27  Correlation and Drivers

27.1 Why this matters

Identifying drivers of turnaround (volume, slide count, scanner, day) helps target interventions that move the needle most.

27.2 Simple Correlations

                 total_turnaround      slides scan_duration lab_to_scan
total_turnaround       1.00000000  0.05020739   -0.11626867  0.71990169
slides                 0.05020739  1.00000000    0.93323388  0.14651723
scan_duration         -0.11626867  0.93323388    1.00000000 -0.06959852
lab_to_scan            0.71990169  0.14651723   -0.06959852  1.00000000
scan_to_pacs           0.37410689 -0.17209285   -0.10824233 -0.37355506
                 scan_to_pacs
total_turnaround    0.3741069
slides             -0.1720929
scan_duration      -0.1082423
lab_to_scan        -0.3735551
scan_to_pacs        1.0000000

27.3 Lightweight Regression

We fit a simple linear model to see relative effect sizes. Categorical predictors use the first level as reference.


Call:
lm(formula = time_total_turnaround ~ slides + scan_duration + 
    lab_to_scan + scan_to_pacs + weekday + scanner, data = model_df)

Residuals:
     16496      16560      16597      16616      16618      16665      18676 
-1.031e-17  2.767e-17 -3.823e-15  3.692e-15  1.130e-16 -1.040e-17 -1.031e-17 
     18727      18768 
 2.071e-17  6.796e-31 

Coefficients:
                 Estimate Std. Error    t value Pr(>|t|)    
(Intercept)     2.805e-14  6.128e-14  4.580e-01    0.727    
slides         -9.169e-15  1.625e-14 -5.640e-01    0.673    
scan_duration   1.000e+00  3.032e-13  3.298e+12 1.93e-13 ***
lab_to_scan     1.000e+00  8.531e-15  1.172e+14 5.43e-15 ***
scan_to_pacs    1.000e+00  1.010e-14  9.901e+13 6.43e-15 ***
weekday.L      -1.685e-15  1.456e-13 -1.200e-02    0.993    
weekday.Q       2.773e-14  5.988e-14  4.630e-01    0.724    
scannerSS45328 -1.824e-14  3.481e-14 -5.240e-01    0.693    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 5.316e-15 on 1 degrees of freedom
  (24876 observations deleted due to missingness)
Multiple R-squared:      1, Adjusted R-squared:      1 
F-statistic: 2.581e+30 on 7 and 1 DF,  p-value: 4.793e-16

27.4 Interpretation Guide

  • Strong positive correlations with lab_to_scan or scan_to_pacs highlight where delays accrue; focus there first.
  • Regression coefficients for scanner or weekday flag specific devices or days that systematically shift turnaround.
  • If slides has a small coefficient, process overhead dominates complexity; streamline handoffs and loading rather than adding capacity.