Department of Global Health and Social Medicine, Harvard Medical School, Boston, MA, USA.
Department of Global Health and Social Medicine, Harvard Medical School, Boston, MA, USA, Partners In Health (PIH), Boston, MA, USA, Division of Global Health Equity, Brigham and Women´s Hospital, Boston, MA, USA.
Int J Tuberc Lung Dis. 2023 Jan 1;27(1):34-40. doi: 10.5588/ijtld.22.0324.
The WHO provides standardized outcome definitions for rifampicin-resistant (RR) and multidrug-resistant (MDR) TB. However, operationalizing these definitions can be challenging in some clinical settings, and incorrect classification may generate bias in reporting and research. Outcomes calculated by algorithms can increase standardization and be adapted to suit the research question. We evaluated concordance between clinician-assigned treatment outcomes and outcomes calculated based on one of two standardized algorithms, one which identified failure at its earliest possible recurrence (i.e., failure-dominant algorithm), and one which calculated the outcome based on culture results at the end of treatment, regardless of early occurrence of failure (i.e., success-dominant algorithm). Among 2,525 patients enrolled in the multi-country endTB observational study, we calculated the frequencies of concordance using cross-tabulations of clinician-assigned and algorithm-assigned outcomes. We summarized the common discrepancies. Treatment success calculated by algorithms had high concordance with treatment success assigned by clinicians (95.8 and 97.7% for failure-dominant and success-dominant algorithms, respectively). The frequency and pattern of the most common discrepancies varied by country. High concordance was found between clinician-assigned and algorithm-assigned outcomes. Heterogeneity in discrepancies across settings suggests that using algorithms to calculate outcomes may minimize bias.
世界卫生组织为利福平耐药（RR）和耐多药（MDR）结核病提供了标准化的结局定义。然而，在某些临床环境中实施这些定义可能具有挑战性，不正确的分类可能会导致报告和研究中的偏倚。通过算法计算的结果可以提高标准化程度，并适应研究问题。我们评估了临床医生分配的治疗结局与基于两种标准化算法之一计算的结局之间的一致性，一种算法在最早可能复发时确定失败（即失败主导算法），另一种算法根据治疗结束时的培养结果计算结局，而不考虑早期失败的发生（即成功主导算法）。在多国家结核病观察性研究中，我们对临床医生分配的和算法分配的结局进行了交叉制表，计算了一致性的频率。我们总结了常见的差异。算法计算的治疗成功率与临床医生分配的治疗成功率高度一致（失败主导算法和成功主导算法分别为 95.8%和 97.7%）。最常见差异的频率和模式因国家而异。在临床医生分配和算法分配的结局之间发现了高度一致性。不同环境中差异的异质性表明，使用算法计算结果可能会最大限度地减少偏差。
