Results
The first part of the conference focused on the scientific
basis for quality measurement and new developments in quality assessment.
Participants debated in particular how to translate guidelines into performance
measures and when it was (or was not) appropriate to use general guideline
targets (such as an A1c<7 percent) as dichotomous threshold definitions of
care quality for use in performance measurement.
Participants also discussed the science and practicality of
using more clinically based measures, such as tightly linked clinical action
measures and weighted continuous measures, as well as measures that track
assessment of performance by individuals of lower socioeconomic position and
low literacy. A session also dealt specifically with applicability of
performance measures to older adults.
These discussions set the stage for the more specific
examinations and recommendations on measures to assess and improve glycemic,
blood pressure, and lipid control, as well as measures to assess and promote
patient self-management support. We will highlight the discussion that centered
around assessing glycemic control because it represents well the challenges we
face in constructing and promoting performance measures that improve any
intermediate outcomes (e.g., glycemic, blood pressure, and lipid outcomes)
without causing unintended consequences. We also will summarize several
additional issues examined during the conference and highlight general
recommendations for practice and research.
Assessing
Glycemic Control
This conference took place immediately following decisions
by the NCQA Committee on Performance Measurement to revise its diabetes
performance measurement set, including adding dichotomous stringent blood
pressure and glycemic control measures. The dichotomous stringent glycemic
control measure specifies that plans should measure what proportion of their
adult members 75 years of age and under have a glycosylated hemoglobin level
(A1c) <7.0 percent, as well as what proportion have poorly controlled A1c
(>9 percent).
Although it was purely coincidental that the scientific
conference on diabetes quality assessment took place 2 weeks after this
decision, it did set the stage for what turned out to be a rather lively and
productive discussion on the benefits and problems with performance measures
that attempt to asses care that is stringently defined by guidelines. In order
to better capture that debate, we present the issues discussed as arguments
supportive of a stringent control measures and arguments opposing this
position. Similar arguments apply to measures specifying stringent blood
pressure and lipid control. Alternative recommendations for assessing good
glycemic control are also discussed.
Arguments supporting a stringent control measure. Several
participants felt strongly that setting stringent performance measures will
help move more patients toward better control, with resulting improvements in
downstream outcomes. They argued that the goal of performance measurement
should be quality improvement, and if a stringent control measure helps to move
the majority of the population toward optimal control, then it will have
benefited the population and improved outcomes. In addition, they suggested
that it is important to align performance measures with recommended guideline
goals, both to minimize confusion on the part of clinicians and to be able to
gauge success in working toward the stringent, guideline level of control.
Having only a poor control measure (A1c >9 percent), which previously had
been the only measure of glycemic control in the diabetes measurement set,
might send the wrong message that getting the A1c lower than 9 percent is
adequate care.
Further, it was suggested that for control measures
achieving 100 percent adherence is not really the goal, but rather the effort
is aimed at gauging relative progress towards a stated guideline. That is, it
would be expected that not all individuals could or should reach the stringent
control standard, and that plans and providers need not strive for that
standard on 100% of patients, but rather some lower percentage. This
understanding of the purpose of control measures, they maintained, should
reduce gaming and unnecessary treatment of patients with contraindications to
tight control or whose disease is so severe or dietary and medication adherence
so poor that tight control is not a realistic goal. A stringent control measure
could continue to use the value of the last A1c in the reporting period, thus
not requiring any additional data collection.
Arguments opposing a stringent control measure. Many participants
at the conference did not agree with the NCQA Committee on Performance
Measurement decision to set dichotomous stringent performance measures (i.e.,
A1c <7.0 percent and BP <130/80). They felt that optimal control
measures represent unadjusted outcome measures. Without any case-mix
adjustment, these measures seemed to them more likely to reflect the underlying
severity of illness in the patient population than the quality of care
delivered by providers. The result would probably be that health plans and
physicians caring for older and sicker patients would not be fairly compared
with those caring for younger patients and patients who are not as sick.
Even more troubling, they argued, setting goals for tight
control in the absence of consideration of current treatment intensity criteria
carries a substantial risk of harming patients by encouraging over-treatment.
This would open up the potential for unwarranted health care costs, patient
burden, and perhaps even patient safety problems resulting from poly-pharmacy.
This is particularly problematic for older patients, who are less likely to
benefit from tight glycemic control but could be subjected to an additional
medication burden and side effects.
In particular, it was brought out that a dichotomous
stringent control measures does not consider patient comorbidities, disease
severity, the amount of treatment the patient is already receiving, and whether
the patient is close to (A1c = 7.2) vs. far away from (A1c = 8.7) the ideal
goal. Because the individual benefits of achieving tight glycemic control vary
widely, mainly based on age and life expectancy, they argued, a single cut-off
for "good control" runs counter to the evidence.
Alternative recommendations for a good control measure.
Despite these diverse viewpoints, even those opposing the dichotomous stringent
control measure agreed that a measure of "good" glycemic control (and blood
pressure control), if implemented correctly, could further improve risk factor
control and outcomes for patients with diabetes. These "good control" measures,
they suggested, should consider treatment regimen intensity (at least for blood
pressure and lipid control) and the likelihood of benefit in downstream
outcomes. For A1c, there was general agreement that a continuous weighted A1c
measure, as described above, would be greatly preferable to a dichotomous
stringent control measure because it would take into account the likely benefit
of tight blood pressure control based on patient age. Thus, it would be less
likely to promote costly and potentially harmful treatment among those who are
not likely to benefit from it.
This measure also would not require any additional data
collection beyond the last A1c of the reporting period. Another alternative
would be higher threshold values to define "good" control (such as 8 percent).
Tightly linked clinical action measures and those based on longitudinal data
(to track improvement among individual patients identified with poor control in
a previous reporting period) are attractive alternatives but would require
access to more detailed clinical data and expansion of the patient population
assessed, thereby increasing the complexity of the measurement process.
Other
Key Discussion Items
Assessing blood pressure and lipid control. As
mentioned earlier, a similar discussion ensued around performance measures
targeted at stringent blood pressure and lipid control. In these cases,
however, there was some debate not only about the performance measure target
itself, but also what guideline goals should be recommended. Nonetheless, the
arguments supporting and opposing stringent control measures were similar to
the ones stated above. Suggested alternatives to dichotomous stringent control
measures to promote good blood pressure and lipid control included:
- Clearly specifying the
population eligible (by age, risk factors, etc.) for different thresholds of
blood pressure and lipid levels.
- Promoting tightly-linked
clinical action measures that take into account current treatment intensity so
that full credit is received if the patient is already on three or four
antihypertensive medications or maximum dose statin medication, even if the
guideline lipid level has not been achieved.
- Considering use of continuous
measures that incorporate the relationship between risk factor control and
downstream outcomes (i.e., through the use of QALY metrics).
Assessing self-management support. Another
important discussion revolved around how to implement measures to assess
patients' perspectives of quality of care. It was noted that assessments of
patient satisfaction with care are already widely used at the health plan level
through the use of the CAHPS® Health Plan (CAHPS-HP) Survey.22 However, this survey does not capture important elements of how patients with
chronic disease perceive the quality of the care that they receive and,
particularly, how they assess the support they get for improving
self-management. While the new CAHPS® Clinician and Group (CAHPS-CG)
Survey does focus more on the physician-patient interaction, especially for
patients with chronic illness, there still are only a few questions that relate
directly to how well self-management support is provided.
Expanded measures currently exist and have been used mainly
in research settings.23 A performance measurement focused on self-management support would encourage
providers to increase the focus, through care reorganization, on patient goal
setting, education, and between-visit care. Because self-management is a
critical component of good patient outcomes, a performance measure based on the
quality of self-management support provided to patients should enhance both
patient care and downstream outcomes.
Although the measures to assess self-management support were
generally perceived to be important and valid, conference participants'
principal main concern with their widespread use was the additional cost of
implementing a survey for patients with diabetes or chronic diseases (the
current CAHPS-HP survey is sent to a random sample of plan participants).
Several alternatives were presented to make gathering patient assessments more
attractive.
- First, the cost could be minimized by keeping the survey short
and administering it every other year instead of yearly.
- Second, some patient assessments could be routinely collected
during the course of clinical care. For example, self-management and treatments
goals could be collected in an automated fashion at the point of a clinic
encounter and entered into an electronic template with extractable data fields.
These variables could then be used in performance assessment.
- Third, to assess and improve self-management support, surveys
could focus not only on how patients perceive that support, but also what
structural improvements medical groups and health plans have made to advance
patient self-management.
Such assessments identify success (or problems) with
progress in instituting the chronic care model—for example, movement toward
team-based care, use of an electronic health record, and so forth.21 While not substituting for patients' assessments, such structural performance
measures could help advance the organization of care in groups and plans to
better support patient-centered care.
The need for better data. Most participants agreed
that getting to the next generation of diabetes measures would require access
to more clinically detailed data than currently available. Some felt that
promoting tightly linked clinical action measures and patient assessments was
impractical because those measures do not use currently available data. Most
participants, however, felt strongly that we need to push for the systematic
collection of detailed patient-centered data in order to construct measures that
are clinically meaningful and actionable. If we persist in promoting only
measures that can be constructed given current data constraints, they argued,
health care organizations will never be motivated to improve their data systems
to allow for the systematic collection of, for example, medication doses, vital
signs, or patient assessments.
Participants recognized that in most health systems, using
more detailed clinical data would increase the complexity and cost of data
collection. To truly assess quality, and not utilization proxies for quality or
outcomes that are not risk-adjusted, participants suggested that a
transformation is needed in the types of data that are available for quality
assessment. Such a transformation will not happen unless we begin to insist on
some measures that utilize more clinically meaningful data.
Return to Contents
Proceed to Next Section