How to get attrition table and chart in HADES?

Developers
4

In the Cohort output, there are three distinct tables that complement the main cohort table. The main cohort table as you has cohort_definition_id, subject_id, cohort_start_date, cohort_end_date. The supporting tables are as follows:

  1. cohort_inclusion ("#cohort_inclusion"): This serves as the rule metadata table, containing rule names and descriptions specific to each cohort. This table helps in identifying which rule is being examined.
  2. cohort_inclusion_stats ("#cohort_inc_stats"): This is essentially the result table, containing fields such as person_count, gain_count, and person_total. The table uses cohort_definition_id as the entity and attributes like rule_sequence and mode_id.
  3. cohort_inclusion_result ("#cohort_inc_result"): Another result table, this is where the ingenuity of @Chris_Knoll becomes apparent. Here, the entity is cohort_definition_id while the attributes include mode_id and inclusion_rule_mask. The value column contains person_count.

Regarding mode_id: The attribute mode_id has two key values:

  • mode_id 0: Denotes all events
  • mode_id 1: Indicates the “best event,” which represents the single event per person meeting the highest number of inclusion criteria.

The cohort_inclusion_result uses inclusion_rule_mask to summatively represent which of the inclusion rules were met for a particular combination. For example, an inclusion_rule_mask of 3 means that rules 1 and 2 were met, and this will be represented in the person_count column for the respective mode_id values. How do I know that:

as.integer(intToBits(0)) = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 i.e. none of max 32 inclusion rules were met.

as.integer(intToBits(3)) = 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 i.e. met rule 1 and 2.

I continue to be amazed by the brilliance of this approach, as initiated by @Chris_Knoll. It by using bitmask numbers to communicate the status of up to 32 inclusion rules efficiently.

Now we can use bit operators in very interesting ways to find count of persons who met only rule 3, or count of persons that met rules 1, 2 but not 3. etc. I heard there is some cool work in OHDSI ShinyModules to use the output to create various types of visualizations of attrition. I cant wait for that innovation.

2 Likes