The study uses a descriptive research model, with transversal evaluation among the participants (July 2018), and longitudinal analysis of costs (12 months prior and 12 months after [July 2017 to June 2019] the date of the transversal evaluation) (Fig. 1). The starting point for sample selection was the medical records of the Regional Hospital (RH) in the city of Presidente Prudente (∼ 200,000 inhabitants), located in the Western region of the State of São Paulo/Brazil, which offers referral care of medium and high complexity, totally free of charge, through the Brazilian Unified Health System (i.e., the SUS health system, which provides free of charge health care for the population using public sector resources), to 45 municipalities in the Western region of the State, serving 447.36 patients/day on average.

Fig. 1.

Study flow diagram and sample selection.

(0.12MB).

The RH was asked to select all medical records of patients who were admitted for cardiovascular reasons (i.e., category I of the International Statistical Classification of Diseases and Related Health Problems – ICD) in the 12 months prior to the first assessment and who met the following inclusion criteria: i) aged between 30 and 65 years (age group with a high prevalence of chronic diseases in Brazil)22 ; ii) used the services offered by the Brazilian Unified Health System due to CVD in the previous 12 months; iii) resident of the city of Presidente Prudente, where the researchers had permission to obtain information regarding healthcare services at primary (PHC), secondary (SHC), and tertiary (THC) levels. A total of 1200 medical records fitting those criteria were identified.

Medical records meeting the criteria mentioned above were randomly selected, with simple random sampling, using a statistical software, until the minimum calculated sample size was reached. Due to the absence of data about the relationship between physical activity and healthcare costs using real world data (assessed directly from the medical records) for adults between 30 and 50 years of age, we ran the sample size calculation using the closest information available in the literature, in this case, for adults over 50 years of age. Thus, to calculate the sample size we used an equation considering a relationship between physical activity and healthcare costs in adults older than 50 years (correlation results: r = −0.23, p-value= 0.011),23 statistical power of 80% and alpha error of 5%.24 An initial minimum sample size of 147 people was estimated. Finally, when adding 40% for losses (estimated from previous studies involving analyses of public health costs),12 a minimum of 206 people was required to participate in the study.

From the list of 1200 patients provided by the RH, random draws in blocks of 300 patients per draw were performed using STATA software, version 16.0. It was necessary to perform 3 draws of 300 patients, with only 194 patients contacted in the final draw to reach the minimum sample size calculation (Fig. 1). Among the 794 phone numbers verified by the researchers, 278 were for patients who participated in the study (among which, none of the participants presented missing data). Sample losses are shown in Fig. 1.

The researchers contacted patients at the beginning of the week, scheduling them for another day in the same week according to their availability. Therefore, in the last week of evaluation, when the minimum number stipulated by the sample size was reached, the researchers continued evaluating all scheduled patients, leading to a total sample of 278 patients.

Eligible patients were contacted by phone and invited to participate in face-to-face interviews and assessments. Those who agreed to participate visited the assessment site (Universidade Estadual Paulista) and signed the consent form. The research proposal was approved by the Research Ethics Committee of the Universidade Estadual Paulista, Presidente Prudente Campus, Brazil (CAAE: 82,767,417.5.0000.5402).

The monetary values were estimated using the perspective of the SUS, based on the sum of resources used directly in the treatment of the patient, micro-costing (bottom up) approaches were used to estimate costs.25

Healthcare services/procedures (PHC, SHC, and THC) listed in the medical records of each patient were collected, in a time horizon of 24 months (12 months prior and 12 months after the date of the initial assessment), this being the best methodological design found by the researchers to obtain a longer follow-up time on the cost variable. There is an absence of widely accepted methodologies to accurately estimate healthcare expenditures. Therefore, with regard to the costs of PHC, we adopted the methodology described in the previous studies,21,23,26 which briefly consists of initially gaining permission from the local municipality health secretariat, after which the researchers obtain access to the medical records of the patients involved in the research and collect the registered information regarding medical consultations, medication dispensed, specific consumables, and diagnostic services. In addition, the Municipal Secretary of Health provided information about attendance services and utility bills of the healthcare unit.21,23,26

Costs with healthcare services relating to PHC, including medical consultations, attendance services, utility bills, medication dispensed, and diagnostic services, are described below:

• i)

  Medical consultations: costs were retrieved from the SUS System Management Procedures Table (SIGTAP), provided by the Ministry of Health;

• ii)

  Attendance services (i.e., scheduling, medication dispensing, management, etc.): costs were calculated using the daily salary rate of the professionals involved in the services provided (monthly salary divided by 30 days) and the average number of patients seen daily (daily salary rate divided by the number of daily visits);

• iii)

  Utility bills of the healthcare unit (electricity, water, and telephone): costs were calculated using the average of the last three months for each utility bill, divided by 30 days. The utility bill value was then divided by the number of patients seen daily.

• iv)

  Medication dispensed, specific consumables, and diagnostic services (laboratory tests and others): costs were calculated using the specific cost of each standard procedure multiplied by the number of procedures performed.

Information regarding salaries, costs of laboratory procedures, consumables, medications, and utility bills were provided by the Municipal Secretary of Health.

Services related to SHC and THC were estimated considering emergency visits, diagnostic imaging and laboratory tests, hospitalizations, surgeries, hemodialysis, as well as medication and consumables used for inpatient care. Costs related to these services/procedures were, in their entirety, provided by the RH, without the need for data collection and calculations performed by the researchers.

In the transversal evaluation, HPA level was measured using the Baecke questionnaire.27 This questionnaire was previously validated for Brazilian Portuguese and recommended for epidemiological studies in Brazil.28 The instrument evaluates HPA in three domains: occupational (8 questions), sports (4 questions), and leisure-time (4 questions). The sum of the three domain scores comprises total level of HPA. The possible HPA score ranges from 3 to 15. The higher the total score obtained with the sum of the three domains, the greater the level of HPA.

All variables described in this session were collected in the transversal evaluation. The patients reported sex and age during the interviews. Information regarding level of education was self-reported. Healthcare services/procedures listed in the patient's medical record were coded according to the International Statistical Classification of Diseases and Related Health Problems (ICD), as follows: I10-I15 = Hypertensive diseases; I20-I25 = Ischemic heart disease; I26-I28 = Pulmonary heart disease and diseases of pulmonary circulation; I30-I52 = Other forms of heart disease; I60-I69 = Cerebrovascular diseases; I70-I79 = Diseases of arteries, arterioles, and capillaries; I80-I89 = Diseases of veins, lymphatic vessels, and lymph nodes, not classified elsewhere.

Diabetes, dyslipidemia, arterial hypertension, and obesity were considered as comorbidities associated with CVD.2 The sum of the comorbidities ranged from 0 to 4 and was considered as a confounding variable in the relationship between HPA and healthcare costs. In a face-to-face interview, the patient was asked about having a diagnosis of and/or any medication taken for these comorbidities.

Height was measured using a stadiometer (Sanny, American Medical do Brasil Ltda, Brazil, 0.1 cm precision). The weight and body fat percentage (%BF) were verified by the device (InBody, model 230), which estimates the percentage of body fat through bioelectrical impedance analysis. To assess %BF, participants were instructed not to drink caffeine, to empty their bladder, and not to eat for two hours before the analysis, to wear light clothing, and no jewelry, accessories, or belts, and not to practice strenuous exercise for 24 h before the analysis. Obesity status was classified according to the recommendations of the American College of Sports Medicine29 for % BF.

Data normality was verified by the Kolmogorov-Smirnov test. Descriptive statistics consisted of mean, standard deviation (SD), median, and interquartile range (IQR), and inferential statistics were represented by the 95% confidence interval (CI).

The questionnaire used in this research to measure HPA does not have a cut-off score for classifying individuals. For this reason, for statistical analysis, the HPA score was divided into percentiles: ≥P75 (more physically active) and

Comparisons between groups were performed using Student's t (comparison between groups30,31 comparisons involving health care costs as the dependent variable are expressed as mean and 95% CI, while the comparisons involved non-parametric tools.

Due to the asymmetric distribution of the dependent variable, the quantile regression multivariate model was adopted in this study.32 This regression analysis focuses on how predictors relate to the occurrence of extreme values (e.g. 10th, 50th, and 99th percentiles), and does not require any assumption about distribution (different from traditional mean regression models).32 Quantile regression was used to measure the relationship between HPA and health care costs considering the simultaneous insertion of different quartiles (quartiles 25, 50, and 75). In addition, the analysis considered the standard errors in bootstrap, which in turn imply the robustness of the model. Initially, potential confounders for the model were identified in the literature,33 including variables related to the outcome. The models were simultaneously adjusted by comorbidities (diabetes, dyslipidemia, arterial hypertension, and obesity), age, and sex. The relationships of all these variables with the outcome were tested and reached the previously established level of significance, set at 0.05. All analyses were performed using Stata statistical software (version 16.0).