Algorithm for Ranking Reliability of Data
Data reliability is set to low, fair or good according to our scoring system from 0 to 100 points.
- Comparable: Data is based on non-targeted screening of entire population or a subset, such as an entry (or discharge) cohort, or of one (or more) facilities. The prevalence is calculated with the numerator (number of HCV positive individuals) corresponding to the denominator (all individuals in the screened group that could be positive). The prevalence can refer to percentage with antibody (seropositive) or virus (viremic), Maximum score for comparability: 50 points if numerator and denominator clearly refer to same population group. If:
- the numerator refers to known positives and denominator is entire population, or
- screening is not consistently routine and non-targeted, then comparability is 0 points. Some data sources are given partial points (40) when they are close to meeting criteria of comparability.
- Contemporary: Prevalence data were collected in the past 5 years: 25 points; >5 years: 0 points.
- Consistent: Percentage positive, for data provided in response to different queries within a time period of 5 years, do not vary. If the largest value is no greater than 130% of the lowest value, we judge the data consistent. 25 points.
Low, Fair or Good Data Reliability
A score of 80 or greater is considered good reliability, a score greater than 25 but lower than 80 is considered fair reliability, and a score of 25 or below is considered low reliability.
Data with a low reliability is often an estimate based on prevalence in a similar state or a national mean. It may also represent data that were collected from national survey by journalism fellow, without clarification of antibody or viremia. Data with a fair reliability may be data that are consistent and contemporary, but it is unclear how testing for HCV is done in the state, or how prevalence was calculated. Or, it may be good data that are not contemporary or consistent with estimates given to the press. Data with a good reliability may be based on non-targeted screening or a non-targeted study, and appropriate methods of calculating prevalence. It is also either contemporary and/or consistent.