First draft of frailty analysis -Descriptive

[RiazAziz]

• [ ]Send SQL codes for Steve to check - Git Hub
• Produce demographic description of the study population
• [ ]- [ ] Outcome : Admission to ICU , for those admitted to ICU ( MV/RRT/Vasopressor support)
• Include a marker for illness severity = NEWS/ APACHE

[docsteveharris]

Issues discussed today

Denominator

• up to 900 admissions in 14m; need to see if we can work out why?
  • check for duplicates incl visit_occurrence, dates of admission etc
  • look for characteristics that might provide a clue: e.g. journey, age, etc
  • get a couple of examples and work with Dave R to double check
• not enough kids and should exclude
• filter all non-ICU out for governance purposes (CCHIC or audit purposes)

Frailty approaches

Polypharmacy definition

• working with unique oral meds on admission after some simple exclusions (e.g. anti-emetics, antibiotics that the inpatient might have started)

ICD-10 codes

• list from Simon that works
• weighted list that is summed to produce score

Co-morbidities

• count within a list of custom codes from ICD10 or SNOMED

Validation suggestions

• backend to frontend reviews
• sense checks e.g. against age
• sense checks e.g. correlations between scores

Confounders

• NEWS on admission and NEWS on ICU admission
• maybe median NEWS on those days

Outcome measure

• ventilation ever
• RRT ever
• hospital mortality
• LoS

Nice to do

• TEPs and DNACPR

[docsteveharris]

already assembled into (several) single tables
including

• one row per patient (first in the time period)
• one row per admission

so work toward a slide presentation for the lab
focus on the cohort building and the features and the checks that have been done for those features
then hopefully have a go at running the model before the presentation

and if all going well then use HYLODE ethics to look at how well these features work for whole hospital questions such as ICU admission and frailty

[RiazAziz]

Quality checks

1. Kids in the database : Total < 18 - 5500