Looking for predictors
(Updated: 2022-02-23 09:11:15 EDT)
Need to define outcomes to predict
1. Trends of shortlisted metrics
New cases per population
The number of adult ICU and acute beds occupied
1.1. New cases per population
Note about positivity rate:
* It correlates with surges, reflecting demand for testing.
* Important for two reasons: to meet the demand for testing, and to guesstimate if/how much the incidence is being underestimated.
* Moving forward, interpretation becomes challenging because of a few factors:
– results of home tests getting into the data system (numerator may be correct, but denominator is likely not)
– accuracy of home tests
– changing severity of cases because of vaccination and/or changing dominant variants (people may not seek testing if no/few symptoms)
* With good vaccination rate (and thus less severe cases overall), focus may become more about tracking the level of people who need hospital care
1.2. The number of adult ICU and acute beds occupied
- All in one scale
- These four correlates with each other, with time lag. New cases peak comes first, followed by hospitalization, and then death
- Same data, just re-scaled to have a similar height across the four indicators during the initial surge/wave - in order to focus on relative changes in each metric.
- The number of ICU beds likely caps at the capacity, and, thus, is not sensitive to real demand for acute care.
- The number of deaths is an outcome, so not much useful if an immediate response is a main goal. Also, the number of deaths is affected by multiple factors: age composition, improving case management, and severity of cases (vaccine, variants)
- NEXT STEP Add: (1) the number of ICU and acute beds occupied for non-COVID patients, and (2) capacity (i.e., total number of ICU and acute beds occupied and ready to receive patients - roughly straight horizontal line) in order to see overall picture of health systems
2. Age composition of new cases
- Explains/predicts mortality trends, because of big differences in case fatality ratio by age
- No need to track frequently. Monthly or whenever the new case volume surges