Community Analytics

Lao Community Data Dashboard

Disaggregated data across health, economic, education, and housing domains for ~15,000 Lao Minnesotans. When data is broken out from the aggregate “Asian” category, significant disparities emerge that are invisible in standard reporting.

6 data sources · 39 indicators · Updated April 2026

Key Insights

The most important findings from disaggregated analysis, curated by Dr. Phouthakannha Nantharath. Each insight pairs a data visualization with context to explain why the finding matters for policy and practice.

1Key Insight

The "Asian" category conceals a 4.8x disparity in gambling addiction

When Lao data is separated from the aggregate "Asian" category, the gambling addiction rate is 24.0% — nearly five times the 5.0% state average. The aggregated "Asian" rate of 8.0% makes it appear Southeast Asian communities have only modestly elevated risk.

4.8xhigher than state average
Aggregated "Asian"Lao (Disaggregated)State avg: 5%8%24%+16.0pp hidden

Source: MDH BRFSS 2024, Community Health Survey 2025

2Key Insight

Linguistic isolation affects 1 in 3 Lao households — 8x the state rate

34.8% of Lao households are linguistically isolated, meaning no one over 14 speaks English "very well." This creates cascading barriers to healthcare access, employment, and civic participation that compound across domains.

34.8%linguistically isolated households
State Average
4.5%
"Asian" Aggregated
15.2%
Lao Community
34.8%

Source: Census ACS 5-year estimates 2020-2024

3Key Insight

45% of Lao elders received no formal schooling — a 22x disparity

Nearly half of Lao elders (65+) report zero years of formal education, reflecting the disruption of the Secret War and refugee experience. This is 22 times the state rate of 2.0% and is invisible in aggregated data.

45.3%elders with no schooling
Aggregated "Asian"Lao (Disaggregated)State avg: 2%10%45.3%+35.3pp hidden

Source: Census ACS 2020-2024, Community Survey 2025

4Key Insight

ML models predict rising elder isolation risk in 3 Minneapolis neighborhoods

The Elder Isolation Index (86.2% accuracy, AUC-ROC 0.89) shows Minneapolis North, Brooklyn Center, and St. Paul East at risk scores exceeding the 65-point community threshold. Early warning triggered for proactive temple-based intervention.

75/100risk score — threshold exceeded
Mpls North
75
Brooklyn Ctr
71
St. Paul E
68
Mpls South
52
Woodbury
38
Above threshold Below threshold Threshold (65)

Source: Lao Community Data Portal ML Pipeline, 2025

Interactive Data Explorer

Select a metric to explore trends, compare populations, and examine statistical significance. Switch between chart and table views to interact with the data.

Gambling Addiction Rate

Time series: Lao community vs. Minnesota state average (2019–2025)

Lao CommunityState Average
Latest Value

24%

95% Confidence

[21.2, 26.8]

Effect Size

d = 1.42 (large)

Significance

p = 0.001 ***

Source: Census ACS 2020-2024, MDH BRFSS, Community Survey 2025

Lao Community Data Portal

Community Disparity Profile

Composite scores across six dimensions (0–100 scale, higher = better)

Lao CommunityMN State Average

Source: Composite index from Census ACS, BRFSS, Community Survey 2025

Lao Community Data Portal

Disaggregation Reveals Hidden Disparities

The same data, disaggregated: rates when Lao data is separated from 'Asian'

Lao CommunityAsian (Aggregated)MN State

Source: Census ACS 2020-2024, MDH BRFSS, Community Survey 2025

Lao Community Data Portal

Methodology

Data sources. This dashboard integrates data from 6 primary sources: U.S. Census American Community Survey (ACS) 5-year estimates (2020–2024), Minnesota Department of Health BRFSS, Bureau of Labor Statistics (BLS), Federal Reserve Economic Data (FRED), Bureau of Economic Analysis (BEA), and the Lao Community Health Survey (n=412, 2025).

Statistical methods. All confidence intervals are computed at the 95% level using Wilson score intervals for proportions. Effect sizes are reported as Cohen's d for continuous measures and odds ratios for binary outcomes. Significance is assessed using two-tailed tests with Bonferroni correction for multiple comparisons across domains.

Machine learning. Four predictive models (gambling risk, dropout prediction, mental health screening, elder isolation) use gradient-boosted trees (XGBoost) with 5-fold stratified cross-validation. All models include SHAP feature importance, fairness auditing (four-fifths rule, disparate impact ratio), and standardized model cards per Mitchell et al. (2019).

Limitations. Community survey data is convenience-sampled through temple networks and community organizations, which may under-represent highly isolated individuals. Census ACS estimates for small populations carry larger margins of error. ML predictions are intended to support — not replace — community judgment and culturally-informed decision-making.

Dr. Phouthakannha Nantharath · Bush Fellowship Community Innovation · Lao Community Data Portal