Supplemental Digital Content
1. CPT and ICD-9 codes used in study
2. Hierarchical model development and calculation of reliability
3. Table of fixed effect odds ratios and between-hospital variance for each model
Supplemental Digital Content 1 - CPT and ICD-9 codes used in study Colorectal procedures were identified using the following CPT codes:
44140, 44141, 44143, 44144, 44145, 44146, 44147, 44150, 44151, 44155, 44156, 44157, 44158, 44160, 44204, 44205, 44206, 44207, 44208, 44210, 44211, 44212, 45110, 45111, 45112, 45113, 45114, 45116, 45119, 45120, 45121, 45123, 45126, 45130, 45135, 45160, 45395, 45397, 45402, 45550
Postoperative diagnosis groups included the following ICD-9 codes:
Diagnosis group ICD-9 codes
Diverticulitis 562.11, 562.13, 562.01, 562.03
Diverticulosis 562.10, 562.12, 562.1, 562, 562.00, 562.02
Fistula 596.1, 569.81, 619.1, 998.6, 565.1, 537.4, 619,
599.1, 619.8
Hemorrhage 578.9, 578.1, 569.85, 569.3, 578, 532.40, 534.40,
569.86
Infectious enteritis/colitis/peritonitis 003.0, 008.4, 008.45, 009.0, 009.1, 569.5, 540.1, 567.22, 540.0, 540.9, 567.9, 567.21, 567.29, 540, 542, 567.89, 541, 567, 567.31, 569.61
Inflammatory bowel disease (Ulcerative colitis and Crohn’s disease)
555, 555.0, 555.1, 555.2, 555.9, 556, 556.0, 556.1, 556.2, 556.3, 556.4, 556.5, 556.6, 556.8, 556.9, 558, 558.1, 558.2, 558.3, 558.9, 567.82
Benign neoplasm 211.3, 235.2, 211.4, 230.3, 239.0, 569.0, 230.4,
238.1, 211.2, 239, 209.57, 215.5, 235.5, 229.8, 230.5, 209.52, 209.60, 211.1, 211.8, 227.0, 228.1, 229.9, 235.3, 238.79, 238.8, 239.7, 239.8, 209.50, 209.53, 209.69, 211.5, 211.7, 211.9, 214.2, 215.6, 217, 230, 235.4, 236.3, 238.9, 214.3, 209.51, 209.43, V12.72, 214.8
Malignant neoplasm 154.1, 153.6, 153.3, 153.4, 153.9, 154.0, 153.1, 153.0, 153.2, 153, 153.7, 153.8, 154, 154.8, 197.5, 153.5, 197.6, 209.03, 154.3, 154.2, 152.2, 197.4, 209.11, 158.0, 202.83, 198.89, 183.0, 202.80, 209.12, 152.9, 171.5, 151.8, 157.2, 197.7, 199.0, 199.1, 209.13, 259.2, 158.8, 196.2, 209.16, 151.9, 157.0, 158.9, 195.3, 198.6, 209.17, 152.8, 156.0, 157, 159.0, 171.9, 174.9, 182.0, 195.2, 200.70, 209.00, 209.15, 150.9, 151.2, 151.3, 151.4, 152.0, 156.1, 157.1, 158, 159.9, 170.6, 171.6, 179, 184.4, 185, 188.9, 189.0, 194.0, 197.8, 198.5, 200.10, 202.03, 203.80, 209, 209.10, 209.29, 209.72, 150.2, 151, 151.6, 152, 152.1, 155, 155.0, 157.9, 161.1, 162.3, 162.9, 172.5, 173.9, 174, 182, 188, 188.8, 193, 195.8, 196.5, 198.82, 200.03, 200.2, 200.23, 200.33, 200.43, 200.7, 202.13, 202.8, 202.87, 202.93, 209.0, 209.01, 209.14, 209.27, 209.3, 209.30, 209.79
Obstruction/perforation 560.0, 560.8, 560.81, 560.89, 560.9, 569.83, 560.2, 552.21, 560, 550.10, 552.8, 552.1, 552.2, 550.0, 551.21, 552.29, 532.50, 537.3, 550.00, 550.11, 551.2, 552.00, 552.20, 532.10, 551.1, 551.8, 552.3, 552.9, 532.11
Rectal prolapse 569.1
Vascular insufficiency 557.0, 557.1, 557.9, 557
Supplemental Digital Content 2 – Hierarchical model development and calculation of reliability John L. Adams, Ph.D.
We developed hierarchical multivariate logistic models for superficial SSI, deep/organ-space SSI, and “any SSI,” with hospital entered as a normally distributed random intercept:
ln ( 1− p
ijp
ij) =β0i+ β
1⃗ X
ij
Where
p
ij is the probability that person j in hospital i will have an infection.X ⃗
ij is a vector of person level covariates.β
1 is a vector of fixed regression coefficients corresponding to the person level covariates inX ⃗
ij (e.g. demographics).β
0i is a hospital level random intercept:β
0i~ Normal( β
0,τ
2log−odds)
Where
β
0 is the mean of the distribution of hospital intercepts andτ
2log−odds is the hospital-to- hospital random effect in the log odds scale.τ
2log−odds is a quantity of fundamental interest in reliability calculationsThe hospital-to-hospital random effect is in the log odds scale. There is a choice to be made whether to calculate the reliability in the log odds or the probability scale. Both approaches have
advantages. The log odds scale is the natural scale for logistic regression. The probability scale is the scale where the hospital profiles will most likely be reported. In the delta calculation reported below you can see that the reliability will be roughly the same if
τ
2 isn’t so large that it spans substantial non-linearity in the logistic transform. We pursue reliability in the probability scale here since it is a better fit to the hospital profiling policy problem.The basic definition of reliability is:
Reliability=
τ
2τ
2+ σ
i2Where
σ
i2 is the hospital specific sampling error variance for the infection rate. The binomial distribution used to represent the number of infections has a mean of^ p
in
i and a variance of^ p
i( 1− ^ p
i) n
i where^ p
i is the proportion of cases at hospital i that have an infection. The rate rather than the total number of infections is the quantity of interest here so we will use the variance of the rate estimate, the squared standard error of the rate estimate:σ ^
i2= ^ p
i( 1 − ^ p
i) /n
is
Since our estimate is in the log odds scale we need to apply a delta method calculation to transform
τ
2log−odds to the probability scale. The expit function (p= e
x1 +e
x ) transforms from the log odds scale to the probability scale. We need the derivative of this function which is p*(1-p). Since this is the multiplier to transformτ
log−odds we must square it to apply toτ
2log−odds so our estimate of τ2 squared will be:τ
2= p
2∗( 1−p )
2τ
log−odds2Where p is the probability where the transformation occurs. For p we use the infection rate for the whole dataset. Note that the inverse of this transformation could be applied to the hospital specific sampling variances to develop a reliability estimate in the log odds scale.
Supplemental Digital Content 3:
Table of fixed effect odds ratios and between-hospital variance for each model Superficial SSI Model Deep/Organ-space SSI
Model “Any SSI” Model
Odds Ratio p-value Odds
Ratio p-value Odds Ratio p-value
Intercept 0.041 <0.001 0.037 <0.001 0.083 <0.001
Procedure: (vs. Open colectomy) Ref Ref ref
Laparoscopic colectomy 0.582 <0.001 0.613 <0.001 0.589 <0.001
Open proctectomy 1.207 0.059 1.290 0.014 1.265 0.002
Laparoscopic proctectomy 0.639 0.096 0.962 0.868 0.766 0.153
Diagnosis: (vs. Benign Neoplasm) Ref Ref ref
Malignant neoplasm 1.295 0.007 1.321 0.022 1.321 <0.001
Diverticulitis 1.686 <0.001 1.137 0.344 1.467 <0.001
Diverticulosis 1.634 0.003 1.153 0.518 1.462 0.007
Ulcerative colitis/Crohns 1.323 0.042 1.986 <0.001 1.631 <0.001
Obstruction/perforation 1.649 <0.001 1.549 0.003 1.654 <0.001
Vascular insufficiency 0.968 0.887 1.101 0.672 1.037 0.827
Fistula 1.518 0.016 2.018 <0.001 1.756 <0.001
Rectal prolapse 0.157 <0.001 0.599 0.081 0.320 <0.001
Infectious colitis/peritonitis 1.184 0.418 1.338 0.163 1.254 0.141
Hemorrhage 3.000 <0.001 2.059 0.027 2.875 <0.001
Other 1.125 0.380 1.751 <0.001 1.413 0.001
Age: (vs. <55 years) Ref Ref Ref
55-64 1.087 0.224 0.759 <0.001 0.933 0.196
65-74 1.052 0.495 0.639 <0.001 0.834 0.002
75-84 0.916 0.318 0.582 <0.001 0.736 <0.001
>84 0.795 0.075 0.480 <0.001 0.622 <0.001
Male 0.982 0.715 1.201 0.001 1.079 0.051
Race: (vs. White) Ref Ref Ref
Black 0.933 0.439 0.921 0.399 0.924 0.256
Other 1.052 0.570 1.375 <0.001 1.213 0.004
Hispanic (vs. Non-Hispanic) 1.134 0.285 0.928 0.583 1.055 0.568
Admission source (vs. Home) Ref Ref Ref
Acute care hospital 0.783 0.124 0.969 0.829 0.858 0.179
Chronic care facility 1.062 0.751 0.800 0.288 0.942 0.682
Other 1.417 0.245 0.944 0.878 1.286 0.304
BMI (vs. Normal 18.5-24.9) Ref Ref Ref
Underweight (<18.5) 0.883 0.452 1.057 0.699 0.975 0.823
Overweight (25-29.9) 1.429 <0.001 1.017 0.807 1.234 <0.001
Class I obesity (30-34.9) 1.741 <0.001 1.197 0.026 1.509 <0.001
Class II obesity (35-39.9) 2.205 <0.001 1.307 0.013 1.851 <0.001
Class III obesity (40) 2.484 <0.001 1.527 <0.001 2.114 <0.001
Functional status (vs. Independent) Ref Ref Ref
Partially dependent 1.088 0.402 1.370 0.002 1.253 0.003
Totally dependent 0.898 0.555 0.959 0.812 0.947 0.679
Smoker 1.069 0.292 1.143 0.047 1.109 0.035
Alcohol use 1.253 0.061 1.338 0.021 1.313 0.004
Renal Failure 0.687 0.067 1.200 0.278 0.951 0.711
Diabetes (vs. None) Ref Ref Ref
Oral medication 1.081 0.314 0.865 0.134 0.987 0.832
Insulin 1.061 0.581 0.846 0.186 0.945 0.514
Dyspnea (vs. None) Ref Ref Ref
Moderate exertion 1.165 0.061 1.032 0.747 1.132 0.060
At rest 1.142 0.466 0.974 0.884 1.024 0.861
COPD 1.163 0.129 1.102 0.390 1.146 0.085
Pneumonia 0.707 0.287 0.747 0.279 0.742 0.165
Steroid Use 0.926 0.459 1.198 0.062 1.075 0.337
Bleeding disorder 1.008 0.939 0.998 0.988 0.993 0.936
Congestive heart failure 0.954 0.834 0.586 0.062 0.771 0.156
Hypertension 1.003 0.957 1.086 0.195 1.033 0.460
Myocardial infarction 0.791 0.403 1.683 0.022 1.158 0.437
Disseminated cancer 0.953 0.694 1.510 <0.001 1.219 0.025
Weight loss 0.850 0.161 0.986 0.901 0.914 0.288
Chemotherapy 0.733 0.140 0.993 0.969 0.884 0.380
Radiotherapy 0.898 0.425 1.434 0.005 1.101 0.332
Preop blood transfusion 0.877 0.637 1.354 0.252 1.097 0.650
Preop Sepsis (vs. None/SIRS) Ref Ref Ref
Sepsis 0.687 0.003 1.351 0.008 1.000 0.998
Septic shock 0.546 0.008 1.201 0.336 0.852 0.294
ASA class (vs. I and II) Ref Ref Ref
Class III 1.254 <0.001 1.339 <0.001 1.311 <0.001
Class IV and V 1.080 0.502 1.115 0.374 1.084 0.364
Wound class (vs. II) Ref Ref Ref
Class III 1.204 0.015 1.306 0.001 1.268 <0.001
Class IV 1.060 0.521 1.420 <0.001 1.221 0.004
Emergency procedure 0.818 0.024 0.819 0.032 0.793 0.001
Between-hospital variance
(probability scale) 1.598 x 10-3 3.997 x
10-4 1.943 x 10-3
Between-hospital variance
(log odds scale) 0.312 0.128 0.145