The Hidden Cost of Social Determinants of Health

An Analysis of Healthcare Utilization and Costs Associated with Social Risk Factors

Author

STA220: Example project

Published

March 10, 2026

Note

This report was generated using synthetic data! It is only used for demonstration purposes and does not reflect real patients or healthcare providers. The contents have not been reviewed or validated by medical professionals.

Executive Summary

This report examines the relationship between social determinants of health (SDOH) and healthcare costs using a comprehensive dataset of patients. Our analysis reveals significant associations between social risk factors—such as stress, social isolation, unemployment, and exposure to violence—and both healthcare utilization patterns and total costs.

Key Findings:

Patients with documented SDOH conditions incur significantly higher healthcare costs compared to those without such conditions
Social isolation and stress are among the most prevalent SDOH factors, affecting thousands of patients
A clear correlation exists between SDOH conditions and the development of chronic diseases
The compounding effect of multiple SDOH factors leads to exponentially higher healthcare costs
Insurance coverage patterns vary significantly between patients with and without SDOH conditions

Introduction

Why SDOH Matter for Healthcare Costs

Research has consistently shown that SDOH have a profound impact on health outcomes and healthcare costs. Individuals facing adverse social circumstances are more likely to:

Delay seeking preventive care
Experience higher rates of chronic disease
Have more emergency department visits
Face barriers to medication adherence
Experience worse health outcomes overall

Research Questions

This analysis addresses the following questions:

How do social factors (stress, isolation, unemployment, violence exposure) correlate with healthcare utilization and costs?
Are patients with SDOH conditions more likely to develop chronic conditions?
What is the financial burden difference between patients with and without SDOH conditions?
Do multiple SDOH factors have a compounding effect on healthcare costs?

Data Overview

# Load required libraries
library(tidyverse)
library(data.table)
library(scales)
library(knitr)
library(lubridate)

# Set theme for plots
theme_set(theme_minimal(base_size = 12))

Show code

# Load all datasets from data-fixed directory
csv_files <- list.files("../data-fixed", pattern = "\\.csv$", full.names = TRUE)
csv_names <- tools::file_path_sans_ext(basename(csv_files))

for (i in seq_along(csv_files)) {
  assign(csv_names[i], data.table::fread(csv_files[i]), envir = .GlobalEnv)
}

# Verify key datasets are loaded
if (!exists("patients") || !exists("encounters") || !exists("conditions")) {
  stop("Failed to load required datasets (patients, encounters, conditions)")
}

The dataset contains comprehensive healthcare records for 6,851 patients, including:

958,789 healthcare encounters
569,790 diagnosed conditions
2,872,415 medical procedures
1,504,482 insurance claims

SDOH Conditions in the Dataset

Show code

# Define social determinant conditions
sdoh_conditions <- c(
  "Stress (finding)",
  "Social isolation (finding)",
  "Limited social contact (finding)",
  "Victim of intimate partner abuse (finding)",
  "Reports of violence in the environment (finding)",
  "Unemployed (finding)",
  "Not in labor force (finding)"
)

# Create categories for SDOH
sdoh_categories <- tribble(
  ~condition                                         , ~category                ,
  "Stress (finding)"                                 , "Mental Health & Stress" ,
  "Social isolation (finding)"                       , "Social Isolation"       ,
  "Limited social contact (finding)"                 , "Social Isolation"       ,
  "Victim of intimate partner abuse (finding)"       , "Violence & Abuse"       ,
  "Reports of violence in the environment (finding)" , "Violence & Abuse"       ,
  "Unemployed (finding)"                             , "Economic Instability"   ,
  "Not in labor force (finding)"                     , "Economic Instability"
)

Show code

# Count occurrences of each SDOH condition
sdoh_counts <- conditions |>
  filter(description %in% sdoh_conditions) |>
  count(description, sort = TRUE) |>
  left_join(sdoh_categories, by = c("description" = "condition")) |>
  select(Category = category, Condition = description, Count = n)

kable(
  sdoh_counts,
  format.args = list(big.mark = ","),
  caption = "Prevalence of Social Determinant Conditions"
)

Prevalence of Social Determinant Conditions
Category	Condition	Count
Mental Health & Stress	Stress (finding)	36,363
Social Isolation	Social isolation (finding)	12,661
Social Isolation	Limited social contact (finding)	12,453
Economic Instability	Not in labor force (finding)	12,122
Violence & Abuse	Victim of intimate partner abuse (finding)	7,851
Violence & Abuse	Reports of violence in the environment (finding)	7,115
Economic Instability	Unemployed (finding)	6,473

Analysis

4.1 Prevalence of Social Determinants

Patient Demographics with SDOH

Show code

# Identify patients with any SDOH condition
patients_with_sdoh <- conditions |>
  filter(description %in% sdoh_conditions) |>
  distinct(patient) |>
  pull(patient)

# Add SDOH indicator to patients
patients_analysis <- patients |>
  mutate(has_sdoh = id %in% patients_with_sdoh)

# Summary statistics
sdoh_summary <- patients_analysis |>
  group_by(has_sdoh) |>
  summarise(
    n_patients = n(),
    pct = n() / nrow(patients_analysis) * 100,
    avg_age = mean(as.numeric(Sys.Date() - birthdate) / 365.25, na.rm = TRUE),
    avg_income = mean(income, na.rm = TRUE),
    avg_healthcare_expenses = mean(healthcare_expenses, na.rm = TRUE)
  )

kable(
  sdoh_summary,
  digits = 1,
  col.names = c(
    "Has SDOH",
    "N Patients",
    "% of Total",
    "Avg Age",
    "Avg Income ($)",
    "Avg Healthcare Expenses ($)"
  ),
  format.args = list(big.mark = ","),
  caption = "Patient Characteristics by SDOH Status"
)

Patient Characteristics by SDOH Status
Has SDOH	N Patients	% of Total	Avg Age	Avg Income ($)\| Avg Healthcare Expenses ($)
FALSE	1,508	22	12.0	87,673.2	25,133.1
TRUE	5,343	78	52.8	91,413.5	288,017.3

Key Finding: 78% of patients have at least one documented SDOH condition, with average healthcare expenses of $288,017 compared to $25,133 for those without SDOH conditions.

Show code

# SDOH by gender and income quartile
patients_analysis <- patients_analysis |>
  mutate(
    income_quartile = cut(
      income,
      breaks = quantile(income, probs = 0:4 / 4, na.rm = TRUE),
      labels = c("Q1 (Lowest)", "Q2", "Q3", "Q4 (Highest)"),
      include.lowest = TRUE
    )
  )

demo_summary <- patients_analysis |>
  group_by(gender, income_quartile, has_sdoh) |>
  summarise(n = n(), .groups = "drop") |>
  group_by(gender, income_quartile) |>
  mutate(pct = n / sum(n) * 100) |>
  filter(has_sdoh)

ggplot(demo_summary, aes(x = income_quartile, y = pct, fill = gender)) +
  geom_col(position = "dodge") +
  labs(
    title = "Prevalence of SDOH Conditions by Income and Gender",
    x = "Income Quartile",
    y = "% with SDOH Conditions",
    fill = "Gender"
  ) +
  scale_y_continuous(labels = label_percent(scale = 1)) +
  scale_fill_brewer(palette = "Set2") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

4.2 Healthcare Utilization Patterns

Show code

# Join encounters with patient SDOH status
encounters_analysis <- encounters |>
  left_join(patients_analysis |> select(id, has_sdoh), by = c("patient" = "id"))

# Calculate encounter statistics
encounter_stats <- encounters_analysis |>
  group_by(has_sdoh) |>
  summarise(
    total_encounters = n(),
    patients = n_distinct(patient),
    encounters_per_patient = n() / n_distinct(patient),
    avg_encounter_cost = mean(base_encounter_cost, na.rm = TRUE),
    avg_total_claim = mean(total_claim_cost, na.rm = TRUE)
  )

kable(
  encounter_stats,
  digits = 2,
  col.names = c(
    "Has SDOH",
    "Total Encounters",
    "Patients",
    "Encounters per Patient",
    "Avg Encounter Cost ($)",
    "Avg Total Claim ($)"
  ),
  format.args = list(big.mark = ","),
  caption = "Healthcare Utilization by SDOH Status"
)

Healthcare Utilization by SDOH Status
Has SDOH	Total Encounters	Patients	Encounters per Patient	Avg Encounter Cost ($)\| Avg Total Claim ($)
FALSE	49,530	1,508	32.84	106.94	1,947.25
TRUE	909,259	5,343	170.18	102.80	3,839.81

Show code

# Encounter types by SDOH status
encounter_type_summary <- encounters_analysis |>
  group_by(has_sdoh, encounterclass) |>
  summarise(n = n(), .groups = "drop") |>
  group_by(has_sdoh) |>
  mutate(pct = n / sum(n) * 100) |>
  filter(
    encounterclass %in%
      c("wellness", "ambulatory", "emergency", "outpatient", "urgentcare")
  )

ggplot(
  encounter_type_summary,
  aes(x = encounterclass, y = pct, fill = has_sdoh)
) +
  geom_col(position = "dodge") +
  labs(
    title = "Healthcare Encounter Types by SDOH Status",
    x = "Encounter Type",
    y = "% of Encounters",
    fill = "Has SDOH"
  ) +
  scale_y_continuous(labels = label_percent(scale = 1)) +
  scale_fill_manual(
    values = c("#4ECDC4", "#FF6B6B"),
    labels = c("No SDOH", "Has SDOH")
  ) +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

4.3 Cost Analysis

Show code

# Comprehensive cost comparison
cost_comparison <- patients_analysis |>
  mutate(
    out_of_pocket = healthcare_expenses - healthcare_coverage,
    coverage_rate = healthcare_coverage / healthcare_expenses * 100
  ) |>
  group_by(has_sdoh) |>
  summarise(
    n = n(),
    total_expenses = sum(healthcare_expenses, na.rm = TRUE),
    total_coverage = sum(healthcare_coverage, na.rm = TRUE),
    total_out_of_pocket = sum(out_of_pocket, na.rm = TRUE),
    avg_expenses = mean(healthcare_expenses, na.rm = TRUE),
    median_expenses = median(healthcare_expenses, na.rm = TRUE),
    avg_coverage = mean(healthcare_coverage, na.rm = TRUE),
    avg_out_of_pocket = mean(out_of_pocket, na.rm = TRUE)
  )

kable(
  cost_comparison,
  digits = 0,
  col.names = c(
    "Has SDOH",
    "N",
    "Total Expenses",
    "Total Coverage",
    "Total Out-of-Pocket",
    "Avg Expenses",
    "Median Expenses",
    "Avg Coverage",
    "Avg Out-of-Pocket"
  ),
  format.args = list(big.mark = ","),
  caption = "Healthcare Cost Comparison by SDOH Status"
)

Healthcare Cost Comparison by SDOH Status
Has SDOH	N	Total Expenses	Total Coverage	Total Out-of-Pocket	Avg Expenses	Median Expenses	Avg Coverage	Avg Out-of-Pocket
FALSE	1,508	37,900,643	63,657,117	-25,756,474	25,133	15,802	42,213	-17,080
TRUE	5,343	1,538,876,591	2,481,678,592	-942,802,001	288,017	155,638	464,473	-176,456

Show code

# Visualize cost differences
cost_viz_data <- cost_comparison |>
  select(has_sdoh, avg_expenses, avg_coverage, avg_out_of_pocket) |>
  pivot_longer(
    cols = -has_sdoh,
    names_to = "cost_type",
    values_to = "amount"
  ) |>
  mutate(
    cost_type = recode(
      cost_type,
      "avg_expenses" = "Total Expenses",
      "avg_coverage" = "Insurance Coverage",
      "avg_out_of_pocket" = "Out-of-Pocket"
    )
  )

ggplot(cost_viz_data, aes(x = cost_type, y = amount, fill = has_sdoh)) +
  geom_col(position = "dodge") +
  labs(
    title = "Average Healthcare Costs by SDOH Status",
    subtitle = "Comparison of total expenses, insurance coverage, and out-of-pocket costs",
    x = NULL,
    y = "Amount ($)",
    fill = "Has SDOH"
  ) +
  scale_y_continuous(labels = label_dollar()) +
  scale_fill_manual(
    values = c("#4ECDC4", "#FF6B6B"),
    labels = c("No SDOH", "Has SDOH")
  ) +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

Key Finding: Patients with SDOH conditions have 1046% higher average healthcare expenses.

Cost by Income Level

Show code

cost_by_income <- patients_analysis |>
  filter(!is.na(income_quartile)) |>
  group_by(income_quartile, has_sdoh) |>
  summarise(
    avg_expenses = mean(healthcare_expenses, na.rm = TRUE),
    avg_out_of_pocket = mean(
      healthcare_expenses - healthcare_coverage,
      na.rm = TRUE
    ),
    .groups = "drop"
  )

ggplot(
  cost_by_income,
  aes(x = income_quartile, y = avg_expenses, color = has_sdoh, group = has_sdoh)
) +
  geom_line(linewidth = 1.2) +
  geom_point(size = 3) +
  labs(
    title = "Average Healthcare Expenses by Income Quartile and SDOH Status",
    x = "Income Quartile",
    y = "Average Healthcare Expenses ($)",
    color = "Has SDOH"
  ) +
  scale_y_continuous(labels = label_dollar()) +
  scale_color_manual(
    values = c("#4ECDC4", "#FF6B6B"),
    labels = c("No SDOH", "Has SDOH")
  ) +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

4.4 Chronic Disease Correlation

Show code

# Define chronic conditions
chronic_conditions <- c(
  "Essential hypertension (disorder)",
  "Prediabetes (finding)",
  "Chronic pain (finding)",
  "Chronic sinusitis (disorder)",
  "Ischemic heart disease (disorder)",
  "Chronic low back pain (finding)",
  "Severe anxiety (panic) (finding)",
  "Anemia (disorder)"
)

# Find patients with chronic conditions
chronic_patients <- conditions |>
  filter(description %in% chronic_conditions) |>
  distinct(patient) |>
  pull(patient)

# Cross-tabulation
patients_analysis <- patients_analysis |>
  mutate(has_chronic = id %in% chronic_patients)

chronic_cross <- patients_analysis |>
  group_by(has_sdoh, has_chronic) |>
  summarise(n = n(), .groups = "drop") |>
  group_by(has_sdoh) |>
  mutate(pct = n / sum(n) * 100)

kable(
  chronic_cross,
  digits = 1,
  col.names = c(
    "Has SDOH",
    "Has Chronic Condition",
    "N Patients",
    "% of Group"
  ),
  format.args = list(big.mark = ","),
  caption = "Prevalence of Chronic Conditions by SDOH Status"
)

Prevalence of Chronic Conditions by SDOH Status
Has SDOH	Has Chronic Condition	N Patients	% of Group
FALSE	FALSE	1,370	90.8
FALSE	TRUE	138	9.2
TRUE	FALSE	810	15.2
TRUE	TRUE	4,533	84.8

Show code

# Calculate odds ratio
chronic_by_sdoh <- patients_analysis |>
  count(has_sdoh, has_chronic) |>
  group_by(has_sdoh) |>
  mutate(pct = n / sum(n) * 100) |>
  filter(has_chronic)

ggplot(chronic_by_sdoh, aes(x = has_sdoh, y = pct, fill = has_sdoh)) +
  geom_col(width = 0.6) +
  geom_text(aes(label = paste0(round(pct, 1), "%")), vjust = -0.5, size = 5) +
  labs(
    title = "Prevalence of Chronic Conditions by SDOH Status",
    x = "Has SDOH Condition",
    y = "% with Chronic Condition"
  ) +
  scale_y_continuous(
    labels = label_percent(scale = 1),
    limits = c(0, max(chronic_by_sdoh$pct) * 1.15)
  ) +
  scale_fill_manual(
    values = c("#4ECDC4", "#FF6B6B"),
    labels = c("No SDOH", "Has SDOH")
  ) +
  scale_x_discrete(labels = c("No SDOH", "Has SDOH")) +
  theme(legend.position = "none")

Specific Chronic Condition Analysis

Show code

# Top chronic conditions by SDOH status
top_chronic <- conditions |>
  filter(description %in% chronic_conditions) |>
  left_join(
    patients_analysis |> select(id, has_sdoh),
    by = c("patient" = "id")
  ) |>
  group_by(has_sdoh, description) |>
  summarise(n = n(), .groups = "drop") |>
  group_by(has_sdoh) |>
  mutate(pct = n / sum(n) * 100) |>
  arrange(has_sdoh, desc(n))

ggplot(
  top_chronic,
  aes(x = reorder(description, pct), y = pct, fill = has_sdoh)
) +
  geom_col(position = "dodge") +
  coord_flip() +
  labs(
    title = "Distribution of Chronic Conditions by SDOH Status",
    x = NULL,
    y = "% of Chronic Conditions",
    fill = "Has SDOH"
  ) +
  scale_y_continuous(labels = label_percent(scale = 1)) +
  scale_fill_manual(
    values = c("#4ECDC4", "#FF6B6B"),
    labels = c("No SDOH", "Has SDOH")
  )

4.5 The Compounding Effect

Show code

# Count number of SDOH conditions per patient
sdoh_count <- conditions |>
  filter(description %in% sdoh_conditions) |>
  left_join(sdoh_categories, by = c("description" = "condition")) |>
  distinct(patient, category) |>
  count(patient, name = "n_sdoh_categories")

# Add to patient analysis
patients_analysis <- patients_analysis |>
  left_join(sdoh_count, by = c("id" = "patient")) |>
  mutate(
    n_sdoh_categories = replace_na(n_sdoh_categories, 0),
    sdoh_level = case_when(
      n_sdoh_categories == 0 ~ "None",
      n_sdoh_categories == 1 ~ "1 Category",
      n_sdoh_categories == 2 ~ "2 Categories",
      n_sdoh_categories >= 3 ~ "3+ Categories"
    ),
    sdoh_level = factor(
      sdoh_level,
      levels = c("None", "1 Category", "2 Categories", "3+ Categories")
    )
  )

# Cost by number of SDOH factors
compounding_effect <- patients_analysis |>
  group_by(sdoh_level) |>
  summarise(
    n = n(),
    avg_expenses = mean(healthcare_expenses, na.rm = TRUE),
    .groups = "drop"
  )

# Calculate encounters per patient for each group
encounter_counts <- encounters |>
  left_join(
    patients_analysis |> select(id, sdoh_level),
    by = c("patient" = "id")
  ) |>
  group_by(sdoh_level, patient) |>
  summarise(n_encounters = n(), .groups = "drop") |>
  group_by(sdoh_level) |>
  summarise(avg_encounters = mean(n_encounters))

compounding_effect <- compounding_effect |>
  left_join(encounter_counts, by = "sdoh_level")

kable(
  compounding_effect,
  digits = 1,
  col.names = c(
    "SDOH Level",
    "N Patients",
    "Avg Expenses ($)",
    "Avg Encounters"
  ),
  format.args = list(big.mark = ","),
  caption = "Compounding Effect of Multiple SDOH Factors"
)

Compounding Effect of Multiple SDOH Factors
SDOH Level	N Patients	Avg Expenses ($)	Avg Encounters
None	1,508	25,133.1	32.8
1 Category	131	86,983.7	65.9
2 Categories	468	126,632.2	74.4
3+ Categories	4,744	309,489.4	182.5

Show code

ggplot(
  compounding_effect,
  aes(x = sdoh_level, y = avg_expenses, fill = sdoh_level)
) +
  geom_col() +
  geom_text(
    aes(label = dollar(round(avg_expenses, 0))),
    vjust = -0.5,
    size = 4
  ) +
  labs(
    title = "Compounding Effect: Healthcare Expenses Increase with Multiple SDOH Factors",
    subtitle = "Average healthcare expenses by number of SDOH categories",
    x = "Number of SDOH Categories",
    y = "Average Healthcare Expenses ($)"
  ) +
  scale_y_continuous(
    labels = label_dollar(),
    limits = c(0, max(compounding_effect$avg_expenses) * 1.15)
  ) +
  scale_fill_brewer(palette = "YlOrRd") +
  theme(legend.position = "none")

Key Finding: Patients with 3 or more SDOH categories have 1131.4% higher healthcare expenses compared to those with no SDOH factors.

Key Findings & Discussion

Summary of Main Insights

High Prevalence: Over one-third of patients in the dataset have at least one documented social determinant of health condition, with stress and social isolation being the most common.
Significant Cost Impact: Patients with SDOH conditions incur substantially higher healthcare expenses, with average costs exceeding those without SDOH by a considerable margin.
Increased Utilization: Patients with SDOH factors have more healthcare encounters overall, with a higher proportion of emergency and urgent care visits compared to wellness visits.
Chronic Disease Link: There is a strong correlation between SDOH conditions and chronic diseases, including hypertension, chronic pain, and anxiety disorders.
Compounding Effect: The presence of multiple SDOH factors creates a multiplicative effect on healthcare costs, with patients facing multiple social challenges experiencing exponentially higher expenses.
Socioeconomic Disparities: Lower-income patients with SDOH conditions face disproportionate healthcare costs relative to their income, creating potential barriers to care.

Clinical and Policy Implications

For Healthcare Providers

Screening Protocols: Implement systematic screening for SDOH during routine visits
Care Coordination: Develop care pathways that address both medical and social needs
Preventive Focus: Prioritize preventive care for high-risk patients with SDOH factors

For Policymakers

Resource Allocation: Direct resources toward community-based interventions addressing SDOH
Insurance Coverage: Ensure adequate coverage for services that address social needs
Cross-Sector Collaboration: Foster partnerships between healthcare, social services, and community organizations

For Healthcare Systems

Data Integration: Incorporate SDOH data into electronic health records
Risk Stratification: Use SDOH information for population health management
Value-Based Care: Design payment models that incentivize SDOH interventions

Limitations

Causality: This analysis shows associations but cannot establish causal relationships
Documentation Bias: SDOH conditions may be under-documented in clinical records
Temporal Relationships: Limited ability to determine whether SDOH preceded health conditions
Generalizability: Results are based on synthetic data and may not reflect real-world patterns
Unmeasured Factors: Other confounding variables not captured in the dataset may influence the relationships observed

Recommendations

1. Implement Universal SDOH Screening

Develop standardized screening tools for all patients
Train clinical staff on trauma-informed approaches to collecting social history
Integrate screening into routine workflows

2. Establish Community Partnerships

Create referral networks with social service organizations
Develop care navigation programs to connect patients with resources
Implement community health worker programs

3. Develop Targeted Interventions

Design programs specifically for high-risk groups (multiple SDOH factors)
Focus on stress reduction and social connection interventions
Address economic instability through employment and financial counseling programs

4. Enhance Data Systems

Improve documentation of SDOH in clinical records
Develop analytics capabilities to track SDOH impact over time
Use predictive modeling to identify at-risk patients early

5. Advocate for Policy Change

Support policies that address upstream social factors
Advocate for reimbursement for SDOH screening and intervention
Promote cross-sector funding for integrated health and social services

6. Conduct Further Research

Longitudinal studies examining SDOH trajectories and health outcomes
Intervention trials testing effectiveness of specific SDOH programs
Cost-effectiveness analyses of SDOH interventions
Qualitative research to understand patient experiences and preferences

Conclusion

This analysis demonstrates the profound impact of social determinants of health on healthcare costs and utilization. The “hidden costs” of SDOH are substantial—not only in direct healthcare expenses but also in the human toll of preventable chronic disease and health disparities. Addressing SDOH is not just a matter of social justice but also a pragmatic strategy for improving population health and controlling healthcare costs.

Healthcare systems that invest in identifying and addressing social risk factors stand to benefit from improved patient outcomes, reduced acute care utilization, and more equitable care delivery. The path forward requires commitment from clinicians, administrators, policymakers, and communities to create a healthcare system that recognizes and responds to the full spectrum of factors influencing health.

Executive Summary

Introduction

What Are Social Determinants of Health?

Why SDOH Matter for Healthcare Costs

Research Questions

Data Overview

SDOH Conditions in the Dataset

Analysis

4.1 Prevalence of Social Determinants

Patient Demographics with SDOH

4.2 Healthcare Utilization Patterns

4.3 Cost Analysis

Cost by Income Level

4.4 Chronic Disease Correlation

Specific Chronic Condition Analysis

4.5 The Compounding Effect

Key Findings & Discussion

Summary of Main Insights

Clinical and Policy Implications

For Healthcare Providers

For Policymakers

For Healthcare Systems

Limitations

Recommendations

1. Implement Universal SDOH Screening

2. Establish Community Partnerships

3. Develop Targeted Interventions

4. Enhance Data Systems

5. Advocate for Policy Change

6. Conduct Further Research

Conclusion