysis and interpretation of data, and the final review of the text. EC Duarte participated in the design of the study and the final review of the text. All authors approved the final version.

acknowledgments
The authors are grateful to the Ministry of Health, Health Surveillance Secretariat, for their financing through TED. DC Malta is grateful to CNPq for the research productivity grant.