COVID Oximetry @home (CO@h) - Feasibility analysis for evaluation of CO@h using regression discontinuity
Links
Summary
This report documents the findings of the Improvement Analytics Unit (IAU)’s feasibility analysis to determine whether a regression discontinuity design would be an appropriate method for evaluating the causal impact of the CO@h programme.
We found that the number and proportion of eligible patients that were onboarded onto the service was low, and that there was very little discontinuity (difference) in onboarding rates above and below the age threshold. For example, approximately 3.5% of eligible patients aged 65–69 were onboarded from January 2021, whereas approximately 2% of patients aged 60–64 were.
The low rates of onboarding and the lack of discontinuity at the age cut-off would result in a lack of statistical power, virtually guaranteeing a null result, even if the intervention was successful in reducing emergency hospital outcomes. Furthermore, the lack of a discontinuity of the magnitude generally required for a high quality RDD would lead one to question the validity of the assumptions of this method relating to non-compliance.
Therefore, we concluded that an RDD was not suitable for the evaluation of the CO@h programme in the current setting.
A number of approaches to evaluate the CO@h programme are being pursued by evaluation partners, including a population-level analysis using generalised synthetic control methods by colleagues within the IAU. Each approach has different assumptions, strengths and weaknesses, but will together aim to provide reliable conclusions about the effect of CO@h on outcomes.