A.2.2 Sensitivity of the arithmetic mean to differing proportions of
This section needs a serious re-work
There are multiple things I am trying to express: 1) How does a treatment effect play into this - But maybe we can forget about this as we want to estimate the bias and that bias is independent of the true treatment effect
2) How to calculate and express differences in proportion
3) Relate this to the main heat map graph
thank the following the for their support.
Typo and formatting.
Discussion
Discussion needs some direction, is this a mathematical methods paper or a commentary on the weakness of research methods paper?
understand if a new formula for the total opioid burden can be generated that will account or control this.
Given the main "heat map" graph, I wonder if it possible to have a "proportion corrected" OME equation. Is there a mathematical way to "correct" for the difference in proportion that would mean that un-balance groups would no longer be based by the conversion value error epsilon?
Whilst datasets that contain pre-calculated MME values offer an attractive data source,
Is there a comparison to a more obviously "wrong" data item that we could draw? Such as a pre-calculated NEWS score in a dataset but with no breakdown of the constant parts. (Although NEWS has one common accepted weighting (read as scoring) system so the translation is more explicit)
a sample of 83
This number is just a limitation of time to extract. I could do more if it would have an impact on the quality of the results.
Can expand this section but not sure if its worth it?
I have tried to compile the "range of values used across the literature" but generally the documentation is poor. For example, no-one lists their routes used and this can have a big bearing on the conversion values.
possible grid values of the Neilsen conversion set.
Generally these grid values are rather narrow so the variation produced is relatively narrow. I could use "the range seen in the literature" but this is sensitive of criticism to using unresisting conversion values does lead to unrealistic results.
Figure 4.2 demonstrates that the observed difference can be described with a rotated hyperbolic parabola with the scaling constant k.
This section needs some work I think.
I have not fully been able to link my basic sum (Equation 2.3) to this graph. There are a few steps that go into defining DeltaP_i and epsilon in how they are generated in the experiment.
gure 4.8
To add caption and swap y axis order to match table order
reported difference in opioid burden (red cross) compared to the de-novo calculated difference
There is some difficulty in estimating the group means as show in the table above. Part of this reason is there is a lot of assumptions made. I have tried to apply blanket rules such as conversion values assume IV route, etc.
I feel the difficult in re-calculating the study results perhaps weakens this section as it makes the de-novo estimates (green cross and orange whisker) look far off (such as for Hall 2025)
The problem with the OME metric
Needs a proper title
The problem with the OME metric
Needs a proper title
ackground drafted, some more citations on the various commentery on OME/metrics/issues from the literature to come.
This is a test comment
Use of Electronic Patient Record Systems for Rapid Response to an MHRA Public Assessment Report
Updated title: Use of Electronic Patient Record Systems for Rapid Response to an MHRA Public Assessment Report: Retrospective Observational Study