How Will You Know Youre

1

• How Will You Know Youre
• Making a Difference?
• Collecting Data and Devising Metrics
• Jeffrey L. Schnipper, MD, MPH
• Director of Clinical Research
• BWH Hospitalist Service
• Brigham and Womens Hospital

2
Overview

• Data Collection
• Measuring quality of glucose control
  glucometrics
• Measuring safety hypoglycemia and extreme
  hyperglycemia
• Measuring processes of care insulin ordering,
  etc.
• Discussion

3
Data Collection

• Ideal
• Automated download of glucose meter data into
  central database
• Link to hospital database with patient, unit, and
  service information
• Automatically generated, customizable reports
• Statistician available when needed

4
Data Collection

• Reality
• No stream of data
• Dont wait, start sampling!
• No link to hospital data
• Create link vs. use all available data
• No reports
• Start with simple analyses (first pass)
• Use Yales glucometrics web site
• No statistician
• Borrow one vs. do without

5
Glucometrics

• Definition
• General Considerations
• What is most clinically relevant?
• What is most sensitive to change?
• What is most likely to be changed by your
  intervention(s)?
• What is least biased?
• What has most statistical power?

6
Glucometrics I Which Patients?

• Who to include? Options
• All patients with discharge dx of DM
• All patients with certain of POC values
• All patients treated with insulin
• All patients with inpatient hyperglycemia
• Who to exclude
• DKA, HHS, Pregnancy
• Patients with fewer than 4-8 readings
• Who to stratify
• ICU vs. Ward

7
Glucometrics I Which Patients?

• Non Critical Care Units
• First pass
• All patients with POC glucose testing
• Exclude DKA, HHS, pregnancy if possible
• Second pass
• All patients with billing codes for diabetes
  (250.xx) or inpatient hyperglycemia (e.g., 2 or
  more glucose readings gt 180 mg/dL)
• Sensitivity analysis exclude patients with fewer
  than 5 evaluable readings

8
Glucometrics I Which Patients?

• Critical Care Units
• First pass
• All patients in each critical care unit
• Second pass
• Patients with DKA, HHS or pregnancy in separate
  analyses
• All patients with fasting glucose gt 126 mg/dL x2
  or random glucose gt 140 mg/dL x2

9
Glucometrics II Which Readings?

• Which to include? Options
• All values
• Day of procedure (or ICU admission) and next 2
  days
• First 10-14 days of ward admission
• 4 standard POC readings on ward
• Which to exclude? Options
• First hospital day?
• Hypoglycemic values when calculating mean?
• Readings within 1 hour of previous reading?

10
Glucometrics II Which Readings?

• Non Critical Care Units
• First pass all POC readings
• Second pass
• Sensitivity analysis exclude readings on
  hospital day 1 and hospital days beyond day 10
• Sensitivity analysis exclude readings taken
  within 1 hour of a previous reading

11
Glucometrics II Which Readings?

• Critical Care Units
• All POC and other glucose values used to guide
  care

12
Glucometrics III Glucose Control

• What unit of analysis?
• Glucose reading
• Pros easy to calculate, most statistical power
• Cons least clinically relevant, skewed data
• Patient Stay
• Pros most clinically relevant
• Cons skewed data by LOS, uneven testing
• Patient-Day
• Pros least biased, good balance of other two
• Cons difficult to calculate and interpret

13
Glucometrics III Glucose Control

• What measures of control?
• Rates of hyper- or hypo-glycemia
• Percent readings within range
• Mean glucose value (/- exclusion of low values)
• Percent of mean values within range
• Rates of being in control (all values within
  range)
• Pros and Cons for each
• Best approach is to choose a few

14
Glucometrics III Glucose Control

• Special Measures
• Patient-Day Weighted Mean
• Three-Day Blood Glucose Average (3BG)
• Hyperglycemic Index
• Measures by hospital day (HD)

15
Glucometrics III Glucose Control

• Establishing definitions
• Hypoglycemia and extreme hypoglycemia
• Hyperglycemia and extreme hyperglycemia
• Ideal glucose range
• Depends on
• Critical Care vs. Not
• Mean vs. All Values
• What will gain consensus at your site
• Consider tightening definitions over time

16
Glucometrics III Glucose Control

• Non Critical Care Units
• First Pass
• Percent patient-days with mean glucose 80-140
  mg/dL
• Percent patient-days with all values 70-180 mg/dL
• Percent patient stays with mean glucose 80-140
  mg/dL

17
Glucometrics III Glucose Control

• Second pass all first-pass analyses plus
• Patient-Day weighted mean glucose
• Mean percent glucose readings per patient that
  are 70-180 mg/dL
• Percent patients with mean glucose 80-140 mg/dL
  by Hospital Day (days 1-7)

18
Glucometrics III Glucose Control

• Critical Care Units
• First pass
• Percent glucose readings lt110 mg/dL, lt140 mg/dL
• Percent patient-days with mean glucose lt 110
  mg/dL, lt 140 mg/dL
• Percent patient-days with all values lt 110 mg/dL,
  lt 140 mg/dL

19
Glucometrics III Glucose Control

• Critical Care Units
• First pass
• 3-day blood glucose average (3BG) for selected
  periop patients
• Percent patients with 3BG lt 110, lt 140 mg/dL
• Mean time to reach glycemic target on insulin
  infusion

20
Glucometrics III Glucose Control

• Critical Care Units
• Second pass
• 3BG as above for all patients in critical care
• Hyperglycemic index for all patients in critical
  care (AUC of glucose values above target)

21
Patient Safety

• Patient-day best measure
• Same for critical care and non critical care
• Percent patient-days with any value
• lt 40 mg/dL (extreme hypoglycemia)
• lt 70 mg/dL (hypoglycemia)
• gt 300 mg/dL (extreme hyperglycemia)

22
Processes of Care Insulin Ordering

• More sensitive to change than outcomes
• Failure to change process ? change the
  intervention
• Selective non-adherence with process
• Improve implementation
• Change protocol
• Accept certain degree to non-adherence

23
Processes of Care Insulin Ordering

• Non Critical Care Units
• First pass
• Among all patients on insulin, percent on basal
• Second pass
• Among patients with hyperglycemia (e.g., 2
  readings gt 180 mg/dL), percent on basal
• Among patients with hyperglycemia and eating,
  percent on basal and nutritional
• Percent patient days when changes made to insulin
  regimen if hypo- or hyperglycemia the day before

24
Processes of Care Insulin Ordering

• Critical Care Units
• First pass
• Percent of patients with 2 or more glucose
  readings gt 140 mg/dL on insulin infusion protocol

25
Processes of Care Other Measures

• Glucose measured within 8 hours of admission
• Hgb A1c measured if not available within 30 days
  of admission
• POC glucose testing for all patients with
  hyperglycemia or diabetes
• Measures of adherence to protocol components
• Use of sliding scale insulin alone
• Use of diabetes order sets
• Use of oral diabetic agents (esp. if
  contraindicated)
• Coordination of POC glucose testing, insulin
  administration, and food delivery
• Patient, physician, and nurse attitudes,
  education, and satisfaction

26
Analyses

• RCT
• Pros easy to interpret and publish results
• Cons doesnt account for CQI methods
• Before-After analysis
• Pros easy to interpret results, may be
  publishable
• Cons doesnt account for CQI methods, difficulty
  adjusting for temporal trends and cointerventions
• Run charts
• Pros best for CQI
• Cons statistics harder to interpret
• Run charts with SPC
• Best of both worlds when using CQI

27
Algorithm
Baseline
Order set
1st order set
28
Discussion

• Experiences at your institutions
• Questions
• Concerns
• Comments
• Thank you!