Multi-Site Analysis for Independent Data Sources

The multi-site analyses included in this suite are intended to be executed against data that are all stored in the same place. However, there may be some instances where the data associated with each site is stored in independent locations. This vignette outlines how the multi-site analysis can be executed in these instances.

All multi-site analyses, regardless of the check type, can reproduced in the same way.

First, execute either of the Single Site, Exploratory analyses, configured appropriately for your study, against each data source.

library(clinicalevents.specialties)

my_table <- cnc_sp_process(cohort = my_cohort,
                           multi_or_single_site = 'single',
                           anomaly_or_exploratory = 'exploratory',
                           time = T / F,
                           ...)

Then, combine both the results with visit counts AND the specialty type results (for labeling) into a single table with the different sites delineated in the site column.

my_final_counts <- my_table1[[2]] %>% dplyr::union(my_table2[[2]]) ... %>%
  dplyr::union(my_table_n[[2]])

my_final_specialties <- my_table1[[1]] %>% dplyr::union(my_table2[[1]]) ... %>%
  dplyr::union(my_table_n[[1]])

Once you have the combined table of counts, you will want to replace the output_function column to reflect the Multi-Site check type you would like to execute. Choose from the table below:

Check Type	output_function
Multi Site, Exploratory, Cross-Sectional	cnc_sp_ms_exp_cs
Multi Site, Exploratory, Longitudinal	cnc_sp_ms_exp_la
Multi Site, Anomaly Detection, Cross-Sectional	cnc_sp_ms_anom_cs
Multi Site, Anomaly Detection, Longitudinal	cnc_sp_ms_anom_la

After this adjustment has been made, the combined & edited table with counts + the combined specialty reference can both be fed into the output function as usual.

my_graph <- cnc_sp_output(cnc_sp_process_output = my_edited_table,
                          cnc_sp_process_names = my_named_specialties,
                          ...)