Household Expenditures for Medicines and the Role of Free Medicines in the Brazilian Public Health System

Abstract

Objectives. We sought to investigate, across different socioeconomic groups, the proportion of household medicine expenses that were paid by households and the proportion paid by the Brazilian national health system.

Methods. We carried out a survey in Porto Alegre, Brazil, that included 2988 individuals of all ages. We defined 2 expenditure variables: “out-of-pocket medicines value” (the sum of retail prices of all medicines used by family members within the previous 15 days and paid for out of pocket) and “free medicines value” (a similar definition for medicines obtained without charge).

Results. In 2003, the Brazilian national health system provided, free of charge, 78% of the monetary value of medicines reported (79% in the bottom wealth quintile and 32% in the top 2 quintiles). The mean out-of-pocket expense for medicines was 6 times greater among the top wealth quintiles compared with those in lower quintiles, but free medicines constituted a 3-times-greater proportion of potential expenditures for medicines among the bottom quintile than among the top 2 quintiles.

Conclusions. Free provision of medicines seems to be saving substantial amounts of medicine expenditures for poor people in Brazil.

Increasing health care expenditures are a worldwide concern.¹ These increases are associated with several factors, including aging populations, demographic and epidemiological transitions, and new expensive medicines and technology.² Medicine expenses contribute significantly to health care expenditures and have been increasing at an even faster rate. In Canada, per capita annual medicine expenses rose more than 4-fold from 1985 to 2004.³ In Portugal, the percentage of medicine costs in relation to overall health care expenses doubled from 1980 to 1998.⁴

Health care expenses are also of significant concern in Brazil. Health care is the fourth largest category of household expenditures, after housing, food, and transportation expenses.^5,6 Among all health care expenses, medicines account for the largest share, particularly among the poorest members of the population.^5–8 According to national household surveys in Brazil, medicines represented 37% of health care expenses in 1995 to 1996⁵ and 41% in 2002 to 2003.⁶ Medicines accounted for 20% of health care expenses among families who belonged to the bottom wealth quintile.¹ From national data, it has been shown that the wealthiest families (15% of total) in Brazil consumed 3 times more medicines than did the poorest families (50% of total).⁹

The Brazilian national universal and integral health system (Sistema Único de Saúde, or SUS) is committed to supporting access to medicines. The Brazilian Ministry of Health spent US $618 million on medicines in 2003, and this expenditure increased to US $2.9 billion in 2006.¹⁰ Half of all medicines used by the population are obtained free of charge from the SUS.¹¹ The Family Health Program (Programa Saúde da Família, or PSF) is a primary health care strategy launched by the SUS in 1994. The founding principles of PSF are to work at the family level, to offer both curative and preventive care, and to actively seek those in need of attention through a multiprofessional team formed by a doctor, nurses, and community health agents.

To date, information is lacking about household spending on health care and medicines, especially on who is getting free medicines supplied by the SUS and how much less households spend because the SUS provides free medicines. Existing data from national population surveys do not provide information on types of medicines used, on indication (professional-prescribed vs self-medication) or on frequency,^5–7 or only represent specific population subgroups.¹² We aimed to investigate, across different socioeconomic groups, how much all households spent on medicines and how much value in the form of free medicines the SUS supplied free of charge to families. These 2 variables were analyzed both as absolute values and as proportions of household income and of total health care expenses.

METHODS

We carried out a cross-sectional study in Porto Alegre, Brazil, from July to September 2003. Porto Alegre, the Rio Grande do Sul state capital with nearly 1.3 million inhabitants,¹³ is one of the wealthiest cities in Brazil. However, a large proportion of Porto Alegre's population is poor and lives in the suburbs. The population covered by 62 PSF units included almost 143 000 people at the time of the survey. For this study, 56 PSF units that had been operating for at least 6 months were eligible. We employed a 2-stage sampling strategy, with areas covered by the PSF constituting the primary sampling units. From the 56 units, 45 were selected with probability proportional to size, and 20 households from each area were sampled.

All individuals living in the selected households were included in the study. Individuals older than age 13 years were interviewed face-to-face, and mothers of younger participants responded for them. Using a precoded questionnaire, we asked individuals about the use of all medicines during the past 2 weeks and requested to see the medicine's package and prescription.¹⁴

The study protocol was approved by the Federal University of Pelotas Medical School Ethics and Research Committee and by the Porto Alegre Municipal Health Secretariat. Written informed consent was obtained before each interview. For respondents younger than age 18 years, one of the parents signed on their behalf.

Variables

For each medicine reported by a respondent, we asked how the medicine was obtained and when. We also collected information on the amount (e.g., number of packages, dose, and package size) bought or obtained free of charge. On the basis of this information, we estimated monthly expenditures on medicines by using retail price tables for July, August, and September 2003. For compounded medicines, prices were estimated directly at the counters of random pharmacies.

When the medicine package was available (76%), we classified products into the following groups for price estimates: compounded medicines, generics, brand-name medicines with prices controlled by the government, freely priced brand-name medicines, and medicines produced by government-owned laboratories. We applied a 5% discount to estimated prices, corresponding to the average discount given by drugstores to customers at the time of the study.

For those medicines with no package available (24%), we estimated the price based on the commercial brand name given by the respondent. When the commercial brand name was not given, we estimated the price as that of the cheapest product available. To estimate the value of medicines obtained free of charge from the SUS, we used the same strategy, assuming that households would have incurred the expenses for such medicines had the medicines not been supplied by the SUS.

We defined 2 expenditure variables: (1) the monetary value of all medicines paid for out of pocket (referred to as “out-of-pocket medicines value”); this variable was calculated as the sum of the 2003 retail prices of all medicines used by family members that were paid for out of pocket; and (2) the monetary value of the medicines that were obtained free of charge (referred to as “free medicines value”), calculated as the sum of the 2003 retail prices of all medicines used by family members that were obtained free of charge. A third variable, total medicines value, represented the sum of out-of-pocket medicines' value and free medicines' value.

For each medicine reported, we obtained detailed information to classify the product according to pharmacologic group. All medicines were grouped according to the World Health Organization Anatomical Therapeutic Chemical Classification System in classification levels 1 and 2.¹⁵ Medicines were also classified according to whether they were for acute or continuous use. Medicines were classified as for continuous use when participants reported that they used the medicines routinely with no date to stop.

We assessed monthly family income as the sum of the incomes of each individual household member during the month prior to the interview. We classified households into socioeconomic status quintiles based on the National Economic Index (Indicador Econômico Nacional, or IEN), which is based on self-reported household assets.¹⁶ The sample was divided into quintiles according to the IEN classification for the Porto Alegre city population, with cutoff points established by the 2000 Brazilian census.¹³ Individuals in the first reference quintile belonged to the 20% poorest of the population in Porto Alegre, and those in the fifth reference quintile represented the 20% wealthiest of the city's population. The sample is representative of the population living in areas covered by the PSF but is not representative of the Porto Alegre population as a whole; therefore, socioeconomic distribution is not equally balanced and includes proportionally more individuals from the lower economic groups. In most analyses, we grouped quintiles 4 and 5 to increase statistical power; in such cases, we refer to “quintiles” instead of “quintile.”

We calculated family health care expenses by asking the household head about any expenses that members of the family had in the past month for health insurance, medical and dentist consultations, examinations, and any other health-related expenses. This value was added to the out-of-pocket expenses on medicines to generate the household's total health care expenses over the past 30 days.

Data Analysis

We carried out double data entry and consistency checks by using Epi Info version 6.04 (Centers for Disease Control and Prevention, Atlanta, GA, 2001). We used Stata version 10.0 (StataCorp, College Station, TX, 2005) for data cleaning and analyses. The main unit of analysis was the household (n = 869), although the total number of medicines mentioned by respondents (n = 3492) was used as the denominator in some analyses. The official exchange rate at the time of the data collection (August 2003) was Brazilian reais (R$) 2.90 = US $1.00.

The goal of the analysis was to calculate the mean values of medicines paid for out of pocket and of medicines provided free of charge by the SUS, at the household level. Because of the strong skewness in amounts spent for medicines, we did not directly calculate means from the population. Instead, we followed an econometric approach that adjusts for highly skewed distributions. We used a generalized linear model with a γ distribution and a log link function.

First, we set up the following model to find the monetary value of medicines used by each socioeconomic group:

where V is the monetary value of medicines in the household, seg1 represents the first socioeconomic quintile (Q1); seg2 represents the second socioeconomic quintile (Q2); seg4 represents the fourth and fifth socioeconomic quintiles (Q4 + Q5); and seg3 (Q3) is the reference group to which the other 3 were compared. The estimate of constant term α represents the logarithm of the mean value of medicines used by families in Q3. By exponentiating, we find the mean value of medicines used by households in Q3. Coefficients of seg1, seg2, and seg4 represent the difference between the mean value of the monetary value of medicines used by families in Q3 and those in Q1, Q2, and Q4 + Q5, respectively. By adding the difference between those mean values to the constant term α and exponentiating, we get the mean medicines values for each of the 3 remaining socioeconomic groups (Q1, Q2, and Q4 + Q5). We calculated medicines' values by socioeconomic group for medicines paid for out of pocket and for medicines obtained free of charge, for acute use and for continuous use. The same technique was used to estimate the monetary value of medicines as a proportion of the monthly family income:

where pinc is the monetary value of medicines as a proportion of the monthly family income. We did the same for the monetary value of medicines as a proportion of the value of total health care expenditures, as shown in equation 3:

where pval is the monetary value of medicines as a proportion of the value of total health care expenditure. We calculated medicines' value as a proportion of income and of total health care expenditures for medicines paid for out of pocket and for medicines obtained free of charge, for acute use and for continuous use. We repeated analyses after adjusting for overall household health status, a summary indicator created on the basis of each household member's answer to a single question on self-rated health status.

We also estimated out-of-pocket medicines value and free medicines value according to type of use (continuous vs acute), indication (professional-prescribed vs self-medication), whether part of the SUS list of essential medicines (yes vs no), and pharmacologic groups, by using the total number of medicines used by the participants as the denominator.

We stratified analyses by socioeconomic status and accounted for household clusters with Stata version 9 survey commands (StataCorp, College Station, TX, 2005). Results of the generalized linear model estimations showing the actual regression coefficients, instead of the means, are available upon request.

RESULTS

A total of 869 households were visited and 2988 individuals were interviewed. The nonresponse rate was 4.4%. The mean value of family expenses for health care over the month before the interview was R $89.43,¹ and this figure was directly associated with both socioeconomic level and medicines expenditures. There were 30 families with no income within the previous month, and 266 families with no health care expenditure within the same period.

As shown in Table 1, the mean overall value of out-of-pocket medicines was R $15.75; this value was more than 6 times greater for families in the top 2 wealth quintiles than for those in the bottom quintile (P < .001). For medicines reported to be used continuously, the expenses of families in the top 2 quintiles were 10.5 times greater than expenses of families in the bottom quintile (P < .001). The mean overall value of free medicines was R $31.35. The richest families received the least amount of free medicines (by value) compared with all other groups. Differences across groups were not as large as in the case of medicines paid for out of pocket but were still statistically significant (P = .01). Adjustment for household health status did not modify the associations described in Table 1 (data not shown).

TABLE 1.

Value of Medicines Paid for Out of Pocket and Obtained Free of Charge, According to Assets Index Quintiles: Porto Alegre, Brazil, 2003

		Quintiles of IEN16 Based on Porto Alegre Distribution
Variables	Overall, Mean (SD)	Q1, Mean (SD)	Q2, Mean (SD)	Q3, Mean (SD)	Q4 + Q5, Mean (SD)	Pa
Value of medicines paid for out of pocket, R$	15.75 (31.61)	5.69 (14.66)	13.69 (24.58)	20.87 (35.26)	36.34 (47.46)	<.001
Acute use	8.21 (18.09)	3.81 (11.27)	8.43 (18.07)	9.46 (18.05)	16.63 (26.13)	<.001
Continuous use	7.54 (23.72)	1.88 (7.36)	5.26 (14.26)	11.41 (31.52)	19.72 (37.98)	<.001
Value of medicines obtained free of charge, R$	31.35 (199.11)	21.84 (39.06)	35.31 (116.64)	56.48 (424.10)	17.05 (28.61)	.01
Acute use	11.25 (47.16)	8.93 (19.60)	15.30 (70.09)	13.15 (61.81)	8.00 (19.28)	.19
Continuous use	20.10 (193.93)	12.92 (33.32)	20.02 (94.35)	43.33 (420.69)	9.05 (22.43)	.02
Total medicines value, R$	40.07 (202.59)	27.53 (43.02)	49.00 (122.66)	77.35 (425.76)	53.39 (59.11)	<.001

Notes. IEN = National Economic Index (based on household assets); R$ = Brazilian reais. Exchange rate: R $2.90 = US $1.00 (August 2003).

^aBy the Kruskal–Wallis test.

Table 2 shows the proportion of the values of out-of-pocket medicines and free medicines in relation to monthly family income and monthly overall health care expenditures. The value of medicines paid for out of pocket represented 1.3% of the total income for the bottom-quintile families and 2.5% for families in the top 2 quintiles (P = .003). The value of medicines obtained free of charge represented 6.2% of income among families in the bottom quintile and 1.3% among families in the top 2 quintiles (P < .001). The value of medicines paid for out of pocket represented 26% of all health care expenditures among the families in the bottom wealth quintile and 39% among families in the top 2 quintiles (P = .009). The value of medicines obtained free of charge represented 65% of total health care expenses among the bottom-quintile families and 23% among families in the top 2 quintiles (P < .001).

TABLE 2.

Proportion of Value of Medicines Paid for Out of Pocket and Value of Medicines Obtained Free of Charge Relative to Family Income and Overall Monthly Health Care Expenses, by Assets Index Quintiles: Porto Alegre, Brazil, 2003

	Monthly Family Income						Monthly Health Care Expenditures
	Quintiles of IEN16 Based on the Porto Alegre Distribution						Quintiles of IEN16 Based on the Porto Alegre Distribution
Variables	Total, %	Q1, %	Q2, %	Q3, %	Q4+Q5, %	Pa	Total, %	Q1, %	Q2, %	Q3, %	Q4+Q5, %	Pa
Value of medicines paid for out of pocket	1.91	1.28	2.22	2.26	2.50	.003	31.45	25.61	33.18	33.04	39.20	.009
Acute use	1.13	0.82	1.51	1.15	1.27	.07	20.13	18.08	23.43	19.33	20.45	.37
Continuous use	0.78	0.46	0.71	1.12	1.23	.008	11.32	7.53	9.75	13.71	18.75	.001
Value of medicines obtained free of charge	5.12	6.22	6.14	4.78	1.34	<.001	49.27	64.61	50.42	43.25	22.95	<.001
Acute use	2.33	2.63	3.48	1.52	0.77	.002	25.56	32.86	27.54	21.64	12.05	<.001
Continuous use	2.79	3.59	2.66	3.26	0.57	<.001	23.71	31.76	22.88	21.61	10.89	<.001

Note. IEN = National Economic Index (based on household assets).

^aBy the Kruskal–Wallis test.

As shown by comparing the data in Table 1, the SUS provided free of charge 78% of the total monetary value of medicines used by this population. In Table 3, our data show that 73% of the medicines for continuous use were provided free of charge, whereas the equivalent proportion of medicines for acute use was 58%. In terms of medicine prescription, 70% of those medicines prescribed were provided free of charge, whereas only 30% of the medicines used for self-medication were provided free of charge. Of the products listed on the SUS list of essential medicines, 86% were provided free of charge. The pharmacologic groups most often provided free of charge were medicines targeting the blood and blood-forming organs (88%) and anti-infectives for systemic use (85%), whereas medicines for the respiratory system (37%) and dermatological products (39%) were less frequently subsidized by SUS.

TABLE 3.

Values (R$) of Medicines Paid for Out of Pocket and Value of Medicines Obtained Free of Charge, by Medicine Characteristics: Porto Alegre, Brazil, 2003

Variable	Medicines Used, No. (%)	Value of Medicines Paid for Out of Pocket, Mean (SD)	P	Value of Medicines Obtained Free of Charge, Mean (SD)	P	% of Total Value That Is Subsidized
Use (n = 3401)			.03a		.01a
Acute	1984 (56.9)	3.8 (9.4)		5.2 (31.2)		57.8
Continuous	1502 (43.1)	4.7 (12.6)		12.5 (108.9)		72.7
Indication (n = 3398)			<.001a		<.001a
Professional-prescribed	2759 (79.2)	4.4 (12.0)		10.1 (84.6)		69.7
Self-medication	724 (20.8)	3.2 (5.1)		1.4 (3.9)		30.4
Part of the list of essential medicines (n = 3368)			<.001a		.65a
No	1113 (32.2)	10.1 (16.6)		7.3 (117.3)		42.0
Yes	2340 (67.8)	1.4 (4.6)		8.9 (43.8)		86.4
Pharmacologic group (n = 3357)			<.001b		<.001b
Alimentary tract and metabolism	320 (9.3)	5.0 (12.7)		8.0 (16.7)		61.5
Blood and blood-forming organs	147 (4.3)	1.4 (4.4)		9.8 (81.7)		87.5
Cardiovascular system	569 (16.6)	4.3 (13.8)		8.6 (31.2)		66.7
Dermatologicals	156 (4.5)	6.4 (13.3)		4.0 (7.7)		38.5
Genitourinary system and sex hormones	286 (8.3)	4.0 (9.0)		4.0 (8.6)		50.0
Systemic hormonal preparations (excluding sex hormones and insulins)	46 (1.3)	7.8 (11.9)		10.5 (19.4)		57.4
Antiinfectives for systemic use	235 (6.8)	5.0 (11.6)		27.8 (126.0)		84.8
Musculoskeletal system	271 (7.9)	6.5 (15.9)		6.9 (10.9)		51.5
Nervous system	919 (26.8)	4.1 (12.1)		4.1 (20.5)		50.0
Antiparasitic products, insecticides, and repellents	40 (1.2)	2.9 (6.5)		5.7 (7.8)		66.3
Respiratory system	398 (11.6)	5.5 (21.7)		3.2 (6.6)		36.8
Sensory organs	48 (1.4)	5.0 (11.0)		5.3 (11.1)		51.5

Notes. IEN = National Economic Index (based on household assets)¹⁶; R$ = Brazilian reais. Exchange rate: R $2.90 = US $1.00 (August 2003).

^aBy the t test for unequal variances.

^bBy the Kruskal–Wallis test.

Figure 1 shows the relative proportions of values of out-of-pocket and free medicines, stratified by quintiles of socioeconomic status. The higher the socioeconomic status, the lower is the contribution of medicines provided free of charge.

FIGURE 1.

Relative proportion of medicines paid for out of pocket and obtained free of charge, according to assets index quintiles: Porto Alegre, Brazil, 2003.

Notes. IEN = National Economic Index (based on household assets)¹⁶; POA = Porto Alegre.

DISCUSSION

We estimated the mean amount of money spent on medicines per month in poor areas of Porto Alegre, and the amount families saved because of the free distribution of medicines by the SUS.

Previous studies^17,18 have shown that overall medicine utilization and the likelihood of paying for drug treatment¹⁷ are directly associated to socioeconomic level. In our sample, families from the top 2 wealth quintiles spent, on average, 6 times more money on medicines than did families from the bottom quintile. This ratio is higher than that observed by Andrade et al.⁷ and by the Brazilian data of the World Health Survey,⁸ which found ratios of 3.1 and 3.9, respectively. Differences in out-of-pocket expenditure ratios may be explained by the greater availability of free medicines to the poor in our sample.

Compared with the 1.9% of income spent on medicines in the present study, previous research found higher shares of family income spent on medicines, such as 5% in the 1995 to 1996 POF (the Family Budget Survey conducted by IBGE, the Brazilian Institute of Geography and Statistics) and 9% in the 1998 PNAD (Pesquisa Nacional por Amostra de Domicílios, a national household survey also conducted by IBGE).⁵ The fact that our sample was poorer than representative national samples may help explain these differences. Another likely explanation is that the 2 national studies were carried out at the early stages of PSF implementation, whereas ours was carried out almost 10 years after PSF was launched. Methods and regional coverage of the 3 studies were also different, thus hampering comparability.

The importance of free provision of medicines was quite evident in our study. The poorest quintile of the population relied on free medicines for 80% of their needs, compared with just 20% among those in the richest quintile. For families in the poorest 2 quintiles in our sample, free provision of medicines accounted for an average of 6% of family income, likely reducing the occurrence of excessive spending and possible financial difficulties for the families. In contrast, free medicines accounted for, on average, only 1.3% of family income for the 2 richest quintiles.

Free medicines mean a reduction in household expenditures and greater access to medicines; if free medicines were not available, families would either have to spend more or forgo taking some medications. Increasing the scope and coverage of the free medicines program could have an even greater impact on the budgets of poor families because poor families are more likely than rich families to use the public health system and consequently, more likely to get medicines free of charge.

The proportion of medicines obtained free of charge varied according to pharmacologic groups. On the one hand, the highest proportion of free use in any therapeutic class was for antibiotics, which is desirable because of the high costs of such medication and the risks associated with incomplete treatments. On the other hand, less than 40% of dermatological products and medicines used for the respiratory system were provided free of charge. The proportion of medicines obtained free of charge was higher for products used continuously than for those for acute use. This finding was expected, because one of the SUS priorities is to provide medicines for people with diabetes, hypertension, and other prevalent chronic diseases, for whom more consistent utilization of medicines may help prevent complications and enhance quality of life. Free medicines are particularly important in such cases because expenses for continuously used medications could compromise family income for long periods.

Interestingly, there was no statistically significant difference in the value of medicines obtained free of charge between medicines that were part of the list of essential medicines and those not a part of that list. It was expected that the mean value of medicines from the list would be much higher than the value of medicines not on the list. However, if an individual needs a medicine not on the essential list (which often happens in the case of diseases for which treatments are very expensive), the individual is able to request the medicine from the government. In our sample, this seems to have been the case, as demonstrated by the high standard deviation of the value of medicines obtained free of charge for those not included on the list of essential medicines.

Some limitations inherent to data collected through interviews should be considered. To avoid recall bias, we used a short recall period, or 15 days, which is compatible with most studies in the field.¹⁴ However, 15-day expenditures may not correctly represent average expenditures over time. To avoid underreporting of medicine consumption and to make it possible to correctly classify medicines into pharmacologic groups, we asked respondents to present the packaging of the medicines used. A challenging methodological aspect was to estimate the prices of the medicines reported by participants. We used retail price lists from the time of the survey, but we are aware that actual prices can vary significantly from one pharmacy to another and that discounts are frequently employed in the pharmaceutical market in Brazil. Although standardized prices may not reflect actual prices paid, standardized prices allowed estimation across socioeconomic groups of the relationships between values of medicines paid for out of pocket and those obtained free of charge. Another limitation was the potential information bias of self-reported income data, especially among the richest households, who tended to underreport.¹⁶

Our study also required the following assumptions. We assumed that medicines not obtained free of charge would otherwise be purchased, although much of this care might have been forgone, especially among the poor. We estimated the value of medicines obtained free of charge by using the cheapest medicines available, which were more likely to be purchased by poorer individuals than were the more expensive brand-name products.

We used means instead of medians to estimate average monthly health care and medicine expenditures in the population. Because there were many households with zero health care expenditures, the use of medians would have resulted in very low estimates of spending. The short data collection period (15 days) led to instability in the expenditures reported, with a few households spending very high amounts. As a consequence, we opted to use an econometric approach that adjusts for highly skewed distributions, but standard deviations presented in our tables are still high, particularly in quintiles 2 and 3, in which some families had very high monetary values of medicines obtained free of charge.

In our study, we gave emphasis to the out-of-pocket expenses on medicines and the savings to families resulting from the free provision of medicines by the Brazilian health system. Although absolute values of free medicines were not markedly different across socioeconomic groups, the proportions that these values represent in relation to total family income were much higher among poor families. Along with other economic incentives, free provision of medicines in Brazil may improve access and avoid high expenditures for medicines for poor people. Future studies will benefit from exploring why some poor families still pay for their medicines out of pocket, despite the government's free provision policy.