Clearly PTSD and mild-moderate TBI are diseases in need of screening and monitoring for the at risk population of combat veterans. The implications of not diagnosing and treating PTSD/TBI in a timely manner may manifest itself in the future in a number of medical and psychological ways, leading to physical and/or mental illness, substance abuse, unemployment, or potentially violence against themselves or others. Detection of signs, symptoms, and manifestations of PTSD and/or TBI can assist with overcoming barriers and gaps with current treatment plans. Earlier detection of cases will prevent downstream effects of more severe and recalcitrant illness, leading to better health and quality of life and cost-effective care. Project Aim: To use common VA compensation and pension examination report documents as the platform for development and testing text-processing to detect terminology for signs, symptoms, and manifestations that represent the case definitions and clinical sequelae of PTSD and mild-moderate TBI, with emphasis on the OIF/OEF combat era veterans.
The objectives are to: (1) Create a viable operational definition of the signs, symptoms, and manifestation terms that correspond to the clinical case definitions for mild TBI and PTSD. (2) Create and evaluate an algorithm of text-processing rules for the likelihood of TBI and PTSD related manifestations using existing SNOMED CT clinical terminology. (3) Execute the PTSD and TBI algorithms on a set of cases with PTSD diagnosis for prevalence of TBI manifestations in PTSD and vice versa.
This study is cross-sectional. It uses the PTSD/TBI C&P reports to conduct natural language concept-based indexing to identify, extract, and codify signs, symptoms, and manifestations of PTSD and TBI. Report documents were divided into a training set of 50 PTSD exams and a test set of 150 PTSD exams and a training set of 50 TBI (Brain) cases and test set of 145 TBI cases. We used the MVCS indexing engine to process words and phrases found in text reports into SNOMED CT encoded concepts. The final rule set for TBI (repeated for PTSD) is applied to a subset of ten human reviewed and indexed exams in a qualitative study for summary counts of true positive (TP), false positive (FP) and false negative (FN) classifications and the statistical estimation of sensitivity (Se) and positive predictive value (PPV) for each of the case definition criteria. The rule set for TBI is applied to the PTSD cases to provide information on the co-occurrence of mild TBI in OEF/OIF veterans with PTSD.
Overall positive predictive value was 77% for TBI and 78% for PTSD. The overall sensitivity was 63% for TBI and 52% for PTSD. Thus, these preliminary findings establish the lower limit for estimating the validity of using concept-based indexing for identification and classification of PTSD and sequelae of TBI. The symptom clusters with greater rule extraction frequency in TBI cases were Cognitive (23%), Affective (21%), and Physical (18%) symptoms. The TBI symptom rule extractions in PTSD cases were Affective (37%), Physical (15%), and Cognitive (13%). Thus, cognitive symptoms predominate in the TBI cases whereas Affective symptoms predominate in PTSD cases. These differences might reflect surveillance bias associated with the type of examination rather than actual case differences.
Concept mapping/ indexing tools and coded text extraction will help researchers answer questions related to the epidemiology, assessment, quality, and early detection of TBI and PTSD. The key features differentiating these veteran groups will inform and enable clinicians to tailor therapy, treatment, and rehabilitation regimens. Our ultimate goal is to create automated text-processing tools that can serve to alert providers to the presence of PTSD and/or mild-moderate MTBI (PTSD/TBI) and the long-term effects of PTSD/TBI.
External Links for this Project
None at this time.