Clinical Summary: Using Natural Language Processing to Evaluate Differences in Psychotherapeutic Services for Posttraumatic Stress Disorder in a Suicide-Risk-Stratified Veteran Sample
Veterans with PTSD carry substantial suicide risk, yet routine encounter counts can miss meaningful differences in what actually happens during psychotherapy. This study shows that among risk-matched Veterans with PTSD, psychotherapy note content differed between those who did and did not die by suicide, pointing to risk-tier-specific gaps in treatment focus.
Key Findings
- The final matched sample consisted of 722 cases and 8,664 controls, including 171 cases and 2052 controls in the high-risk group, 402 cases and 4,824 controls in the moderate-risk group, and 149 cases and 1,788 controls in the low-risk group.
- Based on perplexity curves, LDA visualizations, and minimizing topic overlap and redundancy, 5 was selected as the optimal number of topics, and a 20% topic proportion cutoff was chosen because it resulted in the greatest number of significant differences between cases and controls at each risk tier.
- Case-control differences were significant for the Risk topic at the moderate-risk tier, for the Treatment Planning topic at the moderate- and high-risk tiers, for the Evaluation topic at the high-risk tier, for the Psychosocial topic at the low- and moderate-risk tiers, and for the Medication topic at all patient subgroups.
- When comparing cases and controls across risk tiers, nearly all comparisons showed significant differences, except for low- and moderate-risk cases’ Risk, Psychosocial, and Medication topics; moderate- and high-risk cases’ Evaluation topic; and low- and high-risk cases’ Evaluation topic.
In Veterans with PTSD receiving psychotherapy, unstructured note content distinguished patients who died by suicide from matched controls even when overall service use appeared similar. High-risk documentation shifted toward psychosocial and risk/evaluation themes and away from treatment planning and medication themes, supporting risk-tier-specific psychotherapy assessment and care planning.
Practice Implications
- Do not rely only on psychotherapy visit counts when assessing suicide prevention care in PTSD; review note content for whether treatment planning, medication management, and risk assessment are actually being documented.
- In higher-risk patients, watch for psychotherapy documentation becoming dominated by psychosocial or inpatient-group content while treatment planning and medication themes recede; that pattern was more common as classified risk increased.
- For moderate- and high-risk Veterans with PTSD, consider whether collaborative treatment planning is being maintained during psychotherapy, since Treatment Planning differences separated cases from controls at those tiers.
- Use these findings to support risk-tier-specific quality monitoring in PTSD psychotherapy, while avoiding causal claims about efficacy because the study could not distinguish EBP from non-EBP treatments or specific intervention formats.
