Characteristics and Risk Factors of Medication Incidents Across Stages of Medication Management in Residential Aged Care: A Longitudinal Cohort Study of 5700 Reported Incidents

Abstract

Background

Problems with medication management are consistently identified as key concerns for the quality of residential aged care (RAC). Incident reports can provide valuable information on key issues related to medication management; however, few studies have explored medication incidents in RAC settings.

Objectives

To investigate the characteristics of medication incidents at different stages of medication management and identify the risk factors associated with incidents.

Methods

A retrospective longitudinal cohort study was conducted using medication incidence data from 25 RAC facilities in New South Wales, Australia. All medication incidents between 1 July 2014 and 31 August 2021 relating to 5709 aged care residents aged ≥ 65 years were included. The outcome measure was the medication incidence rate (IR), quantified as the number of medication incidents per 1000 resident days. A multilevel Poisson regression model was performed to identify risk factors associated with exposure to medication incidents.

Results

A total of 5708 medication incidents were analysed. The overall medication IR was 1.81 per 1000 resident days (95% CI 1.76, 1.86). Of 5709 residents, 35% (n = 2016) had at least one recorded medication incident, of which 1095 (> 50%) had more than one. The majority of the incidents were associated with medication administration (3023 incidents, 53%), followed by supply (n = 1546, 27%) and monitoring the response to the medication (n = 548, 9.6%). The outcome of the incident on residents was reported in 5165 (90%) incidents, with 724 (14%) requiring the resident to be monitored by the hospital, general practitioner (GP), or staff. Respite admissions were associated with a higher risk of medication incidents including potentially harmful incidents, compared with permanent admissions (rate ratio (RR) = 1.908, 95% CI 1.646, 2.211, p < 0.01). Residents with Parkinson’s disease had a 1.5-fold greater risk of a medication incident (RR = 1.586, 95% CI 1.318, 1.908) compared with residents without Parkinson’s. The administration of more than five medications (polypharmacy) was associated with an increased risk of medication incidents (RR = 2.019, 95% CI 1.930, 2.111).

Conclusions

Medication incidents affected more than one-third of older adults in RAC facilities. Improvement strategies should focus on medication administration, supply and monitoring, with particular attention given to respite residents and those with multimorbidity and polypharmacy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40264-025-01602-0.

Key Points

More than one-third of older adults in RACF experience medication incidents, with the majority related to medication administration, followed by supply and monitoring residents’ responses.

More than half of residents with a prior history of medication incidents are likely to experience repeated incidents.

Respite admissions have twice the risk of medication incidents, including potentially harmful ones, compared to permanent admissions.

Introduction

Medication errors and adverse drug events often lead to hospital admissions [1]. Annually, approximately 250,000 hospital admissions in Australia stem from medication-related issues [2] and are estimated to account for between 2 and 3% of all hospital admissions [3].

In residential aged care (RAC) facilities, medication management relies on a series of critical tasks, including prescribing, supplying, storing, administering, monitoring responses, communicating and providing consumer advice [4]. Medication incidents can occur at any stage of the medication management process [5] in RAC facilities. The increasing prevalence of multimorbidity [6] and polypharmacy [7], along with diverse administration schedules and complex dosages, collectively create a highly complex environment for medication management [8] in RAC facilities. Moreover, residents in these care facilities demonstrate distinct lifestyle, pharmacokinetic and pharmacodynamic traits compared with those of other age groups [9], and the complexities of medication management and high staff turnover can lead to medication incidents [10–12].

A prospective study from the UK revealed that 69.5% of older adults in RAC facilities had one or more medication incidents involving medication prescription, dispensing or administration [13]. These medication incidents were identified through resident interviews, note review, observation of practice and examination of dispensed items. Another prospective observational study conducted for 30 days in UK RAC facilities revealed that 7% of administered doses led to a medication administration error [14], with error rates ranging from 2.1 to 15.9% at the facility level. An Australian cross-sectional observational study identified issues with more than one in ten doses of administration aids. Half of these incidents were related to unsuitable repacking followed by addition, omission and incorrect quantity [15]. A retrospective study utilising de-identified medication administration electronic records in RAC facilities revealed that 73% of residents had at least one dose omission [16], and a mean rate of 3.59 medication doses were omitted per 100 dispensed doses per resident. The study also revealed that there is a significant relationship between the resident level of care and RAC facility characteristics and the rate of omission. Residents receiving palliative care consistently had a greater percentage of omissions compared with other residents in the facility. In addition, corporate-governed facilities exhibited a higher rate compared with independently owned facilities.

Recognising and reporting medication incidents are pivotal for resident safety and wellbeing [17, 18]. Incident reports provide an important source of information for organisations to inform effective prevention strategies. Analysing medication incidents, including near misses, which occur more frequently than adverse events, offers valuable insights for proactively reducing errors and uncovering hidden dangers that indicate system vulnerabilities [19]. However, studies investigating the incidence characteristics within medication management stages, facility-level differences, and associated resident-level risk factors were infrequent in the RAC setting [20]. To address these gaps, we conducted a retrospective dynamic longitudinal cohort study of medication incidents across 25 RAC facilities to identify where incidents occur in the medication management process, along with residents, facilities and medication-level risk factors associated with those incidents.

Methodology

Study Design

The study period was from 1 July 2014 to 31 August 2021. In this study, residents can have rolling admissions during the study period where they can enter or leave or re-enter the same facility or a different facility at any time.

Study Setting

The study was undertaken using routinely collected data from 25 RAC facilities managed by a large not-for-profit RAC provider who offers a range of aged care services across metropolitan, inner and outer regional areas of New South Wales, Australia. The study received ethics approval from the Macquarie University Human Research Ethics Committee (reference no. 52019614412614).

Participants

This study consisted of both respite and permanent residents aged ≥ 65 years who entered a RAC facility during the study period (n = 6413). The flow chart provided in Supplementary Fig. 1 shows the inclusion and exclusion criteria used for cohort selection. Residents with interim care, without a medication record, same-day discharge or a length of stay less than 7 days were excluded from the study (n = 704). This process yielded a final sample of 5709 participants.

Data Sources, Variables, and Linkages

Electronic incident reporting and electronic medication management systems were implemented in the facilities in early 2014. De-identified data were extracted from the provider’s clinical and care management system, which covers three main data sources: resident profile, daily medication administration, and medication incidents. Resident profile data contain demographic variables, such as age, sex, entry type (i.e., permanent or respite), entry date, departure date, resident status (active, discharged or deceased) and health conditions prevailing at admission (e.g., dementia status).

The medications administered to each resident on a daily basis were captured in the medication administration dataset. This dataset includes the date and time of the medication administered along with the drug name, administered status (administered or missed), drug status (‘as needed medication’ (‘Pro Re Nata’-PRN), packed or unpacked), and dosage information. The Anatomic Therapeutic Chemical (ATC) classification codes were used to extract relevant medication-related data [21].

Medication incidents document the date and time when these incidents occurred or were detected. For most incidents, this reflects the actual date and time of occurrence. However, certain incidents, such as ‘Durogesic Patch found missing,’ may indicate the date and time when the incident was detected. In the analysis, the study will use the ‘incident occurred date’ as reported in the medication incident dataset. The incident data also captured the incident type (categorical variable), description (free text), impact on the resident (categorical variable) and actions taken to avoid future incidents (free text variable).

Medication Incidents

Medication incidents are problems with medicines that could have or did cause unnecessary or unintended harm to a patient [5]. Staff are required to report any medication-related incidents and record them in the provider’s incident reporting system. The types of medication incidents may include adverse drug events, adverse drug reactions, and medication errors such as transcribing, dispensing, administration, monitoring and prescribing errors, along with near misses. The incident reporting system allowed the staff to categorise the types of incidents into 14 main categories (Supplementary Table 1). A significant proportion of the medication incidents were classified as ‘other’ (n = 2188, 38.3%), along with a free-text field for the incident description. Therefore, in this study, the authors with a clinical care background (C.M. and D.C.F) reviewed the free-text incident descriptions for incidents classified as ‘other’ and categorised all these medication incidents into one of six main categories on the basis of the stages of the medication management cycle and identified the incident types (i.e., for those classified as ‘other’) within each stage [22]. The six stages of the medication management cycle are (1) prescribing of medication, (2) supply of medication, (3) storage and accountability of medication, (4) administration of medication (including adverse drug reactions), (5) monitoring of response to medication, and (6) providing consumer advice and information related to the medication (refer to Supplementary Table 2 for definitions).

Outcome Measures

The study considered two medication incident-related outcome measures: (1) medication incident status (i.e., whether a resident experienced any medication incident during their stay–yes/no); and (2) the medication incident rate (IR) for all reported incidents and for potentially harmful incidents. Potentially harmful incidents were identified on the basis of the reported impact on the resident. Incidents where the resident was sent to hospital for treatment or monitoring, required GP attention or needed monitoring by staff were collectively classified as potentially harmful. The observed medication IR was calculated using the total number of medication incidents per 1000 resident days and was utilised to conduct trend analysis. Analyses were conducted for overall incidents and by stage of the medication management process. The expected medication IRs were derived from a multilevel Poisson regression model, as described below.

Statistical Analysis

Descriptive differences in the characteristics of residents with and without reported incidents were analysed. The baseline differences in categorical and continuous variables with respect to medication incidence status (i.e., yes/no) were evaluated using Pearson’s chi-squared test (or Fisher’s exact test) and the Wilcoxon rank sum test, respectively. Descriptive statistics were calculated for type of incident, time of the day, day of the week and impact on residents.

The observed IR per 1000 resident days was calculated as follows.

$$\text{Crude Incidence Rate}(\text{IR})=\frac{\text{Total number of medication incidents}}{\text{Total length of stays of residents in observation period}}\ast \text{1,000 days}$$

For the residents who were admitted to the same facility multiple times, the observed IR was calculated utilising the total number of medication incidents and the total number of resident days. The IRs are reported with 95% confidence intervals (CIs). Unadjusted rates of medication incidents for both overall incidents and incidents within different medication stages were calculated and compared between permanent versus respite residents. Local polynomial regression was performed to visualise trends in IRs for all reported medication incidents over time. Moreover, we applied this analysis to the three most frequently reported medication management stages: medication administration, medication supply and monitoring the response to medication.

Age at admission, entry type, sex, health status comprising 19 comorbidities and number of days with more than five medications were considered as risk factors of interest for outcome measures. More than five medications were selected using the 25th percentile of daily administered unique medication counts. To assess associations between the above risk factors and the reported number of medication incidents and potentially harmful incidents during their stay (i.e., IR), multilevel Poisson regression models were performed with the log form of length of stay as the offset. Facility- and resident-specific random intercepts were included in the model to address correlation within facility level and multiple resident admissions. Univariable and multivariable analyses were performed using the same model structure. The final multivariable model for each outcome was selected using stepwise elimination from the full model after including all variables without multicollinearity. The coefficients of the univariable and final multivariable models are reported as rate ratios (RRs) along with 95% confidence intervals [23]. Multicollinearity, overdispersion and zero inflation were tested using the variance inflation factor (VIF > 5), dispersion ratio and ratio between the predicted and observed zeros, respectively. To identify variation in expected medication IRs at the facility level (i.e., not explained by the fixed effects in the model), the standard deviations and exponential forms of facility-level random intercepts with 95% confidence intervals were extracted from the final multivariable model. P-value < 0.05 was assumed to be statistically significant. All the analyses were performed using R version 4.3.0 [24] in RStudio version 2022.7.2.576 [25].

Results

Resident Baseline Characteristics and Medication Incidence Status

The whole sample (n = 5709) had a median age of 86 (IQR = 81,91) years, with 63% (n = 3587) being females. A total of 5708 medication incidents were reported across 3,152,879 resident days for 2016 (35%) of the 5709 residents between 1 July 2014 and 31 August 2021 (Table 1). Of these 2016 residents, 1095 (54%) residents experienced more than one medication incident.

Table 1.

Characteristics of the residents and their medication incidence status

Variable	Overall, N = 5709a	Medication incident status		Univariable association with medication incident status
		No, N = 3693a	Yes, N = 2016a	p-valueb
Sex				< 0.001
Female	3587 (63%)	2237 (61%)	1350 (67%)
Male	2122 (37%)	1456 (39%)	666 (33%)
Age at admission (categories)				0.14
65–74	525 (9.2%)	319 (8.6%)	206 (10%)
75–84	1750 (31%)	1130 (31%)	620 (31%)
85–94	2899 (51%)	1883 (51%)	1016 (50%)
≥ 95	535 (9.4%)	361 (9.8%)	174 (8.6%)
Age at admission	86 (81, 91)	87 (81, 91)	86 (81, 91)	< 0.01
Total length of stay	308 (55, 889)	124 (32, 530)	783 (3,571,364)	< 0.01
Recurrent incident status
Residents with one reported incident			921 (46%)
Resident with > 1 reported incident			1095 (54%)
Health status
Circulatory disease, any	4863 (85%)	3070 (83%)	1793 (89%)	< 0.01
Cerebrovascular accident	1316 (23%)	775 (21%)	541 (27%)	< 0.01
Endocrine, any	1998 (35%)	1220 (33%)	778 (39%)	< 0.01
Diabetes	1397 (24%)	840 (23%)	557 (28%)	< 0.01
Thyroid disorder	601 (11%)	380 (10%)	221 (11%)	0.4
Chronic respiratory disease	992 (17%)	615 (17%)	377 (19%)	0.051
Neoplasms/cancer	1753 (31%)	1172 (32%)	581 (29%)	0.022
Dementia	2518 (44%)	1479 (40%)	1039 (52%)	< 0.01
Parkinson’s disease	334 (5.9%)	193 (5.2%)	141 (7.0%)	< 0.01
Depression, mood and affective disorders	2,010 (35%)	1088 (29%)	922 (46%)	< 0.01
Anxiety and stress-related disorders	1399 (25%)	720 (19%)	679 (34%)	< 0.01
Peptic Ulcer Disease (PUD) and (gastroesophageal reflux disease) GORD	1665 (29%)	1012 (27%)	653 (32%)	< 0.01
Renal disease	1020 (18%)	652 (18%)	368 (18%)	0.6
Arthritis	2929 (51%)	1754 (47%)	1175 (58%)	< 0.01
Osteoporosis	1457 (26%)	881 (24%)	576 (29%)	< 0.01
Gout	384 (6.7%)	233 (6.3%)	151 (7.5%)	0.089
History of fracture	1735 (30%)	1018 (28%)	717 (36%)	< 0.01
Hearing impairment	1289 (23%)	786 (21%)	503 (25%)	< 0.01
Visual impairment	870 (15%)	498 (13%)	372 (18%)	< 0.01
Number of selected comorbidities presented at admission (multimorbidity)	5 (4, 7)	5 (3, 6)	6 (5, 7)	< 0.01

^an (%); Median (IQR)

^bPearson’s chi-squared test; Wilcoxon rank sum test; Fisher’s exact test

Differences were observed in the characteristics of residents with and without reported incidents. Sex was associated with medication incident status (p < 0.01), with females being more likely to have a reported medication incident. The residents with medication incidents were slightly younger at admission (median age = 86 years (IQR = 81,91, p < 0.01), had a longer LOS (median = 783, IQR = 3,571,364 days, p < 0.01) and had multiple comorbidities (median = 6, IQR = 5,7, p < 0.01) than those in the non-medication incident cohort.

Medication management stages of the reported incidents

Figure 1 represents the ten most frequent incident types reported within each stage of the medication management cycle (refer to Supplementary Table 3 for the complete list of reported incident types). The majority of incidents reported were related to the administration of medication (n = 3023, 53%), followed by the supply of medications (n = 1546, 27%) and monitoring the response to the medication (n = 548, 9.6%). Half of the incidents within the administration stage involved the omission of medication (n = 1494, 49.4%). Dispensing or labelling errors (n = 974, 63%), delays or failures in the monitoring process (n = 481, 87.8%), incorrect counts of schedule 8 (S8) medications (n = 142, 55.9%) and transcription errors (n = 35, 29.4%) were the most frequent incident types within the supply stage, monitoring stage, storage, accountability stage and prescribing stage, respectively. Notably, documentation-related errors were reported, in the administration, supply, consumer advice and prescribing stages (Supplementary Table 3). Examples of the most common incident types are presented in Supplementary Table 4.

Fig. 1.

The ten most frequently reported incident types at various stages of the medication management process. ^aSchedule 8 (s8)/ restricted Schedule 4 (rs4) drugs classified as Drugs of Dependence and Addiction (dda)

The Impact of Incidents on Residents

The impact of the incident on residents was reported in 5165 (90%) incidents (Table 2). Overall, 724 (14%) incidents required the resident to be monitored by the hospital, general practitioner (GP) or staff (Supplementary Table 5). Among them, only 15 (0.3%) incidents required the resident to be hospitalised for treatment or monitoring. The majority of these incidents were due to medication administration (n = 13), with two incidents reported as being due to supply issues. In total, 2.2% (n = 115) of incidents required GP attention, and 11% (n = 594) required residents to be monitored by staff.

Table 2.

Impact of incidents on residents

Characteristic	Overall, N = 5708a	Administration of medication, N = 3023a	Supply of medication, N = 1546a	Monitoring of response to medication, N = 548a	Storage and accountability of medication, N = 254a	Consumer advice and information on medication, N = 220a	Prescribing of medication, N = 117a
Resident sent to hospital for treatment/monitoring	15 (0.3%)	13 (0.5%)	2 (0.1%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)
Resident required GP attention	115 (2.2%)	76 (2.8%)	23 (1.6%)	9 (1.7%)	1 (0.4%)	3 (1.5%)	3 (2.7%)
Resident required monitoring by staff	594 (11%)	453 (17%)	74 (5.2%)	49 (9.5%)	5 (2.1%)	9 (4.6%)	4 (3.6%)
No health impact on resident	4190 (81%)	2007 (75%)	1261 (88%)	436 (84%)	221 (94%)	178 (92%)	87 (79%)
Other	251 (4.9%)	136 (5.1%)	64 (4.5%)	23 (4.4%)	8 (3.4%)	4 (2.1%)	16 (15%)
Unknown	543	338	122	31	19	26	7

^an (%)

Of the medication administration incidents 18% ((13+76+453)/3023) required residents to be monitored by hospital, GP or staff. Of these, 78% (n = 424/542) of incidents were owing to the omission of medication (n = 218), incorrect resident identification (n = 70), incorrect medication administration (n = 70) and incorrect dose (n = 66) (Supplementary Table 5). From the supply stage incidents, pharmacy dispensing/labelling errors (n = 49), failure to order/maintain a stock of medication (n = 27) and delayed or not dispensed medications (n = 14) accounted for 90% (90/99) of the supply incidents that required extra monitoring of residents.

Temporal Patterns in Reported Incidents

As shown in Supplementary Table 6, a greater percentage of daily average medication incidents occurred/were detected on weekdays (15–16%) than on weekends (11–12%), and nearly half of the incidents reported occurred during the morning (44%), followed by afternoon (33%), evening (20%) and night (3.1%). For monitoring-related incidents, the percentage of incidents occurring in the morning (57%) was relatively high compared with that of other incidents (30–47%). However, for storage-related incidents and prescribing-related incidents, the majority occurred during the afternoon.

Trends in the Medication Incident Reporting Rates

During the entire study period, the overall medication IR was 1.81 per 1000 resident days (95% CI 1.76, 1.86). Figure 2 presents the quarterly medication IRs (per 1000 resident days) for all incidents and the three most frequently reported medication stages in our study sample (i.e., administration of medication, supply of medication and monitoring of response to medication).

Fig. 2.

Quarterly reported medication IR for all incidents and the three most common incident types. ^aIncident occurred date has been considered

The data show a notable initial decline in observed IRs relevant to all medication incidents, medication administration incidents and medication supply incidents. Specifically, these observed IRs decreased from 4.1, 2.1 and 1.4 per 1000 resident days to 1.5, 1.1 and 0.3 per 1000 resident days, respectively, between the third quarter of 2014 and the third quarter of 2016. Subsequently, the observed IR relevant to all medication incidents increased, reaching approximately two incidents per 1000 resident days, and then stabilised over the rest of the period after the third quarter of 2018.

Factors Associated with the Rate of Medication Incidents

Univariable expected rate ratios identified by the multilevel Poisson model for medication incidents are included in Supplementary Table 7. Table 3 shows the risk factors retained in the final multivariable model. Multimorbidity was excluded from the full model because it was found to have multicollinearity (VIF > 5) with other comorbidities.

Table 3.

Multilevel multivariable Poisson regression model when considering all medication incidents

Variable	Rate ratio	Standard error	95% CI	p-value
Demographics
Age at admission	0.992	0.003	(0.985, 0.998)	0.014
Entry type: Respite	1.908	0.075	(1.646, 2.211)	< 0.01
Comorbidities
Endocrine, any	1.184	0.049	(1.075, 1.305)	< 0.01
Parkinson’s disease	1.586	0.094	(1.318, 1.908)	< 0.01
Cerebrovascular accident	1.122	0.054	(1.009, 1.248)	0.033
Anxiety and stress-related disorders	1.161	0.052	(1.049, 1.284)	< 0.01
PUD and GORDa	1.133	0.051	(1.025, 1.252)	0.014
Chronic respiratory disease	1.141	0.061	(1.012, 1.286)	0.031
Dementia	0.885	0.049	(0.804, 0.973)	0.012
Medications
Number of days with more than 5 medications	2.019	0.014	(1.930, 2.111)	< 0.01
Random interceptsb	Standard deviation
Facility	0.868		(0.653, 1.154)
Resident	0.928		(0.878, 0.981)

^aPUD & GORD-Peptic ulcer disease (PUD) and gastroesophageal reflux disease (GORD)

^bRandom intercepts were applied for facilities and residents, and the log of stay was adopted as the offset

All risk factors retained in the final model increased expected medication IR except for age at admission and dementia. On average, the expected rate of medication incidents was approximately two times (RR = 1.908, 95% CI 1.646, 2.211) greater for respite residents than for permanent residents (p < 0.01). Similarly, the observed IR for respite residents was greater than that for permanent residents in all medication management stages except for incidents related to monitoring responses to medications (Supplementary Table 8). Residents with Parkinson’s disease had an IR more than 1.5 times (RR = 1.586, 95% CI 1.318, 1.908) higher than residents without Parkinson’s disease. Administering more than five medications increased the expected rate of reported medication incidents by more than two times (RR=2.019, 95% CI 1.930, 2.111) compared with less than five medications. Interestingly, the medication IR was 12% lower for residents with dementia (p = 0.012) and 0.8% lower for each 1-year increase in age at admission (p = 0.014).

Factors Associated with the Rate of Potentially Impactful Medication Incidents

Similar to the pattern observed for all medication incidents, the expected rate of incidents was approximately twice as high for respite residents compared with permanent residents (RR = 2.158, 95% CI: 1.513–3.079; p < 0.01) (Table 4). In addition, residents with a history of cerebrovascular accident and those taking more than five medications were found to have a higher risk of experiencing potentially harmful medication incidents. Residents with Parkinson’s disease had an incident rate more than twice as high as those without the condition (RR = 2.166, 95% CI: 1.590–2.950), which represents a greater relative risk compared with the analysis of all medication incidents.

Table 4.

Multilevel multivariable Poisson regression model when considering potentially impactful medication incident

Variable	Rate ratio	Standard error	95% CI	p-value
Demographics
Entry type: respite	2.158	0.181	(1.513, 3.079)	< 0.01
Comorbidities
Parkinson’s disease	2.166	0.158	(1.590, 2.950)	< 0.01
Cerebrovascular accident	1.300	0.102	(1.065, 1.588)	0.01
Medications
Number of days with more than 5 medications	1.130	0.0289	(1.068, 1.196)	< 0.01
Random interceptsa	Standard deviation
Facility	0.455		(0.298, 0.693)
Resident	1.050		(0.892, 1.235)

^aRandom intercepts were applied for facilities and residents, and the log of stay was adopted as the offset.

In both analyses, there was considerable variation at the resident and facility levels for medication IRs, as denoted by the standard deviation (Tables 3, 4) of the multivariable model. Facility level variability extracted from the random effects of the multivariable model (for all medication incidents) is further visualised aged care planning region wise (de-identified) in Supplementary Fig. 2. The facility level information is reported in Supplementary Table 9.

Discussion

Our findings confirm the high frequency of medication problems in residential aged care, with 35% of residents having a reported medication incident during their stay and 50% of these residents having repeat incidents. Only 0.3% of medication incidents led to hospitalisation, but 14% required additional monitoring. The majority (53%) of medication incidents occurred during the administration stage. Several risk factors were identified. Compared with male residents, female residents had twice as many incident reports, and those with multimorbidity and polypharmacy were also significantly more likely to have an incident. In particular, comorbidities such as endocrine disorders, Parkinson’s disease, cerebrovascular accidents, anxiety and stress-related disorders, PUD and GORD and chronic respiratory disease increased the likelihood of reporting medication IRs. Interestingly, residents with dementia and younger residents at admission showed a lower expected medication IR than non-dementia residents and older residents, respectively. However, it should be noted that the upper confidence intervals of these rate ratios were close to one, which also suggested that little difference was expected between these groups. Importantly, the expected rate of medication incidents for respite residents was twice as high as that for permanent residents.

To the best of our knowledge, this is the first study to show that respite residents have a greater likelihood of medication incidents including potentially impactful incidents. This may be owing to staff’s limited knowledge about respite residents, including their medication management requirements [26]. Moreover, the unplanned nature of admission and communication failures at transition [27] could also increase medication incidents and adverse events. This could be overcome by implementing and promoting regular respite care models, standardising care transition procedures, especially for medication, and developing facilities specialising in respite care [28]. Although a univariable association of sex was found with medication incidence status, there was no association of sex with reported IRs when adjusted for other risk factors. This finding aligns with discordant results found in multiple studies [20]. Furthermore, the lower expected medication IR for dementia residents might be because of close monitoring with specialised care protocols, limited self-administration of medication and simplified medication regimens to reduce complexity and enhance medication administration.

Omitted medication, incorrect resident identification and incorrect medication or dose administration in the administration stage, along with pharmacy dispensing/labelling errors, failure to order/maintain stock and delayed or not dispensed medications in the supplying stage, had a greater impact on residents (i.e., required the resident to be monitored by hospital, GP or staff). Most of these errors could be minimised by enhanced communication strategies facilitated by electronic prescribing (e.g., Electronic National Residential Medication Charts (eNRMCs)), which have the potential to deliver timely and accurate information to all stakeholders. This also implies that streamlining communication channels between stakeholders within and across medication management stages is important for reducing avoidable medication incidents [11]. Importantly, a low proportion (0.3%) of incidents required the resident to be transferred to the hospital and reported an adverse reaction (< 10) in our study sample. This finding is in line with other published literature [20]. The occurrence of serious effects resulting from medication errors was remarkably low, being reported in only a small fraction of cases (ranging from 0 to 1% of all medication errors), and fatalities were exceptionally rare [20]. It is unclear whether medication errors leading to severe consequences are genuinely rare in RAC facilities or if they are underreported due to challenges in detection [20]. Therefore, obtaining clarity on this matter is crucial for developing safer healthcare systems.

Our results showed considerable facility variations in the reported IRs. The greater variability of reported medication incidents in facilities might be owing to other potential contributing factors that are not explained by the measured risk factors in the multivariable model. Some of these factors could include (1) insufficiencies in qualified or well-trained onsite aged care staff, (2) inconsistencies in medication incident reporting practices, (3) a lack of robust quality assurance and monitoring programs to maintain best practices in medication management, (4) geographical locations and environmental factors and (5) greater staff workloads and stress levels. Knowing these variations across facilities is important for effective interventions and for avoiding underreporting.

The study revealed an initial increase in reported medication incident rates in early 2014 followed by a gradual decline overtime. This trend may be explained by the implementation of electronic incident reporting systems, electronic medication management systems and medication incident reporting tools introduced in these facilities in early 2014. The initial rise in incidents reports likely reflects improved reporting capabilities, while the subsequent decline suggests that the electronic medication administration system may have contributed to a reduction in medication administration and supply errors. It is also noted that major aged care reforms, such as the establishment of the Royal Commission into Aged Care Quality and Safety, the Aged Care Quality Standards (Quality Standards) program, and the eNRMC, along with COVID-19, occurred during or after the fourth quarter of 2018. These reforms have initiated quality indicator programs where RAC services are required to collect data and report on different indicators, including medication management and hospitalisation. Although the study has not evaluated the direct impact of these reforms on medication incidents, we observed a stabilisation of observed IRs during this time. This may be owing to the development of a learning culture from these reforms and reported incidents. Thus, a decrease in medication incidence over time may be a positive indication that improving the monitoring and reporting of medication problems is occurring. However, further research is required to confirm this hypothesis.

Implications for Policy, Practice and Research

The study’s findings have significant implications for both RAC practices and policies. The elevated rates of medication incidents among respite residents underscore the critical need for targeted attention and additional investments in this population. Moreover, the results of the study suggest that closer supervision is required for residents who have had a medication incident, as medication incidents reoccurred in half of them. While the study does not analyse the effectiveness of the strategies employed to mitigate future incidents, the observed repeated incidents also raise questions regarding the efficacy of the strategies implemented to prevent future incidents. Staff might be more vigilant when their residents have comorbidities and/or are taking multiple medications.

Furthermore, the study revealed a significant proportion of the medication incidents classified as ‘other’ (n = 2188, 38.3%) along with a free-text field for the incident description. These incident reports hinder the quantitative analysis of trends and measures to identify risk areas that require improvement. This highlights the importance of having a well-structured classification system. An efficient computer-based incident management system for medication incidents [17, 29] should be able to systematically collect hierarchical data related to these incidents. Some studies have suggested aggregated hierarchical classification systems for drug-related problems covering medication incidents [30]. The inclusion of different levels of structured information in a hierarchical format can provide more in-depth insights related to incidents. These hierarchical classifications can also empower staff to report incidents in real time [17] while capturing details related to the stage of medication management, specific incident type within that stage and additional information such as stakeholder causality, involved medications and closely related contributing factors. These classifications enable real-time monitoring of IRs and facility-level variations, facilitating targeted interventions to effectively prevent future occurrences. It should be noted that this does not underestimate the importance of the free-text field. Most importantly, these systems should undertake regular revisions [31], allowing continuous improvements [17] and enabling users to include new incident types within different stages of medication management. Therefore, a national standard should be implemented at RAC facilities on incident classifications and on what fields to be recorded when a medication incident occurs.

Limitations

Analysing reported medication incidents in the RAC setting is crucial for the wellbeing and safety of residents. However, comparisons between rates identified through direct observation and reported incidents reveal that those obtained through reporting tend to underestimate the actual occurrences of such incidents [32]. In such instances, the distribution of reported medication incidents may not accurately reflect the true distribution of all medication incidents. For example, in our study, only 2% of incidents were related to medication prescribing. Given this low prevalence over more than 3 million resident days, it may indicate that important prescribing-related incidents were either not recorded or missed or were managed directly by the care team without formal documentation. This underreporting poses a challenge for preventing similar future medication incidents [32]. Moreover, these medication incidents are classified by RACF staff, and there may be instances where the classifications are subject to bias in identifying the appropriate categories. Nevertheless, a thorough analysis of these incidents can identify the risk areas to reduce more common and less harmful systematic problems associated with near misses, which are mostly reported [22]. Therefore, the study includes a multivariable analysis of all reported medication incidents, as well as a secondary analysis of potentially harmful incidents, identified on the basis of their reported impact on residents. Although such incidents are more likely to be underreported, a multilevel Poisson regression would be able to identify possible risk factors, particularly for potentially harmful medication incidents. This could lead to more targeted interventions for those risk factors which could be implemented to improve medication management.

Conclusions

The administration, supply and monitoring phases are the most commonly reported medication management phases in RAC facilities. Being a respite resident along with multimorbidity and having numerous medications increase the rate of medication incidents. Considerable facility-level variation is also observed at different locations. Identifying these risk areas can address more common and less harmful systematic problems associated with near misses, which are more likely to be reported.

Supplementary Information

Below is the link to the electronic supplementary material.

Declarations

Funding

Open Access funding enabled and organized by CAUL and its Member Institutions. The study was part of a project funded by the National Health and Medical Research Council (NHMRC) Partnership Project Grant in partnership with Anglicare (1170898). The funding organisation did not have any influence on the study design, data collection, analysis, and interpretation as well as the preparation, review or approval of the manuscript for publication.

Ethics Approval

The study received ethics approval from the Macquarie University Human Research Ethics Committee (reference no. 52019614412614). All methods and analysis in this study were carried out in accordance with the principles of the Declaration of Helsinki.

Consent to participate/consent for publication

The study received ethical approval from Macquarie University Human Research Ethics Committee (No. 52019614412614). Since this is a retrospective cohort study utilising, de-identified, existing routinely collected aged care data, a waiver of informed consent was granted by Macquarie University Human Research Ethics Committee following the Australian National Statement on Ethical Conduct in Human Research section 3.3.14 (a) as the study used previously collected de-identified data.

Consent for Publication

Not applicable.

Code Availability

As the analysis includes free-text fields, the statistical R code cannot be publicly shared. However, it can be provided upon request.

Data Availability

The data that support the findings of this study are not openly available owing to reasons of sensitivity and may include some stakeholder details as in free texts. The data will be shared on reasonable request to the corresponding author.

Consent for Publication

Not applicable.

Competing Interest

The authors declare that they have no competing interests.

Author Contributions

The research was designed by S.S.M.S., N.W., and M.Z.R. The analysis was performed by S.S.M.S. The free-text incidents descriptions were coded by C.M., and D.C.F. The manuscript was drafted by S.S.M.S., N.W., M.Z.R., A.D.N., G.H., Y.X. and J.I.W. The manuscript draft and analysis were reviewed by J.I.W., M.Z.R. and N.W. All authors have read and approved the final version.