The study team reported a benefit-transfer modeling strategy to estimate healthcare cost differences associated with residential greenspace across Washington. They used decile-specific relative healthcare-cost ratios previously published from a Northern California cohort of approximately five million Kaiser Permanente members and mapped those ratios onto Washington’s census block groups. The analysis relied on satellite-derived normalized difference vegetation index (NDVI) values to assign block groups to greenness deciles. Ratios were then calibrated to Washington’s reported 2020 statewide per-capita healthcare expenditure so that estimated costs reflected the state’s own spending baseline rather than the origin cohort’s absolute costs. The source did not report individual-level clinical data or new patient-level outcomes; the method transferred effect-size estimates from an earlier, external cohort to a different population and geographic context.
Data indicated a clear modeled gradient in annual per-capita healthcare spending across NDVI deciles. The authors provided estimated per-capita costs of $6,472.75 in the least-green decile and $5,939.39 in the greenest decile, yielding a modeled per-person difference of $533.35 between the extremes. Those figures reflect the calibrated application of the decile-specific cost ratios to Washington’s per-capita expenditure baseline; the preprint did not present new clinical endpoints beyond these modeled cost estimates. The cohort-to-state transfer produced spatially explicit estimates at the census block-group scale, enabling aggregation to statewide totals and to subsets such as urban cores.
The authors ran a set of simulated greening interventions and reported estimated annual statewide savings under each scenario. Raising the lowest decile to the next decile produced an estimated $45.6 million in annual savings. Lifting the four lowest deciles up to Decile five produced a larger modeled benefit, estimated at $157.7 million per year. A targeted urban-core greening scenario—limited to the five largest Washington cities—was associated with modeled annual savings of $56.9 million. These numbers reflect modeled associations produced by applying the transferred ratios and calibrations described above. The preprint did not report the practical costs of implementing greening interventions, nor did it provide cost-effectiveness ratios that combine intervention costs with modeled healthcare savings.
The analysis used three primary data inputs reported in the preprint: satellite-derived NDVI values aggregated to census block groups, U.S. Census block-group population counts for population weighting, and a published statewide per-capita healthcare expenditure for Washington (2020) used as the calibration anchor. Relative healthcare-cost ratios came from a peer-reviewed study of Kaiser Permanente members in Northern California; the preprint stated the originating study cohort comprised roughly five million members. The authors noted that only aggregated, publicly available data and published ratios were used and therefore no institutional review board approval was required. The preprint indicated that data and code are available from the corresponding author upon reasonable request and will be deposited in a public repository upon acceptance.
The authors explicitly framed their results as modeled associations rather than causal forecasts. They reported that estimates depend on the validity of transferring relative cost ratios between regional populations and on NDVI as a proxy for meaningful nature access. NDVI captures vegetation cover but does not measure accessibility, quality of greenspace, or individual-level exposure and use—factors that can influence health outcomes. The preprint did not attempt to quantify implementation costs for greening interventions, nor did it present sensitivity analyses beyond the decile-based simulations described. The approach therefore leaves open key questions: how transferable are the Northern California-derived ratios to Washington’s population composition and healthcare utilization patterns; what are the expected timelines for any health-related spending changes after greening interventions; and how would including intervention costs alter net economic estimates? These uncertainties were acknowledged by the authors as analytic boundaries of the benefit-transfer framework.
Readers should also note the preprint status of this work. The manuscript has not been peer reviewed; as the authors stated, the findings should not be used to guide clinical practice. The last substantive statements in the document emphasized the modeled, non-causal nature of the estimates and the dependency of conclusions on transferability assumptions and on NDVI’s limitations as an exposure proxy. Those open questions define the most immediate needs for follow-up: region-specific empirical studies linking observed greenspace change to health outcomes and healthcare utilization, inclusion of implementation costs in economic assessments, and analyses that use more direct measures of nature access and use to strengthen causal inference.