We gratefully acknowledge support from the Dana-Farber/Boston Children's Cancer and Blood Disorders Center Global Health Initiative and from St. Jude Children's Research Hospital International Outreach Program, and we thank the AHOPCA data managers and clinical staff for their inspiring work.
The delivery of effective treatment for pediatric solid tumors poses a particular challenge to centers in middle-income countries (MICs) that already are vigorously addressing pediatric cancer. The objective of this study was to improve the current understanding of barriers to effective treatment of pediatric solid tumors in MICs.
An ecologic model centered on pediatric sarcoma and expanded to country as the environment was used as a benchmark for studying the delivery of solid tumor care in MICs. Data on resources were gathered from 7 centers that were members of the Central American Association of Pediatric Hematologists and Oncologists (AHOPCA) using an infrastructure assessment tool. Pediatric sarcoma outcomes data were available, were retrieved from hospital-based cancer registries for 6 of the 7 centers, and were analyzed by country. Patients who were diagnosed from January 1, 2000 to December 31, 2009 with osteosarcoma, Ewing sarcoma, rhabdomyosarcoma, and other soft tissue sarcomas were included in the analysis. To explore correlations between resources and outcomes, a pilot performance index was created.
The analyses identified specific deficits in human resources, communication, quality, and infrastructure. The treatment abandonment rate, the proportion of metastatic disease at diagnosis, the relapse rate, and the 4-year abandonment-sensitive overall survival (AOS) rate varied considerably by country, ranging from 1% to 38%, from 15% to 54%, from 24% to 52%, and from 21% to 51%, respectively. The treatment abandonment rate correlated inversely with health economic expenditure per capita (r = −0.86; P = .03) and life expectancy at birth (r = −0.93; P = .007). The 4-year AOS rate correlated inversely with the mortality rate among children aged <5 years (r = −0.80; P = 0.05) and correlated directly with the pilot performance index (r = 0.98; P = 0.005).
There is growing interest in improving our understanding of cancer disparities around the world[1-3] and addressing the high burden of cancer-related mortality faced by low- and middle-income countries (LMCs). Pediatric cancer is no exception, because 80% of children with cancer live in LMCs. In many countries, as standards of living improve and millennium development goals are achieved, the burden of cancer becomes more tangible. However, the purchasing capacity and the allocation of needed technologies and clinical skills may be lagging.
The delivery of effective treatment for pediatric solid tumors poses a particular challenge to middle-income countries (MICs) that already are vigorously addressing pediatric cancer. Successful frameworks for improving outcomes of patients with leukemia in resource-poor settings have been developed; however, these models may not be directly applicable to solid tumors. Experience from the development of pediatric brain tumor programs in MICs has demonstrated the importance of multidisciplinary care, empowering the care team, adhering to protocols, telemedicine, and the twinning model.[8-12] Similar to brain tumors, pediatric extracranial solid neoplasms constitute a heterogeneous group of malignancies with very specific therapeutic principles according to disease and risk group, and they share an inherent need for meticulous, comprehensive, multidisciplinary care.
The Central American Association of Pediatric Hematologists and Oncologists (AHOPCA) offers a unique opportunity to examine the interface between resource availability and pediatric solid tumor outcomes. AHOPCA has a successful track record in delivering protocol-based treatment[13-18] and has prospectively collected outcomes in the Pediatric Oncology Networked Database (POND).[19, 20] However, despite the parallel development of pediatric leukemia and solid tumor programs, improvements in outcomes for children with solid tumors have lagged behind the improvements observed in leukemia.[13, 14, 21] It is believed that this difference is highly influenced by infrastructure shortfalls and difficulty with the implementation of multidisciplinary care.
The objective of this study was to improve our understanding of barriers to effective treatment of pediatric solid tumors in MICs. An ecologic conceptual framework was used as a benchmark for studying the delivery of solid tumor care, identifying nonbiologic factors of interest, and illustrating key components of comprehensive multidisciplinary care. In the absence of an established measure, associations between resources and outcomes were explored in a “proof-of-principle” exercise through a pilot performance index. The results are intended to inform the development of targeted strategies that address identified system-level barriers to pediatric cancer care in MICs and set the stage for further studies on the interface between nonbiologic factors and outcomes.
MATERIALS AND METHODS
A 4-level ecologic model was used to guide data collection and analysis. The model ascends from patient to country and seeks to illustrate influences between the levels (see Fig. 1).
Level 1: The Patients
The shortage of national cancer registries in LMCs documenting survival has limited the study of regional differences in pediatric cancer outcomes. We recently reported on pediatric sarcoma outcomes diagnosed between January 1, 2000 and December 31, 2009 for 6 of the 7 AHOPCA-member countries (excluding the Dominican Republic). Data had been collected previously and stored in hospital-based cancer registries using POND,[19, 20] a web-based, password-protected, data management tool provided free of charge to centers in the developing world and managed by St. Jude Children's Research Hospital International Outreach Program. The diagnoses, classified according to the International Classification of Childhood Cancer, third edition (ICCC-3) diagnostic group for each, included osteosarcoma (diagnostic group VIIIa), Ewing sarcoma (diagnostic group VIIIc), rhabdomyosarcoma (diagnostic group IXa), and other soft tissue sarcomas (diagnostic groups VIIIb, IXa, and IXd). International Classification of Diseases codes were not available in the database. Primary outcomes included the rates of metastatic disease at presentation, treatment abandonment, relapse, and survival. The same population was selected for this study, but the data were analyzed by country, and the survival analysis was updated to reflect follow-up through January 1, 2013 (see Table 1 and Figure 2).
Table 1. Pediatric Sarcoma Outcomes of Interest (Level 1) by Country
Data for levels 2 and 3 (the care team and the centers) were obtained using an exploratory tool, first during January to June 2010; then, the data were updated during January to June 2011. Pediatric oncologist members of AHOPCA and those involved in the treatment of solid tumors served as liaisons for content validity and data collection at their centers. Items were collected by discipline but were analyzed according to level and component in the framework. Regarding level 2 (the care team), the tool addressed human resources (32 questions), communication (8 questions), and perceived quality of care (17 questions). Free-text comments were allowed and are reported along with objective data for each section.
Level 3: The Centers
Seven AHOPCA-member centers were analyzed: National Children's Hospital (HNN), Costa Rica; Benjamin Bloom National Children's Hospital (HNNBB), El Salvador; National Pediatric Oncology Unit (UNOP), Guatemala; Maternal and Child Hospital (HMI), Honduras; “La Mascota” Children's Hospital (HILM), Nicaragua; Children's Hospital of Panama (HNP), Panama; and Roberto Reid Cabral Hospital (HRRC), Dominican Republic. These centers are designated as public national referral centers for pediatric oncology care. However, in some countries, patients alternatively may access care at satellite centers, private institutions, or social security hospitals. Under-diagnosis, under-referral, and efflux to other countries also may occur. This is the case, for example, in the Dominican Republic, where an oncology institute and private hospitals (which primarily treats adults but accept children) exist, as well as a second public pediatric oncology unit in Santiago (the second largest city in the country, located northwest of Santo Domingo). The expected number of children with cancer who are registered in POND in the AHOPCA-member countries has been estimated by other authors to range from 25% to 100% using the observed registrations of patients aged <15 years from 2004 to 2007 and an expected incidence of pediatric cancer of 100 cases per million population aged <15 years (25% in the Dominican Republic, 35% in Panama, 50% in Guatemala, 62% in Honduras, 77% in El Salvador, and 100% in Costa Rica).
The 7 included AHOPCA centers, as designated public national referral centers, were deemed representative and, thus, appropriate for the analysis of resources in the public sector in each country. Infrastructure data from the Dominican Republic were analyzed despite the lack of matching outcomes data, because we believed those data would enrich and balance the granularity of comparative data between upper-middle-income countries (UMICs) and lower-middle-income countries (LMICs). Finally, in the absence of national cancer registries in all countries except Costa Rica, data in POND from the remaining 6 AHOPCA centers were considered representative and appropriate for the analysis of outcomes in the public sector in each country. Regarding level 3, we believed that infrastructure availability was of interest, and the tool addressed diagnostics (8 questions), therapeutics (15 questions), and supportive care (27 questions).
The case-capture rate was evaluated using the rate of observed versus expected cases (the O/E rate). Costa Rica is the only AHOPCA-member country with a national cancer registry. Therefore, O/E rates for pediatric sarcoma were calculated using the average of observed cases between 2005 and 2009, population statistics for 2009, and the reported incidence of pediatric sarcoma in Costa Rica (1984-1992 collection period; 9.7 cases per million), all based on data for children ages birth to 15 years. Costa Rica has approximately 4,000,000 inhabitants, 28% of whom are aged <15 years (about 1,200,000 children).The O/E rate for childhood cancer was similarly calculated using incidence data from Costa Rica (136 cases per million). These incidences are lower than those reported in high-income countries but are higher than in other Latin American countries. Therefore, they may overestimate or underestimate the true O/E rate for other countries in the region. Stability in the incidence of childhood cancers in Costa Rica for 1984 to 1992 (the collection period with data available by ICCC-3 diagnosis group), compared with 1998 to 2002 (the most recent collection period), was confirmed by searching the Cancer Incidence in 5 Continents (CI5) online database for males and females ages birth to 14 years, for “all sites” (C00-96) and “bone” (C40-41). However, incidence rates from the most recent CI5 collection period could not be used to obtain the expected incidence for males and females with pediatric sarcoma needed for this study, because cancers are reported exclusively by topography rather than by diagnostic group. Bone topography (C40-C41) can be selected to represent “bone tumors”; but rhabdomyosarcomas, soft tissue sarcomas, and nonosseous Ewing sarcomas, which can occur in a multitude of topographic sites, cannot be selected and, thus, their observed incidence cannot be retrieved.
Level 4: The Environment
Country-level health and economic indicators and financial mechanisms used to pay for services were considered of interest to evaluate the environment AHOPCA-member centers face while delivering care. Established indicators of interest were obtained from the World Bank online database. Applying the World Bank classification by country income group, Costa Rica, Panama, and the Dominican Republic are classified as upper-middle-income countries (UMICs); and El Salvador, Guatemala, Honduras, and Nicaragua are classified as lower-middle-income countries (LMICs). Each center's reliance on foundations, the public health care system, and out-of-pocket expenditures was qualitatively assessed in the assessment tool.
Correlations and Development of the Pilot Performance Index
An indicator reflecting hospital-level or country-level performance in the delivery of cancer care does not exist. Among the previously analyzed surrogate indicators, the annual governmental health care expenditure per capita (HEC) has had the best correlation with pediatric cancer survival. This and other country-level indicators were tested for correlation with pediatric sarcoma outcomes.
To evaluate performance, 40 items from levels 2 and 3 were selected and aggregated into a pilot performance index. This was a proof-of-principle exercise done to determine whether the incorporated multitude of nonbiologic factors could be aggregated and analyzed in a meaningful way. Composite indicators have become more popular and accepted for the assessment of global health strategies, and guidelines have been published for their development.[30, 31] However, a cancer-oriented composite indicator has not been developed. In the case of pediatric extracranial solid tumors, such an indicator could help define the resources of interest, allow for comparative research between countries, and allow for longitudinal monitoring of the acquisition of better infrastructure and its impact on survival. In the composite indicator literature, item selection and valuation (weighting) is perhaps most controversial.[30, 32, 33]
The proportion between level 2 and 3 items (n = 108) and centers (n = 7) limited the use of regression or factor analysis to identify significant items, as expected. Although alternative approaches for the establishment of priorities and weights for composite indicators have been published,[34, 35] addressing the above-mentioned controversy was beyond the scope of his study. Therefore, items were selected by 2 investigators (P.F. and C.R.-G.) based on probable sensitivity as a screening question, and the list was revised and approved by all other investigators. Items were selected using the investigators' best judgment based on their individual experience practicing pediatric oncology in high-income and lower-income settings. Furthermore, in the absence of quantitative methods to back-up a gradation strategy, all items were weighted equally. This somewhat ad hoc process has been successfully applied in the early stages of development for other composite indicators.[36, 37] With regard to the items, no AHOPCA-member center reported access to all items that were selected as “of interest,” but all incorporated items were present in at least 1 center. Items that were selected to be included in the pilot performance index are marked with asterisk in Figure 3. Cross-sectional correlation of survival for 2000 to 2009 and index scores for 2009 were considered adequate, because pediatric sarcoma outcomes failed to demonstrate a significant improvement within the study period (see Table 1). Therefore, cross-sectional correlation presumably offers the best-powered survival estimate and the best-case scenario for each country.
Survival analysis included standard overall survival (OS), abandonment-sensitive OS (AOS), event-free survival (EFS), and abandonment-sensitive EFS (AEFS). Refusal and abandonment were analyzed as one event (“abandonment”), as recommended by the International Society of Pediatric Oncology Working Group on Treatment Abandonment. Four survival analyses were performed: 1) EFS, with events defined as relapse, progressive disease, secondary malignancy, or death; 2) AEFS, with events defined as abandonment of therapy, relapse, progressive disease, secondary malignancy, or death; 3) OS, with an event defined as death; and 4) AOS, with events defined as abandonment or death (Fig. 2). Time to event was calculated from the time of diagnosis until the first event or the last contact if no event occurred. The Kaplan-Meier method was used to obtain 4-year survival estimates, and the log-rank was used test to assess statistical significance. Descriptive methods summarized the care and infrastructure assessment. Categorical and ordinal items were converted to binary variables and analyzed by each country's income-group using the World Bank classification. The Pearson or Spearman method was used to test for correlations between variables, depending on an evaluation of outcome on scatter plots. P values < .05 were considered significant.
Level 1: The Patients
Center-level results for 785 cases of pediatric sarcoma diagnosed from 2000 to 2009 are presented in Table 1 and Figure 2. The treatment abandonment rate ranged from 1% to 38%, the rate of metastatic disease at diagnosis ranged from 15% to 54%, and the rate of relapse/progressive disease as a first event (which excludes patients who abandon therapy from the denominator) ranged from 24% to 52%. Rates for these outcomes decreased significantly by era in the region, but not at each center. The 4-year AOS estimate for the region was 32%, but it ranged from 21% to 51%. Standard methods overestimated 4-year survival by up to 34% compared with abandonment-sensitive methods. Survival improved somewhat by era in all countries except in Panama and Guatemala; however, the differences did not achieve statistical significance. The difference between AOS and AEFS was minimal.
Level 2: The Care Team
All centers had fully trained pediatric oncologists as well as other providers of interest, as indicated in Table 2 and Figure 3A (human resources). However, subspecialty training and/or pediatric experience was limited among pathology, surgical, and radiation therapy providers. Pediatric oncology staff had large patient panels, particularly in LMICs (>50 new cancer cases per oncologist per year), and was not able to work exclusively at the public institution in 4 of 7 countries. Income inequalities between the private and public health sectors were offered as an explanation by providers in the free-text comments. Pain and palliative care providers were available in 5 centers and 3 centers, respectively. Legal or child-protection teams were available in 4 centers but were used only infrequently according to the free-text comments.
Table 2. Characteristics of Interest From the Care Team, Centers, and Environment by country: Levels 2, 3, and 4
Deaths among children aged <5 y from diarrheal illness, %f
Prevalence of HIV—total, % of population ages 15–49 y
Poverty gap at $2/d—PPP, %
Vulnerable employment—total, % of total
Out-of-pocket health expenditure, % of total
Economically active children—total, % of individuals ages 7–14 y
Formal multidisciplinary care meetings and direct communication were pursued (Fig. 3A, communication and the multidisciplinary team). However, patients were infrequently assigned to a primary oncologist, communication was not always considered effective, and meeting with radiation therapists was considered most challenging. Distance between pediatric oncology and radiation therapy centers was offered as the main reason for the later in the free-text comments.
Quality of care
Treatment delays reportedly occurred more frequently because of bed availability than chemotherapy availability (Fig. 3A, quality of care). In most centers, detailed reporting of surgical and pathologic margins, urgent access to radiation services, and proper timing of radiation services were reported as inconsistently achieved.
Level 3: The Centers
On the basis of population estimates, 2430 new cases of childhood cancer and 173 new cases of pediatric sarcoma would be expected in children aged <15 years in the region each year (Table 2). Variability in capture rates between centers was evidenced by the O/E rate range for all cancers and sarcomas. Pediatric sarcomas appeared to be over-represented only in Costa Rica (120%) and were markedly under-represented in Guatemala and Honduras (42% and 41%, respectively).
Computed tomography and magnetic resonance imaging were available in all centers, but time-sensitive or urgent access was limited (Fig. 3D, diagnostics). Regarding radiology, nuclear medicine services were limited the most, and positron emission tomography was uniformly unavailable. Pathology services were available in all centers, but the quality of slides, access to a full panel of immunohistochemistry, and molecular studies reportedly were limited and of concern. Because of these limitations, all centers sought help from international collaborators in the evaluation of rare or complex cases.
All centers had basic chemotherapy drugs and uniform treatment guidelines available (Fig. 3E, therapeutics). The treatment of bone sarcomas was limited by an inability to monitor methotrexate levels in real time and cost. Access to limb-sparing procedures through international collaborations had recently increased but was only available consistently in Costa Rica. Radiation therapy was available in all countries, but cobalt-based radiation was the only option in 3 of 4 LMICs.
Although pediatric intensive care units were uniformly available (Fig. 3F, supportive care), several centers expressed that access was inadequate for oncology patients, in part because of bed availability but also because of a sense of “fatalism” around the oncologic diagnosis. Permanent vascular devices were infrequently available, and anesthesia services for imaging or radiation sometimes required the physician to travel with the patient and provide the service. Prostheses also were infrequently available after amputation. Finally, limitations in access to blood products during therapy and surgery were reported, particularly regarding platelets.
Level 4: The Environment
The countries' performance on established indicators
Guatemala had the largest pediatric population and ranked lowest for most indicators, except HEC, gross national income per capita, the rate of economically active children (Nicaragua was lowest), and proportion of the population living under $2 per day (Honduras was lowest). The Dominican Republic had the highest percentage of undernourishment (as a percentage of the population), and El Salvador the highest percentage of deaths from diarrheal illness. Costa Rica scored best in all health and economic indicators considered, except for gross national income per capita and the human development index, for which Panama took the lead. Furthermore, Panama had the lowest percent out-of-pocket health expenditure. It is noteworthy that the prevalence of human immunodeficiency virus (HIV) was <1% in all countries analyzed; thus, it is unlikely that HIV/acquired immunodeficiency syndrome (AIDS) was a significant burden on the health care system.
Financing of care
All centers received support from strong local foundations. Through advocacy, local ministries of health and foundations had secured coverage of chemotherapy, supportive care medications, and surgery in all countries, with infrequent need for patient out-of-pocket contribution. However, a contribution by the patient was frequently required to cover the cost of radiology services, radiation therapy, bone grafts, external prostheses, and rehabilitation services.
Pilot performance index
Centers varied in terms of the total score (range, 35%-80%) but varied most drastically on domain subscores. Figure 4 illustrates the distribution of index and domain scores and how they could be interpreted in meaningful ways. For example, in response to acknowledgment of deficits in human resources and communication, the AHOPCA center in Guatemala initiated efforts to incorporate radiation therapists, surgeons, and other specialists with interest in pediatrics formally into their staff and implemented multidisciplinary care meetings with these subspecialties. The total score for LMICs versus UMICs was equal, but UMICs tended to have better infrastructure subscores, and LMICs had better communication subscores.
Correlation of patient outcomes with established indicators
The treatment abandonment rate correlated inversely with the HEC (r = minus;0.86; P = .03) and life expectancy at birth (r = 0.93; P = .007) (see Table 3). The rate of vulnerable employment had a strong correlation (r > 0.75) but failed to achieve significance. The O/E rate correlated significantly with several established indicators associated with better standards of living. AOS at 4 years correlated inversely with the mortality rate among children aged <5 years (r = −0.80; P = .05).
Table 3. Correlation of Established Indicators and Pilot Performance Indicator With Pediatric Sarcoma Outcomes in the 6 Countries With Outcomes Data (Analysis Excludes the Dominican Republic)a
Abbreviations: AOS, abandonment-sensitive overall survival; NA, not available; O/E: observed-to-expected rate; OS, overall survival; PPP, purchasing power parity.
Values in boldface indicate a statistically significant difference.
Treatment abandonment rates improved over time; therefore, the rates for 2005 through 2009 were used in correlations. AOS did not improve over time; therefore, the rates for 2000 through 2009 were used for analysis (see Materials and Methods).
Health expenditure per capita, current US$
Gross national income, per capita
Mortality rate for children aged < 5 y, per 1000
Life expectancy at birth, total y
Literacy rate—adult total, % of individuals aged ≥15 y*
Human Development Index
Prevalence of undernourishment, % of population
Deaths among children aged <5 y from diarrheal illness, %
Poverty gap at $2/d—PPP, %
Vulnerable employment—total, % of total employment
Out-of-pocket health expenditure, % of total expenditure on health
Economically active children—total, % of children ages 7–14 y
Pilot Performance Index score, %
Human resources score
Quality of care score
Supportive care score
Correlation of patient outcomes with the pilot performance index
AOS at 4 years correlated most strongly with the pilot performance index (r = 0.98; P = .005), as indicated in Table 3; and, among the index's 6 subscales, human resources and supportive care scores had a significant, independent correlation with AOS. Diagnostic capacities had a strong correlation (r > 0.75) but did not achieve significance. Figure 5 provides a comparison between total scores on the pilot performance index, the HEC, and AOS, which reveals the following: 1) Some LMICs performed better than expected based on their HEC, as in the case of El Salvador and Guatemala; 2) the HEC and the pilot performance index followed a similar trend, but correlation between the HEC and AOS was not significant (r = 0.55; P = .25), likely because of a large discrepancy between the HEC, the pilot performance index score, and the AOS estimate in Panama; and 3) variability in outcomes was smaller than variability in index scores, particularly among countries with the lowest scores; therefore, it is possible that unaccounted factors may be limiting the center's capacity to compensate for their context.
The treatment of childhood cancer requires multidisciplinary care of high complexity; improved outcomes observed over the last decades in high-income countries can only be interpreted under this premise, and the same principles must guide the development of pediatric cancer programs in countries with limited resources. Thus, as centers in MICs advance care in the management of children with cancer, we must understand unique features, identify strengths and limitations, and develop step-wise initiatives for rational capacity-building and improved outcomes.
The objective of this study was to improve our understanding of barriers to the effective treatment of pediatric solid tumors in MICs. We selected children with sarcomas as the study population, because outcomes for this group of patients reflect the complexity of multidisciplinary care and, thus, could be used as a surrogate to test the strengths and weaknesses of the health care system. We assumed the impact of system-level barriers to care would be similar (or at least consonant) between pediatric sarcoma and other extracranial solid tumors, because the effective treatment of most extracranial solid tumors requires meticulous, comprehensive, multidisciplinary care and a similar pool of human and infrastructure resources.
Our results point to specific human resources, communication, quality, and infrastructure deficits that could explain the observed low survival of children with sarcoma in AHOPCA-member countries, reveal areas to work on, and serve as a starting point for the development of a more meaningful indicator of pediatric cancer center performance. It is noteworthy that we evaluated resources and capacities at public referral centers caring for patients from the general population (not selected centers or patients with means, insurance, etc) and practicing within heterogeneous health care systems. This analysis did not take into account cultural barriers to cancer care, which would be more likely to vary from country to country and region to region. Therefore, we believe that our findings are consonant with what would be observed in other MICs, are generalizable and applicable to other MICs, and could be used to develop similar regional initiatives worldwide.
Pediatric sarcomas collectively represent 6% to 16% of childhood cancers; and, in high-income countries, estimated survival has improved to reach approximately 70%.[39, 40] In AHOPCA-member countries, the AOS rate ranged from 23.5% to 47.1% at 4 years. Figure 1 provides an illustration of the importance of accounting for treatment abandonment in survival analyses in LMCs; an analysis based on the standard definition of OS would noticeably overestimate survival in countries with treatment abandonment rates >20%. Also, AOS was almost identical to AEFS in LMICs, suggesting that survival after relapse was greatly compromised.
Previously identified reasons for observed worse pediatric sarcoma outcomes in AHOPCA-centers included late presentation, a high relapse rate, and a high treatment abandonment rate. The current study suggests that performance factors at the level of the care team and centers also are important. These nonbiologic factors probably are most influential and actionable in resource-limited settings and deserve to be prioritized. Our analysis revealed specific concerns with human resources, clinical space, communication, diagnostic imaging, specialized pathology, specialized surgery, blood banking, coordination of multidisciplinary care, and access to radiation services. Targeted strategies are underway that may have a positive impact on patient care, extend the benefits diagonally to the health care system and other disciplines, and will be monitored prospectively.
The impact of specific socioeconomic, contextual, or infrastructure factors on outcomes can be difficult to assess using single measures. Therefore, composite measures have become more popular and accepted for assessment of global health strategies.[30, 31] In our study, some established health and economic indicators reflected the relation between resources and outcomes very well, as was the case for treatment abandonment and the O/E rate. However, with regard to decision making, the pilot performance index correlated best with AOS, and its subscales were meaningful for determining priorities. Our results suggest that the complex variety of factors involved in the delivery of pediatric cancer care can be conceptualized in discrete components, analyzed using transparent methods, interpreted in ways meaningful to leadership at a pediatric cancer center, and used to promote positive change.
We acknowledge the limitations of our study, including possible ecologic fallacy, the use of grouped data, the scarcity of population-based cancer registries, and its retrospective and cross-sectional nature. However, lessons learned will guide next steps in the goal of developing a measure that meaningfully describes performance in delivering pediatric cancer care and defines resource clustering internationally.
Paola Friedrich's work was supported by training grants from the National Institutes of Health: Pathophysiology of Human Blood Cells (T32 HL 7574) and Program in Cancer Outcomes Research Training (R25 CA092203).