Table 1.
Numbered Summary of the Statistical Reporting Guidelines for CHEST
| 2. Reporting of design and statistical analysis |
| 2.1 Where applicable, follow specific reporting guidelines for the type of study being described |
| 2.2 Describe cohort selection completely |
| 2.3 Describe the study questions and the statistical approaches used to address each question in the statistical methods section |
| 2.4 Describe the statistical methods with sufficient detail to allow replication by an independent statistician given the same dataset |
| 2.5 Provide the sample size calculation for a clinical trial |
| 3. Reporting of inference and P values |
| 3.1 Do not accept the null hypothesis; it is either rejected or not rejected |
| 3.2 Avoid stating that P values just above 5% are a trend or are moving |
| 3.3 Do not quantify the probability of a hypothesis with P values or 95% CIs |
| 3.4 Do not equate a statistically significant P value with clinical significance |
| 3.5 Do not use CIs to test hypotheses. |
| 3.6 Caution is warranted when reporting multiple P values |
| 3.7 Do not report separate P values for each of two different groups to address the question of whether there is a difference between groups |
| 3.8 Use interaction terms in place of subgroup analyses |
| 3.9 Avoid using statistical tests to determine the type of analysis to be conducted |
| 3.10 When reporting P values, be clear about the hypothesis tested and ensure that the hypothesis is sensible |
| 4. Reporting of study estimates |
| 4.1 Use appropriate levels of precision |
| 4.2 Avoid redundant statistics in cohort descriptions |
| 4.3 For descriptive statistics, median and quartiles are preferred over means and SDs |
| 4.4 Report CIs for the main estimates of interest |
| 4.5 Do not treat categorical variables as continuous |
| 4.6 Avoid categorization of continuous variables unless there is a convincing rationale |
| 4.7 Do not use statistical methods to obtain thresholds for clinical practice |
| 4.8 Time-to-event analyses |
| 4.8a Report the number of events but not the proportion |
| 4.8b Report median follow-up separately for patients without the event or the number followed up without an event at a given follow-up time |
| 4.8c Describe when the follow-up period started and when and how patients are censored |
| 4.8d Avoid reporting mean follow-up, mean survival time, or estimates of survival in those who had the event of interest |
| 4.8e Make sure that all predictors are known at baseline or consider alternative approaches such as a landmark analysis or time-dependent covariates |
| 4.8f When presenting Kaplan-Meier figures, present the number at risk and truncate follow-up when low |
| 5. Reporting of multivariable models and diagnostic tests |
| 5.1 Do not assume that multivariable, propensity, and instrumental variable analyses will substitute for randomized trials |
| 5.2 Avoid univariable screening and stepwise selection |
| 5.3 When reporting the effects of continuous predictors, choose two clinically interesting predictor values or a clinically relevant range |
| 5.4 Avoid reporting both univariable and multivariable analyses unless there is a good reason |
| 5.5 Avoid ranking predictors in terms of strength |
| 5.6 Be cautious when comparing models assessed on different datasets |
| 5.7 Correct for overfit when conducting internal validation |
| 5.8 Calibration for a prediction model should be presented graphically |
| 5.9 Report the clinical consequences of using a test or a model |
| 6. Discussion, interpretation, and conclusions |
| 6.1 Draw a conclusion; do not just repeat the results |
| 6.2 Avoid using words such as “may” or “might” |
| 6.3 Avoid pseudo-limitations such as “small sample size” and “retrospective analysis,” and consider instead sources of potential bias and the mechanism for their effect on findings |
| 6.4 Discuss the impacts of missing data and patient selection |