We used a 20% national sample of Medicare enrollees with a diagnosis of AD/ADRD and chronic pain in 2020 based on algorithms from the Chronic Conditions Data Warehouse (CCW) [14, 15]. We used the Master Beneficiary Summary File, Medicare Provider Analysis and Review file, Outpatient Standard Analytic File, Carrier, and Prescription Drug Event files. We included those with continuous Part A, B, and D enrollment from 2018 to 2020 and complete demographic information. AD/ADRD diagnosis required a 3-year reference period [14]. We operationalized chronic pain using the algoritihm from the CCW, requiring one inpatient claim or two other nondrug claims that were at least 1 day apart over a 2-year reference period [15]. Our study was approved by the University of Texas Medical Branch Institutional Review Board (IRB: 16-0247).

Our exposure was the prescription of dementia medications in 2020, specifically memantine and ACHEIs. This was categorized as a four-level variable: none, memantine only, ACHEI only, or both memantine and ACHEI. Our main outcome included prescription pain medications. These included prescription opioids, serotonin and norepinephrine reuptake inhibitors (e.g., duloxetine, venlafaxine), gabapentinoids (gabapentin and pregabalin), and non-steroidal anti-inflammatory drugs. We examined (1) having any pain management prescription; (2) having any opioid prescription; and (3) having a long-term opioid prescription, defined as at least 90 days of consecutive opioid prescriptions in 2020. We identified prescription opioids according to the National Drug Code (complete, partial, and combination opioid agonists were included). Therapeautic classes of prescription medications included in our analyses are included in Supplementary Table 1.

Covariates included age, sex, race and ethnicity, region, original entitlement (disability/end stage renal disease, old age), dual Medicaid coverage, chronic conditions, mental health conditions, opioid use disorder, drug or alcohol abuse, and pain type/location. Chronic conditions included were cancer, hypertension, heart failure, ischemic heart disease, arrhythmia, hyperlipidemia, stroke/transient ischemic attack (TIA), arthritis, asthma, autism, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), diabetes, human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS), liver disease, osteoporosis, and schizophrenia. There was a total of 17 conditions used, and all but arrhythmia were from the CCW [16]. Arrhythmia used an International Classification of Diseases, Tenth Revision (ICD-10) code of I48.x. Cancer included any breast, colorectal, prostate, lung, endometrial, and leukemia/lymphoma cancer diagnosis, but only counted as one condition. Chronic conditions were summed from 0 to 17. Mental health conditions included depression, bipolar disorder, and anxiety. Pain type/locations included back pain, neck pain, headaches, general chronic pain, abdominal/chest pain, cancer, musculoskelal pain, fractures, visceral pain, wound pain, and other pain. A list of ICD codes used to identify pain type/location is detailed in Supplementary Table 2.

We used logistic regression models to assess the odds of (1) having any pain management prescription, (2) having an opioid prescription, and (3) having a long-term (≥ 90 days) opioid prescription, by dementia medication group. We controlled for age, sex, race and ethnicity, region, original entitlement, dual Medicaid coverage, chronic conditions count, mental health conditions (depression, anxiety, and bipolar), opioid use disorder, drug or alcohol abuse, and pain type/location. We conducted sensitivity analyses where we (1) analyzed the three classes of ACHEIs separately and (2) excluded those with cancer. All analysis was conducted with SAS Enterprise Guide 7.1 (SAS Inc., Cary, NC) .