Exposure to High Risk Medications is Associated with Worse Outcomes in Older Veterans with Chronic Pain

Abstract

Background

Chronic pain is common, costly, and leads to significant morbidity in older adults, yet we have limited data on medication safety. We sought to evaluate the association of incident High Risk Medication in the Elderly (HRME) with mortality, emergency department (ED) or hospital care among older adults with chronic pain.

Methods

A retrospective Veterans Health Administration cohort study was conducted examining older Veterans with chronic pain diagnoses and use of incident HRME (opioids, skeletal muscle relaxants, antihistamines, and psychotropics). Outcomes evaluated included all-cause mortality, ED visits, or inpatient hospital care. Descriptive statistics summarized variables for the overall cohort, the chronic pain cohort, and those with and without HRME. Separate generalized linear mixed-effect regression models were used to examine the association of incident HRME on each outcome, controlling for potential confounders.

Results

Among 1,807,404 Veterans who received VA care in 2005–2006, 584,066 (32.3%) had chronic pain; 45,945 Veterans with chronic pain (7.9%) had incident HRME exposure. The strongest significant associations of incident HRME were for: high-risk opioids with all-cause hospitalizations (OR 2.08, 95%CI 1.95–2.23); skeletal muscle relaxants with all-cause ED visits (OR 2.62, 95%CI 2.52–2.73) and mortality (OR 0.80, 95%CI 0.74–0.86); antihistamines with all-cause ED visits (OR 2.82 95%CI 2.72–2.95); and psychotropics with all-cause hospitalizations (OR 2.15, 95%CI 1.96–2.35).

Conclusions

Our data indicate that incident HRME is associated with clinically important adverse outcomes in older Veterans with chronic pain and highlight the importance of being judicious with prescribing certain classes of drugs in this vulnerable population.

INTRODUCTION

Chronic pain is highly prevalent in all adult groups ¹ with over 50% of the US population age 65 and older reporting bothersome pain in the last month ². Among primary care patients in the Veterans Health Administration (VHA), up to 50% of male and 75% of female patients report chronic pain ^3–5. Pain is associated with poorer self-reported health status, higher levels of emotional distress, decreased social and physical activities ^3,6, and greater use of healthcare resources ^7,8. While pain is common, costly, and leads to significant morbidity, older adults are not well represented in clinical trials ⁹; therefore, we have limited evidence to inform decisions especially with respect to medication safety in this population ¹⁰.

Older adults with chronic pain have high rates of analgesic use ^11,12 and are also highly susceptible to adverse effects (ADEs) of analgesic treatments ¹⁰. A meta-analysis of observational studies found that the odds of being hospitalized for ADE-related conditions are four times higher for older compared with younger individuals. Up to 88% of ADE-related hospitalizations in older adults are potentially preventable ¹³. Compared with younger adults, older adults are twice as likely to present to emergency departments for ADE (over 177,000 emergency visits each year) and nearly seven times more likely to be hospitalized after an emergency visit ¹⁴. As highlighted by the Institute of Medicine¹⁵, preventing ADEs in older adults is a public health and patient safety priority.

The risk of ADEs increases in older adults due to a unique combination of age-related physiologic changes including altered drug absorption, decreased renal excretion, multimorbidity, polypharmacy, and functional impairments ^10,16. Exposure to high-risk medications for the elderly (HRME), a Healthcare Effectiveness Data and Information Set (HEDIS) quality measure, was developed by the National Committee on Quality Assurance (NCQA) in 2006 to address this concern. The HEDIS HRME measure ¹⁷ was developed by an expert panel and includes drugs from the Beers criteria ¹⁸ for potentially inappropriate prescribing in the elderly that have been previously identified as being high risk for potentially severe outcomes¹⁹. This measure is now used to benchmark the quality of medication management in older adults enrolled in Medicare and other managed care plans^20,21. Older adults with chronic pain may be at high risk for ADEs related to HRME such as opioids ²², skeletal muscle relaxants ²³, antipsychotics ²⁴, and sedating antihistamines, which are often prescribed in patients with pain although the evidence for their effectiveness and safety is lacking.

The objective of this study is to evaluate the association of incident HRME with mortality, emergency department (ED) or hospital care among older adults (65+) with chronic pain. To accomplish these objectives, data were used from a Department of Veterans Affairs (VA) nationwide cohort of older adults with chronic pain. We hypothesized that, after adjusting for potential confounders, older adults with chronic pain and incident HRME will be more likely than those without incident HRME to experience adverse outcomes after adjusting for potential confounders.

METHODS

For this population-based cohort study, we used data from the VA Health Care System administrative and clinical databases. These databases are the repositories of clinical data from more than 150 VA hospitals and 850 outpatient clinics. Details regarding this study cohort have been previously published ^25,26.

Ethics Statement

The Institutional Review Board of the University of Texas Health Science Center at San Antonio approved this study and granted a waiver of informed consent.

Study Design, Setting, and Sample

The cohort consisted of Veterans 65 years of age and older on October 1, 2005, and who received VA care (n = 1,807,404) during Fiscal Year (FY) 2005 (October 1, 2004– September 30, 2005) and FY 2006 (October 1, 2005– September 30, 2006). From that population, the sample was restricted to Veterans who had diagnoses of chronic pain [neuropathic, nocireceptive, mixed] defined using ICD-9-CM codes ²⁷ at least two times, 7 or more days apart during FY2004-2006 (Appendix 1). To examine incident use, individuals who had prior exposure to outpatient prescriptions for any HRME in FY05 were excluded from the analyses.

Data Sources

We used inpatient and outpatient demographic, healthcare utilization, and comorbidity data from the VA National Patient Care Database. Pharmacy data extracted from VA Pharmacy Benefits Management (PBM) dataset and vital status information obtained from the Vital Status file, were described in greater detail in prior publications ²⁶. Encrypted patient identifiers linked information across these databases.

Outcomes

The outcome measures examined were all-cause mortality, ED visits, or inpatient hospital care. Outcomes were evaluated from initial HRME exposure to one year after the index date. Secondary outcomes were inpatient admissions or ED visits, due to falls or non-spine fractures. Both primary and secondary outcomes were selected based on clinical importance and as potential surrogate markers for falls and fractures.

Independent Variables

We evaluated the four types of most commonly ²⁵ used HEDIS HRME: opioids, skeletal muscle relaxants, antihistamines, and psychotropics. Drug exposure was defined using the VA Product name, which identifies any dose or formulation of that drug (see Table 1). We identified incident HRME by drug category. We attempted to evaluate incident use of indomethacin and oral ketorolac, however, the sample size (n= 925) was too small to evaluate our outcomes of interest.

Table 1.

High Risk Medications in Older Adults with Chronic Pain

Drug Group	Drugs Included	Concerns Regarding Use
Opioids	Propoxyphene, meperidine, pentazocine	Confusion, falls, fractures, dependency, withdrawal
Skeletal muscle relaxants	Methocarbamol, cyclobenzaprine, carisoprodol, chlorzoxazone, metaxalone, orphenadrine	Anticholinergic adverse effects, excessive sedation, and weakness; questionable effectiveness
Antihistamines	Diphendydramine, hydroxyzine, promethazine, cyproheptadine, dexchlorpheniramine, tripelennamine	Confusion and sedation, anticholinergic adverse effects
Psychotropics	Diazepam, chlordiazepoxide, flurazepam, Thioridazine, meprobamate, barbiturates	Excessive sedation, falls, central nervous system and extrapyrimidal side effects, dependency

Covariates

We identified patient demographic characteristics (i.e., age, gender, race/ethnicity) between FY2004 and FY2006. Race/ethnicity was categorized as white, black, Hispanic (of any race), other, and missing. Missing demographic data on race/ethnicity are common in VA files; however, several years of data alleviate this problem. We examined socioeconomic status based on the VA means test; prior studies have shown that income under the VA poverty limit is associated with exposure to HRME ^28–30.

Greater disease burden -- defined by more physical and psychiatric comorbidities, prior health care utilization, and more prescribed medications -- place patients at higher risk for adverse outcomes ^31,32. We used ICD-9-CM codes from VA inpatient and outpatient data (FY04-05) to identify individuals with physical and psychiatric conditions using the Selim comorbidity indices ³³. The Selim measures consist of a continuous count of up to 30 physical disease diagnoses (including stroke, hypertension, diabetes, cardiovascular disease, peripheral vascular disease, osteoarthritis) and a count of up to six mental health diagnoses (including schizophrenia, bipolar disorder, depressive disorder, posttraumatic stress disorder, substance abuse disorder, and anxiety disorders). These comorbidity measures were selected because they were developed and validated in a VA population. We also examined prior healthcare utilization including any emergency department visits or hospital admissions, as well as the number of primary care visits in FY 2005. Finally, we included the number of unique medication classes received in FY 2005.

Statistical Analyses

Descriptive statistics were used to summarize our cohort. Chi-square tests were used to test the relationship between categorical variables and student’s t-test to test the relationship for continuous variables. We used separate generalized linear mixed-effect regression models to examine the association of the incident HRME on each of our [binary] outcomes with the primary medical center as a random effect after adjusting for the variables listed in Table 3.

Table 3.

Chronic Pain Cohort: Characteristics Among Subjects With and Without Incident HRME

Characteristic	No Incident HRME (n = 538,121) n (%)	Incident HRME (n = 45,945) n (%)	P-value
Age groups
65–74	234,643 (43.6)	23,270 (50.7)	<0.001
75–84	250,906 (46.6)	19,211 (41.8)
85+	52,573 (9.8)	3,464 (7.5)
Male	527,637 (98.1)	44,474 (96.8)	<0.001
Race/Ethnicity
White	370,212 (68.8)	33,710 (73.4)	<0.001
African American	37,830 (7.0)	4,362 (9.5)
Hispanic	17,646 (3.3)	2,332 (5.1)
Other/Missing	112,433 (20.9)	5541 (12.1)
Under the poverty limit	349,733 (65.0)	36,617 (79.7)
Selim Physical Comorbidity Index
0–1	50,088 (9.3)	2,751 (6.0)	<0.001
2–3	256,449 (47.7)	17,016 (37.0)
4–5	168,587 (31.3)	16,292 (35.5)
6+	62,997 (11.7)	9,886 (21.5)
Selim Psychiatric Comorbidity Index
0	454,350 (84.4)	34,405 (74.9)	<0.001
1	65,591 (12.2)	8,288 (18.0)
2+	18,180 (3.4)	3,252 (7.1)
Number of unique drugs
0–5	207,216 (28.5)	8,869 (19.3)	<0.001
6–8	142,062 (26.4)	9,976 (21.7)
9–11	94,365 (17.5)	9,754 (21.2)
12+	94,478 (17.6)	17,346 (37.8)
Outcomes
Death within 1 year	24,869 (4.6)	3,817 (8.3)	<0.001
ED visits	88,616 (16.5)	19,473 (42.4)	<0.001
ED visits related to falls/fractures	2,677 (.50)	3,344 (0.57)	<0.001
Hospitalization	32,455 (6.0)	8,311 (18.1)	<0.001
Hospitalization related to falls/ fractures	658 (0.1)	164 (0.4)	<0.001

Statistical significance was defined as a two-tailed p value of ≤ 0.05. STATA 13 (College Station, Texas) was used for all analyses.

RESULTS

Baseline Characteristics

Of the 1,807,404 eligible veterans identified in FY 2006, 584,066 met inclusion criteria for chronic pain and were included in this cohort (see Figure 1). Table 2 shows the difference between the chronic pain cohort and the rest of the cohort. The chronic pain group is more likely to be 85 years of age or older and has a greater number of physical and psychiatric comorbid conditions. Among those with chronic pain, 7.9% had exposure to incident HRME compared with 4% exposure to incident HRME in the non-pain group.

Figure 1.

Derivation of study cohort

Table 2.

Description of Chronic Pain Group vs. Those Without Chronic Pain

Characteristic	Chronic Pain (n =584,066) n (%)	Without Chronic Pain (n =1,223,338)	P-value
Age groups
65–74	257,913 (44.2)	561,992 (45.9)	<0.001
75–84	270,116 (46.3)	558,343 (45.6)
85+	56,037 (9.6)	103,003 (8.4)
Male	572,111 (98.0)	1,206,167 (98.6)	<0.001
Race/Ethnicity
White	403,922 (69.2)	788,732 (64.5)	<0.001
African American	42,192 (7.2)	70,691 (5.8)
Hispanic	19,978 (3.4)	36,309 (3.0)
Other/Missing	117,974 (20.2)	327,606 (26.8)
Below the poverty limit	386,350 (66.2)	673,324 (55.0)	<0.001
Selim Physical Comorbidity Index
0–1	52,839 (9.1)	461,005 (37.7)	<0.001
2–3	273,465 (46.8)	583,124 (47.7)
4–5	184,879 (31.7)	151,606 (12.4)
6+	72,883 (12.5)	27,603 (2.3)
Selim Psychiatric Comorbidity Index
0	488,755 (83.7)	1,089,177 (89.0)	<0.001
1	73,879 (12.7)	109,909 (9.0)
2+	21,432 (3.7)	24,252 (2.0)
Number of unique drugs
0–5	216,085 (37.0)	662,815 (54.2)	<0.001
6–8	152,038 (26.0)	313,917 (25.7)
9–11	104,119 (17.8)	150,710 (12.3)
12+	111,824 (19.2)	95,896 (7.8)
Incident HRME	45,945 (7.9)	48,739 (4.0)	<0.001

In the chronic pain cohort, nearly half (44.2%) were between 65 and 74 years, 46.3% were between 75 and 84 years, and 9.6% were 85 years of age and older. The chronic pain cohort was primarily male (98.0%), white (69.2%), and nearly half had 2–3 physical comorbidities (46.8%). The majority had no psychiatric comorbidities (83.7%). During the 1-year study period, 28,686 (4.9%) died, 40,766 (7.0%) had one or more hospital admissions, and 108,089 (18.5%) received emergency care.

Table 3 lists baseline descriptive characteristics of the chronic pain cohort for those who did and did not receive incident HRME in FY2006 (n=584,066). Osteoarthritis, low back pain, and diabetes were the most frequently reported comorbid conditions regardless of HRME status. The frequencies of outcomes (mortality, ED, and hospitalizations) by incident HRME exposure are shown in Table 3.

In the chronic pain cohort among those who received incident HRME (n=45,945), the incident drug exposure by class of drug was: high-risk opioid use (n=3,609, 7.9%), muscle relaxant use (n=6,932, 15.1%), antihistamine use (n=6,544, 14.2%), and psychotropic use (n=2,252, 4.9%).

Results of Multivariable Regression Models

Table 4 summarizes results for the association between incident HRME drugs and the specified outcomes. We found statistically significant associations of incident HRME with high-risk opioids for all adverse outcomes, especially all-cause hospitalizations (OR 2.08, 95%CI 1.95–2.23), and ED visits related to falls or fracture (OR 2.15, 95%CI 1.78–2.62). The association between exposure to skeletal muscle relaxants and all adverse outcomes was statistically significant, except for hospitalizations related to falls or fracture; the strongest association was for all-cause ED visits (OR 2.62, 95%CI 2.52–2.73). While the association between incident exposure to antihistamines was statistically significant with all adverse outcomes, the strongest was for all-cause ED visits (OR 2.82 95%CI 2.72–2.95). The association between incident exposure to psychotropic medications and adverse outcomes was statistically significant, except for hospitalizations related to falls or fractures (OR 1.09, 95%CI 0.56–2.12). Curiously, exposure to muscle relaxants had a significant association with lower all-cause mortality (OR 0.80, 95%CI 0.74–0.86).

Table 4.

Multilevel Regression Models that Examined the Associations Between Incident HRME Drug Class and Outcomes in Chronic Pain Cohort

	All-Cause Mortality	ED	ED-falls/fractures	Hospitalization	Hospitalization- falls/fracture
Adjusted OR (95% CI)
Opioids	1.13 (1.03–1.24)	1.94 (1.84–2.06)	2.15 (1.78–2.62)	2.08 (1.95–2.23)	2.04 (1.39–3.00)
Skeletal muscle relaxants	0.80 (0.74–0.86)	2.62 (2.52–2.73)	1.81 (1.57–2.09)	1.56 (1.48–1.65)	1.10 (0.77–1.58)
Antihistamines	1.78 (1.68–1.89)	2.83 (2.72–2.95)	1.60 (1.39–1.84)	2.22 (2.12–2.32)	2.03 (1.57–2.62)
Psychotropics	1.15 (1.01–1.32)	2.02 (1.87–2.19)	1.47 (1.11–1.96)	2.15 (1.96–2.35)	1.09 (0.56–2.12)

DISCUSSION

Our study is the first, to our knowledge, to describe the associations of incident HRME with potentially ADE-related acute care in a population-based sample of older Veterans with chronic pain. Prior work from this study found much lower incident HRME exposure in the full cohort of older VA patients. These findings are relevant because they come from a nationally representative sample, allowing for extrapolation to the entire older veteran population and potentially other populations of older men in the United States. Further, our results highlight the importance of being judicious with prescribing certain classes of drugs in older adults with chronic pain.

The associations between incident HRME and ED/hospital visits are consistent with findings from Albert et al. ³⁴ who reported prevalent HRME use with nearly a 2-fold increase in hospital admissions in a retiree cohort receiving employer-based drug benefits ³⁴. However, Albert et al did not specifically evaluate a chronic pain population and examined prevalent HRME rather than incident HRME, which may result in misclassification and potentially underestimate this association.

All four of the HRME drug categories evaluated in this study are frequently prescribed (with the exception of propoxyphene which is no longer available) and often considered benign by both prescribers and patients. Surprisingly, we found that skeletal muscle relaxants were associated with a reduced risk of mortality. It is possible that physicians prescribe muscle relaxants to the most functional patients, or those who request these medications to maintain an active lifestyle, however, underlying mechanisms for this association is unclear. Both skeletal muscle relaxants and antihistamines were associated with significantly higher odds of potentially ADE-related acute care particularly ER visits. Given the relative paucity of data with regard to medication safety (as well as potential efficacy) in older adults, our results, for all four classes of incident HRME, provide evidence for prescribing these medications with caution in older Veterans with chronic pain.

The challenge that clinicians face is choosing a medication that will reduce the pain symptoms, safely, in older populations. Chronic use of non-steroidal anti-inflammatory drugs (NSAIDs) is not recommended in older adults for well-known/established gastrointestinal, renal, and cardiovascular toxicity ^35,36. Capturing NSAID use is a recognized challenge since this class of medications is often purchased over the counter and not always through the VA pharmacy. Our small sample size for those receiving indomethacin and oral ketorolac precluded evaluation of our outcomes of interest. Adjuvant therapies, such as tricyclic antidepressants, for example, are often tried, however, limited evidence for its safety and efficacy exists specifically in older adults, and the tertiary tricyclic antidepressants (including amitriptyline) are not included in the HEDIS measure. Given known (and ongoing research that is uncovering) adverse effects of pharmacologic therapies, further emphasis on non-pharmacological approaches (i.e. physical therapy, cognitive behavioral therapy, transcutaneous electrical nerve stimulation, mindfulness, meditation, relaxation, guided affective imagery, biofeedback, prayer, and music therapy) is encouraged ^7,27,36. Ultimately, future research must focus on determining tolerable and safe combinations of pharmacological and non-pharmacological management approaches that are effective and sustainable in older adults with chronic pain.

Important strengths of our study include large population-based sample of Veterans who receive care from an integrated health care system and the use of EMR data that has been shown to enhance ADE detection ³⁷. Another important strength to highlight is the use of Selim measures of comorbidity to provide a more comprehensive assessment of disease burden (especially the psychiatric component). This is particularly important given the known relationship between chronic pain and psychiatric conditions ³⁸.

This study has several limitations. The outcomes selected were potentially caused by ADEs however we did not validate the likelihood of these encounters being ADE- related by manually reviewing each medical record. Another potential source of underestimation of these problems is our inability to capture potentially ADE-related acute care that occurred outside the VA system. While we evaluated high-risk opioids in this analysis, propoxyphene (on the list of included opioids) has since been removed from the market. Many patients who had been receiving this opioid could have been transitioned to another opioid not listed in Table 1; these opioids (for example, oxycodone) are not included in the HEDIS HRME or Beer’s criteria, and therefore, not the focus of this paper. Despite this, our results do show increased adverse outcomes in chronic pain patients using this class of medication. We would anticipate, based on class effect, that given the wide-spread use of opioids such as oxycodone, the outcomes we report for opioids are potentially an underestimate. This should be evaluated in future research. The sample consisted mostly of community dwelling older male veterans so the results may not be extrapolated to older men who reside in long term care facilities, older females, or younger populations.

CONCLUSIONS

Our data indicate that incident HRME is associated with clinically important adverse outcomes in older Veterans with chronic pain. A better understanding of the outcomes associated with incident HRME in an older chronic pain population is a critical step towards making providers more aware of potentially harmful effects of medications. Further research is needed to develop intervention measures to reduce exposure to these high-risk medications. These are all necessary steps towards proposing effective, safe and cost conscious management approaches for older adults with chronic pain.