Skip to main content
Vellora

Vellora Research · Research report

Geographic Variation in Cardiovascular Health Across Texas Counties

A county-level secondary analysis of CDC PLACES estimates, CDC heart disease death rates and USDA rural-urban codes.

Author
Vellora Research
Published
2026-10-23
Type
Independent secondary-data analysis; not peer reviewed
Data
BRFSS 2023 (via PLACES 2025); deaths 2022–2024

Download PDFData and code

Abstract

Background. Heart disease is a leading cause of death in the United States.[1] This study describes how county-level measures of cardiovascular conditions, their risk factors and heart disease mortality vary across Texas’s 254 counties.

Methods. Ecological, cross-sectional analysis of CDC PLACES 2025 age-adjusted prevalence estimates (BRFSS 2023) for five risk factors and two conditions, CDC heart disease death rates for adults 35 and over (2022–2024) and USDA Rural-Urban Continuum Codes 2023. We report distributions, global Moran’s I with 9,999 permutations, Mann–Whitney tests comparing nonmetropolitan with metropolitan counties, and Spearman correlations with Benjamini–Hochberg adjustment. 253 counties had data; Loving County is suppressed by CDC in every source.

Results. The median county heart disease death rate was 402.1 per 100,000 (interquartile range 345.9–446.1), ranging from 205.3 (Presidio) to 808.0 (Polk). For all 7 PLACES measures, the median county’s 95% confidence interval was wider than the interquartile range across counties. All eight measures were positively spatially autocorrelated (Moran’s I 0.35–0.57, p ≤ 0.0001) and higher in nonmetropolitan counties (adjusted p ≤ 0.0033); the nonmetropolitan median death rate exceeded the metropolitan median by 53.9 per 100,000. Of the risk factors, current smoking had the strongest association with the death rate (ρ = 0.54, 95% CI 0.44 to 0.63); diagnosed diabetes showed no clear association (ρ = 0.09, 95% CI −0.03 to 0.22, adjusted p = 0.1323). Correlations between PLACES risk factors and PLACES condition estimates were very high (ρ 0.83–0.94) but are not independent evidence, because the estimates share a model.

Conclusions. Cardiovascular measures in Texas cluster geographically and are less favourable in nonmetropolitan counties. Current smoking and high blood pressure showed the clearest associations with the independently measured heart disease death rate, with overlapping confidence intervals. These are associations between counties, not causal effects, and they cannot be applied to individuals.

1. Introduction

CDC identifies high blood pressure, high cholesterol and smoking as key risk factors for heart disease, alongside diabetes, obesity, unhealthy diet, physical inactivity and excessive alcohol use.[2] National surveillance has reported higher age-adjusted heart disease death rates in nonmetropolitan than in metropolitan areas.[12] Texas has 254 counties, from large metropolitan counties to very small rural ones, so county measures can show where cardiovascular burden is concentrated within the state.

County estimates for small populations are uncertain, and many are produced by statistical models rather than counted. This report therefore asks narrow, descriptive questions, reports uncertainty alongside every estimate, and separates associations that rest on independent data from those that do not.

2. Research questions

  1. How much do county estimates of cardiovascular conditions, risk factors and heart disease mortality vary across Texas, relative to their uncertainty?
  2. Are these measures geographically clustered?
  3. Do they differ between metropolitan and nonmetropolitan counties?
  4. How strongly do county risk-factor estimates track county condition estimates and heart disease death rates?

3. Data sources

Measures analysed
MeasureRoleDefinitionSource and years
High blood pressure among adultsRisk factorAdults ever told by a health professional they have high blood pressure (excluding only during pregnancy, or borderline)CDC PLACES 2025, age-adjusted prevalence, BRFSS 2023[3, 6, 7]
Diagnosed diabetes among adultsRisk factorAdults told by a health professional they have diabetes (excluding during pregnancy)CDC PLACES 2025, age-adjusted prevalence, BRFSS 2023[3, 6, 7]
Obesity among adultsRisk factorAdults with a body mass index of 30 or more, from self-reported height and weightCDC PLACES 2025, age-adjusted prevalence, BRFSS 2023[3, 6, 7]
Current cigarette smoking among adultsRisk factorAdults who have smoked at least 100 cigarettes and now smoke every day or some daysCDC PLACES 2025, age-adjusted prevalence, BRFSS 2023[3, 6, 7]
No leisure-time physical activity among adultsRisk factorAdults reporting no physical activity or exercise outside their regular job in the past monthCDC PLACES 2025, age-adjusted prevalence, BRFSS 2023[3, 6, 7]
Coronary heart disease among adultsOutcomeAdults ever told they had angina or coronary heart diseaseCDC PLACES 2025, age-adjusted prevalence, BRFSS 2023[3, 6, 7]
Stroke among adultsOutcomeAdults ever told they had a strokeCDC PLACES 2025, age-adjusted prevalence, BRFSS 2023[3, 6, 7]
Heart disease death rate, adults 35+OutcomeDeaths with heart disease as the underlying cause (ICD-10 I00–I09, I11, I13, I20–I51), age-standardized to the 2000 U.S. population, 3-year average, spatially smoothedCDC, National Vital Statistics System, 2022–2024[8, 9]
Metropolitan statusGroupingRural-Urban Continuum Code 1–3 (metropolitan) or 4–9 (nonmetropolitan), from the 2023 OMB metro delineationUSDA ERS, 2023[10]

PLACES estimates are produced by multilevel regression and poststratification: a model fitted to BRFSS survey responses with some or all of individual age, sex, race and ethnicity and education, county poverty and state and county random effects, applied to census population counts. CDC reports each estimate with a 95% confidence interval from 1,000 Monte Carlo simulations.[4] CDC states that age-adjusted estimates can be used for county comparisons, that confidence intervals must be considered, that the estimates should not be used to rank counties’ overall health, and that they cannot track change over time.[5]

All files were downloaded from the publishers and validated: one record per county and measure, Census FIPS codes and names matching Vellora’s county boundaries, a single observation year per measure, and every value inside its published interval. The crude PLACES values also matched Vellora’s previously verified copy exactly, and the 2020 populations in the RUCC file sum to the Texas census total of 29,145,505. Of the 254 counties, 86 are metropolitan and 168 nonmetropolitan; metropolitan counties held 90% of the 2020 population.

4. Methodology

  • Design and unit. Ecological cross-sectional analysis with the county as the unit. Every county counts once, whatever its population; no result is a statewide estimate. Loving County, suppressed by CDC in every source, is excluded, leaving 253 counties.
  • Prevalence type. Age-adjusted PLACES prevalence is used throughout because counties’ age structures differ.[5, 11] Crude prevalence is analysed as a sensitivity check.
  • Variation (RQ1). Median, interquartile range, 10th–90th percentile ratio and range for each measure, with the median width of the published 95% intervals for PLACES measures. Quantiles use linear interpolation (type 7).
  • Clustering (RQ2). Global Moran’s I[13] with row-standardised contiguity weights (counties sharing a boundary within Texas) and a one-sided permutation test with 9,999 permutations, so the smallest attainable p-value is 0.0001.
  • Metropolitan comparison (RQ3). Two-sided Mann–Whitney U tests[14] (normal approximation with tie and continuity corrections) of nonmetropolitan against metropolitan counties, reported with the difference in medians and the probability that a randomly chosen nonmetropolitan county has the higher value.
  • Associations (RQ4). Spearman rank correlations between each of the five risk factors and each of the three outcomes (15 tests), with 95% confidence intervals from the Fisher z transform using the Bonett–Wright standard error.[16] Spearman’s ρ was chosen because it does not assume linear relationships or normal distributions.
  • Multiple testing. Benjamini–Hochberg false discovery rate adjustment[15] within each table; adjusted p-values below 0.05 are described as statistically significant. No regression model was fitted, for reasons given under Limitations.

5. Results

5.1 Variation across counties

The heart disease death rate varied most in relative terms: the county at the 90th percentile had a rate 1.67 times the county at the 10th. Among PLACES measures, the 90th-to-10th percentile ratio ranged from 1.19 (high blood pressure) to 1.54 (diabetes). However, for every PLACES measure the typical county’s 95% interval was wider than the spread between the middle half of counties (Table 1), so most differences between individual counties lie within the uncertainty of the estimates.

Table 1. Distribution across counties. PLACES measures are age-adjusted % of adults; the death rate is per 100,000 adults 35+.
MeasurenMedian (IQR)P90 ÷ P10LowestHighestMedian 95% CI width
High blood pressure25333.5 (32.2–35.1)1.1928.0 (Travis)40.0 (San Augustine)8.2 pts
Diabetes25312.2 (11.0–13.7)1.548.7 (Rockwall)21.5 (Dimmit)3.4 pts
Obesity25337.5 (35.7–39.6)1.2229.4 (Fort Bend)48.2 (Zavala)16.9 pts
Current smoking25315.3 (13.7–16.6)1.429.0 (Collin)22.9 (Newton, San Augustine)5.6 pts
No leisure-time activity25329.6 (26.8–33.0)1.5319.9 (Collin, Kendall)48.9 (Dimmit)10.2 pts
Coronary heart disease2535.9 (5.5–6.4)1.334.3 (Collin)8.5 (Dimmit)1.5 pts
Stroke2533.3 (3.0–3.6)1.412.3 (Collin)4.8 (Dimmit)0.7 pts
Heart disease deaths253402.1 (345.9–446.1)1.67205.3 (Presidio)808.0 (Polk)Not published
Loading Texas county boundaries…
Figure 1. Heart disease death rate, adults 35+ (deaths per 100,000 adults 35+, age-standardized) by county. Darker shades are higher. Classes are quantiles of this measure only, so shades are not comparable between measures. Loving County (hatched) has no published value.

5.2 Geographic clustering

Every measure was positively spatially autocorrelated: neighbouring counties had more similar values than random placement would produce (Table 2). Moran’s I ranged from 0.35 (high blood pressure) to 0.57 (diabetes), against an expected value of −0.004 under no clustering, and no permutation produced a value as large as the observed one for any measure.

Table 2. Global Moran’s I, contiguity weights, 9,999 permutations.
MeasureMoran’s IPermutation p
High blood pressure0.35≤ 0.0001
Diabetes0.57≤ 0.0001
Obesity0.36≤ 0.0001
Current smoking0.38≤ 0.0001
No leisure-time activity0.52≤ 0.0001
Coronary heart disease0.37≤ 0.0001
Stroke0.35≤ 0.0001
Heart disease deaths0.56≤ 0.0001

5.3 Metropolitan and nonmetropolitan counties

All eight measures had higher medians in the 167 nonmetropolitan counties with data than in the 86 metropolitan counties, and every difference was statistically significant after adjustment (Table 3). The probability that a randomly chosen nonmetropolitan county had the higher value ranged from 0.61 to 0.73, where 0.5 would mean no difference. The largest separations were for coronary heart disease (0.73) and current smoking (0.73).

Table 3. Nonmetropolitan (n = 167) against metropolitan (n = 86) counties.
MeasureMedian, nonmetroMedian, metroDifferenceP(nonmetro higher)Adjusted p
High blood pressure33.932.95+0.950.640.0004
Diabetes12.511.6+0.90.610.0033
Obesity38.036.4+1.60.630.0012
Current smoking15.914.1+1.80.73< 0.0001
No leisure-time activity30.527.75+2.750.66< 0.0001
Coronary heart disease6.15.6+0.50.73< 0.0001
Stroke3.43.1+0.30.68< 0.0001
Heart disease deaths415.1361.2+53.90.67< 0.0001

5.4 Risk factors and outcomes

Heart disease death rate. This outcome comes from death records, a source independent of the PLACES model. Its correlations with the risk factors, from strongest to weakest, were: current smoking (ρ = 0.54, 95% CI 0.44 to 0.63); high blood pressure (ρ = 0.37, 95% CI 0.26 to 0.48); obesity (ρ = 0.19, 95% CI 0.07 to 0.31); no leisure-time activity (ρ = 0.16, 95% CI 0.03 to 0.28); diabetes (ρ = 0.09, 95% CI −0.03 to 0.22). All except diabetes were statistically significant after adjustment (diabetes: adjusted p = 0.1323). Only current smoking and high blood pressure exceeded ρ = 0.3; the others were weak.

PLACES condition estimates. Correlations between the five risk factors and age-adjusted coronary heart disease and stroke prevalence were very high (ρ 0.83–0.94). Because all seven PLACES measures are predicted by models fitted to the same survey, drawing on the same set of candidate covariates (individual age, sex, race and ethnicity and education, and county poverty),[4] these correlations partly reflect the model and cannot be read as independent evidence of a relationship.

Table 4. Spearman ρ (95% CI) between risk factors and outcomes, n = 253 counties, age-adjusted prevalence. Adjusted p is Benjamini–Hochberg across the 15 tests. The last column repeats ρ with crude prevalence.
Risk factorOutcomeρ (95% CI)Adjusted pρ, crude
High blood pressureCoronary heart disease0.89 (0.85 to 0.92)< 0.00010.96
High blood pressureStroke0.94 (0.92 to 0.96)< 0.00010.95
High blood pressureHeart disease deaths0.37 (0.26 to 0.48)< 0.00010.31
DiabetesCoronary heart disease0.83 (0.79 to 0.87)< 0.00010.74
DiabetesStroke0.84 (0.79 to 0.88)< 0.00010.79
DiabetesHeart disease deaths0.09 (−0.03 to 0.22)0.13230.14
ObesityCoronary heart disease0.83 (0.78 to 0.87)< 0.00010.27
ObesityStroke0.85 (0.80 to 0.89)< 0.00010.38
ObesityHeart disease deaths0.19 (0.07 to 0.31)0.00240.21
Current smokingCoronary heart disease0.86 (0.82 to 0.89)< 0.00010.53
Current smokingStroke0.86 (0.82 to 0.89)< 0.00010.64
Current smokingHeart disease deaths0.54 (0.44 to 0.63)< 0.00010.56
No leisure-time activityCoronary heart disease0.86 (0.82 to 0.89)< 0.00010.40
No leisure-time activityStroke0.84 (0.79 to 0.88)< 0.00010.48
No leisure-time activityHeart disease deaths0.16 (0.03 to 0.28)0.01300.18

Sensitivity to age adjustment. Several associations changed markedly with crude prevalence. Obesity and coronary heart disease, for example, correlated at ρ = 0.83 with age-adjusted estimates but 0.27 with crude estimates. Crude county prevalence partly reflects age structure, which affects some measures more than others, so the choice of prevalence type matters for this kind of comparison.

Outcomes with each other. Coronary heart disease and stroke prevalence correlated with the death rate at ρ = 0.34 and 0.35, far less than with each other (ρ = 0.96), again consistent with shared modelling among PLACES measures.

Current smoking and heart disease deaths by county

Spearman ρ = 0.54 (95% CI 0.44 to 0.63), n = 253, BH-adjusted p < 0.0001

8.010.012.014.016.018.020.022.024.0200300400500600700800900Current smoking (% of adults, age-adjusted)Heart disease deaths (per 100,000)Anderson County: 17.2, 486Andrews County: 14.4, 415Angelina County: 17.4, 456Aransas County: 14.6, 409Archer County: 14.8, 293Armstrong County: 13.3, 399Atascosa County: 13.8, 355Austin County: 14.9, 347Bailey County: 14.9, 391Bandera County: 14.2, 310Bastrop County: 14.0, 330Baylor County: 16.9, 443Bee County: 15.0, 407Bell County: 13.6, 339Bexar County: 12.3, 328Blanco County: 11.7, 269Borden County: 12.8, 461Bosque County: 14.7, 404Bowie County: 16.3, 565Brazoria County: 11.3, 341Brazos County: 13.8, 363Brewster County: 10.3, 241Briscoe County: 17.3, 432Brooks County: 16.1, 402Brown County: 15.7, 497Burleson County: 14.2, 462Burnet County: 13.5, 305Caldwell County: 14.6, 315Calhoun County: 13.3, 368Callahan County: 14.8, 407Cameron County: 13.6, 289Camp County: 17.7, 472Carson County: 13.2, 429Cass County: 17.7, 490Castro County: 13.8, 346Chambers County: 13.1, 362Cherokee County: 17.3, 452Childress County: 15.9, 446Clay County: 14.7, 376Cochran County: 19.4, 489Coke County: 16.2, 371Coleman County: 16.0, 472Collin County: 9.0, 221Collingsworth County: 16.5, 551Colorado County: 14.1, 375Comal County: 10.4, 288Comanche County: 15.1, 462Concho County: 16.3, 322Cooke County: 14.8, 359Coryell County: 14.6, 343Cottle County: 17.8, 398Crane County: 14.5, 447Crockett County: 14.4, 384Crosby County: 16.4, 415Culberson County: 16.9, 385Dallam County: 17.0, 400Dallas County: 12.5, 312Dawson County: 17.6, 544Deaf Smith County: 16.1, 371Delta County: 16.2, 455Denton County: 9.5, 216DeWitt County: 16.1, 326Dickens County: 15.9, 436Dimmit County: 19.9, 425Donley County: 16.0, 521Duval County: 15.8, 450Eastland County: 16.5, 465Ector County: 14.2, 449Edwards County: 15.6, 383Ellis County: 12.5, 340El Paso County: 12.5, 287Erath County: 14.5, 340Falls County: 17.4, 352Fannin County: 16.7, 484Fayette County: 12.9, 366Fisher County: 13.8, 406Floyd County: 15.7, 412Foard County: 19.2, 364Fort Bend County: 9.5, 251Franklin County: 13.5, 468Freestone County: 16.2, 446Frio County: 15.7, 403Gaines County: 16.7, 426Galveston County: 12.3, 355Garza County: 16.9, 409Gillespie County: 11.9, 297Glasscock County: 10.6, 418Goliad County: 14.7, 340Gonzales County: 16.1, 346Gray County: 16.5, 535Grayson County: 14.6, 380Gregg County: 15.7, 464Grimes County: 15.9, 448Guadalupe County: 12.5, 297Hale County: 16.1, 454Hall County: 16.3, 479Hamilton County: 16.0, 380Hansford County: 14.1, 409Hardeman County: 16.5, 380Hardin County: 15.7, 521Harris County: 12.6, 293Harrison County: 15.7, 387Hartley County: 13.4, 431Haskell County: 16.8, 425Hays County: 11.2, 270Hemphill County: 12.1, 401Henderson County: 16.6, 467Hidalgo County: 13.8, 301Hill County: 15.6, 436Hockley County: 14.7, 494Hood County: 12.4, 289Hopkins County: 15.9, 451Houston County: 18.1, 438Howard County: 14.9, 588Hudspeth County: 18.7, 322Hunt County: 15.4, 408Hutchinson County: 15.0, 512Irion County: 13.0, 358Jack County: 17.7, 465Jackson County: 15.5, 368Jasper County: 19.5, 536Jeff Davis County: 14.9, 282Jefferson County: 16.2, 459Jim Hogg County: 18.2, 337Jim Wells County: 14.9, 460Johnson County: 14.3, 377Jones County: 17.3, 409Karnes County: 16.2, 386Kaufman County: 13.7, 439Kendall County: 10.0, 309Kenedy County: 15.5, 306Kent County: 12.1, 398Kerr County: 13.4, 303Kimble County: 15.8, 302King County: 17.8, 407Kinney County: 13.2, 341Kleberg County: 13.9, 389Knox County: 16.0, 420Lamar County: 17.2, 466Lamb County: 15.6, 374Lampasas County: 14.3, 339La Salle County: 17.0, 285Lavaca County: 15.1, 353Lee County: 16.6, 297Leon County: 16.5, 471Liberty County: 17.8, 523Limestone County: 17.0, 437Lipscomb County: 14.7, 364Live Oak County: 17.2, 430Llano County: 13.7, 308Lubbock County: 13.7, 407Lynn County: 16.0, 426McCulloch County: 15.7, 502McLennan County: 14.5, 361McMullen County: 16.7, 299Madison County: 17.1, 438Marion County: 18.5, 405Martin County: 13.8, 379Mason County: 12.4, 339Matagorda County: 15.5, 358Maverick County: 14.8, 309Medina County: 12.7, 370Menard County: 15.0, 366Midland County: 11.9, 367Milam County: 15.7, 411Mills County: 13.7, 431Mitchell County: 16.7, 438Montague County: 16.1, 467Montgomery County: 11.6, 315Moore County: 15.8, 427Morris County: 17.2, 528Motley County: 15.2, 385Nacogdoches County: 16.6, 428Navarro County: 17.1, 512Newton County: 22.9, 535Nolan County: 17.7, 402Nueces County: 13.4, 341Ochiltree County: 15.2, 372Oldham County: 14.9, 400Orange County: 15.8, 524Palo Pinto County: 16.7, 448Panola County: 16.0, 395Parker County: 12.7, 323Parmer County: 13.7, 339Pecos County: 15.7, 389Polk County: 18.3, 808Potter County: 17.5, 443Presidio County: 16.2, 205Rains County: 14.9, 405Randall County: 12.2, 378Reagan County: 12.9, 363Real County: 17.0, 439Red River County: 18.4, 553Reeves County: 15.9, 404Refugio County: 15.5, 390Roberts County: 12.7, 464Robertson County: 16.0, 417Rockwall County: 9.7, 285Runnels County: 14.7, 352Rusk County: 16.3, 413Sabine County: 17.3, 532San Augustine County: 22.9, 485San Jacinto County: 18.0, 436San Patricio County: 14.5, 433San Saba County: 16.1, 389Schleicher County: 12.6, 353Scurry County: 14.6, 366Shackelford County: 12.8, 424Shelby County: 18.9, 435Sherman County: 12.6, 385Smith County: 14.0, 424Somervell County: 13.3, 325Starr County: 17.6, 339Stephens County: 16.3, 483Sterling County: 11.2, 315Stonewall County: 16.1, 404Sutton County: 14.4, 361Swisher County: 18.3, 435Tarrant County: 12.6, 290Taylor County: 13.9, 409Terrell County: 10.0, 323Terry County: 17.1, 527Throckmorton County: 15.9, 442Titus County: 16.9, 489Tom Green County: 13.3, 351Travis County: 9.3, 233Trinity County: 17.4, 498Tyler County: 20.3, 589Upshur County: 16.8, 458Upton County: 17.0, 421Uvalde County: 13.8, 430Val Verde County: 13.7, 322Van Zandt County: 15.9, 434Victoria County: 14.2, 357Walker County: 17.2, 379Waller County: 13.6, 348Ward County: 14.7, 435Washington County: 13.0, 274Webb County: 13.1, 261Wharton County: 14.7, 417Wheeler County: 16.3, 540Wichita County: 15.3, 431Wilbarger County: 18.6, 352Willacy County: 17.4, 295Williamson County: 9.6, 236Wilson County: 12.3, 330Winkler County: 14.0, 431Wise County: 14.6, 430Wood County: 15.7, 447Yoakum County: 13.3, 523Young County: 14.9, 544Zapata County: 17.8, 263Zavala County: 18.7, 496
Figure 2. Each point is one county. Spearman’s ρ measures how consistently the two measures rank together across counties. It is an ecological association, not evidence that one causes the other.

6. Discussion

Across Texas counties, cardiovascular conditions, their risk factors and heart disease deaths are unevenly distributed, geographically clustered and less favourable outside metropolitan areas. The metropolitan–nonmetropolitan pattern in heart disease mortality is consistent with national CDC surveillance reporting higher age-adjusted heart disease death rates in nonmetropolitan areas.[12] Because the death rate is age-standardized, the difference is not an artefact of nonmetropolitan counties having older populations, though this study cannot say what does explain it.

The most informative associations are those with the death rate, because it is measured independently of the PLACES model. Current smoking and high blood pressure, both identified by CDC as key heart disease risk factors,[2] showed the clearest county-level associations with it. The weak association for diabetes, despite CDC listing it as a heart disease risk factor,[2] illustrates why county correlations should not be used to judge the importance of a risk factor for individuals: county averages can hide very different relationships among the people within them.[17]

The very high agreement between PLACES risk factors and PLACES condition estimates is best read as a property of the estimates rather than a finding about disease. Readers comparing PLACES measures with one another should expect such correlations and look to independent outcomes, such as vital records, for confirmation.

7. Limitations

  • Ecological design. Associations between counties do not show associations between individuals, and none of the results is causal.[17]
  • Modelled estimates. PLACES values are model predictions, not counts, and share inputs. Correlations among them are inflated by design, and the models’ county random effects and shared covariates can also add spatial smoothness to the Moran’s I results.[4]
  • Uncertainty. Many county intervals are wide (Table 1). CDC advises against drawing conclusions from estimates with very broad intervals and against ranking counties.[5] This report names the counties at each extreme only to describe the range, not to rank them.
  • Self-report and diagnosis. Condition and risk-factor measures count adults who report a diagnosis or behaviour; undiagnosed conditions are missed, and access to care can affect diagnosis rates.
  • Different populations and standards. PLACES covers adults 18 and over; the death rate covers adults 35 and over, age-standardized to the 2000 U.S. population.[8, 11] The two kinds of measure are never put on one scale.
  • Smoothing and timing. CDC spatially smooths the death rates, borrowing information from neighbouring counties,[9] which raises their Moran’s I. Survey data are from 2023 and deaths from 2022–2024, so the measures are close in time but not identical.
  • Boundaries and groups. Contiguity uses Texas counties only, so border counties have fewer neighbours (El Paso County has one). The metropolitan split is a coarse two-way grouping of a nine-level code.
  • No adjustment for confounders. No regression was fitted: the PLACES predictors are highly collinear and share model inputs, so coefficients would be unstable and hard to interpret. Income, race and ethnicity, access to care and other factors that differ between counties were not controlled.
  • Unweighted. Each county counts once; the 168 nonmetropolitan counties held 10% of the population, so results describe counties, not the average Texan.

8. Conclusion

Using only public CDC and USDA data, this analysis finds that cardiovascular measures across Texas counties are geographically clustered and consistently less favourable in nonmetropolitan counties, and that current smoking and high blood pressure are the risk-factor estimates most closely associated with county heart disease death rates. The evidence is descriptive and ecological. It can help direct attention to places and questions worth studying further, but it cannot identify causes or describe any individual’s risk.

9. References

  1. [1] Centers for Disease Control and Prevention. Heart Disease Facts. Accessed 2026-10-18. https://www.cdc.gov/heart-disease/data-research/facts-stats/index.html
  2. [2] Centers for Disease Control and Prevention. Heart Disease Risk Factors. Accessed 2026-10-18. https://www.cdc.gov/heart-disease/risk-factors/index.html
  3. [3] Centers for Disease Control and Prevention. PLACES: Local Data for Better Health, County Data, 2025 release (dataset swc5-untb; rows updated 2025-12-04). Accessed 2026-10-22. https://data.cdc.gov/d/swc5-untb
  4. [4] Centers for Disease Control and Prevention. PLACES: Methodology. Accessed 2026-10-22. https://www.cdc.gov/places/methodology/index.html
  5. [5] Centers for Disease Control and Prevention. PLACES: Frequently Asked Questions. Accessed 2026-10-22. https://www.cdc.gov/places/faqs/index.html
  6. [6] Centers for Disease Control and Prevention. PLACES Measure Definitions: Health Risk Behaviors. Accessed 2026-10-22. https://www.cdc.gov/places/measure-definitions/health-risk-behaviors.html
  7. [7] Centers for Disease Control and Prevention. PLACES Measure Definitions: Health Outcomes. Accessed 2026-10-22. https://www.cdc.gov/places/measure-definitions/health-outcomes.html
  8. [8] Centers for Disease Control and Prevention, Division for Heart Disease and Stroke Prevention. Heart Disease Mortality Data Among US Adults (35+) by State/Territory and County – 2022–2024 (dataset mri7-5jtw). Accessed 2026-10-18. https://data.cdc.gov/d/mri7-5jtw
  9. [9] Centers for Disease Control and Prevention. Atlas of Heart Disease and Stroke: Statistical Methods. Accessed 2026-10-22. https://www.cdc.gov/heart-disease-stroke-atlas/statistical-methods/index.html
  10. [10] U.S. Department of Agriculture, Economic Research Service. Rural-Urban Continuum Codes, 2023 (data and documentation). Accessed 2026-10-22. https://www.ers.usda.gov/data-products/rural-urban-continuum-codes
  11. [11] National Center for Health Statistics. Age adjustment. Health, United States: Sources and Definitions. Accessed 2026-10-22. https://www.cdc.gov/nchs/hus/sources-definitions/age-adjustment.htm
  12. [12] Moy E, Garcia MC, Bastian B, et al. Leading causes of death in nonmetropolitan and metropolitan areas — United States, 1999–2014. MMWR Surveill Summ. 2017;66(1):1–8. https://doi.org/10.15585/mmwr.ss6601a1
  13. [13] Moran PAP. Notes on continuous stochastic phenomena. Biometrika. 1950;37(1/2):17–23. https://doi.org/10.2307/2332142
  14. [14] Mann HB, Whitney DR. On a test of whether one of two random variables is stochastically larger than the other. Ann Math Stat. 1947;18(1):50–60. https://doi.org/10.1214/aoms/1177730491
  15. [15] Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B. 1995;57(1):289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x
  16. [16] Bonett DG, Wright TA. Sample size requirements for estimating Pearson, Kendall and Spearman correlations. Psychometrika. 2000;65(1):23–28. https://doi.org/10.1007/BF02294183
  17. [17] Robinson WS. Ecological correlations and the behavior of individuals. Am Sociol Rev. 1950;15(3):351–357. https://doi.org/10.2307/2087176

10. Reproducibility

Every number in this report, its PDF and its figures is read from one results file, generated by the analysis code below from the published input files. The permutation tests use a fixed seed (20261022), so rerunning the analysis gives identical results. An independent Python implementation using SciPy and NumPy recomputed every descriptive statistic, correlation, confidence interval, Mann–Whitney test, adjusted p-value and Moran’s I and matched the results file; its own permutation tests also gave p ≤ 0.001 for every measure.

To reproduce: download the inputs with node scripts/fetch-cardio-study.mjs, regenerate the results with npm run research:cardio, and check them with uvx --with scipy --with numpy python scripts/verify-cardio-study.py. CDC may revise its datasets; results here use the versions recorded in the inputs file.

This report is an independent secondary-data analysis by Vellora Research. It has not been peer reviewed, is not affiliated with any university or public agency, and is not medical advice. CDC and USDA data are used without endorsement by either agency.