Length of Stay and Imminent Discharge Probability Distributions from Multistage Models

Appendix - Page 1 of 37

LENGTH OF STAY AND IMMINENT DISCHARGE PROBABILITY DISTRIBUTIONS FROM MULTISTAGE MODELS: VARIATION BY DIAGNOSIS, SEVERITY OF ILLNESS, AND HOSPITAL

Section 1: Description of key predictors (Primary Condition, LAPS, COPS)

Section 2: Population description from our original report

Section 3: LOS probability distributions for all hospitalizations

Supplementary Figure 1a:

All patients surviving to discharge, stratified by mortality risk

Supplementary Figure 1b:

All patients surviving to discharge, stratified by admission physiologic derangement

Supplementary Figure 1c:

All patients surviving to discharge, stratified by admission pre-existing comorbidity burden

Section 4: LOS probability distributions for patients with pneumonia (Figures 2a - 2c) or appendicitis/cholecystitis (ICD codes 540-543, 550-553, 574-576; Figure 2d)

Supplementary Figure 2a:

All pneumonia patients surviving to discharge, stratified by mortality risk

Supplementary Figure 2b:

All pneumonia patients surviving to discharge, stratified by admission physiologic derangement

Supplementary Figure 2c:

All pneumonia patients surviving to discharge, stratified by admission pre-existing comorbidity

Supplementary Figure 2d:

All appendicitis/cholecystitis patients surviving to discharge, stratified by mortality risk

Section 5: Supplementary figure showing 48 hour incipient release probabilities for pneumonia patients by predicted mortality terciles

For reviewers and interested readers – to be made available via world wide web

Appendix - Page 1 of 37

SECTION 1, PART A:GROUPING OF INTERNATIONAL CLASSIFICATION OF DISEASES CODES INTO ADMIT DIAGNOSES

In order to have a manageable number of diagnostic categories for our regression models, we divided all 16,090 possible International Classification of Diseases codes, including the V and E codes, into 44 mutually exclusive Primary Conditions (every ICD code was assigned to one and only one category). The approach we followed is similar conceptually to that followed by Render et al. (see text for citations). However, we did not employ Render’s scheme because their groupings did not include all possible ICD codes and because the Veterans Administration population was primarily male. Consequently, we defined our own groupings, which were based on biological plausibility (we tried, insofar as possible, to group diseases with similar pathophysiology) as well as similar overall inpatient mortality and length of stay.

Table A-1 summarizes these 44 Primary Conditions and lists the specific ICD codes and/or ICD code ranges included in each group, the inpatient mortality rate (with 95% confidence interval for the Primary Condition), and the hospital length of stay (in hours) for survivors (median, mean  standard deviation). The only codes that are not included here are those for certain conditions originating in the perinatal period (ICD codes 760 through 779). Note that ICD codes for pregnancy are included. However, we excluded delivery records from our cohort, so the presence of these codes reflects the fact that women who were hospitalized either before or after their delivery hospitalization are included in our cohort.

TABLE A-1: GROUPING OF INTERNATIONAL CLASSIFICIATION OF DISEASES EMPLOYED IN STUDY
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
Congestive Heart Failure
010 CHF / Congestive heart failure & some related illnesses
Major codes are 425, 428, and miscellaneous (398.91, 402s, 422s, and some 429s, incl. ‘429’) / 5.53(5.13,5.94) / 75.90; 109.99 +/- 136.42
SEPSIS
20 SEPSIS / Sepsis, meningitis, septic shock, and major catastrophic infections (003.1, 003.21, 027.0, 036-038, 040, 320-326, 422.92, 728.86, 785.4, 785.59, 790.7, 995.92, 9993) / 15.95(15.01,16.93) / 118.80; 185.18 +/- 235.43
CATASTROPHIC CONDITIONS
030 CATAST / Catastrophic conditions, incl. dissecting aneurysms, cardiac arrest, respiratory arrest, all forms of shock except septic shock; intracranial & subdural hemorrhages (multiple ICD codes) / 24.68(23.58,25.81) / 115.80; 190.46 +/- 260.15
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
PNEUMONIA
40 PNEUM / All forms of pneumonia (480-487); empyema (510); pleurisy (511); and lung abscess (513); also includes pulmonary TB (011, 012.8);pulmonary congestion and hypostasis (514) / 9.56(9.14,9.99) / 92.10; 132.38 +/- 167.30
DKA & RELATED METABOLIC
50 METAB1 / Diabetic ketoacidosis, with and without coma; hypoglycemic coma; unspecified coma & alteration of consciousness
Misc. 250s, 251, 780.0x / 7.01(6.38,7.68) / 71.80; 119.80 +/- 238.30
INGESTIONS BENIGN TUMORS
66 OD&BNCA / Non-gynecologic benign neoplasms
210-217, 222-239, 610, 611
Drug overdoses, drug abuse, adverse drug reactions, and poisonings
291, 292, 303-305, 790.3, 796, 960-989, 995.2 / 0.86(0.66,1.10) / 48.70; 77.70 +/- 105.11
FLUID AND ELECTROLYTE
71 FL&ELEC / Typical fluid & electrolyte disorders & dehydration
275.2 – 276.9 / 5.77(5.22,6.35) / 69.40; 104.69 +/- 119.47
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
OTHER METABOLIC
72 METAB3 / All other endocrine, metabolic & miscellaneous immune disorders (but not including SLE or RA)
240-255, 257-272, 274-275.1, 277-279, misc. 790s / 0.79; (0.58,1.05) / 50.70; 77.13 +/- 126.07
URINARY TRACT INFECTIONS
80 UTI / Urinary tract infections, not including pregnancy related ones
590, 595, 597, 599, 601, 604, misc. 996s / 3.80; (3.39,4.25) / 81.00; 110.06 +/- 124.20
ALL OTHER INFECTIONS
90 INFEC4 / All other infections with the exception of hepatitis; unspecified fever
001-139, multiple others, incl. joint infections & muscle infections (711 & 728); 780.6 (fever) / 2.12; (1.92,2.34) / 84.30; 121.73 +/- 149.68
STROKE
110 STROKE / Stroke & post-stroke complications
434-438, 997.0x / 5.30; (4.86,5.76) / 61.20; 95.93 +/- 168.16
ACUTE MYOCARDIAL INFARCTION
121 AMI / Myocardial infarction
410-414 / 2.76; (2.57,2.96) / 43.40; 76.15 +/- 119.88
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
CHEST PAIN
122 ROAMI / Chest pain, MI not specified
Misc. 786.5X, V71.7 / 0.62; (0.53,0.73) / 28.50; 48.08 +/- 66.65
OTHER CARDIAC CONDITIONS
130 HEART2 / Diseases of pulmonary circulation & cardiac dysrhythmias
415-417, 426, 427, misc. 785s, misc. 996s / 2.54; (2.26,2.85) / 47.90; 77.52 +/- 96.50
GYNECOLOGY
140 GYNEC1 / Non-malignant, non-infectious gynecologic diseases, incl. benign neoplasms
Must be female patient
218-221, 256 & multiple miscellaneous codes (including V codes). / 0.04; (0.02,0.10) / 51.80; 55.96 +/- 41.94
ATHEROSCLEROSIS AND PVD
150 HEART4 / Atherosclerosis (including that affecting precerebral arteries) & other forms of peripheral vascular disease
429.2, 433, 440-459 / 3.42; (3.04,3.83) / 53.40; 98.23 +/- 129.06
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
OTHER RENAL
170 RENAL3 / All other renal diseases other than infections
Miscellaneous 405s, 591-608, misc. other codes / 0.68; (0.51,0.88) / 38.20; 63.04 +/- 75.29
GYNECOLOGIC CANCERS
180 GYNECA / Gynecologic malignancies other than ovarian cancer; female breast cancer
Must be female patient
174, 179-182, 184 / 1.64; (1.28,2.08) / 29.80; 53.07 +/- 55.02
PREGNANCY
190 PRGNCY / Pregnancy & related conditions
Must be female patient
630-677, V22 through V28 / 0 (no deaths occurred) / 23.90; 40.61 +/- 65.45
CANCER A
201 CANCRA / Malignant neoplasms of respiratory tract & intrathoracic organs; leukemias, non-Hodgkin’s lymphomas, & other histiocytic malignancies
160-165, 202-208 / 13.33; (12.12,14.62) / 122.80; 175.07 +/- 199.72
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
CANCER B
202 CANCRB / All other malignant neoplasms not in 202-208 or gynecologic ones (incl. Hodgkin’s disease); radiation therapy & chemotherapy encounters where CA not specified
140-159, 170-173, 175, 176, 185-195, 200, 201, V58.0, V58.1, V66.1, V66.2 / 3.99; (3.60,4.41) / 98.70; 131.75 +/- 172.51
OVARIAN AND METASTATIC CANCER
210 CANCRM / Ovarian cancer & metastatic cancer
183, 196-199 / 12.19; (10.74,13.76) / 99.80; 134.86 +/- 146.30
NON-MALIGNANT HEMATOLOGIC
230 HEMTOL / Hematologic problems other than malignancies
273, 280-289, misc 790s, 996.85 / 2.81; (2.41,3.25) / 56.00; 92.98 +/- 136.82
SEIZURES
240 SEIZURE / Seizure disorders
345, misc. 780.1-780.4 / 1.55; (1.33,1.79) / 42.10; 67.57 +/- 108.91
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
OTHER NEUROLOGICAL
251 NEUMENT / All other neurologic problems and mental disorders (other than drug overdoses); senility
290-319, 327-344, 346-389, 781, 797, V71.0 / 2.71; (2.31,3.17) / 65.60; 122.88 +/- 283.17
ACUTE RENAL FAILURE
270 RENAL1 / Acute renal failure, nephrotic syndrome, & related conditions
580, 581, 584 / 8.36; (7.34,9.49) / 96.10; 136.92 +/- 172.89
CHRONIC RENAL FAILURE
280 RENAL2 / Chronic renal failure, ESRD, & kidney transplants
582, 583, 585-589, 996.81, V42.0xx / 3.49; (2.73,4.40) / 48.10; 88.98 +/- 138.83
MISCELLANEOUS CARDIAC
290 MISCHRT / Miscellaneous cardiac conditions & congenital heart disease
392-405, 745-747 / 2.13; (1.54,2.88) / 49.20; 83.42 +/- 126.90
COPD
300 COPD / COPD & some less common respiratory conditions
490-496, 500-508, 512, 515, 517-519 / 3.70; (3.36,4.06) / 74.95; 103.40 +/- 116.37
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
ACUTE RESPIRATORY
320 RESPR4 / Acute respiratory infections & miscellaneous respiratory diseases
460-478, 786 / 6.19; (5.68,6.73) / 63.30; 95.83 +/- 124.34
HIP FRACTURE
350 HIPFX / Hip fracture
Some 733s, 808, 820, 821, some 905s, 959.6 / 2.81; (2.42,3.24) / 100.45; 120.66 +/- 97.13
ARTHROPATHIES
361 ARTHSPIN / Arthropathies and spine disorders (but no infections or autoimmune conditions)
712, 715-729, most 731-739 (except for 733.1xx, pathologic fracture) / 0.37; (0.30,0.45) / 78.60; 89.95 +/- 89.28
FRACTURES AND DISLOCATIONS
381 FXDISLC / All other fractures & dislocations, incl. pathologic fractures
733.1xx, 805-807, 809-819, 822-839, misc. 905, 907, 952 / 0.66; (0.49,0.87) / 45.10; 67.41 +/- 137.06
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
ALL OTHER TRAUMA
390 TRAUMA / Traumatic injuries not included elsewhere, including head injuries without intracranial or subdural bleeds
800-804, 840-848, 850-854, 860-904, most of 905-959 / 1.77; (1.38,2.23) / 36.65; 69.45 +/- 111.22
APPENDICITIS
411 APPCHOL / Appendicitis, hernias, cholecystitis, & cholangitis
540-543, 550-553, 574-576 / 0.55; (0.44,0.68) / 46.10; 71.64 +/- 89.05
PANCREATIC DISORDERS
440 PNCRDZ / Pancreatic disorders
577 / 1.75; (1.38,2.19) / 90.30; 138.24 +/- 207.62
GI IBD & OBSTRUCTION
451 GIOBSENT / Inflammatory bowel disease and malabsorption; GI obstruction; enteritides
555-558,560, 568, 579 / 3.82; (3.39,4.28) / 93.60; 143.65 +/- 171.51
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
GI BLEEDING
461 GIBLEED / GI hemorrhage; misc. disorders of stomach & duodenum; diverticulitis; abdominal symptoms, nausea with vomiting; blood in stool
530-537, 562, 564, 565, 569, 578, 787, 789, 792.1 / 2.86; (2.68,3.05) / 67.00; 102.16 +/- 133.76
LIVER DISORDERS
510 LIVERDZ / Liver disorders, including hepatitis
570-573 / 10.82; (9.40,12.36) / 67.70; 104.51 +/- 124.61
MISCELLANEOUS # 1
520 MISCL1 / Miscellaneous conditions not classified previously
990-999 / 1.07; (0.83,1.36) / 75.10; 111.53 +/- 156.85
MISCELLANEOUS # 2
531 MSC2&3 / Remaining V codes; remaining 790-796; all E codes. / 1.33; (0.99,1.74) / 25.55; 66.41 +/- 169.07
PERICARDITIS
550 PERVALV / Pericarditis & valvular heart disease
391, 423, 424 / 3.28; (2.32,4.50) / 124.05; 156.28 +/- 179.56
SKIN & AUTOIMMUNE DISORDERS
560 SKNAUT / SLE, rheumatoid arthritis, skin disorders, & related autoimmune diseases, sialoadenitis
690-710, 713, 714, 782 / 2.38; (1.85,3.01) / 79.35; 141.29 +/- 238.13
PRIMARY CONDITION / DESCRIPTION & INCLUDED ICD CODES / % IN-HOSPITAL MORTALITY RATE
POINT ESTIMATE (95% CI) / LENGTH OF STAY IN HOURS: MEDIAN, MEAN ± SD
MISCELLANEOUS # 3
591 MISCL5 / Miscellaneous non-cardiac congenital anomalies; miscellaneous symptoms other than fever; miscellaneous tooth & tongue disorders
520-529 (tooth & tongue disorders); 740-759, 780 (except for 780.6), 783-785 (if not found elsewhere) / 4.70; (4.38,5.03) / 63.00; 106.79 +/- 175.48

For reviewers and interested readers – to be made available via world wide web

Harrison – WEB APPENDIX

SECTION 1, PART B: DEVELOPMENT AND VALIDATION OF LABORATORY-BASED ACUTE PHYSIOLOGY SCORE (LAPS)

For the development of the LAPS, our cohort consisted of all 895,159 hospital stays that occurred in the 17 KPMCP Northern California Region hospitals where the patient's initial admission took place at a KPMCP hospital between 11/01/1999 and 06/30/2005. These individual stays were concatenated to create a dataset of 854,139 linked hospitalizations where the start date was when the patient was first admitted to the hospital and the end date was when the patient went home, was sent to a nursing home, or died. This dataset is larger than the one employed in our report because DxCG data did not become available to us until 2002.

The dataset of linked hospitalizations was split randomly into a derivation dataset of 427,070 hospitalizations and a validation dataset of 427,069 hospitalizations. The in-hospital mortality rate for the derivation dataset was 3.39% and that of the validation dataset was 3.43%.

We scanned the NC-KPMCP laboratory database and found all test results for the 13 laboratory tests employed to assign our laboratory-based acute physiology score (LAPS). We linked all test results obtained in the 24 hours preceding hospitalization to the patients in our cohort. To assign point values for the LAPS, we employed two sequential regression models. We employed this approach for two reasons: 1) a number of exploratory analyses we conducted suggested that, under certain conditions, the absence of certain laboratory tests was associated with increased mortality following hospital admission as well as increased rates of physiologic deterioration leading to transfer to the intensive care unit after initial admission to a general medical-surgical ward; and 2) the management of hospitalized patients in the NC-KPMCP falls under two departmental jurisdictions. Initial evaluation of hospitalized patients is done by either clinic or emergency department physicians, while the actual care of hospitalized patients is the responsibility of full time hospital-based specialists (hospitalists).

In the first regression model (Laboratory Score Preliminary Model), we employed a patient's age, admission type, ratio of blood urea nitrogen to serum creatinine, sodium, and ratio of the anion gap to the serum bicarbonate to predict in-hospital death. We then divided patients into two groups based on their predicted mortality: low (< 6%) or high ( 6%). In this model, if a laboratory test was missing, we imputed a normal value. When more than one laboratory test result was available, we employed the worst (most unphysiologic) test result. The Laboratory Score Preliminary Model had a c statistic of 0.79 in the validation dataset but had very poor calibration above a mortality risk of 25-30%. Its performance characteristics in the validation dataset with respect to the 6% vs. 6% distinction are shown in Table B-1. The two risk groupings were retained for imputation purposes for the final LAPS Model.

TABLE B-1: LABORATORY SCORE PRELIMINARY MODEL
Predicted mortality risk stratum / Number of patients / Number of inpatient deaths
Predicted / Observed
Low risk (< 6%) / 353,803 / 6,926 / 6,914
High risk ( 6%) / 73,266 / 7,736 / 7,748

For the final LAPS model, we employed the following laboratory test results to calculate an acute physiology score: serum albumin; arterial pH, PaCO2, and PaO2; total serum bilirubin; blood urea nitrogen; serum creatinine; serum glucose; hematocrit; serum sodium, serum troponin I, and total white blood cell count in thousands. In this model, patients who were classified as “high risk” using the first model had a different imputation approach for three test results (pH, total white blood cell count, and troponin I) than patients who were classified as “low risk.” For patients in the “high risk” group, missing data for these three laboratory test results had their own risk band, whereas for patients in the “low risk” group, missing data were imputed as being normal.

We employed the beta coefficients from this second regression model to assign point values to specific laboratory test results, and the final point scoring scheme is shown in Table B-2. The LAPS is thus a continuous variable that can range between a low of zero and a theoretical maximum of 256, although < 0.05% of patients in our cohort had LAPS exceeding 120 and none had a LAPS > 165. The final LAPS model had a c statistic of 0.73 in the validation dataset. The Figureshows the relationship between the LAPS and inpatient mortality.

TABLE B-2: SCORING SCHEME FOR LABORATORY-BASED ACUTE PHYSIOLOGY SCORE (LAPS)
LABORATORY TEST RESULT / POINTS ASSIGNED TO LAPS
SODIUM (mEq/L)
< 129
129 – 131
132 – 134
135 – 145
146 – 148
149 – 154
 155 / 10
7
6
0
6
7
10
BLOOD UREA NITROGEN (mg/dL)
< 18
18 – 19
20 – 39
40 – 79
 80 / 0
4
12
19
24
TABLE B-2: SCORING SCHEME FOR LABORATORY-BASED ACUTE PHYSIOLOGY SCORE (LAPS)
LABORATORY TEST RESULT / POINTS ASSIGNED TO LAPS
CREATININE (mg/dL)
< 1.0
1.0– 1.9
2.0– 3.9
 4.0 / 0
1
7
5
BUN / CREATININE
< 25
 25 / 0
6
ALBUMIN
 2.5
2.0 – 2.4
< 2.0 / 0
18
23
HEMATOCRIT (%)
< 20.0
20.0 – 39.9
40.0 – 49.9
50.0 – 59.9
 60.0 / 7
5
0
6
23
TABLE B-2: SCORING SCHEME FOR LABORATORY-BASED ACUTE PHYSIOLOGY SCORE (LAPS)
LABORATORY TEST RESULT / POINTS ASSIGNED TO LAPS
WHITE BLOOD CELL COUNT (1000s/mm3)
< 2.0
2.0 – 4.9
5.0 – 12.9
 13.0
Missing and in high risk group / 29
6
0
15
23
ARTERIAL pH
< 7.15
7.15 – 7.24
7.25 – 7.34
7.35 – 7.44
7.45 – 7.54
 7.55
Missing and in high risk group / 30
23
16
0
11
14
11
ARTERIAL PaCO2 (mm Hg)
< 25
25 – 34
35 – 44
45 – 54
55 – 64
 65 / 5
12
0
10
9
13
TABLE B-2: SCORING SCHEME FOR LABORATORY-BASED ACUTE PHYSIOLOGY SCORE (LAPS)
LABORATORY TEST RESULT / POINTS ASSIGNED TO LAPS
ARTERIAL PaO2 (mm Hg)
< 50
50 – 119
 120 / 13
0
18
SERUM GLUCOSE (mg/dL)
< 40
40 – 59
60 – 199
 200 / 16
12
0
3
TOTAL SERUM BILIRUBIN (mg/dL)
< 2.0
2.0 – 2.9
3.0 – 4.9
5.0 – 7.9
 8.0 / 0
10
16
22
32
TROPONIN I (pg/mL)
0
0.01– 0.19
0.20 – 0.99
1.00 – 2.99
3.00 – 5.99
 6.00
Missing and in high risk group / 0
2
6
18
20
25
9

Figure

Relationship between the Laboratory-based Acute Physiology Score (LAPS) and inpatient mortality. The LAPS integrates information from 14 laboratory tests into a single continuous variable with a value that can range between a minimum of 0 and a theoretical maximum of 256 (in our dataset, however, patients with LAPS values > 100 are uncommon). For example, a hematocrit of < 20% contributes 7 points to the total LAPS, while a hematocrit of 40-49% contributes zero points. Laboratory test results for the LAPS were obtained from the 24 hours preceding hospitalization.

SECTION1, PART C: DEVELOPMENT AND VALIDATION OF THE COMORBIDITY POINT SCORE (COPS)

We employed Diagnostic Cost Groups (DxCG) software to group diagnoses that occurred prior to the hospitalization. Every month, the NC-KPMCP Management Information and Analysis Department employs DxCG software (see text for citations)to scan data from outpatient and inpatient encounters from the entire KPMCP membership. Based on the ICD codes from these encounters, the DxCG software assigns patients to 184 possible Hierarchical Condition Categories (HCCs) based on their inpatient and outpatient utilization during the 12 month period preceding each monthly scan. A given patient may have multiple HCC assignments. We examined the relationship between having any of these 184 HCCs and inpatient mortality. Based on the mortality risk and biological plausibility, we grouped 147 of these 184 HCC’s into 41 comorbidity groups. For example, we grouped three HCCs related to head injury into a single comorbidity group, which we labeled CMGR_HEDINJ, as is shown in Table C-1, below.

We did not employ 37 of the 184 HCCs because they either did not have relevance to the study cohort (e.g., some of the HCCs that are specifically focused on pediatric patients) or because they did not have any significant relationship to mortality.

TABLE C-1: EXAMPLE OF DxCG HIERARCHICAL CONDITION CATEGORIES GROUPING IN THE DERIVATION DATASET
HCC / NAME / N in cohort withwithout the HCC / Death rate if HCC is present or absent / Mean length of stay (in hours) if condition present or absent
154 / Severe head injury / 26
197,035 / 7.7%
3.4% / 252.1
99.6
155 / Major head injury / 960
196,101 / 5.2%
3.4% / 112.9
99.6
156 / Concussion or unspecified head injury / 1,542
195,519 / 4.5%
3.4% / 117.4
99.5
--- / CMGR_HEDINJ (any of 154, 155, or 156 is present) / 2,528
194,533 / 4.8%
3.4% / 117.1
99.4

We then employed these 41 comorbidity groups as dichotomous predictor variables in a logistic regression model using the derivation dataset. We then assigned a point value to each comorbidity group based on the beta coefficients from the regression model. This integrated all comorbidity information into a single continuous variable that can range from 0 to a theoretical maximum of 701 (however, few observations in our dataset had scores > 300). We then applied these point values to the validation dataset, which resulted in a model with a c statistic of 0.67.

Table C-2 summarizes the 41 comorbidity groupings (CMGR_’s) we employed. The table, which is based on the entire dataset, shows the groupings, the point score assigned to that comorbidity, the mortality among hospitalizations where a patient did or did not have that comorbidity, and the length of stay among hospitalizations where a surviving patient did or did not have that comorbidity. Note that, in 15,978 hospitalizations where we could not obtain comorbidity data on a patient because that individual was not a Kaiser Foundation Health Plan, Inc. member, which was associated with increased mortality, this was grouped under the category CMGR_UNK, which contributes 42 points towards the COPS. The Figureshows the relationship between the COPS and inpatient mortality.