MFMatúš Falis

Matúš Falis

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Matúš Falis has a total of 13 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Treatment resistant depression in electronic health records: definitions matter

    by Matthew H. Iveson, Emily L. Ball, Chris Wai Hang Lo, Matúš Falis, Cathryn M. Lewis, Heather C. Whalley

    Background: Many people with depression do not respond well to the first antidepressant prescribed. Treatment Resistant Depression (TRD) refers to depression which does not respond to multiple subsequent antidepressant treatments. Identifying TRD in routinely-collected health records is challenging due to limited response-related data. Previous studies have used definitions based on the number of antidepressant switches observed. However, these do not account for other features clinically indicative of treatment resistance, such as augmentation of antidepressants with lithium or antipsychotics and switches between antidepressant classes. This study examined definitions of TRD and their impact on the resulting sample across three cohorts. Methods: Across the DataLoch, UK Biobank, and Generation Scotland cohorts, we identified cases of depression from primary and secondary care record codes and extracted antidepressant treatment patterns from dispensing/prescribing records ( N = 51,283, N = 10,556, and N = 649 respectively). We examined 9 TRD definitions that varied along two axes: the minimum number of switches required (1+, 2 + or 3 + switches), and the inclusion of other clinical features (augmentation and one or more between-class switches) as alternative routes to TRD. We contrasted sample size and characteristics between definitions, and examined factors associated with inclusion versus a reference definition of 2 + switches. Results: The reference TRD definition included 10% of depression cases in the routine data collection, but substantially fewer cases (4%) in consented cohorts. More inclusive definitions that required fewer switches or included a between-class switch classified more individuals as TRD, but resulted in a proportionally older, more deprived sample with fewer depression-related health record codes, older age of depression onset, lower symptom severity, and greater use of first-line antidepressants. Requiring more switches (3 + switches) classified fewer individuals as TRD, but resulted in a proportionally younger sample, with more depression-related health record codes, younger age of depression onset, and greater use of antidepressants associated with later in the treatment line (e.g., Tricyclics). Definitions including augmentations resulted in a small increase in sample size without notable change in sample characteristics. Conclusions: TRD is underrepresented in consented cohort studies. A definition of TRD that includes 2 + antidepressant switches or augmented antidepressant treatment as indicators balances sample size with depression severity, while incorporating features from real-world treatment journeys. Clinical trial number: Not applicable.


  • Open Access

    Treatment Resistant Depression in electronic health records: Definitions Matter

    by Matthew H Iveson, Emily L Ball, Chris Wai Hang Lo, Matúš Falis, Cathryn M Lewis, Heather C Whalley

    Background: Many people with depression do not respond well to the first antidepressant prescribed. Treatment Resistant Depression (TRD) refers to depression which does not respond to multiple subsequent antidepressant treatments. Identifying TRD in routinely-collected health records is challenging due to limited response-related data. Previous studies have used definitions based on the number of antidepressant switches observed. However, these do not account for other features clinically indicative of treatment resistance, such as augmentation of antidepressants with lithium or antipsychotics and switches between antidepressant classes. This study examined definitions of TRD and their impact on the resulting sample across three cohorts. Methods: Across the DataLoch, UK Biobank, and Generation Scotland cohorts, we identified cases of depression from primary and secondary care record codes and extracted antidepressant treatment patterns from dispensing/prescribing records (N = 51,283, N = 10,556, and N = 649 respectively). We examined 9 TRD definitions that varied by: the number of switches required (1+, 2+ or 3+ switches), augmentation and between-class switches. We contrasted sample size and characteristics between definitions, and examined factors associated with inclusion versus a reference definition of 2+ switches. Results: The reference TRD definition included 10% of depression cases in the routine data collection, but substantially fewer cases (4%) in consented cohorts. More inclusive definitions that required fewer switches or included a between-class switch classified more individuals as TRD, but resulted in a proportionally older, more deprived sample with fewer depression-related health record codes, older age of depression onset, lower symptom severity, and greater use of first-line antidepressants. Requiring more switches (3+ switches) classified fewer individuals as TRD, but resulted in a proportionally younger sample, with more depression-related health record codes, younger age of depression onset, and greater use of antidepressants associated with later in the treatment line (e.g., Tricyclics). Definitions including augmentations resulted in a small increase in sample size without notable change in sample characteristics. Conclusions: TRD is underrepresented in consented cohort studies. A definition of TRD that includes 2+ antidepressant switches or augmented antidepressant treatment as indicators balances sample size with depression severity, while incorporating features from real-world treatment journeys. Key summary: Treatment Resistant Depression (TRD) is often identified in health records using number of switches in antidepressant treatment, but this misses other important indicators of treatment resistance We examined 9 TRD definitions across 3 cohorts, varying in the number of antidepressant switches, the inclusion of augmented treatment, and the inclusion of between-class switches A TRD definition that included 2+ switches or augmented treatment balanced sample size and severity


  • Open Access

    Can GPT-3.5 generate and code discharge summaries?

    by Matúš Falis, Aryo Pradipta Gema, Hang Dong, Luke Daines, Siddharth Basetti, Michael Holder, Rose S Penfold, Alexandra Birch, Beatrice Alex

    Objectives: The aim of this study was to investigate GPT-3.5 in generating and coding medical documents with International Classification of Diseases (ICD)-10 codes for data augmentation on low-resource labels. Materials and Methods: Employing GPT-3.5 we generated and coded 9606 discharge summaries based on lists of ICD-10 code descriptions of patients with infrequent (or generation) codes within the MIMIC-IV dataset. Combined with the baseline training set, this formed an augmented training set. Neural coding models were trained on baseline and augmented data and evaluated on an MIMIC-IV test set. We report micro- and macro-F1 scores on the full codeset, generation codes, and their families. Weak Hierarchical Confusion Matrices determined within-family and outside-of-family coding errors in the latter codesets. The coding performance of GPT-3.5 was evaluated on prompt-guided self-generated data and real MIMIC-IV data. Clinicians evaluated the clinical acceptability of the generated documents. Results: Data augmentation results in slightly lower overall model performance but improves performance for the generation candidate codes and their families, including 1 absent from the baseline training data. Augmented models display lower out-of-family error rates. GPT-3.5 identifies ICD-10 codes by their prompted descriptions but underperforms on real data. Evaluators highlight the correctness of generated concepts while suffering in variety, supporting information, and narrative. Discussion and Conclusion: While GPT-3.5 alone given our prompt setting is unsuitable for ICD-10 coding, it supports data augmentation for training neural models. Augmentation positively affects generation code families but mainly benefits codes with existing examples. Augmentation reduces out-of-family errors. Documents generated by GPT-3.5 state prompted concepts correctly but lack variety, and authenticity in narratives.