Opioid therapy is widely accepted for the treatment of moderate to severe pain. In 2012 alone, US physicians wrote 259 million prescriptions for opioid analgesics.1 It has been estimated that more than 4.3 million American adults regularly take an opioid in any given week.2 A substantial proportion (47%) of these opioid users have been taking opioids regularly for 2 years, and one-fifth have been taking them regularly for 5 years or longer.2 Although effective for pain management, side effects are common with opioid use, particularly constipation and other gastrointestinal symptoms. Estimates of the prevalence of opioid-induced constipation vary from 15% to 95%.3-5 Opioid-induced constipation results from the binding of opioids to µ-opioid receptors in the gastrointestinal tract, which reduces gut motility and secretions, leading to hard, dry stools.6 Rarely are patients tolerant of opioid-induced constipation,4 which may occur after a single opioid dose and is often refractory to treatment.4,6
Opioid-induced constipation negatively affects pain management and patients’ quality of life,7,8 which may result in increased healthcare utilization and costs.9-11 In a multinational survey of the impact of opioid-induced bowel dysfunction, constipation was the most common symptom, was most frequently reported as severe, and was ranked as the most distressing symptom.12 The majority of surveyed patients reported that the constipation had a moderate-to-great or a great effect on their quality of life, which led to poor adherence and/or discontinuation of their opioid therapy.12
Reduced adherence to opioid therapy could decrease the analgesic efficacy of pain management, which may lead to additional healthcare utilization and costs. Although healthcare resource utilization has been shown to increase after a patient begins opioid therapy for noncancer pain,13 little is known about the proportion of increased costs that can be attributed to constipation in this population. A study of patients who received opioid therapy for noncancer- or cancer-related pain showed that patients with constipation during therapy had significantly more hospital admissions, more emergency department visits, greater use of medical resources, and higher mean all-cause costs than those without constipation.11
It is important to quantify the potential effect on healthcare utilization and cost that constipation has on individuals who initiate opioid therapy to understand the economic burden of constipation in this population. This study was designed to assess healthcare utilization and costs, including all-cause, constipation-related, and pain-related costs, for patients with versus those without opioid-induced constipation during the 12 months after the initiation of opioid therapy for noncancer pain.
Data Source and Patient Identification
This retrospective cohort study was conducted using administrative claims data from the HealthCore Integrated Research Environment (HIRE). HIRE is a diverse longitudinal administrative claims repository containing data from commercial health plans in the Northeast, Midwest, South, and West regions of the United States. Researchers had access only to deidentified patient data, and patient anonymity and confidentiality were safeguarded in compliance with the Health Insurance Portability and Accountability Act. Institutional review board approval was not required for this observational study.
Inclusion and Exclusion Criteria
To be included in the study, patients had to have ≥1 prescription claims for an opioid (hydrocodone, propoxyphene, codeine, dihydrocodeine, fentanyl, hydromorphone, levorphanol tartrate, methadone, morphine, oxycodone, oxymorphone, tramadol, tapentadol, or meperidine) during the intake period between July 1, 2006, and June 30, 2013. The date for the first opioid prescription fill was considered the index date, and patients had to continuously take the opioid for ≥28 days starting on the index date to be included. In addition, patients in the study were required to be aged ≥18 years and have ≥6 months of continuous health plan eligibility before the index date (the preindex period) and ≥12 months of eligibility after the index date (the postindex period), which ended by June 30, 2014.
Patients were excluded from the study if they had a diagnosis of cancer at any point during the study period (January 1, 2006–June 30, 2014), which was determined by the existence of ≥2 claims within 60 days that specified a diagnosis of the same type of cancer.
Also excluded were patients who had ≥1 prescriptions for methadone and ≥1 medical claims with a diagnosis of addiction or substance abuse disorder during the study period. In addition, patients with ≥1 claims containing a diagnosis of inflammatory bowel disease (including Crohn’s disease or ulcerative colitis) were excluded, as were patients who had a diagnosis of constipation, a procedure related to constipation, a pharmacy claim for a constipation medication, or previous use of an opioid during the 6-month preindex period.
The study population was grouped into 2 cohorts (Figure). Cohort 1 consisted of patients who initiated opioid therapy and experienced constipation; cohort 2 comprised patients who initiated opioid therapy but did not experience constipation. The patients assigned to cohort 1 were required to have ≥1 medical claims with an International Classification of Diseases, Ninth Revision (ICD-9) diagnosis code for constipation, a pharmacy claim for a constipation medication, or a procedure related to constipation at any point during continuous opioid therapy or until a gap of ≥30 days occurred between prescription fills. (See Appendix Tables 1 and 2, for ICD-9 codes for constipation and pain.) Patients who were receiving opioid therapy but did not have constipation during the 12-month postindex period were assigned to cohort 2.
For all patients, continuous opioid use was defined as the time interval between filling the first prescription until discontinuing opioid therapy. Discontinuation was defined as not filling an opioid prescription within 30 days after exhausting the previously filled supply. The discontinuation date was calculated by using the fill date of the most recent opioid prescription and adding the number of supply days covered by that prescription. A patient might have discontinued opioid use and restarted it later, resulting in multiple opioid regimens during the 12-month postindex period (Figure).
The primary outcome was healthcare utilization and associated healthcare costs. Healthcare utilization was reported as constipation-related, pain-related, and all-cause. Constipation-related utilization referred to events with ≥1 claims containing a diagnosis, procedure, or medication ICD-9 code for constipation or constipation treatment. Pain-related utilization represented events with ≥1 diagnoses for a pain condition. All-cause utilization encompassed all healthcare utilization events, regardless of diagnosis or reason for visit. Utilization and costs were reported overall and by place of service, including physician office, other outpatient facility, emergency department, inpatient, skilled nursing facility, and pharmacy. Costs were adjusted for inflation (according to inflation rates provided by the Bureau of Labor Statistics14) and are reported in 2014 US dollars.
The strength of the index-prescribed opioid was categorized as strong or weak based on the mechanism of action and the affinity for binding with the μ-, κ-, and δ-opioid receptors.15 Weak opioids included codeine, tramadol, dihydrocodeine, tapentadol, hydrocodone, and propoxyphene. Strong opioids included fentanyl, hydromorphone, methadone, morphine, oxycodone, meperidine, oxymorphone, and levorphanol. The index opioid dose was converted to its oral morphine equivalent to allow comparison across opioid types.
To minimize selection bias, propensity scores were used to match each patient in cohort 1 to a patient in cohort 2 (in a 1:1 ratio). The propensity score for each patient was estimated as the probability of having constipation based on observed baseline characteristics. Prespecified variables used to balance the 2 cohorts included index month, index year, geographic region, health plan type, weak or strong opioid use at index, preindex Deyo-Charlson Comorbidity Index (DCI) score, specialty of prescribing physician, preindex all-cause healthcare resource utilization, and preindex all-cause healthcare costs. Logistic regression models were used to calculate the propensity scores, and a greedy matching algorithm was used to match the 2 cohorts.16 (Results of the matching procedure appear in the Appendix Table 3)
Statistical models were used to analyze differences in healthcare utilization and costs (all-cause and pain-related) and in opioid treatment patterns between the 2 cohorts. Healthcare costs were analyzed through generalized linear models with a gamma distribution and log link. Count variables, such as the number of visits and the length of inpatient stays, were analyzed via generalized linear models with a negative binomial distribution and log link. Analysis of differences in the number of patients with and without any event was done by logistic regression. The models were adjusted for their analogous preindex utilization or cost variable to adjust for the possibility that some residual differences in preindex costs could be contributing to the differences in postindex costs. For example, models used to assess differences in postindex pain-related costs were adjusted for preindex pain-related costs. Other baseline characteristics were not included, because successful balance of potential confounding variables was achieved by matching the propensity scores.
A sensitivity analysis was performed after excluding patients who were identified as having constipation based solely on a claim for constipation medication. Those patients were excluded because they may have been prescribed constipation treatment for prophylactic reasons but were not currently experiencing symptoms. Hence, only the patients who had a diagnosis of constipation or had undergone a constipation-related procedure were included in the sensitivity analysis.
A total of 390,874 patients met the inclusion criteria and were deemed eligible for study participation (18,008 in cohort 1; 372,866 in cohort 2; see Figure in the Appendix). After matching and successfully balancing all prespecified variables (P >.05), 17,384 patients were retained in each cohort. At the index date, the distribution of each cohort was similar in terms of age, sex, health plan type, and geographic region of residence. Preferred provider organizations were the most common health plan type (71.1% for both cohorts), and the majority of patients were from the Midwest and South (Table 1).
During the 6-month preindex period, the mean DCI score was 0.7 for both cohorts, and the distribution of prespecified comorbid conditions was similar for the 2 cohorts. The most common chronic disease comorbidities during preindex in each cohort were hypertension and diabetes mellitus.
Index Opioid Use
The distribution of the index opioid strength and type was similar for the study cohorts, as was the distribution of the prescribing physicians (Table 2).
The majority (75%) of patients in each cohort received a weak opioid at the index date, most frequently hydrocodone (43% in each cohort). Among the opioids classified as weak, propoxyphene was associated with the highest mean index dose (oral morphine equivalent) in both cohorts (64.3 mg in cohort 1; 62.8 mg in cohort 2). Oxycodone was the most common strong opioid at the index date, and methadone was associated with the highest mean dose (153.2 mg in cohort 1; 156.5 mg in cohort 2). In both cohorts, approximately 37% of index opioids were prescribed by family physicians, general practitioners, or internal medicine physicians.
12-Month Postindex Healthcare Resource Utilization and Costs
During the 12-month postindex period, the patients who had constipation had nearly twice the risk of having an all-cause inpatient hospitalization, emergency department visit, and office or other outpatient visit compared with those who did not (Table 3). Among the patients with constipation, 26.6% had a pain-related inpatient hospitalization, 17.6% had a pain-related emergency department visit, and 84% had a pain-related outpatient or office visit. After adjusting for the analogous preindex utilization, it was noted that patients with constipation had twice the risk of having a pain-related inpatient hospitalization as those without, 1.4 times the risk of having a pain-related emergency department visit, and 1.3 times the risk of having a pain-related outpatient or office visit.
The total mean adjusted overall costs (medical plus pharmacy) during the 12-month postindex period were $12,413 higher for the opioid users who had constipation than for those who did not (95% confidence interval [CI], $11,726-$13,116; Table 4).
Medical costs were $11,558 higher and pharmacy costs were $723 higher for patients with constipation than for those without. The mean adjusted plan-paid costs were $11,533 (95% CI, $10,855-$12,228) higher for patients with constipation than for those without, and the total mean adjusted patient-paid costs were $818 (95% CI, $767-$869) higher for patients who had constipation.
Compared with patients who did not have constipation, the total mean adjusted overall pain-related costs for patients with constipation were $6778 (95% CI, $6293-$7279) higher; mean plan-paid costs were $6619 (95% CI, $6114-$7143) higher, and patient-paid costs were $254 (95% CI, $228-$280) higher. Moreover, mean pain-related medical costs were $6894 higher and pharmacy costs were $270 higher for patients with constipation. The total overall constipation-related costs among patients using opioids were $4646 per patient (medical costs, $4585; pharmacy costs, $61); the total average plan-paid costs were $4424 per patient, and the total patient-paid costs were $222.
The sensitivity analysis included 6666 matched patients in the cohort of patients using opioids who had constipation, identified by a diagnosis of constipation or a constipation-related procedure, and patients using opioids who had no constipation. Although the degree of balance in covariates achieved by matching propensity scores was lower for the sensitivity analysis, no major deviations in estimates from the original analysis were observed for preindex demographics, comorbidities, medications, or type of index opioid among patients in cohort 1 (data not shown).
The 12-month postindex healthcare utilization pattern in the sensitivity analysis was similar to that in the original analysis, but the differences were of higher magnitude in the sensitivity analysis. All-cause healthcare utilization differences were more pronounced in the sensitivity analysis: 51.7% of patients in cohort 1 had ≥1 inpatient visits during the postindex period compared with just 23.5% of patients in cohort 2. Between-cohort differences of similar magnitude were noted in the sensitivity analysis for other types of utilization, including pain-related events.
The 12-month cost differences between cohorts showed similar patterns in the 2 analyses, again with differences in the sensitivity analysis being of greater magnitude. The average all-cause costs were $34,055 in cohort 1 and $15,776 in cohort 2, and pain-related costs were 2.5 times higher (on average) for patients in cohort 1 ($18,457 vs $7333 in cohort 2; adjusted mean difference, $11,042; 95% CI, $10,048-$12,091). The total overall constipation costs among patients using opioids were $9287 per patient, with the total average plan-paid costs being $8880, and the total patient-paid costs being $407 per patient.
In this retrospective, real-world study of 34,768 patients using opioids—17,384 with constipation and 17,384 without constipation—we found that patients who had constipation had significantly more inpatient hospitalizations, emergency department visits, and outpatient visits than those who did not have constipation. The mean adjusted differences between patients with and without constipation also were significant for plan-paid, patient-paid, and overall costs.
Although our results are consistent with those of previous claims analyses of healthcare utilization and cost burden of opioid-induced constipation, most of those analyses were not limited to patients with noncancer pain.9-11 Iyer and colleagues compared patients using opioids who had constipation with those who did not, and also concluded that healthcare utilization and costs were higher for patients with constipation in all categories, including hospitalization, office visits, emergency department visits, other outpatient services, and nursing facility care.11 Similar results were found in a claims analysis of patients with cancer pain using opioids who had constipation versus those who did not.9 However, the patient populations in these studies were much smaller than the sample in the current study. A small European study of the direct and indirect costs of opioid-induced constipation demonstrated that costs increased with severity and that patients with severe opioid-induced constipation had significantly higher costs than those with moderate, mild, or no constipation.10 The researchers concluded that opioid-induced constipation interferes with pain management.10
Although our results are consistent with previous findings, our study is unique in assessing healthcare utilization and cost of constipation among a large population of patients with noncancer pain who were using opioids. Moreover, our study patients were required to have no previous evidence of constipation, and we identified new-onset constipation after the initiation of opioid therapy during an active opioid regimen. This limited the possibility of the constipation having been caused by factors other than opioid therapy. Opioid users with constipation were older and had higher DCI scores than opioid users without constipation before matching. However, we elected to use matched pairs to achieve comparable groups for understanding the impact of opioid-induced constipation on the healthcare utilization and costs. Thus, by using propensity scores to match the opioid users with constipation to those without constipation on a comprehensive list of variables and confounders, we were able to estimate the excess burden of constipation on healthcare utilization and costs.
Our results highlight the importance of managing constipation in patients using opioids to improve the quality of pain treatment. Our results also suggest that advising patients on the possible side effects of opioid therapy—including constipation—and promptly diagnosing and treating opioid-induced constipation may help to reduce the burden and increased costs associated with opioid-induced constipation.
Potential limitations of the study should be noted. All patients in the analysis were members of large commercial US health plans. Therefore, the results may not be generalizable to patients with other types of insurance, uninsured patients, or patients in other countries.
Data were obtained from administrative claims, which are intended for reimbursement use rather than for research purposes.
Undetected coding errors or omissions might have been present.
Moreover, specific ICD-9 diagnosis codes did not exist for opioid-induced constipation, and the claims did not contain information on the use of over-the-counter medications, such as laxatives. This might have resulted in underestimation of the true economic burden of opioid-induced constipation.
It is possible that some patients in cohort 2 actually had constipation. The utilization and cost estimates for cohort 1 likely pertain more to patients with severe constipation who seek professional treatment than to all patients with constipation while taking opioid medication.
Future efforts may focus on understanding the economic burden by improving the identification of patients with constipation, including those who use over-the-counter laxatives.
Finally, several variables contribute to determining the dose of an opioid, including patient genetics, opioid tolerability, combination pain pharmacotherapy, multiple prescribers, and the underlying pain condition. Little information was available on the clinical drivers of the need for different opioid doses in specific clinical cases.
Patients using opioids with newly diagnosed constipation had significantly greater healthcare utilization and costs than patients using opioids who did not have constipation. Constipation-related costs accounted for approximately 16% of the total healthcare costs per patient during the 12-month study period. Patients being prescribed opioid therapy should be educated on possible side effects, including constipation. Prompt recognition and treatment of opioid-induced constipation may improve outcomes for patients with chronic pain and limit the costs and burden of untreated opioid-induced constipation.
Editorial support for the manuscript was provided by Cheryl Jones, BA, from HealthCore, Inc, and Regina E. Burris, PhD, and Diane DeHaven-Hudkins, PhD, from Complete Healthcare Communications, LLC. Research support for the study was provided by Suma Vupputuri from HealthCore, Inc.
This study was funded by AstraZeneca Pharmaceuticals LP.
Author Disclosure Statement
Dr Fernandes, Dr Datto, and Dr McLeskey are employees of AstraZeneca. Mr Kern, Dr Chen, and Dr Tunceli are employees of HealthCore, Inc, which received funding from AstraZeneca for this research.
1. Centers for Disease Control & Prevention. Opioid painkiller prescribing: where you live makes a difference. Updated July 1, 2014. www.cdc.gov/vitalsigns/opioid-prescribing/. Accessed April 15, 2016.
2. Parsells Kelly J, Cook SF, Kaufman DW, et al. Prevalence and characteristics of opioid use in the US adult population. Pain. 2008;138:507-513.
3. Benyamin R, Trescot AM, Datta S, et al. Opioid complications and side effects. Pain Physician. 2008;11(2 suppl):S105-S120.
4. Swegle JM, Logemann C. Management of common opioid-induced adverse effects. Am Fam Physician. 2006;74:1347-1354.
5. Moore RA, McQuay HJ. Prevalence of opioid adverse events in chronic non-malignant pain: systematic review of randomised trials of oral opioids. Arthritis Res Ther. 2005;7:R1046-R1051.
6. Sternini C. Receptors and transmission in the brain-gut axis: potential for novel therapies. III. μ-opioid receptors in the enteric nervous system. Am J Physiol Gastrointest Liver Physiol. 2001;281:G8-G15.
7. Bell T, Annunziata K, Leslie JB. Opioid-induced constipation negatively impacts pain management, productivity, and health-related quality of life: findings from the National Health and Wellness Survey. J Opioid Manag. 2009;5:137-144.
8. Panchal SJ, Müller-Schwefe P, Wurzelmann JI. Opioid-induced bowel dysfunction: prevalence, pathophysiology and burden. Int J Clin Pract. 2007;61:1181-1187.
9. Candrilli SD, Davis KL, Iyer S. Impact of constipation on opioid use patterns, health care resource utilization, and costs in cancer patients on opioid therapy. J Pain Palliat Care Pharmacother. 2009;23:231-241.
10. Hjalte F, Berggren A-C, Bergendahl H, Hjortsberg C. The direct and indirect costs of opioid-induced constipation. J Pain Symptom Manage. 2010;40:696-703.
11. Iyer S, Davis KL, Candrilli S. Opioid use patterns and health care resource utilization in patients prescribed opioid therapy with and without constipation. Manag Care. 2010;19:44-51.
12. Bell TJ, Panchal SJ, Miaskowski C, et al. The prevalence, severity, and impact of opioid-induced bowel dysfunction: results of a US and European Patient Survey (PROBE 1). Pain Med. 2009;10:35-42.
13. Kern DM, Zhou S, Chavoshi S, et al. Treatment patterns, healthcare utilization, and costs of chronic opioid treatment for non-cancer pain in the United States. Am J Manag Care. 2015;21:e222-e234.
14. Bureau of Labor Statistics; US Department of Labor. CPI inflation calculator. www.bls.gov/data/inflation_calculator.htm. Accessed April 15, 2016.
15. Freynhagen R, Geisslinger G, Schug SA. Opioids for chronic non-cancer pain. BMJ. 2013;346:f2937. Erratum in: BMJ. 2014;349:g4641.
16. Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. In: Proceedings of the Twenty-Sixth Annual SAS Users Group International Conference. Cary, NC: SAS Institute Inc; 2001. Paper 214-26.