Assessment of Joint and Interactive Effects of Multimorbidity and Chronic Pain on ADRD Risk in the Elder Population

Objective Multimorbidity and non-cancer chronic pain conditions (NCPC) are independently linked to elevated risk for cognitive impairment and incident Alzheimer’s Disease and Related Dementias (ADRD)-both - We present the study of potential joint and interactive effects of these conditions on the risk of incident ADRD in older population. Methods This retrospective-cohort study drew baseline and 2-year follow-up data from linked Medicare claims and Medicare Current Beneficiary Survey (MCBS). Baseline multimorbidity and NCPC were ascertained using claims data. ADRD was ascertained at baseline and follow-up. Results NCPC accompanied by multimorbidity (vs. absence of NCPC or multimorbidity) had a significant and upward association with incident ADRD (adjusted odds ratio (AOR): 1.72, 95% CI 1.38, 2.13, p < 0.0001). Secondary analysis by number of comorbid conditions suggested that the joint effects of NCPC and multimorbidity on ADRD risk may increase with rising number contributing chronic conditions. Interaction analyses indicated significantly elevated excess risk for incident ADRD.


Introduction
A dramatic rise in prevalence of Alzheimer's disease and related dementias (ADRD), both worldwide and in the United States (U.S.), is projected in the next decade owing in part to increasing longevity and growing number of older adults ("Alzheimer's Association Facts and Figures," 2021). Paralleling the increase in ADRD prevalence is a corresponding escalation in disability, dependence, and institutionalization rates accompanied by overwhelming healthcare costs (Nichols et  Multimorbidity is on the rise globally as well as in the U.S. Multimorbidity, de ned as the concurrent presence of 2 or more chronic conditions (Johnston, Crilly, Black, Prescott, & Mercer, 2018), has traditionally been overlooked in ADRD epidemiology where individual risk factors were included and adjusted for in the epidemiologic analyses. More recent developments suggest that the study of interaction between chronic comorbid conditions is as important as their individual effects on the risk for ADRD. In addition, underreported chronic conditions often get overlooked while studying multimorbidity. For example, non-cancer chronic pain, a highly frequent yet understudied condition in older population.
Accumulating evidence suggests that both multimorbidity and certain non-cancer chronic pain conditions (NCPC), for example headache, bromyalgia, knee and joint pain, and neuropathic pain may likewise signi cantly increase risk for cognitive decline and new-onset dementia (Innes & Sambamoorthi, 2020; Wei, Levine, Zahodne, Kabeto, & Langa, 2020). Like that of ADRD, the prevalence of both multimorbidity and NCPCs increases steeply with advancing age, and there is also growing evidence that multimorbidity is strongly and reciprocally associated with chronic pain and related functional impairment, psychological distress, and other adverse sequalae (Nakad et al., 2020), factors also linked to poor cognitive outcomes (Lacreuse, Raz, Schmidtke, Hopkins, & Herndon, 2020). Extant evidence, although limited, suggests that multimorbidity can exacerbate chronic pain and vice versa (Nakad et al., 2020; Scherer et al., 2016), contributing to a vicious cycle of deteriorating health and function which may, in turn, further increase the risk for accelerated cognitive decline, cognitive impairment, and incident ADRD.
However, although there is some documented evidence of independent associations of NCPCs ) and multimorbidity (Wei et al., 2020) to ADRD risk, and growing evidence presents clues of an important and potentially synergistic role of both multimorbidity and chronic pain in the development of cognitive decline and ADRD in older adults (Jellinger & Attems, 2015;Nakad et al., 2020), the potential joint and interactive effect of chronic pain and multimorbidity involving other chronic health conditions on risk for incident ADRD remains unexplored. While chronic pain, a chronic health condition itself, is a leading cause of years lived with disability in the population, it suffers an underrepresentation in the published multimorbidity literature. In this study, we used linked data from a large, nationally representative sample of older United States Medicare bene ciaries to investigate the potential joint and interactive effects of multimorbidity from other chronic health conditions and NCPCs on incident ADRD. We hypothesized that co-occurring NCPCs and multimorbidity would amplify the effects of both conditions on ADRD risk, and that the joint effects of NCPC and multimorbidity would be more pronounced with a higher number of chronic conditions contributing to multimorbidity.

Study Design and Data Source
Multiple Medicare Current Bene ciary Survey (MCBS) cohorts from several years (2001-2013) were pooled to generate a large study sample with su cient study power to evaluate the joint effect of baseline multimorbidity and NCPCs on risk for incident ADRD. We employed a retrospective cohort study design, and our study cohorts were drawn from a nationally representative sample of Medicare fee-for-service bene ciaries 65 + years of age who completed the Medicare Current Bene ciary Survey (MCBS). MCBS uses a complex strati ed, three-stage probability sampling design for participant recruitment, and collects both cross-sectional and longitudinal data on health status, health services utilization, prescriptions, and healthcare costs using a series of survey interviews and administrative records; comprehensive explanation of procedures can be found in previous publications (MCBS; Mues et al., 2017). Our study speci cally used MCBS Cost and Use les linked with Medicare fee-for-service claims to ascertain demographics, access to care, lifestyle factors, medical conditions and medication use.

Eligibility Criteria and Analytic sample
The study analytic sample comprised non-institutionalized MCBS participants aged 65 + years at baseline, alive at the end of follow-up (de ned as the time of incident ADRD diagnosis or end of the 3-year study period) and enrolled in Medicare feefor-service (FFS) throughout the 3-year study period or until incident ADRD diagnosis. Participants with diagnosed ADRD at baseline were excluded. Applying a priori exclusion criteria yielded a nal analytic sample size of 16,934 adults. . Using a combination of claims and survey data to ascertain ADRD has been recommended by MCBS investigators to increase capture of ADRD, and has been shown to yield results similar to those of expert in-person-assessments (Rose et al., 2016).

Presence of Non-Cancer Chronic Pain Conditions (NCPCs)
FFS claims were used to ascertain NCPC status at baseline as either two outpatient claims (90 days apart) or one inpatient claim using ICD-9-CM codes as recommended by CMS ("Chronic Conditions Data Warehouse: Your source for national CMS Medicare and Medicaid research data," Access Yr 2020) and consistent with prior studies of NCPC (Sullivan, Turner, & Romano, 1991). Any NCPC was assessed as a binary variable (yes/no) during baseline, including ve common non-cancer pain conditions: back or neck pain, headache, joint pain, neuropathic pain, and osteoarthritis.

Presence of Multimorbidty
Multimorbidity was ascertained using FFS claims during the baseline year using ICD-9-CM ("Chronic Conditions Data Warehouse: Your source for national CMS Medicare and Medicaid research data," Access Yr 2020). The presence of multimorbidity was de ned as having 2 or more co-existing chronic physical health conditions (excluding NCPCs) in addition to the index disease (ADRD) (Buntinx & Knottnerus, 1996). The construction of multimorbidity variable was guided by the conceptual framework on multiple chronic conditions detailed by Richard Goodman et al. (Goodman et al., 2016), which provided standardized de nitions and identi cation approaches to improve the understanding and surveillance of individual as well as multiple co-occurring conditions (multimorbidity) implicated in high disease burden and escalating healthcare costs in the United States. Analytical variables for multimorbidity were constructed using 1) an aggregate binary variable (yes/no) de ned using a comprehensive set of diagnosed chronic conditions, other than NCPCs, at baseline, including: asthma, cardiac arrythmias, cerebrovascular disease, coronary artery disease, congestive heart failure, chronic obstructive pulmonary disease, cancer, diabetes, hypercholesterolemia, and hypertension; 2) to study the effects of number of chronic conditions comprising multimorbidity, a secondary variable was created with 3 categories: multimorbidity with 2-3 chronic conditions, multimorbidity with ≥ 4 chronic conditions, and no multimorbidity (reference category, < 2 chronic conditions).

Covariates
Factors widely reported or suspected to be associated with ADRD risk, chronic pain, and/or multimorbidity, were accounted for, using a priori inclusion criteria in our multivariable models. These included biological factors (age, sex and race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, and other)), demographic characteristics (marital status, educational level, family income, health insurance status), lifestyle factors (smoking status, body mass index), History of other chronic conditions at baseline not included in the multimorbidity variable, such as traumatic brain injury (TBI) and speci c auto-

Statistical Analysis
Differences in baseline characteristics by multimorbidity were assessed using Chi-square tests. Logistic regressions were used to assess the unadjusted and adjusted individual and joint associations of NCPCs and multimorbidity to incident ADRD. In the multivariable logistic regressions, we carried out incremental block-wise adjustment for demographic and socioeconomic factors, lifestyle factors, certain other chronic health conditions, and analgesics use. Effect of mood and sleep disorders on the adjusted joint associations of NCPC and multimorbidity to ADRD risk was assessed separately.
Investigation of joint and interactive effects in this study was guided by the recommendations for reporting effect modi cation and interactions in epidemiologic research by Knol and VanderWeele (Knol & VanderWeele, 2012). Speci cally, the recommendations involve reporting of 1) individual associations (odds ratios (OR)) of both the key independent variables; 2) joint effects (OR) of the key independent variables on the association with the outcome; 3) the relative excess risk due to interaction (RERI) for reporting excess risk on additive scale due to the presence of another key variable, as well as the risk due to interaction on multiplicative scale. To investigate the added risk of NCPC speci cally for each level of multimorbidity, i.e., multimorbidity with 2-3 chronic conditions and with ≥ 4 chronic conditions, we performed regression analyses with contrast using respective reference categories for each contrast.
For assessment of potential interactions in this study, we calculated individual (OR NCPC, no Multimorbidity , OR Multimorbidity, no NCPC ) and joint (OR NCPC, Multimorbidity ) effects, using the same reference category (no multimorbidity and no NCPC). Additive risk due to interaction was calculated using RERI, whereas separate logistic regression contrast models for 2-3 chronic conditions and ≥ 4 chronic conditions were used to compare the increment in the risk in the presence of NCPC.

Results
Our analytic sample comprised 16,934 eligible participants. Table 1 summarizes the study characteristics distributed by the presence or absence of multimorbidity. The majority of our analytic sample with multimorbidity were female (58%), white (84%), married (52%), lived in metropolitan areas (73%), had private insurance (80%), and reported a family income at or above 200% of the FPL (51%). Participants with multimorbidity had higher use of opioid analgesics (22% vs. 12%) and NSAIDs (21% vs. 13%). Overall baseline prevalence of NCPCs in the multimorbidity group was 68% (weighted). The percentage of participants with baseline NCPCs varied signi cantly by number of co-occurring chronic conditions ((76% vs. 62% vs. 39%, for ≥ 4 vs. 2-3 vs. no multimorbidity, respectively, p < 0.0001) (data not shown). The proportion of participants with diagnosed depression, anxiety, or sleep disorders at baseline also rose with increasing number of comorbid conditions (p's < 0.0001). Participants with ≥ 4 co-occurring chronic conditions (vs. 2-3 and no multimorbidity) were also more likely to be older (80+), on Medicaid insurance, and to report lower family income. At the end of study follow-up period, 68 for every 1000 participants were diagnosed with incident ADRD (N = 1149). Incident ADRD rates were signi cantly higher in those with NCPC vs. without baseline NCPCs (7.3% vs 4.7%) and multimorbidity (6.3% vs 4.0%) (p's < 0.0001). Incident ADRD rates were also higher in women, non-Hispanic blacks, widowed, poorly educated, and participants < 200% FPL, Medicaid insurance and those who lacked private insurance (p's < 0.0001). Those with a history of TBI, as well as those who reported opioid analgesics use had a higher incidence rate compared to those without (p's < 0.0001). Similarly, those who were diagnosed with sleep, depression or anxiety disorders at baseline included a signi cantly higher proportion of incident ADRD cases (p's < 0.0001) (see supplementary table).   smoking status, body mass index; Also including history of traumatic brain injury, use of nonsteroidal antiin ammatory drugs and opioid analgesics, **contrast using logistic regression models; p-value from multiplicative interactions from fully adjusted logistic regression p = 0.773 indicates absence of multiplicative interaction Inclusion of potential mediators, i.e., sleep disorders, depression, and anxiety, in the fully adjusted models attenuated but did Additional analysis assessing relative excess risk for incident ADRD due to interaction between multimorbidity and NCPC (RERI) indicated that the additive interaction of NCPC with multimorbidity accounted for a relative excess risk of 36% (29.3, 42.6) ( Table-2c  *Study sample comprised 11 pooled cohorts (2001-2013) of 16,934 older community-dwelling adults aged 65 + years and enrolled in fee-for-service Medicare. Odds ratios (OR) and adjusted odds ratios (AOR) with 95% con dence intervals (CI) were calculated using separate logistic regression models. ¥Including socio-demographics: sex, age, race/ethnicity, education, income, Medicaid insurance, private insurance, marital status, region. ¥¥ Also including lifestyle factors: smoking status, body mass index; Also including history of traumatic brain injury, use of nonsteroidal antiin ammatory drugs and opioid analgesic

Discussion
To our knowledge, this investigation is the rst large retrospective cohort study to assess the joint and interactive effects of NCPC and multimorbidity, highly prevalent conditions in older adults, on the risk for incident ADRD, and to explore the potential synergistic effects of NCPC and chronic condition burden on incident ADRD risk in community-dwelling older adults. We found that participants with both NCPCs and multimorbidity at baseline were at signi cantly elevated risk for incident ADRD after adjustment for multiple potential confounders, with risk estimates substantially exceeding those for either NCPC or multimorbidity alone. Further analysis regarding the joint effects of NCPCs and multimorbidity by number of co-occurring chronic conditions showed similar results i.e., NCPC in the presence of both low (2-3 conditions) and high (≥ 4) chronic condition burden was signi cantly and positively associated with risk for incident ADRD, with risk estimates particularly pronounced in those with NCPC and ≥ 4 chronic conditions. Additional analyses regarding the interactive effects of NCPC's and multimorbidity indicated that the additive interaction of these conditions accounted for a signi cant positive excess risk for incident ADRD after adjustment for potential confounders. Excess risk measure (RERI) yielded ndings consistent with the primary analyses across all adjusted models; risk for ADRD remained signi cantly higher in the presence of co-occurring NCPC's and multimorbidity compared to the presence of either condition alone. In agreement with our primary hypothesis, these ndings suggest that multimorbidity may amplify the effects of NCPC on risk for ADRD and vice versa. Our ndings also suggest that the joint effects of NCPC and multimorbidity on ADRD risk may increase with rising number of contributing chronic conditions. . Collectively, these ndings support a likely contribution of multimorbidity and chronic pain in promoting cognitive dysfunction and neuropathologic changes, and, together with the probable reciprocal relationship between chronic pain and multimorbidity, suggest that these conditions may have adverse synergistic effects on ADRD risk.
In this study, we found that the joint effects of NCPC's and multimorbidity on ADRD risk were attenuated in strength and magnitude after inclusion of depression, anxiety, and insomnia-related sleep disorders, suggesting that these conditions may partially mediate the observed effects of co-occurring NCPC and multimorbidity on risk for incident ADRD. Multimorbidity, like NCPCs, has also been strongly and reciprocally associated with mental health conditions such as While it can be argued that analgesic medications may in part explain the observed associations of NCPCs to ADRD risk, ndings to date regarding the effects of opioid and NSAIDs analgesics use on cognitive function have been inconsistent (Etminan et al., 2003;Kurita, Pimenta, & Nobre, 2008). In this study, the use of neither NSAIDs nor opioid analgesics was related to risk for ADRD, nor did adjustment for these factors affect the joint or interactive effects of NCPC and multimorbidity on ADRD risk.

Limitations
One of the major limitations of this study was the shorter duration of follow-up. As we know of ADRD being an underdiagnosed condition with ambiguous manifestations, the shorter follow-up duration may have led to an underascertainment of ADRD, pulling the risk toward the null. Moreover, claims data were used to ascertain multimorbidity status, which may suffer from under-reporting, potentially diluting the effect of multimorbidity in the analysis. Furthermore, the measure of number of chronic conditions, implying high load of multimorbidity, does not account for the type and different combinations of chronic health conditions constituting multimorbidity load, and that certain combinations contributing to multimorbidity may carry more weight in terms of risk factors for ADRD than others. Therefore, multimorbidity categories (2-3 and ≥ 4 chronic conditions) may not always re ect relative disease burden. Since we desired to assess the contribution of NCPCs and multimorbidity, rst separately and then in tandem, it is important to note that our construct of multimorbidity only included select chronic physical health conditions while excluding any non-cancer chronic pain conditions as well as potential effect mediators including mental-and insomnia-related disorders. Additionally, the information on severity and stage of baseline chronic conditions was not available, and hence could not be included in our assessment of multimorbidity. Similarly, insu cient information on the severity or duration of NCPCs or chronic health conditions at the time of ascertainment hindered an in-depth assessment of the potentially dynamic in uence of multimorbidity on the association of NCPCs to ADRD risk. Moreover, ADRD, multimorbidity and chronic pain have been associated with increased mortality, thereby, introducing potential survivor bias through overrepresentation of low-risk individuals (alive at the end of study follow up) in our analyses.

Conclusion
In this large population-based study of older Medicare bene ciaries, co-occurring NCPC and multimorbidity were associated with signi cantly elevated risk for ADRD, with ADRD risk highest in those with the greatest number of chronic conditions. If con rmed in future longitudinal investigations with longer follow-up durations and sophisticated measures of multimorbidity burden, these ndings may aid in targeting at risk populations, help further motivate interventions designed to reduce NCPC and multimorbidity, and ultimately, inform strategies for decreasing ADRD risk. These ndings also argue that the potential modifying effects of multimorbidity should be considered when assessing the effects of chronic pain on risk for cognitive impairment and ADRD.