Antimuscarinic Discontinuation in Patients with Overactive Bladder in Nursing Homes: A Retrospective Study of Medicare Beneficiaries

The Minimum Data Set (MDS) linking 2013–2015 Medicare claims data involving Part A, B and D claims [17, 18] was used to examine treatment patterns and predictors of antimuscarinic discontinuation among LTNH residents with OAB. The MDS is a federally mandated nursing home health assessment tool for all residents in Medicare/Medicaid-certified nursing facilities. It is administered to all nursing home residents within 14 days of admission and at quarterly intervals thereafter. It captures detailed information on physical, psychologic and psychosocial functioning and active clinical diagnosis, health conditions, treatments and services. It also provides information on patient’s cognitive functioning and behavioral problems. Medicare Part A covers hospital care, initial care in skilled nursing facilities, hospice care and home health care. Part B covers services such as laboratory, ambulance, outpatient mental health and other physician services that are not included in Part A. Each Part A and Part B record contains up to ten diagnoses recorded according to International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. Part D, launched in 2006, covers prescription benefits for Medicare beneficiaries. Each Part D Prescription Drug Event claim comprises information on prescription drug fills, including drug name, fill date, days of supply and quantity.

The study protocol was approved by the University of Houston Committee for the Protection of Human Subjects. Informed consent was not required given the deidentified nature of the data and feasibility considerations for a large retrospective database such as Medicare.

The retrospective cohort consisted of LTNH residents diagnosed with OAB. LTNH residents were defined based on a previously validated MDS/Skilled Nursing Facility algorithm as residents with at least one nursing home episode lasting at least 101 consecutive days [19]. Patients were included in the cohort using a step-wise process if they met the following criteria: (1) ≥ 65 years of age at index date, which was defined as the first long-stay nursing home admission date; (2) at least one claim with an OAB diagnosis either during the nursing home episode or prior (identified based on previously published lists of ICD-9-CM codes) [12, 20, 21] or at least one prescription claim for OAB-specific medication (antimuscarinics, mirabegron or onabotulinumtoxinA specific for bladder usage) during nursing home stay (Supplementary Table 1); (3) Parts A, B and D coverage in the 3 months before the nursing home admission and until the end of the nursing home stay; (4) newly initiated antimuscarinic medication claims lasting at least 30 days continuously during nursing home stay.

Antimuscarinics initiated during the nursing home stay were used to examine treatment patterns. Antimuscarinic medications were operationally defined using the American Hospital Formulary Service classification system and identified using National Drug Codes (NDC) [22]. Antimuscarinic medications included darifenacin, fesoterodine, oxybutynin, solifenacin, tolterodine and trospium. Treatment discontinuation was defined as the treatment gap of ≥ 30 days between prescription refills prior to nursing home discharge date [23, 24]. Accordingly, patients were categorized into two groups: those who discontinued the index antimuscarinic medication and those who did not discontinue the index antimuscarinic medication [25].

A multivariable Cox proportional hazards (PH) regression analysis was performed to assess the time to discontinuation of antimuscarinic medication among OAB patients. The Andersen Behavioral Model (ABM) was used to guide selection of independent variables in the analysis [26, 27]. This model postulates that an individual’s use of health services is a function of predisposing, enabling and need characteristics [26, 27]. Variables corresponding to these factors were selected based on the published literature and data available in the Medicare claims and the MDS [11, 28]. Predisposing factors are the characteristics of an individual that were in existence before the illness and included age, gender, race/ethnicity and marital status. Enabling factors are related to the ability of an individual to secure the healthcare services and included region, Medicare-Medicaid dual eligibility and urban/rural area. Urban/rural area was captured using MDS as the data source. The urban/rural distinction was based on the location of nursing home facility. The need factors describe the perceived and actual health status of an individual. The need variables were captured from both the Medicare claims and the MDS. These included index antimuscarinic agent, comorbidities measured using the Elixhauser comorbidity scores [29] and falls, prescription medication use (such as antidepressants, diuretics, beta-blockers, calcium channel blockers, anticonvulsants, angiotensin converting enzyme [ACE] inhibitors, antipsychotics, antiparkinson agents and alpha-blockers), anticholinergic burden (measured using the Anticholinergic Cognitive Burden [ACB] scale [30]) and body mass index (BMI). All variables were measured 3 months prior to the first long-stay nursing home admission date. Need characteristics for the ABM were also determined based on the MDS assessment recorded at nursing home admission. These included the presence of urinary or bowel incontinence (defined as always continent, occasionally incontinent, frequently incontinent, always incontinent or not rated, the former of which has been used to characterize patients with urinary incontinence in several studies [31]), bladder continence management (i.e., use of an indwelling/external catheter, ostomy and/or intermittent catheterization), urinary toileting program, response to toileting program, current toileting program or trial, depressed mood indicator and scores on the Cognitive Performance Scale (CPS) [32] and Activities of Daily Living (ADL) scale [33]. Using the CPS, a higher score is indicative of greater impairment, ranging from 0 (intact cognition) to 6 (severe cognitive impairment). Mild cognitive impairment is associated with a score of 2. A higher score on the ADL is indicative of greater impairment/dependence [33]. Using the ADL, an individual’s independence is graded as dependent [4‐6], limited/extensive assistance [7‐9] and independent [10‐18]. Due to the higher number of missing values, UI-specific measures (i.e., use of an indwelling/external catheter, ostomy, intermittent catheterization) were not included in the multivariable Cox PH regression model.

Descriptive statistics were used to characterize LTNH residents with OAB and antimuscarinic use. For the study cohort, the discontinuation rate based on the index antimuscarinic medication was reported. Categorical and continuous patient characteristics were compared among patients who discontinued the index antimuscarinic medication versus those who did not discontinue, using chi-square and student’s t-tests, respectively. Time to discontinuation between different antimuscarinic users was compared using Kaplan-Meier curves and log-rank test. Proportionality hazards assumption was confirmed using the Schoenfeld test. Multivariable Cox proportional hazards regression analysis was performed to assess the risk of discontinuation among different antimuscarinic medications with time to discontinuation as the dependent variable. Independent variables were predisposing, enabling and need characteristics based on the ABM framework. Death and loss to follow-up due to lack of insurance coverage were considered as censoring criteria. All relevant variables available in the data set were included in the model.

All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC) with statistical significance set at an a priori- specified level of 0.05.

Figure 1 describes the process of cohort derivation. A total of 250,016 LTNH residents ≥ 65 years of age with evidence of OAB and continuous enrollment were initially identified. From this group, 11,012 (4.4%) LTNH residents were newly initiated with antimuscarinic prescription for OAB and had continuous supply of medication for at least 30 days dispensed during the LTNH stay. The mean (± SD) age of the cohort was 81.6 (± 8.5) years. The cohort consisted of 74.6% women and 89.8% non-Hispanic whites. Most of the patients were from South (37.4%) and Midwest (33.5%) regions and resided in urban areas (70.2%). Frequently reported comorbidities included hypertension (80.4%), cardiac arrhythmias (40.7%), diabetes (37.6%) and fluid and electrolyte disturbances (33.4%). The most frequently prescribed medications were antidepressants (47.4%), beta-blockers (41.5%) and diuretics (36.3%). The mean (± SD) ACB score of the cohort pre-index was 0.7 (± 1.4). This was a pre-index measure of anticholinergic burden. Additional characteristics of the study cohort are summarized in Table 1. The mean duration of nursing home stay was 530.1 (± 268.4) days.

Admission MDS assessments of LTNH residents with OAB revealed that 23.5% had UI frequently (i.e., incontinent seven or more episodes over a 7-day look-back period) and 37.1% had bowel continence constantly. Data on UI were missing for 22.2%. Approximately two-thirds (66.8%) of patients did not require any bladder continence management and 11.1% required some kind of continence management—including an indwelling/external catheter (9.7%), ostomy (1.9%) and intermittent catheterization (0.3%). Approximately 10.1% of the patients required a urinary toileting program (i.e., scheduled training, prompted voiding and bladder training). However, only 1.1% were completely dry (i.e., continent) or showed decreased wetness (i.e., whereby urinary incontinence frequently decreased) following the toileting program. The majority of patients had intact cognition (88.3%) but just over half (51.1%) were dependent on their ADLs (Table 1).

About 4.4% (n = 11,012) of the LTNH residents were newly initiated with antimuscarinics with supply lasting at least 30 days continuously. Oxybutynin (69.8%) was the most commonly prescribed antimuscarinic, followed by tolterodine (13.0%) and solifenacin (12.3%) (Table 2). About 67% (n = 7319) of the study cohort was identified to have discontinued the index antimuscarinic medication during their nursing home stay, and while the median days to discontinuation was 61 (interquartile range [IQR]: 31–121) days, the mean (standard deviation) days to discontinuation was 99.9 (± 105.8) days. Discontinuation rate based on the index medication in the order were darifenacin (78.8%), trospium (78.2%), tolterodine (70.8%), oxybutynin (66.4%), fesoterodine (61.4%) and solifenacin (59.8%) (Table 2).

The Kaplan-Meier curve with discontinuation probabilities of antimuscarinic medication over time is shown in Fig. 2. Discontinuation of antimuscarinics varied solely based on the need characteristics (Table 3). Results from the multivariable Cox PH regression model revealed that LTNH residents who initiated with solifenacin (Hazards Ratio (HR) 0.79, 95% CI 0.73–0.85) and darifenacin (HR 0.82, 95% CI 0.68–0.99) were less likely to discontinue medication, while patients who started with tolterodine (HR 1.11, 95% CI 1.03–1.18), darifenacin (HR 1.33, 95% CI 1.06–1.67) and trospium (HR 1.44, 95% CI 1.24–1.66) were more likely to discontinue medication compared to those who were started with oxybutynin. Patients who had liver disease (HR 0.82, 95% CI 0.71–0.94), antipsychotic use (HR 0.93, 95% CI 0.86–0.99) and participated in a urinary toileting program (HR 0.75, 95% CI 0.58–0.97) were less likely to discontinue antimuscarinics than their counterparts, whereas LTNH residents with severe cognitive impairment (HR 1.25, 95% CI 1.10–1.42), dependent on ADL (HR 1.13, 95% CI 1.04–1.23) and occasionally bowel incontinent (HR 1.09, 95% CI 1.01–1.19) were more likely to discontinue antimuscarinics compared to their counterparts. For every unit increase in the pre-index ACB burden, there was a 4% increased likelihood of discontinuing the antimuscarinic medication.