Screening criteria of potentially inappropriate medications and associations of exposure timing with incident hospital admissions among older adults in a rural community hospital Authors: Chih-Hsueh Lin, Hsin-Hui Tsai, Chun-Wei Tu, Shelly L. Gray, Hsiang-Wen Lin Published: 2025 ABSTRACT BackgroundLimited studies consider resource disparity on the impact of potentially inappropriate medications (PIMs) in rural areas. This study aimed to explore PIMs prevalence and its characteristics using different screening criteria and examine the associations of incident hospital admissions and different PIM exposure timing for older adults living in a rural area.MethodsWe conducted a retrospective cohort study to examine the one-year prevalence of PIM prevalence based on 2018 PIM-Taiwan and 2019 Beers criteria using the data retrieved from a local community hospital in a rural area during July 2018 to June 2019. We conducted a nested case-crossover study to evaluate the associations between PIM exposure in different exposure windows (most recent one-month versus 3 earlier control exposure periods) and the incident hospital admission by performing conditional logistic regressions.ResultsOf 8,234 older adult outpatients, 58% were exposed to at least one PIM (51.4% to PIM-Taiwan and 52.1% to Beers criteria). Of 384 older adult patients prescribed with medications and ever urgent hospital admissions, the most recent one-month exposure to PIM, and either or both criteria, as compared with the three earlier control periods, were statistically associated with incident hospital admissions but not for any class or individual PIMs, except Benzodiazepines (BZDs).ConclusionMost recent one-month exposure to either criteria of PIM and BZDs are the significant factors associated with incremental incident hospital admissions. Further studies to investigate the undesired outcomes associated with expose timing of specific class of PIM(s) within older adult individuals in the underserved areas than urban areas to consider the health disparity for older adults are needed.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12877-025-06663-5. Background While Asia was expected to be the second fastest area with growing aging population in the world [1], Taiwan had become the aged society (i.e., people aged 65 or older accounted for more than 14%) since 2018 [2]. In 2019, the older adults accounted for 15% of total population in Taiwan and 85% of them had at least one chronic disease [2]. The average annual expenditure per older adults increased by a margin of 7.7% from 2014 [3]. As a result, all the older adults consumed approximately 40% of National Health Insurance (NHI) medical expenditure [4]. Although patients are not restricted to access any scales of medical institutes under the implementation of NHI in Taiwan [5], older adults living in the “rural areas” were usually less likely to access the needed medical care units due to the limited medical resources, social isolation, poverty, or a lack of easy transportation, or relative less medical institutes and health care providers, as compared to that in the urban areas [6]. Thus, patients tended to visit and receive their medical care and medications in the closest medical institutes, regardless the scale of hospitals, in the rural areas in Taiwan, as that occurred in the other countries [7]. Given the older adults consumed greater proportion of health care cost or medical resources not only in Taiwan but also worldwide [8, 9], appropriate comprehensive medication review (including polypharmacy, potentially inappropriate medications [PIMs]) for the medication use among the older adults could save sufficient money (i.e., US$ 12.8 per capita cost monthly) [10]. Explicitly, PIM is defined as the use of those medications with a high risk-benefit ratio, with uncertain therapeutic effects, or with safer alternatives for the older adults [11]. PIMs increase the risks of adverse outcomes, such as falling events or gastrointestinal bleeding [12, 13]. However, the prevalence of PIM use was varied widely across studies (e.g., 62.6% according to Beers criteria in outpatient clinics of a geriatric hospital [14], or 18.5% to 82.6% in long-term care facilities [15]). PIM prevalence varies among studies likely because of the differences in study designs, practice settings (e.g., in community-dwelling units [16], geriatric outpatient clinics [10], primary care center in rural area [17], long-term care facilities, or acute care settings [18]), and used various screening criteria or tools [15, 17, 19]. For instance, a prospective study in a geriatric rural primary care center in Greece used both Screening Tool to Alert Doctors to the Right Treatment (START) and Screening Tool of Older Persons’ Prescription (STOPP) showed the significant associations between PIM and increasing emergency hospitalizations and death [17]. Although numerous studies have examined PIM use in different settings and countries [10–24], some were conducted in Taiwan [20–24]. For instance, two retrospective cohort studies using National Health Insurance Research Database (NHIRD) in Taiwan reported that PIM exposure increased hospitalization risks [23] and 90-day rehospitalization [24] but was not associated with death as compared to no use [23]. Given 2012 updated Beers Criteria was used in Lu’s study, PIM-Taiwan, PRISCUS and Beers criteria were used in Hsieh’s study [23, 24]. A limitation of these studies is that they did not consider the disparity of resources in small scale hospitals or in rural areas because these studies mainly focused on the whole population in Taiwan. Further, the potential confounding bias within the subjects for the identified cohorts might compromise the interpretation of study findings due to assuming exposures to PIM were time invariant during 10-year observation periods across different age groups of the older adults [23] or among hospitalized older adult cohorts identified by different PIM criteria between 2001 and 2018 [24]. One approach for overcoming this limitation is use a “self-matched case-control” study design (i.e., case-crossover design) using electronic healthcare databases to control for time-invariant confounders within the individuals [25]. Focusing on older adults who visited a local community hospital in a rural area, the specific aims of this study were to: (1) describe the prevalence of PIMs according to the two criteria; (2) compare the characteristics among those older adults who were exposed to PIMs or not; (3) examine the associations between PIM use and risk of hospital admissions using the case-crossover design where individuals serve as their own control. Methods Data source This study utilized the data from the electronic medical records (EMRs) from a local community hospital located in the relative rural area of “central Taiwan” (Changhua), where the proportion of older adults accounted for the total population was 17.4% and there were relative fewer medical institutes (31 various scale of hospitals) in this region. In contrast, there were less proportion of older adults across all counties in Taiwan (i.e., 16.8%), and even less in the most urbanized county in central Taiwan, “Taichung” (14.4% of the entire population and more than 65 medical institutes) [26]”. We believe older adults living in this area tended to visit this local hospital on the regular basis because of the limited medical resources in such rural areas (i.e., neither having big scale medical center nor convenient transportation). Accordingly, we retrieved the data of patient demographics, medication history, comorbidities, and healthcare utilization from the inpatient and outpatient databases during January 2018 to June 2019. Study design We conducted a retrospective cohort study to examine the one-year prevalence and PIM use based on the two different PIM criteria. Furthermore, we used a self-matched case-control design to overcome between subject confounding biases. In this case, we conducted a nested case-crossover study to evaluate the associations between PIM exposure in one case exposure period and three control exposure periods and the incident hospital admissions (i.e., hospitalization [HP] or emergency room [ER] visit) during the whole observation period (January 2018 to June 2019) (Fig. 1). In particular, the “intermittent exposures and abrupt-onset event” are the two major assumptions in the case-crossover study design [25, 27]. We assumed the following scenarios to meet the requirements of the case-crossover design: (1) PIM exposure could vary according the study period (i.e., not sustained exposure) as that presented in Supplement Fig. 1, and (2) the outcomes of interest (i.e., first urgent hospital admission [named as incident hospital admission) is a distinct event. Thus, we defined the index date as the date of first hospital admission during July 2018 to June 2019. Given medications prescribed for the older adults were usually prepared as every month or up to every three month for the chronic diseases under NHI policy in Taiwan, we decided to choose 1–30 days (most recent one-month) prior to the index date as the case period [28, 29], and three earlier control periods were identified as 61–90 days, 91–120 days, and 151–180 days (2- to 3-, 3- to 4- and 5- to 6-month) prior to the index date, respectively (Fig. 1). Fig. 1The study design of retrospective cohort (A) and nested case-crossover (B) studies. † The date of first urgent hospital admission (hospitalization or emergency room visits) during July 2018 to June 2019 The study design of retrospective cohort (A) and nested case-crossover (B) studies. † The date of first urgent hospital admission (hospitalization or emergency room visits) during July 2018 to June 2019 Specifically, we came up with three hypothetical case scenarios to showed their various exposures to PIM among those older adults who were prescribed with various medications across times within the defined case and control periods in order to demonstrate the needs to perform case-crossover design (Supplement Fig. 1). In this case, the Patient 2 ever exposed to some PIMs (melitracen, etoricoxib, ranitidine, and cytoproheptadine) during all four periods. However, Patient 1 and Patient 3 exposed to the other different PIMs in case and control 2 and 3 periods (but not during control 1). Study participants Older adult (i.e., aged 65 years and older) patients who visited the outpatient units in this community hospital during July 2018 to June 2019 were identify as our cohort 1 to further examine the objective 1 and 2, where their characteristics and health conditions within the 6 months of January 2018 to June 2018 were compared. Then, those who were hospitalized or visited ER in this community hospital during July 2018 to June 2019 were identify as our cohort 2 to further explore the objective 3. If the patients had more than one hospital admission, only the first event was considered. In addition, we excluded those who had a history of HP or ER visit within the 6 months prior to the index date. Since the exposure of PIM during case and control periods were compared in the case-crossover study, those patients without prescription records of any medication in the outpatient units in this hospital during these four study periods were also excluded. PIM exposure The PIM exposure was identified according to the 2018 PIM-Taiwan criteria [30] and 2019 American Geriatrics Society Beers criteria [31]. The PIM-Taiwan was first developed to come up with country-specific explicit PIM criteria in 2010 based on the lists of PIMs in seven published PIM criteria (including 2003 version of the Beers criteria [32]) and comprehensive reviewed by 21 experts from various specialties using Delphi approach [33]. Only the PIM independent of diseases (Table 1 in PIM-Taiwan 2018 [30] and Table 2 in Beers criteria 2019 [31], which were similar) were utilized in this study. That is because not all of the patients’ diagnoses could be identified in the databases in this hospital, which is also the common approach taken in the other studies [28, 34–36]. Specifically, those older adult outpatients who had ever been prescribed with a PIM on either list based on the proportion of day covered during the one-year period [37] were identified as PIM users. The majority of the PIM lists do not consider the drug dosage and use duration, except digoxin with dosage more than 0.125 mg per day [30, 31] and proton-pump inhibitors (PPI) with long-term duration (more than 8 weeks) [31]. Thus, we also excluded PPI more than 8 weeks from the lists of PIMs to further evaluate the association of hospital admissions and PIM use.Table 1The one-year prevalence of potentially inappropriate medications (PIM) according to the PIM-Taiwan and beers criteria†among eligible older adult patientsCategoryIndividual drug items had been prescribedPrevalence‡n (%)PIM TaiwanBeers criteriaAny PIMAs below4,737(57.5%)§4,229 (51.4%)¶4,290 (52.1%)††Cardiovascular system Aldosterone antagonistsSpironolactone111 (1.4%)✓ Digitalis glycosideDigoxin >0.125 mg/day3 (0.04%)✓✓ Peripheral alpha-1 blockersDoxazosin, Terazosin145 (1.8%)✓✓ OthersNifedipine (immediate release)28 (0.3%)✓Endocrine system EstrogensConjugated estrogens, Estradiol (combinations)16 (0.2%)✓✓ Short/rapid insulin‡‡Insulin lispro, Regular Insulin4 (0.05%)✓ Sulfonylureas, long-duration§§Glimepiride177 (2.2%)✓ ThiazolidinedionesPioglitazone37 (0.5%)✓Gastrointestinal system H2-receptor antagonistsRanitidine1,022 (12.4%)✓ PropulsiveMetoclopramide147 (1.8%)✓✓ Proton-pump inhibitors >8 weeksDexlansoprazole, Esomeprazole, Lansoprazole, Omeprazole400 (4.9%)✓Genitourinary system Drugs for urinary frequency and incontinenceOxybutynin63 (0.8%)✓ OtherDesmopressin28 (0.3%)✓Nervous system Antidepressants, SSRIsParoxetine11 (0.1%)✓✓ Antidepressants, tricyclicImipramine112 (1.4%)✓✓ Antidepressants, otherMelitracen178 (2.2%)✓ Antiparkinsonian agentsBiperiden52 (0.6%)✓✓ Antipsychotics, first-generation (conventional)Chlorpromazine, Flupentixol, Haloperidol, Sulpiride254 (3.1%)✓✓ Antipsychotics, second-generation (atypical)Olanzapine, Risperidone, Aripiprazole86 (1.0%)✓✓ BenzodiazepinesAlprazolam, Bromazepam, Clonazepam, Diazepam, Estazolam, Fludiazepam, Flunitrazepam, Lorazepam, Oxazolam1,667 (20.2%)✓✓ Nonbenzodiazepine hypnoticsZopiclone, Zolpidem232 (2.8%)✓✓ Ergoloid mesylatesDihydroergotoxine Mesylate70 (0.9%)✓Pain medications Non-COX-2 selective NSAIDAcemetacin, Diclofenac, Etodolac, Ibuprofen, Indomethacin, Mefenamic Acid, Meloxicam1,922 (23.3%)✓✓ Ketorolac (parenteral)Ketorolac (parenteral)41 (0.5%)✓ Skeletal muscle relaxantsChlorzoxazone, Cyclobenzaprine1,372 (16.7%)✓Respiratory system First-generation antihistaminesCyproheptadine, Dimenhydrinate, Meclizine, Triprolidine363 (4.4%)✓✓Others AntispasmodicsHyoscyamine109 (1.3%)✓ AntithromboticsDipyridamole244 (3.0%)✓Abbreviations: NSAIDs Nonsteroidal anti-inflammatory drugs, PIM Potentially inappropriate medications, SSRIs Selective serotonin reuptake inhibitors†According to the 2018 PIM-Taiwan and the 2019 Beers criteria, there were 53 and 63 individual PIM independent of diseases identified in the hospital formulary, respectively‡The total number of older adult patients who ever visited outpatient clinics from July 2018 to June 2019 was 8,234§Older adult patients used any PIM according to the PIM-Taiwan or the Beers criteria. There were 58 individual drugs identified in the prescriptions§Older adult patients used any PIM according to the PIM-Taiwan or the Beers criteria. There were 58 individual drugs identified in the prescriptions¶Older adult patients used any PIM according to the PIM-Taiwan criteria. There were 43 individual drugs identified in the prescriptions††Older adult patients used any PIM according to the Beers criteria. There were 53 individual drugs identified in the prescriptions‡‡Insulin regimens contained only short or rapid insulin without concurrent use of basal or long-acting insulin§§Only glyburide was included in the PIM-Taiwan criteria, but there was no older adult prescribed glyburide during the study periodTable 2Comparison of patient characteristics, health conditions and medication use between older adults with potentially inappropriate medications (PIM) and without PIM between July 2018 and June 2019PIM groupn=4,737Non-PIM group n=3,497SMDGender Females2,644 (55.8%)1,823 (52.2%)0.04Average of age74.9±7.575.4±7.9−0.07Age group* 65–742,488 (52.5%)1,780 (50.9%)0.02 75–841,653 (34.9%)1,180 (33.7%)0.01 ≥85596 (12.6%)537 (15.4%)−0.03Number of specialist visits 12,610 (55.1%)2,844 (81.3%)−0.26 21,237 (26.1%)521 (14.9%)0.11 ≥3890 (18.8%)132 (3.8%)0.15Average number of comorbidities6.7±6.84.3±4.60.41Number of comorbidities 1–2654 (13.8%)916 (26.2%)−0.12 3–41,810 (38.2%)1,747 (50.0%)0.14 ≥52,273 (48.0%)834 (23.9%)0.24Comorbidities Cancer250 (5.3%)120 (3.4%)0.02 Cardiovascular disease461 (9.7%)251 (7.2%)0.03 Cognitive impairment213 (4.5%)82 (2.3%)0.02 Congestive heart failure245 (5.2%)129 (3.7%)0.02 Coronary artery disease470 (9.9%)221 (6.3%)0.04 Diabetes1,213 (25.6%)793 (22.7%)0.03 Hypertension1,897 (40.1%)1,231 (35.2%)0.05 Infection disorders517 (10.9%)249 (7.1%)0.04 Mental disorders791 (16.7%)184 (5.3%)0.11 Peptic ulcer disease895 (18.9%)292 (8.4%)0.11 Renal disease697 (14.7%)412 (11.8%)0.03CCI <11,808 (38.2%)1,570 (44.9%)−0.07 1–21,927 (40.7%)1,445 (41.3%)−0.01 ≥31,002 (21.2%)482 (13.8%)0.07Polypharmacy †2,836 (59.9%)853 (24.4%)0.36Number of prescribed, distinct medications ≤2497 (10.5%)1235 (35.3%)−0.25 3–4824 (17.4%)981 (28.1%)−0.11 5–91848 (39.0%)1092 (31.2%)0.08 ≥101568 (33.1%)189 (5.4%)0.28Abbreviations: PIM Potentially inappropriate medications, CCI Charlson comorbidity index†The definition of polypharmacywas concurrent use of five or more medications. SMD= standard mean difference of mean or proportion (=(P1-P2)/((P1*(1-P1)+P2*(1-P2))/2)^(1/2), where P1=proportion of PIM group and P2= proportion of Non-PIM group), respectively Table 1The one-year prevalence of potentially inappropriate medications (PIM) according to the PIM-Taiwan and beers criteria†among eligible older adult patientsCategoryIndividual drug items had been prescribedPrevalence‡n (%)PIM TaiwanBeers criteriaAny PIMAs below4,737(57.5%)§4,229 (51.4%)¶4,290 (52.1%)††Cardiovascular system Aldosterone antagonistsSpironolactone111 (1.4%)✓ Digitalis glycosideDigoxin >0.125 mg/day3 (0.04%)✓✓ Peripheral alpha-1 blockersDoxazosin, Terazosin145 (1.8%)✓✓ OthersNifedipine (immediate release)28 (0.3%)✓Endocrine system EstrogensConjugated estrogens, Estradiol (combinations)16 (0.2%)✓✓ Short/rapid insulin‡‡Insulin lispro, Regular Insulin4 (0.05%)✓ Sulfonylureas, long-duration§§Glimepiride177 (2.2%)✓ ThiazolidinedionesPioglitazone37 (0.5%)✓Gastrointestinal system H2-receptor antagonistsRanitidine1,022 (12.4%)✓ PropulsiveMetoclopramide147 (1.8%)✓✓ Proton-pump inhibitors >8 weeksDexlansoprazole, Esomeprazole, Lansoprazole, Omeprazole400 (4.9%)✓Genitourinary system Drugs for urinary frequency and incontinenceOxybutynin63 (0.8%)✓ OtherDesmopressin28 (0.3%)✓Nervous system Antidepressants, SSRIsParoxetine11 (0.1%)✓✓ Antidepressants, tricyclicImipramine112 (1.4%)✓✓ Antidepressants, otherMelitracen178 (2.2%)✓ Antiparkinsonian agentsBiperiden52 (0.6%)✓✓ Antipsychotics, first-generation (conventional)Chlorpromazine, Flupentixol, Haloperidol, Sulpiride254 (3.1%)✓✓ Antipsychotics, second-generation (atypical)Olanzapine, Risperidone, Aripiprazole86 (1.0%)✓✓ BenzodiazepinesAlprazolam, Bromazepam, Clonazepam, Diazepam, Estazolam, Fludiazepam, Flunitrazepam, Lorazepam, Oxazolam1,667 (20.2%)✓✓ Nonbenzodiazepine hypnoticsZopiclone, Zolpidem232 (2.8%)✓✓ Ergoloid mesylatesDihydroergotoxine Mesylate70 (0.9%)✓Pain medications Non-COX-2 selective NSAIDAcemetacin, Diclofenac, Etodolac, Ibuprofen, Indomethacin, Mefenamic Acid, Meloxicam1,922 (23.3%)✓✓ Ketorolac (parenteral)Ketorolac (parenteral)41 (0.5%)✓ Skeletal muscle relaxantsChlorzoxazone, Cyclobenzaprine1,372 (16.7%)✓Respiratory system First-generation antihistaminesCyproheptadine, Dimenhydrinate, Meclizine, Triprolidine363 (4.4%)✓✓Others AntispasmodicsHyoscyamine109 (1.3%)✓ AntithromboticsDipyridamole244 (3.0%)✓Abbreviations: NSAIDs Nonsteroidal anti-inflammatory drugs, PIM Potentially inappropriate medications, SSRIs Selective serotonin reuptake inhibitors†According to the 2018 PIM-Taiwan and the 2019 Beers criteria, there were 53 and 63 individual PIM independent of diseases identified in the hospital formulary, respectively‡The total number of older adult patients who ever visited outpatient clinics from July 2018 to June 2019 was 8,234§Older adult patients used any PIM according to the PIM-Taiwan or the Beers criteria. There were 58 individual drugs identified in the prescriptions§Older adult patients used any PIM according to the PIM-Taiwan or the Beers criteria. There were 58 individual drugs identified in the prescriptions¶Older adult patients used any PIM according to the PIM-Taiwan criteria. There were 43 individual drugs identified in the prescriptions††Older adult patients used any PIM according to the Beers criteria. There were 53 individual drugs identified in the prescriptions‡‡Insulin regimens contained only short or rapid insulin without concurrent use of basal or long-acting insulin§§Only glyburide was included in the PIM-Taiwan criteria, but there was no older adult prescribed glyburide during the study period The one-year prevalence of potentially inappropriate medications (PIM) according to the PIM-Taiwan and beers criteria†among eligible older adult patients 4,737 (57.5%)§ Abbreviations: NSAIDs Nonsteroidal anti-inflammatory drugs, PIM Potentially inappropriate medications, SSRIs Selective serotonin reuptake inhibitors †According to the 2018 PIM-Taiwan and the 2019 Beers criteria, there were 53 and 63 individual PIM independent of diseases identified in the hospital formulary, respectively ‡The total number of older adult patients who ever visited outpatient clinics from July 2018 to June 2019 was 8,234§Older adult patients used any PIM according to the PIM-Taiwan or the Beers criteria. There were 58 individual drugs identified in the prescriptions §Older adult patients used any PIM according to the PIM-Taiwan or the Beers criteria. There were 58 individual drugs identified in the prescriptions ¶Older adult patients used any PIM according to the PIM-Taiwan criteria. There were 43 individual drugs identified in the prescriptions ††Older adult patients used any PIM according to the Beers criteria. There were 53 individual drugs identified in the prescriptions ‡‡Insulin regimens contained only short or rapid insulin without concurrent use of basal or long-acting insulin §§Only glyburide was included in the PIM-Taiwan criteria, but there was no older adult prescribed glyburide during the study period Table 2Comparison of patient characteristics, health conditions and medication use between older adults with potentially inappropriate medications (PIM) and without PIM between July 2018 and June 2019PIM groupn=4,737Non-PIM group n=3,497SMDGender Females2,644 (55.8%)1,823 (52.2%)0.04Average of age74.9±7.575.4±7.9−0.07Age group* 65–742,488 (52.5%)1,780 (50.9%)0.02 75–841,653 (34.9%)1,180 (33.7%)0.01 ≥85596 (12.6%)537 (15.4%)−0.03Number of specialist visits 12,610 (55.1%)2,844 (81.3%)−0.26 21,237 (26.1%)521 (14.9%)0.11 ≥3890 (18.8%)132 (3.8%)0.15Average number of comorbidities6.7±6.84.3±4.60.41Number of comorbidities 1–2654 (13.8%)916 (26.2%)−0.12 3–41,810 (38.2%)1,747 (50.0%)0.14 ≥52,273 (48.0%)834 (23.9%)0.24Comorbidities Cancer250 (5.3%)120 (3.4%)0.02 Cardiovascular disease461 (9.7%)251 (7.2%)0.03 Cognitive impairment213 (4.5%)82 (2.3%)0.02 Congestive heart failure245 (5.2%)129 (3.7%)0.02 Coronary artery disease470 (9.9%)221 (6.3%)0.04 Diabetes1,213 (25.6%)793 (22.7%)0.03 Hypertension1,897 (40.1%)1,231 (35.2%)0.05 Infection disorders517 (10.9%)249 (7.1%)0.04 Mental disorders791 (16.7%)184 (5.3%)0.11 Peptic ulcer disease895 (18.9%)292 (8.4%)0.11 Renal disease697 (14.7%)412 (11.8%)0.03CCI <11,808 (38.2%)1,570 (44.9%)−0.07 1–21,927 (40.7%)1,445 (41.3%)−0.01 ≥31,002 (21.2%)482 (13.8%)0.07Polypharmacy †2,836 (59.9%)853 (24.4%)0.36Number of prescribed, distinct medications ≤2497 (10.5%)1235 (35.3%)−0.25 3–4824 (17.4%)981 (28.1%)−0.11 5–91848 (39.0%)1092 (31.2%)0.08 ≥101568 (33.1%)189 (5.4%)0.28Abbreviations: PIM Potentially inappropriate medications, CCI Charlson comorbidity index†The definition of polypharmacywas concurrent use of five or more medications. SMD= standard mean difference of mean or proportion (=(P1-P2)/((P1*(1-P1)+P2*(1-P2))/2)^(1/2), where P1=proportion of PIM group and P2= proportion of Non-PIM group), respectively Comparison of patient characteristics, health conditions and medication use between older adults with potentially inappropriate medications (PIM) and without PIM between July 2018 and June 2019 Abbreviations: PIM Potentially inappropriate medications, CCI Charlson comorbidity index †The definition of polypharmacywas concurrent use of five or more medications. SMD= standard mean difference of mean or proportion (=(P1-P2)/((P1*(1-P1)+P2*(1-P2))/2)^(1/2), where P1=proportion of PIM group and P2= proportion of Non-PIM group), respectively Statistical analyses The prevalence of overall, specific class, or individual PIM use within one year for the cohort 1 was estimated, where the number of the older adults who ever visited outpatient units was identified as the denominator and the number of older adults ever prescribed with PIM was considered as the numerator. Furthermore, chi-square/Fisher Exact tests and t-tests were used to compare patient characteristics, health conditions and prescribed medications during the six-months prior the observation period between PIM users and non-users. We estimated the corresponding standard mean difference (SMD) of mean or proportion and considered the range within − 0.1 and 0.1 as a negligible difference between the two groups and otherwise [38], instead of using p-values. For cohort 2, we used chi-square/Fisher Exact tests to compare the differences of health conditions during case and control periods in the case-crossover study. Conditional logistic regressions were performed to determine the associations between incident hospital admissions and the use of PIM in different comparisons of timing. A univariate logistic regression model was used to estimate the crude odds ratios (ORs). In multivariate logistic model, those variates listed in the available literature (Charlson comorbidity index [CCI], number of prescribed medicines, and number of comorbidities) were included [29, 39]. In addition, those covariates, including specific class or individual PIMs with SMD more than |±0.1 [38] which were derived from the univariate analysis, were also included to minimally and fully adjusted model due to the limited sample size. The ORs and 95% CIs were calculated for overall exposure to any PIM according to PIM-Taiwan, Beers criteria, or both criteria and those individual classes or items of PIM. All analyses were performed using SAS version 9.4. Results Use of PIM Of the 8,234 older adults, 4,737 (57.5%) had been prescribed with at least one PIM (Supplement Fig. 2). The prevalence of PIM use was 51.4% and 52.1% according to PIM-Taiwan and Beers criteria, respectively. The most prevalent classes of PIM were non-Cyclooxygenase-2 (COX-2) selective NSAIDs (23.3%), BZDs (20.2%), skeletal muscle relaxants (16.7%), and H2-receptor antagonists (12.4%) (Table 2). The top five prescribed individual items of PIM were chlorzoxazone (15.5%), ranitidine (12.4%), diclofenac (12.0%), alprazolam (9.5%), and meloxicam (7.7%), respectively (data not shown). Among those PIM users, 46.1% (n=2,181) and 27.3% (n=1,295) were exposed to one and two PIMs, respectively. More than one fourth of the PIM users (26.6%, n=1,261) received three or more PIM prescriptions (data not shown). The average age of PIM users was 74.9±7.5 years-old and more than half of the PIM users were female (55.8%, n=2,644) (Table 2). In contrast with non-PIM users, PIM users were slightly younger, more were female, or encountered severe diseases (CCI≥3) (all SMDs not ³|±0.1|), were prescribed with more medications, encountered polypharmacy, had more comorbidities, and ever visited more different specialists (all SMDs ³|±0.1|. PIM and hospital admissions Overall, 2,543 older adult patients had either a hospital or ER admission. After excluding those patients with hospital admission history before the index date and without prescription of any medication during the four periods, a total of 384 older adult patients (79.83% ever visited ER urgently and 29.17% had incident hospitalizations, as cohort 2) were eligible in the case-crossover study (Supplement Fig. 2). The prevalence of polypharmacy, number of prescribed/distinct medications, number of specialty visits, number of comorbidities, comorbidities of interest, and CCI during case period were similar with the other three control periods (Supplement Table 1). Given all patients in cohort 2 were prescribed with medications across times, almost all of them also exposed to various types of PIM in the four periods (100%), except that in control 1 periods (91.5%). Use of PIM identified by either or both criteria in the case period significantly increased the risks of incident hospital admissions when comparing with the other three earlier control periods (Comparison 1 through 3 in Table 3). Further, the results were statistically significant when examining each set of criteria separately. For example, the adjusted ORs (95% CI) for PIM specific according to PIM-Taiwan criteria were 2.24 (1.25–4.00.25.00) and 2.07 (1.17–3.65.17.65) between case period and control period 1 and 2, respectively, while that were 1.98 (1.12–3.50.12.50) and 1.81 (1.02–3.22.02.22) based upon Beers criteria (Table 3). Table 3Potential risk of incident and urgent hospital admissions associated with exposure to potentially inappropriate medications among older adult patients in cohort 2†PIMComparison 1Comparison 2Comparison 3Case period‡ vs control 1 period§Exposure OR (95% CI)Case period vs control 2 period¶Exposure OR (95% CI)Case period vs control 3 period††Exposure OR (95% CI)CrudeAdjusted‡‡CrudeAdjusted‡‡CrudeAdjusted‡‡Any of PIM§§1.82 (1.08- 3.06)*1.80 (1.04- 3.11)*1.96 (1.17–3.27.17.27)*1.91 (1.05–3.47.05.47)*1.55 (0.99–2.43.99.43)1.86 (1.12–3.07.12.07)*PIM according to PIM-Taiwan2.21 (1.29–3.80.29.80)*2.24 (1.25–4.00.25.00)*2.18 (1.32–3.61.32.61)*2.07 (1.17–3.65.17.65)*1.63 (1.04–2.57.04.57)*1.89 (1.15–3.12.15.12)*PIM according to Beers criteria¶¶1.95 (1.14–3.34.14.34)*1.98 (1.12–3.50.12.50)*2.05 (1.20–3.50.20.50)*1.81 (1.02–3.22.02.22)*1.52 (0.96–2.39.96.39)1.89 (1.14–3.14.14.14)*Peripheral alpha-1 blockers0.33 (0.04–3.21.04.21)0.19 (0.02–2.15.02.15)1.00 (0.06–15.99.06.99)0.26 (0.01–7.64.01.64)2.00 (0.37–10.92.37.92)1.30 (0.23–7.42.23.42)Ranitidine0.77 (0.34–1.75.34.75)0.65 (0.27–1.59.27.59)1.18 (0.53–2.64.53.64)1.04 (0.44–2.50.44.50)1.14 (0.56–2.34.56.34)1.22 (0.58–2.57.58.57)Melitracen0.40 (0.08–2.06.08.06)0.21 (0.03–1.29.03.29)0.67 (0.11–3.99.11.99)0.31 (0.04–2.35.04.35)0.50 (0.09–2.73.09.73)0.43 (0.08–2.39.08.39)Imipramine0.67 (0.11–3.99.11.99)0.38 (0.06–2.47.06.47)0.67 (0.11–3.99.11.99)0.32 (0.04–2.33.04.33)1.00 (0.14–7.10.14.10)0.86 (0.10–7.13.10.13)Antipsychotics, first-generation0.43 (0.11–1.66.11.66)0.30 (0.07–1.32.07.32)0.50 (0.13–2.00.13.00)0.40 (0.09–1.83.09.83)0.38 (0.10–1.41.10.41)0.34 (0.09–1.29.09.29)Antipsychotics, second-generation1.50 (0.25–8.98.25.98)1.57 (0.26–9.63.26.63)1.50 (0.25–8.98.25.98)1.43 (0.23–8.75.23.75)1.50 (0.25–8.98.25.98)1.94 (0.31–12.04.31.04)Benzodiazepines2.44 (1.13–5.31.13.31)*2.41 (1.08–5.41.08.41)*2.27 (1.12–4.62.12.62)*2.37 (1.08–5.18.08.18)*1.59 (0.87–2.91.87.91)1.66 (0.87–3.15.87.15)Non-benzodiazepine hypnotics (Z-drugs)3.00 (0.31–28.84.31.84)1.63 (0.15- 17.56)5.00 (0.58–42.80.58.80)3.45 (0.38–31.32.38.32)2.00 (0.50–8.00.50.00)1.93 (0.458–8.32.458.32)Dihydroergotoxine3.00 (0.31–28.84.31.84)2.14 (0.20- 22.48)3.00 (0.31–28.84.31.84)1.84 (0.17–20.06.17.06)3.00 (0.31–28.84.31.84)5.58 (0.53–58.69.53.69)Non-COX-2 selective NSAID1.86 (0.97–3.56.97.56)1.71 (0.86–3.41.86.41)1.71 (0.89–3.31.89.31)1.40 (0.68 - 2.89)1.29 (0.73–2.27.73.27)1.48 (0.81–2.71.81.71)Skeletal muscle relaxants1.57 (0.61–4.05.61.05)1.06 (0.38–2.93.38.93)1.00 (0.46–2.16.46.16)0.81 (0.35–1.85.35.85)1.00 (0.48–2.10.48.10)1.12 (0.50–2.49.50.49)First-generation antihistamines0.86 (0.29–2.55.29.55)0.83 (0.26–2.63.26.63)1.00 (0.32–3.10.32.10)0.75 (0.23–2.47.23.47)0.60 (0.22–1.65.22.65)0.66 (0.23–1.90.23.90)Hyoscyamine0.50 (0.05–5.51.05.51)0.41 (0.03–4.76.03.76)1.00 (0.06–15.99.06.99)1.51 (0.09–25.10.09.10)1.00 (0.06–15.99.06.99)1.00 (0.06–15.99.06.99)Abbreviations: PIM Potentially inappropriate medications, OR Odds ratio, CI Confidence interval, COX Cyclooxygenase, NSAID Nonsteroidal anti-inflammatory drugs†The total number of participants included in case-crossover study was 384‡ Case period: 1–30 days prior to the index date (i.e., the first date of hospitalization or emergency department visiting during July 2018 to June 2019)§Control 1 period: 61–90 days prior to the index date¶Control 2 period: 91–120 days prior to the index date††Control 3 period: 151–180 days prior to the index date‡‡Adjusted for polypharmacy, Number of prescribed, distinct medications, Number of visiting division, Number of comorbidities, Charlson comorbidity index, and those covariates with standard mean difference (SMD) >|±0.1| in univariate analysis§§Any PIM according to the PIM-Taiwan or Beers criteria except proton-pump inhibitors use for long-term (>8weeks)¶¶Any PIM according to the Beers criteria except proton-pump inhibitors use for long-term (>8weeks) Table 3Potential risk of incident and urgent hospital admissions associated with exposure to potentially inappropriate medications among older adult patients in cohort 2†PIMComparison 1Comparison 2Comparison 3Case period‡ vs control 1 period§Exposure OR (95% CI)Case period vs control 2 period¶Exposure OR (95% CI)Case period vs control 3 period††Exposure OR (95% CI)CrudeAdjusted‡‡CrudeAdjusted‡‡CrudeAdjusted‡‡Any of PIM§§1.82 (1.08- 3.06)*1.80 (1.04- 3.11)*1.96 (1.17–3.27.17.27)*1.91 (1.05–3.47.05.47)*1.55 (0.99–2.43.99.43)1.86 (1.12–3.07.12.07)*PIM according to PIM-Taiwan2.21 (1.29–3.80.29.80)*2.24 (1.25–4.00.25.00)*2.18 (1.32–3.61.32.61)*2.07 (1.17–3.65.17.65)*1.63 (1.04–2.57.04.57)*1.89 (1.15–3.12.15.12)*PIM according to Beers criteria¶¶1.95 (1.14–3.34.14.34)*1.98 (1.12–3.50.12.50)*2.05 (1.20–3.50.20.50)*1.81 (1.02–3.22.02.22)*1.52 (0.96–2.39.96.39)1.89 (1.14–3.14.14.14)*Peripheral alpha-1 blockers0.33 (0.04–3.21.04.21)0.19 (0.02–2.15.02.15)1.00 (0.06–15.99.06.99)0.26 (0.01–7.64.01.64)2.00 (0.37–10.92.37.92)1.30 (0.23–7.42.23.42)Ranitidine0.77 (0.34–1.75.34.75)0.65 (0.27–1.59.27.59)1.18 (0.53–2.64.53.64)1.04 (0.44–2.50.44.50)1.14 (0.56–2.34.56.34)1.22 (0.58–2.57.58.57)Melitracen0.40 (0.08–2.06.08.06)0.21 (0.03–1.29.03.29)0.67 (0.11–3.99.11.99)0.31 (0.04–2.35.04.35)0.50 (0.09–2.73.09.73)0.43 (0.08–2.39.08.39)Imipramine0.67 (0.11–3.99.11.99)0.38 (0.06–2.47.06.47)0.67 (0.11–3.99.11.99)0.32 (0.04–2.33.04.33)1.00 (0.14–7.10.14.10)0.86 (0.10–7.13.10.13)Antipsychotics, first-generation0.43 (0.11–1.66.11.66)0.30 (0.07–1.32.07.32)0.50 (0.13–2.00.13.00)0.40 (0.09–1.83.09.83)0.38 (0.10–1.41.10.41)0.34 (0.09–1.29.09.29)Antipsychotics, second-generation1.50 (0.25–8.98.25.98)1.57 (0.26–9.63.26.63)1.50 (0.25–8.98.25.98)1.43 (0.23–8.75.23.75)1.50 (0.25–8.98.25.98)1.94 (0.31–12.04.31.04)Benzodiazepines2.44 (1.13–5.31.13.31)*2.41 (1.08–5.41.08.41)*2.27 (1.12–4.62.12.62)*2.37 (1.08–5.18.08.18)*1.59 (0.87–2.91.87.91)1.66 (0.87–3.15.87.15)Non-benzodiazepine hypnotics (Z-drugs)3.00 (0.31–28.84.31.84)1.63 (0.15- 17.56)5.00 (0.58–42.80.58.80)3.45 (0.38–31.32.38.32)2.00 (0.50–8.00.50.00)1.93 (0.458–8.32.458.32)Dihydroergotoxine3.00 (0.31–28.84.31.84)2.14 (0.20- 22.48)3.00 (0.31–28.84.31.84)1.84 (0.17–20.06.17.06)3.00 (0.31–28.84.31.84)5.58 (0.53–58.69.53.69)Non-COX-2 selective NSAID1.86 (0.97–3.56.97.56)1.71 (0.86–3.41.86.41)1.71 (0.89–3.31.89.31)1.40 (0.68 - 2.89)1.29 (0.73–2.27.73.27)1.48 (0.81–2.71.81.71)Skeletal muscle relaxants1.57 (0.61–4.05.61.05)1.06 (0.38–2.93.38.93)1.00 (0.46–2.16.46.16)0.81 (0.35–1.85.35.85)1.00 (0.48–2.10.48.10)1.12 (0.50–2.49.50.49)First-generation antihistamines0.86 (0.29–2.55.29.55)0.83 (0.26–2.63.26.63)1.00 (0.32–3.10.32.10)0.75 (0.23–2.47.23.47)0.60 (0.22–1.65.22.65)0.66 (0.23–1.90.23.90)Hyoscyamine0.50 (0.05–5.51.05.51)0.41 (0.03–4.76.03.76)1.00 (0.06–15.99.06.99)1.51 (0.09–25.10.09.10)1.00 (0.06–15.99.06.99)1.00 (0.06–15.99.06.99)Abbreviations: PIM Potentially inappropriate medications, OR Odds ratio, CI Confidence interval, COX Cyclooxygenase, NSAID Nonsteroidal anti-inflammatory drugs†The total number of participants included in case-crossover study was 384‡ Case period: 1–30 days prior to the index date (i.e., the first date of hospitalization or emergency department visiting during July 2018 to June 2019)§Control 1 period: 61–90 days prior to the index date¶Control 2 period: 91–120 days prior to the index date††Control 3 period: 151–180 days prior to the index date‡‡Adjusted for polypharmacy, Number of prescribed, distinct medications, Number of visiting division, Number of comorbidities, Charlson comorbidity index, and those covariates with standard mean difference (SMD) >|±0.1| in univariate analysis§§Any PIM according to the PIM-Taiwan or Beers criteria except proton-pump inhibitors use for long-term (>8weeks)¶¶Any PIM according to the Beers criteria except proton-pump inhibitors use for long-term (>8weeks) Potential risk of incident and urgent hospital admissions associated with exposure to potentially inappropriate medications among older adult patients in cohort 2† Abbreviations: PIM Potentially inappropriate medications, OR Odds ratio, CI Confidence interval, COX Cyclooxygenase, NSAID Nonsteroidal anti-inflammatory drugs †The total number of participants included in case-crossover study was 384 ‡ Case period: 1–30 days prior to the index date (i.e., the first date of hospitalization or emergency department visiting during July 2018 to June 2019) §Control 1 period: 61–90 days prior to the index date ¶Control 2 period: 91–120 days prior to the index date ††Control 3 period: 151–180 days prior to the index date ‡‡Adjusted for polypharmacy, Number of prescribed, distinct medications, Number of visiting division, Number of comorbidities, Charlson comorbidity index, and those covariates with standard mean difference (SMD) >|±0.1| in univariate analysis §§Any PIM according to the PIM-Taiwan or Beers criteria except proton-pump inhibitors use for long-term (>8weeks) ¶¶Any PIM according to the Beers criteria except proton-pump inhibitors use for long-term (>8weeks) When examining specific PIM criteria, only benzodiazepines (BZDs) were associated with increased risk for incident hospital admissions. All of the 95% CIs for ORs were all overlapping although 95% CIs for exposure to any or either PIM criteria were more precise than the specific class or individual items of PIM. To compare with control 1 or control 2 periods, specifically, use of BZDs in the case period had more than doubled risks of incident hospital admissions, while there were no significant differences in the comparisons between case period and control 3 period. Discussion Approximately 58% of the older adult outpatients ever exposed to at least one PIM, with the most common prescribed classes of PIM, including non-COX-2 selective NSAIDs, BZDs, and skeletal muscle relaxants. Among those older adults who had incident hospital or ER admissions in this community hospital in a rural area, PIM use was significantly associated with incident urgent hospital admission, as identified by either criteria, or exposure to BZDs. In the other words, this study showed that the exposure to PIM within the most recent one-month than the other three earlier control periods and using either criteria would influence the interpretation of association extents between incident hospital admissions and exposure to PIM. The prevalence of PIM ranges considerably wide among studies (range from ~18% to 82% in one systematic review [15], ~47% to 72% using 2001-2018 NHIRD [24], 61% in a Greek rural primary care center [17], to the others of 0–98.0.0% [19, 40]). One of the major reasons might be due to the various used screening criteria or tools (e.g., PIM-Taiwan, PRICUS, and Beers criteria in Hsieh’s study [24], STOPP criteria [39], START/STOPP version 2 [17]). However, the prevalence of PIM defined by 2018 PIM-Taiwan was slightly lower than 2019 Beers criteria, which was consistent with the aforementioned study [24]. While 2010 PIM-Taiwan criteria had lower levels of sensitivity and higher levels of specificity [41], using the updated version of PIM-Taiwan and carried out in Taiwan seemed to have better sensitivity for Taiwanese older adults who used medications. Further, the prevalence of PIM in this study (~58%) seems lower than what has been reported by the other studies conducted in Taiwan (62.5–86.2.5.2%) [24, 36, 42, 43]. This may be related to the fact that we only included one local community hospital, whereas the other studies focused the population prevalence using NHIRD or including urban medical center patients. Further, we assumed different version of criteria/tools and datasets might also cause the difference of prevalence, i.e., the overall prevalence increased from 46.1% based on 2010 version to 74.6% based on 2018 version of PIM-Taiwan criteria [20]. Given there is almost no restriction to access various scales of medical institutes in Taiwan [5], our PIM prevalence focusing on those older adults who tended to visit a local community hospital in a rural area was similar to that in a prospective cohort study conducted in a rural primary care center in Greek (i.e., 61%) [17]. More studies to explore the impact due to health disparity of seeking or accessing medical care among those older adults prescribed with medications in rural areas as compared to urban areas are needed. Our study showed that those older adults with polypharmacy or having more comorbidities tended to be prescribed with PIM, which was consistent with the other studies [17, 35, 44–46]. The PIM defined by the STOPP criteria was associated with hospital admissions, but not as that defined by Beers criteria, in a nested case-control study [39]. Our study showed that PIM defined by either 2018 PIM-Taiwan or 2019 Beers criteria significantly increased the risk of incident hospital or ER admission. The association was similar for these two criteria in our study, which was in contrast with the most recent retrospective cohort study for different outcome of interest using NHIRD in Taiwan (i.e., 2019 Beers criteria seemed more associated with 90-day rehospitalization) [24]. Given exposing to inappropriate medications would increase the risk of adverse events (e.g., hospital admission, and mortality, etc.) [17, 22, 28, 29, 34, 36] and healthcare utilization (e.g., number of prescribed medications, and costs [28, 47]), but not the needs of long-term care [47]), patients’ or drug-related factors (e.g., drug classes or numbers) should be considered about its contribution of PIM encounters [35, 47]. Heider’s cohort study found that older adults exposed to PIM had 2.9 additional different drug classes [47], Roux’s population-based study showed that those older adults prescribed with more medications were exposed to higher risk of using PIM [35]. Further, patients’ characteristics (e.g., age, gender, and socioeconomic status, etc.) were usually statistically associated with both PIM and polypharmacy exposure [14, 35, 48, 49]. In the other words, the relationships between the interested outcomes (i.e., incident and urgent hospital admission) with PIM and polypharmacy, respectively, were correlated but might be complicated. The associations between PIM use and hospitalization has been reported in a systematic review and meta-analysis studies [50, 51]. In contrast to the other studies, we performed the self-controlled study design to try to control for the characteristics among the individuals over time (including unstable diseases and changes of medications) and utilized various earlier control periods to examine the associations of most recent one-month PIM exposure and incident hospital admissions among the older adults. While Sato’s study designed with five short case periods (1, 2, 4, 8, and 12 weeks) to compare with the corresponding farther control period as 1 year prior to unplanned hospitalization for their case-crossover study [29], we choose to go for only one case period (most recent one-month) to compare with the three prior and earlier control periods. As a result, the ORs of unplanned hospitalization in the 8-weeks and 12-week case periods than the control period ranged from 1.48 to 2 based upon the records of more than 240,000 Japanese older adults in Sato’s study [29]. This is very similar to the ORs of out case period to the three earlier control periods prior the incident hospital admission (around 1.8 after controlling for the other factors) in our study (i.e., among 384 older adults). Overall, these evidences showed that time when exposure to the PIM (most recent one month as compared to the other earlier control periods) is one of the important factors associated with the extent of undesired consequences of PIM exposure. Interestingly, the most frequently prescribed PIMs identified in this community hospital were non-COX-2 selective NSAIDs and BZDs, which were similar to the findings obtained from the other population-based studies conducted either in Taiwan or in the other countries [35, 42, 43, 52]. Since non-COX-2 selective NSAIDs might increase risk of gastrointestinal bleeding or peptic ulcer disease for the older adults, both criteria have listed these drugs as inappropriate medications [30, 31]. Thus, those COX-2 selective NSAIDs are considered as the alternatives. However, non-selective NSAIDs with cheaper price were the first choice due to the cost and payment restriction of NHI policy in Taiwan [53]. Although there were nearly one-fourth older adult outpatients had ever used non-selective NSAIDs, we did not identify a significant association between incident hospital admission and prescription of non-selective NSAIDs. Since H2-receptor antagonists and PPIs are proved to reduce gastrointestinal adverse effects associated NSAIDs [54], the non-significant risk might be due to the prevalent use of H2-receptor antagonists and PPIs in Taiwan. Furthermore, this study revealed that older adults prescribed with BZDs but not Z-drugs were more likely to encounter the risk of incident hospital admissions. In contrast, the other studies revealed the contributions of both BZDs and Z-drugs on the negative outcomes of hip fractures, fall requiring hospitalization, or ED visits [55–57]. Overall, there are some limitations in this study. First, only the most important diseases could be identified from the electronic databases in this local community hospital due to the EMR format restriction. Thus, we try to exclude those PIMs with disease-interactions to reduce the possible confounding biases. Second, only those PIMs included in this hospital formulary could be investigated even if the prevalent drug classes of PIMs detected in this study were similar to the other population-based studies. Third, PIM exposure might be underestimated due to lack of information about patients’ compliance and complete medication history outside this local hospital. As a result, we perform the case-crossover study design, as a self-controlled design, to investigate the risk associations through reducing the influence of between-subject confounders, and to reflect the real conditions and serve as a reference for clinical care in the real-world practice. Fourth, the number of selected control periods would influence the sample size of participants in the case-crossover study because only those older adults still used prescriptions in all three earlier control periods were considered for further analysis. The sample size might increase if the number of control periods were reduced to two or one, but the comparisons between exposure windows might be reduced. Fifth, there were limited evidences to explore the outcomes after exposing to the PIM criteria so that we can only choose to go for more concrete outcomes (i.e., falls). Last but not the least, the generalizability of this study findings to the other community hospitals in different rural areas in Taiwan or in other countries is still limited and might compromise the interpretation given this is just a convenient sample of database from the local hospital during 2018 to 2019 in a rural area of central Taiwan. Although American Geriatrics Society updated Beers criteriaÒ in 2023 [58], there was no way to get the most current and updated EMR in this community hospital to perform the retrospective data analysis (i.e., data only available up to 2019). Further study to confirm such findings and its implication using the updated data might be needed. In conclusion, most recent one-month exposure to PIM, and both used criteria as a whole to assess PIM are the main factors associated with incident hospital admissions among the older adults, given patients were self-compared their exposure to most recent one-month to different earlier control periods. The exposure to BZDs tended to increase the risk of incident hospital admissions significantly. Such findings suggest that healthcare professionals, administrators in medical institutes, and those insurance officers either in Taiwan and/or in the other countries should aware the possible risks associated with PIMs, monitor the adverse effects carefully, and consider prescribing alternative drugs, and its undesired patients’ consequences. Further studies are needed to investigate the undesired outcomes associated with exposed timing of PIMs within individuals in the underserved areas compared with urban areas in order to consider the impact of health disparity in different scales of medical institutes and different areas or regions on medication use for the older adults. Supplementary Information Supplementary Material 1 Supplementary Material 1