Table 1. Strengths and weaknesses of strategies for handling missing data
Methodological Considerations in Generating Provider Performance Scores
| Reason data are missing | Data missing in a way that is not related to true performance (e.g., missing at random) | Data missing in a way that is related to true performance (e.g., low performers not reporting data) |
|---|---|---|
| Option 1: Imputation | Stronger: may reduce the risk of misclassification due to chance | Weaker: may introduce systematic performance misclassification |
| Option 2: Report the average score | Stronger: will not result in systematic performance misclassification | Weaker: high likelihood of resulting in systematic performance misclassification; and creates incentive not to report low performance |
| Option 3: Report only the available data | Stronger: will not result in systematic performance misclassification | Weaker: may create incentive not to report low performance |
| Option 4: Report only when providers are not missing any data | Stronger: will not result in systematic performance misclassification | Weaker: may create incentive not to report low performance |
| Option 5: Report the lowest possible score | Weaker: high likelihood of resulting in systematic performance misclassification* | Stronger: potentially less likely to result in systematic performance misclassification; creates incentive to report all performance data |
| Option 6: Report the lowest observed score | Weaker: high likelihood of resulting in systematic performance misclassification* | Stronger: potentially less likely to result in systematic performance misclassification; creates incentive to report all performance data |
* For example, if providers with fewer patients are more likely to have missing data (regardless of their performance), then these providers will be systematically misclassified as low performers.


- 880.13 KB]
5600 Fishers Lane Rockville, MD 20857