Comparing Clinical Risk Profiles of Adolescent and Adult Epilepsy
Undergraduate Research Team
2026-07-08
epilepsy <- read_csv("/Users/drlakeshialegettejones/Downloads/dataset.csv") %>%
clean_names()
glimpse(epilepsy)## Rows: 1,000
## Columns: 30
## $ age <dbl> 15, 4, 36, 32, 29, 18, 14, 70, 12, 76…
## $ gender <chr> "Male", "Female", "Female", "Female",…
## $ weight <dbl> 48.3, 56.0, 44.8, 70.4, 54.5, 62.1, 5…
## $ height <dbl> 168.2, 174.5, 156.2, 180.4, 161.7, 16…
## $ medication_status <chr> "On Medication", "On Medication", "On…
## $ alcohol_or_drug_use <chr> "Yes", "No", "No", "No", "No", "No", …
## $ eeg_abnormality_detected <chr> "Yes", "Yes", "Yes", "Yes", "Yes", "Y…
## $ mri_ct_scan_result <chr> "Normal", "Abnormal", "Abnormal", "No…
## $ seizure_frequency <dbl> 1, 1, 3, 0, 1, 1, 2, 1, 1, 5, 3, 1, 0…
## $ seizure_duration <dbl> 60, 30, 30, 60, 60, 30, 30, 30, 90, 3…
## $ seizure_type <chr> "Tonic-Clonic", "Generalized", "Tonic…
## $ aura_before_seizure <chr> "No", "Yes", "Yes", "Yes", "No", "Yes…
## $ loss_of_consciousness <chr> "Yes", "Yes", "No", "Yes", "Yes", "Ye…
## $ muscle_stiffness <chr> "No", "Yes", "Yes", "No", "Yes", "No"…
## $ jerky_movements <chr> "Yes", "Yes", "Yes", "Yes", "Yes", "Y…
## $ postictal_confusion <chr> "Yes", "Yes", "No", "Yes", "Yes", "Ye…
## $ blank_stare_episodes <chr> "No", "No", "Yes", "Yes", "No", "Yes"…
## $ eye_rolling <chr> "Yes", "No", "Yes", "No", "No", "No",…
## $ stress_or_anxiety_before_episode <chr> "Yes", "No", "Yes", "Yes", "Yes", "Ye…
## $ lack_of_sleep_before_episode <chr> "Yes", "No", "No", "Yes", "No", "Yes"…
## $ flashing_lights_sensitivity <chr> "Yes", "No", "No", "Yes", "No", "Yes"…
## $ loud_sound_sensitivity <chr> "Yes", "No", "Yes", "Yes", "Yes", "No…
## $ missed_medication <chr> "No", "No", "Yes", "No", "No", "No", …
## $ family_history_of_epilepsy <chr> "No", "Yes", "Yes", "No", "Yes", "Yes…
## $ head_injury_history <chr> "No", "Yes", "No", "No", "No", "No", …
## $ brain_tumor <chr> "No", "Yes", "No", "No", "Yes", "No",…
## $ history_of_stroke <chr> "No", "No", "No", "No", "No", "No", "…
## $ genetic_disorder <chr> "No", "No", "No", "No", "No", "No", "…
## $ developmental_delay_in_children <chr> "Yes", "No", "No", "No", "No", "No", …
## $ target_epilepsy_type <chr> "Generalized", "Generalized", "Genera…epilepsy <- epilepsy %>%
mutate(
age_group = case_when(
age >= 12 & age <= 19 ~ "Adolescent",
age >= 20 ~ "Adult",
age < 12 ~ "Child"
),
age_group = factor(age_group, levels = c("Adolescent", "Adult", "Child"))
)
analysis_data <- epilepsy %>%
filter(age_group %in% c("Adolescent", "Adult")) %>%
mutate(age_group = droplevels(age_group))
analysis_data %>%
count(age_group) %>%
mutate(percent = round(100 * n / sum(n), 1))## # A tibble: 2 × 3
## age_group n percent
## <fct> <int> <dbl>
## 1 Adolescent 92 10.6
## 2 Adult 773 89.44. Exploratory Data Analysis
analysis_data %>%
group_by(age_group) %>%
summarise(
n = n(),
mean_age = mean(age, na.rm = TRUE),
sd_age = sd(age, na.rm = TRUE),
mean_weight = mean(weight, na.rm = TRUE),
sd_weight = sd(weight, na.rm = TRUE),
mean_height = mean(height, na.rm = TRUE),
sd_height = sd(height, na.rm = TRUE)
)## # A tibble: 2 × 8
## age_group n mean_age sd_age mean_weight sd_weight mean_height sd_height
## <fct> <int> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 Adolescent 92 15.2 2.11 64.1 15.3 165. 9.27
## 2 Adult 773 49.7 17.6 66.1 15.3 165. 10.3analysis_data %>%
count(age_group, gender) %>%
group_by(age_group) %>%
mutate(percent = round(100 * n / sum(n), 1))## # A tibble: 6 × 4
## # Groups: age_group [2]
## age_group gender n percent
## <fct> <chr> <int> <dbl>
## 1 Adolescent Female 30 32.6
## 2 Adolescent Male 60 65.2
## 3 Adolescent Other 2 2.2
## 4 Adult Female 332 42.9
## 5 Adult Male 424 54.9
## 6 Adult Other 17 2.2ggplot(analysis_data, aes(x = age, fill = age_group)) +
geom_histogram(bins = 30, alpha = 0.7, position = "identity") +
labs(
title = "Age Distribution by Group",
x = "Age",
y = "Count"
)
ggplot(analysis_data, aes(x = gender, fill = age_group)) +
geom_bar(position = "dodge") +
labs(
title = "Gender Distribution by Age Group",
x = "Gender",
y = "Count"
)
symptom_vars <- c(
"aura_before_seizure",
"loss_of_consciousness",
"muscle_stiffness",
"jerky_movements",
"postictal_confusion",
"blank_stare_episodes",
"eye_rolling"
)
risk_vars <- c(
"alcohol_or_drug_use",
"stress_or_anxiety_before_episode",
"lack_of_sleep_before_episode",
"flashing_lights_sensitivity",
"loud_sound_sensitivity",
"missed_medication",
"family_history_of_epilepsy",
"head_injury_history",
"brain_tumor",
"history_of_stroke",
"genetic_disorder"
)
clinical_vars <- c(
"medication_status",
"eeg_abnormality_detected",
"mri_ct_scan_result",
"seizure_frequency",
"seizure_duration",
"seizure_type",
"target_epilepsy_type"
)analysis_data %>%
select(age_group, all_of(symptom_vars)) %>%
pivot_longer(
cols = all_of(symptom_vars),
names_to = "symptom",
values_to = "response"
) %>%
count(age_group, symptom, response) %>%
group_by(age_group, symptom) %>%
mutate(percent = 100 * n / sum(n)) %>%
filter(response == "Yes") %>%
ggplot(aes(x = symptom, y = percent, fill = age_group)) +
geom_col(position = "dodge") +
coord_flip() +
labs(
title = "Symptom Profiles by Age Group",
x = "Symptom",
y = "Percent Reporting Symptom"
)
analysis_data %>%
select(age_group, all_of(risk_vars)) %>%
pivot_longer(
cols = all_of(risk_vars),
names_to = "risk_factor",
values_to = "response"
) %>%
count(age_group, risk_factor, response) %>%
group_by(age_group, risk_factor) %>%
mutate(percent = 100 * n / sum(n)) %>%
filter(response == "Yes") %>%
ggplot(aes(x = risk_factor, y = percent, fill = age_group)) +
geom_col(position = "dodge") +
coord_flip() +
labs(
title = "Risk Factors by Age Group",
x = "Risk Factor",
y = "Percent Reporting Risk Factor"
)
5. Statistical Analysis
vars_to_test <- c(clinical_vars, symptom_vars, risk_vars)
chi_square_results <- map_df(vars_to_test, function(var) {
test_data <- analysis_data %>%
select(age_group, all_of(var)) %>%
drop_na()
tbl <- table(test_data$age_group, test_data[[var]])
if (nrow(tbl) < 2 || ncol(tbl) < 2) {
return(tibble(
variable = var,
chi_square = NA,
df = NA,
p_value = NA
))
}
test <- suppressWarnings(chisq.test(tbl))
tibble(
variable = var,
chi_square = round(as.numeric(test$statistic), 3),
df = as.numeric(test$parameter),
p_value = round(test$p.value, 4)
)
})
chi_square_results %>%
arrange(p_value)## # A tibble: 25 × 4
## variable chi_square df p_value
## <chr> <dbl> <dbl> <dbl>
## 1 seizure_type 41.1 3 0
## 2 target_epilepsy_type 24.0 5 0.0002
## 3 aura_before_seizure 2.03 1 0.154
## 4 seizure_duration 4.45 3 0.217
## 5 jerky_movements 1.48 1 0.224
## 6 loss_of_consciousness 1.39 1 0.239
## 7 postictal_confusion 1.30 1 0.255
## 8 history_of_stroke 0.863 1 0.353
## 9 mri_ct_scan_result 1.54 2 0.463
## 10 family_history_of_epilepsy 0.438 1 0.508
## # ℹ 15 more rowsmodel_data <- analysis_data %>%
mutate(
adolescent = ifelse(age_group == "Adolescent", 1, 0)
) %>%
select(
adolescent,
gender,
weight,
height,
medication_status,
alcohol_or_drug_use,
eeg_abnormality_detected,
mri_ct_scan_result,
seizure_frequency,
seizure_duration,
seizure_type,
aura_before_seizure,
loss_of_consciousness,
muscle_stiffness,
jerky_movements,
postictal_confusion,
blank_stare_episodes,
eye_rolling,
stress_or_anxiety_before_episode,
lack_of_sleep_before_episode,
missed_medication,
family_history_of_epilepsy,
head_injury_history,
genetic_disorder,
target_epilepsy_type
) %>%
mutate(across(where(is.character), as.factor)) %>%
drop_na()logit_model <- glm(
adolescent ~ .,
data = model_data,
family = binomial
)
logit_results <- tidy(
logit_model,
exponentiate = TRUE,
conf.int = TRUE
) %>%
arrange(p.value)
logit_results## # A tibble: 33 × 7
## term estimate std.error statistic p.value conf.low conf.high
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 seizure_typeFocal 0.121 0.524 -4.03 5.64e-5 0.0425 0.334
## 2 seizure_typeGenerali… 0.230 0.500 -2.94 3.30e-3 0.0854 0.610
## 3 seizure_typeTonic-Cl… 0.359 0.511 -2.00 4.52e-2 0.130 0.973
## 4 genderMale 1.61 0.250 1.91 5.67e-2 0.994 2.65
## 5 seizure_duration 0.992 0.00432 -1.77 7.64e-2 0.984 1.00
## 6 jerky_movementsYes 0.611 0.298 -1.66 9.77e-2 0.340 1.10
## 7 postictal_confusionY… 0.655 0.261 -1.62 1.05e-1 0.394 1.10
## 8 weight 0.989 0.00781 -1.40 1.63e-1 0.974 1.00
## 9 loss_of_consciousnes… 1.41 0.294 1.16 2.44e-1 0.797 2.53
## 10 mri_ct_scan_resultNo… 0.745 0.257 -1.14 2.52e-1 0.449 1.24
## # ℹ 23 more rowsrf_data <- model_data %>%
mutate(
adolescent = factor(
adolescent,
levels = c(0, 1),
labels = c("Adult", "Adolescent")
)
)
set.seed(123)
rf_model <- randomForest(
adolescent ~ .,
data = rf_data,
importance = TRUE,
ntree = 500
)
importance_df <- importance(rf_model) %>%
as.data.frame() %>%
rownames_to_column("variable") %>%
arrange(desc(MeanDecreaseGini))
importance_df## variable Adult Adolescent
## 1 weight 1.21225533 0.41573211
## 2 height -0.29711248 1.09277451
## 3 seizure_type 3.39799091 8.69402249
## 4 seizure_frequency -0.76198080 -2.32756338
## 5 target_epilepsy_type 1.24128761 1.44187408
## 6 seizure_duration -1.20245006 -0.80013987
## 7 mri_ct_scan_result -0.68716582 -0.34488143
## 8 eye_rolling 1.18580219 4.09113751
## 9 gender -1.41888108 -0.25370992
## 10 muscle_stiffness 2.50258063 -1.03090560
## 11 jerky_movements 1.82391510 3.01553318
## 12 lack_of_sleep_before_episode -0.38653867 -0.77670047
## 13 aura_before_seizure 0.08913831 -0.62671879
## 14 postictal_confusion 1.92697174 -0.05460891
## 15 family_history_of_epilepsy -1.99128271 1.50908312
## 16 stress_or_anxiety_before_episode -2.98509348 1.45899237
## 17 medication_status -1.69400793 -0.13358916
## 18 head_injury_history 0.06250515 0.82008398
## 19 blank_stare_episodes -3.25933420 0.22925981
## 20 loss_of_consciousness 0.51963175 0.20733471
## 21 missed_medication 0.78761861 -1.31314838
## 22 eeg_abnormality_detected -0.90039778 -1.10666739
## 23 genetic_disorder -2.89393562 2.55601013
## 24 alcohol_or_drug_use -1.37169823 -2.12010005
## MeanDecreaseAccuracy MeanDecreaseGini
## 1 1.22935484 24.996569
## 2 0.04969975 23.477850
## 3 6.30307802 11.374333
## 4 -1.50160457 10.863905
## 5 1.69840313 9.637369
## 6 -1.41328297 8.079422
## 7 -0.76666042 7.247477
## 8 2.55016127 4.766428
## 9 -1.50951374 4.584075
## 10 1.88172618 4.310033
## 11 2.67638799 4.174650
## 12 -0.59280314 4.133665
## 13 -0.08336847 4.095214
## 14 1.77196560 4.013726
## 15 -1.29984603 3.994467
## 16 -2.19681833 3.921945
## 17 -1.63109118 3.540770
## 18 0.36480064 3.534973
## 19 -3.03923438 3.493657
## 20 0.59903114 3.453896
## 21 0.34830511 3.416778
## 22 -1.24630948 2.762019
## 23 -1.88584648 2.569827
## 24 -2.11521592 2.108855varImpPlot(rf_model)
6. Topological Characterization of Clinical Risk Profiles
tda_data <- analysis_data %>%
select(
age_group,
gender,
medication_status,
alcohol_or_drug_use,
eeg_abnormality_detected,
mri_ct_scan_result,
seizure_frequency,
seizure_duration,
seizure_type,
aura_before_seizure,
loss_of_consciousness,
muscle_stiffness,
jerky_movements,
postictal_confusion,
blank_stare_episodes,
eye_rolling,
stress_or_anxiety_before_episode,
lack_of_sleep_before_episode,
missed_medication,
family_history_of_epilepsy,
head_injury_history,
genetic_disorder,
target_epilepsy_type
) %>%
drop_na()
age_labels <- tda_data$age_group
risk_profiles <- tda_data %>%
select(-age_group) %>%
mutate(across(where(is.character), as.factor))
str(risk_profiles)## tibble [865 × 22] (S3: tbl_df/tbl/data.frame)
## $ gender : Factor w/ 3 levels "Female","Male",..: 2 1 1 2 2 2 1 2 1 2 ...
## $ medication_status : Factor w/ 2 levels "Not on Medication",..: 2 2 1 2 2 1 2 1 2 2 ...
## $ alcohol_or_drug_use : Factor w/ 2 levels "No","Yes": 2 1 1 1 1 1 1 1 1 1 ...
## $ eeg_abnormality_detected : Factor w/ 2 levels "No","Yes": 2 2 2 2 2 1 2 1 2 2 ...
## $ mri_ct_scan_result : Factor w/ 3 levels "Abnormal","Normal",..: 2 1 3 3 1 3 2 2 1 2 ...
## $ seizure_frequency : num [1:865] 1 3 0 1 1 2 1 1 5 3 ...
## $ seizure_duration : num [1:865] 60 30 60 60 30 30 30 90 30 30 ...
## $ seizure_type : Factor w/ 4 levels "Absence","Focal",..: 4 4 2 4 4 1 1 4 4 3 ...
## $ aura_before_seizure : Factor w/ 2 levels "No","Yes": 1 2 2 1 2 1 2 2 1 2 ...
## $ loss_of_consciousness : Factor w/ 2 levels "No","Yes": 2 1 2 2 2 1 2 2 1 2 ...
## $ muscle_stiffness : Factor w/ 2 levels "No","Yes": 1 2 1 2 1 2 1 1 1 1 ...
## $ jerky_movements : Factor w/ 2 levels "No","Yes": 2 2 2 2 2 1 1 1 2 2 ...
## $ postictal_confusion : Factor w/ 2 levels "No","Yes": 2 1 2 2 2 2 1 1 2 1 ...
## $ blank_stare_episodes : Factor w/ 2 levels "No","Yes": 1 2 2 1 2 2 2 1 1 1 ...
## $ eye_rolling : Factor w/ 2 levels "No","Yes": 2 2 1 1 1 2 2 2 2 2 ...
## $ stress_or_anxiety_before_episode: Factor w/ 2 levels "No","Yes": 2 2 2 2 2 1 1 2 2 2 ...
## $ lack_of_sleep_before_episode : Factor w/ 2 levels "No","Yes": 2 1 2 1 2 1 2 2 2 2 ...
## $ missed_medication : Factor w/ 2 levels "No","Yes": 1 2 1 1 1 1 1 1 1 1 ...
## $ family_history_of_epilepsy : Factor w/ 2 levels "No","Yes": 1 2 1 2 2 2 1 2 2 1 ...
## $ head_injury_history : Factor w/ 2 levels "No","Yes": 1 1 1 1 1 2 1 1 1 2 ...
## $ genetic_disorder : Factor w/ 2 levels "No","Yes": 1 1 1 1 1 2 1 1 1 1 ...
## $ target_epilepsy_type : Factor w/ 6 levels "Absence","Complicated",..: 4 4 3 4 4 1 1 2 2 4 ...gower_dist <- daisy(risk_profiles, metric = "gower")
gower_matrix <- as.matrix(gower_dist)mds <- cmdscale(gower_dist, k = 2)
mds_df <- data.frame(
Dim1 = mds[, 1],
Dim2 = mds[, 2],
age_group = age_labels
)
ggplot(mds_df, aes(x = Dim1, y = Dim2, color = age_group)) +
geom_point(alpha = 0.7) +
labs(
title = "Risk Profile Space Using Gower Distance",
x = "Dimension 1",
y = "Dimension 2"
)
filter_values <- mds[, 1]
mapper_result <- mapper1D(
distance_matrix = gower_matrix,
filter_values = filter_values,
num_intervals = 10,
percent_overlap = 50,
num_bins_when_clustering = 10
)
mapper_vertices <- mapper_result$points_in_vertex
mapper_edges <- mapper_result$adjacency
g <- graph_from_adjacency_matrix(
mapper_edges,
mode = "undirected"
)
node_adolescent_prop <- sapply(mapper_vertices, function(indices) {
mean(age_labels[indices] == "Adolescent")
})
V(g)$adolescent_prop <- node_adolescent_propplot(
g,
vertex.size = 8 + 25 * V(g)$adolescent_prop,
vertex.label = NA,
main = "Mapper Graph of Epilepsy Risk Profiles"
)
mapper_summary <- tibble(
node = seq_along(mapper_vertices),
node_size = sapply(mapper_vertices, length),
adolescent_proportion = node_adolescent_prop
) %>%
arrange(desc(adolescent_proportion))
mapper_summary## # A tibble: 10 × 3
## node node_size adolescent_proportion
## <int> <int> <dbl>
## 1 4 176 0.148
## 2 9 250 0.132
## 3 10 166 0.127
## 4 8 201 0.119
## 5 5 203 0.0985
## 6 3 164 0.0976
## 7 7 160 0.0938
## 8 6 192 0.0677
## 9 1 45 0.0667
## 10 2 109 0.0642Interpretation Notes
The exploratory analysis describes differences between adolescents and adults in the dataset.
The statistical analysis identifies individual clinical, demographic, and behavioral variables that distinguish adolescent epilepsy patients from adult epilepsy patients.
The TDA analysis treats each patient as a multivariate clinical risk profile. Rather than studying variables one at a time, Mapper studies the shape of the risk-profile space and identifies clusters of patients with similar combinations of symptoms and risk factors.