| Number (percent) of interviewees
making the recommendation |
Recommendation |
| 27 |
(50%) |
Develop indicator specifications that rely
on incorporating additional data elements with the administrative data
(including present-on-admission flag, do not resuscitate order flag, clinical
data elements, etc.) |
| 13 |
(24%) |
Perform validation studies |
| 12 |
(22%) |
Develop composite indices |
| 12 |
(22%) |
Improve risk adjustment (alignment with other indicator system, non-proprietary
system) |
| 7 |
(13%) |
Add more analytic tools to the software, such as various levels of
significance testing |
| 5 |
(9%) |
Improve identification of "avoidable" admissions using
secondary diagnoses or other methods |
| 12 |
(22%) |
Develop composite indices |
| 12 |
(22%) |
Improve risk adjustment (alignment with other indicator system, non-proprietary
system) |
| 7 |
(13%) |
Add more analytic tools to the software, such as various levels of significance
testing |
| 5 |
(9%) |
Improve identification of "avoidable" admissions using secondary
diagnoses or other methods |
| 5 |
(9%) |
Improve obstetric PSIs |
| 5 |
(9%) |
Periodically assess the applicability of some of the IQIs to the inpatient
setting, in particular for procedures that are now mostly done on an outpatient
basis, like laparoscopic cholecystectomy |
| 4 |
(7%) |
Provide guidance on how to perform trend analysis over time given changes
in indicator definitions |
| 3 |
(6%) |
Provide guidance on appropriate coding of source of admission |
| 2 |
(4%) |
Adapt PQIs for hospital-level analyses |
| 2 |
(4%) |
Calculate and disseminate cost-effectiveness of quality improvement using
each indicator |
| 1 |
(2%) |
Assume Poisson distribution for counts of infrequent events rather than
normal distribution |
| 1 |
(2%) |
Change smoothing procedure for PSIs so results are not over-smoothed |
| 1 |
(2%) |
Develop a common minimum set of checks of data quality ("common
minimum edits") |
| 1 |
(2%) |
Develop open-source methods for probabilistic data linkages |
| 1 |
(2%) |
Exclude cancer patients from failure to rescue indicator |
| 1 |
(2%) |
Exclude patients transferred in from another hospital from all indicators |
| 1 |
(2%) |
Improve handling of zero numerator events |
| 1 |
(2%) |
Improve risk adjustment for IQI 33—it groups women 18 and younger
and women 35 and older together despite the fact that the different groups
have different risks |
| 1 |
(2%) |
Provide PSIs for all ages—not split pediatric/adult |
| 1 |
(2%) |
Study how coding practices vary across hospitals |
540 Gaither Road Rockville, MD 20850 
