Comparing prevalence and types of potentially inappropriate medications among patient groups in a post-acute and secondary care hospital

Reducing potentially inappropriate medications (PIMs) is a challenge in post-acute care hospitals. Some PIMs may be associated with patient characteristics and it may be useful to focus on frequent PIMs. This study aimed to identify characteristic features of PIMs by grouping patients as in everyday clinical practice. A retrospective review of medical records was conducted for 541 patients aged 75 years or older in a Japanese post-acute and secondary care hospital. PIMs on admission were identified using the Screening Tool for Older Person’s Appropriate Prescriptions for Japanese. The patients were divided into four groups based on their primary disease and reason for hospitalization: post-acute orthopedics, post-acute neurological disorders, post-acute others, and subacute. Approximately 60.8% of the patients were taking PIMs, with no significant difference among the four patient groups in terms of prevalence of PIMs (p = 0.08). However, characteristic features of PIM types were observed in each patient group. Hypnotics and nonsteroidal anti-inflammatory drugs were common in the post-acute orthopedics group, multiple antithrombotic agents in the post-acute neurological disorders group, diuretics in the post-acute others group, and hypnotics and diuretics in the subacute group. Grouping patients in clinical practice revealed characteristic features of PIM types in each group.


Evaluation of medication.
The total number of medications was counted for each patient.We included medications that were considered to be for transient use.We also included inhalants and patch medications for the treatment of internal diseases, but we excluded eye drops, nose drops, patch medications, and ointments for eye diseases, otolaryngological diseases, orthopedic diseases, and skin diseases.Intravenous or pro re nata medications were also excluded.We defined taking 5 or more medications as polypharmacy 21 .We evaluated PIMs according to the Screening Tool for Older Person's Appropriate Prescriptions for Japanese (STOPP-J) 17 .Although other tools, such as the Beers criteria 22 , are available for evaluation of PIMs, we used STOPP-J because a previous study suggested that country-oriented criteria would be clinically useful 23 .According to STOPP-J, loop diuretics or aldosterone antagonists are deemed to be PIMs regardless of the patient's condition.Thiazides were not considered to be PIMs.

Classification of patients into groups.
We divided patients into four groups based on the patient's primary disease and reason for hospitalization.The four groups are subacute, post-acute orthopedics, post-acute neurological disorders, and post-acute others.Patients in the subacute group are directly admitted to our hospital from the patient's home or nursing home for treatment of acute disease such as pneumonia."Post-acute" in the present study means transfer from an acute hospital for rehabilitation or transitional care after acute inpatient treatment.In everyday practice at our hospital, a patient's ward and treatment team are determined in this way.
In Japan, indications for hospitalization in convalescent rehabilitation wards are defined by the Ministry of Health, Labour and Welfare 24,25 .Briefly, the indications contain three disease categories: neurological disorders, including stroke and spinal cord injury; orthopedic diseases, including hip fracture, pelvic fracture, and vertebral fracture; and disuse syndrome after surgery or pneumonia.In practice, patients are rarely transferred to rehabilitation wards for disuse syndrome, and more than 90% of rehabilitation wards in Japan are occupied mostly (> 80%) by patients needing rehabilitation for neurological disorders and orthopedic diseases 26 .In addition, approximately 30% of rehabilitation wards are occupied mostly (> 80%) by either neurological or orthopedic patients 26 .Approximately 90% of post-acute hospitals have wards in addition to rehabilitation wards, and patients who need rehabilitation or transitional care but do not meet the indications are admitted to these wards 27 .Therefore, the method used in this study of dividing patients into four groups is relatively standard in Japan.

Statistical analysis.
The descriptive statistics were used to summarize the patient data.
We compared the four groups in terms of total number of medications, frequency of polypharmacy, total number of PIMs, and frequency of taking PIMs.We also compared the groups in terms of frequency of PIMs by medication category, limited to frequently taken medication categories (5% or more of all patients).Diuretics (loop diuretics and/or aldosterone antagonists) are sometimes appropriate for patients with heart failure.Therefore, we also analyzed the data for patients without heart failure who were taking diuretics.Furthermore, we investigated the frequency of proton-pomp inhibitors (PPIs) use.This was because PPIs are prescribed quite extensively in Japan, and while they are listed as PIMs according to the 2019 Beers criteria 22 , they are not included in STOPP-J 23 .It should be noted that PPIs were not included in the overall PIMs frequency calculation in the present study.Prior research has indicated a higher prescription rate of PIMs for female patients compared to male patients 11 .Therefore, we performed additional analyses to explore differences in the frequency of PIMs based on sex, both within the entire patient cohort and within each of the four patient groups.These comparisons were conducted using one-way analysis of variance, the Kruskal-Wallis test, or Fisher's exact test as appropriate.
Furthermore, to investigate the patient characteristics associated with the use of PIMs, both crude and adjusted logistic regression analyses were performed.The adjusted model included, with reference to previous study 11 , age, sex, living situation before hospitalization (at home or elsewhere), daily functional status (ISDE category), total number of medications, CCI, and the four patient groups.We also conducted similar logistic analyses for benzodiazepine receptor agonists (BZRAs, including benzodiazepines and so-called Z-drugs), because they are one of the most common and important PIMs.
All statistical analyses were performed using EZR version 1.55 (Saitama Medical Center, Jichi Medical University, Saitama, Japan).EZR is a graphical user interface for R version 4.1.2(The R Foundation for Statistical Computing, Vienna, Austria) 28 .A sample size calculation was not conducted a priori for this study.This decision was made based on our belief that any findings detectable in data from one year would hold significance in clinical practice on a ward-by-ward basis or within small hospitals.The analyses were performed without imputation of missing values.A p-value of less than 0.05 was considered statistically significant.
When comparing the four patient groups, significant differences in patient characteristics were observed (Table 1).The subacute group had a higher proportion of females, were less likely to be living at home prior to hospitalization, and had a higher prevalence of heart failure and dementia.The post-acute orthopedics group had a higher proportion of females, greater physical function (FIM), and had fewer comorbidities.Patients in the post-acute neurological disorders group were younger, had a higher body mass index, and a higher CCI (almost all patients had stroke).The post-acute others group had a higher prevalence of heart failure and malignant tumors.
The medications taken in the four patient groups are compared in Table 1.There were no significant differences among the four patient groups in the total number of medications (p = 0.18) or prevalence of PIMs  www.nature.com/scientificreports/(p = 0.08).However, characteristic features of PIM types were observed in each patient group.Patients in the subacute group were taking BZRAs (20.7%) and diuretics (32.4%) more frequently than patients in the other groups.The post-acute orthopedics group was frequently taking BZRAs (21.6%) and NSAIDs (20.7%), the postacute neurological disorders group was taking 2 or more antithrombotic agents (22.9%), and the post-acute others group were taking diuretics (31.3%).In additional analyses on diuretics, the prevalence of diuretic use without heart failure was highest (12.4%) in the subacute group.The frequency of taking PPIs was higher overall (35.9%), particularly in the post-acute neurological disorders group (62.9%).There was no sex difference in the frequency of PIMs use within each patient group (Fig. 2).Table S1 of the Supplementary Information shows the most common primary diseases in the four patient groups.The most common medications and PIMs are shown in Supplementary Tables S2 and S3, respectively.The details of multiple use of antithrombotic agents are described in Table 2.

Discussion
This study found a high prevalence of polypharmacy and PIMs in older inpatients admitted to a post-acute and secondary care hospital.When the patients were divided into four groups based on their primary disease and reason for hospitalization in the same way as in everyday clinical practice, there was no difference in the total number of medications or in the prevalence of PIMs among the groups.However, types of PIMs showed characteristic features in each group.
PIMs in all patients.In this study, 60.8% of patients met the STOPP-J criteria for PIMs.This figure is within the range of 42.3%-72.4% reported in previous studies that have used STOPP-J [29][30][31][32] .The PIMs most frequently used in our study were also in accordance with those studies [29][30][31][32] .
There were no sex differences in the use of PIMs on admission in the present study.However, it is plausible that statistically significant differences could emerge with an increase in the number of participants.Previous studies reporting sex differences in the frequency of PIMs were conducted on a large scale (n > 250,000) 33,34 .
Table 1.Demographic and clinical characteristics and medications according to patient group.Data are presented as number (%), mean ± standard deviation, or median (interquartile range) unless indicated otherwise.Cre creatinine, FIM Functional Independence Measure, ISDE Independence Scale of the Disabled Elderly, PIMs potentially inappropriate medications.a Comparison among the four subgroups.b Data on both height and weight were missing in 7 patients, and data on only height were missing in 3 patients.c In ISDE, all patients were categorized into four groups: Rank J (independent), Rank A (house-bound), Rank B (chairbound), and Rank C (bed-bound).d FIM ranges from 18 to 126; a higher score indicates better function.e Data on FIM were missing in 58 of 541 patients overall, in 56 of 275 patients in the post-acute group, and 2 of 80 patients in the post-acute others group.f Charlson Comorbidity Index ranges from 0 to 37; a higher score indicates more comorbidities.g Polypharmacy was defined as 5 or more medications.h Thiazides were not considered to be PIMs in this study.As the determinant of the use of PIMs on admission, solely the total number of medications was extracted.A prior systematic review has also identified the number of medications as the most prevalent factor associated with PIMs use 11 , and the result of the present study is concurrent with that finding.
Regarding the differences in frequency and types of PIM in post-acute or secondary care hospitals when patients are categorized into multiple groups, we were unable to identify any comparable studies to benchmark against the present study.If the number of participants were slightly larger, there might be differences in the frequency of PIMs among the patient groups.

Subacute group.
Patients in this group were frequently taking diuretics (loop diuretics and/or aldosterone antagonists) (32.4%).Diuretics can cause several complications, including falls, fractures, dehydration, and electrolyte imbalance 17 .In previous studies that used the STOPP-J criteria, 12.1%-25.6% of patients were taking diuretics 23,[29][30][31] .The difference in diuretics use in our study may reflect the prevalence of heart failure.However, it should be noted that 12.4% of patients in our subacute group were taking diuretics without a diagnosis of heart Table 2. Details of multiple use of antithrombotic agents.a All anticoagulants were direct oral anticoagulants (DOAC).b Antiplatelet plus warfarin.c Regarding anticoagulants, one patient was taking DOAC, and the other patient was taking warfarin.d Regarding anticoagulants, two patients were taking DOAC, and one patient was taking warfarin.e One patient was taking aspirin, clopidogrel, and cilostazol, and the other patient was taking these three antiplatelets and DOAC.f DOAC for deep vein thrombosis.g Warfarin for mechanical aortic valve.www.nature.com/scientificreports/failure in contrast to 3.8% in the post-acute others group.This suggests that many patients in the subacute group were taking diuretics without a clear indication, or were not recognized as having heart failure 35 .Special attention may be necessary when an unexpectedly hospitalized older patient is taking diuretics.
Post-acute orthopedics group.Many patients in this group were taking BZRAs (21.6%) on admission.
The prevalence of BZRA use in this group was almost the same as that in the subacute group (20.7%) but was higher than that in the other two groups (approximately 9 to 10%) [It should be noted that this difference may not be truly statistically significant due to multiple comparisons and a relatively high p-value (0.014)].A previous study has reported that females have higher odds of being prescribed benzodiazepines 36 .Interestingly, a higher proportion of females was noted in both the post-acute orthopedics group and the subacute group.However, the results from the multiple logistic regression analyses in the present study indicated that the use of BZRAs was more closely associated with patient group rather than sex.Another previous study reported that no medication adjustments were made during hospital stays on a conventional trauma ward 37 .BZRAs may have been discontinued before referral to our hospital in the patients in post-acute neurological disorders and postacute others groups, which might explain their lower BZRA use.Many patients in the post-acute orthopedics group were hospitalized because of fall-related fractures.As is well known, BZRAs are associated with adverse events such as falls 38 .Discontinuation of BZRAs is especially important in this population.
Post-acute neurological disorders group.Patients in this group were frequently taking multiple antithrombotic medications (22.9%).Most patients (13 out of 16 patients) were taking concomitant antiplatelet medications without any anticoagulants (Table 2).At least in the acute phase, the benefits would have outweighed the risks.However, it is better to consider reducing those medications during or after hospitalization, depending on the patient's condition, because of the potential risk of bleeding 39,40 .We collected data on the details of the antithrombotic medications and on the diagnostic names.Unfortunately, however, we did not gathered data regarding the length of stay at the referring hospital or the number of days since stroke onset.Consequently, we were unable to thoroughly assess the appropriateness of the multiple antithrombotic medications at the time of admission to our hospital.Rather, considering the frequency of prescriptions and clinical importance, the findings of the current study might indicate a necessity for intervention against hypnotics and PPIs.
Post-acute others group.Many patients in this group were taking diuretics (31.3%), and most (22 out of 25 who were taking diuretics) had heart failure.This finding suggests that, unlike in the subacute group, most patients in the post-acute others group were prescribed diuretics based on clinical necessity, and these patients were identified as having heart failure.However, diuretics should be used at the smallest dose possible with monitoring for dehydration and electrolyte abnormalities 41 .Implications for clinical practice and healthcare policy.The findings of this study might not be revolutionary; however, they do hold implications for clinical practice.Hospitals sharing similar characteristics with ours could focus on the PIMs identified in the present study.Other hospitals could develop more practical PIMs countermeasures by grouping patients in a manner suitable to each hospital and identifying the most common PIMs (which would not be overly burdensome in itself).This approach might be more feasible than intervening equally for all PIMs in hospitalized patients.The present study also has implications for healthcare policy.Since 2016, Japan has been implementing a policy that offers incentives to hospitals that succeed in reducing two or more medications per patient 42 .As a result of this policy, the prevalence of polypharmacy seems to have decreased 42 .However, conversely, the prevalence of PIMs has not decreased; in fact, it has increased 43 .A recent review indicated that healthcare policies aimed at promoting the deprescribing of specific PIMs might lead to unintended consequences 44 .One potential approach could involve assisting in the investigation and intervention of PIMs on a hospital-by-hospital or wardby-ward basis.This would also enhance the staff 's sense of participation in combating against PIMs, compared to if the government were to take the lead in reducing specific PIMs.
Limitations.This study has several limitations.If we had used other criteria for PIMs, such as the 2019 Beers criteria, the frequently taken PIMs might have been different from those identified in the present study.For instance, under the 2019 Beers criteria, multiple antithrombotic medications are not classified as PIMs.However, under the criteria, PPIs are considered as PIMs, and it would form the characteristic features of PIMs in each patient group.In essence, even with variations in PIMs criteria, the primary conclusion might remain unchanged, that is, each patient group would have characteristic features in types of PIMs.Another limitation is that this study was conducted at a single center.Therefore, caution is necessary when generalizing its results.Patients in another hospital may have to be divided in another way, and common PIMs might differ by hospital and country.Especially, providing subacute care, rehabilitation care, and other post-acute transitional care within a single hospital might be specific to Japan.However, the results from the present study suggest that grouping patients in a way suitable to each hospital may generally be helpful in understanding the status of PIMs.

Conclusion
In this study, we found a high prevalence of PIMs in older inpatients on admission in a post-acute and secondary care hospital.When we divided patients into four groups based on actual clinical practice, there was no difference in the prevalence of PIMs among the groups, but types of PIMs showed characteristic features in each group.

Figure 2 .
Figure 2. Differences in the frequency of PIMs use based on sex.

Table 3 .
Factors associated with the use of PIMs on admission (n = 538 CCI ranges from 0 to 37; a higher score indicates more comorbidities.d Adjusted model includes age, sex, living situation before hospitalization, ISDE, total number of medications, CCI, and patient group.e Post-acute neurological disorders group was selected as a reference because this group had the lowest frequency of PIMs. a ).CCI Charlson Comorbidity Index, CI confidence interval, ISDE Independence Scale of the Disabled Elderly, PIMs potentially inappropriate medications.* , ** , ***Represent p < 0.05, p < 0.01, and p < 0.001, respectively.a Out of a total of 541 patients, 3 patients with no ISDE data were excluded.b In ISDE, all patients were categorized into four groups: Rank J (independent), Rank A (house-bound), Rank B (chair-bound), and Rank C (bed-bound).c Vol.:(0123456789) Scientific Reports | (2023) 13:14543 | https://doi.org/10.1038/s41598-023-41617-0

Table 4 .
Factors associated with the use of benzodiazepine receptor agonists (BZRAs) a on admission (n = 538 b ).CCI Charlson Comorbidity Index, CI confidence interval, ISDE Independence Scale of the Disabled Elderly.* , ** , ***Represent p < 0.05, p < 0.01, and p < 0.001, respectively.aBZRAsinclude benzodiazepines and Z-drugs.bOut of a total of 541 patients, 3 patients with no ISDE data were excluded.cInISDE, all patients were categorized into four groups: Rank J (independent), Rank A (house-bound), Rank B (chair-bound), and Rank C (bed-bound).d CCI ranges from 0 to 37; a higher score indicates more comorbidities.e Adjusted model includes age, sex, living situation before hospitalization, ISDE, total number of medications, CCI, and patient group.f Post-acute others group was selected as a reference because this group had the lowest frequency of BZRAs use.