Identifying pre-diabetes 'hotspots' in Northern California using geospatial analysis: opportunities to target diabetes prevention strategies and improve health equity.

Objectives: The US Preventive Services Task Force recommends screening of adults aged 35-70 with a body mass index ≥25 kg/m for type 2 diabetes and referral of individuals who screen positive for pre-diabetes to evidence-based prevention strategies. The diabetes burden in the USA is predicted to triple by 2060 necessitating strategic diabetes prevention efforts, particularly in areas of highest need. This study aimed to identify pre-diabetes hotspots using geospatial mapping to inform targeted diabetes prevention strategies. A 'hotspot' is defined as a cluster of 3 or more neighbouring census tracts with elevated pre-diabetes prevalence.

Design: A cross-sectional study using ArcGIS software to geospatially map pre-diabetes prevalence hotspots. We used health system and census data to identify pre-diabetes hotspots using a systematic five-step geoprocessing approach that made use of incremental spatial autocorrelation and Getis-Ord Gi*.

Setting: This study was set in Kaiser Permanente Northern California (KPNC), an integrated health delivery system with over four million members.

Participants: KPNC adults ages 35-70 who underwent a haemoglobin A1c (HbA1c) or fasting plasma glucose (FPG) screening test in 2019 were mapped to census tracts in Northern California. People were considered to have pre-diabetes with an HbA1c of 5.7%-6.4% (39-46 mmol/mol) or FPG 100-125 mg/dL.

Primary And Secondary Outcome Measures: Individual and census-level characteristics were compared between hotspots and non-hotspots using χ and Wilcoxon rank sum tests, as well as risk differences (RDs) and Hodges-Lehmann (HL) estimates of location shift. Individual-level characteristics were derived from electronic health records and administrative data, while census-level characteristics were derived from the 2019 American Community Survey.

Results: A total of 760 044 adults met the study inclusion criteria and 40% had pre-diabetes. Individuals in pre-diabetes hotspots were less likely to be non-Hispanic white (33.6% vs 50.6%, RD: -17.04%, 95% CI -17.26% to -16.81%, p<0.0001) and more likely to have overweight or obesity (72.2% vs 69.2%, RD: 2.95%, 95% CI 2.73% to 3.16%, p<0.0001). Census tracts within hotspots had lower levels of household income (HL estimate: -3651.00, 95% CI -7256.00 to -25.00), per cent of adults with bachelor's degrees or higher (HL estimate: -9.08, 95% CI -10.94 to -7.24) and median home values (HL estimate: -113 200.00, 95% CI -140 600.00 to -85 700.00) and higher rates of household poverty (HL estimate: 0.96, 95% CI 0.55 to 1.37), unemployment (HL estimate: 0.39, 95% CI 0.24 to 0.54), household public assistance (HL estimate: 0.97, 95% CI 0.76 to 1.18) and per cent receiving Medicaid (HL estimate: 4.56, 95% CI 3.40 to 5.76) (p<0.05 for all).

Conclusions: We found that individual-level and census tract-level socioeconomic status, obesity prevalence and race and ethnicity categories of patients living in pre-diabetes hotspots differed from those not identified as a hotspot. Policy-makers and care providers can use this information to target diabetes prevention resources and outreach by enacting policies that provide insurance coverage for low-income populations and placing diabetes prevention programmes in communities with highest need.