RESUMO
Patients' narratives about traumatic experiences and symptoms are useful in clinical screening and diagnostic procedures. In this study, we presented an automated assessment system to screen patients for posttraumatic stress disorder via a natural language processing and text-mining approach. Four machine-learning algorithms-including decision tree, naive Bayes, support vector machine, and an alternative classification approach called the product score model-were used in combination with n-gram representation models to identify patterns between verbal features in self-narratives and psychiatric diagnoses. With our sample, the product score model with unigrams attained the highest prediction accuracy when compared with practitioners' diagnoses. The addition of multigrams contributed most to balancing the metrics of sensitivity and specificity. This article also demonstrates that text mining is a promising approach for analyzing patients' self-expression behavior, thus helping clinicians identify potential patients from an early stage.
Assuntos
Mineração de Dados , Diagnóstico por Computador , Programas de Rastreamento , Narração , Processamento de Linguagem Natural , Autorrelato , Transtornos de Estresse Pós-Traumáticos , Adolescente , Adulto , Algoritmos , Árvores de Decisões , Diagnóstico Precoce , Feminino , Humanos , Determinação da Personalidade/estatística & dados numéricos , Reprodutibilidade dos Testes , Transtornos de Estresse Pós-Traumáticos/classificação , Transtornos de Estresse Pós-Traumáticos/diagnóstico , Transtornos de Estresse Pós-Traumáticos/psicologiaRESUMO
OBJECTIVE: To evaluate the reliability of a crosswalk, developed in The Netherlands, between the Health Assessment Questionnaire (HAQ) disability index (DI) and the Short Form 36 physical functioning scale (PF-10) in a sample of patients with various rheumatic diseases in the US. METHODS: Baseline data from patients with rheumatoid arthritis (RA; n = 29,020), fibromyalgia (FM; n = 3,776), and systemic lupus erythematosus (SLE; n = 1,609) participating in the National Data Bank for Rheumatic Diseases were analyzed. Reliability of the crosswalk was evaluated by calculating intraclass correlation coefficients (ICCs), and agreement between observed and predicted scores was evaluated using the Bland-Altman approach. RESULTS. The crosswalk produced reliable conversions for both the HAQ DI (ICC range 0.70-0.77) and PF-10 (ICC range 0.73-0.78) in all 3 disease groups. The mean difference between observed and expected scores was close to zero in US patients with RA. For all 3 disease groups, the limits of agreement were fairly wide and conversion at the level of individual patients is not recommended. CONCLUSION: The crosswalk produced reliable conversions at the group level in a crosscultural setting and can be used to convert HAQ DI to PF-10 scores and vice versa in US patients with RA, FM, or SLE.
Assuntos
Avaliação da Deficiência , Matemática/métodos , Avaliação de Resultados da Assistência ao Paciente , Doenças Reumáticas/diagnóstico , Doenças Reumáticas/fisiopatologia , Índice de Gravidade de Doença , Inquéritos e Questionários , Adulto , Idoso , Artrite Reumatoide/diagnóstico , Artrite Reumatoide/fisiopatologia , Estudos de Coortes , Comparação Transcultural , Feminino , Fibromialgia/diagnóstico , Fibromialgia/fisiopatologia , Humanos , Lúpus Eritematoso Sistêmico/diagnóstico , Lúpus Eritematoso Sistêmico/fisiopatologia , Masculino , Pessoa de Meia-Idade , Países Baixos , Reprodutibilidade dos Testes , Estados UnidosRESUMO
OBJECTIVE: To determine the psychometric properties of the Academic Medical Center (AMC) Linear Disability Scale (ALDS) item bank in a population of patients with rheumatoid arthritis (RA). METHODS: 129 patients with RA completed the ALDS and Health Assessment Questionnaire Disability Index (HAQ-DI) at baseline, and after 8 and 16 weeks of anti-tumor necrosis factor-alpha treatment. Disease activity assessments at these timepoints included serum levels of C-reactive protein, Disease Activity Score 28, morning stiffness, and visual analog scales for global disease activity and fatigue. RESULTS: Reliability of the ALDS was excellent (homogeneity, Cronbach's alpha = 0.95; test-retest, intraclass correlation coefficient = 0.93). The ALDS results at baseline were strongly correlated with the HAQ-DI (r = -0.75). With regard to known group validity, both instruments discriminated between higher and lower disease activity (ALDS, p < 0.0001; HAQ-DI, p = 0.002) and between non-, moderate, and good responders (ALDS, p = 0.002; HAQ-DI, p < 0.0001), indicating that both instruments differentiate between groups. The ALDS was moderately to highly responsive to changes between baseline and after 8 weeks and 16 weeks of treatment (standardized response mean, range = 0.71-1.19). No substantial floor or ceiling effects were found. CONCLUSION: Our results show that the ALDS is a promising new instrument, with at least equivalent psychometric properties compared to the HAQ-DI. Advantages of the ALDS item bank are its linear structure and an item bank that can be adapted depending on the ability level of the patient.