Analysis Of Stroke Burden Indices

In this notebook, we analyze the final indices developed for our Stroke Burden Index project. The objective is to evaluate county-level patterns in stroke risk, healthcare resource availability, geographic accessibility, and overall stroke burden across New York, New Jersey, and Connecticut.

The analyses presented in this notebook are performed using the finalized county-level database and the three main indices:

• Stroke Risk Index (SRI): Measures demographic, socioeconomic, and health-related vulnerability to stroke.
• Stroke Care Access Index (SCAI): Measures the availability of healthcare resources related to stroke prevention and treatment.
• Geographic Accessibility Index (GAI): Measures physical accessibility to stroke care using travel time and distance to certified stroke centers.

Unlike the methodology notebooks, which focus on data preparation and index construction, this notebook examines the characteristics of the completed indices and explores how they vary across the tri-state region.

---

Objectives

The primary objectives of this notebook are to:

• Examine the distribution of each index.
• Compare index values across counties and states.
• Identify counties with the highest and lowest index scores.
• Evaluate relationships between the indices using correlation and visualization.
• Explore geographic patterns in stroke burden and healthcare access.
• Identify counties that may represent priorities for future public health interventions.

---

Data

This notebook uses the finalized county-level database generated during the data integration and index construction process. The database contains one observation for each of the 91 counties in New York, New Jersey, and Connecticut and includes demographic, socioeconomic, health, healthcare resource, geographic accessibility, and mortality variables, along with the final SRI, SCAI, GAI, and SBPI scores.

Importing Packages

import sqlite3
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.lines import Line2D
from IPython.display import display
import seaborn as sns
from matplotlib.colors import Normalize

Loading Data

We load the database to get state and county names, as well as all other demographic, socioeconomic, geographic, mortality, heathcare, and accessibility variables for a complete analysis.

con = sqlite3.connect("../data/stroke_burden.db")

master = pd.read_sql("""
SELECT *
FROM master
""", con)

Next, we load our csv file containing the calculated index values.

indices = pd.read_csv("../data/indices.csv")

# Making sure fips matches

master["fips"] = master["fips"].astype(str).str.zfill(5)
indices["fips"] = indices["fips"].astype(str).str.zfill(5)

We merge these together dataframes together to obtain a full dataframe containing:

• FIPS
• County
• State
• SRI
• SCAI
• GAI
• SBPI
• SBPI_class

df = master.merge(indices, on="fips", how="inner")

Summary Statistics

In this section, we will look at the distributions of the four indices using descriptive statistics to compare their ranges and variability across counties.

summary = df[["sri","scai","gai","sbpi"]].describe().T
summary

	count	mean	std	min	25%	50%	75%	max
sri	91.0	53.007877	24.596933	0.0	35.13945	54.4988	72.66540	100.0000
scai	91.0	35.480171	20.205070	0.0	21.67875	33.7155	49.63950	100.0000
gai	91.0	50.812421	21.613708	0.0	34.07680	50.6657	66.59595	100.0000
sbpi	91.0	55.697404	19.268748	0.0	42.02055	53.5334	75.10880	92.0213

All indices except sbpi have a minimum of 0 and a maximum of 100.

Distribution of Index Scores

Histograms and boxplots will be generated to examine the distributions of the four indices and identify potential outliers.

Style Choices

index_cols = ["sri", "scai", "gai", "sbpi"]

colors = {
    "sri": "firebrick",
    "scai": "steelblue",
    "gai": "forestgreen",
    "sbpi": "darkorchid"
}

titles = {
    "sri": "Stroke Risk Index (SRI)",
    "scai": "Stroke Care Access Index (SCAI)",
    "gai": "Geographic Accessibility Index (GAI)",
    "sbpi": "Stroke Burden Priority Index (SBPI)"
}

fig, axes = plt.subplots(2, 2, figsize=(14, 10))
axes = axes.flatten()

for ax, col in zip(axes, index_cols):

    data = df[col]

    mean = data.mean()
    median = data.median()
    std = data.std()
    minimum = data.min()
    maximum = data.max()
    skew = data.skew()

    # Histogram
    ax.hist(data,
            bins=15,
            edgecolor="black",
            color=colors[col],
            alpha=0.8)

    # Mean & Median
    ax.axvline(mean,
               color="purple",
               linestyle="--",
               linewidth=2)

    ax.axvline(median,
               color="blue",
               linestyle=":",
               linewidth=2)

    # Title
    ax.set_title(titles[col], fontsize=14, fontweight="bold")

    ax.set_xlabel("Index Score")
    ax.set_ylabel("Number of Counties")

    ax.grid(alpha=0.3)

    # Statistics box
    stats = (
        f"Mean: {mean:.2f}\n"
        f"Median: {median:.2f}\n"
        f"SD: {std:.2f}\n"
        f"Min: {minimum:.2f}\n"
        f"Max: {maximum:.2f}\n"
        f"Skew: {skew:.2f}"
    )

    ax.text(
        0.97, 0.97,
        stats,
        transform=ax.transAxes,
        fontsize=10,
        verticalalignment="top",
        horizontalalignment="right",
        bbox=dict(
            boxstyle="round",
            facecolor="white",
            alpha=0.85
        )
    )

    # Mean/Median legend
    ax.plot([], [], color="purple", linestyle="--", label="Mean")
    ax.plot([], [], color="blue", linestyle=":", label="Median")
    ax.legend(loc="upper left")

plt.tight_layout()

plt.savefig(
    "../docs/images/index_histograms.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

The histograms summarize the distributions of the four indices across all counties. The SRI, GAI, and SBPI exhibit relatively symmetric distributions centered near the middle of the standardized 0–100 scale, indicating that counties span the full spectrum of stroke risk and accessibility. The SCAI distribution is more concentrated toward lower values, reflecting that relatively few counties possess abundant healthcare resources. Mean and median values are generally similar, suggesting limited skewness after transformation and standardization.

fig, axes = plt.subplots(2, 2, figsize=(14, 8))
axes = axes.flatten()

for ax, col in zip(axes, index_cols):

    data = df[col]

    # Summary statistics
    mean = data.mean()
    median = data.median()
    q1 = data.quantile(0.25)
    q3 = data.quantile(0.75)
    iqr = q3 - q1
    minimum = data.min()
    maximum = data.max()

    # Create boxplot
    bp = ax.boxplot(
        data,
        vert=False,
        patch_artist=True,
        widths=0.6,
        showmeans=True,
        meanline=False,
        meanprops=dict(
            marker="D",
            markerfacecolor="#7a0404",
            markeredgecolor="black",
            markersize=8
        ),
        medianprops=dict(
            color="black",
            linewidth=2.5
        ),
        whiskerprops=dict(
            color="dimgray",
            linewidth=1.5
        ),
        capprops=dict(
            color="dimgray",
            linewidth=1.5
        ),
        flierprops=dict(
            marker="o",
            markersize=6,
            markerfacecolor="darkorange",
            markeredgecolor="black",
            alpha=0.8
        )
    )

    # Color the box
    bp["boxes"][0].set(
        facecolor=colors[col],
        edgecolor="black",
        linewidth=1.5
    )

    # Titles and labels
    ax.set_title(titles[col], fontsize=14, fontweight="bold")
    ax.set_xlabel("Index Score")
    ax.set_yticks([])
    ax.grid(axis="x", alpha=0.3)

    # Statistics box
    stats = (
        f"Min: {minimum:.2f}\n"
        f"Q1: {q1:.2f}\n"
        f"Median: {median:.2f}\n"
        f"Q3: {q3:.2f}\n"
        f"Max: {maximum:.2f}\n"
        f"IQR: {iqr:.2f}"
    )

    ax.text(
        0.98,
        0.95,
        stats,
        transform=ax.transAxes,
        fontsize=10,
        va="top",
        ha="right",
        bbox=dict(
            boxstyle="round",
            facecolor="white",
            edgecolor="gray",
            alpha=0.9
        )
    )

    # Legend
    legend_elements = [
        Line2D([0], [0],
               color="black",
               lw=2.5,
               label="Median"),
        Line2D([0], [0],
               marker="D",
               color="#7a0404",
               markeredgecolor="black",
               linestyle="",
               markersize=8,
               label="Mean")
    ]

    ax.legend(handles=legend_elements, loc="lower left", fontsize=9)

plt.tight_layout()

plt.savefig(
    "../docs/images/index_boxplots.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

The boxplots provide a concise summary of each index's central tendency and variability. The relatively wide interquartile ranges indicate meaningful differences between counties, supporting the usefulness of the indices for distinguishing geographic variation. SCAI contains a small number of high-scoring outliers, representing counties with substantially greater healthcare resource availability than the regional average. Overall, the indices exhibit sufficient variability to support meaningful comparisons across counties.

County Rankings

The highest- and lowest-scoring counties were identified for each index.

Top 10:

header_colors = {
    "sri": "#C44E52",
    "scai": "#4C72B0",
    "gai": "#55A868",
    "sbpi": "#8172B2"
}

for col in index_cols:

    print(f"\n{titles[col]} - Top 10 Counties\n")

    top10 = (
        df.nlargest(10, col)[["county", "state", col]]
        .reset_index(drop=True)
    )

    top10.index += 1
    top10.index.name = "Rank"

    # Choose color map based on interpretation of the index
    
    if col in ["sri", "sbpi"]:
        cmap = "RdYlGn_r"   
    else:
        cmap = "RdYlGn"    

    display(
        top10.style
            .format({col: "{:.2f}"})
            .background_gradient(
                subset=[col],
                cmap=cmap,
                vmin=0,
                vmax=100
            )
            .set_caption(f"Top 10 Counties by {titles[col]}")
            .set_properties(**{
                "text-align": "center"
            })
            .set_table_styles([
    {
        "selector": "th",
        "props": [
            ("background-color", header_colors[col]),
            ("color", "white"),
            ("font-weight", "bold"),
            ("text-align", "center")
        ]
    },
    {
        "selector": "caption",
        "props": [
            ("font-size", "14pt"),
            ("font-weight", "bold")
        ]
    }
])
    )

Stroke Risk Index (SRI) - Top 10 Counties

Top 10 Counties by Stroke Risk Index (SRI)

	county	state	sri
Rank
1	Bronx	NY	100.00
2	Cumberland	NJ	95.12
3	Cattaraugus	NY	91.76
4	Hamilton	NY	91.32
5	Delaware	NY	89.39
6	Franklin	NY	87.57
7	Chautauqua	NY	86.23
8	St. Lawrence	NY	85.87
9	Fulton	NY	85.73
10	Chenango	NY	84.67

Stroke Care Access Index (SCAI) - Top 10 Counties

Top 10 Counties by Stroke Care Access Index (SCAI)

	county	state	scai
Rank
1	New York	NY	100.00
2	Warren	NY	78.91
3	Otsego	NY	76.00
4	Albany	NY	73.47
5	Nassau	NY	71.49
6	Westchester	NY	71.28
7	Monroe	NY	68.92
8	Onondaga	NY	65.68
9	Bergen	NJ	63.41
10	Monmouth	NJ	59.20

Geographic Accessibility Index (GAI) - Top 10 Counties

Top 10 Counties by Geographic Accessibility Index (GAI)

	county	state	gai
Rank
1	New York	NY	100.00
2	Kings	NY	96.88
3	Bronx	NY	87.04
4	Broome	NY	86.25
5	Essex	NJ	86.03
6	Queens	NY	84.93
7	Hartford	CT	83.95
8	Passaic	NJ	82.07
9	Onondaga	NY	79.01
10	Hudson	NJ	77.55

Stroke Burden Priority Index (SBPI) - Top 10 Counties

Top 10 Counties by Stroke Burden Priority Index (SBPI)

	county	state	sbpi
Rank
1	Hamilton	NY	92.02
2	Delaware	NY	87.99
3	Chenango	NY	83.51
4	Essex	NY	83.35
5	Seneca	NY	82.08
6	Orleans	NY	81.85
7	Fulton	NY	81.52
8	St. Lawrence	NY	81.19
9	Lewis	NY	80.89
10	Cattaraugus	NY	80.27

Bottom 10

for col in index_cols:

    print(f"\n{titles[col]} - Bottom 10 Counties\n")

    bottom10 = (
        df.nsmallest(10, col)[["county", "state", col]]
        .reset_index(drop=True)
    )

    bottom10.index += 1
    bottom10.index.name = "Rank"

    # Choose color map based on index direction
    
    if col in ["sri", "sbpi"]:
        cmap = "RdYlGn_r"
    else:
        cmap = "RdYlGn"

    display(
        bottom10.style
            .format({col: "{:.2f}"})
            .background_gradient(
                subset=[col],
                cmap=cmap,
                vmin=0,
                vmax=100
            )
            .set_caption(f"Bottom 10 Counties by {titles[col]}")
            .set_properties(**{
                "text-align": "center"
            })
            .set_table_styles([
                {
                    "selector": "th",
                    "props": [
                        ("background-color", header_colors[col]),
                        ("color", "white"),
                        ("font-weight", "bold"),
                        ("text-align", "center"),
                        ("font-size", "11pt")
                    ]
                },
                {
                    "selector": "td",
                    "props": [
                        ("text-align", "center"),
                        ("padding", "6px")
                    ]
                },
                {
                    "selector": "caption",
                    "props": [
                        ("caption-side", "top"),
                        ("font-size", "14pt"),
                        ("font-weight", "bold"),
                        ("padding", "8px")
                    ]
                }
            ])
    )

Stroke Risk Index (SRI) - Bottom 10 Counties

Bottom 10 Counties by Stroke Risk Index (SRI)

	county	state	sri
Rank
1	New York	NY	0.00
2	Morris	NJ	1.19
3	Somerset	NJ	6.08
4	Tompkins	NY	7.59
5	Hunterdon	NJ	8.35
6	Bergen	NJ	12.60
7	Nassau	NY	15.17
8	Fairfield	CT	16.38
9	Westchester	NY	17.11
10	Putnam	NY	18.02

Stroke Care Access Index (SCAI) - Bottom 10 Counties

Bottom 10 Counties by Stroke Care Access Index (SCAI)

	county	state	scai
Rank
1	Hamilton	NY	0.00
2	Orleans	NY	0.26
3	Seneca	NY	0.87
4	Tioga	NY	3.41
5	Washington	NY	4.04
6	Greene	NY	4.83
7	Herkimer	NY	5.27
8	Schoharie	NY	8.72
9	Cayuga	NY	9.54
10	Wyoming	NY	10.55

Geographic Accessibility Index (GAI) - Bottom 10 Counties

Bottom 10 Counties by Geographic Accessibility Index (GAI)

	county	state	gai
Rank
1	Clinton	NY	0.00
2	Franklin	NY	4.95
3	Essex	NY	6.29
4	Delaware	NY	15.50
5	Chautauqua	NY	17.50
6	Hamilton	NY	18.20
7	Steuben	NY	21.19
8	Lewis	NY	23.48
9	Schoharie	NY	24.27
10	Chenango	NY	24.74

Stroke Burden Priority Index (SBPI) - Bottom 10 Counties

Bottom 10 Counties by Stroke Burden Priority Index (SBPI)

	county	state	sbpi
Rank
1	New York	NY	0.00
2	Nassau	NY	21.66
3	Westchester	NY	21.69
4	Morris	NJ	21.98
5	Bergen	NJ	22.45
6	Somerset	NJ	24.02
7	Albany	NY	27.37
8	Hunterdon	NJ	28.90
9	Monmouth	NJ	30.82
10	Fairfield	CT	32.24

Relationships Between the Indices

Pearson correlations were calculated to evaluate how the four indices relate to one another.

# Constructing correlation matrix

corr = df[
    ["sri","scai","gai","sbpi"]
].corr()

corr

	sri	scai	gai	sbpi
sri	1.000000	-0.543384	-0.517596	0.925314
scai	-0.543384	1.000000	0.544884	-0.783637
gai	-0.517596	0.544884	1.000000	-0.726108
sbpi	0.925314	-0.783637	-0.726108	1.000000

# Mask the UPPER triangle so the LOWER triangle is shown

mask = np.triu(np.ones_like(corr, dtype=bool))

plt.figure(figsize=(8, 7))

ax = sns.heatmap(
    corr,
    mask=mask,
    annot=True,
    fmt=".2f",
    cmap="vlag",
    center=0,
    vmin=-1,
    vmax=1,
    square=True,
    linewidths=2,
    linecolor="white",
    cbar_kws={
        "label": "Pearson Correlation",
        "shrink": 0.8
    },
    annot_kws={
        "size": 12,
        "weight": "bold"
    }
)

# Title
ax.set_title(
    "Correlation Between Indices",
    fontsize=18,
    fontweight="bold",
    pad=20
)

# Tick labels
labels = ["SRI", "SCAI", "GAI", "SBPI"]

ax.set_xticklabels(
    labels,
    fontsize=12,
    fontweight="bold",
    rotation=30,
    ha="right"
)

ax.set_yticklabels(
    labels,
    fontsize=12,
    fontweight="bold",
    rotation=0
)

label_colors = {
    "SRI": "#C44E52",
    "SCAI": "#4C72B0",
    "GAI": "#55A868",
    "SBPI": "#8172B2"
}

for tick in ax.get_xticklabels():
    tick.set_color(label_colors[tick.get_text()])

for tick in ax.get_yticklabels():
    tick.set_color(label_colors[tick.get_text()])

# Make annotation colors dynamic

for text in ax.texts:
    try:
        value = float(text.get_text())
        text.set_color("white" if abs(value) > 0.65 else "black")
    except ValueError:
        pass

plt.tight_layout()

plt.savefig(
    "../docs/images/index_correlation_heatmap.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

The correlation matrix demonstrates that the indices capture related but distinct dimensions of stroke burden. SRI is moderately negatively correlated with both SCAI and GAI, indicating that counties with higher stroke risk generally have poorer healthcare access. SCAI and GAI exhibit a moderate positive correlation, suggesting that counties with stronger healthcare infrastructure also tend to have shorter travel times to stroke care. As expected, SBPI is strongly positively correlated with SRI because stroke risk is the primary driver of overall priority while still incorporating accessibility measures.

Note: For SRI and SBPI, a higher score indicates worse access.

Pairwise Relationships

Scatterplots were generated to visualize relationships among the component indices.

pairs = [
    ("sri", "scai"),
    ("sri", "gai"),
    ("scai", "gai"),
    ("sbpi", "sri")
]

titles = {
    ("sri", "scai"): "SRI vs SCAI",
    ("sri", "gai"): "SRI vs GAI",
    ("scai", "gai"): "SCAI vs GAI",
    ("sbpi", "sri"): "SBPI vs SRI"
}

colors = {
    ("sri", "scai"): "#4C72B0",
    ("sri", "gai"): "#55A868",
    ("scai", "gai"): "#8172B2",
    ("sbpi", "sri"): "#C44E52"
}

fig, axes = plt.subplots(2, 2, figsize=(12, 10))
axes = axes.flatten()

for ax, (x, y) in zip(axes, pairs):

    sns.regplot(
        data=df,
        x=x,
        y=y,
        ax=ax,
        scatter_kws={
            "s": 60,
            "alpha": 0.75,
            "edgecolor": "white",
            "linewidths": 0.6,
            "color": colors[(x, y)]
        },
        line_kws={
            "color": "black",
            "linewidth": 2
        }
    )

    # Correlation coefficients
    
    r = df[[x, y]].corr().iloc[0, 1]

    ax.text(
        0.05,
        0.95,
        f"$r$ = {r:.2f}",
        transform=ax.transAxes,
        fontsize=11,
        fontweight="bold",
        va="top",
        bbox=dict(
            facecolor="white",
            alpha=0.9,
            edgecolor="gray"
        )
    )

    ax.set_title(
        titles[(x, y)],
        fontsize=14,
        fontweight="bold"
    )

    ax.grid(alpha=0.3)

plt.tight_layout()

plt.savefig(
    "../docs/images/index_relationships.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

The pairwise regression plots further illustrate the relationships among the indices. Negative associations between SRI and both accessibility indices reinforce the observation that higher-risk counties often experience reduced access to care. Conversely, SCAI and GAI display a positive relationship, indicating that counties with greater healthcare capacity also tend to have better geographic accessibility. The strong relationship between SBPI and SRI confirms that the composite priority index effectively incorporates stroke risk while still reflecting accessibility information.

Comparison by State

Average index scores were compared across New York, New Jersey, and Connecticut.

state_summary = df.groupby("state")[
    ["sri","scai","gai","sbpi"]
].agg(["mean","std"])

state_summary

	sri		scai		gai		sbpi
	mean	std	mean	std	mean	std	mean	std
state
CT	35.925612	14.756776	39.329225	13.910687	51.898988	14.950187	45.784238	10.069573
NJ	38.569557	23.982614	41.174671	13.486522	63.666486	12.137348	44.199081	14.653719
NY	60.102439	22.756853	33.054737	22.392013	46.318423	23.188991	60.871116	19.497387

# Creating state color palette

colors = {
    "NY": "#4C72B0",
    "NJ": "#55A868",
    "CT": "#C44E52"
}

# Setting titles

titles = {
    "sri": "Stroke Risk Index (SRI)",
    "scai": "Stroke Care Access Index (SCAI)",
    "gai": "Geographic Accessibility Index (GAI)",
    "sbpi": "Stroke Burden Priority Index (SBPI)"
}

fig, axes = plt.subplots(2, 2, figsize=(14, 10))
axes = axes.flatten()

for ax, col in zip(axes, index_cols):

    sns.violinplot(
        data=df,
        x="state",
        y=col,
        hue="state",
        palette=colors,
        inner=None,
        legend=False,
        ax=ax
    )

    sns.boxplot(
        data=df,
        x="state",
        y=col,
        width=0.18,
        showfliers=False,
        boxprops=dict(facecolor="white"),
        ax=ax
    )

    ax.set_title(
        titles[col],
        fontsize=14,
        fontweight="bold"
    )

    ax.set_xlabel("")
    ax.set_ylabel("Index Score")
    ax.grid(alpha=0.3)

plt.tight_layout()

plt.savefig(
    "../docs/images/state_comparison.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

The violin plots compare index distributions across New York, New Jersey, and Connecticut. New York exhibits the greatest variability across all indices, reflecting the coexistence of highly urbanized counties with excellent access and remote rural counties with substantial barriers to care. New Jersey generally demonstrates stronger healthcare access and lower variability due to its dense healthcare network, while Connecticut occupies an intermediate position. These state-level differences highlight how geography contributes to disparities in stroke risk and accessibility.

Relationship With Stroke Mortality

Correlations between the composite indices and county-level acute stroke mortality were examined to evaluate whether the indices capture meaningful variation in stroke outcomes.

mortality_corr = df[
    [
        "acute_stroke_mortality_per_100k",
        "sri",
        "scai",
        "gai",
        "sbpi"
    ]
].corr()

mortality_corr

	acute_stroke_mortality_per_100k	sri	scai	gai	sbpi
acute_stroke_mortality_per_100k	1.000000	0.432161	-0.228511	-0.347383	0.425647
sri	0.432161	1.000000	-0.543384	-0.517596	0.925314
scai	-0.228511	-0.543384	1.000000	0.544884	-0.783637
gai	-0.347383	-0.517596	0.544884	1.000000	-0.726108
sbpi	0.425647	0.925314	-0.783637	-0.726108	1.000000

# Setting titles 

titles = {
    "sri": "Stroke Risk Index (SRI)",
    "scai": "Stroke Care Access Index (SCAI)",
    "gai": "Geographic Accessibility Index (GAI)",
    "sbpi": "Stroke Burden Priority Index (SBPI)"
}

# Setting colors

colors = {
    "sri": "#C44E52",
    "scai": "#4C72B0",
    "gai": "#55A868",
    "sbpi": "#8172B2"
}

fig, axes = plt.subplots(2, 2, figsize=(14, 10))
axes = axes.flatten()

for ax, col in zip(axes, index_cols):

    sns.regplot(
        data=df,
        x=col,
        y="acute_stroke_mortality_per_100k",
        ax=ax,
        ci=95,
        scatter_kws={
            "s": 60,
            "alpha": 0.75,
            "color": colors[col],
            "edgecolor": "white",
            "linewidths": 0.6
        },
        line_kws={
            "color": "black",
            "linewidth": 2
        }
    )

    # Pearson correlation
    
    r = df[[col, "acute_stroke_mortality_per_100k"]].corr().iloc[0, 1]
    r2 = r**2

    ax.text(
        0.05,
        0.95,
        f"$r$ = {r:.2f}\n$R^2$ = {r2:.2f}",
        transform=ax.transAxes,
        fontsize=11,
        fontweight="bold",
        va="top",
        bbox=dict(
            facecolor="white",
            edgecolor="gray",
            alpha=0.9
        )
    )

    ax.set_title(
        f"{titles[col]} vs Acute Stroke Mortality",
        fontsize=13,
        fontweight="bold"
    )

    ax.set_xlabel(titles[col])
    ax.set_ylabel("Acute Stroke Mortality\n(per 100,000)")
    ax.grid(alpha=0.3)

plt.tight_layout()

plt.savefig(
    "../docs/images/mortality_relationships.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

The scatterplots above compare each index with observed county-level acute stroke mortality. SRI demonstrates the strongest positive relationship with mortality, suggesting that higher estimated stroke risk corresponds with higher observed mortality rates. Both SCAI and GAI exhibit negative relationships with mortality, indicating that counties with better healthcare resources and shorter travel times generally experience lower mortality. SBPI also demonstrates a positive association with mortality, providing evidence that the integrated index successfully identifies counties with greater overall stroke burden.

Stroke Burden Priority Counties

The Stroke Burden Priority Index (SBPI) combines stroke risk, healthcare resource availability, and geographic accessibility into a single measure for identifying counties that may benefit from targeted public health interventions.

Ranked Table

priority = (
    df.sort_values("sbpi", ascending=False)
      [["county", "state", "sbpi", "sbpi_class"]]
      .head(20)
)

priority.index += 1
priority.index.name = "Rank"

display(
    priority.style
    .format({"sbpi": "{:.2f}"})
    .background_gradient(
        subset=["sbpi"],
        cmap="RdYlGn_r",
        vmin=0,
        vmax=100
    )
    .set_caption("Top 20 Counties by Stroke Burden Priority Index")
)

Top 20 Counties by Stroke Burden Priority Index

	county	state	sbpi	sbpi_class
Rank
21	Hamilton	NY	92.02	4
13	Delaware	NY	87.99	4
9	Chenango	NY	83.51	4
16	Essex	NY	83.35	4
50	Seneca	NY	82.08	3
37	Orleans	NY	81.85	4
18	Fulton	NY	81.52	4
45	St. Lawrence	NY	81.19	3
25	Lewis	NY	80.89	4
5	Cattaraugus	NY	80.27	3
7	Chautauqua	NY	79.59	3
17	Franklin	NY	79.42	2
22	Herkimer	NY	78.81	3
62	Yates	NY	78.31	4
49	Schuyler	NY	78.30	3
68	Cumberland	NJ	77.63	3
20	Greene	NY	77.30	2
58	Washington	NY	77.03	2
48	Schoharie	NY	77.01	2
61	Wyoming	NY	76.79	2

Bar Chart

top20 = (
    df.nlargest(20, "sbpi")
      .sort_values("sbpi")
)

plt.figure(figsize=(10,8))

sns.barplot(
    data=top20,
    x="sbpi",
    y="county",
    hue="state",
    dodge=False,
    palette={
        "NY":"#4C72B0",
        "NJ":"#55A868",
        "CT":"#C44E52"
    }
)

plt.title(
    "Worst 20 Counties by Stroke Burden Priority Index",
    fontsize=16,
    fontweight="bold"
)

plt.xlabel("SBPI Score")
plt.ylabel("County")
plt.grid(axis="x", alpha=.3)

plt.legend(title="State")

plt.tight_layout()
plt.show()

Showing SBPI Classes

class_counts = (
    df["sbpi_class"]
      .value_counts()
      .sort_index()
)

class_counts

sbpi_class
1    55
2    16
3    10
4    10
Name: count, dtype: int64

plt.figure(figsize=(7,5))

sns.countplot(
    data=df,
    x="sbpi_class",
    hue="sbpi_class",
    palette="RdYlGn_r",
    legend=False
)

plt.title(
    "Distribution of Counties Across SBPI Priority Classes",
    fontsize=15,
    fontweight="bold"
)

plt.xlabel("Priority Class")
plt.ylabel("Number of Counties")

plt.grid(axis="y", alpha=.3)


plt.savefig(
    "../docs/images/sbpi_priority_counties.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

Summary Statistics by Class

df.groupby("sbpi_class")[
    [
        "sbpi",
        "sri",
        "scai",
        "gai",
        "acute_stroke_mortality_per_100k"
    ]
].mean().round(2)

	sbpi	sri	scai	gai	acute_stroke_mortality_per_100k
sbpi_class
1	43.60	39.26	45.09	62.52	14.05
2	66.78	64.07	26.23	36.93	14.67
3	78.13	82.78	20.35	35.78	17.23
4	82.08	81.15	12.54	23.66	18.12

SBPI State Comparison

sns.boxplot(
    data=df,
    x="state",
    y="sbpi",
    hue="state",
    palette={
        "NY":"#4C72B0",
        "NJ":"#55A868",
        "CT":"#C44E52"
    },
    legend=False
)

<Axes: xlabel='state', ylabel='sbpi'>

top20 = (
    df.nlargest(20, "sbpi")
      .sort_values("sbpi")
)

plt.figure(figsize=(10,8))

plt.hlines(
    y=top20["county"],
    xmin=0,
    xmax=top20["sbpi"],
    color="lightgray",
    linewidth=2
)

class_colors = {
    4: "#B2182B",
    3: "#EF8A62",
    2: "#FDDBC7",
}

colors = top20["sbpi_class"].map(class_colors)

plt.scatter(
    top20["sbpi"],
    top20["county"],
    s=140,
    c=colors,
    edgecolor="black"
)

plt.xlabel("SBPI Score")
plt.title(
    "Worst 20 Counties by Stroke Burden Priority Index",
    fontsize=16,
    fontweight="bold"
)

plt.grid(axis="x", alpha=.3)

plt.tight_layout()

plt.savefig(
    "../docs/images/sbpi_priority_counties.png",
    dpi=300,
    bbox_inches="tight"
)

plt.show()

The lollipop chart above ranks the twenty counties with the highest Stroke Burden Priority Index scores. Nearly all of the highest-priority counties are located in rural upstate New York, where elevated stroke risk coincides with reduced healthcare resources and longer travel times to stroke centers. Counties such as Hamilton, Delaware, Chenango, and Essex consistently emerge as the highest-priority locations for potential intervention. This ranking provides a practical framework for identifying counties where investments in stroke prevention, healthcare resources, or geographic accessibility may have the greatest impact.