RESUMO
Discharge summaries are an important clinical narrative as they include the continuity of care information. Identification of data contained in their text is a difficult task due to its freeform text and lack of consensus on essential content. This research proposes a rule-based method to verify the presence of information about continuity of care in Portuguese texts, applying Natural Language Processing (NLP) techniques, and based on an annotated medical corpus. After the experiments, 4 rules were defined and applied in the text of 200 summaries to identify if they have or not the continuity of care information. This process had resulted in Precision value of 84%, Recall value of 70%, Specificity value of 97% and F-Measure value of 76% related to algorithm evaluation.
Assuntos
Algoritmos , Continuidade da Assistência ao Paciente/classificação , Processamento de Linguagem Natural , Sumários de Alta do Paciente Hospitalar/classificação , Garantia da Qualidade dos Cuidados de Saúde/métodos , Vocabulário Controlado , Inteligência Artificial , Continuidade da Assistência ao Paciente/estatística & dados numéricos , Mineração de Dados/métodos , Sumários de Alta do Paciente Hospitalar/estatística & dados numéricos , PortugalRESUMO
French Guiana is the region of France where the HIV epidemic is most prevalent. To determine the risk factors for being lost for follow-up, we followed a cohort of 1,213 patients between 1992 and 2002 and determined which variables were related to two definitions of being lost to follow-up: permanently disappearing from HIV clinics and coming back after more than 1 year of missed appointments. The incidence rate for permanent follow-up interruption was 17.2 per 100 person-years. The median time to lost to follow-up was 4.3 years (interquartile range = 1.4-8.4 years). Cox modeling showed that the younger age groups, foreigners, patients with initial CD4 counts at the time of HIV diagnosis less than 500/mm3, and patients followed before the availability of highly active antiretroviral therapy (HAART) were significantly more likely to be permanently lost to follow-up, suggesting that some of the patients may have died. When looking at temporary loss to follow-up, younger age groups, untreated patients, patients consulting before the availability of HAART, and patients with CD4 counts more than 500/mm3 were more likely to not come back for a period of more than 1 year.