PRACTICE REPORT Postsurgical opioids
ar
PRACTICE REPORT
Layar
Adverse drug events among patients receiving postsurgical opioids in a large health system: Risk factors and outcomes Harold S. Minkowitz, Stephen K. Gruschkus, Manan Shah, and Aditya Raju
A
n estimated 25 million inpatient surgeries and 35 million ambulatory care surgeries are performed annually in the United States.1,2 Severe pain after surgery remains a serious problem, affecting approximately 20–40% of patients.3 Poorly controlled pain management has physical and psychological consequences for patients and a negative impact on the healthcare system, including greater costs, increased length of stay (LOS) or delayed discharge, and higher rates of hospital readmission. Current guidelines from the American Society of Anesthesiologists (ASA) advocate the use of multimodal approaches to postsurgical pain management that typically include opioid analgesics such as morphine and fentanyl.4 Opioids remain firstline agents for providing superior postsurgical analgesia.5 Opioids have demonstrated efficacy for pain relief after surgery and often are the analgesics of choice for postoperative pain, but their use is often accompanied by
Purpose. Results of a study of postsurgical opioid-related adverse drug events (ORADEs) within a large health system are reported. Methods. In a retrospective cohort study, data from the information database of an 11-hospital Texas health system were analyzed to (1) describe postsurgical opioid use among adult patients undergoing elective or emergency surgery over a one-year period, (2) identify ORADE risk factors and associated costs, and (3) compare clinical and economic outcomes in patients who experienced ORADEs and those who did not. Multivariate logistic regression was used to identify ORADE risk factors. Propensity score–matched comparisons of outcomes in patients with and without ORADEs were conducted. Results. Among 6,285 patients in the study population, 6,274 (99.8%) received postsurgical opioids; 11.5% of those
adverse drug events (ADEs) and other negative consequences, including increased mortality.5
Harold S. Minkowitz, M.D., is Anesthesiologist, Memorial Hermann Memorial City Medical Center, Houston, TX. Stephen K. Gruschkus, Ph.D., M.P.H., was, when this article was written, Assistant Director, Applied Data Analytics, Xcenda, AmerisourceBergen Consulting Services, Palm Harbor, FL. Manan Shah, Pharm.D., Ph.D., is Director, Health Services and Outcomes Research, Bristol-Myers Squibb, Plainsboro, NJ; when this article was written, he was Director, Applied Data Analytics, Xcenda. Aditya Raju, M.S., is Manager, Global Health Economics and Outcomes Research, Xcenda. Address correspondence to Mr. Raju ([email protected]). Research and manuscript preparation funded by Pacira Pharmaceuticals.
1556
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
patients experienced an ORADE. ORADE risk factors included age (≥65 years), male sex, prior opioid use, chronic obstructive pulmonary disease, cardiac dysrhythmias, regional enteritis, diverticulitis, and ulcerative colitis. Patients with multiple risk factors had higher mean hospitalization costs ($21,073) relative to patients with one risk factor ($14,110) or no risk factor ($11,433) and accounted for a disproportionately large share of overall costs; patients who experienced ORADEs were more likely to be cost and length of stay (LOS) outliers. Conclusion. Analysis of information from a large database demonstrated that opioidtreated postsurgical inpatients who had multiple risk factors for ORADEs were more likely to have higher mean costs, greater readmission rates, and longer LOS than patients with fewer risk factors. Am J Health-Syst Pharm. 2014; 71:155665
Opioids provide analgesic activity by binding to opioid receptors found throughout the central nervous sys-
Dr. Minkowitz has performed other clinical research funded by Pacira. At the time of this research, Drs. Gruschkus and Shah were employed by Xcenda, a consulting company that received funding from Pacira. Mr. Raju is employed by Xcenda, a consulting company that received funding from Pacira to support development of this article. Copyright © 2014, American Society of Health-System Pharmacists, Inc. All rights reserved. 1079-2082/14/0902-1556. DOI 10.2146/ajhp130031
PRACTICE REPORT Postsurgical opioids
tem (CNS). However, their binding affinity is not limited to the CNS, and activation of receptors in other areas of the body may lead to adverse events. Wheeler et al.6 found that 29% of ADEs were associated with analgesic use, with the majority of those ADEs involving opioids. Opioid-related ADEs (ORADEs) such as CNS impairment and respiratory depression may lead to marked patient impairment and even death.7,8 Several patient demographic and clinical characteristics have been linked to ADE risk, such as age, sex, ethnicity, insurance status, type of physician visit (specialty versus primary care), number of medications used, weight, smoking history, and renal and hepatic functions.9-11 Other factors such as medication errors, which are common with opioid delivery systems, also contribute substantially to the risk of developing an ADE.6 While opioids have low acquisition costs, the frequency and severity of postsurgical ADEs pose a significant economic burden for hospitals and healthcare providers. Oderda et al.12 reported that patients with an ADE had a significant increase in overall hospital costs, with their median costs being 7.4% higher than those of patients not experiencing an ADE. Significant increases were seen in both LOS and cost among patients with postsurgical ADEs for each of the evaluated types of surgery (general, obstetric–gynecologic, orthopedic) relative to matched patients without ADEs. Some ADEs (e.g., respiratory depression, nausea, vomiting) required increased monitoring by the nursing staff and administration of supportive care therapies such as antiemetics. In another study, the overall rate of readmission after various types of same-day surgery was found to be low (5.7%), but among the 1.5% of patients who were readmitted for reasons associated with the original surgery, pain accounted for more than a third of the readmissions.13 Ultimately, the
additional personnel time, moreintensive patient management, longer LOS, and greater readmission rates associated with ADEs led to increased burdens on the healthcare system in terms of cost12 and diminished quality of life for the patient.14 Consequently, appropriate use of this important class of analgesics with the aim of balancing analgesia and adverse events is critical. Efforts to identify specific segments of the surgical population at increased risk for ADEs provide an opportunity to determine the safest and most effective pain management strategy and may help to decrease the economic burden of ADEs. The purposes of the study described here were to characterize postsurgical opioid utilization, estimate rates of ADEs among opioid users, identify risk factors for ADEs, and evaluate the impact of ADEs on cost and resource utilization using data from a large nonprofit healthcare system. Methods Data source. This retrospective cohort study was based on analysis of data from the Eclipsys Sunrise (EPSI) database (Eclypsis Corporation, Atlanta, GA), which is a large, service-level, all-payer (including Medicaid, Medicare, and commercial) comparative database containing administrative and payment data on patients receiving care within the Memorial Hermann Hospital System (MHHS), the largest nonprofit healthcare system in Texas, comprising 11 hospitals and accounting for approximately 3500 inpatient beds. Additionally, EPSI provides financial transactions for patient-level services, including inpatient and outpatient procedures and medications, laboratory tests, diagnostic and therapy tests, discharge diagnoses and status, and physician-input International Classification of Diseases, Ninth Revision (ICD-9) codes for documentation of adverse events.
Study design and sample selection. Patients 18 years of age or older with a primary inpatient surgical procedure occurring between January 1 and December 31, 2010, were eligible for inclusion in the study population. Specific surgical procedures (Table 1) were selected to represent common orthopedic and soft tissue surgeries and to align with previous research.15-17 For all patients, the first hospitalization during the study period was designated as the index hospitalization. Using information provided with the EPSI database, patients were then followed from the admission date through postdischarge day 30 to describe patient characteristics and to evaluate postsurgical pain management and outcomes. Postsurgical opioid pain management was defined as the administration of parenteral or oral opioid analgesics on or after the procedure date and before discharge. The following opioid analgesics commonly used in the surgical setting within MHHS were considered: codeine, fentanyl, hydrocodone, hydromorphone, meperidine, methadone, morphine, oxycodone, and propoxyphene. Once the rate of opioid utilization was determined, all subsequent analyses were limited to those patients who received postsurgical opioids. All patients in the study cohort had a primary inpatient surgical procedure and hence associated charges or costs. ADE identification and risk factor assessment. The frequency of ADEs was estimated by identifying specific ICD-9 diagnoses (Table 2) within the EPSI database, including gastrointestinal (GI), respiratory, genitourinary (GU), CNS, and other ADEs (postoperative bradycardia, fall from bed, rash or itching). As ICD-9 coding within large databases typically has some variability, three-digit codes were included to capture a broader group of codes; where specificity was required, four- and five-digit codes were used. Patients were then placed
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
1557
PRACTICE REPORT Postsurgical opioids
into cohorts based on whether they had experienced an ADE or not. The characteristics considered for the ADE risk factor assessment included age, sex, opioid use prior to surgery, and other comorbidities based on their relevance to medication use and ADE risk; this approach was consistent with previous research in this area17 (Table 3). Multivariate logistic regression was used to evaluate odds ratios (ORs) and accompanying 95% confidence intervals (CIs) and to identify factors significantly associated with ADE risk in the overall population and by sex. Within the multivariate model, the occurrence of an ADE was specified as the dependent variable, with risk factors specified as the predictor variables. Only significant variables (p < 0.05)
were retained in the final model and reported. Outcomes assessment. Outcomes after ADEs evaluated in this study included LOS, total hospitalization costs, 30-day readmission rates, inpatient mortality, and outlier status. LOS was defined as the time in days from the index admission date to the discharge date. Total hospitalization cost was measured from the admission index date to the discharge date and was based on charges for the following: operating room, charge room, charge supplies, cardiology services, respiratory and pulmonary services, emergency services, radiology services, laboratory services, pharmacy services, and other ancillary services. To provide a better estimate of true costs, 2011 Medicare
Table 1.
Postsurgical Opioid Use in Base Study Population, by Procedurea
Procedure Laparoscopic cholecystectomy Total abdominal hysterectomy Laparoscopic gastric bypass Hip replacement, total Hip fracture reduction with fixation Open colectomy, partial Other partial gastrectomy Laparoscopic colectomy, partial Laparoscopic total abdominal hysterectomy Laparoscopic assisted vaginal hysterectomy Open cholecystectomy Laparoscopic gastric restrictive procedure Open gastric bypass Ileostomy reversal Hip replacement, partial Open colectomy, total Laparoscopic colectomy, total Mastectomy Abdominoplasty Hip fracture reduction without fixation All procedures
ICD-9 Procedure Code(s) 51.23 68.4, 68.49 44.38 81.51 79.35 45.7, 45.71–45.79 43.89 17.31–17.39
% Receiving Opioids
1,730 99.7 938 99.9 665 99.7 634 100 523 99.8 482 100 285 100 251 100
68.41
213
100
68.31 51.22
139 111
100 100
44.95 44.39 46.51, 46.52 81.52 45.8, 45.82, 45.83 45.81 85.4, 85.41–85.48 86.83 79.25 . . .
ICD-9 = International Classification of Diseases, Ninth Revision. Not applicable.
a
b
1558
No. Patients Identified
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
91 100 68 100 55 100 49 98.0 40 100 5 100 4 100 2 100 0 . . .b 6285 99.8
charge-to-cost ratios were applied to the charge data. Readmission rates were based on documentation of subsequent admission to a participating hospital within 30 days of the discharge date for the index procedure. Inpatient mortality was defined by patient discharge status. Additionally, patients were identified as LOS or cost outliers if their LOS or costs were greater than one standard deviation above the corresponding mean values for the overall MHHS population. A matched analysis using propensity scores was employed to control for underlying differences in demographic and patient characteristics between patients who experienced ADEs and those who did not. A propensity score is a composite score representing the probability (based on baseline characteristics) of patients falling into comparison groups of interest (in this case, patients with or without ADEs). Using this probability to adjust the estimated effect of adverse events on outcomes has the advantage of creating a “quasirandomized” experiment. For this study, propensity scores were obtained from procedure-specific logistic regression models that predicted the probability of having an adverse event. The selection of factors for inclusion in the risk assessment model was based on previous literature and included age, sex, Charlson Comorbidity Index score, obesity, degenerative joint disease (DJD), and prior opioid use. Patients with ADEs were matched by propensity score to patients without ADEs using a 1:1 ratio with the technique of nearest available match for the propensity scores. A narrow caliper (± 0.001 S.D.) was used to ensure that good matching was achieved and that the covariates were well balanced across the groups. Prematch and postmatch baseline demographic and patient characteristics were evaluated to identify any imbalance in the distribution of potential confounders between the
PRACTICE REPORT Postsurgical opioids
cohorts. Prematch comparisons between cohorts were conducted using t tests for continuous variables and chi-square tests for categorical variables. Postmatch comparisons were done using paired t tests for continuous variables and McNemar’s test for categorical variables. Matched inferential comparisons were conducted for the outcomes analysis. Wilcoxon signed rank tests were used to assess the differences in LOS and total hospitalization costs. Rate ratios and corresponding 95% CIs for LOS were generated using negative binomial regression models, a method appropriate for modeling count variables that accounts for overdispersion. Cost ratios and 95% CIs for total hospitalization costs were generated by a generalized linear model approach with a log-link function and gamma distribution. The 30-day readmission and inpatient mortality rates were evaluated using McNemar’s tests. Relative risk values and corresponding 95% CIs were generated via logistic regression models. All statistical analyses tested a two-sided hypothesis of no difference between cohorts at a significance level of 0.05 and were carried out using SAS statistical software, version 9.2 (SAS Institute Inc., Cary, NC). Results A total of 51,298 patients were identified in the initial MHHS data set; of these, 44,251 (86.3%) did not receive a surgical procedure of interest. Another 3,149 patients (6.1%) were excluded because they were younger than 18 years of age, and 649 (1.3%) were excluded due to incomplete capture of medication data (these reported criteria were not mutually exclusive). After application of all the inclusion and exclusion criteria, a total of 6,285 surgical patients were considered eligible for inclusion in the study population. The mean age of the final study population was 52.2 years, and 72.1% (n = 4,522) were female. Among all the surgical
procedures evaluated, laparoscopic cholecystectomy, total abdominal hysterectomy, laparoscopic gastric bypass, and total hip replacement were the most common procedures and accounted for 63% of the study population (Table 1). Nearly all patients (n = 6,274, 99.8%) in the base study population received postsurgical opioids; financial records could not be used to evaluate reasons why the remaining 11 patients did not receive postsurgical opioids. Among patients who received postsurgical opioid-based regimens, 11.0% (n = 689) experienced a documented ADE. GI ADEs were the
most common, occurring in 6.3% (n = 396) of patients. Less common ADEs included respiratory (n = 182, 2.9%), GU (n = 94, 1.5%), CNS (n = 38, 0.6%), and other (n = 57, 0.9%) ADEs. Results of the risk factor evaluation are presented in Table 4. In the overall population, patients 65 years of age or older were twice as likely to have an ADE compared with younger patients. This association persisted by sex: Older men were 50% more likely than younger men to experience an ADE, while older women had a greater than twofold increased risk relative to younger women. In
Table 2.
Diagnosis Codes Used to Identify Postsurgical Opioid-Related Adverse Drug Events in Study Populationa Adverse Event
ICD-9 Code
Respiratory Bradypnea 786.09 Pulmonary insufficiency 518.5 Respiratory complications 997.3 Asphyxia 799.01 Hypoxemia 799.02 Gastrointestinal (GI) Constipation 564.09 Constipation, narcotic related E937.9 Dizziness/vertigo 386.2 Dry mouth 527.7 Ileus, postoperative 997.4 Paralytic Ileus 560.1 Nausea with vomiting 787.01 Nausea with vomiting after GI surgery 564.3 Central nervous system Nervousness 799.2 Delirium 780.09 Confusion, postoperative 293.9 Confusion due to condition classified elsewhereb 293 Altered mental status 780.97 Cerebral hypoxia 997.01 Genitourinary system Urinary retention 788.2 Oliguria 997.5 Other Bradycardia, postoperative 997.1 Rash/itching 698.9 Drugs causing adverse effects in therapeutic use E935.2 Fall from bed E884.4 ICD-9 = International Classification of Diseases, Ninth Revision. Confusion due to ICD-9–coded condition.
a
b
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
1559
PRACTICE REPORT Postsurgical opioids
Table 3.
Demographic and Clinical Factors Included in Assessment of Risk of Postsurgical Opioid-Related Adverse Drug Eventsa Factor
ICD-9 Diagnosis Code(s)
Age . . .b Sex ...b Opioid use prior to surgery ...b Obesity 278 DJD 715 COPD 490, 493 Asthma 493 Pulmonary hypertension 416.18 Congestive heart failure 428.0 Benign prostatic hypertrophy 600 Coronary atherosclerosis 414.0 Cardiac dysrhythmias 427, 427.89, 427.9 Hypertension 401, 401.1, 401.9 Dementia 290, 290.1, 294, 294.1, 294.11, 294.2 Depression 296, 296.2 Diabetes mellitus 249, 250 Irritable bowel syndrome 564.1 Regional enteritis 555.0, 555.1, 555.9 Diverticulitis 562 Ulcerative colitis 556 GERD 530.11, 530.81 a ICD-9 = International Classification of Diseases, Ninth Revision, DJD = degenerative joint disease, COPD = chronic obstructive pulmonary disease, GERD = gastroesophageal reflux disease. b Not applicable.
general, men had a higher ADE risk than women. Comorbidities associated with increased ADE risk in the overall population included chronic obstructive pulmonary disease (COPD), cardiac dysrhythmia, regional enteritis, diverticulitis, and ulcerative colitis. While not statistically significant, there was some evidence of an association between ADE risk and asthma, diabetes mellitus, and depression. Among men, we found that patients with benign prostatic hyperplasia (BPH) had a greater than fivefold increased risk of ADEs compared with patients without BPH. Interestingly, the presence of DJD was associated with a lower risk of ADEs in the overall population, though this association may have been confounded by the possibility that these patients were managed with greater vigilance in the postsurgical setting. In addition to demographics and comorbidities, postsurgical opioid use was significantly associated with ADE risk in the overall population.
Table 4.
Frequency of Risk Factors for Opioid-Related Adverse Drug Events (ADEs) in All Patientsa No. (%) Patients Risk Factor Age of ≥65 yr Male sex Obesity DJD COPD Asthma Atherosclerosis Cardiac dysrhythmias Hypertension Diabetes mellitus Regional enteritis Diverticulitis Ulcerative colitis GERD Dementia Depression Prior opioid use
Without ADE (n = 5585)
With ADE (n = 689)
All (n = 6274)
1263 (22.6) 1440 (25.8) 1554 (27.8) 691 (12.4) 47 (0.8) 258 (4.6) 308 (5.5) 103 (1.8) 2268 (40.6) 1027 (18.4) 23 (0.4) 257 (4.6) 20 (0.4) 810 (14.5) 133 (2.4) 61 (1.1) 1845 (33.0)
290 (42.1) 312 (45.3) 134 (19.4) 72 (10.7) 22 (3.2) 40 (5.8) 74 (10.7) 81 (11.8) 310 (45.0) 166 (24.1) 12 (1.7) 87 (12.6) 9 (1.3) 105 (15.2) 39 (5.7) 13 (1.9) 314 (45.6)
1553 (24.8) 1752 (27.9) 1688 (0.93) 763 (12.2) 69 (1.1) 298 (4.7) 382 (6.1) 184 (2.9) 2578 (41.1) 1193 (19.0) 55 (0.9) 344 (5.5) 29 (0.5) 915 (14.6) 172 (2.7) 74 (1.2) 2159 (34.4)
Odds Ratio (95% CI)
p
2.02 (1.65–2.47) 2.00 (1.69–2.38) 0.93 (0.74–1.16) 0.73 (0.55–0.97) 2.34 (1.35–4.07) 1.43 (0.99–2.06) 1.06 (0.79–1.43) 5.39 (3.92–7.42) 0.92 (0.77–1.10) 1.21 (0.99–1.49) 4.30 (2.05–9.04) 2.35 (1.79–3.09) 3.77 (1.65–8.63) 1.10 (0.86–1.39) 1.37 (0.92–2.05) 1.88 (0.98–3.60) 1.27 (1.07–1.52)
ar
PRACTICE REPORT
Layar
Adverse drug events among patients receiving postsurgical opioids in a large health system: Risk factors and outcomes Harold S. Minkowitz, Stephen K. Gruschkus, Manan Shah, and Aditya Raju
A
n estimated 25 million inpatient surgeries and 35 million ambulatory care surgeries are performed annually in the United States.1,2 Severe pain after surgery remains a serious problem, affecting approximately 20–40% of patients.3 Poorly controlled pain management has physical and psychological consequences for patients and a negative impact on the healthcare system, including greater costs, increased length of stay (LOS) or delayed discharge, and higher rates of hospital readmission. Current guidelines from the American Society of Anesthesiologists (ASA) advocate the use of multimodal approaches to postsurgical pain management that typically include opioid analgesics such as morphine and fentanyl.4 Opioids remain firstline agents for providing superior postsurgical analgesia.5 Opioids have demonstrated efficacy for pain relief after surgery and often are the analgesics of choice for postoperative pain, but their use is often accompanied by
Purpose. Results of a study of postsurgical opioid-related adverse drug events (ORADEs) within a large health system are reported. Methods. In a retrospective cohort study, data from the information database of an 11-hospital Texas health system were analyzed to (1) describe postsurgical opioid use among adult patients undergoing elective or emergency surgery over a one-year period, (2) identify ORADE risk factors and associated costs, and (3) compare clinical and economic outcomes in patients who experienced ORADEs and those who did not. Multivariate logistic regression was used to identify ORADE risk factors. Propensity score–matched comparisons of outcomes in patients with and without ORADEs were conducted. Results. Among 6,285 patients in the study population, 6,274 (99.8%) received postsurgical opioids; 11.5% of those
adverse drug events (ADEs) and other negative consequences, including increased mortality.5
Harold S. Minkowitz, M.D., is Anesthesiologist, Memorial Hermann Memorial City Medical Center, Houston, TX. Stephen K. Gruschkus, Ph.D., M.P.H., was, when this article was written, Assistant Director, Applied Data Analytics, Xcenda, AmerisourceBergen Consulting Services, Palm Harbor, FL. Manan Shah, Pharm.D., Ph.D., is Director, Health Services and Outcomes Research, Bristol-Myers Squibb, Plainsboro, NJ; when this article was written, he was Director, Applied Data Analytics, Xcenda. Aditya Raju, M.S., is Manager, Global Health Economics and Outcomes Research, Xcenda. Address correspondence to Mr. Raju ([email protected]). Research and manuscript preparation funded by Pacira Pharmaceuticals.
1556
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
patients experienced an ORADE. ORADE risk factors included age (≥65 years), male sex, prior opioid use, chronic obstructive pulmonary disease, cardiac dysrhythmias, regional enteritis, diverticulitis, and ulcerative colitis. Patients with multiple risk factors had higher mean hospitalization costs ($21,073) relative to patients with one risk factor ($14,110) or no risk factor ($11,433) and accounted for a disproportionately large share of overall costs; patients who experienced ORADEs were more likely to be cost and length of stay (LOS) outliers. Conclusion. Analysis of information from a large database demonstrated that opioidtreated postsurgical inpatients who had multiple risk factors for ORADEs were more likely to have higher mean costs, greater readmission rates, and longer LOS than patients with fewer risk factors. Am J Health-Syst Pharm. 2014; 71:155665
Opioids provide analgesic activity by binding to opioid receptors found throughout the central nervous sys-
Dr. Minkowitz has performed other clinical research funded by Pacira. At the time of this research, Drs. Gruschkus and Shah were employed by Xcenda, a consulting company that received funding from Pacira. Mr. Raju is employed by Xcenda, a consulting company that received funding from Pacira to support development of this article. Copyright © 2014, American Society of Health-System Pharmacists, Inc. All rights reserved. 1079-2082/14/0902-1556. DOI 10.2146/ajhp130031
PRACTICE REPORT Postsurgical opioids
tem (CNS). However, their binding affinity is not limited to the CNS, and activation of receptors in other areas of the body may lead to adverse events. Wheeler et al.6 found that 29% of ADEs were associated with analgesic use, with the majority of those ADEs involving opioids. Opioid-related ADEs (ORADEs) such as CNS impairment and respiratory depression may lead to marked patient impairment and even death.7,8 Several patient demographic and clinical characteristics have been linked to ADE risk, such as age, sex, ethnicity, insurance status, type of physician visit (specialty versus primary care), number of medications used, weight, smoking history, and renal and hepatic functions.9-11 Other factors such as medication errors, which are common with opioid delivery systems, also contribute substantially to the risk of developing an ADE.6 While opioids have low acquisition costs, the frequency and severity of postsurgical ADEs pose a significant economic burden for hospitals and healthcare providers. Oderda et al.12 reported that patients with an ADE had a significant increase in overall hospital costs, with their median costs being 7.4% higher than those of patients not experiencing an ADE. Significant increases were seen in both LOS and cost among patients with postsurgical ADEs for each of the evaluated types of surgery (general, obstetric–gynecologic, orthopedic) relative to matched patients without ADEs. Some ADEs (e.g., respiratory depression, nausea, vomiting) required increased monitoring by the nursing staff and administration of supportive care therapies such as antiemetics. In another study, the overall rate of readmission after various types of same-day surgery was found to be low (5.7%), but among the 1.5% of patients who were readmitted for reasons associated with the original surgery, pain accounted for more than a third of the readmissions.13 Ultimately, the
additional personnel time, moreintensive patient management, longer LOS, and greater readmission rates associated with ADEs led to increased burdens on the healthcare system in terms of cost12 and diminished quality of life for the patient.14 Consequently, appropriate use of this important class of analgesics with the aim of balancing analgesia and adverse events is critical. Efforts to identify specific segments of the surgical population at increased risk for ADEs provide an opportunity to determine the safest and most effective pain management strategy and may help to decrease the economic burden of ADEs. The purposes of the study described here were to characterize postsurgical opioid utilization, estimate rates of ADEs among opioid users, identify risk factors for ADEs, and evaluate the impact of ADEs on cost and resource utilization using data from a large nonprofit healthcare system. Methods Data source. This retrospective cohort study was based on analysis of data from the Eclipsys Sunrise (EPSI) database (Eclypsis Corporation, Atlanta, GA), which is a large, service-level, all-payer (including Medicaid, Medicare, and commercial) comparative database containing administrative and payment data on patients receiving care within the Memorial Hermann Hospital System (MHHS), the largest nonprofit healthcare system in Texas, comprising 11 hospitals and accounting for approximately 3500 inpatient beds. Additionally, EPSI provides financial transactions for patient-level services, including inpatient and outpatient procedures and medications, laboratory tests, diagnostic and therapy tests, discharge diagnoses and status, and physician-input International Classification of Diseases, Ninth Revision (ICD-9) codes for documentation of adverse events.
Study design and sample selection. Patients 18 years of age or older with a primary inpatient surgical procedure occurring between January 1 and December 31, 2010, were eligible for inclusion in the study population. Specific surgical procedures (Table 1) were selected to represent common orthopedic and soft tissue surgeries and to align with previous research.15-17 For all patients, the first hospitalization during the study period was designated as the index hospitalization. Using information provided with the EPSI database, patients were then followed from the admission date through postdischarge day 30 to describe patient characteristics and to evaluate postsurgical pain management and outcomes. Postsurgical opioid pain management was defined as the administration of parenteral or oral opioid analgesics on or after the procedure date and before discharge. The following opioid analgesics commonly used in the surgical setting within MHHS were considered: codeine, fentanyl, hydrocodone, hydromorphone, meperidine, methadone, morphine, oxycodone, and propoxyphene. Once the rate of opioid utilization was determined, all subsequent analyses were limited to those patients who received postsurgical opioids. All patients in the study cohort had a primary inpatient surgical procedure and hence associated charges or costs. ADE identification and risk factor assessment. The frequency of ADEs was estimated by identifying specific ICD-9 diagnoses (Table 2) within the EPSI database, including gastrointestinal (GI), respiratory, genitourinary (GU), CNS, and other ADEs (postoperative bradycardia, fall from bed, rash or itching). As ICD-9 coding within large databases typically has some variability, three-digit codes were included to capture a broader group of codes; where specificity was required, four- and five-digit codes were used. Patients were then placed
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
1557
PRACTICE REPORT Postsurgical opioids
into cohorts based on whether they had experienced an ADE or not. The characteristics considered for the ADE risk factor assessment included age, sex, opioid use prior to surgery, and other comorbidities based on their relevance to medication use and ADE risk; this approach was consistent with previous research in this area17 (Table 3). Multivariate logistic regression was used to evaluate odds ratios (ORs) and accompanying 95% confidence intervals (CIs) and to identify factors significantly associated with ADE risk in the overall population and by sex. Within the multivariate model, the occurrence of an ADE was specified as the dependent variable, with risk factors specified as the predictor variables. Only significant variables (p < 0.05)
were retained in the final model and reported. Outcomes assessment. Outcomes after ADEs evaluated in this study included LOS, total hospitalization costs, 30-day readmission rates, inpatient mortality, and outlier status. LOS was defined as the time in days from the index admission date to the discharge date. Total hospitalization cost was measured from the admission index date to the discharge date and was based on charges for the following: operating room, charge room, charge supplies, cardiology services, respiratory and pulmonary services, emergency services, radiology services, laboratory services, pharmacy services, and other ancillary services. To provide a better estimate of true costs, 2011 Medicare
Table 1.
Postsurgical Opioid Use in Base Study Population, by Procedurea
Procedure Laparoscopic cholecystectomy Total abdominal hysterectomy Laparoscopic gastric bypass Hip replacement, total Hip fracture reduction with fixation Open colectomy, partial Other partial gastrectomy Laparoscopic colectomy, partial Laparoscopic total abdominal hysterectomy Laparoscopic assisted vaginal hysterectomy Open cholecystectomy Laparoscopic gastric restrictive procedure Open gastric bypass Ileostomy reversal Hip replacement, partial Open colectomy, total Laparoscopic colectomy, total Mastectomy Abdominoplasty Hip fracture reduction without fixation All procedures
ICD-9 Procedure Code(s) 51.23 68.4, 68.49 44.38 81.51 79.35 45.7, 45.71–45.79 43.89 17.31–17.39
% Receiving Opioids
1,730 99.7 938 99.9 665 99.7 634 100 523 99.8 482 100 285 100 251 100
68.41
213
100
68.31 51.22
139 111
100 100
44.95 44.39 46.51, 46.52 81.52 45.8, 45.82, 45.83 45.81 85.4, 85.41–85.48 86.83 79.25 . . .
ICD-9 = International Classification of Diseases, Ninth Revision. Not applicable.
a
b
1558
No. Patients Identified
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
91 100 68 100 55 100 49 98.0 40 100 5 100 4 100 2 100 0 . . .b 6285 99.8
charge-to-cost ratios were applied to the charge data. Readmission rates were based on documentation of subsequent admission to a participating hospital within 30 days of the discharge date for the index procedure. Inpatient mortality was defined by patient discharge status. Additionally, patients were identified as LOS or cost outliers if their LOS or costs were greater than one standard deviation above the corresponding mean values for the overall MHHS population. A matched analysis using propensity scores was employed to control for underlying differences in demographic and patient characteristics between patients who experienced ADEs and those who did not. A propensity score is a composite score representing the probability (based on baseline characteristics) of patients falling into comparison groups of interest (in this case, patients with or without ADEs). Using this probability to adjust the estimated effect of adverse events on outcomes has the advantage of creating a “quasirandomized” experiment. For this study, propensity scores were obtained from procedure-specific logistic regression models that predicted the probability of having an adverse event. The selection of factors for inclusion in the risk assessment model was based on previous literature and included age, sex, Charlson Comorbidity Index score, obesity, degenerative joint disease (DJD), and prior opioid use. Patients with ADEs were matched by propensity score to patients without ADEs using a 1:1 ratio with the technique of nearest available match for the propensity scores. A narrow caliper (± 0.001 S.D.) was used to ensure that good matching was achieved and that the covariates were well balanced across the groups. Prematch and postmatch baseline demographic and patient characteristics were evaluated to identify any imbalance in the distribution of potential confounders between the
PRACTICE REPORT Postsurgical opioids
cohorts. Prematch comparisons between cohorts were conducted using t tests for continuous variables and chi-square tests for categorical variables. Postmatch comparisons were done using paired t tests for continuous variables and McNemar’s test for categorical variables. Matched inferential comparisons were conducted for the outcomes analysis. Wilcoxon signed rank tests were used to assess the differences in LOS and total hospitalization costs. Rate ratios and corresponding 95% CIs for LOS were generated using negative binomial regression models, a method appropriate for modeling count variables that accounts for overdispersion. Cost ratios and 95% CIs for total hospitalization costs were generated by a generalized linear model approach with a log-link function and gamma distribution. The 30-day readmission and inpatient mortality rates were evaluated using McNemar’s tests. Relative risk values and corresponding 95% CIs were generated via logistic regression models. All statistical analyses tested a two-sided hypothesis of no difference between cohorts at a significance level of 0.05 and were carried out using SAS statistical software, version 9.2 (SAS Institute Inc., Cary, NC). Results A total of 51,298 patients were identified in the initial MHHS data set; of these, 44,251 (86.3%) did not receive a surgical procedure of interest. Another 3,149 patients (6.1%) were excluded because they were younger than 18 years of age, and 649 (1.3%) were excluded due to incomplete capture of medication data (these reported criteria were not mutually exclusive). After application of all the inclusion and exclusion criteria, a total of 6,285 surgical patients were considered eligible for inclusion in the study population. The mean age of the final study population was 52.2 years, and 72.1% (n = 4,522) were female. Among all the surgical
procedures evaluated, laparoscopic cholecystectomy, total abdominal hysterectomy, laparoscopic gastric bypass, and total hip replacement were the most common procedures and accounted for 63% of the study population (Table 1). Nearly all patients (n = 6,274, 99.8%) in the base study population received postsurgical opioids; financial records could not be used to evaluate reasons why the remaining 11 patients did not receive postsurgical opioids. Among patients who received postsurgical opioid-based regimens, 11.0% (n = 689) experienced a documented ADE. GI ADEs were the
most common, occurring in 6.3% (n = 396) of patients. Less common ADEs included respiratory (n = 182, 2.9%), GU (n = 94, 1.5%), CNS (n = 38, 0.6%), and other (n = 57, 0.9%) ADEs. Results of the risk factor evaluation are presented in Table 4. In the overall population, patients 65 years of age or older were twice as likely to have an ADE compared with younger patients. This association persisted by sex: Older men were 50% more likely than younger men to experience an ADE, while older women had a greater than twofold increased risk relative to younger women. In
Table 2.
Diagnosis Codes Used to Identify Postsurgical Opioid-Related Adverse Drug Events in Study Populationa Adverse Event
ICD-9 Code
Respiratory Bradypnea 786.09 Pulmonary insufficiency 518.5 Respiratory complications 997.3 Asphyxia 799.01 Hypoxemia 799.02 Gastrointestinal (GI) Constipation 564.09 Constipation, narcotic related E937.9 Dizziness/vertigo 386.2 Dry mouth 527.7 Ileus, postoperative 997.4 Paralytic Ileus 560.1 Nausea with vomiting 787.01 Nausea with vomiting after GI surgery 564.3 Central nervous system Nervousness 799.2 Delirium 780.09 Confusion, postoperative 293.9 Confusion due to condition classified elsewhereb 293 Altered mental status 780.97 Cerebral hypoxia 997.01 Genitourinary system Urinary retention 788.2 Oliguria 997.5 Other Bradycardia, postoperative 997.1 Rash/itching 698.9 Drugs causing adverse effects in therapeutic use E935.2 Fall from bed E884.4 ICD-9 = International Classification of Diseases, Ninth Revision. Confusion due to ICD-9–coded condition.
a
b
Am J Health-Syst Pharm—Vol 71 Sep 15, 2014
1559
PRACTICE REPORT Postsurgical opioids
Table 3.
Demographic and Clinical Factors Included in Assessment of Risk of Postsurgical Opioid-Related Adverse Drug Eventsa Factor
ICD-9 Diagnosis Code(s)
Age . . .b Sex ...b Opioid use prior to surgery ...b Obesity 278 DJD 715 COPD 490, 493 Asthma 493 Pulmonary hypertension 416.18 Congestive heart failure 428.0 Benign prostatic hypertrophy 600 Coronary atherosclerosis 414.0 Cardiac dysrhythmias 427, 427.89, 427.9 Hypertension 401, 401.1, 401.9 Dementia 290, 290.1, 294, 294.1, 294.11, 294.2 Depression 296, 296.2 Diabetes mellitus 249, 250 Irritable bowel syndrome 564.1 Regional enteritis 555.0, 555.1, 555.9 Diverticulitis 562 Ulcerative colitis 556 GERD 530.11, 530.81 a ICD-9 = International Classification of Diseases, Ninth Revision, DJD = degenerative joint disease, COPD = chronic obstructive pulmonary disease, GERD = gastroesophageal reflux disease. b Not applicable.
general, men had a higher ADE risk than women. Comorbidities associated with increased ADE risk in the overall population included chronic obstructive pulmonary disease (COPD), cardiac dysrhythmia, regional enteritis, diverticulitis, and ulcerative colitis. While not statistically significant, there was some evidence of an association between ADE risk and asthma, diabetes mellitus, and depression. Among men, we found that patients with benign prostatic hyperplasia (BPH) had a greater than fivefold increased risk of ADEs compared with patients without BPH. Interestingly, the presence of DJD was associated with a lower risk of ADEs in the overall population, though this association may have been confounded by the possibility that these patients were managed with greater vigilance in the postsurgical setting. In addition to demographics and comorbidities, postsurgical opioid use was significantly associated with ADE risk in the overall population.
Table 4.
Frequency of Risk Factors for Opioid-Related Adverse Drug Events (ADEs) in All Patientsa No. (%) Patients Risk Factor Age of ≥65 yr Male sex Obesity DJD COPD Asthma Atherosclerosis Cardiac dysrhythmias Hypertension Diabetes mellitus Regional enteritis Diverticulitis Ulcerative colitis GERD Dementia Depression Prior opioid use
Without ADE (n = 5585)
With ADE (n = 689)
All (n = 6274)
1263 (22.6) 1440 (25.8) 1554 (27.8) 691 (12.4) 47 (0.8) 258 (4.6) 308 (5.5) 103 (1.8) 2268 (40.6) 1027 (18.4) 23 (0.4) 257 (4.6) 20 (0.4) 810 (14.5) 133 (2.4) 61 (1.1) 1845 (33.0)
290 (42.1) 312 (45.3) 134 (19.4) 72 (10.7) 22 (3.2) 40 (5.8) 74 (10.7) 81 (11.8) 310 (45.0) 166 (24.1) 12 (1.7) 87 (12.6) 9 (1.3) 105 (15.2) 39 (5.7) 13 (1.9) 314 (45.6)
1553 (24.8) 1752 (27.9) 1688 (0.93) 763 (12.2) 69 (1.1) 298 (4.7) 382 (6.1) 184 (2.9) 2578 (41.1) 1193 (19.0) 55 (0.9) 344 (5.5) 29 (0.5) 915 (14.6) 172 (2.7) 74 (1.2) 2159 (34.4)
Odds Ratio (95% CI)
p
2.02 (1.65–2.47) 2.00 (1.69–2.38) 0.93 (0.74–1.16) 0.73 (0.55–0.97) 2.34 (1.35–4.07) 1.43 (0.99–2.06) 1.06 (0.79–1.43) 5.39 (3.92–7.42) 0.92 (0.77–1.10) 1.21 (0.99–1.49) 4.30 (2.05–9.04) 2.35 (1.79–3.09) 3.77 (1.65–8.63) 1.10 (0.86–1.39) 1.37 (0.92–2.05) 1.88 (0.98–3.60) 1.27 (1.07–1.52)
