All patient Refined-Diagnosis Related Group (APR-DRG) Severity of illness and risk of mortality as in-hospital mortality predictors

• Autor(a) principal: Santos, João Vasco Nunes dos
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/134915
• topic: In-hospital mortality; prediction; Diagnosis-Related Groups; Severity of Illness; Risk of Mortality; Charlson Comorbidity Index; Elixhauser Comorbidity Index; reliability; agreement; Mortalidade intra-hospitalar; predição; Grupos de Diagnóstico Homogéneo; Domínio/Área Científica::Ciências Sociais::Outras Ciências Sociais

ABSTRACT - Background: Predicting in-hospital mortality can be important for several purposes. Several countries relying on All-Patient Refined Diagnosis-Related Groups (APR-DRG) payment schemes calculate Severity of Illness (SOI) and Risk of Mortality (ROM) scores. Comorbidity measures such as the Charlson Comorbidity Index (CCI) or the Elixhauser Comorbidity Index (ECI) are commonly used. The aims of this thesis were (1) to assess the level of reliability and agreement between SOI and ROM, (2) to assess SOI and ROM as predictors of in-hospital mortality, (3) to compare SOI and ROM with Charlson Comorbidity and Elixhauser Comorbidity weighted scores as in-hospital mortality predictors; and (4) to identify the specific DRGs in which SOI and/or ROM perform better. Methods: A retrospective observational study was performed, using mainland Portuguese public hospitalizations of adult patients from 2011 to 2016. Reliability and agreement between SOI and ROM were assessed by means of quadratic weighted kappa and proportion of agreement, respectively, overall and by APR-DRG. For each APR-DRG, original and updated CCI and ECI were calculated and compared to SOI and ROM, each combined with age and sex. Model discrimination (C-statistic/ area under curve) and goodness-of-fit (R-squared) were calculated. Results: Considering 4,467,198 hospitalizations, overall reliability and agreement between SOI and ROM were high (weighted kappa: 0.717, 95%CI 0.717-0.718; proportion of agreement: 69.0%, 95%CI 69.0-69.0). However, there was high heterogeneity across APR-DRGs, ranging from 0.016 to 0.846 on reliability and from 23.1% to 94.8% on agreement. Most of APR-DRGs (263 out of 284) showed a higher proportion of episodes with ROM level above the SOI level than the contrary. Regarding the in-hospital mortality prediction, the results comprised 4,176,142 hospitalizations with 5.9% in-hospital deaths. Compared to the CCI and ECI models, the model that considered SOI, age and sex showed a statistically significantly higher discrimination in 49.6% (132 out of 266) of APR-DRGs, while this value was only 33.9% of APR-DRGs for the model with ROM. Between these two models, SOI was the best performer for nearly 20% of APR-DRGs. Some particular APR-DRGs have showed good discrimination such as those related to burns, viral meningitis or specific transplants. Conclusions: Severity of Illness and Risk of Mortality measures must be clearly distinguished and are “two scales of different concepts” rather than “two sides of the same coin”. However, this is more evident for some APR-DRGs than for others. As inhospital mortality predictors, SOI or ROM, combined with age and sex, perform better than more widely used comorbidity indices. While ROM is the only score specifically designed for in-hospital mortality prediction, SOI performed better than ROM. These findings can be helpful for hospital or organizational models benchmarking or epidemiological analysis.