Ideas for additional analysis

[JaidipGill]

• How to describe rate ratios on sparse practice level measures? Use rate difference? Adjust baseline rates by offset?
• [ ] Add analysis of appointment types/statuses and SRO measures to look at changes in proportion rather than rate since the rates are confounded by total appt number -> Do this by adjusting for total no. of appts in a glm, in stats_test.r.
  • ~[ ] How would this differ from restriction to _appts in interval ~
• Clustering measures in backlog
• Backlog more rigorous pressure evaluation
• Further analysis of best measure -> variation between years and within winter -> Variance decomposition?

Further discussion needed

• [ ] Stratfication by quintiles of electronic frailty score and 20-comorbidity score
• [ ] Practice-stratification (as described in original approval text: "how those pressures may vary by practice characteristics (e.g. list size, staff composition, etc)". -> Practice subgroup analysis #143
• A measure that captures pressure with respect to capacity (e.g. propn of max year) as discussed originally here
• Potential comorbidity analysis - Number of comorbidities
  over time, which would confirm duplication of referral request dates.
• The bottom decile of start/seen_in_interval has a very low rate, which is unusual. Only including practices with registrations the whole interval made them non-zero, but its still lower than expected and should be explained.
• Consider moving epidemic method approach - For each measure, create average curve across seasons (around each years peak week) and compare current levels to it.
• Future service idea: A dashboard kinda thing with a timeseries where one line = weekly average rate and second line = current weekly rate. Can be done at national level, or practice level to allow practices to self-evaluate changes in their measures.