Sage Journals: Discover world-class research

Abstract

Background

Prescription opioids are essential in managing pain among adults with chronic pain conditions. However, persistent use over time can lead to negative health consequences. Identifying individuals with persistent use over time and their characteristics can inform clinical decision-making and aid in reducing the risk of abuse and overdose deaths.

Objective

This study aims to examine trajectories of prescription opioid use over time and factors associated with these trajectories among older cancer survivors with any non-cancer pain conditions (NCPC).

Methods

We conducted a retrospective cohort study design with longitudinal data of older (age at cancer diagnosis ≥67 years) cancer (incident breast, colorectal, and prostate cancers, or non-Hodgkin lymphoma) survivors with any NCPC. Data were derived from the 2007-2015 linked Surveillance, Epidemiology, and End Results (SEER)-Medicare dataset (N = 35,071). Group-Based Trajectory Modeling (GBTM) was used to identify homogeneous subgroups (distinct trajectories) of individuals based on every 90-day prescription opioid use during pre-cancer diagnosis (t₁-t₄), acute cancer treatment (t₅-t₈), and post-cancer treatment (t₉-t₁₂) periods. Biological factors, social determinants of health (SDoH), physical and mental health, medication use, health care use, and external factors associated with a trajectory membership were analyzed with multivariable multinomial logistic regressions.

Results

Four distinct trajectories of opioid use were identified: (1) increase-decrease use (6.1%); (2) short-term use after cancer diagnosis (40.6%); (3) low-use (41.0%); and (4) persistent use (12.3%). In the fully-adjusted multinomial logistic regression, the SDoH such as Non-Hispanic Black [adjusted odds ratios (AOR) = 1.69; 95%CI = 1.48, 1.93)] and rural residence (AOR = 1.49; 95%CI = 1.15, 1.94)], comorbid anxiety (AOR = 1.33; 95%CI = 1.18, 1.51), and medication use (NSAIDs - AOR = 1.20; 95%CI = 1.10, 1.30) were associated with membership in the persistent use group. Persistent use was less likely among those with higher fragmented care index (AOR = 0.95, 95%CI = 0.93, 0.97) and those living in counties with higher Medicare advantage penetration (AOR = 0.96; 95%CI = 0.95, 0.97).

Conclusions

One in eight older adults had persistent opioid use over time. The profile characteristics of this group were different from the other trajectory groups. Policies and programs to reduce chronic opioid use need to consider the intra- and inter-individual variability to reduce opioid-related morbidity and mortality.

Plain Language Summary

This study looked at how older adults with cancer and long-term non-cancer pain used prescription opioids before and after their cancer diagnosis. The goal was to understand the patterns of opioid use over time and identify factors linked to long-term use. We analyzed data from over 35,000 cancer survivors with chronic pain conditions, focusing on their opioid prescriptions for the year before and two years after their cancer diagnosis. We identified four main patterns (or “trajectories”) of opioid use: (1) Increase-decrease use (6%): Use increased but later decreased; (2) Short-term use after cancer diagnosis (41%): Opioid use was brief and followed the cancer diagnosis; (3) Low-use (41%): Very little opioid use; (4) Persistent use (12%): Consistent use over time. Certain groups, including Black individuals, rural residents, and those with anxiety, were more likely to have persistent opioid use. People with more fragmented healthcare or living in areas with more Medicare Advantage coverage were less likely to stay on opioids long-term. The study highlights that 1 in 8 older cancer survivors continued using opioids for an extended period. Tailored programs are needed to address individual differences in order to reduce the risks of opioid misuse and overdose.

Keywords

opioids persistent use trajectory group-based trajectory modeling chronic pain cancer

Introduction

Among cancer survivors, prescription opioids continue to remain as a primary therapy to treat pain associated with cancer growth, surgical, and other cancer treatments.^1,2 Furthermore, a subpopulation that may have high rates of prescription opioid use is individuals with non-cancer pain conditions (NCPC) for their pain management. Nearly one in two adults (44%) with NCPCs filled at least one opioid prescription in a year,³ and 14% reported long-term use.⁴ Moreover, older adults are more prone to use prescription opioids.^5-7 Ramachandran and colleagues found a significant increase in new prescriptions of opioids annually from 2013 to 2015 (from 6.6% to 10.3%) among older Medicare beneficiaries.⁸

During the cancer treatment period, older individuals with both NCPCs and cancer may require higher doses of opioids for a longer term, even in the absence of disease progression due to opioid-induced hyperalgesia and the effects of tolerance.^9,10 Clinical guidelines recommend that opioids should be prescribed for only short-term periods and should be tapered off to prevent the prescription opioids from triggering opioid use disorders.^11,12 However, research has shown that many patients persist in long-term prescription opioid use. For example, a study on older Medicare beneficiaries with NCPCs found that 25.0% of patients were being prescribed opioids at a consistently moderate dose over a period of six months,¹³ while among hospitalized opioid users (mean age = 59.2 years, SD = 14.4), 19.2% remained on the stable high dose for a year since the day of the first hospital opioid prescription.¹⁴

Opioid-trajectory-wise, only one study was found conducted on older adults with NCPCs. It identified four prescription opioid dose trajectories: gradual dose discontinuation (23.5%), gradual dose increase (30.3%), consistent low dose (24.3%), and consistent moderate dose (22.0%).¹³ Of these four patterns, older beneficiaries with a gradual dose increase had an increased risk of having opioid adverse effects compared to those with opioid dose discontinuation.¹³

To date, no investigations have systematically characterized the distinct trajectories of prescription opioid use over time before and after cancer diagnosis or the factors associated with the trajectories among older adults with any NCPC. Therefore, this study aims to identify the opioid use trajectories and to profile the characteristics associated with these trajectories before and after cancer diagnosis utilizing the population-based, Surveillance, Epidemiology, and End Results (SEER)-Medicare claims linked database in the US. This study can inform clinical decision-making to reduce the potential for abuse and the risk of opioid-related overdose deaths.

Methods

Study Design

We employed a retrospective cohort study design with longitudinal data. The cohort consisted of older adults (age >67 years at the time of cancer diagnosis) with any NCPC and an incident primary cancer (breast, prostate, colorectal cancer, or Non-Hodgkin Lymphoma (NHL)) between 2009 and 2013. The NCPCs consisted of headaches, nervous system disorders, systemic lupus erythematosus, osteoarthritis, rheumatoid arthritis, gout, spondylosis, fractures, and other musculoskeletal and connective tissue disorders. These NCPCs were carefully selected based on their prevalence among the elderly population, as reflected in the dataset utilized in this study.

We selected age 67 years or older at cancer diagnosis to allow for Medicare enrollment 2-year before cancer diagnosis. We observed the selected individuals for 48 months consisting of a 24-month pre-cancer diagnosis (12 months of NCPC identification +12 months of opioid use and other independent variable data collection) and a 24-month post-cancer diagnosis period (24 months of opioid use follow-up).

West Virginia University’s Institutional Review Board (IRB) approved this study's ethics with Acknowledgement of Not Human Subjects Research (NHSR) - protocol # 2109427002 - as no individually identifiable private information was involved.

Data Sources

This study utilized the linked SEER-Medicare database. This linked database merged large population-based data sources that provide patient- (SEER-registry and Medicare claims), zip code- (census track), and county-level (Area Health Resource File) data. The population-based SEER cancer registry contains demographic and clinical information about the cancer diagnosis, including type, subtype, stage, tumor size, lymph node involvement, tumor receptors and other prognostic features, and cause of death.¹⁵ Medicare claims provide data on the date of diagnosis, presence of other health conditions, services and treatments for cancer and non-cancer conditions, health care expenditures, and beneficiary enrollment information.¹⁵ We obtained county-level information on population characteristics, economics, health care professions, health professions training, health facilities, hospital utilization and expenditures, and the environment from the area health resource files (AHRF).¹⁶ The zip code-level neighborhood profiles of economic and demographic data were derived from the census tract files.

Identification of NCPCs

We identified the older individuals (age at cancer diagnosis ≥67 years) with any NCPC before the cancer diagnosis date using one inpatient or two outpatient visits for NCPC. NCPC conditions were identified based on the AHRQ Clinical Classifications Software (CCS) codes mapped to ICD-9 and ICD-10.^17,18 To preserve temporality, opioid use was identified after NCPC diagnosis. Thus, the NCPC identification period was between 24 and 12 months before cancer diagnosis or 12 months before opioid use.

Identification of Older Adults with Cancer

We included the older adults diagnosed with the following incident and primary cancers: breast, colorectal cancer, prostate, or NHL between 2009 and 2013. These cancers were identified from the SEER-cancer registry.

Inclusion and Exclusion Criteria

We required the study cohort to be continuously enrolled in Fee-for-Service Parts A and B for the entire observation period (48 months), continuously enrolled in Part D (Medicare Prescription Drug Program) for 36 months (12 months before cancer and 24 months after cancer diagnosis). We excluded individuals who were diagnosed with cancer at autopsy. The final study cohort consisted of 35,071 older adults with any NCPC and primary incident breast cancer (42.0%), prostate (32.8%), colorectal (17.0%) cancers, or NHL (8.2%).

Measures

Dependent Variable: Opioid Use Every 90 days (yes/no)

We measured any prescription opioid use every 90 days as a dichotomous variable (yes/no) during the 36 months (12 months before cancer diagnosis and 24 months after cancer diagnosis). The opioids were identified using the generic names (GNN)¹⁹ from the Medicare Part D prescription events file. The opioids included are codeine, dihydrocodeine, hydrocodone, oxycodone, hydromorphone, oxymorphone, morphine, tramadol, buprenorphine, pentazocine, fentanyl, sufentanil, methadone, and meperidine.

Independent Variables

All independent variables were measured 12 months before cancer diagnosis. The selection of independent variables was guided by the conceptual framework adapted from the modified Determinants of Health model by Park et al.²⁰ This model suggests that health is multifactorial, and that the health of individuals may be considered to be the result of interactions of many variables. Using this model, we classified the independent variables into the following domains: biological, social determinants of health (SDoH), physical health, mental health, health care use, medication use, and external factors.

The biological factors included age at cancer diagnosis and sex (male or female). SDoH factors encompassed zip code-level percentages of residents living below poverty and those with a college education, patient-level race and ethnicity (non-Hispanic white, non-Hispanic black, Latino/Hispanic, or others), Medicare and Medicaid dual eligibility, marital status, rural-urban commuting area (metro, urban, or rural), and geographic region (Northeast, South, North-Central, or West). Physical health factors included cancer types and stages, physical comorbidities (≥2) (thyroid, asthma, coronary artery disease, arrhythmia, diabetes, congestive heart failure, chronic kidney disease, chronic obstructive pulmonary disease, dementia, hepatitis, hyperlipidemia, HIV, hypertension, osteoporosis, stroke). Mental health was measured by diagnosed depression, anxiety, and substance abuse. Medication use consisted of polypharmacy (≥6 drugs, excluding opioids) and the use of NSAIDs, steroids, antidepressants, and benzodiazepines. Health care use comprised visits to pain specialists, anesthesiologists, primary care physicians (PCP), and fragmentation care index (FCI). The FCI was computed based on a modified version of a validated continuity of care index,²¹ based on the number of health care visits, providers seen, and the proportion of visits with each provider. The ranges of FCI were from 0 (all visits with the same provider) to 10 (each visit with a different provider), and higher values suggest fragmented care or care discontinuation. For ease of interpretation, we multiplied the FCI by 10. External factors included diagnosis year and market characteristics, specifically the county-level percentages of Medicare Advantage penetration. The Medicare Advantage penetration variable was determined using the county data from SEER-Medicare linked to an external source of AHRF data.

Statistical Analyses

Identification of Opioid Use Trajectories

GBTM groups individuals into subgroups that show statistically similar patterns over time for the primary variable under consideration (in our case, opioid use) based on the data.²² It is an alternative approach to multi-level modeling used to study the intra- and inter-individual variabilities.²³ GBTM is a statistical method to identify individuals’ subgroups based on time functions as determined by maximum likelihood estimation. The GBTM assigns group membership by ranking probabilities and assigning members with the highest probability to a specific group.

We followed the best practices of GBTM to identify opioid use trajectories.²² For example, we determined the necessary number of groups before choosing the optimal functional model for each group. We selected the optimal number of groups based on clinical meaningfulness, distinct patterns, and statistical significance. We explored different combinations of linear, quadratic, cubic, or quartic terms.

After identifying the opioid use trajectory patterns, we used multivariable multinomial logistic regressions (MLR) (covariate adjustment method) to analyze the unadjusted and adjusted associations of the study cohort’s characteristics with the trajectory groups. The utilization of the MLR covariate adjustment method was intended to reduce potential selection bias in this study. All data management and all analyses except GBTM were conducted with SAS 9.4 (SAS Institute Inc., Cary, NC). Stata 14 (Stata Corp LLC, College Station, TX) was used for GBTM.

The reporting of this study has followed relevant Equator guidelines,²⁴ and complied with The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.²⁵

Results

Study Cohort Characteristics

A majority of the study cohort was female (58.5%), Non-Hispanic White (75.9%), and living in metropolitan areas (84.2%). An overwhelming majority (79.3%) had pre-existing physical comorbidities (excluding NCPCs). Of all older adults, 9.1% had anxiety, and 29.3%, 46.3%, 23.7% were prescribed NSAIDs, steroids, and antidepressants, respectively. The fragmented care index averaged 0.73 with SD = 0.248. The mean Medicare Advantage Penetration rate was 24.99% with SD = 12.84%.

Opioid Use Trajectories

The final GBTM resulted in four distinct patterns of opioid use that were anchored to cancer diagnosis and treatment. For example, the use patterns followed three distinct time periods: pre-cancer diagnosis (t₁-t₄), acute cancer treatment (t₅-t₈), and post-acute cancer treatment (t₉-t₁₂). The trajectories were: (1) increase-decrease use (6.1%); (2) short-term use after cancer diagnosis (40.6%); 3); low-use (41.0%); and (4) persistent use (12.3%) (refer to Table 1). The selection of these groups was based on the following criteria: (1) significant P-values (<0.05) for the highest-order polynomial parameter; (2) at least 5% of the population assigned to each trajectory group; (3) no polynomial overfitting; and (4) clinically meaningful grouping.^22,26,27 The final trajectory model satisfied Nagin’s diagnostic criteria of the “average of posterior probabilities (groupAPP) > 0.7″ for all groups.^26,27

Table 1.

Statistics of Opioid Use Trajectory Groups by GBTM Among Older Medicare Beneficiaries With Any NCPC, 12 Months before and 24 Months after Incident Primary Breast, Colorectal, Prostate Cancers, or NHL Cancers, and Non-hodgkin Lymphoma (NHL) Linked Surveillance, Epidemiology, and End Results (SEER) and Medicare, 2007-2015.

Group	Parameter	Estimate	Standard Error	P-Values
1	Intercept	−4.36182	0.14503	<0.0001
	Linear	1.48009	0.04219	<0.0001
	Quadratic	−0.09873	0.00269	<0.0001
2	Intercept	−78.83808	2.02968	<0.0001
	Linear	30.66687	0.75859	<0.0001
	Quadratic	−2.98561	0.07055	<0.0001
3	Intercept	−2.14057	0.02536	<0.0001
	Linear	0.45497	0.00920	<0.0001
	Quadratic	−0.04428	0.00074	<0.0001
4	Intercept	1.46626	0.07452	<0.0001
	Linear	0.73675	0.02190	<0.0001
	Quadratic	−0.08054	0.00156	<0.0001

Group membership		% of Pop Assigned	Standard Error	P-Values	GroupAPP
1	Increase-decrease use	6.1	0.23991	<0.0001	0.821366
2	Short-term use after cancer diagnosis	40.6	0.33671	<0.0001	0.827613
3	Low use	41.0	0.36048	<0.0001	0.945373
4	Persistent use	12.3	0.20208	<0.0001	0.965117
Entropy = 0.806

Note: The final model was run with the all-quadratic polynomial combination.

Based on 35,071 adults with NCPCs 67 years or older, diagnosed with incident primary breast, prostate, colorectal cancers, and NHL between 2009 and 2015, continuously enrolled in Fee-for-Service Medicare Parts A and B for 24 months, and continuously enrolled in Medicare part D for 12 months before and 24 months after cancer diagnosis. Group differences by opioid use categories were tested with Chi-Square test.

Abbreviations: GBTM: Group-Based Trajectory Modeling, GroupAPP: average posterior probability for each group, OCC: odds of correct classification, NHL: Non-Hodgkin Lymphoma.

Opioid Use Trajectories

Figure 1 displays the trajectories of opioid use over time estimated from the GBTM. During the pre-cancer diagnosis period (t₁ through t₄), the following groups: (1) “increase-decrease use”; (2) “short-term use after cancer diagnosis”; and (3) “low use” had low probability of any opioid use. The opioid use pattern for the “increase-decrease use” group consisted of a gradually increasing probability of opioid use before cancer diagnosis (t₁-t₄), with the highest probability during the acute cancer treatment (t₅-t₈) and tapering off during the post-acute cancer treatment (t₉-t₁₂). The group with “short-term use after cancer diagnosis” did not have any opioid use either before a diagnosis of cancer (t₁-t₄) or during the post-acute cancer treatment phase (t₉-t₁₂) and had a high probability of opioid use during the acute cancer treatment phase (t₅-t₈). The “low-use” group had a consistently low probability of any opioid use before cancer diagnosis (t₁-t₄) and during the post-acute cancer treatment (t₉-t₁₂) period.

Figure 1.

Trajectories of any opioid use among older medicare beneficiaries with any NCPC, 12 months before and 24 months after incident primary breast, colorectal, prostate cancers, or non-hodgkin lymphoma (NHL) linked surveillance, epidemiology, and end results (SEER) and medicare, 2007-2015.

On the other hand, the “persistent use” group was more likely to use opioids before cancer diagnosis (t₁-t₄) and during the acute treatment phase (t₅-t₈). In this group, although the probability of opioid use declined, it never reached the lower probability of the other groups. For all groups, the probability of opioid use peaked immediately after cancer diagnosis (t₅).

Associations of Independent Variables with Opioid Use Trajectory Groups

Before computing the Odds Ratios, we explored the correlation between the independent variables and the opioid use trajectory groups descriptively listed in Table 2. Subsequently, the unadjusted odds ratios (UOR) and adjusted odds ratios (AOR) (with “low use” as the reference) from the multinomial logistic regressions on opioid use trajectory groups are summarized in Tables 3 and 4, respectively.

Table 2.

Description of Older Medicare Beneficiaries with any NCPC Diagnosed with Breast, Colorectal, Prostate Cancers, or NHL by Opioid Use Trajectory Groups Linked Surveillance, Epidemiology, and End Results (SEER) and Medicare, 2007-2015.

	Increase-Decrease Use	%	Short-Term Use After Cancer Diagnosis	%	Low Use	%	Persistent Use	%	Sig
All	2,150	6.1	14,244	40.6	14,371	41.0	4,306	12.3
Sex									<0.001
Female	1,373	6.7	8,138	39.6	8,256	40.2	2,760	13.4
Male	777	5.3	6,106	42.0	6,115	42.0	1,546	10.6
Race and ethnicity									<0.001
Non-hispanic white	1,604	6.0	10,905	41.0	10,975	41.3	3,121	11.7
Non-hispanic black	258	8.1	1,061	33.2	1,222	38.2	659	20.6
Latino/hispanic	198	7.3	965	35.6	1,191	43.9	356	13.1
Other races	80	3.9	1,067	51.5	803	38.8	120	5.8
Marital status									<0.001
Married	996	5.8	7,280	42.3	7,250	42.1	1,688	9.8
Wid/Sep/Div	756	6.7	4,262	37.9	4,453	39.6	1,775	15.8
Never married	208	7.1	1,171	39.8	1,147	39.0	417	14.2
Substance abuse									<0.001
Yes	151	8.6	407	23.2	679	38.7	519	29.6
No	1,999	6.0	13,837	41.5	13,692	41.1	3,787	11.4
Comorbidities (≥2)									<0.001
Yes	1,765	6.3	11,034	39.7	11,476	41.3	3,536	12.7
No	385	5.3	3,210	44.2	2,895	39.9	770	10.6
Depression									<0.001
Depression	328	8.2	1,101	27.6	1,665	41.7	902	22.6
No depression	1,822	5.9	13,143	42.3	12,706	40.9	3,404	11.0
Anxiety									<0.001
Yes	248	7.8	901	28.3	1,294	40.6	744	23.3
No	1,902	6.0	13,343	41.8	13,077	41.0	3,562	11.2
Prescription NSAIDs									<0.001
Yes	836	8.1	2,906	28.3	4,712	45.9	1,813	17.7
No	1,314	5.3	11,338	45.7	9,659	38.9	2,493	10.1
Steroids									<0.001
Yes	1,122	6.9	5,694	35.1	7,017	43.2	2,409	14.8
No	1,028	5.5	8,550	45.4	7,354	39.1	1,897	10.1
Antidepressants									<0.001
Yes	663	8.0	2,278	27.4	3,478	41.9	1,884	22.7
No	1,487	5.6	11,966	44.7	10,893	40.7	2,422	9.0
Benzodiazepine									<0.001
Yes	152	7.5	595	29.4	834	41.3	440	21.8
No	1,998	6.0	13,649	41.3	13,537	41.0	3,866	11.7
Polypharmacy (≥6)									<0.001
Yes	1,309	8.0	5,144	31.3	6,995	42.5	2,994	18.2
No	841	4.5	9,100	48.8	7,376	39.6	1,312	7.0
Cancer stage									<0.001
Early stage	1,577	5.6	11,519	41.2	11,432	40.9	3,425	12.3
Advanced stage	440	9.0	1,714	35.3	2,081	42.8	627	12.9
Type of cancer									<0.001
Breast	939	6.4	5,954	40.4	5,868	39.8	1,983	13.4
Prostate	567	4.9	4,914	42.7	4,827	42.0	1,195	10.4
Colorectal	422	7.1	2,342	39.3	2,427	40.7	773	13.0
Non-hodgkin lymphoma	222	7.7	1,034	35.8	1,249	43.2	355	12.3
Medicaid eligibility									<0.001
Yes	334	7.8	1,419	33.1	1,594	37.2	937	21.9
No	1,816	5.9	12,825	41.7	12,777	41.5	3,369	10.9
Pain specialist visit									<0.001
Pain specialist visit	516	6.6	2133	27.5	3,892	50.1	1,229	15.8
No pain specialist visit	1,634	6.0	12,111	44.4	10,479	38.4	3,077	11.3
Diagnosis year									0.0005
2009	367	5.9	2,625	41.9	2568	41.0	699	11.2
2010	456	6.9	2,696	40.9	2,693	40.9	744	11.3
2011	430	6.2	2,735	39.7	2,811	40.8	916	13.3
2012	418	5.8	2,892	40.0	2,979	41.2	935	12.9
2013	479	5.9	3,296	40.7	3,320	41.0	1,012	12.5
Metropolitan status									<0.001
Metropolitan	1,739	5.9	12,209	41.3	12,164	41.2	3,416	11.6
Non-metropolitan	363	7.5	1,788	37.1	1,929	40.0	743	15.4
Rural	47	6.7	238	34.1	267	38.3	145	20.8
Region									<0.001
Northeast	320	4.5	3,506	49.6	2,761	39.1	478	6.8
South	686	8.4	2,573	31.5	3,487	42.7	1,412	17.3
North central	253	5.9	1,882	43.8	1,673	39.0	486	11.3
West	891	5.7	6,283	40.4	6,450	41.5	1,930	12.4

	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Age at cancer diagnosis	75.61	6.35	76.37	6.44	75.77	6.19	75.35	6.15
Zip Code Level Median Income	39,699.50	23,344.20	43,326.60	25,804.21	42,274.19	25,129.04	37,781.15	22,307.44
Education (% less than HS education)	25.45	16.74	23.29	16.59	23.92	16.71	26.84	16.91
Poverty percentage	15.44	12.67	13.87	12.23	14.35	12.41	16.39	12.85
Fragmented care index (x 10)	7.30	2.56	7.24	2.38	7.38	2.52	7.23	2.63
% medicare advantage penetration	24.41	12.82	24.83	12.84	25.22	12.82	25.06	12.93

Note: Based on 35,071 adults with NCPCs 67 years or older, diagnosed with incident primary breast, prostate, colorectal cancers, and NHL between 2009 and 2015, continuously enrolled in fee-for-service medicare parts a and B for 24-month, and continuously enrolled in medicare part D for 12 months before and 24 months after cancer diagnosis. Group differences by opioid use categories were tested with chi-square test.

Abbreviations: NHL: Non-hodgkin lymphoma, Wid/Sep/Div: Widow/Separate/Divorce, HS: High School, SD: Standard deviation.

Table 3.

Unadjusted Odds Ratios (AOR) and 95% Confidence Intervals (CI) from Multivariable Logistic Regressions on Opioid Trajectory Groups with “Low Use” as Reference Group Among Older Medicare Beneficiaries with any NCPC and Breast, Colorectal, Prostate, or NHL Linked Surveillance, Epidemiology, and End Results (SEER) and Medicare, 2007-2015.

	Increase-Decrease Use			Short-term Use after Cancer Diagnosis			Persistent Use
	UOR	95% CI	Sig	UOR	95% CI	Sig	UOR	95% CI	Sig
Sex
Female (ref)
Male	0.75	[0.68, 0.83]	<0.0001	1.00	[0.95, 1.05]	0.9814	0.75	[0.70, 0.81]	<0.0001
Age
67-74 (ref)
>=75	0.96	[0.88, 1.05]	0.3864	1.16	[1.10, 1.21]	<0.0001	0.91	[0.85, 0.97]	0.0050
Race and ethnicity
Non-hispanic white (ref)
Non-hispanic black	1.48	[1.28, 1.71]	<0.0001	0.88	[0.80, 0.96]	0.0039	1.90	[1.70, 2.11]	<0.0001
Latino/hispanic	1.08	[0.91, 1.28]	0.3727	0.81	[0.74, 0.89]	<0.0001	1.06	[0.93, 1.21]	0.3464
Other races	0.68	[0.53, 0.87]	0.0025	1.33	[1.21, 1.47]	<0.0001	0.51	[0.41, 0.63]	<0.0001
Cancer type
Breast	0.87	[0.74, 1.02]	0.0904	1.26	[1.15, 1.38]	<0.0001	1.16	[1.02, 1.33]	0.0265
Colorectal	0.63	[0.53, 0.75]	<0.0001	1.25	[1.13, 1.37]	<0.0001	0.85	[0.74, 0.98]	0.0240
Prostate	0.95	[0.80, 1.14]	0.6122	1.20	[1.08, 1.33]	0.0007	1.10	[0.95, 1.28]	0.1983
NHL (ref)
Metropolitan status
Metro (ref)
Nonmetro	1.31	[1.16, 1.49]	<0.0001	0.93	[0.87, 1.00]	0.0616	1.35	[1.22, 1.48]	<0.0001
Rural	1.29	[0.94, 1.78]	0.1167	0.89	[0.74, 1.07]	0.2236	1.90	[1.54, 2.35]	<0.0001
Marital status
Married (ref)
Wid/sep/div	1.25	[1.13, 1.39]	<0.0001	0.96	[0.91, 1.02]	0.1591	1.74	[1.61, 1.88]	<0.0001
Never married	1.32	[1.12, 1.56]	0.0012	1.03	[0.94, 1.13]	0.4851	1.58	[1.39, 1.80]	<0.0001
Region
South	1.75	[1.51, 2.02]	<0.0001	0.58	[0.54, 0.62]	<0.0001	2.34	[2.08, 2.63]	<0.0001
North central	1.35	[1.12, 1.61]	0.0012	0.88	[0.81, 0.96]	0.0029	1.65	[1.43, 1.90]	<0.0001
West	1.23	[1.06, 1.41]	0.0049	0.76	[0.71, 0.81]	<0.0001	1.72	[1.54, 1.93]	<0.0001
Northeast (ref)
Medicaid eligibility
Yes	1.21	[1.06, 1.37]	0.0037	0.77	[0.71, 0.83]	<0.0001	2.00	[1.83, 2.19]	<0.0001
No (ref)
Pain specialist visit
Yes	0.85	[0.76, 0.94]	0.0027	0.47	[0.45, 0.50]	<0.0001	1.08	[0.99, 1.16]	<0.0001
No (ref)
Substance abuse
Yes	1.52	[1.26, 1.84]	<0.0001	0.59	[0.52, 0.67]	<0.0001	2.76	[2.44, 3.12]	<0.0001
No (ref)
Depression
Yes	1.35	[1.18, 1.54]	<0.0001	0.62	[0.57, 0.67]	<0.0001	2.02	[1.84, 2.21]	<0.0001
No (ref)
Anxiety
Yes	1.30	[1.12, 1.51]	0.0005	0.68	[0.62, 0.74]	<0.0001	2.09	[1.89, 2.31]	<0.0001
No (ref)
Comorbidities (>=2)
Yes	1.15	[1.02, 1.30]	0.0224	0.87	[0.82, 0.92]	<0.0001	1.16	[1.06, 1.27]	0.0014
No (ref)
Polypharmacy
Yes	1.67	[1.52, 1.83]	<0.0001	0.60	[0.57, 0.63]	<0.0001	2.45	[2.27, 2.64]	<0.0001
No (ref)
Prescription NSAIDs
Yes	1.30	[1.18, 1.43]	<0.0001	0.52	[0.49, 0.55]	<0.0001	1.49	[1.38, 1.60]	<0.0001
No (ref)
Steroids
Yes	1.14	[1.04, 1.26]	0.0049	0.70	[0.67, 0.73]	<0.0001	1.33	[1.24, 1.43]	<0.0001
No (ref)
Antidepressants
Yes	1.39	[1.25, 1.54]	<0.0001	0.59	[0.56, 0.63]	<0.0001	2.48	[2.30, 2.66]	<0.0001
No (ref)
Benzodiazepine
Yes	1.21	[1.01, 1.46]	0.0405	0.72	[0.64, 0.80]	<0.0001	1.86	[1.64, 2.11]	<0.0001
No (ref)
Fragmented care index (x 10)	0.99	[0.97, 1.01]	0.1501	0.98	[0.97, 0.99]	<0.0001	0.98	[0.96, 0.99]	0.0016

Note: Based on 35,071 adults with NCPCs (67 years at cancer diagnosis) or older, diagnosed with incident primary breast, prostate, colorectal cancers, and NHL between 2009 and 2015, continuously enrolled in Fee-for-Service Medicare Parts A and B for 24-month, and continuously enrolled in Medicare part D for 12 months before and 24 months after cancer diagnosis.

Abbreviations: NHL: Non-Hodgkin Lymphoma, Wid/Sep/Div: Widow/Separate/Divorce.

Table 4.

Fully-Adjusted Odds Ratios (AOR) and 95% Confidence Intervals (CI) from Multivariable Logistic Regressions on Opioid Trajectory Groups with “Low Use” as Reference Group among Older Medicare Beneficiaries with any NCPC and Breast, Colorectal, Prostate, and NHL Linked Surveillance, Epidemiology, and End Results (SEER) and Medicare, 2007-2015.

	Increase-Decrease Use			Short-Term Use after Cancer Diagnosis			Persistent Use
	AOR	95% CI	Sig	AOR	95% CI	Sig	AOR	95% CI	Sig
Sex
Female (ref)
Male	0.75	[0.67, 0.85]	<0.0001	0.99	[0.93, 1.05]	0.6354	0.92	[0.84, 1.01]	0.0723
Race and ethnicity
Non-hispanic white (ref)
Non-hispanic black	1.24	[1.04, 1.49]	0.0190	0.91	[0.82, 1.02]	0.1080	1.69	[1.48, 1.93]	<0.0001
Latino/hispanic	1.07	[0.87, 1.30]	0.5351	0.85	[0.76, 0.96]	0.0059	0.86	[0.73, 1.01]	0.0654
Other races	0.65	[0.49, 0.87]	0.0039	1.50	[1.32, 1.69]	<0.0001	0.45	[0.35, 0.57]	<0.0001
Metropolitan status
Metro (ref)
Non-metro	1.10	[0.93, 1.30]	0.2579	1.06	[0.96, 1.17]	0.2486	1.07	[0.94, 1.22]	0.3076
Rural	1.07	[0.74, 1.54]	0.7200	1.08	[0.87, 1.35]	0.4710	1.49	[1.15, 1.94]	0.0027
Marital status
Married (ref)
Wid/Sep/Div	1.06	[0.93, 1.21]	0.3616	0.95	[0.89, 1.02]	0.1640	1.41	[1.28, 1.56]	<0.0001
Never married	1.16	[0.95, 1.40]	0.1399	1.01	[0.91, 1.12]	0.8513	1.18	[1.02, 1.38]	0.0290
Region
Northeast (ref)
South	1.59	[1.32, 1.91]	<0.0001	0.55	[0.50, 0.60]	<0.0001	1.82	[1.57, 2.11]	<0.0001
North central	1.41	[1.15, 1.75]	0.0012	0.74	[0.67, 0.82]	<0.0001	1.63	[1.37, 1.93]	<0.0001
West	1.42	[1.18, 1.71]	0.0003	0.66	[0.60, 0.72]	<0.0001	2.05	[1.77, 2.39]	<0.0001
Medicaid eligibility
Yes	1.26	[1.08, 1.49]	0.0038	1.08	[0.97, 1.20]	0.0921	1.73	[1.54, 1.94]	<0.0001
No (ref)
Pain specialist visit
Yes	0.91	[0.80, 1.04]	0.1569	0.47	[0.44, 0.51]	<0.0001	1.16	[1.05,1.28]	0.0022
No (ref)
Substance abuse
Yes	1.46	[1.19, 1.81]	0.0004	0.71	[0.62, 0.83]	<0.0001	2.20	[1.91, 2.55]	<0.0001
No (ref)
Depression
Yes	1.04	[0.87, 1.23]	0.6730	0.87	[0.78, 0.97]	0.0120	1.06	[0.94, 1.20]	0.3481
No (ref)
Anxiety
Yes	1.02	[0.85, 1.22]	0.8229	0.89	[0.80, 0.99]	0.0401	1.33	[1.18, 1.51]	<0.0001
No (ref)
Comorbidities (>=2)
Yes	1.02	[0.87, 1.20]	0.7861	0.96	[0.89, 1.04]	0.3645	0.94	[0.83, 1.07]	0.3580
No (ref)
Polypharmacy
Yes	1.48	[1.31, 1.66]	<0.0001	0.72	[0.68, 0.77]	<0.0001	1.84	[1.68, 2.02]	<0.0001
No (ref)
Prescription NSAIDs
Yes	1.17	[1.05, 1.31]	0.0047	0.58	[0.54, 0.62]	<0.0001	1.20	[1.10, 1.30]	<0.0001
No (ref)
Steroids
Yes	1.07	[0.96, 1.19]	0.2507	0.86	[0.81, 0.91]	<0.0001	1.15	[1.06, 1.25]	0.0013
No (ref)
Antidepressants
Yes	1.13	[0.99, 1.30]	0.0658	0.77	[0.71, 0.84]	<0.0001	1.79	[1.62, 1.97]	<0.0001
No (ref)
Benzodiazepine
Yes	1.01	[0.82, 1.25]	0.9115	0.89	[0.79, 1.01]	0.0824	1.40	[1.22, 1.62]	<0.0001
No (ref)
Age at cancer diagnosis (x5)	0.97	[0.96, 0.98]	0.1709	1.06	[1.05, 1.07]	<0.0001	0.94	[0.93, 0.95]	0.0002
Zipcode-level percentage below poverty line	1.00	[0.99, 1.00]	0.5456	1.00	[1.00, 1.00]	0.9087	1.00	[1.00, 1.01]	0.5105
Zip code level education (% less than HS education)	1.00	[1.00, 1.01]	0.3860	1.00	[1.00, 1.00]	0.5861	1.00	[1.00, 1.01]	0.0075
Fragmented care index (x 10)	0.98	[0.95, 1.00]	0.0567	1.02	[1.01, 1.04]	0.0011	0.95	[0.93, 0.97]	<0.0001
Medicare advantage penetration (x5)	0.96	[0.96, 0.97]	0.0104	1.01	[1.00, 1.01]	0.1431	0.96	[0.95, 0.97]	0.0004

Note: Based on 35,071 adults with NCPCs 67 years or older, diagnosed with incident primary breast, prostate, colorectal cancers, and NHL between 2009 and 2015, continuously enrolled in Fee-for-Service Medicare part A and B for 24 months, and continuously enrolled in Medicare part D for 12 months before cancer diagnosis. Group differences by opioid use group trajectory were tested with Chi-Square test.

Abbreviations: NHL: Non-Hodgkin Lymphoma, Wid/Sep/Div: Widow/Separate/Divorce.

The fully adjusted multivariable multinomial logistic regression (MLR) revealed that the group membership differed by some SDoH, mental health, pain medication, and health care use factors. For example, Non-Hispanic Black (AOR = 1.69; 95%CI = 1.48, 1.93, compared to Non-Hispanic White) and rural residents (AOR = 1.49; 95%CI = 1.15, 1.94, compared to metropolitan) were more likely to be in “persistent use.” Older adults with anxiety (AOR = 1.33; 95%CI = 1.18, 1.51)) and pain medications (NSAIDs (AOR = 1.20; 95%CI = 1.10, 1.30, P < 0.0001), steroids (AOR = 1.15; 95%CI = 1.06, 1.25, P = 0.0013), polypharmacy (AOR = 1.84; 95%CI = 1.68, 2.02, P < 0.0001)) were among some of the factors associated with membership in the “persistent use.” Group membership in the “persistent use” was less likely among those with fragmented care (AOR = 0.95, 95%CI = 0.93, 0.97, P < 0.0001) and those living in higher Medicare advantage penetration counties (AOR = 0.96; 95%CI = 0.95, 0.97, P = 0.0004).

On the other hand, some SDoH, anxiety and pain medications, and health care use were associated with lower odds of short-term opioid use after cancer diagnosis. For example, Latino/Hispanics (AOR = 0.85; 95%CI = 0.76, 0.96, P = 0.0059, compared to Non-Hispanic Whites) were less likely to have short-term opioid use after cancer diagnosis. Individuals with anxiety (AOR = 0.89, 95%CI = 0.80, 0.99, P = 0.0401) and pain medication use (NSAIDs (AOR = 0.58; 95%CI = 0.54, 0.62, P < 0.0001), steroids (AOR = 0.86; 95%CI = 0.81, 0.91, P < 0.0001) were less likely to be in “short-term use after cancer diagnosis” group. Group membership in the “short-term use after cancer diagnosis” was more likely among those with fragmented care (AOR = 1.02, 95%CI = 1.01, 1.04, P = 0.0011).

Discussion

This study examined distinctive trajectories of prescription opioid use over time (before and after cancer) and factors associated with the trajectories among older NCPC adults with incident breast, colorectal, non-Hodgkin’s lymphoma, and prostate cancers. This study identified trajectory models that further elucidated our understanding of the pattern of older adults’ opioid use before and after cancer. In this study cohort, we identified four distinct patterns of opioid use. We found that 40.6% of older adults with NCPC and incident cancer used opioids for a short-term, in accordance with the guidelines that suggest discontinuing opioid use after the acute treatment phase. We also found that 41.0% had a low probability of opioid use over time. However, 12.4% belonged to the “persistent use” group. Taken together, these findings suggest that there is heterogeneity in opioid use patterns and an overwhelming majority of older adults with NCPC and cancer may be prescribed opioids consistent with the guidelines. In addition, we also observed that in all trajectory groups, opioid use peaked during the 90 days after a cancer diagnosis. The high probability of opioid use during the acute phase of cancer treatment may be warranted because most cancer patients experience pain associated with cancer growth or cancer treatments.^28,29

The high probability of opioid use, even after the acute treatment phase, in some individuals is concerning because these individuals are more likely to experience opioid use disorder, abuse, and overdose deaths.^30-32 The high probability of persistent opioid use may reflect inappropriate use and, as stated in the introduction, can be explained from the neuropsychological perspective.³³ This finding also reinforces the need for linking opioid prevention and control strategies across multiple sectors (community/social support, infrastructure, education, and employers)³⁴ and integrating “health in all policies.”^35,36 Future studies need to examine and closely monitor the subgroup population more likely to be in the “persistent use” group to provide necessary strategies to begin and maintain opioid high-use recoveries.

This study found that race was one of the factors associated with being persistent opioid users. Although numerous studies have shown the relationships between race and health outcomes,³⁷ in our case, it is strenuous to use this type of database analysis to pinpoint or establish the causal effect of race on the outcome. The reason is that the health outcome measured might also be interconnected with genetic, biological, or other factors.^37,38 Future studies integrating a racial viewpoint in this study area are needed to confirm the findings.³⁸

A noteworthy finding was that several SDoH were associated with the persistent use of opioids. For example, rural residents were more likely to be persistent users of opioids. It has been reported that opioid prescribing rates are higher in rural counties than in urban areas.³⁹ Persistent opioid use can be partially explained by the lack of affordable alternative treatment options, such as integrative health therapies for pain management in rural communities.^40,41 Therefore, some studies have recommended efforts to prevent and overcome opioid use issues in rural areas, including improving local access to treatment by minimizing travel distances^42,43 and reducing stigma through public education about the complexities of opioid use and the value of harm reduction.^43,44

Anxiety was associated with the persistent use of opioids over time. Anxiety and chronic pain may have a reciprocal relationship⁴⁵ This reciprocity of worry and pain among people with NCPCs and cancer may create a vicious cycle that would create a higher psychological and physiological need for opioids that make them require more time to taper off.^45,46

Pain medications (NSAIDs, steroids, and antidepressants) were among the risk factors of persistent opioid use. Individuals prescribed NSAIDs, steroids, and antidepressants were less likely to be in the “short-term use after cancer diagnosis” group. NSAIDs, steroids, and antidepressants are often used as supportive medications for pain, along with opioids.⁴⁷ It is plausible that the persistent users have greater pain intensity, necessitating additional nonopioid pain medications. Future studies are needed to explore the interactions of nonopioid pain medications and pain intensity on the persistent use of opioids.

Studies have shown that the relationship between polypharmacy (in our case, defined as ≥6 drugs) and opioid use is complex.⁴⁸ The speculation is that opioids are associated with their adverse effects, which lead to the need for more drugs for the patients. Polypharmacy may also cause adverse effects that result in pain depending on the drugs used.^48,49 In addition, the side effects associated with opioids can interfere with treatments for comorbid conditions not associated with pain, especially among patients with cancer.⁵⁰

This study found that patients with a higher fragmentation of care were less likely to have “persistent use.” This is in contrast to findings from a study based on data from 20% random sample of Medicare beneficiaries,⁵¹ which suggested that the risk of opioid use was higher in those seeking care from multiple providers. Our finding is also inconsistent with published studies, although not specific to older adults with any NCPCs and cancer, that have reported undesirable outcomes with fragmented care.^52,53 It is plausible seeing multiple providers for cancer may trigger opportunities for closer surveillance of opioid use and prescriptions. We cannot also rule out the possibility that for patients with complex chronic conditions such as cancer and NCPC, the fragmented care index may reflect appropriate care due to the need for care from multiple providers.

We also observed that higher county-level Medicare Advantage penetration rates were associated with lower odds of being in the “persistent use” group. This is consistent with the National Bureau of Economic Research report that found a reduction in the likelihood of opioid prescription in managed care prescription drug plans.⁵⁴ The reduction could be due to integrated care in Medicare Advantage plans leading to improved quality of care.⁵⁵ Plausible explanations for these area-level effects may be related to the so-called “spillover effects.”⁵⁶ An increase in the proportion of managed care plans in an area can affect physician practice patterns both for managed care and fee-for-service patients; the resultant spillover of managed care policies eventually benefits all patients in their care.⁵⁷

This study has several strengths: (1) this is the first US study to examine trajectories of prescription opioid use over time and factors associated with the trajectories among older cancer survivors with any NCPC and filled a critical knowledge gap; (2) opioid use trajectories use may be of interest to providers, patients, policymakers, and payers; (3) use of nationally representative real-world data from publicly insured older adults; and (4) comprehensive list of factors at the individual-, county-, and zip code-levels to profile the opioid trajectory groups.

Limitations of this study are: (1) As our study cohort was limited to FFS older adults, results are not generalizable to all Medicare beneficiaries; (2) This study is restricted to general opioids due to the data availability; therefore, unable to quantify the effect of different opioid class; and (3) Reliance on prescription fill data may not reflect the actual use of the medication.

Conclusion

One in eight older adults had persistent use of opioids over time. Policies and programs to reduce chronic opioid use need to consider the intra- and inter-individual variability to reduce opioid-related morbidity and mortality.

Supplemental Material

Supplemental Material - Prescription Opioid Use before and after Diagnosis of Cancer Among Older Cancer Survivors With Non-Cancer Chronic Pain Conditions (NCPCs): An Application of Group-Based Trajectory Modeling (GBTM)

Supplemental Material for Prescription Opioid Use before and after Diagnosis of Cancer Among Older Cancer Survivors With Non-Cancer Chronic Pain Conditions (NCPCs): An Application of Group-Based Trajectory Modeling (GBTM) by Rudi Safarudin, Traci LeMasters, Salman Khan, and Usha Sambamoorthi in Cancer Control.

Footnotes

Acknowledgments

The collection of cancer incidence data used in this study was supported by the California Department of Public Health pursuant to California Health and Safety Code Section 103885; Centers for Disease Control and Prevention’s (CDC) National Program of Cancer Registries, under cooperative agreement 1NU58DP007156; the National Cancer Institute’s Surveillance, Epidemiology and End Results Program under contract HHSN261201800032I awarded to the University of California, San Francisco, contract HHSN261201800015I awarded to the University of Southern California, and contract HHSN261201800009I awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the author(s) and do not necessarily reflect the opinions of the State of California, Department of Public Health, the National Cancer Institute, and the Centers for Disease Control and Prevention or their Contractors and Subcontractors.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

This research was, in part, funded by the National Institutes of Health (NIH) Agreement No. 1OT2OD032581 (Usha Sambamoorthi). The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the NIH.

IRB Statement

West Virginia University’s Institutional Review Board (IRB) approved this study’s ethics with Acknowledgement of Not Human Subjects Research (NHSR) - protocol # 2109427002 - as no individually identifiable private information was involved.

ORCID iDs

Rudi Safarudin

Traci LeMasters

Salman Khan

Usha Sambamoorthi

Supplemental Material

Supplemental material for this article is available online.

Baseline Characteristics of Older Medicare Beneficiaries with any NCPC Diagnosed with Breast,Colorectal,Prostate Cancers,or NHL Linked Surveillance,Epidemiology,and End Results (SEER) and Medicare,2007-2015.

ALL	N	%
ALL	35,071	100
Sex
Female	20,527	58.5
Male	14,544	41.5
Race and ethnicity
Non-hispanic white	26,605	75.9
Non-hispanic black	3,200	9.1
Latino/hispanic	2,710	7.7
Other races	2,070	5.9
Marital status
Married	17,214	49.1
Wid./Sep/Div	11,246	32.1
Never married	2,943	8.4
Substance abuse
Yes	1,756	5.0
No	33,315	95.0
Type of cancer
Breast	14,744	42.0
Prostate	11,503	32.8
Colorectal	5,964	17.0
Non-hodgkin lymphoma	2,860	8.2
Cancer stage
Early stage	27,953	79.7
Late stage	4,862	13.9
Comorbidity (>=2)
Yes	27,811	79.3
No	7,260	20.7
Depression
Yes	3,996	11.4
No	31,075	88.6
Anxiety
Yes	3,187	9.1
No	31,884	90.9
Prescription NSAIDs
Yes	10,267	29.3
No	24,804	70.7
Steroids
Yes	16,242	46.3
No	18,829	53.7
Antidepressant
Yes	8,303	23.7
No	26,768	76.3
Benzodiazepine
Yes	2,021	5.8
No	33,050	94.2
Polypharmacy (≥6)
Yes	16,442	46.9
No	18,629	53.1
Medicaid dual eligibility
Yes	4,284	12.2
No	30,787	87.8
Pain specialist visit
Yes	7,770	22.2
No	27,301	77.8
Diagnosis year
2009	6,259	17.8
2010	6,589	18.8
2011	6,892	19.7
2012	7,224	20.6
2013	8,107	23.1
Metropolitan status
Metropolitan	29,528	84.2
Non-metropolitan	4,823	13.8
Rural	697	2.0
Region
Northeast	7,065	20.1
South	8,158	23.3
North central	4,294	12.2
West	15,554	44.4

	Mean	SD
Age	75.95	6.31
Fragmented care index (x 10)	7.3	2.48
Numaber of physical comorbidities	3.05	1.91
Neighborhood characteristics
Zip code level median income	$41,992.13	$25,038
Zip code level education (% less than HS education)	24.12%	16.73%
Percentage below poverty line	14.47	12.43
% medicare advantage penetration	24.99	12.84

Abbreviations: NHL: Non-Hodgkin Lymphoma, Wid/Sep/Div : Widow/Separate/Divorce, HS : High School, SD: Standard Deviation.

References

Colvin

Bull

Hales

. Perioperative opioid analgesia—when is enough too much? A review of opioid-induced tolerance and hyperalgesia. Lancet. 2019;393(10180):1558-1568. doi:10.1016/S0140-6736(19)30430-1

Lee

Silverman

Hansen

Patel

Manchikanti

. A comprehensive review of opioid-induced hyperalgesia - PubMed. Pain Physician. 2011. https://pubmed.ncbi.nlm.nih.gov/21412369/ Accessed 17 August 2020.

Miller

Guy

Zhang

Mikosz

. Prevalence of nonopioid and opioid prescriptions among commercially insured patients with chronic pain. Pain Med. 2019;20(10):1948-1954. doi:10.1093/pm/pny247

Patel

Shah

Sambamoorthi

. Healthcare expenditures associated with persistent opioid use among adults with chronic non-cancer pain conditions: a retrospective cohort study. J Pain Palliat Care Pharmacother. 2020;34(3):127-140. doi:10.1080/15360288.2020.1734710

Marshall

Hale

. Older adults and the opioid crisis. Home Healthc Nurse. 2019;37(2):117. doi:10.1097/NHH.0000000000000743

Larsson

Hansson

Sundquist

Jakobsson

. Chronic pain in older adults: prevalence, incidence, and risk factors. Scand J Rheumatol. 2017;46(4):317-325. doi:10.1080/03009742.2016.1218543

Prostran

Vujović

Vučković

, et al. Pharmacotherapy of pain in the older population: the place of opioids. Front Aging Neurosci. 2016;8(JUN):144. doi:10.3389/FNAGI.2016.00144

Ramachandran

Salkar

Bentley

Eriator

Yang

. Patterns of long-term prescription opioid use among older adults in the United States: a study of medicare administrative claims data. Pain Physician. 2021;24(1):31-40. doi:10.36076/ppj.2021.24.31-40

Raffa

Pergolizzi

. Opioid-induced hyperalgesia: is it clinically relevant for the treatment of pain patients? Pain Manag Nurs. 2013;14(3):e67-e83. doi:10.1016/j.pmn.2011.04.002

10.

Hayhurst

Durieux

. Differential opioid tolerance and opioid-induced hyperalgesia: a clinical reality. Anesthesiology. 2016;124(2):483-488. doi:10.1097/ALN.0000000000000963

11.

Dowell

Haegerich

Chou

. CDC Guideline for Prescribing Opioids for Chronic Pain — United States, 2016. New York, NY: CDC. doi:10.15585/mmwr.rr6501e1externalicon

12.

Dowell

Ragan

Jones

Baldwin

Chou

. CDC clinical practice guideline for prescribing opioids for pain — United States, 2022. MMWR Recomm Rep (Morb Mortal Wkly Rep). 2022;71(3):1-95. doi:10.15585/MMWR.RR7103A1

13.

Wei

YJJ

Chen

Lewis

Schmidt

Winterstein

. Trajectories of prescription opioid dose and risk of opioid-related adverse events among older Medicare beneficiaries in the United States: a nested case–control study. PLoS Med. 2022;19(3):e1003947. doi:10.1371/JOURNAL.PMED.1003947

14.

Binswanger

Shetterly

, et al. Opioid dose trajectories and associations with mortality, opioid use disorder, continued opioid therapy, and health plan disenrollment. JAMA Netw Open. 2022;5(10):e2234671. doi:10.1001/JAMANETWORKOPEN.2022.34671

15.

Institute

. Brief Description of SEER-Medicare Database. Bethesda, MA: National Institute of Health; 2019. https://healthcaredelivery.cancer.gov/seermedicare/overview/, Accessed 6 January 2023.

16.

US-DHHR . Area health resources files. Health resources and services administration. https://data.hrsa.gov/topics/health-workforce/ahrf. Accessed 6 September 2022.

17.

Clinical classifications software (CCS) for ICD-9-CM. Accessed January 22, 2023.https://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp

18.

Clinical classifications software (CCS) for ICD-10-PCS (beta version). Accessed January 22, 2023.https://www.hcup-us.ahrq.gov/toolssoftware/ccs10/ccs10.jsp

19.

Generic name | ResDAC. Accessed January 23, 2023.https://resdac.org/cms-data/variables/generic-name

20.

Park

Park-Park’s textbook of preventive and social medicine-banarsidas bhanot (2015) Ayurveda | Epidemiology. Jabalpur, India: Bhanot. Accessed May 20, 2021. https://www.scribd.com/document/326461178/K-Park-Park-s-Textbook-of-Preventive-and-Social-Medicine-Banarsidas-Bhanot-2015-1

21.

Saultz

. Defining and measuring interpersonal continuity of care. Ann Fam Med. 2003;1(3):134-143. doi:10.1370/AFM.23

22.

Nagin

. Group-Based Modeling of Development. Cambridge, MA: Harvard University Press; 2005.

23.

Bryk

Raudenbush , SW. Hierarchical linear models: Applications and data analysis methods. Sage Publications, Inc. American Psychological Association-PsycNet. 1992. Accessed January 26, 2023. https://psycnet.apa.org/record/1992-97494-000

24.

Equator . Selecting the appropriate reporting guideline for your article | EQUATOR Network. UK EQUATOR Centre - Centre for Statistics in Medicine (CSM), NDORMS, University of Oxford. UK. https://www.equator-network.org/toolkits/selecting-the-appropriate-reporting-guideline/. Accessed August 5, 2024.

25.

von Elm

Altman

Egger

Pocock

Gøtzsche

Vandenbroucke

STROBE Initiative . The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol. 2008;61(4):344-349. doi:10.1016/J.JCLINEPI.2007.11.008

26.

Nagin

. Analyzing developmental trajectories: a semiparametric, group-based approach. Psychol Methods. 1999;4(2):139-157. doi:10.1037/1082-989X.4.2.139

27.

Nagin

Odgers

. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol. 2010;6:109-138. doi:10.1146/annurev.clinpsy.121208.131413

28.

Novy

Nelson

Koyyalagunta

Cata

Gupta

. Pain, opioid therapy, and survival: a needed discussion. Pain. 2020;161(3):496-501. doi:10.1097/J.PAIN.0000000000001736

29.

Page

Blanchard

. Opioids and cancer pain: patients’ needs and access challenges. J Oncol Pract. 2019;15(5):229-231. doi:10.1200/JOP.19.00081

30.

Karila

Marillier

Chaumette

Billieux

Franchitto

Benyamina

. New synthetic opioids: Part of a new addiction landscape. Neurosci Biobehav Rev. 2019;106:133-140. doi:10.1016/j.neubiorev.2018.06.010

31.

Blendon

Benson

. The public and the opioid-abuse epidemic. N Engl J Med. 2018;378(5):407-411. doi:10.1056/nejmp1714529

32.

Bohnert

ASB

Valenstein

Bair

, et al. Association between opioid prescribing patterns and opioid overdose-related deaths. JAMA. 2011;305(13):1315-1321. doi:10.1001/jama.2011.370

33.

Volkow

McLellan

. Opioid abuse in chronic pain—misconceptions and mitigation strategies. N Engl J Med. 2016;374(13):1253-1263.

34.

ARC . Investments supporting partnerships in recovery ecosystems initiative - appalachian regional commission. Appalachian Regional Commission. https://www.arc.gov/sud/ Accessed 25 January 2023.

35.

Health in all policies. Accessed January 26, 2023.https://www.apha.org/topics-and-issues/health-in-all-policies

36.

Health in All Policies | AD for Policy and Strategy. Atlanta, Georgia, United States: CDC. Accessed January 26, 2023.https://www.cdc.gov/policy/hiap/index.html

37.

Flanagin

Frey

Christiansen

Bauchner

. The reporting of race and ethnicity in medical and science journals: comments invited. JAMA. 2021;325(11):1049-1052. doi:10.1001/JAMA.2021.2104

38.

Boyd

Lindo

Weeks

McLemore

. On racism: a new standard for publishing on racial health inequities. Health Affairs Forefront 2020. https://www.healthaffairs.org/content/forefront/racism-new-standard-publishing-racial-health-inequities.

39.

García

Heilig

Lee

, et al. Opioid prescribing rates in nonmetropolitan and metropolitan counties among primary care providers using an electronic health record system — United States, 2014–2017. MMWR Morb Mortal Wkly Rep. 2019;68(2):25-30. doi:10.15585/MMWR.MM6802A1

40.

Guy

Zhang

Bohm

, et al. Vital signs: changes in opioid prescribing in the United States, 2006–2015. MMWR Morb Mortal Wkly Rep. 2019;66(26):697-704. doi:10.15585/MMWR.MM6626A4

41.

Palombi

St Hill

Lipsky

Swanoski

Lutfiyya

. A scoping review of opioid misuse in the rural United States. Ann Epidemiol. 2018;28(9):641-652. doi:10.1016/J.ANNEPIDEM.2018.05.008

42.

Liu

Zhang

Bonny

Nahata

. Strategies to Improve Access to Care for Patients With Opioid Use Disorder. https://doi.org/10.1177/10600280241273258. September 2024. doi:10.1177/10600280241273258

43.

Stopka

Estadt

Leichtling

, et al. Barriers to opioid use disorder treatment among people who use drugs in the rural United States: A qualitative, multi-site study. Soc Sci Med. 2024;346:116660. doi:10.1016/J.SOCSCIMED.2024.116660

44.

Ezell

Walters

Friedman

, et al. Stigmatize the use, not the user? Attitudes on opioid use, drug injection, treatment, and overdose prevention in rural communities. Soc Sci Med. 2021;268:113470. doi:10.1016/J.SOCSCIMED.2020.113470

45.

Thielking

. Cancer pain and anxiety. Curr Pain Headache Rep. 2003;7(4):249-261. doi:10.1007/S11916-003-0045-X

46.

Langdon

Dove

Ramsey

. Comorbidity of opioid-related and anxiety-related symptoms and disorders. Curr Opin Psychol. 2019;30:17-23. doi:10.1016/J.COPSYC.2018.12.020

47.

Yang

Bauer

Wahner-Roedler

Chon

Xiao

. The modified WHO analgesic ladder: is it appropriate for chronic non-cancer pain? J Pain Res. 2020;13:411-417. doi:10.2147/JPR.S244173

48.

Matos

Bankes

Bain

Ballinghoff

Turgeon

. Opioids, polypharmacy, and drug interactions: a technological paradigm shift is needed to ameliorate the ongoing opioid epidemic. Pharmacy: Journal of Pharmacy Education and Practice. 2020;8(3):154. doi:10.3390/PHARMACY8030154

49.

Giummarra

Gibson

Allen

Pichler

Arnold

. Polypharmacy and chronic pain: harm exposure is not all about the opioids. Pain Med. 2015;16(3):472-479. doi:10.1111/PME.12586/SUPPINFO

50.

Kotlinska-Lemieszek

Paulsen

Kaasa

Klepstad

. Polypharmacy in patients with advanced cancer and pain: a European cross-sectional study of 2282 patients. J Pain Symptom Manag. 2014;48(6):1145-1159. doi:10.1016/J.JPAINSYMMAN.2014.03.008

51.

Jena

Goldman

Weaver

Karaca-Mandic

. Opioid prescribing by multiple providers in Medicare: retrospective observational study of insurance claims. BMJ. 2014;348:g1393. doi:10.1136/BMJ.G1393

52.

Juo

Sanaiha

Khrucharoen

Chang

Dutson

Benharash

. Care fragmentation is associated with increased short-term mortality during postoperative readmissions: a systematic review and meta-analysis. Surgery. 2019;165(3):501-509. doi:10.1016/J.SURG.2018.08.021

53.

Castro-Sánchez

Charani

Drumright

Sevdalis

Shah

Holmes

. Fragmentation of care threatens patient safety in peripheral vascular catheter management in acute care--a qualitative study. PLoS One. 2014;9(1):e86167. doi:10.1371/JOURNAL.PONE.0086167

54.

Baker

Bundorf

Kessler

. The effects of medicare advantage on opioid use. J Health Econ. 2020;70:102278. doi:10.1016/J.JHEALECO.2019.102278

55.

DuGoff

Tabak

Diduch

Garth

. Quality, health, and spending in medicare advantage and traditional medicare. Am J Manag Care. 2021;27(9):395-400. doi:10.37765/AJMC.2021.88641

56.

Kane

Butler

Fink

, et al. Interventions to Prevent Age-Related Cognitive Decline, Mild Cognitive Impairment, and Clinical Alzheimer’s-type Dementia. Rockville, MA: Agency for Healthcare Research and Quality (US); 2017.

57.

Baicker

Chernew

Robbins

. The spillover effects of medicare managed care: medicare advantage and hospital utilization. J Health Econ. 2013;32(6):1289-1300. doi:10.1016/J.JHEALECO.2013.09.005

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.25 MB