Calculate pairwise measures of health inequality from a fitted surveil time series model, with credible intervals and MCMC samples. Calculates absolute and fractional rate differences (RD and population attributable risk (PAR)), rate ratios, and excess cases.
group_diff(x, target, reference)
# S3 method for surveil
group_diff(x, target, reference)
# S3 method for list
group_diff(x, ...)Menvielle, G, Kulhanaova, I, Machenbach, JP. Assessing the impact of a public health intervention to reduce social inequalities in cancer. In: Vaccarella, S, Lortet-Tieulent, J, Saracci, R, Conway, D, Straif, K, Wild, CP, editors. Reducing Social Inequalities in Cancer: Evidence and Priorities for Research. Geneva, Switzerland: WHO Press, 2017:185-192.
Either a fitted surveil time series model, or a list of two stand_surveil objects (i.e., surveil models with age-standardized rates, as returned by standardize). If x is a list of stand_surveil objects, see details below and note that the models must contain the same number of MCMC samples---to ensure this is the case, when using stan_rw set iter and chains to the same values for each of the two models.
The name (character string) of the disadvantaged group that is the target of inference. If x is a list of stand_surveil objects, the target argument is ignored and the first listed model will serve as the target group.
The name (character string) of the reference group to which target will be compared. If x is a list of stand_surveil objects, the reference argument is ignored and the second listed model will serve as the reference group.
Additional arguments (not used).
A list, also of class "surveil_diff", with the following elements:
A tibble with a summary of posterior distributions (mean and 95 percent cred. intervals) for the target group incidence rate, the RD, RR, PAR, and excess cases.
Summary of the posterior distribution for the cumulative number of excess cases and the PAR (mean and 95 percent cred. intervals)
Character string with target and reference population names
A data frame of MCMC samples for each quantity of interest (target and reference rates, RD, RR, PAR, and EC, as well as Trend_Cases = Rate * Population). Indexed by time.
MCMC samples of the cumulative number of excess cases.
For the following calculations, the terms reference and target refer to incidence rates for the respective groups; p is the size of the target population. (Target is the group that is the 'target' of our inferences, so that it is the numerator in rate ratios, etc.) The following measures are calculated by group_diff:
# rate difference
RD = target - reference
# population attributable fraction
PAR = RD/target = (RR - 1)/RR
# rate ratio
RR = target/reference
# excess cases
EC = RD * pAs the math communicates, the PAR is the rate difference expressed as a fraction of total risk for the target population. This could also be read as the fraction of risk in the target population that would have been removed had the target rate equaled the reference rate (Menvielle et al. 2017).
If the user provides a list of stand_surveil objects with age-standardized rates (instead of a single surveil model), then the exact calculations will be completed as follows. The RR is simply the ratio of age-standardized rates, and the rate difference is similarly the difference between age-standardized rates. However, excess cases is calculated for each age group separately, and the total excess cases across all age groups is returned. Similarly, the attributable risk is calculated by taking the total excess cases across all age groups per year and dividing by the total risk (i.e., by the sum of the whole number of cases across all age groups). Cumulative excess cases is the sum of the time-period specific total number of excess cases. (Notice that the PAR is not equal to (RR-1)/RR when the PAR is derived from a number of age-specific rates and the RR is based on age-standardized rates.)