diff --git a/src/patientflow/train/classifiers.py b/src/patientflow/train/classifiers.py
index 2d0120d..ef50387 100644
--- a/src/patientflow/train/classifiers.py
+++ b/src/patientflow/train/classifiers.py
@@ -60,6 +60,23 @@
 )
 
 
+def _is_string_like_column(series: Series) -> bool:
+    """Check whether a Series holds string-like values that need encoding.
+
+    Uses an exclusion-based approach so that any current or future
+    string-like dtype (``object``, ``StringDtype``, ``ArrowDtype("string")``,
+    ``CategoricalDtype`` with string categories, etc.) is detected without
+    needing to enumerate each one explicitly.
+    """
+    if isinstance(series.dtype, pd.CategoricalDtype):
+        series = series.cat.categories.to_series()
+    return not (
+        pd.api.types.is_numeric_dtype(series)
+        or pd.api.types.is_bool_dtype(series)
+        or pd.api.types.is_datetime64_any_dtype(series)
+    )
+
+
 class FeatureColumnTransformer(BaseEstimator, TransformerMixin):
     """
     Ensure that input data has exactly the columns seen during training.
@@ -118,7 +135,7 @@ def fit(
                 default = 0.0
             elif pd.api.types.is_datetime64_any_dtype(series):
                 default = pd.NaT
-            elif series.dtype == "object":
+            elif _is_string_like_column(series):
                 mode = series.mode(dropna=True)
                 default = mode.iloc[0] if not mode.empty else "Unknown"
             else:
@@ -328,8 +345,7 @@ def create_column_transformer(
         # Keep boolean columns as single 0/1 features rather than one-hot encoding
         elif df[col].dtype == "bool":
             transformers.append((col, "passthrough", [col]))
-        # One-hot encode string/object categoricals
-        elif df[col].dtype == "object":
+        elif _is_string_like_column(df[col]):
             transformers.append((col, OneHotEncoder(handle_unknown="ignore"), [col]))
         # Keep non-boolean binary (0/1) columns as single features
         elif df[col].nunique() == 2:
diff --git a/tests/test_classifiers.py b/tests/test_classifiers.py
index 881bb50..36083f6 100644
--- a/tests/test_classifiers.py
+++ b/tests/test_classifiers.py
@@ -214,6 +214,42 @@ def test_invalid_parameters(self):
                 single_snapshot_per_visit=True,
             )
 
+    def test_string_dtype_columns(self):
+        """Test training when categorical columns use pandas StringDtype."""
+        visits = self.train_visits.copy()
+        visits["sex"] = visits["sex"].astype("string")
+        visits["arrival_method"] = visits["arrival_method"].astype("string")
+
+        model = train_classifier(
+            train_visits=visits,
+            valid_visits=visits.copy(),
+            prediction_time=self.prediction_time,
+            exclude_from_training_data=self.exclude_from_training_data,
+            grid=self.grid,
+            ordinal_mappings=self.ordinal_mappings,
+            visit_col="visit_number",
+        )
+        self.assertIsInstance(model, TrainedClassifier)
+        self.assertIsNotNone(model.pipeline)
+
+    def test_categorical_dtype_columns(self):
+        """Test training when categorical columns use CategoricalDtype."""
+        visits = self.train_visits.copy()
+        visits["sex"] = visits["sex"].astype("category")
+        visits["arrival_method"] = visits["arrival_method"].astype("category")
+
+        model = train_classifier(
+            train_visits=visits,
+            valid_visits=visits.copy(),
+            prediction_time=self.prediction_time,
+            exclude_from_training_data=self.exclude_from_training_data,
+            grid=self.grid,
+            ordinal_mappings=self.ordinal_mappings,
+            visit_col="visit_number",
+        )
+        self.assertIsInstance(model, TrainedClassifier)
+        self.assertIsNotNone(model.pipeline)
+
     def test_feature_importance(self):
         """Test that feature importance is captured when available."""
         model = train_classifier(