New use of psychotropic medication after hospitalization among people with dementia

Psychotropic medication is commonly used among people with dementia (PWD), but it shows modest efficacy and it has been associated with severe adverse events. Hospitalizations are an opportunity for medication management as well as treatment recommendations for outpatient physicians. The aim of this study was to asses factors associated with new use of psychotropic medication after hospitalization among PWD.

accumulation of hospital days, 12 and deterioration. 13,14 Consequently, lists of potentially inappropriate medication for older adults, such as the Beers criteria and the Screening Tool of Older Persons' potentially inappropriate prescriptions/screening tool to alert to right treatment (STOPP/START) generally advice against prescription of these drugs for PWD, unless stringently indicated. 15, 16 Nevertheless, psychotropic medications are commonly prescribed among PWD in the community, nursing home, and during hospitalization. [17][18][19][20][21] Hospitalizations are common among PWD and can lead to an increase in prescribed medication, [22][23][24] but hospital stays could also be regarded as an opportunity for an improved medication management. Since hospitalizations can also have an immediate impact on outpatient care and prescribing of medication, it is of utmost importance to understand which factors influence the prescription of psychotropic medications and to identify high-risk groups. This would contribute to developing appropriate measures to reduce the initiation of inappropriate prescribing of psychotropic medications.
In general, studies have shown that psychotropic medication use increases with care dependency, 3 history of psychiatric diseases, 25,26 and behavioral and psychological symptoms. 2,3 Nevertheless, there is a lack of studies assessing risk factors for new use of psychotropic medication among PWD specifically after hospitalization. A recent study among hospitalized PWD focused on anticholinergic agents which include some of the psychotropic medications, and discovered that the presence of comorbid psychiatric conditions played a major role in prescribing. 27 However, this study only focused on a single hospital stay and prevalent use of medication.
Hence the aim of this study was to comprehensively assess factors associated with new use of psychotropic medication after hospitalization across multiple hospital stays among PWD.

| Study Design
A retrospective dynamic cohort study was conducted based on claims data of a German health insurance plan from 2004 to 2015.
The database itself consists of a statutory health insurance sample beginning in 1998 (18.75% random sample of all subjects insured by "Allgemeine Ortskrankenkasse [AOK] Hessen"). 28 Patient informed consent was not required by law as the study was based on pseudonymous data and the utilization of the database for research purposes was approved by the Ministry of Social Affairs of Hesse.
Starting in 2006, cohort entry and cohort exit was possible in every year of the study period. Eligible for cohort entry were all members of the statutory health insurance aged ≥55 years during the study period with information on age and sex, and a continuous insurance period of at least 2 years prior to cohort entry. The baseline period started 2 years prior to cohort entry to assess pre-existing comorbidities and medication use. Reasons for cohort exit were limited to death, end of insurance period, or end of study period. End of insurance period which was very rare (1.9% of patients during the period of investigation) relates to participants switching to a different health insurance plan.

| Ascertainment of dementia, delirium and neuropsychiatric symptoms
A detailed description of the algorithm used to identify PWD has been previously provided elsewhere. 29 In brief, dementia was defined as two confirmed outpatient diagnoses in the same or consecutive 3-month periods or one inpatient diagnosis. The application of at least one appropriate diagnostic measurement was required as well and included testing of cerebrospinal fluid, computed tomography scan and magnetic resonance imaging of the head, or positronemission tomography of the brain. To determine the type of dementia all dementia diagnoses in the first year after cohort entry were considered. Since no specific ICD-10 code for mixed dementia exists, it was defined as the presence of Alzheimer's disease (AD) and vascular dementia types according to ICD-10 coding.
Repeat switching of specific dementia types or switching from a specific to an unspecific dementia type lead to an assignment to other/ unknown dementia.
The following coded diagnoses of delirium and neuropsychiatric
• Delirium and neuropsychiatric symptoms were associated with significantly increased odds of new psychotropic medication use.
• Hospital stays due to dementia and the need of care were predictors for new use of psychotropic medication.

| Dispensation of psychotropic medications
Outpatient dispensation of psychotropic medication was assessed using the anatomical therapeutic chemical classification system. Psychotropic medication was differentiated according to the following medication classes: antidepressants (N06A), antipsychotics (N05A), anxiolytics or hypnotics/sedatives (N05B, N05C), and Alzheimer's medication (N06DA02-04, N06DX01). The exact date of the dispensation from the pharmacy was used to determine new use for each class of medication within 30 days after discharge from the hospital.
New use for each psychotropic medication class was defined as the first dispensation of the medication. Hence, participants were eligible for new use when there was no dispensation during the baseline period or prior to the respective hospitalization. There was no information available on medications given to PWD by hospital physicians during the hospital stay. The duration of psychotropic medication prescribing was assessed by adding the package sizes in form of the DDD (eg, 100 DDDs last theoretically for 100 days) to the date of the initial dispensation from the pharmacy. New users were then grouped into short-term (< 6 weeks) and long-term (≥6 weeks) users.

| Statistical analysis
The associations between potential determinants with new use of each of the medication classes after hospitalization were analyzed using logistic regression models to compute odds ratios (OR) with corresponding 95% confidence interval (CI). The observation time began at the day of discharge from the hospital and ended 30 days after discharge, or on the day of the first dispensation of the analyzed medication, a new hospitalization or cohort exit (death or end of insurance period), whichever came first. First, the estimates were derived separately for up to four rounds of hospitalizations per participant after dementia diagnosis to explore possible trends, and second the association of possible predictors with new use of psychotropic medication after any of the four hospitalizations was estimated jointly to increase the power of the analyses. To account for clustering effects in the joint analyses, logistic regression models with a sandwich estimate using the covariance matrix were computed. 30 In case the first dementia diagnosis was an inpatient diagnosis, this was considered the first hospitalization.
Potential determinants considered included age, sex, comorbidities present at baseline, delirium and NPS during hospitalization, main discharge diagnosis (reflects the main reason for the hospital stay as determined by the hospital physicians), year of hospitalization, type of dementia, time since cohort entry, and need of care. The need of care is based on care services reimbursed by the statutory care insurance, is established by a qualified nurse or physician by assessing the ability to perform activities of daily living, and reflects the dependency on care. At the time of the study the need of care was classified into three levels in the German health care system. All statistical analyses were done using SAS 9.4 (SAS Institute Inc., Cary, N.C., USA).

| Sample characteristics
Characteristics of PWD identified through the algorithm are shown in

| New use of psychotropic medication
The number of new users, short-and long-term users for each medication class according to the different hospitalizations are presented in

| Predictors of new use of psychotropic medication
Predictors of new use of the various classes of psychotropic medication are shown in Tables 3-5 and S1. No associations were seen between age and sex and the outcomes in any of the models, and therefore results for these sociodemographic variables are not displayed.
The factors associated with new use of antidepressants are displayed in Table 3. PWD with anxiety during any hospitalization had a significantly higher odds for new use of antidepressants antipsychotics. The current evidence regarding the efficacy of antipsychotics to treat or prevent delirium among hospitalized older adults does not support the use of antipsychotics. 33,34 Antipsychotics should only be used in PWD with agitation or psychotic symptoms with the potential of causing harm to the person and/or others and if they did not respond to non-pharmacological treatments. 35 According to a recent consensus recommendation, citalopram and analgesia should be prioritized ahead of antipsychotics for agitation, if pharmacologic strategies were needed. In contrast, for psychosis, pharmacologic options should be prioritized following the assessment of underlying causes. 36 Interestingly, people with delusion, hallucination or an acute psychotic state had a significantly higher odds of new Alzheimer's medication use. The most likely explanation is that Alzheimer's medication might reduce NPS, including delusions, which is why treatment might have been initiated. [37][38][39] The higher odds of receiving psychotropic medication among PWD with delirium and NPS in general highlights the need to establish early detection and (non-pharmacological) treatment of those symptoms, especially during hospitalization.
People with Lewy body dementia had a significantly higher odds of new use of antipsychotics compared to people with AD. Studies have shown a significantly greater total score and sub-scores for hallucinations on the neuropsychiatric inventory among people with Lewy body dementia as well as a higher risk of delirium compared to people with AD. [40][41][42] Although we adjusted the analysis for delirium and NPS there might be cases with delirium or NPS that were not coded which led to new use of antipsychotics. Another explanation might be that delirium and Lewy body dementia share a number of clinical features that make a differentiation between those two challenging. 43 Hence, people with Lewy-Body-Dementia might appear to have delirium and thus get prescribed antipsychotics. Nevertheless, the use of antipsychotics in Lewy body dementia is not advised. 44 T A B L E 5 Factors associated with new use of anxiolytics or hypnotics/sedatives after hospitalization displayed as odds ratios (OR)

| Strengths and weaknesses
Our study has several limitations. We analyzed data from a single German region, which limits the generalizability of our results. The analysis relied on ICD-10-GM codes for billing purposes in the health insurance system, which should ensure a high degree of validity but was not externally validated. In order to ensure a valid diagnosis of dementia, as well as the underlying etiology, diagnostic measurements as a part of our case definition and all diagnoses made in the first year after cohort entry were used. Nevertheless, the distribution of dementia subtypes seems unusual, which might impede validity of the associations between different subtypes of dementia and new use of psychotropic medication. Also, delirium and NPS are underreported when using the ICD-10 system. 45 This might have led to underestimating effects in regards to delirium and NPS and new use of psychotropic medication. Although the aim of this study was to assess new use after hospitalization, it is to be noted that we did not have access to the use of medication during the hospital stay. Hence, new use might actually be a continuation of treatment beginning during the hospitalization. Furthermore, information on possible confounders including clinical parameters and family support were not available.
Among the strengths of our study is the large sample size, the novel assessment of new use of psychotropic medication classes, the differentiation between multiple hospitalizations, and the inclusion of people independently of their living situation, health status or nationality. Finally, based on the nature of the data, recall or interviewer bias was avoided.
In conclusion, we identified delirium and NPS during hospitalization, main discharge diagnoses of dementia, and the need of care as main predictors for new use of psychotropic medication after hospitalization among PWD. This highlights the need for future research to develop and implement appropriate interventions to detect delirium and NPS in early stages as well as to establish medication management across the different levels of need of care among PWD.