diff --git a/pyhealth/data/__init__.py b/pyhealth/data/__init__.py
index 1cf916bfe..ed8bf8d87 100755
--- a/pyhealth/data/__init__.py
+++ b/pyhealth/data/__init__.py
@@ -5,4 +5,4 @@ class Visit:
     """This class is deprecated and should not be used."""
     def __init__(self, *args, **kwargs):
         import warnings
-        warnings.warn("The Visit class is deprecated and will be removed in a future version.", DeprecationWarning)
\ No newline at end of file
+        warnings.warn("The Visit class is deprecated and will be removed in a future version.", DeprecationWarning, stacklevel=2)
\ No newline at end of file
diff --git a/pyhealth/datasets/__init__.py b/pyhealth/datasets/__init__.py
index c29955e7d..6835363d7 100644
--- a/pyhealth/datasets/__init__.py
+++ b/pyhealth/datasets/__init__.py
@@ -7,6 +7,7 @@ def __init__(self, *args, **kwargs):
         warnings.warn(
             "The BaseEHRDataset class is deprecated and will be removed in a future version.",
             DeprecationWarning,
+            stacklevel=2,
         )
 
 
@@ -19,6 +20,7 @@ def __init__(self, *args, **kwargs):
         warnings.warn(
             "The BaseSignalDataset class is deprecated and will be removed in a future version.",
             DeprecationWarning,
+            stacklevel=2,
         )
 
 
@@ -31,6 +33,7 @@ def __init__(self, *args, **kwargs):
         warnings.warn(
             "The SampleEHRDataset class is deprecated and will be removed in a future version.",
             DeprecationWarning,
+            stacklevel=2,
         )
 
 
@@ -43,6 +46,7 @@ def __init__(self, *args, **kwargs):
         warnings.warn(
             "The SampleSignalDataset class is deprecated and will be removed in a future version.",
             DeprecationWarning,
+            stacklevel=2,
         )
 
 
diff --git a/pyhealth/datasets/mimic3.py b/pyhealth/datasets/mimic3.py
index 7e569d2f3..150db2892 100644
--- a/pyhealth/datasets/mimic3.py
+++ b/pyhealth/datasets/mimic3.py
@@ -60,6 +60,7 @@ def __init__(
                 "Events from prescriptions table only have date timestamp (no specific time). "
                 "This may affect temporal ordering of events.",
                 UserWarning,
+                stacklevel=2,
             )
         super().__init__(
             root=root,
diff --git a/pyhealth/datasets/mimic4.py b/pyhealth/datasets/mimic4.py
index 9d1aa55d8..c5392fb1e 100644
--- a/pyhealth/datasets/mimic4.py
+++ b/pyhealth/datasets/mimic4.py
@@ -122,12 +122,14 @@ def __init__(
                 "Events from discharge table only have date timestamp (no specific time). "
                 "This may affect temporal ordering of events.",
                 UserWarning,
+                stacklevel=2,
             )
         if "discharge_detail" in tables:
             warnings.warn(
                 "Events from discharge_detail table only have date timestamp (no specific time). "
                 "This may affect temporal ordering of events.",
                 UserWarning,
+                stacklevel=2,
             )
         log_memory_usage(f"Before initializing {dataset_name}")
         super().__init__(
diff --git a/pyhealth/models/concare.py b/pyhealth/models/concare.py
index c642ee810..c4274a856 100644
--- a/pyhealth/models/concare.py
+++ b/pyhealth/models/concare.py
@@ -954,6 +954,7 @@ def __init__(
                     "and sum-pooled per visit for memory efficiency. This reduces "
                     "per-visit code representations to a single scalar.",
                     UserWarning,
+                    stacklevel=2,
                 )
             else:
                 input_dim = proc.size()
diff --git a/pyhealth/models/stagenet.py b/pyhealth/models/stagenet.py
index 79900e134..11ad3fb73 100644
--- a/pyhealth/models/stagenet.py
+++ b/pyhealth/models/stagenet.py
@@ -440,6 +440,7 @@ def forward_from_embedding(
                     f"StageNet format with time intervals for "
                     f"better performance.",
                     UserWarning,
+                    stacklevel=2,
                 )
             else:
                 time = time.to(self.device)
@@ -453,6 +454,7 @@ def forward_from_embedding(
                     f"Feature '{feature_key}' does not have mask "
                     f"information. Default mask will be created from "
                     f"embedded values. But it may not be accurate.",
+                    stacklevel=2,
                 )
                 mask = (value.abs().sum(dim=-1) != 0).int()
             elif not processor.is_token() and value.dim() == mask.dim():
diff --git a/pyhealth/models/stagenet_mha.py b/pyhealth/models/stagenet_mha.py
index 637f2fdc5..5e3d5acef 100644
--- a/pyhealth/models/stagenet_mha.py
+++ b/pyhealth/models/stagenet_mha.py
@@ -516,6 +516,7 @@ def forward_from_embedding(
                     f"StageNet format with time intervals for "
                     f"better performance.",
                     UserWarning,
+                    stacklevel=2,
                 )
             else:
                 time = time.to(self.device)
@@ -528,6 +529,7 @@ def forward_from_embedding(
                     f"Feature '{feature_key}' does not have mask "
                     f"information. Default mask will be created from "
                     f"embedded values. But it may not be accurate.",
+                    stacklevel=2,
                 )
                 mask = (value.abs().sum(dim=-1) != 0).int()
             elif not processor.is_token() and value.dim() == mask.dim():
diff --git a/pyhealth/models/text_embedding.py b/pyhealth/models/text_embedding.py
index 57549fea5..15e388555 100644
--- a/pyhealth/models/text_embedding.py
+++ b/pyhealth/models/text_embedding.py
@@ -244,6 +244,7 @@ def _chunk_and_encode(
                 f"Text produced {original_chunks} chunks, truncated to {self.max_chunks}. "
                 f"Consider increasing max_chunks or summarizing input.",
                 UserWarning,
+                stacklevel=2,
             )
 
         # Step 4: Pad chunks to uniform length for batched encoding