Radiologist Utilization of the EMR

Author

Lu Mao

Statistical analysis

Placeholder for description of methods.

User characteristics and user-level analysis

We first aggregate data at the user level. For each user, the number of months during which they have been using the system is defined as the difference between the latest and earliest dates of their records (ual_date). The total number of hours active is the sum of the active times (number_of_seconds_active), overall and by category (ual_category). Finally, the overall and category-specific hours active per month is the total number of hours divided by the months of usage.

By training status

Table 1:

User characteristics and EMR usage by training status.
Characteristic Physician, N = 1431 Physician Assistant, N = 51 Nurse Practitioner, N = 41 p-value2
Years at UW 3 (1, 13) 7 (4, 9) 6 (5, 7) 0.8
    Unknown 1 0 0
Sex


0.005
    Female 43 (30%) 3 (60%) 4 (100%)
    Male 99 (70%) 2 (40%) 0 (0%)
    Unknown 1 0 0
Months of usage 21.2 (8.6, 21.6) 21.5 (21.4, 21.5) 21.6 (16.5, 21.6) 0.7
Monthly hours active - Overall 6 (3, 12) 33 (18, 49) 39 (34, 52) <0.001
Monthly hours active - Clinical Review 2.43 (1.22, 4.79) 7.37 (7.11, 11.18) 9.01 (8.60, 12.66) <0.001
Monthly hours active - In Basket 0.24 (0.11, 0.48) 3.46 (0.84, 3.95) 1.50 (0.80, 3.74) 0.017
Monthly hours active - Notes/Letters 0.1 (0.0, 0.4) 7.3 (1.0, 11.8) 12.0 (9.5, 14.2) <0.001
Monthly hours active - Orders 0.04 (0.00, 0.20) 1.91 (0.37, 4.75) 4.62 (2.63, 6.44) <0.001
Monthly hours active - Schedule/Patient Lists 0.51 (0.13, 1.30) 4.93 (2.85, 5.46) 5.33 (4.04, 10.00) <0.001
Monthly hours active - Visit Navigator 0.01 (0.00, 0.05) 1.54 (0.82, 1.62) 1.24 (0.94, 1.83) <0.001
Monthly hours active - Other/Unmapped 1.49 (0.66, 3.01) 4.01 (2.75, 5.21) 4.66 (4.32, 6.43) 0.002
1 Median (IQR); n (%)
2 Kruskal-Wallis rank sum test; Fisher’s exact test

Comments:

  1. User ID 29 misses years at UW; ID 34 misses sex.
  2. Physicians spend far less time using EMR than non-physicians, both overall and for each category.
  3. NP generally spends more time than PA, largely due to longer time in Notes/Letters and in Orders (despite shorter time in Basket).

The next two sub-sections focus on physicians only.

By physician sex

There is little difference between the sexes.

Table 2:

Physician characteristics and EMR usage by sex.
Characteristic Female, N = 431 Male, N = 991 p-value2
Years at UW 4 (0, 12) 3 (1, 13) 0.4
    Unknown 1 0
Months of usage 19.8 (10.5, 21.6) 21.4 (7.3, 21.6) >0.9
Monthly hours active - Overall 7 (4, 13) 6 (3, 11) 0.4
Monthly hours active - Clinical Review 2.70 (1.47, 4.39) 2.42 (1.20, 5.00) >0.9
Monthly hours active - In Basket 0.17 (0.07, 0.69) 0.26 (0.12, 0.44) 0.6
Monthly hours active - Notes/Letters 0.08 (0.00, 1.04) 0.05 (0.00, 0.33) 0.2
Monthly hours active - Orders 0.01 (0.00, 0.16) 0.04 (0.00, 0.20) 0.5
Monthly hours active - Schedule/Patient Lists 0.44 (0.13, 1.12) 0.53 (0.11, 1.39) 0.7
Monthly hours active - Visit Navigator 0.01 (0.00, 0.06) 0.01 (0.00, 0.04) 0.6
Monthly hours active - Other/Unmapped 2.14 (0.87, 3.59) 1.35 (0.58, 2.89) 0.3
1 Median (IQR)
2 Wilcoxon rank sum test

By physician years at UW

Table 3:

Physician characteristics and EMR usage by years at UW.
Characteristic 0 - 5, N = 821 5 - 15, N = 371 15+, N = 231 p-value2
Sex


0.6
    Female 25 (30%) 12 (33%) 5 (22%)
    Male 57 (70%) 24 (67%) 18 (78%)
    Unknown 0 1 0
Months of usage 11.9 (6.0, 21.5) 21.6 (21.5, 21.6) 21.6 (21.2, 21.6) <0.001
Monthly hours active - Overall 9 (5, 13) 5 (2, 7) 4 (2, 5) <0.001
Monthly hours active - Clinical Review 3.81 (1.95, 5.87) 1.83 (1.04, 3.08) 1.57 (1.06, 2.07) <0.001
Monthly hours active - In Basket 0.17 (0.09, 0.35) 0.35 (0.12, 0.78) 0.34 (0.23, 0.54) 0.007
Monthly hours active - Notes/Letters 0.11 (0.01, 0.43) 0.02 (0.00, 0.57) 0.03 (0.00, 0.15) 0.2
Monthly hours active - Orders 0.09 (0.01, 0.25) 0.01 (0.00, 0.07) 0.02 (0.00, 0.05) 0.004
Monthly hours active - Schedule/Patient Lists 0.74 (0.28, 1.57) 0.28 (0.06, 0.71) 0.11 (0.02, 0.53) <0.001
Monthly hours active - Visit Navigator 0.02 (0.00, 0.05) 0.00 (0.00, 0.03) 0.00 (0.00, 0.02) 0.11
Monthly hours active - Other/Unmapped 2.47 (1.15, 3.85) 0.92 (0.44, 1.76) 0.65 (0.32, 1.33) <0.001
1 n (%); Median (IQR)
2 Pearson’s Chi-squared test; Kruskal-Wallis rank sum test

Comments:

  1. New hires (< 5 years at UW) tend to spend more time, mainly driven by clinical review, schedule/patient lists, and other/unmapped.
    • Unclear whether it’s cohort or temporal effect (will be addressed in next section).

Figure 1: Overall and category-specific monthly hours active by physician years at UW.

Temporal analysis

In this section we map out the monthly hours active over time.

By training status

The temporal trend of overall monthly hours active by training status is plotted in Figure 2, and decomposed by category in Figure 3.

Figure 2: Overall median monthly hours active (IQR shaded) over time by training status.

Figure 3: Category-specific median monthly hours active over time by training status.

Comments:

  1. PA and NP spend consistently and substantially more time on EHR than physicians do.
  2. There is no obvious time trend for physicians.

By physician sex

The temporal trend of overall monthly hours active by physician sex is plotted in Figure 4, and decomposed by category in Figure 5.

Figure 4: Overall median monthly hours active (IQR shaded) over time by physician sex.

Figure 5: Category-specific median monthly hours active over time by physician sex.

Comments:

  1. Females tend to spend slightly more time than males, particularly on Notes/Letters.
  2. There is an increase in Other/Unmapped and a decrease in Clinical Review in later months.

By physician years at UW

The temporal trend of overall monthly hours active by physician years at UW is plotted in Figure 6, and decomposed by category in Figure 7.

Figure 6: Overall median monthly hours active (IQR shaded) over time by physician years at UW.

Figure 7: Category-specific median monthly hours active over time by physician years at UW .

Comments:

  1. Recent hires (\(<\) 5 yrs) tend to spend more time consistently than old hires (cohort effect, not temporal effect);
  2. In all groups there is an increase in Other/Unmapped and decrease in Clinical Review in recent months.

Physician seniority and division

Data in this section are restricted to 2021-12-01 – 2023-04-30.

Physician seniority

Frequency by training status and period
Characteristic Before 2022-6-30, N = 1061 After 2022-7-1, N = 1131
Training

    Resident 10 (9.8%) 6 (5.4%)
    Fellow 11 (11%) 17 (15%)
    Attending 81 (79%) 89 (79%)
1 n (%)

The temporal trend of overall monthly hours active by physician seniority is plotted below.

(Exploratory) Overall median monthly hours active (IQR shaded) over time by physician seniority.

Resident has too few data points. Combine it with fellow (Figure 8).

Figure 8: Overall median monthly hours active (IQR shaded) over time by physician seniority.

Summary statistics and formal tests on training status by two periods are tabulated in Table 4. Residents/Fellows spend substantially and significantly more time than attendings.

Table 4:

Summary statistics of average monthly hours active by physician seniority and period.
Period Resident/Fellow, N = 361 Attending, N = 901 p-value2
Before 2022-6-30 11.2 (8.1, 13.3) 4.4 (2.5, 6.7) <0.001
After 2022-7-1 12.6 (9.9, 15.0) 4.2 (2.0, 6.7) <0.001
Whole period 12.0 (9.6, 13.6) 4.2 (2.1, 6.7) <0.001
1 Median (IQR)
2 Wilcoxon rank sum test

Physician divison

Frequency by division and period
Characteristic Before 2022-6-30, N = 1061 After 2022-7-1, N = 1131
Division

    IR/NeuroIR 13 (13%) 14 (13%)
    MRI/Abdominal 22 (22%) 24 (21%)
    Neuro 11 (11%) 13 (12%)
    MSK 13 (13%) 13 (12%)
    Peds 5 (4.9%) 8 (7.1%)
    Chest/CV 11 (11%) 12 (11%)
    Trainee 9 (8.8%) 7 (6.3%)
    Breast 8 (7.8%) 10 (8.9%)
    Community 6 (5.9%) 8 (7.1%)
    Nucs 4 (3.9%) 3 (2.7%)
1 n (%)

Because there is no switch in division, we focus on the whole period. Breast spends more time than other divisions (Figure 9 and Table 5).

Figure 9: Boxplots of whole-period average monthly hours active by physician division.

Table 5: Summary statistics of average monthly hours active by physician division.

Period Breast, N = 111 Chest/CV, N = 121 Community, N = 81 IR/NeuroIR, N = 141 MRI/Abdominal, N = 281 MSK, N = 181 Neuro, N = 141 Nucs, N = 41 Peds, N = 81 Trainee, N = 91 p-value2
Whole period 14.4 (12.7, 25.6) 3.4 (2.2, 6.7) 2.0 (1.8, 3.5) 7.0 (5.3, 10.9) 4.9 (3.0, 10.4) 10.8 (4.9, 13.3) 2.6 (1.2, 5.1) 5.1 (3.4, 6.5) 3.0 (1.8, 4.1) 8.5 (8.1, 9.7) <0.001
1 Median (IQR)
2 Kruskal-Wallis rank sum test
(a) Breast vs other
Period Breast, N = 111 Other, N = 1151 p-value2
Whole period 14.4 (12.7, 25.6) 5.3 (2.8, 9.5) <0.001
1 Median (IQR)
2 Wilcoxon rank sum test

The temporal trend of overall monthly hours active for by Breast vs other is plotted in Figure 10, and decomposed by category in Figure 11.

Figure 10: Overall median monthly hours active (IQR shaded) over time by physician division.

Figure 11: Category-specific median monthly hours active over time by physician division.
Table 6:

User characteristics and EMR usage by division.
Characteristic Breast, N = 111 Other, N = 1151 p-value2
Years at UW 5 (1, 12) 5 (2, 14) 0.7
    Unknown 1 0
Sex

0.004
    Female 8 (73%) 31 (27%)
    Male 3 (27%) 83 (73%)
    Unknown 0 1
Monthly hours active - Overall 14.4 (12.7, 25.6) 5.3 (2.8, 9.5) <0.001
Monthly hours active - Clinical Review 2.41 (2.10, 5.16) 2.75 (1.38, 5.08) 0.6
Monthly hours active - In Basket 0.43 (0.29, 0.83) 0.21 (0.09, 0.45) 0.023
Monthly hours active - Notes/Letters 9.00 (7.67, 15.29) 0.05 (0.01, 0.32) <0.001
Monthly hours active - Orders 0.00 (0.00, 0.02) 0.07 (0.01, 0.22) 0.006
Monthly hours active - Schedule/Patient Lists 1.38 (0.88, 2.00) 0.35 (0.07, 0.95) 0.001
Monthly hours active - Visit Navigator 0.00 (0.00, 0.01) 0.01 (0.01, 0.07) 0.006
Monthly hours active - Other/Unmapped 1.89 (1.12, 2.24) 0.84 (0.41, 1.91) 0.024
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Fisher’s exact test

Comment:

  1. The difference between breast vs other is driven by Notes/Letters.

Breast

Table 7:

User characteristics in Breast division and EMR usage by training status.
Characteristic Fellow, N = 41 Attending, N = 71 p-value2
Years at UW 1.0 (1.0, 1.0) 10.0 (4.5, 12.0) 0.021
    Unknown 1 0
Sex

>0.9
    Female 3 (75%) 5 (71%)
    Male 1 (25%) 2 (29%)
Monthly hours active - Overall 26 (24, 27) 13 (12, 14) 0.073
Monthly hours active - Clinical Review 5.16 (4.80, 5.54) 2.24 (1.55, 2.40) 0.073
Monthly hours active - In Basket 0.24 (0.19, 0.29) 0.80 (0.43, 0.99) 0.006
Monthly hours active - Notes/Letters 15.3 (13.8, 16.8) 7.9 (7.0, 8.7) 0.073
Monthly hours active - Orders 0.007 (0.003, 0.032) 0.004 (0.002, 0.019) >0.9
Monthly hours active - Schedule/Patient Lists 2.00 (1.87, 2.08) 1.09 (0.65, 1.24) 0.073
Monthly hours active - Other/Unmapped 2.10 (1.94, 2.42) 1.13 (1.11, 1.84) 0.2
Monthly hours active - Visit Navigator 0.001 (0.000, 0.004) 0.007 (0.000, 0.009) 0.5
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Fisher’s exact test; Wilcoxon rank sum exact test

The temporal trend of overall monthly hours active for Breast physicians by training status is plotted in Figure 12, and decomposed by category in Figure 13.

Figure 12: Overall monthly hours active (thick line, median) over time for Breast physicians by training status.

Comments:

  1. Generally attending << fellow. But there is an outlying attending (GEGIOS, ALISON R), which makes the test only borderline significant given the small sample size (Table 7).

Figure 13: Category-specific median monthly hours active over time for Breast physicians by training status.