Targeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health records
Observational causal inference is useful for decision-making in medicine when randomized clinical trials (RCTs) are infeasible or nongeneralizable. However, traditional approaches do not always deliver unconfounded causal conclusions in practice. The rise of “doubly robust” nonparametric tools coupl...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
IEEE
2022
|
| _version_ | 1811139481601835008 |
|---|---|
| author | Rao, S Mamouei, M Salimi-Khorshidi, G Li, Y Ramakrishnan, R Canoy, D Hassaine, A Rahimi, K |
| author_facet | Rao, S Mamouei, M Salimi-Khorshidi, G Li, Y Ramakrishnan, R Canoy, D Hassaine, A Rahimi, K |
| author_sort | Rao, S |
| collection | OXFORD |
| description | Observational causal inference is useful for decision-making in medicine when randomized clinical trials (RCTs) are infeasible or nongeneralizable. However, traditional approaches do not always deliver unconfounded causal conclusions in practice. The rise of “doubly robust” nonparametric tools coupled with the growth of deep learning for capturing rich representations of multimodal data offers a unique opportunity to develop and test such models for causal inference on comprehensive electronic health records (EHRs). In this article, we investigate causal modeling of an RCT-established causal association: the effect of classes of antihypertensive on incident cancer risk. We develop a transformer-based model, targeted bidirectional EHR transformer (T-BEHRT) coupled with doubly robust estimation to estimate average risk ratio (RR). We compare our model to benchmark statistical and deep learning models for causal inference in multiple experiments on semi-synthetic derivations of our dataset with various types and intensities of confounding. In order to further test the reliability of our approach, we test our model on situations of limited data. We find that our model provides more accurate estimates of relative risk least sum absolute error (SAE) from ground truth compared with benchmark estimations. Finally, our model provides an estimate of class-wise antihypertensive effect on cancer risk that is consistent with results derived from RCTs. |
| first_indexed | 2024-03-07T07:16:34Z |
| format | Journal article |
| id | oxford-uuid:42dca2d1-4e89-46de-83af-c30ccf68a63f |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-09-25T04:06:47Z |
| publishDate | 2022 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:42dca2d1-4e89-46de-83af-c30ccf68a63f2024-06-05T11:43:29ZTargeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health recordsJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:42dca2d1-4e89-46de-83af-c30ccf68a63fEnglishSymplectic ElementsIEEE2022Rao, SMamouei, MSalimi-Khorshidi, GLi, YRamakrishnan, RCanoy, DHassaine, ARahimi, KObservational causal inference is useful for decision-making in medicine when randomized clinical trials (RCTs) are infeasible or nongeneralizable. However, traditional approaches do not always deliver unconfounded causal conclusions in practice. The rise of “doubly robust” nonparametric tools coupled with the growth of deep learning for capturing rich representations of multimodal data offers a unique opportunity to develop and test such models for causal inference on comprehensive electronic health records (EHRs). In this article, we investigate causal modeling of an RCT-established causal association: the effect of classes of antihypertensive on incident cancer risk. We develop a transformer-based model, targeted bidirectional EHR transformer (T-BEHRT) coupled with doubly robust estimation to estimate average risk ratio (RR). We compare our model to benchmark statistical and deep learning models for causal inference in multiple experiments on semi-synthetic derivations of our dataset with various types and intensities of confounding. In order to further test the reliability of our approach, we test our model on situations of limited data. We find that our model provides more accurate estimates of relative risk least sum absolute error (SAE) from ground truth compared with benchmark estimations. Finally, our model provides an estimate of class-wise antihypertensive effect on cancer risk that is consistent with results derived from RCTs. |
| spellingShingle | Rao, S Mamouei, M Salimi-Khorshidi, G Li, Y Ramakrishnan, R Canoy, D Hassaine, A Rahimi, K Targeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health records |
| title | Targeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health records |
| title_full | Targeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health records |
| title_fullStr | Targeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health records |
| title_full_unstemmed | Targeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health records |
| title_short | Targeted-BEHRT: deep learning for observational causal inference on longitudinal electronic health records |
| title_sort | targeted behrt deep learning for observational causal inference on longitudinal electronic health records |
| work_keys_str_mv | AT raos targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords AT mamoueim targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords AT salimikhorshidig targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords AT liy targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords AT ramakrishnanr targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords AT canoyd targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords AT hassainea targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords AT rahimik targetedbehrtdeeplearningforobservationalcausalinferenceonlongitudinalelectronichealthrecords |