Skip to main content

Advertisement

Table 2 Associations between alert acceptance and clinician characteristics and workload

From: Effects of workload, work complexity, and repeated alerts on alert fatigue in a clinical decision support system

Predictor Median value of continuous predictors Acceptance rates (yes vs. no or above vs. below median) Bivariate models Multivariable model
    IRR (95% CI) P IRR (95% CI) p
 For best practice advisories (BPAs)
  Clinician characteristics
   Nurse practitioner (vs physician) -- 22.7% vs 21.0% 1.08 (0.74, 1.57) .69 0.79 (0.57-1.10) .16
   Attested for meaningful use (vs not attested) -- 22.9% vs 19.3% 1.19 (0.94, 1.51) .15 1.06 (0.88-1.28) .50
   Female (vs male) -- 23.0% vs 18.4% 1.25 (0.98, 1.59) .07 1.06 (0.88-1.27) .54
  Markers of workload
   Patients per year 790 20.7% vs 21.7% 1.002 (0.98, 1.02)c .85 ---f  
   Encounters per year 1357 21.8% vs 20.6% 0.999 (0.99, 1.01)d .92 1.01 (1.00-1.02) .21
  Markers of complexity
   BPAs received per encounter 3.2 15.8% vs 26.7% 0.65 (0.55, 0.77)a <.001 0.70 (0.60-0.82) <.001
   Drug alerts received per encounter 0.7 22.7% vs 19.8% 1.17 (0.80, 1.72)a .41 1.91 (1.39-2.64) <.001
   Average ADG comorbidity count of patients 0.3 18.9% vs 23.5% 0.27 (0.10, 0.71)b .008 0.49 (0.23-1.03) .06
   Percent of BPAs that are within-patient repeats 26.2 17.0% vs 25.4% 0.90 (0.87, 0.94)e <.001 0.90 (0.86-0.95) <.001
 For drug alerts (DDI and DAI)
  Clinician characteristics
   Nurse practitioner (vs physician) -- 0.9% vs 0.2% 3.97 (1.64, 9.62) .002 4.56 (1.72-12.06) .002
   Attested for meaningful use (vs not attested) -- 0.3% vs 0.3% 0.92 (0.40, 2.11) .85 0.96 (0.49-1.88) .91
   Female (vs male) -- 0.4% vs 0.2% 1.72 (0.82, 3.60) .15 0.85 (0.37-1.95) .71
  Markers of workload
   Patients per year 790 0.3% vs 0.3% 0.99 (0.94, 1.04)c .60 -----f  
   Encounters per year 1357 0.3% vs 0.4% 0.98 (0.96, 1.01)d .12 1.00 (0.97-1.04) .82
  Markers of complexity
   BPAs received per encounter 3.2 0.3% vs 0.3% 1.10 (0.78, 1.56)a .59 1.49 (1.03-2.17) .04
   Drug alerts received per encounter 0.7 0.2% vs 0.4% 0.69 (0.29, 1.67)a .42 1.39 (0.44-4.39) .57
   Average ADG comorbidity count of patients 0.3 0.3% vs 0.4% 0.15 (0.03, 0.80)b .03 0.37 (0.05-2.74) .33
   Percent of alerts that are within-patient repeats 31.8 0.3% vs 0.4% 0.87 (0.76, 0.998)e .046 0.84 (0.68-1.03) .09
   Order sets used per encounter 1.1 0.3% vs 0.3% 1.20 (0.44, 3.23)a .72 1.68 (0.76-3.71) .20
   Labs ordered per encounter 0.9 0.3% vs 0.3% 0.89 (0.43, 1.83)a .75 0.80 (0.41-1.56) .51
  1. a. IRR computed per 1 additional unit
  2. b. IRR computed per 1 additional ADG
  3. c. IRR computed per 100 additional patients
  4. d. IRR computed per 100 additional encounters per year
  5. e. IRR computed per 5 percentage point increment
  6. f. Patients per year omitted from multivariable model to avoid collinearity with encounters per year
  7. NOTE: An IRR (incident rate ratio) is the ratio of 2 rates and is interpreted similarly to a relative risk or odds ratio. For example, the adjusted IRR of 4.56 above indicates that nurse practitioners are more than 4 times as likely to accept drug alerts as physicians are, in a model that controlled for patient comorbidity and proportion of repeated alerts