Understanding and using sensitivity, specificity and predictive values

Skaik, Younis

doi:10.4103/0301-4738.41424

Cited by 23 publications

(10 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A logistic regression model was developed in the modeling sample, and model performance was evaluated based on the c-statistic and the Hosmer-Lemeshow test (39) in the validation sample. Sensitivity, specificity, PPV, and NPV were estimated for various cutoffs of the predictive failure rate that defined fluconazole failure, using the validation sample (40).…”

Section: Methodsmentioning

confidence: 99%

A Risk Score for Fluconazole Failure among Patients with Candidemia

Ostrosky-Zeichner¹,

Harrington

Azie

et al. 2017

Antimicrob Agents Chemother

View full text Add to dashboard Cite

This study aimed to develop a prediction model to identify patients with candidemia who were at high risk of failing fluconazole treatment. Adult patients in the United States with candidemia who received fluconazole during hospitalization were selected from the Cerner Health Facts Hospital Database (04/2004 to 03/2013). Fluconazole failure was defined as switching/adding another antifungal, positive Candida culture Ն10 days after fluconazole initiation, or death during hospitalization. Patients were randomized into modeling and validation samples. Using the modeling sample, a regression analysis of least absolute shrinkage and selection operator was used to select risk predictors of fluconazole failure (demographics, Candida species, initiation of fluconazole before positive culture and after admission, and comorbidities, procedures, and treatments during the 6 months before admission and fluconazole initiation). The prediction model was evaluated using the validation sample. We found that of 987 identified patients (average age of 61 years, 51% male, 72% Caucasian), 49% failed and 51% did not fail fluconazole treatment. Of those who failed, 70% switched or added another antifungal, 21% had a second positive Candida test, and 42% died during hospitalization. Nine risk factors were included in the prediction model: days to start fluconazole after admission, Candida glabrata or Candida krusei infection, hematological malignancy, venous thromboembolism (VTE), enteral nutrition, use of nonoperative intubation/irrigation, and other antifungal use. All but VTE were associated with a higher risk of failure. The model's c-statistic was 0.65, with a Hosmer-Lemeshow test P value of 0.23. In summary, this prediction model identified patients with a high risk of fluconazole failure, illustrating the potential value and feasibility of personalizing candidemia treatment.

show abstract

Section: Methodsmentioning

confidence: 99%

A Risk Score for Fluconazole Failure among Patients with Candidemia

Ostrosky-Zeichner¹,

Harrington

Azie

et al. 2017

Antimicrob Agents Chemother

View full text Add to dashboard Cite

show abstract

“…However, the PPV and NPV are influenced by the prevalence of the disease. The more rare the cancer is, the more sure we can be that a negative predictive value indicates no abnormality, and the less sure we can be that a positive predictive value really indicates an abnormality …”

Section: Discussionmentioning

confidence: 99%

Comparison of two lung cancer screening scores among patients with chronic obstructive pulmonary disease: A community study

Gagnat

Gulsvik

Bakke

et al. 2019

Clinical Respiratory J

View full text Add to dashboard Cite

Introduction Based on the National Lung Cancer Screening Trial (NLST), guidelines on screening programs for lung cancer have recommended low‐dose computed tomography (LDCT). De Torres et al made a score for COPD patients (COPD‐LUCSS) to improve their selection criteria. Objective To examine and compare the discriminating value of both scores in a community‐based cohort of COPD patients. Methods Four hundred and twenty‐two ever‐smokers with COPD from the GenKOLS study in Bergen were merged with the Cancer Registry of Norway. We divided the patients into groups of high and low risk according to the COPD‐LUCSS and the NLST criteria. Cox regression and logistic regression were used to analyse the associations between the scores and lung cancer. We used Harrell's C and area under the curve (AUC) to estimate discriminating values and to compare the models. Results Hazard ratio for the high risk vs the low risk in the COPD‐LUCSS was 3.0 (1.4‐6.5 95% CI), P < 0.01. Hazard ratio for the NLST criteria was 2.2 (95% CI 1.1‐4.5), P < 0.05. Harrell's C was 0.63 for the COPD‐LUCSS and 0.59 for the NLST selection criteria. AUC was 0.61 for COPD‐LUCSS and 0.59 for NLST criteria. Comparing tests showed no differences (P = 0.76). Conclusion Although the COPD‐LUCSS and the NLST criteria were associated with increased risk of lung cancer, the AUC and Harrell's C values showed that these models have poor discriminating abilities in our cohort of COPD patients. The COPD‐LUCSS was not significantly better than the NLST criteria.

show abstract

“…The purpose of diagnostic or screening tests depends on its validity, considering its sensitivity and specificity [77] . Even a small fault of test specificity could bring to false positive results, provoking worry patient concern and enormous costs for subsequent diagnosis.…”

Section: Conclusion and Future Perspectivementioning

confidence: 99%