You are on page 1of 21

Utility of the Healthcare Cost and Utilization Project

National Inpatient Sample (HCUP-NIS) for outcome-


focused spine surgery research

Andrew Caras
VCU Department of Neurosurgery Resident Research Day
June 17, 2022
Impact of Opioid Dependence on Complications, Discharge Disposition
and Resource Utilization Following Spinal Decompression and Fusion
Introduction
• Opioid dependence (OpiD) in the spinal surgery population is increasingly prevalent
• Up to 10% of patients misuse opioids after spinal surgery1
• Over half of patients who receive a fusion have previously used opioids 2
• Prescribing post-operative opioids is itself a risk factor for chronic opioid use 3
• Several retrospective studies demonstrate links between OpiD and poorer outcomes following spinal
surgery4-6
• Database-driven analysis of outcomes following spinal decompression, fusion and revision including all
spinal levels in OpiD patients may be the best representation
• e.g. includes nearly all payers and hospitals; 10-20% of all inpatient stays within a given year, multiple patient populations
• Multivariate analysis can clarify the potential independent impact of OpiD in the short-term postoperative
course
• Socioeconomic-focused analyses can identify patients most at risk

1. Sharma, M., Ugiliweneza, B., Aljuboori, Z., & Boakye, M. (2018). Health care utilization and overall costs based on opioid dependence in patients undergoing surgery for degenerative spondylolisthesis. Neurosurgical Focus, 44(5).
2. Hilliard PE, Waljee J, Moser S, et al. Prevalence of Preoperative Opioid Use and Characteristics Associated With Opioid Use Among Patients Presenting for Surgery. JAMA Surg. 2018;153(10):929-937.
3. Hah, J. M., Bateman, B. T., Ratliff, J., Curtin, C., & Sun, E. (2017). Chronic Opioid Use After Surgery. Anesthesia & Analgesia, 125(5), 1733-1740.
4. Tank, A., Hobbs, J., Ramos, E., & Rubin, D. S. (2018). Opioid Dependence and Prolonged Length of Stay in Lumbar Fusion. Spine, 43(24), 1739-1745.
5. Jain, N., Phillips, F. M., Weaver, T., & Khan, S. N. (2018). Preoperative Chronic Opioid Therapy. Spine, 43(19), 1331-1338.
6. Walid, M., Hyer, L., Ajjan, M., Barth, A. C., & Robinson, J. S. (2007). Prevalence of opioid dependence in spine surgery patients and correlation with length of stay. Journal of Opioid Management, 3(3), 127-128.
Methods and Design
• Healthcare Cost and Utilization Project National Inpatient Sample (HCUP-NIS) database,
Agency for Healthcare Research and Quality
• Years 2008-2014, adults ≥ 18 years old, any spinal level
• Procedures selected via ICD-9 nomenclature
• Spinal fusion (ICD-9 81.00-81.08)
• Spinal decompression (ICD-9 03.09)
• Spinal revision (ICD-9 81.30-81.39)
• Exclusion:
• vertebral fracture (ICD-9 805.00-806.09)
• cancer (ICD-9 140.0-239.9, 173.00-173.99, 198.81-198.89, 200.00-209.36)
• Opioid + other or unspecified drug dependence (ICD-9 304.70-73, 304.90-93)
• Cases divided into opioid-dependent (OpiD; ICD-9 304.0x) and non-dependent (ND) cohorts
Methods and Design
Paraplegia (344.1) Nervous system complications
• Demographics, hospital (997.00-9)
Pulmonary embolism (415.1)
characteristics, insurance, Cardiac complications (997.1)
household income, and AHRQ in- Lower extremity phlebitis &
thrombophlebitis (451.1) Respiratory complications (997.31-9)
hospital mortality score were
compared between OpiD and ND Acute post-hemorrhagic anemia Digestive system complications
groups (285.1) (997.4)
Post-operative shock (998.00-2, Urinary complications (997.5)
• Indices of outcome compared
998.09)
between groups including: Vascular complications of other
Hemorrhage/hematoma/seroma vessels (997.79)
• In-hospital mortality (MR) (998.11-3)
Red blood cell transfusion (99.04)
• Length of Stay (LOS)
Post-operative infection (998.51-9)
Other mechanical complication of
• Discharge disposition (DISP)
Non-healing surgical wound (998.83) internal orthopedic device (996.49)
• Total hospital charge (TOTCHG)
• Complications (ICD-9) 
Statistical Methods
• Continuous data presented as either mean ± SD or median/interquartile range (IQR).
• Student’s T-test or nonparametric test were used to compare the differences
between groups depending on whether the data was normally distributed.
• Categorical data were presented as rate/ratio, and the comparisons between
different groups were done by Chi-square test.
• For multivariable analysis, general linear models were used for continuous
dependent variables, and binary logistic regression models were used for binomial
dependent variables.
• Confounding factors used in multivariable analysis included: type of procedure,
elective admission, demographics, payer, income, hospital characteristics, and in-
hospital mortality index score.
• Type I error set at 0.05. All analyses were 2-sided.
Results Overview Prevalence, Type of Procedure
Opioid dependence No opioid dependence

• Total of 574,730 cases 100%


*
were identified 90% 86.3%
*
83.9%
90.8%

79.4%
80%
• After exclusion, 491,919 70%
patients underwent spinal 60%
fusion, decompression or 50%
revision and met inclusion 40%
criteria 30%

• 2,021 (0.4%) were opioid- 20% *


11.8%
8.8% 9.0%
dependent patients 10% 4.7%
0%
Fusion Fusion + Decompression Revision Elective Admission

*p < 0.001
Demographics: Age and Gender

Total N OpiD ND p
(Proportional %) (n=2,021) (n=489,898)
Age (years, mean ± SD) 56.8 ± 13.4 52.7 ± 12.1 56.9 ± 13.4 < .001
Gender .942
Male 222,806 (45.3%) 917 41.2
Female 269,113 (54.7%) 1104 41.0
White patients more likely than African-Americans or
Hispanics to have comorbid opioid dependence
45
*
40

35
Prevalence of OpiD (per 10,000)

30

25
42.6
20 38.1
33.2
15 28.8

10

0
White African-American Hispanic Other
OpiD prevalence is greatest for Medicaid
enrollees & top 25th percentile income
100 *†§ 60
90 *
Prevalence of OpiD (per 10,000)

Prevalence of OpiD (per 10,000)


80 50
70
40
60
50 †
† 90.9 30
40 50.3
30 20 42.1
36.2 36.0
20 40.9 45.7
34.1
10 10
0
Medicare Medicaid Private Other 0
0-25th 26-50th 51-75th 76-100th
perce... perce... perce... perce...
* significantly higher prevalence than Medicare
† significantly higher prevalence than Private * significantly higher prevalence than 0-25th, 26th to 50th,
§ significantly higher prevalence than Other and 51st to 75th percentile individual cohorts
Each significant at p < 0.01 Each significant at p < 0.05
Western US has highest prevalence of OpiD patients
receiving spinal surgery
100
90
*†§
Prevalence of OpiD (per 10,000)

80
70
60
50
87.9
40 †
30
20
33.6 29.4
10
26.1

0
Northeast Midwest South West

* significantly higher prevalence than Northeast hospital


† significantly higher prevalence than Midwest hospital Regions are defined by U.S. Census Bureau
§ significantly higher prevalence than South hospital
Each significant at p < 0.05
OpiD associated with increased incidence of several
post-operative complications
Complication OpiD (%) ND (%) OR (95% CI) P-value
Any complication 39.30 18.70 2.82 (2.58-3.08) < .001
Acute post-hemorrhagic anemia 17.71 8.91 2.2 (1.96-2.47) < .001
RBC transfusion 14.84 6.29 2.6 (2.30-2.94) < .001
Mechanical complication of internal 14.00 4.42 3.52 (3.10-4.00) < .001
orthopedic device
Digestive system complications 1.73 1.09 1.6 (1.14-2.24) .006
Hemorrhage/hematoma/seroma 2.62 0.95 2.8 (2.13-3.69) < .001
Respiratory complications 1.58 0.60 2.67 (1.88-3.79) < .001
Paraplegia 1.44 0.35 4.11 (2.84-5.95) < .001
Postoperative infection 1.04 0.28 3.78 (2.45-5.82) < .001
Acute deep vein thrombosis 0.50 0.26 1.93 (1.04-3.60) .035
Pulmonary embolism 0.69 0.23 3.03 (1.79-5.15) < .001
Postoperative shock 0.50 0.11 4.73 (2.52-8.85) < .001
Non-healing surgical wound 0.25 0.02 12.79 (5.20-31.46) < .001
Urinary complications 0.35 0.60 0.58 (0.27-1.21) .140
Cardiac complications 0.69 0.56 1.24 (0.73-2.11) .418
Nervous system complications 0.74 0.51 1.46 (0.88-2.44) .140
Vascular complications 0.05 0.01 3.73 (0.52-26.90) .238
Significantly more likely non-routine discharge;
mortality risk is unclear
Proportion of cases, per discharge metric
80% OpiD ND
OR Adjusted OR Adjusted
Discharge Metric p-value (95% CI) p-value
73.0%
(95% CI) 70%
SNF = skilled nursing facility
Mortality
2.37
.032
1.677
.197 60%
* ICF = intermediate care facility
(1.13-5.00) (0.764-3.680) 55.5%
HHC = home health care
0.46 0.374
50% STH = short-term hospital
Routine discharge <.001 <.001
(0.42-0.50) (0.339-0.413) 40%

Discharge to short- 2.95 3.434 30% *


term care facility (2.04-4.27)
.001
(2.352-5.013)
<.001 23.9% *
20% 18.3%
2.02 2.542 13.5% 12.8%
Discharge to <.001 <.001
SNF/ICF/Other (1.82-2.23) (2.259-2.859) 10%
*
1.4% 0.5%
1.53 1.663 0%
Received home Routine Discharge SNF/ICF/Other HHC STH
<.001 <.001
health care (1.36-1.71) (1.482-1.866)

*p < 0.001
Mean length of stay: 4 days
2.49 days longer adjusted mean length of stay for OpiD cases
5 2.8
* 2.6

Adjusted Difference, Mean Length of Stay


2.49 2.4
4
4 2.2
Mean Length of Stay (Days)

2
3 1.8
3
† 1.6

(Days)
1.4
1.2
2
1
0.8
1 0.6
0.4
0.2
0 0

OpiD ND Adjusted Difference

*p < 0.001
†p = 0.012
$117,400 mean total charges, per case, for OpiD patient
$23,280 more expensive compared to ND patient
140K * 25K
$23,280
$117,400
120K
*
Mean Total Charges ($ thousand)

Adjusted Total Charge Difference


20K
100K

($ thousand)
15K
80K $73,000

60K
10K

40K
5K
20K

K K

OpiD ND Adjusted Difference


$44,400 mean difference
*p < 0.001 (pre-adjustment)
≈ +$47 million total charges
Cervical-only cohort analysis
• METHODS:
• Patients ≥ 18 years of age undergoing cervical spinal fusion (ICD-9-CM codes 81.02 [anterior], 81.03
[posterior]), decompression (03.09) or revision operations (81.32 [anterior], 81.33 [posterior]) from
years 2008-2014
• Cohorts were compared via propensity-score matching for age, type of procedure, approach, revision
versus non-revision procedure, insurance status, number of chronic medical conditions, comorbid
alcohol use, comorbid tobacco use, elective procedure, hospital geographic region, hospital size, and
hospital teaching status.
• Matched cohorts were compared using Chi-square tests for ordinal variables and either t-test for
normally distributed data (mean and 95% CI) or Mann-Whitney U test for nonparametric data
distribution (median and IQR).
• Odds ratios were calculated to approximate relative risk of experiencing an outcome variable depending
on OpiD status was calculated, with 95% confidence intervals.
• Demographic and baseline variables were compared between matched cohorts to assess for adequate
matching. Significance was interpreted at p < .05 for all analyses.
Cervical-only cohort analysis
• RESULTS:
• 180,530 patients received cervical spinal operations; 490 (0.27%) had comorbid OpiD
• The annual proportion of patients with OpiD doubled from 2008 (0.17%) to 2014 (0.34%).
• Matched cohort comparisons:
• similar complication rates (23.3% versus 19.9%; p = .100)
• longer median length of stay (LOS; 3 days [IQR 2-7] versus 2 days [IQR 1-4]; p < .001)
• greater median total inpatient charges (TIC; $80,215 [IQR $50,614-$134,118] versus $61,189 [IQR $40,502-
$100,952]; p < .001) for patients with OpiD.
• Non-home discharge was more frequent in opioid-dependent patients (33.7% versus 26.9%;
p = .001).
  Non-matched Matched Median household - - .225 - - .515
Variable OpiD (n=490) Non-OpiD P-value OpiD Non-OpiD P-value income
(n=180,040) (n=490) (n=490)

Age, mean ± SD 52.03 ± 10.56 54.92 ± 12.01 < .001 52.03 ± 52.39 ± .322 0-25th percentile 123 (25.1%) 44, 426 (24.7%) - 123 120 -
10.56 11.31 (25.1%) (24.5%)
Percent female 249 (50.8%) 93,900 (52.7%) .554 249 (50.8%) 247 (50.4%) .898 26-50th percentile 110 (22.4%) 47,672 (26.5%) - 110 111 -
(22.4%) (22.7%)
Procedure - - - - - -
Fusion 443 (90.4%) 174,569 (97.0%) < .001 443 (90.4%) 441 (90.0%) .830 51-75th percentile 133 (27.1%) 45,922 (25.5%) - 133 151 -
(27.1%) (30.8%)
Fusion + decompression 34 (6.9%) 7199 (4.0%) .001 34 (6.9%) 18 (3.7%) .023
76-100th percentile 124 (25.3%) 52,020 (23.3%) - 124 108 -
Revision 70 (14.3%) 7547 (4.2%) < .001 70 (14.3%) 61 (12.4%) .398
(53.4%) (46.6%)
Elective admission 388 (79.2%) 160,366 < .001 388 402 .258
Anterior cervical fusion 333 (68.0%) 156,076 (86.7%) < .001 333 (68.0%) 332 (67.8%) .945 (89.1%) (79.2%) (82.0%)
(ACF)
Posterior cervical fusion 96 (19.6%) 17,716 (9.8%) < .001 96 (19.6%) 101 (20.6%) .690 Hospital region - - < .001 - - .745
(PCF) Northeast 74 (15.1%) 32,124 (17.8%) - 74 (15.1%) 68 (13.9%) -
Revision ACF 39 (8.0%) 3,936 (2.2%) < .001 39 (8.0%) 33 (6.7%) .463
Midwest 48 (9.8%) 29,696 (16.5%) - 48 (9.8%) 41 (8.4%) -
Revision PCF 22 (4.5%) 2,312 (1.3%) < .001 22 (4.5%) 24 (4.9%) .763

Race - - .490 - - .192 South 177 (36.1%) 87,844 (48.8%) - 177 177 -
White 404 (82.4%) 146,184 (81.2%) - 404 (82.4%) 382 (78.0%) - (36.1%) (36.1%)
 
Black 49 (10.0%) 17,396 (9.7%) - 49 (10.0%) 52 (10.6%) -
West 191 (39.0%) 30,376 (16.9%) - 191 204 -
(39.0%) (41.6%)
Hispanic 24 (4.9%) 9264 (5.1%) - 24 (4.9%) 34 (6.9%) - Hospital bed size - - < .001 - - .176
Others 13 (2.7%) 7196 (4.0%) - 13 (2.7%) 22 (4.5%) -
Small 46 (9.4%) 28,322 (15.7%) - 46 (9.4%) 38 (7.8%) -
Insurance - - < .001 - - .520
Medicare 177 (36.1%) 55,010 (30.6%) - 177 (36.1%) 191 (39.0%) - Medium 101 (20.6%) 42,545 (23.6%) - 101 83 (16.9%) -
(20.6%)
Medicaid 72 (14.7%) 11,551 (6.4%) - 72 (14.7%) 59 (12.0%) -
Large 343 (70.0%) 109,173 (60.6% - 343 369 -
Private insurance 185 (37.8%) 92,177 (51.2%) - 185 (37.8%) 178 (36.3%) - (70.0%) (75.3%)
Teaching hospital - - < .001 - - .213
Self-pay 56 (11.4%) 21,302 (11.8%) - 56 (11.4%) 62 (12.7%) - status
Number of chronic 6 [5-9] 3 [2-5] < .001 6 [5-9] 6 [4-9] .673 Rural 5 (1.0%) 7635 (4.2%) - 5 (1.0%) 1 (0.2%) -
conditions, median
[IQR]
Urban non- 242 (49.4%) 77.820 (43.2%) - 242 254 -
Alcoholism 30 (6.1%) 2417 (1.3%) < .001 30 (6.1%) 40 (8.2%) .215 teaching (49.4%) (51.8%)
Smoking 225 (45.9%) 56,774 (31.5%) < .001 225 (45.9%) 224 (45.7%) .949 Urban teaching 243 (49.6%) 94,585 (52.5%) - 243 235 -
(49.6%) (48.0%)
  Non-matched Matched
Variable OpiD (n=490) Non-OpiD P-value OpiD Non-OpiD P-value Relative risk
(n=180,040) (n=490) (n=490) (95% CI)
Any complication 114 (23.3%) 13346 (7.4%) < .001 114 (23.3%) 93 (19.0%) .100 1.294 (0.951-
1.761)
Post-hemorrhagic 26 (5.3%) 464 (0.3%) < .001 26 (5.3%) 19 (3.9%) .285 1.389 (0.758-
anemia 2.544)
Received red blood cell 27 (5.5%) 463 (0.3%) < .001 27 (5.5%) 19 (3.9%) .227 1.446 (0.793-
transfusion 2.636)
Mechanical complication 54 (11.0%) 436 (0.2%) < .001 54 (11.0%) 44 (9.0%) .287 1.255 (0.825-
of internal orthopedic 1.909)
device
Length of stay, median 3 [2-7] 1 [1-2] < .001 3 [2-7] 2 [1-4] < .001 -
[IQR]
Discharge disposition - - < .001 - - .002  
Home 325 (66.3%) 154,065 (85.6%) - 325 (66.3%) 358 (73.1%) -  
Short-term inpatient 11 (2.2%) 624 (0.3%) - 11 (2.2%) 1 (0.2%) -  
care
SNF, intermediate care 81 (16.5%) 12,606 (7.0%) - 81 (16.5%) 88 (18.0%) -  
facility, or other
Home health care 70 (14.3%) 12,354 (6.9%) - 70 (14.3%) 41 (8.4%) -  
AMA 2 (0.4%) 131 (0.1%) - 2 (0.4%) 2 (0.4%) -  
Total inpatient charges, $80,215 [$50,614- $49,805 [$33,570- < .001 $80,215 $61,189 < .001 -
median [IQR] $134,118] $75,035] [$50,614- [$40,502-
$134,118] $100,952]
Mortality NR 249 (0.1%) - NR NR - -
Limitations
• Variables within the database are largely limited to identification via ICD-9-CM code until 2014
• inclusion of data after this timepoint was deferred to minimize bias and improve internal validity of patient selection
• Non-longitudinal
• Some patients within the HCUP-NIS likely suffered from OpiD yet lacked formal diagnosis (and therefore a
corresponding ICD-9-CM diagnosis code) and were not included in the OpiD cohort
• Relatively low OpiD patients compared to the general population which may indicate: 1) only a few providers
were regularly diagnosing patients with OpiD or 2) only patients who were overtly symptomatic were likely
to be diagnosed with OpiD.
• Several ICD-9-CM diagnosis codes for “opioid abuse” exist (as opposed to “opioid dependence” which this
study utilized); however, we felt these codes were vague, both in definition and severity, compared to opioid
dependence and may be arbitrarily applied to patients.
• The HCUP-NIS does not contain prescription data which would permit calculation of morphine-equivalent
dose for patients taking opioids prior to surgery, identification of the most common opioids used, and
analysis of a dose and duration-dependent relationship
CONCLUSIONS
• Medicaid participants and higher income individuals associated with increased likelihood of opioid misuse
• Stereotypical regions of high-prevalence opioid use disorder (e.g. Midwest) are not reflected
• Poorly controlled post-operative pain  increased LOS  increased costs and non-routine discharge
• Outcome and resource utilization following cervical fusion is probably less impacted by the independent effect of
OpiD than lumbar fusion
• OpiD is probably not an independent risk factor for complications following cervical spinal decompression and fusion

Future Directions
• Individualization: comparison of cervical vs. lumbar cohorts, approaches, construct length
• Analysis of ICD-10 social determinant “Z” codes
• Study of outpatient functionality, cost, and readmission: socioeconomic impacts of non-routine discharge
• Expanded consideration of multidisciplinary perioperative interventions for OpiD patients receiving spinal surgery
may alleviate these risks, significantly reduce LOS and decrease inpatient costs.
• Early-and-often physical therapy
• Pre-operative behavioral psychology and pain management consultation
• Implementation of ERAS techniques during admission and creation of custom ERAS protocols for chronic opioid users

You might also like