Access to pediatric cancer care in Brazil : mapping origin-destination fl ows

OBJECTIVE: To analyze fl ows of travel between place of residence and health care services by children and adolescents with cancer. METHODS: The fl ows of travel between place of residence and the health care service for children and adolescents receiving care in Brazil’s Unifi ed Health System (SUS) were monitored between 2000 and 2007. The unit of analysis was the health care district. The geographical information system data and network methodology, by type of treatment received (chemotherapy and radiotherapy) and hospital admissions were used. RESULTS: The SUS made 465,289 authorizations for chemotherapy, 29,151 for radiotherapy and 383,568 for hospital admissions for the treatment of children and adolescents with a diagnosis of cancer. The dominant fl ow formed 48 networks for chemotherapy, 53 for radiotherapy and 112 for hospital admissions. Most of the volume of treatment occurred in the health districts of Brazil’s 12 largest cities (with strong links between them and each having an extensive area of direct infl uence accompanying the structure of the Brazilian urban system. CONCLUSIONS: Identifying the networks formed by utilization of SUS facilities providing care for children and adolescents with cancer shows that overall most patients are covered by the existing networks. However, about 10% of travel occurs outside the dominant structure, indicating the need for alternative regionalization. These results show the importance of planning the distribution of services to meet the population’s needs. DESCRIPTORS: Child. Adolescent. Neoplasms. Cancer Care Facilities, supply & distribution. Oncology Service, Hospital. Health Services Accessibility. Geographic Information Systems. Unifi ed Health System. 2 Cancer care for children and adolescents Grabois MF et al Cancer in childhood comprises a set of relatively rare pathologies, but is the principal cause of death from illness among children and adolescents in high income countries.11,14,17 The mean annual incidence for all types of cancer in this age group, under 20, is of 18.8 cases/100,000 people – year in the United States.23 In Brazil, in 14 of its population based records, the median incidence of cancer in childhood is 154.3/thousand,6 representing the fourth and fi fth most common causes of death in those aged under 18 for females and males respectively, and the most common cause of death from illness in over 5s.a The capacity for diagnosing and treating childhood cancer 50 years ago was rudimentary, and survival rates were < 10%. Today, more than 70% of children diagnosed with cancer in high income countries survive and the majority are considered cured thanks to advances in diagnosis and appropriate treatment of the disease in its early stages. However, the majority of people live in low and middle income countries, where the rates of cure for children are disappointing.17 Late diagnosis of cancer due to diffi culties accessing medical care, abandoning treatment or lack of appropriate support are all motives which contribute to this discrepancy. Studies report that early diagnosis and immediate access to specialist centers are critical for appropriate treatment and are associated with higher survival rates.9,11,12,14,16,17 Access to health care services is one of the essential elements in caring for cancer. Geographical accessibility, one of the components of access, is infl uenced by the distribution of resources and services. Only citizens who are able to travel to the locations where these resources and services can be found are in a condition to make use of them, as accessibility is not the same everywhere in the country.22 Service users’ displacements are best measured in terms of their cost in time and money. Lacking any other more specifi c measure, distance can be used as an indicator of ‘diffi culty’.20 Brazil is extensive and very diverse and “access to cancer diagnosis and treatment is not suffi cient as it is centralized in the state capitals and in the more economically developed states”.18 Evaluating the location of health care services should include the conditions of accessibility for the local population. Examining networks is one way of investigating relationships between the location of residence and of the health care service and highlighting problems of access, identifying INTRODUCTION areas with few options and bottle necks or the need for regionalized alternatives, which show the need for care policies to be re-thought. The Unifi ed Health System (SUS), in Brazil was created in 1988. Its basic principle is to guarantee access to health care to all citizens. Almost all hospitals, including public, private and university hospitals form part of the SUS, with different fi nancial support agreements. The Ministry of Health has been organizing oncological care in Brazil since 1990. It created authorization for highly complex oncology treatments (AHCP) and established the National Oncological Care Policy in 2005, which includes promotion, prevention, diagnosis, treatment, rehabilitation and palliative care. In addition, the “need to structure a network of regionalized, hierarchized services which guarantee comprehensive care for the population, as well as access to consultations and tests for diagnosing cancer” was emphasizedb and the minimum criteria for registering with the Centers for Highly Complex Oncological Procedures (CACON) were defi ned.c A Brazilian study on children with cancer showed similarities in the spatial patterns of chemotherapy and radiotherapy treatments paid for by the SUS. There was a greater concentration of chemotherapy in the CenterSouth of Brazil, and reduced access to treatment for those in the North and the more peripheral regions of the Northeast, compared with residents in the Southeast, South and Midwest.13 Differences in childhood cancer incidence rates between high income countries and the majority of low and middle income countries are related, with lower rates in the latter, suggesting under reporting and under diagnosis of the disease due to insuffi cient access to specialist services.3,4,16,17,24 This study aimed fl ows of travel between place of residence and health care services by children and adolescents with cancer.

Cancer in childhood comprises a set of relatively rare pathologies, but is the principal cause of death from illness among children and adolescents in high income countries. 11,14,17The mean annual incidence for all types of cancer in this age group, under 20, is of 18.8 cases/100,000 people -year in the United States. 23 Brazil, in 14 of its population based records, the median incidence of cancer in childhood is 154.3/thousand, 6 representing the fourth and fi fth most common causes of death in those aged under 18 for females and males respectively, and the most common cause of death from illness in over 5s. a The capacity for diagnosing and treating childhood cancer 50 years ago was rudimentary, and survival rates were < 10%.Today, more than 70% of children diagnosed with cancer in high income countries survive and the majority are considered cured thanks to advances in diagnosis and appropriate treatment of the disease in its early stages.However, the majority of people live in low and middle income countries, where the rates of cure for children are disappointing. 17Late diagnosis of cancer due to diffi culties accessing medical care, abandoning treatment or lack of appropriate support are all motives which contribute to this discrepancy.Studies report that early diagnosis and immediate access to specialist centers are critical for appropriate treatment and are associated with higher survival rates. 9,11,12,14,16,17cess to health care services is one of the essential elements in caring for cancer.Geographical accessibility, one of the components of access, is infl uenced by the distribution of resources and services.Only citizens who are able to travel to the locations where these resources and services can be found are in a condition to make use of them, as accessibility is not the same everywhere in the country. 22Service users' displacements are best measured in terms of their cost in time and money.Lacking any other more specifi c measure, distance can be used as an indicator of 'diffi culty'. 20azil is extensive and very diverse and "access to cancer diagnosis and treatment is not suffi cient as it is centralized in the state capitals and in the more economically developed states". 18Evaluating the location of health care services should include the conditions of accessibility for the local population.Examining networks is one way of investigating relationships between the location of residence and of the health care service and highlighting problems of access, identifying

INTRODUCTION
areas with few options and bottle necks or the need for regionalized alternatives, which show the need for care policies to be re-thought.
The Unifi ed Health System (SUS), in Brazil was created in 1988.Its basic principle is to guarantee access to health care to all citizens.Almost all hospitals, including public, private and university hospitals form part of the SUS, with different fi nancial support agreements.
The Ministry of Health has been organizing oncological care in Brazil since 1990.It created authorization for highly complex oncology treatments (AHCP) and established the National Oncological Care Policy in 2005, which includes promotion, prevention, diagnosis, treatment, rehabilitation and palliative care.In addition, the "need to structure a network of regionalized, hierarchized services which guarantee comprehensive care for the population, as well as access to consultations and tests for diagnosing cancer" was emphasized b and the minimum criteria for registering with the Centers for Highly Complex Oncological Procedures (CACON) were defi ned.c A Brazilian study on children with cancer showed similarities in the spatial patterns of chemotherapy and radiotherapy treatments paid for by the SUS.There was a greater concentration of chemotherapy in the Center-South of Brazil, and reduced access to treatment for those in the North and the more peripheral regions of the Northeast, compared with residents in the Southeast, South and Midwest. 13Differences in childhood cancer incidence rates between high income countries and the majority of low and middle income countries are related, with lower rates in the latter, suggesting under reporting and under diagnosis of the disease due to insuffi cient access to specialist services. 3,4,16,17,24is study aimed fl ows of travel between place of residence and health care services by children and adolescents with cancer.

METHODS
The health care service for children and adolescents receiving care in Brazil's Unifi ed Health System were monitored between 2000 and 2007.
Secondary data were examined, which were taken from authorizations for hospital admissions (AHA) and authorizations for highly complex procedures (AHCP) in oncology: for chemotherapy (CT) and radiotherapy (RT).The analysis was based on data from the Health Care Information Systems (SIS) and from the available programs: TabWin and TerraView.The data were tabulated with TabWin, version 3.6, developed by the Ministry of Health, which includes a routine for creating tables of origin-destination.Analysis of the travel fl ows was done using the Geographical Information System (GIS): TerraView version 3.2.1, which imported the tables generated in Tabwin and had a special module for classifying the fl ows.d The population of the study included children and adolescents (aged up to 19 years old) with malignant neoplasia, covered by codes C00 to C97 in the International Classifi cation of Diseases (ICD-10). 25he data from 2000 to 2007 were aggregated and the mean of the 2003 to 2004 populations used to calculate rates.The aggregated data for an eight-year-period were analyzed, as childhood cancer is a rare event.The geographical unit of analysis was the health district, which numbered 352 in 2008.Health districts are heterogeneous, especially in terms of size, varying between two and 54 municipalities (as well as the Federal District, which was one health district), and population size. 13The health districts refl ected the logic of comprehensive health care in linked networks constituting a unit of analysis appropriate for this study.
Hospitalizations referred to clinical and surgical procedures with neoplasia as the principal ICD in hospitals regardless of whether or not they were equipped for highly complex procedures.e As the same patient could have been admitted to a hospital several times, the total value referred to the number of hospitalizations and not to the number of patients.
Authorization for CT is valid for three months.The programmed chemotherapy may, depending on the evolution of the clinical case, continue or be substituted or suspended.AHCP refers to the monthly costs of a treatment scheme and not to the cost of a cycle.For example, treating adolescents and children with acute lymphocytic leukemia (ALL) -the most common cancer among children -the protocol of the Brazilian Group for the Treatment of Childhood Leukemia (BGTCL-99), lasts 24 months, 8 i.e., in order to complete the treatment the patient needs eight AHCP.Authorization for RT has a maximum validity of three months to conclude the planned treatment and cannot be repeated on the same target-organ.Consequently, an AHCP is necessary for RT of this target organ.f The travel fl ow of children and adolescents with cancer between health districts for CT, RT and hospitalizations was mapped.The maps of the fl ow show the movement between points (or nodes) of origin and destination, i.e., the journey between two points and the connection between them, forming interconnected structures named networks.Each link corresponds (in kilometers) to the distance in a straight line between the regional health care centroids.The links which make up a network can be classifi ed according to the volume and direction of the fl ows between different nodes.To identify the framework of network connections, the method of dominant fl ow, proposed by Nystuen & Dacey, g was used, which establishes a hierarchy among the nodes of the network based on three properties of the fl ow: "(1) a city is independent if its greatest fl ow is directed towards a smaller city, and subordinate if the fl ow goes towards a larger city; (2) transitivity: if A is subordinate to B, and B is subordinate to C, then A is subordinate to C; and (3) a city may not be subordinate to any of its subordinates".
The dominant fl ow is defi ned as the greatest fl ow from each origin, and is the base on which the network is structured.Other fl ows establish internal and external relationships for each network.To avoid loss of information caused by only examining the dominant fl ows, the connections were classifi ed according to the system used by Oliveira et al, 20 based on the original proposal by Rabino & Occelli.h The fl ow is named "local connection" when origin and destination coincide.The method establishes hierarchy between the nodes in the networks formed, which are classifi ed in different levels and types of relationships, according to the degree of "closure" or interconnection between the networks.Hierarchical fl ows accompany the structure of the network and transversal fl ows connect different networks or different branches of the same network.The fl ows may be horizontal, when linking nodes at the same level; ascendant, when they link to a higher level node; or descendant, when linking to a node at a lower level.The fl ows are called "short circuits" when they jump the closest level.The classifi cation combines the dimensions of hierarchy and direction.d The nodes were represented by the health district in which the patient was resident (origin) and the health district where health care was provided (destination), and were connected by links.For each network investigated, the matrix of fl ows is composed of the origin-destination nodes (regional health care centroids), with the cells containing the number of authorizations (hospitalizations or chemotherapy or radiotherapy procedures) The research was approved by the Committee for Ethical Research of the Escola Nacional de Saúde Pública Sergio Arouca, process nº 0248.0.0.031.000-10.

RESULTS
There were 465,289 AHCPs for CT, 29,151 for RT and 383,568 hospitalizations for treating children and adolescents diagnosed with cancer in the SUS between 2000 and 2007.There were records of CT for patients resident in all health districts, even if the treatment was carried out in a different district.
Around 60% of the volume of CT and RT care was provided locally (Table ), i.e., the majority of children and adolescents had access to medical care in the same health district in which they lived, with half of this care taking place in the Southeast.The links relating to the dominant fl ow represented 88.6% for CT and 86.1% for RT and conformed to 48 and 53 networks respectively.The majority of procedures (67.9% CT and 68.3% RT) were on patients who lived within a 100 km radius of the care-providing service.The median distances covered were around 230 km for CT and for RT (Table ).
Figure 1 shows the dominant fl ows for CT and RT.The majority of networks in the North and Northeast were polarized around the state capitals, which served the territory of their states.Some places showed a different pattern.The network in Goiania, the geographical center of Brazil, went beyond the state and received fl ows from 42 regions, covering a large area with low demographic density.The opposite pattern was found in the Southeast, with 20 networks, twelve of which were in the state of Sao Paulo.A mixed, regionalized pattern was found in Rio Grande de Sul and Minas Gerais, without crossovers or overlaps.
The transversal fl ows represent a large number of the links, although they correspond to a smaller volume of fl ows (9.4% of the CT fl ows and 11,3% of RT).These fl ows are spread out across the whole country and cover greater distances: median of almost 500 km for CT and around 360 km for RT (Table ).In Figure 2, the transversal displacements are shown.These fl ows provoked an enormous entanglement, showing evidence of problems in the organization of the network.For the same population and period, the volume of patients treated with RT was lower than that of those treated with CT.
Almost 68% of the volume of hospitalizations occurred locally (Table ), i.e., in the health district in which the patient lived.Around 40% of this care took place in the Southeast and 48.3% in the 12 largest cities in Brazil.
The pattern of hospitalizations is less concentrated than that of CT and RT.
The distribution of hospitalizations fi nanced by the SUS was more widespread that the CT and RT distribution, as the admissions included clinical and surgical procedures which were not necessarily carried out in hospitals equipped for highly complex procedures.
The dominant fl ow in hospital admissions represented 80.3% of the total volume and formed 112 networks.The transversal fl ows represented 12.7% of the total volume, a little higher than that of CT and RT, and refl ected spatially fragmented access (Table ).Relatively large volumes (fi ve hospital admissions per year or 40 in the eight year period of the study) are highlighted in Figure 3.Of the total volume of hospital admissions, 75.8% included patients living less than 100 km away from the health district of the hospital.The median distance covered corresponding to the dominant fl ows was of 215 km, almost doubling in the case of transversal fl ows (377 km).
Large distances characterized the North.For CT and RT, around 58% of residents were obliged to travel more than 1,000 km to reach a specialist health care unit, and approximately 25% had to leave their region to receive treatment, travelling, on average, more than 1,600 km.This suggests that these patients lacked adequate access to health care services.

DISCUSSION
The pattern of dominant fl ow of children and adolescents with cancer shows that the SUS is reasonably well organized and structured regionally and that the established networks in this environment include the majority of patients.However, a considerable portion of the origin-destination fl ow (10% to 20%) occurs outside of this structure, as shown in the mapping of transversal fl ows.A limitation of the study is that displacements within each health district or within each municipality were not investigated.Moreover, they were measured as straight lines between the regional centroids, not taking into account geographical barriers (mountains and rivers, amongst others) or the quality of the links.The results found certainly underestimate the real picture and indicate the necessity for regionalized alternatives.
Access to specialist health care services at the correct time is crucial in increasing survival and improving prognosis, as it makes accurate diagnosis and appropriate treatment viable, especially when the disease is still in the early stages.However, access to these centers in terms of distance, time and cost of transport has a profound impact on the quality of life for these children and their families, and the geographical pattern of access may be informative in the planning and allocation of resources.Research on delays in diagnosing cancer in children highlights that those resident in rural or distant areas have diffi culty in accessing care, as they need to travel long distances to reach specialist health care services. 2,11,12,14,16,19e majority of CT, RT and hospitalizations occurred in the health districts of the 12 main Brazilian cities: São Paulo, Rio de Janeiro, Belo Horizonte, Curitiba, Porto Alegre, Brasília, Goiania, Fortaleza, Recife, Salvador, Manaus and Belem -following Brazil's urban network structure, i and suggesting the existence of a defi cit in access to cancer treatment for children and adolescents living in the North and Northeast.This is especially evident in the cases of those living in the North, who are obliged to travel long distances to reach specialist treatment, and refl ects the inequality between the different regions of Brazil.Many families move due to the illness, or provide false information to facilitate access to specialist centers, which leads to the differences between places with and without access to increase still further.Patients residing in a radius of 100 km from health care services have greater access to treatment.Treatment protocols for CT and RT generally require patients to visit the specialist service several times a week, 10,17 making it diffi cult to administer treatment even to patients living closer than 100 km from the service.Support homes to provide accommodation for children and adolescents with cancer may reduce treatment dropout rates and provide support to families on a low income living long distances away from the service.A Brazilian study showed that this strategy, combined with better training and cooperation between pediatric oncologists, diminishes treatment dropout rates and increases survival rates for children with acute lymphocytic leukemia (ALL). 15,21eveloping policies which would increase routine mechanisms for providing children and adolescents, and their families, in the travelling and accommodation is critical to the success of the treatment.
The scale and volume of care in the health care units are considered indirect measures of the quality of care of patients with complex illnesses, 7 especially in the case of children and adolescents with cancer. 1,5,10As childhood cancer is a rare event, and demographic density is variable, depending on the region, it does not make sense to establish a CACON in each place, as they would be underused. 13The distribution of CACONs equipped to treat childhood cancer corresponds to the density of children and adolescents per km 2 , in Brazil.
It is critical to guarantee access to specialist centers through reference and social support mechanisms, especially for those who live far away from a CACON, enabling accurate diagnosis, appropriate treatment and improved rates of survival.
In conclusion, this study showed the importance of planning in the distribution of health care services according to the needs of the population and will help the health care authorities assess condition of access to services, as the tools and methodology used here are freely available.
Early diagnosis and timely treatment are key elements in the prognosis of childhood cancer.Higher chances of success depend on specialized care.
In this country, there are huge regional inequalities in childhood cancer care, with higher numbers of appointments and a wider range of resources in the richer regions.
Chemotherapy and radiotherapy protocols usually require the young patients to visit specialist health care services several times a week for relatively long periods.Therefore, accommodation for patients during the periods of treatment may help to reduce levels of abandoning treatment and provide support to less well-off families The study shows that between 10% and 20% of the fl ows of patients between their place of residence and the center where care is provided occurs outside of the structured networks, obliging patients to cover large distances in order to obtain suitable treatment.These results are, probably, underestimated.
The study points out the need to design regionalized alternatives in addition to the provision of accommodation, nutrition and transport solutions in order to reduce regional inequalities and provide equal opportunities for treatment.

Figure 2 .Figure 3 .
Figure 2. Transversal fl ow for chemotherapy and radiotherapy in children and adolescents aged up to 19 with cancer.Brazil, 2000 to 2007.

Table .
Classifi cation of travel fl ows for children with cancer between health districts according to type of treatment.Brazil, 2000 to 2007. 13 i Instituto Brasileiro de Geografi a e Estatística.Regiões de infl uência das cidades: 2007.Rio de Janeiro; 2008.Available from: http://www.ibge.gov.br/home/geociencias/geografi a/regic.shtm Figure 1.Dominant fl ows for chemotherapy and radiotherapy in children and adolescents aged up to 19, with cancer.Brazil, 2000 to 2007.