Thanks for that introduction, an extraordinarily tough act to follow there and I very much appreciate the lessons learned that you've just showed. I have no conflicts of interest and the disclosures relate to the funding from the BCH, ICCTR and Society of Critical Care Medicine, which I'm super grateful for. Now where you live is a key social determinant of health with many pathways to influence health outcomes and while centralisation promotes quality for infrequently used therapies that are complex like ECMO and PICU, it is at the expense of geographic access with distance unknown impediment to access life-saving therapies and so uneven geospatial distribution of healthcare resources according to social determinants of health may intensify inequitable access. But at the same time social barriers also shape access to care and non-geographic factors contribute to cross-border patient flows and there's more to patient care than just a PICU bed. And so in an attempt to study both geographic and non-geographic facets of access we apply complementary geographic information system and market share approaches to comprehensively analyse distribution and quality of PICU and ECMO capacity in the continental United States. And so we performed a cross-sectional vector based GIS analysis delineating drive times of 30, 60 and 120 minutes to hospital locations with more than or equal to one PICU bed or those providing ECMO support and this was based on 2019 AHA survey data and the Extracorporeal Life Support Organisation. And so this creates four catchments at increasing distance from a service with direct geographic access defined by a less than or equal to 60 minute drive time in the literature. And we then described the percent of paediatric population, population density as well as social determinants of health for each catchment and the social determinants were based on census data and publicly available metrics and this was performed at each catchment to compare social determinants of health according to drive time at national, regional and state levels as well as cartographically. And in the interest of time here I'll focus on the Child Opportunity Index and the proportion of underrepresented or underserved residents as being representative of neighbourhood level resource availability and diversity respectively. For the market share approach National Healthcare Utilisation Project or HCUP data was used to define bed capacity by allocating ICU capacity to zip codes based on actual PICU admissions in patients less than 14 years. And we did this for admissions to hospital with greater than or equal to one PICU bed based on the AHA data and to examine more specialised PICU bed capacity we restricted the PICU admissions to dedicated children's hospitals and hospitals with the top 25% of paediatric discharges. And for ECMO capacity again we used ELSA. And we then tested for disparities in access to capacity by studying the hospital serving patients with different racial and ethnic backgrounds and according to measures of social advantage and here again I'll focus on COI as representative of these social metrics. And finally to explore quality of each hospital we examined variables available in the AHA, HCUP and also three years of CMMS beneficiary data aggregated at the hospital level. And that included the structural variables, inputs or outputs per bed as well as some patient outcomes. And I'll start out by describing our geospatial results. And we identified 337 PICUs and 199 paediatric ECMO centres. And 80% of children have direct geographic access to PICU services while 5% of children reside more than two hours from PICU. Just under 70% have direct geographic access to an ECMO centre and almost 10% live more than two hours away. And at a national level neighbourhoods with direct geographic access to PICU services have statistically greater social advantage and this was indicated by a six point higher child opportunity index. Closer areas also had a 7% higher proportion of underrepresented races compared with children living more than 70 minutes away. And ECMO results really mirrored those for ICU. Obviously there's substantial overlap between those cohorts. Looking at this graphically across the four distance increments we can see that for both PICU and ECMO the COI is substantially higher for the less than or equal to 60 minute drive times. In comparison the percent underrepresented races has a very steep decrement beyond the 30 minutes. And what was really interesting is that at a state level while the socioeconomic measures formed a more consistent trend as you can see with the dark line here, the relationships more substantially differed between states in the race and diversity measures. And the South and West or the North East and Mid West tracked together regionally. Now maps can identify gaps in coverage and access and here in these maps the coloured areas have a high population density so that is more than 1,000 paediatric persons per square mile. And the green areas represent 30 minute drive times from a service. The red areas there represent areas with low COI that have drive times more than 60 minutes and the blue represents high COI areas with long drive times. In both PICU and ECMO maps there are substantially more red areas than blue which reflects the national findings that I just described but allows a deeper dive into the regional differences and also at a city level where using the same colour scheme you can see the red low COI areas with high drive times to ECMO in New York City and LA. One of the innovative aspects of this study was incorporating quality metrics to see if that changed any of the relationships between social determinants of health and geographic access. And we looked at statistical metrics including missingness, distribution, collinearity in an exploratory data analysis and also weighed clinical relevance, objectiveness and data accuracy for all these variables and ultimately it led to these variables in red being chosen as the quality metrics to use for adjustment. And interestingly it resulted in really minimal changes to the mean difference in social determinants of health scores between direct access and more than 60 minute drive times for both PICU and ECMO and this probably emphasises both the strength of the baseline relationships we identified but also some of the challenges in defining and quantifying quality. Using the market share methodology, paediatric ICU capacity for all children was 1.75 beds per 10,000 paediatric residents and that fell to 1.5 for specialised PICU capacity and 56% of PICU hospital capacity offered ECMO. And so first looking at capacity by race, the number of stars here indicates greater statistical significance with the reference groups shown in blue. And so it can be considered with these numbers as for the average child in this social determinant of health group, what is the ICU capacity serving them? And we can see consistent with some of the geospatial results, Hispanic children had higher local PICU capacity but this was much less substantive in the specialised analysis where actually non-Hispanic black children had less access to specialised ICU capacity. Hispanic and Asian children had less ECMO access while that was actually higher for the non-Hispanic black children. There were no clear trends for COI except ECMO access was much higher for the very high COI cohort, again consistent with some of the geospatial findings. Aggregate inputs via pay-a-mix, outputs per bed as well as the rates of CLABSI are shown here for all children from HCAP and notably hospitals serving Hispanic or non-Hispanic black children had substantially lower capital, expenditure and FTE per bed. Now these financial dollars are in the range of multiples of 1,000 so represent a really massive financial deficit, a difference in expenditure of up to $130,000 and 1.2 FTE per bed for hospitals serving underrepresented races and ethnicities. And these lower outputs were mirrored by lower inputs with a much higher proportion of Medicaid pay-a-mix in hospitals serving black or Hispanic children. Relative to the very high COI groups, more socially disadvantaged children were also served by less well-resourced hospitals. Children from very high COI areas attended hospitals with on average $370,000 more dollars expended and 1.5 FTE per bed compared with patients from the very low COI area. And the difference in proportion of Medicaid pay-a-mix was 13% between the two. And finally CLABSI rates were pretty similar across the board. A few limitations to note, our analyses exclude NICU, AHA and ELSO data is self-reported and so there's potential that this doesn't reflect exact contemporary care provision. And HCOP data unfortunately cannot be linked to external sources and their output and input data includes both paediatric and adult revenues but the specialised paediatric results that were restricted to high proportion of paediatric discharges showed very similar trends. So in summary, critically ill children living further from acute care tended to have less neighbourhood level resources. Consistent with urbanisation patterns and historic migration, underrepresented races and ethnicities often live in cities closer to PICUs, however there's wide variability between states and regions and PICU capacity specifically differs within the individual underrepresented racial groups. Patients from under-resourced communities or under-represented racial and ethnic backgrounds are more likely to be served by substantially less well-resourced hospitals and these investment trends were mirrored by the revenue those beds generate based on pay-a-mix. So some conclusions, all PICU beds are not created equal, potentially contributing to disparate outcomes and so we need better ways to define quality to better assess that. Inequitable access is probably context-specific, requiring analyses that look beyond aggregated national results. As paediatric acute care continues to consolidate, resource allocation requires critical review with a health equity lens and this is going to be important because focusing exclusively on geographic distribution may not be adequate because healthcare consumption is dynamic and quality varies. Ensuring all children have access to high-quality acute care requires more bottom-up approaches to break the cycle of disinvestment. I really want to thank all the people on this list and many, many others who've helped complete this. Thanks very much.

pediatric intensive care

ECMO services

healthcare disparities

resource allocation

equitable access