Analyzing Analgesic Medications in Community-Dwelling Older Adults

Abstract

Objectives

Analyzing medication data for research purposes is complex, and methods are rarely described in the literature. Our objective was to describe methods of quantifying opioid and nonopioid analgesics and to compare the utility of five different analgesic coding methods when analyzing relationships between pain, analgesic use, and clinical outcomes. In this study, we used physical function as the outcome variable for its clinical relevance and its relationship to pain in older adults.

Design

Secondary analyses of baseline cross-sectional data from the Advanced Cognitive Training Interventions for Vital Elders (ACTIVE) study.

Setting

Community settings in six regions of the United States.

Subjects

A total of 2,802 community-residing adults older than age 65 years.

Methods

A medication audit was conducted. Analgesics were coded as any pain medication, counts (total analgesics, number of opioids and nonopioids), equianalgesics (oral morphine equivalents, oral acetaminophen equivalents), and dose categories. Adjuvant medications used to treat pain (e.g., tricyclic antidepressants and anticonvulsants) and low-dose aspirin typically used for cardiovascular conditions were excluded from these analyses. To examine the utility of these various approaches, a series of hierarchical regression models were conducted with pain and analgesics as predictors and physical functioning as the dependent variable.

Results

Eighty-one point nine percent of participants reported experiencing recent pain, but 26% reported analgesic use. Nonopioids were the most common drug class used. Models revealed that pain was significantly associated with worse physical function (β = –0.45, P = 0.001), after controlling for demographic and analgesic variables. Two basic drug coding methods (e.g., any pain medication, number of pain medications) were equivalent in their explanatory power (β = –0.12, P = 0.001) and were slightly stronger predictors of function than the more complex coding procedures.

Conclusions

Analgesic medications are important variables to consider in community-based studies of older adults. We illustrate several methods of quantifying analgesic medications for research purposes. In this community-based sample, we found no advantage of complex equianalgesic coding methods over simple counts in predicting physical functioning. The results may differ depending on the research question or clinical outcome studied. Thus, methods of analyzing analgesic drug data warrant further research.

Introduction

There is a high prevalence of pain in older adults [1,2]. Chronic diseases, including cardiovascular disease, degenerative joint disease, osteoporosis, cancer, and peripheral neuropathies are more prevalent in older adults and are often associated with persistent pain [3]. It is estimated that as many as 76% of community-dwelling elders experience persistent noncancer pain [4]. Untreated or ineffectively treated moderate to severe persistent pain has implications for older adults’ functioning and quality of life [5,6]. Thus, in community-based studies of older adults, many study participants may be experiencing pain. Given that pain has documented effects on physical, cognitive, and social functioning, it is important to consider the potential role of pain when investigating outcomes in community-based studies of aging [5,7].

Analgesic medications are the mainstay of pain management [6,8]. These medications reduce or relieve pain, but also have potential effects on cognition, gait and balance, nutrition, mood, and daily activity and functioning [9,10]. Thus, it is important to assess and analyze analgesic medication use in community-based studies of older adults.

Analyzing medication data for research purposes is a complex and time-consuming process that is rarely described in the literature. In community-based studies, medication review is often conducted using a “brown bag survey” in which participants are asked to bring all medications that they currently take to a study session. Drug names are standardized and coded to reflect the major class and subclass of drugs. This enables investigators to focus on specific drug classes, such as analgesics, and their relationship to clinical outcomes. Depending on the research question, analgesic use variables may be conceptualized differently in clinical research studies. In studies focused on pain, some ways that analgesic use can be considered include the following: 1) to understand the impact of pain on an outcome after adjusting for pain treatment such as analgesic use, 2) to understand the impact of pain treatment on an outcome, 3) to adjust for analgesic use (as a co-therapy) when examining the independent effect of another treatment on an outcome, 4) to investigate analgesic use as a mediator in the relationship between pain and an outcome, or 5) to investigate predictors of analgesic use. For researchers, the challenge is how to best quantify analgesic medications for analytical purposes.

In the pain and aging literature, several approaches to quantifying analgesic medications have been used. One common approach is to count the number of analgesics per person. This provides a simplistic view of analgesic use because it fails to consider the dose, number, and frequency of analgesic medications prescribed and taken. This variable is often dichotomized into an indicator of whether or not study participants are taking any analgesic drug (yes or no). A second approach is to consider subcategories of analgesics (e.g., opioid or nonopioid drugs). The numbers of opioids and nonopioid analgesic medications are summed and analyzed. A third approach is to convert analgesic medications to equianalgesic doses. Conceptually, this approach is preferable because the sum of any two analgesic drugs may have very different effects. For example, consider the case in which Person A takes acetaminophen and ibuprofen and Person B takes acetaminophen plus codeine and oral morphine. In both cases, the sum score equals two medications, yet these medication profiles reflect very different pain levels, analgesic effects, and potential side effects [9,10]. Computing equianalgesic doses standardizes drugs and doses to approximately equivalent analgesic effects. Equianalgesic dosing charts for opioids are commonly used in clinical practice to select appropriate doses when changing from one opioid to another or finding optimal drug combinations [8,11]. Opioid drugs are standardized to oral morphine equivalents (OMEs), with 30 mg by mouth as the referent. For nonopioid medications, equianalgesic conversion tables are relatively rare and difficult to ascertain, even for commonly used nonsteroidal anti-inflammatory drug (NSAID) medications. Acetaminophen is sometimes used as the referent for nonopioid drugs, but the computation of oral acetaminophen equivalents (OAEs) is difficult because there is not a definitive reference that equilibrates all nonopioid analgesics. In addition, most drugs in this category, acetaminophen and NSAIDs, belong to different classes of analgesics, raising questions about the applicability of this conversion. The equianalgesic approaches, both OME and OAE computations, have been used in studies of older adults residing in nursing homes [9,12,13] and of adults undergoing knee replacement surgery [14] to control for the effects of pain treatment. Finally, we propose an alternative method of quantifying nonopioid analgesic dose intensity that categorizes drugs into low, medium, or high doses, rather than equating doses.

We found no published studies that directly compared different approaches to quantifying analgesic drugs or that evaluated whether more complex approaches (morphine and acetaminophen equivalents, nonopioid analgesic dose intensity calculations) are more predictive of important clinical outcomes than a simpler approach (counting number of analgesics). We hypothesize that more complex coding methods that take into account dose intensity (e.g., morphine and acetaminophen equivalents or nonopioid dose intensity rankings) will more strongly relate to clinical outcomes than other methods because they would be more accurate and precise estimates of actual analgesia. In addition, dose intensity variables would have a wider distribution of scores than count variables, which would enhance statistical validity. To test this hypothesis, a series of hierarchical regression models were conducted with pain and analgesic drugs, coded in different ways, as predictors, with physical functioning as the dependent variable. Physical function was selected as the outcome because it is an important and commonly studied clinical outcome in community-based studies of elderly adults and because pain has an established relationship with functioning in this population [5,6]. Thus, analgesic medication can be expected to have an influence in these relationships. Thus, the purpose of this manuscript is to a) describe methods to quantify opioid and nonopioid analgesic drugs and b) to compare the utility of five different analgesic coding approaches in models examining the relationships between pain, analgesics, and physical functioning in community-dwelling older adults.

Methods

Design, Setting, and Participants

This study was comprised of participants from the Advanced Cognitive Training Interventions for Vital Elders (ACTIVE) study. ACTIVE was a National Institutes of Health (NIH)–funded multisite clinical trial designed to test the effects of cognitive interventions on cognitive and functional outcomes at seven time points from 1998 to 2010, with an additional follow-up data collection currently underway. Participants were enrolled at six sites in the United States [15]. The study was approved by the institutional review board at each site, and informed consent was obtained from all study participants. Secondary analyses of baseline (1998) cross-sectional data are presented here.

The targeted population was community-residing adults aged 65 years and older. Participants were included in the study if they did not have substantial cognitive decline (Mini Mental State Exam ≤ 23 or self-reported Alzheimer’s disease), substantial functional disability (e.g., needed assistance with activities of daily living of dressing, personal hygiene, or bathing), history of cerebral vascular accident, or severe sensory losses that impaired communication.

Participants were not selected based on pain or medications; these variables were collected as covariates in the trial. A total of 2,802 participants were randomized to one of three cognitive training interventions or the control group. The interventions were not targeted at pain, pain treatment, or medications.

Measures

Medication Assessment

As part of a comprehensive test battery, a brown bag survey of participants’ medications was conducted. All medications used by participants, both prescription and nonprescription, were recorded on a detailed medication administration form. Drug name, dose, frequency, route of administration, and routine or as-needed use were recorded. In addition, participants’ comments about the purpose of the drug or specific administration instructions were recorded in a text field. Drug names were standardized and converted to generic drug names. Medications were categorized using the American Hospital Formulary System Drug Information 2015 [16], which classifies central nervous system agents potentially prescribed for pain control (code 2808). Opioid medications are subclassified as opioid agonists, opioid partial agonists, and opioid antagonists. No participants reported using opioid antagonists; thus this drug category was excluded from this study. Nonopioid analgesic medications included NSAIDS (COX-2 inhibitors, salicylates, and other nonsteroidal anti-inflammatory agents) and miscellaneous analgesics and antipyretics (which included acetaminophen). For combination drugs, subcomponents were identified and classified according to their primary action (e.g., acetaminophen with codeine was deconstructed into acetaminophen and codeine) and coded accordingly when calculating equianalgesic doses. When creating summed variables, combination medications were treated as a single drug and counted once. Adjuvant medications used to treat pain (e.g., tricyclic antidepressants and anticonvulsants) and low-dose (81 mg) aspirin, typically used for cardiovascular conditions, were excluded from these analyses. All other analgesic medications, including routine and as-needed drugs, were included in the analysis. Analgesic medications were coded to create the following variables:

1. Any analgesic. This dichotomous variable indicated whether participants reported any analgesic medication.

2. Number of analgesics. This variable reflects the total number of analgesic medications reported.

3. Number of opioids and number of nonopioid analgesics. Based on the American Hospital Formulary Service drug classifications, medications were coded as either opioid or nonopioid analgesics. Within each category, the total number of medications reported was summed.

4. Opioid (OME) and nonopioid (OAE) equianalgesic doses. Due to differences in the mechanisms of action and relative potencies of nonopioids and opioids, two separate equianalgesic variables were created and summed: oral morphine equivalents and acetaminophen equivalents, respectively. These calculations used standardized conversion procedures that considered the type, dosage, and frequency of each medication.

5. Number of low-, medium-, and high-dose nonopioid analgesics. Due to the lack of a definitive and comprehensive equianalgesic table for nonopioid medications, we created a coding approach that classified the average daily doses as low, medium, or high. This approach provides dose comparative information, is comprehensive, and is more easily replicable than equianalgesic conversions [17].

Pain

The Bodily Pain Scale of the Medical Outcomes Study Short-Form 36 (SF-36) was used to assess pain intensity and interference. Participants were asked to rate how much bodily pain they had during the past four weeks. Responses were scored on a 1–6 scale, where 1 = no pain and 6 = very severe pain. Pain interference with normal work (housework or outside the home) was also assessed on a five-point scale (1 = not at all and 5 = extremely). If participants reported pain, they were asked to indicate the locations; the total number of reported pain locations was summed. A composite pain index was created from the pain intensity, interference, and number of painful body site variables using factor analysis and used as the pain indicator in these analyses.

Covariates

Age was included as a continuous variable and sex as a categorical variable. Race was analyzed as self-reported black or white. Educational level was assessed as years of education completed.

Outcome Variable

The main study outcome was physical functioning. This variable was measured with the SF36 Physical Functioning Subscale, a 10-item scale that asks participants to rate the extent to which health limits participation in activities. The activities assessed include vigorous or moderate activities, lifting, climbing stairs, walking, bending/kneeling, and bathing/dressing. Items are rated on a three-point scale, 1 = limited a lot, 2 = limited a little, and 3 = not limited at all. Using published scoring algorithms, scores are summed to create a total subscale score and transformed into a 0–100 scale. A score of zero indicates maximum disability, and 100 indicates no disability. This scale is widely used and has established reliability (alpha = 0.93) [18].

Procedures

Calculating Opioid Equianalgesic Dosages

Opioid dosages were converted to oral morphine equivalents (30 mg by mouth as the referent) [8]. Table 1 provides a list of commonly used opioid medications in older adults and the OME.

Table 1.

Equianalgesic conversions: Approximate equivalent dose of opioid and nonopioid drugs for mild to moderate pain

Opioids	Oral Morphine Equivalents (Equivalent to 30 mg PO MSO4) [8,19–21]
Codeine sulfate	200
Hydrocodone bitartrate*	30
Hydromorphone hydrochloride	7.5
Oxycodone hydrochloride*	20
Propoxyphene hydrochloride*,†	390
Propoxyphene napsylate*,†	600
Tramadol hydrochloride*	300
Nonopioids	Oral Acetaminophen Equivalents (Equivalent to 325 mg PO Acetaminophen) [8,18]
Aspirin	325
Celecoxib	1,300
Etodolac	50
Ibuprofen	200
Naproxen	125
Naproxen sodium	137.5
Oxaprozin	2,600
Piroxicam	325
Rofecoxib†	650
Sodium salicylate	500
Sulindac	75

OME = oral morphine equivalent; PO = by mouth.

^*OME dose is based only on the opioid component of these combination drugs.

^†Drugs are no longer approved in the United States, but were widely used at the time of data collection.

Calculating Nonopioid Equianalgesic Doses

Nonopioid analgesic doses were converted to oral acetaminophen equivalents (325 mg by mouth as the referent). This was challenging due to the lack of a comprehensive reference that included all of the nonopioid analgesics and inconsistencies in dose conversion ratios in different published sources. When this occurred, we searched the primary literature, consulted with a pharmacist, and used conversion ratios that were supported by at least two references. Table 1 provides a list of commonly used nonopioid analgesic medications in community-dwelling older adults and the OAE. Note that this table includes some medications that have been discontinued (e.g., rofecoxib). These discontinued drugs represent a small proportion (7%) of the drugs reported in this sample. Therefore, we included them in this analysis because they were commonly used by older adults at the time the baseline data were collected for the ACTIVE trial. While no longer commercially available in the United States, this information may be relevant for other researchers who are analyzing drug data from longitudinal studies of older adults.

To compute total equianalgesic doses, average daily dose (ADD) was computed for all analgesic medications, including analgesic subcomponents of combination drugs, by multiplying the number of doses taken (frequency) by the drug strength (dose) [10]. The ADD was converted to OME and OAE and summed to create a total OAE or OME dose. It is important to emphasize that this equianalgesic table was constructed for research purposes. Because of the equivalence conversions, some of the computed drug doses far exceed normal or safe values and should not be considered appropriate for clinical purposes.

Calculating the Number of Low-, Medium-, and High-Dose Nonopioid Analgesics

To create these variables, average daily doses for each nonopioid analgesic medication were calculated. These doses were compared with Food and Drug Administration–approved dosing ranges and coded as low-dose, medium-dose, or high-dose using available pharmacy resources and consultation with a pharmacist [22]. The total number of nonopioid analgesic medications in each dose level was summed (Table 2).

Table 2.

Nonopioid drug dose classifications, expressed in average daily doses

Generic Drug	Low Dose, mg	Medium Dose, mg	High Dose, mg
Acetaminophen	1,950	2,600	3,250
Aspirin	1,950	2,600	3,250
Celecoxib	200	400	800
Diclofenac sodium	100	150	200
Etodolac	600	800	1,200
Ibuprofen	1,200	1,800–2,400	3,200
Ketoprofen	75–150	225	300
Naproxen	750	1,000	1,250
Naproxen sodium	825	1,100	1,375
Oxaprozin	600	1,200	1,800
Piroxicam	10	20	40
Rofecoxib*	12.5	25	50
Sulindac	300	N/A	400

^*Rofecoxib is no longer approved for use in the United States.

Statistical Analyses

Baseline characteristics were analyzed using descriptive statistics. Frequencies and proportions or means and standard deviations are reported, as appropriate. Of the 2,802 participants, 54 cases had missing data on the physical functioning or pain variables. Thus, chi-square tests and t tests were used to investigate whether there were significant differences between those with and without complete data. The results revealed no significant differences in sex or race, but those with missing data were significantly older (mean age = 76 years) than those with complete data (mean age = 73.6 years, t = 2.57, df = 54.28, P = 0.01). Thus, age was included as a control variable in the subsequent analyses. Hierarchical regression analyses were conducted with physical functioning as the dependent variable. Demographic characteristics were entered in the first block. The pain indicator was entered in the second block. In the third block, analgesic medications were added. Five separate models were tested, with the different formulation of analgesics added in the last block. Data were analyzed using SPSS v.24.

Results

Sample Characteristics

The mean age of the sample was 73.6 years (range = 65–95 years). The sample was predominantly female (N = 2,126, 76%) and white (N = 2,038, 73%). Participants had a mean of 13.5 years of education (range = 4–20 years). The majority (N = 2,254, 81.9%) reported experiencing pain in the last four weeks, with most reporting mild (N = 1,370, 48.9%) or moderate (N = 667, 24%) pain. Almost half of the sample (N = 1,375, 49.9%) stated that pain interfered with daily activities, with most reporting mild interference (N = 698, 25%) or moderate interference (N = 445, 16%). Participants reported mild to moderate physical disability (mean = 68.8, SD = 24.1; 0= maximum disability, 100 = no disability).

Analgesic Medications

In this community-based sample, 26% (N = 733) of participants reported taking at least one analgesic drug. Of those who reported one or more analgesic drugs, 671 (91.5%) reported taking a nonopioid drug and 104 (14.2%) reported taking an opioid medication. NSAIDs were the most commonly reported category of analgesic drug, with aspirin as the most reported drug in that category. Propoxyphene napsylate was the most commonly reported opioid analgesic medication. When considering equianalgesic dosages, participants reported means of 4.5 mg OME (range = 0–600 mg) and 469.43 mg OAE (range = 0–16, 250 mg).

Bivariate Relationships Between Demographic Characteristics, Pain, Analgesic Medications, and Physical Function

Bivariate correlations were examined between physical functioning and demographic characteristics, pain, and analgesic medication variables. Age, race, and sex were all significantly correlated with physical functioning; participants who were older, black, or female had worse functioning (rs ranged from –0.08 for sex to –0.19 for age; all P = 0.000). More intense and interfering pain was significantly correlated with worse physical functioning (r = –0.50, P = 0.000). The relationship between pain and physical functioning was statistically significant for both males (r = –0.40, P = 0.000) and females (r = –0.52, P = 0.000), even though the sample was largely female. All of the analgesic medication variables, across the five coding strategies, were significantly correlated with physical functioning (Table 3). Thus, all demographic variables were included in the regression models. In addition, differences in pain and functioning between persons taking any analgesic and those taking no analgesics were examined. Participants who took any analgesic had significantly higher pain intensity (mean = 3.4 vs 2.5, t = –16.5, df = 2,749, P = 0.000), higher pain interference (mean = 2.3 vs 1.7, t = –16.0, df = 2,752, P = 0.000), and lower physical functioning (mean = 57.7 vs 72.8, t = 14.9, df = 2,758, P = 0.000) than those who took no analgesic medications.

Table 3.

Descriptive statistics and correlations with physical functioning for analgesic drugs across 5 different coding approaches (N = 2,802)

Code	Variable		Frequency, No. (%)	Mean (Median)	Range	r
1.	Any pain med	No	2,069 (73.8)			−0.28*
		Yes	733 (26.2)
	Any opioid	No	2,698 (96.3)			−0.17*
		Yes	104 (3.7)
	Any nonopioid	No	2,131 (76.1)			−0.24*
		Yes	671 (23.9)
2	Total number of analgesic medications	0	2,069 (73.8)	0.31 (0)		−0.28*
		1	623 (22.2)
		2	95 (3.4)
		3	14 (0.5)
		4	1 (0.1)
3.	Number of opioids	0	2,698 (96.3)	0.04 (0)		−0.17*
		1	100 (3.6)
		2	3 (0.1)
		3	1 (0.0)
	Number of nonopioids	0	2,131 (76.1)	0.27 (0)		−0.24*
		1	597 (21.3)
		2	70 (2.5)
		3	4 (0.1)
4.	Morphine equivalents (OME)			4.50 (0)	0–600	−0.10*
	Acetaminophen equivalents (OAE)			469.43 (0)	0–16, 250	−0.23*
5.	Number of low-dose nonopioids	0	2,309 (82.4)	0.21 (0)		−0.19*
		1	457 (16.3)
		2	33 (1.2)
		3	3 (0.1)
	Number of medium-dose nonopioids	0	2,619 (93.6)	0.12 (0)		−0.15*
		1	2 (0.1)
		2	181 (6.5)
		3	0
	Number of high-dose Nonopioids	0	2,775 (99.0)	0.03 (0)		−0.06*
		1	2 (0.1)
		2	0
		3	25 (0.9)

OAE = oral acetaminophen equivalent; OME = oral morphine equivalent.

^* P < 0.001.

Pain and Analgesic Drugs as Predictors of Physical Function, Controlling for Age, Race, and Sex

The results of hierarchical regression models are shown in Table 4. Pain was a significant predictor of worse physical function (β = –0.45, P = 0.001) in all models, after controlling for demographic variables. Models 1–5 added different configurations of the analgesic drug variables to determine whether analgesics added to the prediction of physical functioning. As seen in Table 4, the dichotomous variable indicating any pain medication reported was equivalent to the number of pain medications reported in terms of their explanatory power (β for both models = –0.12, P = 0.001). Taking any pain medications and taking more pain medications were associated with lower physical functioning. When considering types of analgesics, both the number of nonopioids (β = –0.09, P = 0.000) and the number of opioids (β = –0.098, P = 0.000) were significantly associated with lower physical function. Model 4 included the number of opioids and nonopioids (subcategorized into low, medium, or high dose). In this model, the number of opioids was a significant predictor of physical functioning (β = –0.06, P = 0.001) in addition to the number of low-dose (β = –0.07, P = 0.001) and medium-dose (β = –0.06, P = 0.001) nonopioid analgesic medications. Finally, Model 5 included morphine and acetaminophen equivalents. This model revealed that the amount of OAE was a statistically significant predictor of lower physical functioning (β = –0.09, P = 0.000), but the amount of OME was not significantly significant.

Table 4.

Hierarchical regression models predicting physical functioning from demographic characteristics and pain, controlling for analgesic drugs (N = 2,802)

Step		Model 1: Any Pain Drug		Model 2: Total No. of Pain Drugs		Model 3: No. of Opioids and Nonopioids		Model 4: No. of Low-, Medium-, and High-Dose Nonopioids		Model 5: OME and OAE
	Variable	B (SE)	β	B (SE)	β	B (SE)	β	B (SE)	β	B (SE)	β
1	Sex, % female	−4.38 (0.90)	−0.08*	−4.38 (0.90)	−0.08*	−4.35 (0.90)	−0.08*	−4.25 (0.90)	−0.08*	−4.23 (0.90)	−0.08*
	Age, y	−0.76 (0.07)	−0.19*	−0.76 (0.07)	−0.19*	−0.76 (0.07)	−0.19*	−0.76 (0.07)	−0.19*	−0.79 (0.07)	−0.19*
	Race, % white	5.77 (0.87)	0.11*	5.81 (0.87)	0.11*	5.74 (0.87)	0.11*	5.70 (0.87)	0.11*	5.70 (0.87)	−0.11*
2	Pain	−10.85 (0.40)	−0.45*	−10.82 (0.41)	−0.45*	−10.79 (0.40)	−0.45*	−10.80 (0.41)	−0.45*	−11, 11 (0.40)	−0.46*
3a	Any pain drug	−6.52 (0.91)	−0.12*
3b	No. of pain drugs			−5.14 (0.72)	−0.12*
3c	No. of opioids					−9.28 (1.9)	−0.08*
	No. of nonopioids					−4.47 (0.79)	−0.09*
3d	No. of opioids							−7.48 (2.0)	−0.06*
	No. of low-dose nonopioids							−3.68 (0.89)	−0.07*
	No. of medium-dose nonopioids							−3.14 (0.86)	−0.06*
	No. of high-dose nonopioids							−2.36 (1.47)	−0.03
3e	No. of OMEs									−0.01 (0.01)	−0.02
	No. of OAEs									−0.002 (0.00)	−0.09*
	No. of observations	2,736		2,736		2,736		2,736		2,736
	R2	0.31		0.31		0.31		0.31		0.31
	F for change in R2	50.86*		59.71*		28.37*		13.52*		16.88*

OAE = oral acetaminophen equivalent; OME = oral morphine equivalent.

^*P < 0.001.

Discussion

Pain and analgesic medications are part of everyday life for many older adults. This study focused on these concepts in a community-based sample of older adults. Not surprisingly, we found that 81.9% of the sample reported experiencing pain in the last four weeks and 49.9% reported that pain interfered with activities. As expected, pain was a significant predictor of worse physical functioning among older adults. These findings are congruent with the extant literature on the prevalence and consequences of pain in this population [2].

Only 26% of the older adults in this study reported taking any analgesic medication on a scheduled or as-needed basis. These findings are somewhat surprising given that more than two-thirds of the participants reported pain during the last four weeks. This finding lends further evidence to the prevalence of pain in older adults and the fact that pain may be undertreated in this population [6,23]. Pain remains a common problem among older adults and continues to have an impact on their everyday functioning.

In this study, we compared different approaches of quantifying analgesic medications. We compared simple approaches (number of analgesics, number of opioid and nonopioid analgesics) with more complex approaches that considered dosage and frequency information (equianalgesic doses and number of low-, medium-, and high-dose nonopioids). We hypothesized that morphine and acetaminophen equivalents would be better predictors of physical functioning than simpler approaches (e.g., any analgesic [yes/no] or number of medications) because equianalgesic doses are a more precise estimate of the amount of analgesia “on board.” This hypothesis was not supported. Equianalgesic calculations standardize analgesics to a common metric (morphine or acetaminophen) and consider the relative dose and frequency of analgesics in the calculation as opposed to a simple pill count. In addition, equianalgesic variables have more variability than count variables, which are prone to limited or skewed distributions. The fact that the complex approach was not superior to the simpler approaches is somewhat surprising, but there are several possible explanations. First, these findings may be specific to the model that we tested (e.g., pain and pain treatment as predictors) and may have functioned differently when treated as a covariate or as an outcome variable. Second, the use of physical functioning as the outcome variable and how this variable was measured may have influenced the relationships. Physical functioning is a relatively global concept and was measured with a tool that assessed the extent to which health limits activities. This concept, as measured, may not have been sensitive to differences in how pain treatment was classified. In this context, global assessments of analgesic medications (any analgesic drug or number of drugs) may have been sufficient indicators. In other clinical studies with nursing home residents and postoperative patients, equianalgesic doses were important covariates [13,14]. These studies, however, focused on specific pain outcomes, asked different research questions, and measured pain and pain medications simultaneously. Thus, different research questions, dependent variables, and measurements may yield different results. Third, there was a relatively low rate of analgesic medication use in this sample in general (only 25% participants reported using any pain drug) and a low reported use of opioid medications (14%). These findings are consistent with the reported opioid use in the last decade [7], but may have contributed to the relatively low explanatory power of these variables.

The result that all five analgesic coding methods yielded similar results may be a positive finding for researchers. The procedure for calculating equianalgesic variables is time-consuming and challenging. There are many reference sources for equianalgesic conversions for opioids, but there is a lack of consistency across references [17]. For nonopioid analgesic medications, there is a dearth of equiananalgesic conversion data and inconsistencies in the information that are available. Given that the vast majority of the analgesics reported in this sample were nonopioids, this is an important limitation. In an effort to develop a coding method for nonopioids that was more robust than a simple count and more reliable than computing equianalgesic doses, we developed an approach that categorized the doses of nonopioid drugs into low, medium, and high and counted the number of doses reported at each dose range. This approach, while novel, has not been empirically validated.

There are strengths and limitations to this study. The strengths are a large sample size of seniors drawn from different regions of the country. Using a community-based approach, we were able to gain insight into the analgesic use of relatively healthy older adults. Because we had a complete profile of medications used, we were able to consider drug category, dose, frequency, route, scheduled or as-needed use, and qualitative data about drug administration or instructions. This enabled us to analyze the analgesic drugs and dosages with some confidence. Despite the richness of these data, they are self-reported and we do not have drug administration data. That is, medications were assessed during a study session, but data are not available on whether or how often the drugs were actually taken. Another limitation, as previously mentioned, is the computation and interpretation of the acetaminophen equivalents. We were unable to find a single comprehensive source of nonopioid analgesic drug conversions to acetaminophen equivalents. In contrast to the OME, the acetaminophen conversion approach has not been widely validated. Thus, the equianalgesic values presented are conservative estimates. When conflicting conversion ratios were found in the literature, we opted to use the lowest values so as to underestimate rather than overestimate analgesic effects. The true amount of analgesia experienced by patients would vary based on biological and psychological factors that influence the perception of pain and pain relief [19] and individual differences in pharmacokinetics and pharmacodynamics.

In this study, we analyzed medication data from the baseline session of a longitudinal study of community-based older adults. The data set includes drugs that have since been discontinued (e.g., Celebrex). We opted to include these drugs in this analysis because they represent the true state of pharmacological pain management at the time the study was conducted. These drugs represented a relatively small proportion of reported analgesics in this study and are not likely to have significantly inflated the analgesic variables or altered the results. In future research, as we analyze the analgesic drugs in this sample longitudinally, these discontinued medications will be naturally dropped from analysis as drug formulations change over time.

In summary, the results of this study indicate that more complex drug coding methodologies were not superior to a simple dichotomous variable (any pain med: yes/no) or the sum of the number of analgesics when quantifying analgesic medications in community-based studies of older adults. We tested five approaches to quantifying analgesic drugs in statistical models predicting physical function and found them to be approximately equivalent in terms of their predictive ability. In the empirical literature on pain or aging, there are very few studies that describe the complex and time-consuming process of collecting and preparing drug data for analysis. Pain treatment, and the ability to quantify it in noncontrolled, community-based studies, is important in understanding analgesic medication practices and addressing the challenges of pain management in older adults. Future research should continue to investigate and explicate strategies for analyzing analgesic use in different samples, with different conceptual models and outcome variables, and addressing different research questions. Clarifying and standardizing approaches to quantifying analgesic drugs would advance our understanding of pharmacological pain treatment and its effects in older adults.

Funding sources: Drs Horgas and Marsiske were funded the ACTIVE trial described herein. The ACTIVE cognitive training trial Study was supported by grants from the National Institutes of Health (National Institute on Aging & National Institute of Nursing Research) to six field sites and the coordinating center, including Hebrew Senior-Life, Boston (NR04507), the Indiana University School of Medicine (NR04508), the Johns Hopkins University (AG14260), the New England Research Institutes (AG14282), the Pennsylvania State University (AG14263), the University of Alabama at Birmingham (AG14289), and the University of Florida (AG014276). Ms. Snigurska received financial support from the University of Florida, via the University Scholars Program and College of Nursing Scholars Program. The opinions expressed here are those of the authors and do not necessarily reflect those of the funding agencies, academic, research, or governmental institutions involved.