Abstract
Background:
Literature supports an increasing number of older patients living with neurocognitive disorders alongside with their annual worldwide costs. Therapeutic management of behavioral and psychological symptoms includes the use of anticholinergic and sedative drugs for which significant exposure is negatively associated with clinical outcomes.
Objective:
The aim of this study was to assess the healthcare costs differences related to an increase in the exposure to anticholinergic and sedative drugs in older patients with neurocognitive disorder.
Methods:
A longitudinal study was conducted during 3 years on 1,604 participants of the MEMORA cohort linked with both regional public health insurance and hospital discharge databases between 2012 and 2017. Direct medical and non-medical costs were included. Exposure to anticholinergic and sedative drugs was measured by the drug burden index (DBI).
Results:
Costs difference associated with a DBI≥0.5 were + 338€ (p < 0.001). After adjustment on comorbidities, NCD stage, cognitive impairment, functional limitation, polypharmacy, and sociodemographic characteristics, a DBI≥0.5 was found to be an independent predictor of an increase of total healthcare costs by 22%(p < 0.001).
Conclusion:
Anticholinergic and sedative drugs have a substantial economic burden among older patients with neurocognitive disorder. More studies are required to assess the clinical and economic impact of an efficient strategy based on the reduction of the exposure to anticholinergic and sedative drugs and the promotion of non-pharmacological interventions.
Keywords
INTRODUCTION
Decline in many physiological systems develops as a consequence of aging. Older people have a substantially increased risk of falls, disability, long-term care, and death [1, 2]. While old age does not imply chronic diseases, older people experience high comorbidity prevalence associated with polypharmacy and therefore an increased risk of adverse drug events (ADEs) [3, 4]. Many of these ADEs are caused by potentially inappropriate medications (PIM) and are preventable [5–7]. ADEs are also associated with a high economic burden on patients, caregivers, and healthcare systems [8]. Older patients with cognitive disorders are more likely to develop ADEs and contribute to higher use of healthcare resources and greater costs [9]. Additionally, studies have demonstrated that patients presenting preventable ADEs make greater use of healthcare services and incur higher related costs [10, 11].
Anticholinergic and sedative drugs (ASD) are common PIM prescribed to relieve behavioral and psychological symptoms in people living with neurocognitive disorder (NCD) [12]. Nearly 40%of patients attending memory clinics used one or more PIM and these PIM were mostly ASD [13]. Moreover, ASD are also frequently involved in increasing ADEs and mortality in this specific population [13–15]. Their exposure can be assessed using the Drug Burden Index (DBI) [16]. The DBI is a tool developed to measure individuals’ exposure to both ASD considering dose-response and cumulative effect [17]. A higher DBI score, i.e., a greater exposure to ASD, has been shown to negatively affect the cognitive performance and the functional status of older adults and to increase the risk of falls [14, 17–25]. In older people, it is also linked to frailty and all-cause mortality [26–28]. Higher DBI scores have been correlated with greater numbers of general practitioner visits and longer and more frequent hospital stays [26, 29]. But the economic consequences of a pharmacological burden, like the DBI score, on healthcare costs have not been yet reported and quantified.
The number of people living with major NCD is estimated to triple by 2050 and the current annual worldwide cost of major NCD, estimated at US $1trillion, is expected to double by 2030 [30]. Therefore, understanding determinants of healthcare cost in people with a NCD is the first step to design future cost-effective interventions.
This study aimed to assess the difference in healthcare costs associated to a higher exposure to ASD measured by the DBI in patients living with NCD.
METHODS
MEMORA cohort and source of data
Data used in the current study were extracted from the MEMORA real-life cohort [31]. This cohort aims to assess the determinants of functional decline and cognitive impairment over time. The MEMORA study relies on data from medical records on outpatients from the Clinical and Research Memory Centre. Data are also linked with the regional Primary Health Fund (public health insurance) and with the hospital data of the hospital discharge database PMSI (Programme de medicalisation des systèmes d’information).
Inclusion criteria in the MEMORA cohort are outpatients attending a medical evaluation with a neurologist, a geriatrician, or a psychiatrist at the memory clinic, whatever the cognitive stage. Exclusion criteria are patients living in a nursing home or being under guardianship and patients with significant hearing or visual impairment preventing cognitive evaluation.
The current study included outpatients followed at least 3 consecutive years between 2012 and 2017, and for whom public health insurance data were available over the 3-year period. Data from the public health insurance are aggregated every 6 months providing 6 repeated paired estimations of the DBI and associated healthcare costs over the 3-year period.
Written information regarding the collection of clinical data was provided to the patient and caregivers. Authorization for handling personal data has been granted by the national data protection authority, August 6, 2010.
Exposure to anticholinergic and sedative drugs and polypharmacy
Exposure to ASD was quantified using the DBI which was developed in older people based on pharmacological principles [23]. We did not use prescriptions to estimate daily doses required for DBI calculation. Indeed, we considered that the number of medication units delivered in community pharmacies to the patient represented a better estimate of the real patient drugs consumption. Thus, we calculated the consumption for each reimbursed medication over 6 months and estimated a mean daily dose. Reimbursed medications that are delivered in community pharmacies are recorded in the public health insurance database through the dispensing software. Medications with anticholinergic and/or sedative effects were extracted from the literature [32–37].
Polypharmacy was defined regarding the mean number of medications (not only ASD) delivered over 6 months and stratified as patients with 4 or less delivered medications, patients with 5 to 9 delivered medications and those with 10 or more delivered medications.
Cognitive and functional impairment
Overall baseline cognitive impairment was ass-essed using the standardized Mini-Mental State Examination (MMSE) and range from 0 (severe cognitive impairment) to 30 (no impairment) [38]. Diagnosis stage was determined by a neurologist, a geriatrician, or a psychiatrist during outpatient visits and was coded in 4 categories according to Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) nomenclature: subjective cognitive decline, mild NCD, and major NCD.
Autonomy was assessed at baseline using the score of the instrumental activities of daily living scale (IADL) and range from 0 (severe dependency) to 8 (autonomy) [39].
Sociodemographic characteristics and comorbidities
The sociodemographic characteristics considered in the study were age, gender, educational level (nil, primary, secondary, tertiary). The presence of com-orbidities was retrieved from the national hospital discharge database (diagnosis) and from the public health insurance database (medications). Comorbidities considered in the study were selected according to their economic burden [40]. Patients with stroke, heart disease, cancer, chronic kidney disease, cirrhosis, alcohol use disorders, diabetes, respiratory illness, and depression were identified using ICM-10 codes [40]. Patients with diabetes, respiratory illness and depression were also identified according to delivered medications and their Anatomical Therapeutic Chemical subgroups: respectively “A10–Drugs used in diabetes”, “R03–Drugs for obstructive airway diseases” and “N06A–Antidepressants”.
Healthcare costs
Information on healthcare costs was retrieved from the public health insurance database. Direct costs were collected for the following areas: inpatient care from public and private hospitals, outpatient care (including diagnostic, i.e., imaging, and laboratory tests), consultations with general practitioners, specialists, and other office-based practitioners (nurses, physiotherapists, speech pathologists, and orthoptists), medication costs, and transport services. Medication costs consist of reimbursed drugs based on medications delivered to patients by community pharmacists.
A health insurance perspective included direct medical and nonmedical costs (transportation costs) was adopted. Cumulative costs over 6 months were determined by multiplying the number of units used for each relevant resource with the corresponding unit cost. All costs were inflated to a 2017 level using Consumer Price Index according to the OECD recommendations (OECD. Stat web site–OECD-CPI, 2017). All costs are presented in mean monthly costs and in euros (€).
Statistical analyses
Patients’ characteristics are described using means and standard deviation for continuous variables and numbers and percentages for categorical variables. Results are shown depending on the type of DBI: DBI Total (all sedative and anticholinergic drugs), DBI AC (only anticholinergic drugs), and DBI SED (only sedative drugs). As a continuous variable, DBI was categorized according to the target attribute of total costs using a regression tree (supervised discretization): patients with DBI < 0.5 (less exposed) and DBI≥0.5 (more exposed). The FCS MI method (Multiple Imputation by Fully Conditional Specification) was applied to replace missing values for the following variables (%of missing values): educational level (13%), MMSE score (8%), and IADL score (4%) [41]. Monthly mean costs per patient were presented with estimated 95%confidence intervals. Extreme values of costs were kept in the analyses. In the costs differences analysis, we considered all 6 months paired estimations regardless of patients. Differences in means between patients with DBI < 0.5 and DBI≥0.5 in every type of DBI were tested using independent sample χ2 test (for categorical variables) or t-tests (for continuous variables).
A log link generalized linear model for repeated data (generalized estimating equations methodology) assuming a gamma distribution was implemented to estimate main predictors of total costs. This type of model was chosen to account for variations in anticholinergic and sedative exposure and associated costs and the diagnosis of new comorbidities every 6 months. A modified parks test was conducted to select appropriate distribution of total costs. The predictor of interest was the DBI score. MMSE score, IADL score, gender, age, educational level, disease stage, comorbidities, and polypharmacy were also included in the model. Statistical interactions between the DBI and disease stage, depression and polypharmacy were also tested.∥Sensitivity analyses were performed to assess the impact of DBI calculation on the model: DBI calculation that included furosemide and inhaled anticholinergic drugs (drugs with controversial anticholinergic burden), excluded antidepressants and DBI OMS calculation [42]. We also performed analyses with continuous DBI.∥Statistical analyses were performed with SAS 9.4 software (SAS Institute Inc. Cary, NC) and SPSS V.19.0 (SPSS Inc.). All tests were two tailed and a p-value of less than 0.05 was considered statistically significant.
RESULTS
Population characteristics
A total of 1,604 patients were included in this study. Patients’ characteristics at baseline are summarized in Table 1. Among these patients, 879 had a DBI total < 0.5 and 725 had a DBI total≥0.5, 1,468 patients had a DBI AC < 0.5 and 136 a DBI AC≥0.5, and 918 patients had a DBI SED less than 0.5 and 686 had a DBI SED of 0.5 or more. Mean age was 80 years (SD±6), and no differences were observed between groups of patients with DBI (total, AC and SED) < 0.5 and DBI≥0.5. Women were more numerous in the study (63%) and were more likely to be exposed to sedative drugs. Major NCD represented 34%of patients, Mild NCD 30%, and subjective cognitive disorder 36%. Mean MMSE was 21 (SD±6). Educational level, diagnosis stage, and MMSE score did not differ between groups. IADL scores were higher for patients with DBI total and SED < 0.5. Highly polymedicated patients were more likely to have a DBI≥0.5. Patients with depression were also more likely to a DBI≥0.5 given that antidepressants have commonly anticholinergic or/and sedative properties. Patients with a DBI≥0.5 also were affected by diabetes and respiratory illness.
Characteristics of participants at baseline
DBI, Drug Burden Index; IADL, Instrumental Activities of Daily Living; MMSE, Mini-Mental Status Exam; SD, standard deviation.
Costs differences
Table 2 reports monthly mean costs and costs differences per patients and by type of DBI and by level of exposure. Monthly total costs differences related to a DBI≥0.5 were + 338€ when using the DBI total, + 302€ with the DBI AC, and + 342€ with the DBI SED. A higher exposure to sedatives drugs did not lead to significant costs differences in outpatients. Transportation costs were similar regardless of the exposure to anticholinergic drugs. Inpatient costs were the main driver of costs differences (57%of total additional costs) in every type of DBI. A higher exposure to sedative drugs accounted for more additional total costs than anticholinergic drugs.
Monthly mean costs and cost differences (€, 2017)
DBI, Drug Burden Index.
Results presented in the Supplementary Table 1 show the impact of DBI calculation on total costs differences. Adding furosemide and inhaled anticholinergics raised the monthly total costs difference by 18%(400€). This increase is mainly supported by a greater difference in inpatient care costs. Using the DBI OMS calculation raised the monthly total costs difference by 14%(385€).
Multivariate analysis
The results of the adjusted generalized linear model are presented in Table 3. Model 1 focused on the DBI total while Model 2 and 3 respectively focused on sedatives drugs and anticholinergic drugs. Model 1 showed the DBI total as an independent predictor of total healthcare costs (OR 1.22, 95%CI 1.10–1.37) after adjustments on age, gender, educational level, MMSE, disease stage, IADL, polypharmacy, and comorbidities. Results of model 2 and 3 indicated that the exposure to sedative drugs were a predictor of higher total healthcare costs (OR 1.26, 95%CI 1.13–1.41) contrary to the exposure to anticholinergic drugs of which no association was found. Running models with DBI considered as a continuous variable did not significantly change the results (Supplementary Table 2). Also, in Supplementary Table 2, considering furosemide and inhaled anticholinergic in the DBI calculation significantly increased the impact of anticholinergic exposure on total healthcare costs (OR 1.22, 95%CI 1.07–1.39).
Predictors for total direct healthcare costs over the 3-year follow-up
DBI, Drug Burden Index; IADL, Instrumental Activities of Daily Living; MMSE, Mini-Mental Status Exam, aper 1-year increase in age, bper 1-unit increase in MMSE score, cper 1-unit increase in IADL score.
Other independent predictors of total healthcare costs in the 3 models were being a woman, the IADL limitation score, having 5 and more medications and suffering from chronic kidney disease or heart disease. No tested statistical interactions were found.
DISCUSSION
In this longitudinal study of outpatients of a memory center, the main findings show that the monthly mean total costs difference associated with a higher exposure to ASD (DBI ≥0.5) were + 338€. Inpatient costs represented 57%of total costs difference. After adjustment, a higher exposure to ASD (DBI ≥0.5) was found to be an independent predictor of an increase of total healthcare costs by 22%.
This study is the first to highlight extra costs related to an exposure to ASD regardless of comorbidities and NCD stage. Patients with comorbid conditions are more likely to experience higher use of healthcare services and higher healthcare costs [40]. Models in this study were adjusted on comorbidities that have a high economic burden. Furthermore, the apparition of these comorbidities over the 3 years of the study was accounted in these longitudinal models. However, a selection bias remained in patients included in our study; Patients consulting in memory centers are outpatients and are more likely to have less severe comorbidities at this stage than inpatients or patients not consulting. Yet MacNeil-Vroomen et al find similar association between chronic condition and health-care costs and also a significant impact of chronic kidney disease and heart disease on costs [43].
Another main strength is that we have suggested a discretization of the DBI variable thus defining a 3-year high-cost risk threshold based on the DBI score. This threshold of 0.5 was statistically defined. According to DBI calculation, it also refers to a daily prescription of one anticholinergic or sedative drug at a recommended dosage. A higher DBI refers either to an overuse or an association of several ASD.
Functional limitations was also associated to higher total costs which is consistent with previous findings also conducted on patients with NCD [44]. In the same study, results on cognitive impairment differed from our study. This can be explained by our methodology focusing on comorbidities with a high economic burden: patients with poor MMSE score had significantly less heart disease which is also associated with higher costs in our study. Mild NCD patients also have a significant impact on total costs with a higher proportion of cancer and heart disease in our population whereas direct medical costs did not differ significantly in the literature between subjective cognitive disorders and mild NCD. However major NCD was linked to higher inpatient costs [45].
In our population most of patients have an altered cognitive function. These patients are more likely to have experienced early signs of chronic diseases leading to a NCD such as mood changes and depression which can be treated by anticholinergic or sedative drugs. Moreover, cognitive decline can be a consequence of ADEs or related to prodromes or clearly identified neurodegenerative disease such as Alzhei-mer’s disease [46–48]. However, results of the current study showed that MMSE and diagnosis stage did not differ between DBI < 0.5 and DBI ≥0.5 groups. Depression was significantly higher in patients with higher DBI but were not associated with higher costs. Looking at functional limitations, results show that IADL were significantly lower in patient more exposed to sedative drugs. Both IADL and higher exposure to sedative drugs were associated to increasing costs.
Although there is no published work reporting differences in healthcare costs linked to the use of ASD against which to compare our results, several studies did measure the additional costs of inappropriate prescribing, which include most of the ASD. One study, based on Finnish Prescription Register data, found more PIM in high-cost polypharmacy patients than all drug users (28.0%versus 19.9%, p < 0.001) [49]. In this study, patients taking anticholinergic drugs were more likely to have high-cost polypharmacy (26.7%versus 9.6%, p < 0.001). In a U.S. hospital study, individuals with three or more PIMs had significantly higher hospital costs (€ 2539.53) than those with one PIM (€ –113.01) [50]. The same trend towards higher costs for hospitalizations with PIM than without PIM was found in a Swedish study [51]. The same trend was observed in our study; however, no adjustment on detailed comorbidities was conducted in these three studies. Inpatient costs were found to be cost drivers in the literature which is in line with our results [52]. Regarding healthcare utilization, Nishtala et al. reported that exposure to ASD was independently associated with fall-related hospitalizations, and the number of GP visits [26]. Gnjidic et al. observed a relationship between cumulative anticholinergic and sedative drug use and all type hospitalizations [27]. These conclusions are in line with significant costs differences for inpatients and GP visits.
Several limitations in this study should be considered. First, a measurement bias may be found in the way of calculating the DBI score in the national insurance database: daily doses estimated according to delivered drugs may be subject to inaccuracies as information in the database was reported as 6-month aggregate data. For drugs that were not regularly delivered, exposure time might be overestimated, and daily dose might be underestimated. Then, data on the costs of informal caregivers were not available and thus, were not accounted in the models [53]. Lastly, our population came from patients referred to memory center based on a suspicion of NCD. Hence, the proportion of patients with major NCD in our population is higher than in the general population (34% versus between 7.4%and 12.9%) [54]. Also, the use of drugs and associated costs might be overestimated in our population compared to general population. Therefore, these results should be carefully generalized.
When considering delivered drugs and the chronic exposure to ASD in our population, we found that a DBI score higher than 0.5 was associated to a positive costs difference. We also found that patients with a DBI higher than 0.5 had significantly more chronic kidney disease and depression alongside with diabetes, heart disease and respiratory illness. These can explain the significantly high rate of patients with more than 10 medications delivered in the DBI ≥0.5 group. Moreover, diagnosis stages were equally distributed between patients with a DBI ≥0.5 and patients with a DBI < 0.5. Suggestion might be that there is a potential overuse of ASD that may not be related to the progression of the NCD. Considering the high rate of depression and the equal distribution of MMSE, these drugs may be those used to treat depression. But removing antidepressant from the DBI calculation did not change the association significantly which is in line with the statistical adjustment on depression. Lastly, further data are needed to investigate behavioral disorders and related therapeutic management in our population. These findings might suggest that a proper medication order review made by trained healthcare professionals might be an interesting economic strategy to early reduce the potential inappropriate exposure to ASD [55, 56]. Indeed the number of aging patients with cognitive impairment is expected to increase considerably in the coming decades alongside with the cost of their management and therefore the economic burden of ASD.
In conclusion, the exposure to ASD in aging patients with a cognitive impairment has a substantial economic burden. Interventional and cost-utility studies are required to assess the clinical and economic impact of an efficient strategy based on the reduction of the exposure to ASD and the promotion of non-pharmacological interventions in older patients.
Footnotes
ACKNOWLEDGMENTS
We thank Pascale Gauthier-Robino and Laurent Colas who provided us with healthcare costs (Pri-mary Health Insurance Fund of the Rhône - CPAM Rhône, Lyon, France), Michel Kossovsky who worked on preliminary data and database quality control (Rehabilitation and Geriatrics, Geneva University Hospital, Geneva, Switzerland), Aline Dorey and Claire Moutet who worked on patients selections (Memory Research Centre of Lyon (CMRR); Geriatrics Unit, Charpennes Hospital, University Hospital of Lyon, Villeurbanne, France). We also thank Xavier Dode who provided us with Theriaque database and Erin Salmon and Philip Robinson who helped us with English proofreading. The MEMORA study received a grant from the MSD Avenir fund and Biogen US fund.
