A Group-based Trajectory Analysis of Longitudinal Psychotropic Agent Use and Adverse Outcomes Among Older People

JAMDA xxx (2019) 1e8

JAMDA journal homepage: www.jamda.com

Original Study

A Group-based Trajectory Analysis of Longitudinal Psychotropic Agent Use and Adverse Outcomes Among Older People Shih-Tsung Huang PharmD a, Yu-Wen Wen PhD b, Susan Shur-Fen Gau PhD c, Liang-Kung Chen PhD d, e, f, Fei-Yuan Hsiao PhD a, g, h, * a

Graduate Institute of Clinical Pharmacy, College of Medicine, National Taiwan University, Taipei, Taiwan Clinical Informatics and Medical Statistics Research Center, Chang Gung University, Taoyuan, Taiwan c Department of Psychiatry, National Taiwan University Hospital, Taipei, Taiwan d Institute of Public Health, School of Medicine, National Yang-Ming University, Taipei, Taiwan e Aging and Health Research Center, National Yang-Ming University, Taipei, Taiwan f Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, Taiwan g School of Pharmacy, College of Medicine, National Taiwan University, Taipei, Taiwan h Department of Pharmacy, National Taiwan University Hospital, Taipei, Taiwan b

a b s t r a c t Keywords: Psychotropic agents group-based trajectory model older people mortality unplanned hospitalization

Objective: Existing evidence on long-term psychotropic polypharmacy in older people is scarce despite serious safety concerns regarding this issue. This study aims to identify distinct trajectories of longitudinal psychotropic agent use and to examine the association with the risk of mortality and unplanned hospitalization according to distinct trajectories. Design: A retrospective longitudinal cohort study based on data from Taiwan’s National Health Insurance Research Database (NHIRD). Setting and participants: A population-based study including 39,803 eligible participants aged 65 years or older who were newly prescribed oral psychotropic agents in 2004. Measures: We assessed 5 years of monthly consumption of psychotropic agents among eligible participants and used group-based trajectory modeling to identify distinct groups of longitudinal psychotropic use. Cox proportional hazards models were used to examine the association between distinct trajectories of longitudinal psychotropic agent use and subsequent 1- and 3-year unplanned hospitalization and allcause mortality. Results: Among 39,803 eligible participants, we identified 5 trajectories of longitudinal psychotropic agent use over a 5-year follow-up period: sustained intense (7.1%), moderate increasing (12.1%), decreasing (13.4%), slowly increasing (15.7%), and infrequent users (51.6%). Moderate increasing and sustained intense use were associated with significantly higher risks of unplanned hospitalization [moderate increasing use: hazard ratio [HR] 1.30, 95% confidence interval (CI) 1.18-1.44; sustained intense use: HR 1.40, 95% CI 1.24-1.58] and all-cause mortality (moderate increasing use: HR 1.25, 95% CI 1.09-1.44; sustained intense use: HR 1.32, 95% CI 1.12-1.56) than infrequent use. Similar results were observed at 3 years of follow-up. Conclusions/Implications: Older people with moderately increasing and sustained intense use of psychotropic agents over time had higher risks of unplanned hospitalization and mortality. Tracking older people’s use of psychotropic agents over time may help to identify individuals at greatest risk of unplanned hospitalization and mortality. Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.

F-Y.H. received research assistantships from a research project (MOST 1042410-H-002-225-MY3) sponsored by the Ministry of Science and Technology,Taiwan. The authors declare no conflicts of interest.

https://doi.org/10.1016/j.jamda.2019.05.012 1525-8610/Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.

* Address correspondence to Fei-Yuan Hsiao, PhD, Associate professor Graduate Institute of Clinical Pharmacy College of Medicine, National Taiwan University Room 220, 33, Linsen S. Road, Taipei, Taiwan 10050. E-mail address: [email protected] (F.-Y. Hsiao).

2

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

Psychotropic agents, including antipsychotics, antidepressants, anxiolytics, and hypnotics (benzodiazepines and z-hypnotics), are widely used to treat short-term symptoms of depression, anxiety, and insomnia in older people.1e5 Safety regarding the use of these drugs in older people, however, has been called into question.6,7 Notably, inappropriate long-term use of these agents has been reported to be associated with harmful adverse effects, such as impaired cognitive function,8,9 falls, hospitalization, and death.10,11 Because of higher sensitivity and prolonged central nervous system effects related to aging pharmacokinetics and pharmacodynamics,7 these adverse effects may be more predominant in older people. Another safety concern regarding the use of psychotropic agents in older people is the epidemic of psychotropic polypharmacy,12e14 which refers to the increasing number of older people using multiple psychotropic agents concurrently. Psychotropic polypharmacy may be associated with more adverse outcomes because of the additive effects of each psychotropic agent. However, real-world data regarding this important issue are scarce. Most previous studies examining the association between psychotropic agent use and clinical adverse outcomes have been limited to a single category of psychotropic agents,11 1-time measurement of drug exposure, or the assumption that patients are continuous users based on several crosssectional surveys during the long-term follow-up period.11,15 These approaches do not reflect the intra-individual variability of psychotropic agent use among older people or the longitudinal course of psychotropic agent use because older people may often discontinue or switch their psychotropic agents in real-world situations.11 Furthermore, it remains unclear whether different use patterns exist for psychotropic agents over time among older people and whether distinct use patterns confer different risks for clinical adverse outcomes, including death and hospitalization, in older people. To fill this knowledge gap, the aim of this study was to identify distinct trajectories of longitudinal psychotropic agent use and to examine the association between these distinct trajectories and clinical adverse outcomes, including the risks of all-cause mortality and unplanned hospitalization. Methods Data Source We performed a population-based retrospective longitudinal cohort study based on data from Taiwan’s National Health Insurance Research Database (NHIRD). The NHIRD is a nationwide claims-based database comprising anonymous eligibility and enrollment information as well as claims for outpatient visits, admissions, procedures, and medication prescriptions for more than 99% of the entire population (23 million) of Taiwan.8 The “outpatient visits” in the NHIRD include general practitioner and primary care visits as well as long-term residential care. We used a subset of the NHIRD that contains claims data for 20% of randomly selected beneficiaries aged 65 years and older in 2004, and created a 10-year (2003-2012) panel of claims for analysis. The identification numbers for all entries in the NHIRD were encrypted to protect the privacy of individual patients. The study protocol was approved by the Institutional Review Board of the National Taiwan University Hospital (No. 201403069W). Study Population From the NHIRD, we identified individuals aged 65 years and older who initiated oral psychotropic agents in the outpatient setting in 2004 as our study cohort. The index date was defined as the date of the first prescription for oral psychotropic agents. Individuals who had ever received oral psychotropic agents in an outpatient or inpatient setting within 1 year prior to the index date were excluded.

For the present study, we determined whether the trajectories of long-term psychotropic agent use over a 5-year (2005-2010) period were associated with subsequent adverse outcomes (unplanned hospitalization and mortality) through year 8 (2011-2013). The study design is shown in Supplementary Figure 1. Because we aimed to characterize longitudinal use patterns of psychotropic agent use, we excluded individuals who died during the trajectory period (year 1 through year 5) to guarantee that we fully assessed the participants’ prescription records for psychotropic agents. Measures Long-term psychotropic agent use We assessed 5 years of monthly consumption of psychotropic agents (60 measurements) during the trajectory period among our study participants. The psychotropic agents examined in this study included antipsychotics [Anatomical Therapeutic Chemical (ATC) code: N05A-], anxiolytics (N05B-), hypnotics (benzodiazepine derivatives (N05CD-), and non-benzodiazepine hypnotics (z-drugs; N05CF-), and antidepressants (N06A-) (Supplementary Table 1). To assess the dose effect of long-term psychotropic agent use, we calculated the monthly consumption of psychotropic agents during the 5-year trajectory period as the defined daily dose (DDD). The DDD system was developed by the World Health Organization (WHO) Collaborating Center for Drug Statistics Methodology to establish a standardized unit for quantifying drug utilization and cumulative doses across different drug categories and strengths. The definition of DDD is the assumed average maintenance dose per day for a drug used for its main indication in 70-kg adults.16 We adopted this classification to determine cumulative exposure to all psychotropic agents (by DDDs) in the 60 measurements of the data collection period for our study subjects. Although older people are not generally the average 70-kg adult (as in the assumption of the DDD system), this did not affect our study (because we sum the cumulative DDDs of all psychotropic agents). The cumulative DDD of psychotropic agents during each month of the trajectory period (60 months) for each participant was used in the group-based trajectory model (GBTM) to identify distinct groups of longitudinal psychotropic agent use. Clinical adverse outcomes The clinical adverse outcomes of interest were unplanned hospitalization and all-cause mortality. Unplanned hospitalization was defined as admission via the emergency department. The reason we focused on unplanned hospitalization rather than all-cause hospitalization was that most psychotropic agenteassociated hospitalizations, including falls, cardiovascular events, cerebrovascular events, and acute infection, are unplanned. All deaths occurring during the 3-year outcome measurement period were also identified. Other Variables The age and sex of participants were documented at the index date. Information regarding the first prescription of oral psychotropic agents, including primary and all diagnoses, drug category, and number of days supplied, as well as the type of specialist who issued the prescription (psychiatry or nonpsychiatry), was also collected. We also retrieved the participants’ medical history within 1 year before the index date and 1 year before the outcome assessment period (years 6-8) for general indicators of comorbidity, psychiatric comorbidities, and other comorbidities. General indicators of comorbidity, including the number of hospitalizations and the Charlson Comorbidity Index (CCI),17 were assessed. The CCI is the most widely used comorbidity index to predict mortality and contains 19 conditions, including diabetes with diabetic complications, congestive heart failure, peripheral vascular disease, chronic pulmonary disease, mild

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

Psychotropic agents use (monthly cumulative DDD)

30

3

Group 1: infrequent users

Group 2: slowly increasing users

Group 3: decreasing users

Group 4: moderate increasing users

Group 5: sustained intense users

25

20

15

10

5

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Month Fig. 1. Five-year trajectories of longitudinal psychotropic agent use among older people.

and severe liver disease, hemiplegia, renal disease, leukemia, lymphoma, metastatic tumor, and acquired immunodeficiency syndrome (AIDS), each of which was weighted according to their potential influence on mortality. Mental illness was defined as depression, psychosis, schizophrenia, delirium, and anxiety. Other comorbidities included insomnia, dementia, hypertension, osteoporosis, osteoarthritis, and Parkinson’s disease. Statistical Analysis The 5 trajectory groups were identified according to group-based trajectory analysis (GBTM),18 which is a cluster analysis method. Briefly, the parameters we used to “distinguish” these groups were based on the 5 years of monthly consumption of psychotropic agents (60 measurements) among our study participants. Some participants may have gradually increased their use of psychotropic agents, and some may have maintained their use in the 5 years of the observation period. By adopting GBTM, we grouped participants in a statistically meaningful way based on the 60 time-varying measurements of psychotropic agent use, which enabled clinical interpretation.19,20 All other statistical analyses, including the process of model selection, are detailed below. To select the best-fitting model, we evaluated the appropriate number of trajectories and trajectory shapes according to the recommended procedures.19,20 We first decided on an appropriate analysis mode, which was related to the data distribution for the variable of interest, before fitting a trajectory model. Because cumulative DDDs per month are count data, we used the zero-inflated Poisson (ZIP) mode. For trajectory model building, we determined the number of groups first and then determined what shapes best fit the data by changing the polynomial order and zero-inflated polynomial order, including linear, quadratic, and cubic terms, for each group (terms were considered significant at the confidence level of alpha ¼ 0.05 for each group). To select the best model with good intergroup heterogeneity, we used the Bayesian information criterion (BIC) value as an index of model selection to compare different models with a varying number of trajectory groups and shapes. Each trajectory group was required to include at least 5% of participants, and the model with the highest BIC values was chosen.

Following the procedures outlined above, we found that the model with 5 trajectories (both polynomial order and zero-inflated polynomial order were cubic for all trajectories) was superior to models with fewer trajectories (BIC values: 5-trajectory model ¼ 1,777,275; 4-trajectory model ¼ 1,806,502; 3-trajectory model ¼ 1,849,027). Although the 6-trajectory model was associated with better model fit (BIC value ¼ 1,760,193), this model yielded a trajectory with less than 5% of the study population. Therefore, we selected the 5-trajectory model as the final model for use in subsequent analyses. After the model-building process, each participant received posterior probabilities of membership in multiple trajectories and was assigned to the trajectory with the highest probability of membership. We used analyses of variance (ANOVAs) for continuous variables and c2/Fisher exact test for categorical variables. We further used multivariate Cox proportional hazards models to examine the association between distinct trajectories of longitudinal psychotropic agent use and clinical adverse outcomes [presented as the hazard ratio (HR) with a 95% confidence interval (CI)]. All models were adjusted for demographics and comorbidities within 1 year before the outcome period. Considering that older people have a higher mortality rate than the general population, death before hospitalization is a competing risk and thus may have biased the results. We therefore also conducted a subdistribution proportional hazards model, a competing risk analysis, as a sensitivity analysis.21,22

Results We identified 50,346 people aged 65 years and older who initiated an oral psychotropic agent in the outpatient setting in 2004. We excluded 10,543 individuals who died during the trajectory period (year 1 though year 5) to guarantee that we fully assessed their prescription records, resulting in 39,803 people. As summarized in Supplementary Table 2, age, sex, category, division, diagnosis and duration of first prescription of a psychotropic agent, number of hospitalizations within 1 year before the index date, CCI, psychiatric comorbidity, and other comorbidity differed between those who died during the 5-year data collection period and our study participants (P < .01).

4

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

Table 1 Participant Baseline Characteristics by Trajectory Group of Longitudinal Psychotropic Agent Use Variable name

Infrequent Users, n (%*) (n ¼ 20,543)

Slowly Increasing Users, n (%*) (n ¼ 6237)

Decreasing Users, n (%*) (n ¼ 5353)

Age, y Mean  SD 72.2  5.7 72.3  5.7 72.6  5.8 65-74 14,256 (69.4) 4258 (68.3) 3533 (66.0) 75-84 5584 (27.2) 1781 (28.6) 1624 (30.3) 85 703 (3.4) 198 (3.2) 196 (3.7) Sex Female 10,080 (49.1) 3356 (53.8) 2802 (52.3) Male 10,463 (50.9) 2881 (46.2) 2551 (47.7) First prescription of psychotropic agent Antipsychotics 2574 (12.5) 666 (10.7) 415 (7.8) Anxiolytics 5633 (27.4) 1731 (27.8) 1165 (21.8) Antidepressants 1227 (6.0) 364 (5.8) 418 (7.8) Sedative drug 11,109 (54.1) 3476 (55.7) 3355 (62.7) BZD 10,247 (49.9) 3192 (51.2) 2873 (53.7) Z-drug 862 (4.2) 284 (4.6) 482 (9.0) Division of the first prescription of psychotropic agents in the index date Psychiatry 185 (0.9) 69 (1.1) 122 (2.3) Nonpsychiatry 20,358 (99.1) 6168 (98.9) 5231 (97.7) Diagnosis of the first prescription of psychotropic agents in the index date (primary diagnosis) Mental illness 398 (1.9) 141 (2.3) 219 (4.1) Depression 58 (0.3) 26 (0.4) 49 (0.9) Psychosis 94 (0.5) 39 (0.6) 64 (1.2) Schizophrenia 13 (0.1) 2 (0.0) 3 (0.1) Delirium 5 (0.0) 1 (0.0) 2 (0.0) Bipolar disorder 4 (0.0) 5 (0.1) 0 (0.0) Anxiety 267 (1.3) 89 (1.4) 132 (2.5) Dementia 77 (0.4) 40 (0.6) 65 (1.2) Insomnia 289 (1.4) 91 (1.5) 135 (2.5) Without any of the above diagnoses 19,853 (96.6) 6002 (96.2) 4994 (93.3) Diagnosis of the first prescription of psychotropic agents in the index date (all diagnoses) Mental illness 15,737 (76.6) 4682 (75.1) 3634 (67.9) Depression 140 (0.7) 42 (0.7) 99 (1.9) Psychosis 15,120 (73.6) 4469 (71.7) 3360 (62.8) Schizophrenia 15,072 (73.4) 4451 (71.4) 3327 (62.2) Delirium 15,073 (73.4) 4451 (71.4) 3328 (62.2) Bipolar disorder 15,076 (73.4) 4452 (71.4) 3329 (62.2) Anxiety 15,670 (76.3) 4660 (74.7) 3595 (67.2) Dementia 103 (0.5) 49 (0.8) 80 (1.5) Insomnia 15,727 (76.6) 4698 (75.3) 3616 (67.6) Without any of the above diagnoses 4211 (20.5) 1329 (21.3) 1453 (27.1) Duration of the first prescription of psychotropic agent in the index date, d Mean  SD 7.5  7.8 8.0  8.2 11.5  10.0 1-7 16,472 (80.2) 4870 (78.1) 3293 (61.5) 8-14 1853 (9.0) 582 (9.3) 799 (14.9) 15-21 239 (1.2) 91 (1.5) 120 (2.2) 22-28 1118 (5.4) 403 (6.5) 676 (12.6) >28 861 (4.2) 291 (4.7) 465 (8.7) Baseline comorbidities General indicator of comorbidity No. of hospitalizationsy, 0.2  0.5 0.2  0.5 0.2  0.6 mean  SD Charlson Comorbidity Index, 0.5  0.9 0.6  1.0 0.7  1.1 mean  SD Psychiatric comorbidity Depression 18 (0.1) 3 (0.1) 14 (0.3) Psychosis 94 (0.5) 38 (0.6) 49 (0.9) Schizophrenia 3 (0.0) 0 (0.0) 2 (0.0) Delirium 4 (0.0) 1 (0.0) 4 (0.1) Anxiety 138 (0.7) 50 (0.8) 63 (1.2) Other comorbidity Dementia 94 (0.5) 36 (0.6) 48 (0.9) Insomnia 206 (1.0) 91 (1.5) 77 (1.4) Hypertension 6387 (31.1) 2210 (35.4) 2082 (38.9) Osteoporosis 619 (3.0) 267 (4.3) 252 (4.7) Osteoarthritis 1719 (8.4) 768 (12.3) 710 (13.3) Parkinson’s disease 81 (0.4) 43 (0.7) 59 (1.1)

Moderate Increasing Users, n (%*) (n ¼ 4831)

Sustained Intense Users, n (%*) (n ¼ 2839)

P Value

72.8  5.8 3125 (64.7) 1536 (31.8) 170 (3.5)

73.1  6.0 1761 (62.0) 938 (33.0) 140 (4.9)

<.01

2554 (52.9) 2277 (47.1)

1422 (50.1) 1417 (49.9)

434 881 384 3132 2701 431

236 392 230 1981 1480 501

<.01

(9.0) (18.2) (8.0) (64.8) (55.9) (8.9)

(8.3) (13.8) (8.1) (69.8) (52.1) (17.7)

<.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01

114 (2.4) 4717 (97.6)

186 (6.6) 2653 (93.5)

194 40 64 8 6 0 109 61 125 4507

(4.0) (0.8) (1.3) (0.2) (0.1) (0.0) (2.3) (1.3) (2.6) (93.3)

245 71 69 19 5 3 128 63 144 2442

(8.6) (2.5) (2.4) (0.7) (0.2) (0.1) (4.5) (2.2) (5.1) (86.0)

<.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01

3179 85 2939 2905 2905 2905 3136 81 3196 1386

(65.8) (1.8) (60.8) (60.1) (60.1) (60.1) (64.9) (1.7) (66.2) (28.7)

1828 104 1657 1624 1625 1622 1786 87 1845 822

(64.4) (3.7) (58.4) (57.2) (57.2) (57.1) (62.9) (3.1) (65.0) (29.0)

<.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01

11.6  10.4 2992 (61.9) 623 (12.9) 104 (2.2) 669 (13.9) 443 (9.2)

13.3  10.6 1521 (53.6) 428 (15.1) 78 (2.8) 481 (16.9) 331 (11.7)

<.01

0.2  0.6

0.3  0.6

<.01

0.8  1.1

0.8  1.1

<.01

14 70 1 3 92 66 107 2162 250 608 63

(0.3) (1.5) (0.0) (0.1) (1.9) (1.4) (2.2) (44.8) (5.2) (12.6) (1.3)

12 51 5 0 52 52 75 1342 140 320 43

(0.4) (1.8) (0.2) (0.0) (1.8)

<.01 <.01 <.01 .13 <.01

(1.8) (2.6) (47.3) (4.9) (11.3) (1.5)

<.01 <.01 <.01 <.01 <.01 <.01

*The percentage of older people who had the characteristics in each group, and the denominator is the number of old people in each of those groups. y The average number of hospitalization in the year prior to index date in each trajectory groups.

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

5

Table 2 Five Most Frequently Combined Uses of Psychotropic Agents (Psychotropic Polypharmacy), Stratified by Trajectory Group Infrequent Users Composition

%*

Average Duration, mo

Slowly Increasing Users

Decreasing Users

Composition

Combined use of 2 or more psychotropic agents Total 15.1 1.2 Total AX þ BZD 5.2 1.2 AX þ BZD AP þ BZD 4.2 1.2 AP þ BZD AD þ BZD 2.6 1.2 BZD þ Z BZD þ Z 2.3 1.2 AD þ BZD AP þ AX 1.1 1.1 AX þ AD Combined use of 3 or more psychotropic agent Total 1.0 1.1 Total AD þ 0.2 1.1 AX þ BZD þ Z AD þ BZD AX þ 0.2 1.0 AX þ AD þ BZD BZD þ Z AP þ 0.2 1.1 AD þ BZD þ Z BZD þ Z AX þ 0.2 1.1 AP þ BZD þ Z BZD þ Z AP þ 0.1 1.1 AP þ AD þ BZD AD þ BZD

Moderate Increasing Users

Sustained Intense Users

%*

Average Duration, mo

Composition

%*

Average Duration, mo

Composition

%*

Average Duration, mo

Composition

%*

Average Duration, mo

51.8 21.5 13.8 12.5 12.4 5.0

1.8 1.8 2.0 1.7 2.1 1.6

Total AX þ BZD BZD þ Z AD þ BZD AP þ BZD AD þ Z

50.1 17.5 16.2 13.7 13.0 5.3

2.1 2.0 2.0 2.5 2.2 2.4

Total BZD þ Z AX þ BZD AD þ BZD AP þ BZD AD þ Z

74.6 36.1 28.8 27.9 22.6 13.4

3.4 3.3 2.9 4.5 4.0 2.9

Total BZD þ Z AD þ BZD AD þ Z AP þ BZD AX þ BZD

85.8 61.0 43.6 33.5 31.7 25.2

8.1 9.1 11.5 9.0 8.4 3.8

7.1 1.7

1.3 1.2

9.0 3.0

1.6 1.8

1.9 2.1

4.7 6.1

2.0

1.5

5.1

2.4

16.7

4.1

1.4

1.5

2.0

2.0

5.1

1.7

15.7

6.3

1.2

1.3

1.5

1.2

4.9

1.6

11.0

2.6

1.2

1.7

1.4

1.4

4.0

1.9

Total AD þ BZD þ Z AP þ BZD þ Z AP þ AD þ BZD AX þ BZD þ Z AP þ AD þ Z

44.8 25.5

1.2

Total AD þ BZD þ Z AP þ AD þ BZD AP þ BZD þ Z AX þ BZD þ Z AX þ AD þ BZD

21.8 8.3

1.5

Total AD þ BZD þ Z AP þ BZD þ Z AP þ AD þ BZD AX þ BZD þ Z AX þ AD þ BZD

11.0

6.2

AD, antidepressants; AP, antipsychotics; AX, anxiolytics; Z, z-drug. *The percentage of older people who ever used a certain category of combination of psychotropic agents during the trajectory period in each group.

After the GBTM analysis and subsequent procedures of model selection, we identified 5 trajectories of longitudinal psychotropic agent use over a 5-year follow-up period as the final model. We labeled these 5 trajectories as infrequent users (n ¼ 20,543, 51.6%), slowly increasing users (n ¼ 6237, 15.7%), decreasing users (n ¼ 5353, 13.4%), moderate increasing users (n ¼ 4831, 12.1%), and sustained intense users (n ¼ 2839, 7.1%) (Figure 1). The mean posterior probabilities of group membership for each trajectory suggested high reliability and intragroup homogeneity (0.98, 0.95, 0.94, 0.97, and 0.98 for infrequent, slowly increasing, decreasing, moderate increasing, and sustained intense users, respectively).

As summarized in Table 1, demographics and comorbidities at the index date and the first measurement of psychotropic agent use differed among the distinct trajectory groups. Overall, sustained intense users were older [infrequent use: mean 72.2 (SD 5.7), slowly increasing use: mean 72.3 (SD 5.7), decreasing use: mean 72.6 (SD 5.8), moderate increasing use: mean 72.8 (SD 5.8), and sustained intense use: mean 73.1 (SD 6.0)], had the highest mean CCI, and had more admissions, psychiatric comorbidities, and other comorbidities than the other groups at baseline (P < .01). The number of days supplied of the first psychotropic agents also differed across the different trajectory groups (P < .01).

Table 3 Incidence of Unplanned Hospitalization and All-Cause Mortality Variable name

1-year follow-up period Unplanned hospitalization Event Follow-up years, mean Follow-up period, patient-years Incident, per 100 patient-years* All-cause mortality Event Follow-up years, mean Follow-up period, patient-years Incident, per 100 patient-years* 3-year follow-up period Unplanned hospitalization Event Follow-up years, mean Follow-up period, patient-years Incident, per 100 patient-years* All-cause mortality Event Follow-up years, mean Follow-up period, patient-years Incident, per 100 patient-years*

Infrequent Users

Slowly Increasing Users

Decreasing Users

Moderate Increasing Users

Sustained Intense Users

(n ¼ 20,543)

(n ¼ 6237)

(n ¼ 5353)

(n ¼ 4831)

(n ¼ 2839)

1645 1.0 19,721.3 8.3

689 0.9 5862.8 11.8y

523 1.0 5085.4 10.3y

593 0.9 4492.8 13.2y

390 0.9 2640.3 14.8y

788 1.0 20,132.1 3.9

287 1.0 6112.3 4.7y

248 1.0 5245.9 4.7y

304 1.0 4686.1 6.5y

208 1.0 2725.4 7.6y

4316 2.7 54,644.4 7.9

1655 2.6 15,966.7 10.4y

1306 2.6 13,917.8 9.4y

1387 2.5 12,125.8 11.4y

894 2.5 6983.9 12.8y

2597 2.8 57,931.3 4.5

937 2.8 17,338.9 5.4y

821 2.8 14,881.3 5.5y

907 2.7 13,092.0 6.9y

614 2.7 7580.1 8.1y

*The overall P values for crude incident rates are <.01. y The subgroup P values for crude incident rates compared to infrequent users are <.01.

6

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

Table 4 Hazard Ratios of Unplanned Hospitalization and All-Cause Mortality Model

Infrequent Users, HR (95% CI) (n ¼ 20,543)

1-year follow-up period Unplanned hospitalization Unadjusted Ref Adjusted Ref. SDH Ref. All-cause mortality Unadjusted Ref. Adjusted Ref. 3-year follow-up period Unplanned hospitalization Unadjusted Ref. Adjusted Ref. SDH Ref. All-cause mortality Unadjusted Ref. Adjusted Ref.

Slowly Increasing Users, HR (95% CI) (n ¼ 6237)

Decreasing Users, HR (95% CI) (n ¼ 5353)

Moderate Increasing Users, HR (95% CI) (n ¼ 4831)

Sustained Intense Users, HR (95% CI) (n ¼ 2839)

1.41 (1.29-1.54) 1.28 (1.17-1.40) 1.28 (1.17-1.40)

1.23 (1.12-1.36) 1.14 (1.03-1.26) 1.16 (1.03-1.26)

1.57 (1.43-1.72) 1.30 (1.18-1.44) 1.30 (1.18-1.44)

1.78 (1.60-1.99) 1.40 (1.24-1.58) 1.38 (1.22-1.57)

1.20 (1.05-1.38) 1.06 (0.92-1.21)

1.21 (1.05-1.40) 1.07 (0.92-1.23)

1.66 (1.45-1.89) 1.25 (1.09-1.44)

1.95 (1.67-2.23) 1.32 (1.12-1.56)

1.31 (1.24-1.39) 1.22 (1.15-1.30) 1.22 (1.16-1.30)

1.19 (1.12-1.26) 1.12 (1.05-1.20) 1.12 (1.05-1.19)

1.45 (1.36-1.54) 1.26 (1.19-1.35) 1.25 (1.17-1.33)

1.62 (1.50-1.74) 1.35 (1.24-1.46) 1.33 (1.22-1.44)

1.20 (1.12-1.30) 1.09 (1.01-1.18)

1.23 (1.14-1.33) 1.12 (1.04-1.21)

1.54 (1.43-1.66) 1.22 (1.13-1.33)

1.81 (1.66-1.98) 1.28 (1.16-1.41)

Ref., reference group; SDH, subdistribution hazard mode.

Most of the study participants received sedative drugs as their first psychotropic agent prescription, especially among the sustained intense users (70% vs 54%-65%). In addition, we found that more than 90% of the study participants received their first psychotropic prescription from physicians without a specialty in psychiatry and had no primary diagnosis of mental illness, dementia, or insomnia (Table 1). We also analyzed all diagnoses at the time of the first prescription and found that 23% had no diagnosis of mental illness, dementia, or insomnia. The pattern of long-term psychotropic agent use differed between the trajectory groups. Sustained intense users had the longest duration of psychotropic agent use, especially when they used sedative drugs (24.4  17.2 months) and BZDs (24.1  17.2 months) during the trajectory period (Supplementary Figure 2). In addition, we found that sustained intense users used z-drugs markedly longer than others (21.8 months vs 1.6-9.0 months). With regard to psychotropic polypharmacy, we found that 40.4% of the study participants had used 2 psychotropic agents at the same time. As shown in Table 2, sustained intense users had the highest rate of psychotropic polypharmacy (P < .01). Furthermore, 85.8% of sustained intense users had used 2 psychotropic agents, and 44.8% of them had used 3 psychotropic agents at the same time. The most common pattern of psychotropic polypharmacy in sustained intense users was a combination of benzodiazepines and z-drugs (61.0%). Table 3 shows that sustained intense users had the highest crude incidence rates of unplanned hospitalization and all-cause mortality compared to all other groups, both at 1 year of follow-up (unplanned hospitalization: 14.8 per 100 patient-years; all-cause mortality: 7.6 per 100 patient-years) and at 3 years of follow-up (unplanned hospitalization: 12.8 per 100 patient-years; all-cause mortality: 8.1 per 100 patient-years). The results of unadjusted and adjusted Cox proportional hazards models are summarized in Table 4. At the 1-year follow-up, compared to infrequent users, participants in the other trajectory groups were significantly more likely to have unplanned hospitalization after adjusting for age, sex, and comorbidity, especially among the sustained intense users (adjusted HR 1.40, 95% CI 1.24-1.58). Furthermore, moderate increasing users and sustained intense users were associated with a significantly higher risk of all-cause mortality than infrequent users (moderate increasing users: HR 1.25, 95% CI 1.091.44; sustained intense users: HR 1.32, 95% CI 1.12-1.56). Similar results were observed at 3 years of follow-up. A sensitivity analysis using the subdistribution hazards model, a competing risk analysis, tested

the influence of competing risk events and yielded results similar to our primary findings. Discussion To our knowledge, this is the first population-based longitudinal study using a GBTM approach to examine the long-term use patterns of psychotropic agents and clinical outcomes in older people. In addition, our study may be the first to examine psychotropic polypharmacy in this population based on 5 years of monthly consumption of psychotropic agents. We found that 85.8% of sustained intense users had used 2 or more psychotropic agents, and 44.8% of them had used 3 or more psychotropic agents at the same time. In this study, we identified 5 trajectory groups and found that sustained intense users, who used psychotropic agents not only with the longest duration but also with the highest proportion of psychotropic polypharmacy, were associated with the highest risk of unplanned hospitalization and death compared with infrequent users. Slowly increasing and moderate increasing users of psychotropic agents were also associated with increased risks of unplanned hospitalization and death. Our study has several methodological merits in its examination of psychotropic polypharmacy and clinical adverse outcomes compared to existing studies. First, instead of using the number of psychotropic agents,23 we used the monthly cumulative DDD of psychotropic agents received by our study participants over 5 years to measure the drug burden. We believe that this approach more accurately captures the dose-response effect of the longitudinal use of psychotropic agents, which was also confirmed by our previous study.24 Second, because most hospitalizations related to the adverse events of psychotropic agents, including falls, cardiovascular events, cerebrovascular events, and acute infection, are unplanned,11 our study focused only on unplanned hospitalization rather than all-cause hospitalization to yield more unbiased results. Third, most existing studies have only captured short-term (ie, 1-year) exposure to psychotropic agents. However, our study included a very long study period that combined a 5-year trajectory period for exposure measurement and a 3-year outcome assessment period. This longitudinal cohort study design was able to assess the long-term effects of psychotropic polypharmacy. Our finding regarding the dose-response relationship between distinct trajectory groups and adverse clinical outcomes is consistent with that of previous studies.23,24 Johnell et al reported a significantly strong dose-response relationship between psychotropic

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

polypharmacy and the risk of death and hospitalization in their casecontrol study. In their study, subjects who used more than 4 psychotropic agents concurrently had the highest risk of hospitalization (adjusted odds ratio 1.27, 95% CI 1.22-1.33) and death (adjusted odds ratio 2.50, 95% CI 2.33-2.69).23 In our study, even slowly increasing and decreasing users, who used a lower number, a lower cumulative monthly DDD, and a shorter duration of psychotropic agents during the trajectory period, were associated with both increased risks of unplanned hospitalization and death. These findings are consistent with the findings of previous studies using a 1-time measurement design and shorter exposure period.11,15 In addition, this study found that most older people in all of the trajectory groups received their first prescription of psychotropic agents from physicians without a psychiatry specialty and without a primary diagnosis of mental illness, dementia, or insomnia. The analysis of all diagnoses showed that 23% of all participants and 29% of sustained intense users did not have a diagnosis of mental illness, dementia, or insomnia. The results indicated that the primary reason our study participants went to their doctors was not for mental illness. This is highly possible because older people may go to their doctors for hypertension or diabetes mellitus and ask for help regarding insomnia or mental problems, and it may be very easy for them to receive psychotropic agents or even combinations, especially a combination of sedative agents, without consulting a psychiatric specialist. These findings are based on the “real-world” prescription patterns in Taiwan and echo previous studies in other countries. For example, Maust et al conducted a national community-based study and found that most older people received psychotropic agents from nonpsychiatrists, ranging from 82.7% to 96.5%, and had no diagnosis of mental disorder, ranging from 65.2% to 85.0%, in the United States.25 A nationwide study in the United States also found that 67% to 76% of patients received psychotropic agents from nonpsychiatrists,26 and another study of older people with information collected through telephone interviews reported that nearly 48% of older people without a psychiatric disorder were prescribed a new psychotropic agent.27 Because older people are more likely than younger adults to experience psychotropic-related adverse events when using psychotropic agents28 and because the American Geriatrics Society Beers Criteria include every class of psychotropic agent as potentially inappropriate medication for older people,10 further efforts are warranted to address psychotropic use in older people. Another important finding of our study is that many older people may have combined use of double or triple psychotropic agents for a long-term period, particularly among the moderate increasing and sustained intense users. For example, 61% of sustained intense users had combined use of BZDs and z-hypnotics for an average of 9.1 months, and 25.5% of them had combined use of antidepressants, BZDs, and z-hypnotics for an average of 6.1 months. This situation raises serious concern regarding psychotropic polypharmacy, which could be complex and must take into account multiple diagnoses, comorbid conditions, mental health, and drug interactions associated with this use pattern. For example, several previous studies23,29 and 1 of our studies8 have indicated that the load of psychotropic agents is associated with increased risk of dementia. Clinicians should pay more attention to potential psychotropic polypharmacy. Overall, the strength of this study is the novel analysis used to identify different trajectories of long-term use of psychotropic agents and to examine the association with the risks of unplanned hospitalization and mortality. Our study, which used a nationwide longitudinal database, benefits from the ability to divide the study period into 2 segments (5 years of trajectory and 3 years of follow-up) to adopt the GBTM approach to achieve our aims. Despite the effort that went into this study, there are still several limitations. First, although we adjusted for a wide range of potential covariates, we were unable to access variables not routinely captured in a claims database, such as

7

disease severity. For example, although we adjusted psychiatric comorbidities, the impact of the severity of mental diseases on longterm psychotropic agent use and clinical adverse outcomes remains unclear. Because of the nature of a retrospective observational study design, we also could not identify the causal relationship between psychotropic agent use and hospitalization or death. Although psychotropic agents may be associated with hospital admission because of inappropriate use and adverse effects, it could also be true that patients who are sicker or even under palliative care may be prescribed psychotropic agents more frequently or may concomitantly use several psychotropic agents. We attempted to capture the impacts of longitudinal use of psychotropic agents on adverse outcomes with a sophisticated study design. We believe that our study provides valuable insights into this issue. Second, our study measured only oral psychotropic agents in outpatient settings, not injected psychotropic agents or those used during admission. However, less than 3% of psychotropic agents were injected or given during inpatient visits; therefore, we believe that this influence could be negligible in our study. Finally, in our study design, we excluded older people who died during the trajectory period to fully measure the participants’ longterm exposure to psychotropic agents. Therefore, our study participants may be healthier and younger than the general population of older people. However, among these healthier participants, we still found that distinct trajectories of psychotropic agent use are associated with clinical adverse outcomes, especially for sustained intense users who persistently had a high burden of multiple psychotropic agents. We believe that this association also exists in the general older population and may be even more serious. Based on our findings, we suggest that tracking older people’s use of psychotropic medications over time, especially for those without a mental illness diagnosis from a psychiatric specialist or those not receiving palliative care, may help to identify individuals at greatest risk of unplanned hospitalization and mortality. In particular, we found that sustained intense use of psychotropic agents was associated with the highest risk of subsequent clinical adverse outcomes. We further recommend that when titrating or adding multiple psychotropic agents physicians may consider consulting psychiatric specialists regarding the history and current use of psychotropic agents to avoid inappropriate psychotropic prescribing. Conclusions and Relevance We identified 5 distinct trajectories of longitudinal psychotropic agent use in older people. Those with a persistently high burden of multiple psychotropic agents, especially sustained intense users, had the highest risk of unplanned hospitalization and mortality. Tracking older people’s use of psychotropic medication use over time may help to identify individuals at the greatest risk of unplanned hospitalization and mortality. Acknowledgments The authors thank the Ministry of Science and Technology, Taiwan, for the funding of this project (MOST 104-2410-H-002-225-MY3) and the National Health Insurance Association and National Health Research Institutes for making available the databases for this study. The content of this article, however, in no way represents any official position of the National Health Insurance Association or National Health Research Institutes. References 1. Cerejeira J, Lagarto L, Mukaetova-Ladinska EB. Behavioral and psychological symptoms of dementia. Front Neurol 2012;3:73.

8

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

2. Beck CA, Williams JV, Wang JL, et al. Psychotropic medication use in Canada. Can J Psychiatry 2005;50:605e613. 3. Richter T, Mann E, Meyer G, et al. Prevalence of psychotropic medication use among German and Austrian nursing home residents: A comparison of 3 cohorts. J Am Med Dir Assoc 2012;13. 187.e7e187.e13. 4. Chien IC, Bih SH, Chou YJ, et al. Trends in the use of psychotropic drugs in Taiwan: A population-based national health insurance study, 1997-2004. Psychiatr Serv 2007;58:554e557. 5. Frenk SM, Sautter JM, Paulose-Ram R. Prevalence and trends in psychotropic medication use among US male veterans, 1999-2010. Pharmacoepidemiol Drug Saf 2015;24:1215e1219. 6. Klotz U. Pharmacokinetics and drug metabolism in the elderly. Drug Metab Rev 2009;41:67e76. 7. Sergi G, De Rui M, Sarti S, et al. Polypharmacy in the elderly: Can comprehensive geriatric assessment reduce inappropriate medication use? Drugs Aging 2011;28:509e518. 8. Hsiao F-Y, Yang C-L, Huang Y-T, et al. Using Taiwan’s national health insurance research databases for pharmacoepidemiology research. J Food Drug Anal 2007;15:99e108. 9. Koyama A, Steinman M, Ensrud K, et al. Ten-year trajectory of potentially inappropriate medications in very old women: Importance of cognitive status. J Am Geriatr Soc 2013;61:258e263. 10. American Geriatrics Society 2015 Beers Criteria Update Expert Panel. American Geriatrics Society 2015 Updated Beers Criteria for Potentially Inappropriate Medication Use in Older Adults. J Am Geriatr Soc 2015;63: 2227e2246. 11. Huybrechts KF, Rothman KJ, Silliman RA, et al. Risk of death and hospital admission for major medical events after initiation of psychotropic medications in older adults admitted to nursing homes. CMAJ 2011;183: E411eE419. 12. Mojtabai R, Olfson M. National trends in psychotropic medication polypharmacy in office-based psychiatry. Arch Gen Psychiatry 2010;67:26e36. 13. Fulton MM, Allen ER. Polypharmacy in the elderly: A literature review. J Am Acad Nurse Pract 2005;17:123e132. 14. Wu YH, Chen CC, Wu TY. Geriatric polypharmacy in Taiwan. J Formos Med Assoc 2016;115:891e892. 15. Frandsen R, Baandrup L, Kjellberg J, et al. Increased all-cause mortality with psychotropic medication in Parkinson’s disease and controls: A

16.

17.

18. 19. 20. 21. 22. 23.

24.

25.

26.

27.

28. 29.

national register-based study. Parkinsonism Relat Disord 2014;20: 1124e1128. Guidelines for ATC Classification and DDD Assignment, 2019. Oslo, Norway: World Health Organization Collaborating Centre for Drug Statistics Methodology. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987;40:373e383. Traj: Group-based modeling of longitudinal data. Available at: https://www. andrew.cmu.edu/user/bjones/. Accessed May 31, 2017. Jones BL, Nagin DS. Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol Methods Res 2007;35:542e571. Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol 2010;6:109e138. Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J Epidemiol 2009;170:244e256. Berry SD, Ngo L, Samelson EJ, et al. Competing risk of death: An important consideration in studies of older adults. J Am Geriatr Soc 2010;58:783e787. Johnell K, Jonasdottir Bergman G, Fastbom J, et al. Psychotropic drugs and the risk of fall injuries, hospitalisations and mortality among older adults. Int J Geriatr Psychiatry 2017;32:414e420. Hsiao FY, Peng LN, Lin MH, et al. Dose-responsive effect of psychotropic drug use and subsequent dementia: A nationwide propensity score matched casecontrol study in Taiwan. J Am Med Dir Assoc 2014;15:509e513. Maust DT, Oslin DW, Marcus SC. Effect of age on the profile of psychotropic users: Results from the 2010 National Ambulatory Medical Care Survey. J Am Geriatr Soc 2014;62:358e364. Pincus HA, Tanielian TL, Marcus SC, et al. Prescribing trends in psychotropic medications: Primary care, psychiatry, and other medical specialties. JAMA 1998;279:526e531. Maust DT, Mavandadi S, Eakin A, et al. Telephone-based behavioral health assessment for older adults starting a new psychiatric medication. Am J Geriatr Psychiatry 2011;19:851e858.     Curkovi c M, Dodig-Curkovi c K, Petek Eri c A, et al. Psychotropic medications in older adults: A review. Psychiatr Danub 2016;28:13e24. Shash D, Kurth T, Bertrand M, et al. Benzodiazepine, psychotropic medication, and dementia: A population-based cohort study. Alzheimers Dement 2016;12: 604e613.

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

8.e1

Supplementary Figure 1. Study design.

30

25

20

Duration, mo

Infrequent users Slowly increasing users 15

Decreasing users Moderate increasing users Sustained intense users

10

5

0 Antipsychotics

Anxiolytics

Antidepressants

Sedative drug

BZD

Z drug

Supplementary Figure 2. Mean duration of psychotropic agents during the trajectory period, categorized by psychotropic agents. BZD, benzodiazepine; Z-drug, nonbenzodiazepine hypnotics.

8.e2

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

Supplementary Table 1 Study Medication Category, Medication Class, and ATC Codes Medication Category

Medication Class

ATC Codes

Antipsychotics

Typical

N05AA01, N05AA02, N05AB02, N05AB03, N05AB04, N05AB06, N05AC02, N05AC04, N05AD01, N05AF01, N05AF02, N05AF03, N05AF05, N05AG02, N05AH01, N05AL01, N05AX09 N05AE04, N05AH02, N05AH03, N05AH04, N05AL05, N05AX08, N05AX11, N05AX12, N05AX13 N06AA, N06CA

Antidepressants

Anxiolytics Sedative drug

Atypical Tricyclics or tetracyclics SSRIs SNRIs SARIs Other antidepressants BZDs

Z-hypnotics

N06AB N06AX16, N06AX17, N05AX21 N06AX05 N06AG02, N06AX11, N06AX12 N05BB01, N05BB02, N05BC01, N05BC03, N05BC04, N05BD01, N05BE01, N05BX01, N05BX02, N05BX03, N05BX04 N03AE01, N05BA01, N05BA02, N05BA03, N05BA05, N05BA09, N05BA16, N05BA22, N05CD01, N05CD02, N05CD03, N05CD91, N06CA01, N05BA04, N05BA06, N05BA08, N05BA12, N05BA17, N05BA56, N05BA91, N05CD04, N05CD05, N05CD06, N05CD08, N05CD09 N05CF

BZD, benzodiazepine; SARI, serotonin antagonist and reuptake inhibitor; SNRI, serotonin norepinephrine reuptake inhibitor; SSRI, selective serotonin reuptake inhibitor; Z-hypnotics, nonbenzodiazepine hypnotics.

S.-T. Huang et al. / JAMDA xxx (2019) 1e8

8.e3

Supplementary Table 2 The Comparison Table of Baseline Characteristics for Those Who Died During the 5-Year Data Collection Period (Trajectory Period) and Study Participants Variable name

Final Study Population in This Study, n (%*)

Older People Who Died During the Trajectory Period, n (%*)

Total, P Value

(n ¼ 39,803)

(n ¼ 10,543)

(n ¼ 50,346)

77.5  7.4 3947 (37.5) 4678 (44.4) 1918 (18.2)

<.01

Age, y Mean  SD 72.4  5.8 65-74 26,933 (67.7) 75-84 11,463 (28.8) >85 1407 (3.5) Sex Female 20,214 (50.8) Male 19,589 (49.2) First prescription of psychotropic agent Antipsychotics 4325 (10.9) Anxiolytics 9802 (24.6) Antidepressants 2623 (6.6) Sedative drug 23,053 (57.9) BZD 20,493 (51.5) Z-drug 2560 (6.4) Division of the first prescription of psychotropic agents in the index date Psychiatry 676 (1.7) Nonpsychiatry 39,128 (98.3) Diagnosis of the first prescription of psychotropic agents in the index date (primary diagnosis) Mental illness 1197 (3.0) Depression 244 (0.6) Psychosis 330 (0.8) Schizophrenia 45 (0.1) Delirium 19 (0.1) Bipolar disorder 12 (0.0) Anxiety 725 (1.8) Dementia 306 (0.8) Insomnia 784 (2.0) Without any of the above diagnoses 37,798 (95.0) Diagnosis of the first prescription of psychotropic agents in the index date (all diagnoses) Mental illness 29,060 (73.0) Depression 470 (1.2) Psychosis 27,545 (69.2) Schizophrenia 27,379 (68.8) Delirium 27,382 (68.8) Bipolar disorder 27,384 (68.8) Anxiety 28,847 (72.5) Dementia 400 (1.0) Insomnia 29,082 (73.1) Without any of the above diagnoses 9201 (23.1) Duration of the first prescription of psychotropic agent in the index date, d Mean  SD 7.5  7.8 1-7 29,148 (73.2) 8-14 4285 (10.8) 15-21 632 (1.6) 22-28 3347 (8.4) >28 2391 (6.0) Baseline comorbidities General indicator of comorbidity No. of hospitalizations, mean  SD 0.2  0.5 Charlson Comorbidity Index, mean  SDy 0.6  1.0 Psychiatric comorbidity Depression 61 (0.2) Psychosis 302 (0.8) Schizophrenia 11 (0.0) Delirium 12 (0.0) Anxiety 395 (1.0) Dementia 296 (0.7) Insomnia 556 (1.4) Hypertension 14,183 (35.6) Osteoporosis 1528 (3.8) Osteoarthritis 4125 (10.4) Parkinson’s disease 289 (0.7)

<.01 4130 (39.2) 6413 (60.8) 1588 2008 941 6006 4921 1085

(15.1) (19.1) (8.9) (57.0) (46.7) (10.3)

<.01 <.01 <.01 <.01 .08 <.01 <.01 <.01

359 (3.4) 10,184 (96.6) 602 68 384 8 24 5 177 341 246 9681

(5.7) (0.6) (3.6) (0.1) (0.2) (0.1) (1.7) (3.2) (2.3) (91.8)

<.01 .71 <.01 .30 <.01 .39 .32 <.01 .02 <.01

6894 146 6549 6370 6385 6366 6707 453 6919 3150

(65.4) (1.4) (62.1) (60.4) (60.6) (60.4) (63.6) (4.3) (65.6) (29.9)

<.01 .09 <.01 <.01 <.01 <.01 <.01 <.01 <.01 <.01

8.0  8.2 6882 (65.3) 1563 (14.8) 255 (2.4) 1213 (11.5) 630 (6.0)

<.01

0.6  1.1 1.4  1.7

<.01 <.01

30 392 7 8 98 374 122 4422 438 1014 271

<.01 <.01 .06 .04 .56 <.01 .06 <.01 .14 .02 <.01

(0.3) (3.7) (0.1) (0.1) (0.9) (3.6) (1.2) (41.9) (4.2) (9.6) (2.6)

BZD, benzodiazepine; SD, standard deviation; Z-drug, nonbenzodiazepine hypnotics. *The percentage of older people who had the characteristics in each group and the denominator is the number of old people in each of those groups. y The average number of hospitalizations in the year before the index date in each trajectory group.