Medication Discrepancies at Hospital Discharge Among Adults with and Without Mental Health Conditions: A Retrospective Cohort Study
1
Department of Medicine, Faculty of Medicine and Health Sciences, McGill University, Montréal, QC H4A 3J1, Canada
2
Clinical & Health Informatics Research Group, McGill University, Montréal, QC H3A 1G1, Canada
3
McGill University Health Centre (MUHC), Montréal, QC H4A 3J1, Canada
4
Department of Epidemiology, Biostatistics and Occupational Health, School of Population and Global Health, Faculty of Medicine and Health Sciences, McGill University, Montréal, QC H3A 1G1, Canada
*
Author to whom correspondence should be addressed.
Pharmacoepidemiology 2026, 5(2), 17; https://doi.org/10.3390/pharma5020017
Submission received: 25 February 2026
/
Revised: 25 May 2026
/
Accepted: 28 May 2026
/
Published: 4 June 2026
Abstract
Background/Objectives: Medication discrepancies at hospital discharge are common and may contribute to adverse drug events and avoidable healthcare use. Patients with mental health conditions may be at increased risk because of greater clinical complexity, polypharmacy, and fragmented care, but comparative evidence during general hospital admissions is limited. Our primary objective was to determine whether adults with mental health conditions were more likely than those without such conditions to experience unintended medication discrepancies at hospital discharge. Secondary objectives were to examine discrepancy subtypes, assess whether associations differed for serious mental illness versus other mental health conditions, and explore whether associations varied by reconciliation arm. Methods: We conducted a retrospective cohort study using linked data from the RightRx cluster-randomized trial at the McGill University Health Centre (2014–2016) and Quebec administrative databases. Adults with continuous provincial drug coverage for at least 12 months before admission who met study eligibility criteria were included. The primary exposure was any documented mental health condition; secondary analyses distinguished serious mental illness (SMI) from other mental health conditions. The primary outcome was any unintended medication discrepancy at discharge; subtype analyses examined omissions, therapeutic duplications, and unintended dose changes. Results: Among 3567 patients, 877 (24.6%) had a mental health condition. Crude discrepancy prevalence was similar between groups. In the prespecified primary analysis, mental health condition status was associated with lower observed odds of any unintended discrepancy at discharge. This unexpected inverse association should not be interpreted as evidence of a protective effect and may reflect differences in documentation, residual confounding, selection, or other unmeasured processes. Secondary and supplementary analyses, including omission and SMI subgroup comparisons, did not remain statistically significant after Holm correction. Conclusions: These findings suggest that documentation-based discrepancy measures may relate to mental health status in heterogeneous ways, but they require confirmation in independent settings.
1. Introduction
Medication discrepancies at hospital discharge are common and clinically important because transitions of care are periods in which medication regimens often change and communication failures may occur. Unintended discrepancies, including omissions, duplications, and dose changes without documented clinical rationale, have been associated with adverse drug events and post-discharge healthcare use, including emergency visits and readmissions. Recent studies continue to show that medication-related harm and readmission risk remain substantial after discharge, especially in patients with comorbidity, polypharmacy, and medication changes [1,2].
Medication reconciliation is a central strategy for reducing discharge discrepancies, but its effectiveness depends on how it is implemented. Electronic reconciliation tools and structured reconciliation programs can reduce discrepancies, particularly omissions, by improving access to community medication information and supporting workflow standardization. However, improvements in discrepancy rates do not always translate into lower adverse drug event rates or readmissions, suggesting that broader transitional care interventions, including pharmacist involvement and post-discharge follow-up, may still be required to improve clinical outcomes [3,4].
Patients with mental health conditions may be especially vulnerable to medication discrepancies because of complex regimens, multiple prescribers, communication barriers, and fragmented care across psychiatric and general medical settings. Prior studies in mental health hospitals and outpatient psychiatric settings have documented high rates of prescribing, clerical, and communication errors, as well as frequent clinically relevant omissions in medication lists [5,6]. However, direct comparisons of discharge discrepancies during general hospital admissions between patients with and without mental health conditions remain limited.
To address this gap, we conducted a retrospective cohort study using linked data from the RightRx study, a cluster-randomized trial of electronic versus usual-care medication reconciliation at a large academic health centre [7]. Our primary objective was to determine whether adults with mental health conditions were more likely than those without such conditions to experience unintended medication discrepancies at hospital discharge. Secondary objectives were to examine discrepancy subtypes, assess whether associations differed for serious mental illness versus other mental health conditions, and explore whether associations varied by reconciliation arm.
2. Results
2.1. Study Population and Baseline Characteristics
The analytic cohort included 3567 patients, of whom 877 (24.6%) met the study definition of a mental health condition. Patients with mental health conditions were slightly younger on average and included a higher proportion of women than patients without mental health conditions. They also had greater baseline psychiatric medication exposure and greater clinical complexity across several measured domains. Baseline characteristics are summarized in Table 1.
Among patients with any mental health condition (n = 877), 346 (39.5%) met criteria for serious mental illness (SMI) and 531 (60.5%) had other mental health conditions (Table 2). Within this subgroup, patients with SMI were older (mean 73.47 vs. 65.70 years), had a higher comorbidity burden (mean Elixhauser count 3.0 vs. 2.6), more baseline ED use (66.2% vs. 57.8%), and a longer mean length of stay (20.08 vs. 10.51 days).
2.2. Diagnostic Composition of the Mental Health Subgroup
Within the mental health group, diagnostic categories were not mutually exclusive. The most frequently observed categories were anxiety disorders (31.6%), delirium/confusion (30.9%), depressive disorders (22.0%), psychotic disorders (20.1%), and addictive disorders (14.8%). Approximately one third of patients in the mental health group had two or more distinct mental health diagnosis categories (32.6%). A concise summary is provided in Table 3, and the full breakdown remains available in Table S18.
2.3. Descriptive Prevalence of Medication Discrepancies at Discharge
Crude discrepancy prevalence at discharge was similar between patients with and without mental health conditions. Any unintended discrepancy occurred in 42.1% of patients with mental health conditions and 40.9% of those without. Omissions were also similar between groups (25.3% vs. 26.0%). Therapeutic duplication was observed in 6.8% of patients with mental health conditions and 6.1% of those without, whereas unintended dose changes were somewhat more frequent in the mental health group (23.5% vs. 20.0%). These descriptive results are shown in Table 4; the corresponding 3-level mental health breakdown is shown in Table S1.
In the 3-level descriptive analysis, the prevalence of any discrepancy was 40.9% in patients with no mental health condition, 44.3% in those with other mental health conditions, and 35.1% in those with serious mental illness. Similar patterns were seen for omissions (26.0%, 26.4%, and 21.8%, respectively), while therapeutic duplications were less frequent than omissions and dose changes across all groups. These results are detailed in Table S1.
2.4. Multivariable Associations with Unintended Discrepancies at Discharge
In multivariable logistic regression models adjusting for demographics, comorbidity, baseline healthcare use, baseline medication burden and care fragmentation, trial arm, baseline high-risk medication exposure, and discharge medication count, mental health condition status was associated with lower observed odds of any unintended discrepancy at discharge (aOR 0.81, 95% CI 0.67–0.98; p = 0.027), which was the prespecified primary confirmatory analysis. The corresponding subtype-specific analyses were secondary; for omission, the point estimate was also lower (aOR 0.77, 95% CI 0.62–0.95; raw p = 0.016), but this association did not remain statistically significant after Holm correction. There was no statistically significant association between binary mental health status and therapeutic duplication or unintended dose change. These adjusted estimates are summarized in Table 5 and detailed in Tables S2–S5.
Within the mental health subgroup, the point estimate for serious mental illness (SMI) suggested lower odds of any unintended discrepancy than for other mental health conditions (aOR 0.55, 95% CI 0.37–0.82; raw model p = 0.003), but this association did not remain statistically significant after Holm correction. Similarly, omission, therapeutic duplication, and unintended dose change did not remain statistically significant after multiplicity adjustment. These subgroup findings should therefore not be given major interpretive weight. These results are summarized in Table 5 and reported in detail in Tables S2–S5.
In sensitivity analyses applying the Holm correction to the family of secondary and supplementary inferential models, no secondary or supplementary association remained statistically significant after multiplicity adjustment. This included subtype-specific models, interaction models, arm-stratified analyses, and SMI subgroup analyses. Full raw and Holm-adjusted p-values are presented in Table S21.
Across models, greater care fragmentation, reflected by higher numbers of distinct prescribers and pharmacies, and a higher number of discharge medications were consistently associated with greater odds of discrepancies. The intervention arm also showed substantially lower discrepancy odds than the control arm across the adjusted models for the main outcome and related analyses. Full covariate estimates are reported in Tables S2–S5.
2.5. Effect Modification by Trial Arm
Interaction testing did not provide robust evidence of effect modification. For omission, the raw interaction p-value was nominally significant, but this association did not remain statistically significant after Holm correction. Interaction tests for the other discrepancy subtypes were also not statistically significant after multiplicity adjustment. Results are reported in Tables S6–S9.
In arm-stratified analyses, point estimates varied by trial arm, but none of these stratified associations remained statistically significant after Holm correction. These analyses should therefore not be given major interpretive weight. These stratified estimates are shown in Tables S10–S13.
2.6. Discrepancy Burden and Supplementary 3-Level Mental Health Analyses
To quantify discrepancy burden, we modeled a discrepancy count ranging from 0 to 3 using negative binomial regression. In this analysis, binary mental health status was not materially associated with discrepancy count. Although the point estimate for SMI suggested lower discrepancy burden than for other mental health conditions, this finding did not remain statistically significant after Holm correction. These findings should therefore not be given major interpretive weight and are reported in Tables S14 and S15.
In the supplementary 3-level mental health exposure model for the primary outcome, point estimates suggested that serious mental illness (SMI) may differ from both other mental health conditions and no mental health condition more than the broad binary mental health classification suggests. However, because this analysis was supplementary and secondary/supplementary findings were not robust after multiplicity adjustment, these results should not be given major interpretive weight. The operational definition of this hierarchical 3-level exposure is detailed in Table S17, and the corresponding adjusted model is shown in Table S16.
3. Discussion
In this retrospective cohort study of adults discharged from general medical and surgical units, crude discrepancy prevalence was similar between patients with and without mental health conditions despite greater baseline clinical and medication complexity in the mental health group. After adjustment, the prespecified primary analysis showed a lower observed inverse association between mental health condition status and any unintended medication discrepancy at discharge. However, this unexpected result should not be interpreted as evidence that mental health conditions are protective against discharge discrepancy risk. Rather, it should be interpreted cautiously because it was contrary to the a priori expectation generated by prior literature describing high rates of prescribing, clerical, and communication errors in psychiatric settings and may reflect documentation-related processes, residual confounding, selection, or other unmeasured factors rather than lower true vulnerability [2,8].
This cautious interpretation is important because prior work has generally suggested that patients with mental health conditions, especially those with serious mental illness, experience greater multimorbidity, more fragmented care, and higher nonpsychiatric health service utilization, all of which would ordinarily be expected to increase transition-related risk [8]. At the same time, discharge medication discrepancies remain clinically important because they have been linked to post-discharge emergency visits and to medication-related readmissions, a meaningful proportion of which may be preventable [8]. Thus, the inverse association observed here should be viewed as unexpected and provisional, not as evidence that patients with mental health conditions are somehow safer at discharge.
Importantly, no secondary or supplementary analyses remained statistically significant after Holm correction. Accordingly, subtype-specific, subgroup, interaction, arm-stratified, discrepancy-count, and 3-level exposure findings should not carry major interpretive weight in the present study. Those analyses remain useful for transparency and sensitivity assessment, but they do not provide robust inferential support beyond the prespecified primary analysis. In particular, the corrected analyses do not support a stable claim that omission risk, serious mental illness subgroup status, or trial-arm interaction meaningfully explains the observed primary inverse association.
The practical contribution of this study may therefore lie less in identifying a clinically actionable “lower-risk” mental health subgroup and more in showing that documentation-based discrepancy measures may not map straightforwardly onto underlying clinical vulnerability. In this dataset, patients with mental health conditions appeared more clinically complex at baseline, yet did not show higher observed discrepancy prevalence. One implication is that medication discrepancy measures derived from documented lists may capture only one dimension of transition quality, while broader medication-related harm may depend on other mechanisms not fully reflected in list concordance alone. This interpretation is consistent with evidence showing that medication reconciliation interventions can reduce discrepancies as process outcomes without consistently reducing adverse drug events, emergency department visits, or readmissions [3,4].
Several covariate patterns were more consistent with expected medication-safety mechanisms than the primary mental health contrast itself. Greater care fragmentation, reflected by more distinct prescribers and pharmacies, and higher discharge medication counts were consistently associated with higher discrepancy risk or burden. These findings support the interpretation that system-level coordination problems and regimen complexity remain central drivers of discharge discrepancies across patient groups, a conclusion that is consistent with broader transition-of-care literature emphasizing the importance of coordinated reconciliation workflows and implementation quality [4]. For clinicians and quality-improvement teams, this means that the most actionable signals from the study remain the familiar ones: improve the reliability of medication reconciliation, reduce fragmentation across settings, and pay particular attention to patients with high medication burden and multiple care interfaces. Recent evidence continues to support the value of structured reconciliation programs and medication review interventions for reducing discrepancies and, in some settings, lowering readmission risk [9].
The intervention context also helps frame these results. This study was nested within the RightRx trial, in which electronic medication reconciliation reduced medication discrepancies but did not significantly reduce adverse drug events or downstream utilization outcomes [3]. Our findings are consistent with that distinction. In the present analysis, the intervention arm was associated with lower discrepancy risk and burden, but the multiplicity-adjusted secondary analyses did not support strong evidence that the intervention’s association with the primary outcome differed meaningfully by mental health status. This reinforces the interpretation that reconciliation processes may improve documentation and list accuracy without necessarily resolving the broader determinants of medication-related harm.
Taken together, the study contributes in three ways. First, it shows that an unexpected inverse association between mental health status and a documentation-based discrepancy outcome can arise even in a cohort where patients with mental health conditions have greater baseline complexity. Second, it shows that these unexpected signals weaken considerably once multiplicity is handled rigorously. Third, it suggests that for clinicians and quality-improvement audiences, the more stable targets remain care fragmentation, medication burden, and reconciliation process quality, rather than broad assumptions that mental health status alone defines either a uniformly higher-risk or lower-risk subgroup.
4. Limitations
This study has several limitations. First, mental health status was based on documented diagnostic codes and may not fully capture undiagnosed conditions, illness severity, symptom burden, or outpatient psychiatric care not reflected in the available data. Second, discrepancy definitions were operationalized using the documentation available in the linked data sources. Accordingly, the absence of a documented clinical rationale does not necessarily mean that a medication change was clinically inappropriate; some omissions or dose changes may have been intentional but incompletely documented. Third, although the validated community medication list strengthened ascertainment, the study still depended on routine clinical and administrative data, which are subject to coding and documentation limitations. Fourth, the primary analysis was prespecified around the binary mental health exposure and the any-discrepancy outcome, whereas subtype, interaction, count, and 3-level exposure analyses were secondary or supplementary; these additional analyses should therefore not be given major interpretive weight. Finally, this was a study conducted within a single academic health network and within the context of the RightRx trial, which may limit generalizability to other healthcare systems and reconciliation workflows. In addition, the unexpected inverse association observed in the prespecified primary analysis may reflect differences in documentation practices, residual confounding, selection processes, or other unmeasured factors rather than a true reduction in underlying discrepancy risk among patients with mental health conditions; this finding therefore requires confirmation in independent datasets and clinical settings. Although the primary analysis was prespecified, the large number of secondary and supplementary comparisons increases the possibility of false-positive findings. In sensitivity analyses using Holm correction, none of the secondary or supplementary associations remained statistically significant; therefore, these findings should not be given major interpretive weight.
5. Materials and Methods
5.1. Study Design and Setting
We conducted a retrospective cohort study using linked data from the RightRx cluster-randomized trial at the McGill University Health Centre (MUHC), Montréal, Québec, Canada, between October 2014 and November 2016. The MUHC is a large tertiary academic health network. RightRx enrolled adult patients admitted to the study units represented in this analytic cohort, including internal medicine and cardiac surgery at the Royal Victoria Hospital and the Montreal General Hospital.
RightRx compared usual-care discharge medication reconciliation with an electronic medication reconciliation intervention integrated into hospital workflow. In the intervention arm, clinicians had access to electronically retrieved community medication information and a structured reconciliation process to support review of admission and discharge medications. In the control arm, medication reconciliation was performed according to usual care at the participating units, without the full electronic reconciliation workflow used in the intervention arm. In the parent trial, the intervention reduced medication discrepancies but did not significantly reduce adverse drug events, emergency department visits, or readmissions [3].
Québec’s healthcare system is publicly funded. Medical services are covered by the provincial health insurance plan, the Régie de l’assurance maladie du Québec (RAMQ), and prescription drug coverage is provided through a combination of public and private insurance, with RAMQ drug coverage applying primarily to older adults, individuals receiving social assistance, and those without private or employer-sponsored plans.
5.2. Data Sources
Patient-level information was retrieved from two main linked data sources. The provincial health insurance agency (RAMQ) administrative databases included the beneficiary file (date of birth, sex, and drug coverage status), the outpatient pharmacy claims database (drugs dispensed including dispensation date, quantity, strength, and pharmacy/prescriber identifiers), the physician claims database (medical services including location, date, and type of service), and the hospitalization database (MED-ÉCHO; dates of service, diagnostic codes, procedures, and admission/discharge details). The RightRx trial database included the validated community medication list compiled at admission and the hospital discharge prescription generated for each eligible admission [7] (Figure 1).
For each patient, the linked dataset provided information on mental health diagnoses, comorbidities, healthcare utilization, medication exposures before admission, and the complete set of community and discharge medications required to construct discrepancy measures.
5.3. Study Population
The analytic cohort included adult patients (aged ≥ 18 years) enrolled in RightRx who had continuous RAMQ drug coverage for at least 12 months before the index admission, to permit ascertainment of pre-admission outpatient medication use, and who had the validated community medication list and discharge prescription required to assess discrepancies. Admissions missing the validated community medication list or the discharge prescription required to assess discrepancies were excluded. The final cohort derivation is summarized in Figure 1.
5.4. Exposure: Mental Health Conditions
The primary exposure was the presence versus absence of any mental health condition. Mental health conditions included mood disorders, anxiety disorders, psychotic disorders, trauma-related disorders, obsessive–compulsive disorder, eating disorders, neurodevelopmental disorders, addictive disorders, personality disorders, neurotic and somatoform disorders, adjustment disorders, miscellaneous psychiatric conditions, suicidal ideation or attempts, and delirium/confusion, as defined using ICD-9 and ICD-10 diagnostic codes. A patient was classified as having a mental health condition if at least one relevant diagnosis code was recorded in RAMQ physician claims or hospital discharge data during the period from 12 months before admission through 7 days after discharge. Patients without any qualifying code during that window were classified as having no mental health condition.
For secondary analyses, we used a hierarchical 3-level mental health exposure: no mental health condition, other mental health condition, and serious mental illness (SMI). SMI was defined as the presence of a psychotic spectrum disorder or bipolar disorder, with SMI taking precedence over other diagnoses. Patients with at least one qualifying mental health diagnosis who did not meet SMI criteria were classified as having other mental health conditions (Table S17). This hierarchical approach is consistent with prior work distinguishing serious mental illness from other psychiatric disorders [10].
5.5. Outcome Measures
The primary outcome was the presence of at least one unintended medication discrepancy at hospital discharge. Discrepancies were identified by comparing the validated community medication list compiled at admission with the discharge prescription recorded in the RightRx database. We examined three discrepancy subtypes:
Omissions: Community medications absent from the discharge prescription without a documented clinical rationale in the available discharge documentation.
Therapeutic duplications: A newly prescribed discharge medication from the same therapeutic class as a pre-admission medication, without documentation indicating intended therapeutic overlap.
Unintended dose changes: Discharge prescriptions in which the prescribed daily dose differed by at least 25% from the dose on the community medication list, without a documented clinical explanation.
These definitions were operationalized using the documentation available in the linked study data. Accordingly, absence of a documented rationale may not in all cases indicate that a medication change was clinically inappropriate; rather, it indicates that no justification was documented in the data sources used for this study.
To quantify discrepancy burden, we also modeled a discrepancy count ranging from 0 to 3, defined as the sum of the three subtype indicators.
5.6. Potential Confounders
5.6.1. Age and Sex
5.6.2. Healthcare Utilization Before Admission
Healthcare utilization before admission may reflect clinical complexity and care patterns that influence medication safety. ED utilization and prior hospitalization were captured in the 3 months prior to admission using administrative data sources (RAMQ claims for ED; MED-ÉCHO for hospitalizations) [13,14,15,16].
Because the analytic cohorts may encode utilization in different ways, baseline ED utilization and prior hospitalization were modeled using the operationalization available in the analytic dataset (e.g., count of ED visit days when available; otherwise, an indicator of any ED use; similarly for hospitalization).
To characterize baseline care fragmentation and medication burden, dispensing data were used to define the number of distinct prescribers, the number of distinct community pharmacies, and the number of distinct drugs dispensed in the baseline period.
5.6.3. Comorbidity
Comorbidity burden was measured using the Elixhauser comorbidity framework derived from ICD-9/ICD-10 codes in physician claims and hospitalization data [17]. Comorbidity burden was summarized using both a weighted Elixhauser comorbidity score, derived using published weights, and a simple count of comorbidities [18].
5.6.4. Medication-Related Covariates
Psychotropic medication use in the 3 months prior to admission was summarized descriptively using categories reflecting increasing medication burden (1, 2–4, and 5+ distinct psychotropic medications), based on the count of distinct dispensed medications (CODEGEN) within the baseline window. Psychotropic medications were identified using ATC-based definitions (Table S19). For regression adjustment, psychotropic exposure was modeled using the corresponding categorical indicator variables, with 1 psychotropic medication used as the reference category. Baseline high-risk medication exposure was captured as the count of distinct high-risk medications (CODEGEN) dispensed in the 3 months prior to admission, with high-risk classes defined using ATC groupings (Table S20). For descriptive reporting, high-risk medications were also summarized in categories (0, 1, 2–4, 5+). The total number of discharge medications was measured as the count of distinct medications (CODEGEN) on the discharge prescription (control and intervention lists combined), after applying the same discharge-list validation/standardization rules used for discrepancy identification and excluding items flagged for removal.
5.6.5. Trial Arm and Hospital Unit
Trial arm (electronic vs. usual-care medication reconciliation) was included as an adjustment covariate and as an effect modifier in interaction analyses. Hospital unit was summarized descriptively but excluded from adjusted models due to sparse strata and model stability considerations.
5.6.6. Hospitalization-Related Variables
5.7. Statistical Analysis
We first described baseline characteristics and crude discrepancy prevalence by mental health status. The prespecified primary analysis examined the association between binary mental health status (any mental health condition vs. no mental health condition) and any unintended medication discrepancy at discharge using multivariable logistic regression. Secondary logistic models examined each discrepancy subtype separately. Additional analyses within the mental health subgroup compared serious mental illness (SMI) with other mental health conditions. We also fit a negative binomial model for discrepancy count and conducted a supplementary 3-level exposure analysis (no mental health condition/other mental health condition/SMI). To assess whether associations differed by trial arm, we fit interaction models including a mental health status × trial arm term and arm-stratified models. All multivariable models adjusted for the covariates listed above. Because concerns were raised regarding multiplicity, we conducted an additional sensitivity analysis applying the Holm correction to the family of secondary and supplementary inferential analyses. The single prespecified primary confirmatory test was kept unadjusted. Adjusted p-values for secondary and supplementary analyses are reported in the Supplementary Materials. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Statistical significance is defined using a two-sided α of 0.05.
6. Conclusions
In this cohort of adults discharged from general medical and surgical units, the prespecified primary analysis showed a lower observed inverse association between mental health condition status and any unintended medication discrepancy at discharge. However, this result was contrary to the a priori expectation suggested by prior literature on mental health populations and medication-related risk and should not be interpreted as evidence of a protective effect. No secondary or supplementary analyses remained statistically significant after Holm correction. Taken together, these findings suggest that documentation-based discrepancy measures may not correspond straightforwardly to underlying clinical vulnerability and that discrepancy risk in patients with mental health conditions may be more heterogeneous than often assumed. The observed primary association should therefore be confirmed in independent settings before any causal or operational interpretation is made.
Supplementary Materials
The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/pharma5020017/s1, Table S1: Descriptive prevalence of unintended medication discrepancies at discharge by 3-level mental health status (No MH, Other MH, SMI); Table S2: Multivariable logistic regression for any unintended discrepancy at discharge (Panel A: full cohort, MH vs. No MH; Panel B: mental health subgroup, SMI vs. Other MH); Table S3: Multivariable logistic regression for omission at discharge (Panel A: MH vs. No MH; Panel B: SMI vs. Other MH); Table S4: Multivariable logistic regression for therapeutic duplication at discharge (Panel A: MH vs. No MH; Panel B: SMI vs. Other MH); Table S5: Multivariable logistic regression for unintentional dose change at discharge (Panel A: MH vs. No MH; Panel B: SMI vs. Other MH); Table S6: Any unintended discrepancy at discharge with Exposure × Arm interaction (MH×Arm in full cohort; SMI×Arm in mental health subgroup); Table S7: Omission at discharge with Exposure × Arm interaction; Table S8: Therapeutic duplication at discharge with Exposure × Arm interaction; Table S9: Unintentional dose change at discharge with Exposure × Arm interaction; Table S10A: Any unintended discrepancy at discharge stratified by trial arm (Control vs. Intervention), MH vs. No MH; Table S10B: Any unintended discrepancy at discharge stratified by trial arm (Control vs. Intervention), SMI vs. Other MH; Table S11: Omission at discharge stratified by trial arm (Control vs. Intervention), MH vs. No MH and SMI vs. Other MH; Table S12: Therapeutic duplication at discharge stratified by trial arm (Control vs. Intervention), MH vs. No MH and SMI vs. Other MH; Table S13: Unintentional dose change at discharge stratified by trial arm (Control vs. Intervention), MH vs. No MH and SMI vs. Other MH; Table S14: Negative binomial regression for discrepancy count at discharge (MH vs. No MH; SMI vs. Other MH); Table S15: Negative binomial regression for discrepancy count using a 3-level mental health exposure with Arm and MH3×Arm interaction; Table S16: Any unintended discrepancy at discharge using a 3-level mental health exposure (No MH/Other MH/SMI); Table S17: Operational definition of the 3-level mental health exposure (SMI vs. Other MH vs. No MH); Table S18: Specific mental health condition indicators and burden of distinct mental health problems; Table S19: ATC-based definition of baseline psychotropic medication classes (3 months pre-admission); Table S20: ATC-based definition of baseline high-risk medication exposure (3 months pre-admission). Table S21: Raw and Holm-adjusted p-values for secondary and supplementary inferential analyses.
Author Contributions
Conceptualization, N.N. and R.T.; methodology, N.N. and R.T.; software, N.N.; validation, N.N. and R.T.; formal analysis, N.N.; investigation, N.N.; resources, R.T.; data curation, N.N.; writing—original draft preparation, N.N.; writing—review and editing, N.N. and R.T.; visualization, N.N.; supervision, R.T.; project administration, R.T.; funding acquisition, R.T. All authors have read and agreed to the published version of the manuscript.
Funding
This research received no external funding. The APC was funded by the authors.
Institutional Review Board Statement
The study was conducted in accordance with the Declaration of Helsinki and was approved by the McGill University Health Centre (MUHC) Research Ethics Board (CTGQ panel), project number 2011-930 (also listed as 10-180). The initial project REB approbation date was 9 March 2011, with the most recent annual renewal approved on 24 November 2025 (renewal period 21 December 2025 to 20 December 2026).
Informed Consent Statement
Informed consent was obtained from all subjects involved in the RightRx trial. For the current retrospective cohort analysis using linked administrative data, consent requirements were handled in accordance with Research Ethics Board approvals and applicable provincial privacy regulations.
Data Availability Statement
The data used in this study are not publicly available due to legal and ethical restrictions related to privacy and the conditions of access to Quebec administrative health data (RAMQ/MED-ÉCHO) and the RightRx trial dataset. Access may be possible through the appropriate data custodians and institutional approvals, subject to data sharing agreements and Research Ethics Board authorization.
Acknowledgments
The authors thank the RightRx trial team and the McGill University Health Centre (MUHC) clinical and informatics teams for supporting data access and linkage processes, as well as the provincial data custodians for the use of RAMQ and MED-ÉCHO data.
Conflicts of Interest
The authors declare no conflicts of interest.
Abbreviations
The following abbreviations are used in this manuscript:
| ADE | Adverse drug event |
| ATC | Anatomical Therapeutic Chemical (classification system) |
| CI | Confidence interval |
| CODEGEN | Drug code identifier used in dispensing/prescribing data (study operational variable) |
| ED | Emergency department |
| IRR | Incidence rate ratio |
| LOS | Length of stay |
| MED-ÉCHO | Québec hospitalization administrative database |
| MH | Mental health condition(s) |
| No MH | No documented mental health condition(s) |
| MUHC | McGill University Health Centre |
| RAMQ | Régie de l’assurance maladie du Québec |
| REB | Research Ethics Board |
| RightRx | Cluster-randomized trial of electronic vs. usual-care medication reconciliation |
| SMI | Serious mental illness |
| aOR | Adjusted odds ratio |
References
- Glans, M.; Ekstam, A.K.; Jakobsson, U.; Bondesson, Å.; Midlöv, P. Medication-related hospital readmissions within 30 days of discharge-A retrospective study of risk factors in older adults. PLoS ONE 2021, 16, e0253024. [Google Scholar] [CrossRef] [PubMed]
- Park, J.; Kim, A.J.; Cho, E.-J.; Cho, Y.S.; Jun, K.; Jung, Y.S.; Lee, J.-Y. Unintentional medication discrepancies at care transitions: Prevalence and their impact on post-discharge emergency visits in critically ill older adults. BMC Geriatr. 2024, 24, 1000. [Google Scholar] [CrossRef] [PubMed]
- Tamblyn, R.; Abrahamowicz, M.; Buckeridge, D.L.; Bustillo, M.; Forster, A.J.; Girard, N.; Habib, B.; Hanley, J.; Huang, A.; Kurteva, S.; et al. Effect of an Electronic Medication Reconciliation Intervention on Adverse Drug Events: A Cluster Randomized Trial. JAMA Netw. Open 2019, 2, e1910756. [Google Scholar] [CrossRef] [PubMed]
- Schnipper, J.L.; Nieva, H.R.; Yoon, C.; Mallouk, M.; Mixon, A.S.; Rennke, S.; Chu, E.S.; Mueller, S.K.; Smith, G.R.; Williams, M.V.; et al. What works in medication reconciliation: An on-treatment and site analysis of the MARQUIS2 study. BMJ Qual. Saf. 2023, 32, 457–469. [Google Scholar] [CrossRef] [PubMed]
- Simoons, M.; Mulder, H.; Risselada, A.J.; Wilmink, F.W.; Schoevers, R.; Ruhé, H.G.; Roon, E.N. Medication Discrepancies at Outpatient Departments for Mood and Anxiety Disorders in the Netherlands: Risks and Clinical Relevance. J. Clin. Psychiatry 2016, 77, 1511–1518. [Google Scholar] [CrossRef] [PubMed]
- Keers, R.N.; Williams, S.D.; Vattakatuchery, J.J.; Brown, P.; Miller, J.; Prescott, L.; Ashcroft, D.M. Medication safety at the interface: Evaluating risks associated with discharge prescriptions from mental health hospitals. J. Clin. Pharm. Ther. 2015, 40, 645–654. [Google Scholar] [CrossRef] [PubMed]
- Tamblyn, R.; Winslade, N.; Lee, T.C.; Motulsky, A.; Meguerditchian, A.; Bustillo, M.; Elsayed, S.; Buckeridge, D.L.; Couture, I.; Qian, C.J.; et al. Improving patient safety and efficiency of medication reconciliation through the development and adoption of a computer-assisted tool with automated electronic integration of population-based community drug data: The RightRx project. J. Am. Med. Inform. Assoc. 2018, 25, 482–495. [Google Scholar] [CrossRef] [PubMed]
- Uitvlugt, E.B.; Janssen, M.J.A.; Siegert, C.E.H.; Kneepkens, E.L.; van den Bemt, B.J.F.; van den Bemt, P.M.L.A.; Karapinar-Çarkit, F. Medication-Related Hospital Readmissions Within 30 Days of Discharge: Prevalence, Preventability, Type of Medication Errors and Risk Factors. Front. Pharmacol. 2021, 12, 567424. [Google Scholar] [CrossRef] [PubMed]
- Degen, M.; Chen, L.J.; Schöttker, B. Medication reviews in hospitalised patients for reduced hospital readmission and mortality. Systematic review, meta-analysis and meta-regression of RCTs. Ageing Res. Rev. 2025, 104, 102661. [Google Scholar] [CrossRef] [PubMed]
- Gaynes, B.; Brown, C.; Lux, L.J.; Sheitman, B.; Ashok, M.; Boland, E.; Morgan, L.; Swinson-Evans, T.; Whitener, L.; Viswanathan, M. Relationship Between Use of Quality Measures and Improved Outcomes in Serious Mental Illness, 2015. Available online: https://www.ncbi.nlm.nih.gov/books/NBK274091/ (accessed on 24 March 2026).
- Brabete, A.C.; Greaves, L.; Maximos, M.; Huber, E.; Li, A.; Lê, M.-L. A Sex-and Gender-Based Analysis of Adverse Drug Reactions: A Scoping Review of Pharmacovigilance Databases. Pharmaceuticals 2022, 15, 298. [Google Scholar] [CrossRef] [PubMed]
- Franconi, F.; Campesi, I. Sex and gender influences on pharmacological response: An overview. Expert Rev. Clin. Pharmacol. 2014, 7, 469–485. [Google Scholar] [CrossRef] [PubMed]
- Canadian Institute for Health Information. ICD-10-CA/CCI Classifications. Available online: https://secure.cihi.ca/estore/productSeries.htm?pc=PCC1860 (accessed on 25 May 2026).
- Kawasumi, Y.; Abrahamowicz, M.; Ernst, P.; Tamblyn, R. Development and Validation of a Predictive Algorithm to Identify Adult Asthmatics from Medical Services and Pharmacy Claims Databases. Health Serv. Res. 2011, 46, 939–963. [Google Scholar] [CrossRef] [PubMed]
- Tamblyn, R.; Huang, A.; Kawasumi, Y.; Bartlett, G.; Grad, R.; Jacques, A.; Dawes, M.; Abrahamowicz, M.; Perreault, R.; Taylor, L.; et al. The development and evaluation of an integrated electronic prescribing and drug management system for primary care. J. Am. Med. Inform. Assoc. 2006, 13, 148–159. [Google Scholar] [CrossRef] [PubMed]
- Roy, L.; Zappitelli, M.; White-Guay, B.; Lafrance, J.P.; Dorais, M.; Perreault, S. Agreement Between Administrative Database and Medical Chart Review for the Prediction of Chronic Kidney Disease G category. Can. J. Kidney Health Dis. 2020, 7, 2054358120959908. [Google Scholar] [CrossRef] [PubMed]
- Elixhauser, A.; Steiner, C.; Harris, D.R.; Coffey, R.M. Comorbidity measures for use with administrative data. Med. Care 1998, 36, 8–27. [Google Scholar] [CrossRef]
- Van Walraven, C.; Austin, P.C.; Jennings, A.; Quan, H.; Forster, A.J. A modification of the elixhauser comorbidity measures into a point system for hospital death using administrative data. Med. Care 2009, 47, 626–633. [Google Scholar] [CrossRef]
- Ahmed, S.; Tamblyn, R.; Winslade, N. Using decision support for population tracking of adherence to recommended asthma guidelines. BMJ Open 2014, 4, e003759. [Google Scholar] [CrossRef] [PubMed]
- Tamblyn, R.; Huang, A.; Perreault, R.; Jacques, A.; Roy, D.; Hanley, J.; McLeod, P.; Laprise, R. The medical office of the 21st century (MOXXI): Effectiveness of computerized decision-making support in reducing inappropriate prescribing in primary care. CMAJ 2003, 169, 549–556. [Google Scholar] [PubMed]
Figure 1.
Flow diagram of cohort derivation from the RightRx trial (October 2014–November 2016), adapted for the present study.
Figure 1.
Flow diagram of cohort derivation from the RightRx trial (October 2014–November 2016), adapted for the present study.
Table 1.
Baseline patient characteristics by mental health status (binary exposure). Cohort: N = 3567 (MH n = 877; No MH n = 2690).
Table 1.
Baseline patient characteristics by mental health status (binary exposure). Cohort: N = 3567 (MH n = 877; No MH n = 2690).
| Characteristic | Overall (N = 3567) | MH (n = 877) | No MH (n = 2690) |
|---|---|---|---|
| Patient gender | |||
| Female | 1507 (42.3%) | 411 (46.9%) | 1096 (40.7%) |
| Male | 2060 (57.8%) | 466 (53.1%) | 1594 (59.3%) |
| Age at admission | |||
| Age (years), mean (SD) | 70.0 (15.0) | 68.8 (15.3) | 70.1 (14.9) |
| Comorbidity | |||
| Elixhauser comorbidity count, mean (SD) | 2.5 (1.7) | 2.7 (2.2) | 2.4 (2.0) |
| Health care use: 3 months before admission | |||
| Any ED visit | 1938 (54.3%) | 536 (61.1%) | 1402 (52.1%) |
| Any hospitalization | 831 (23.3%) | 211 (24.1%) | 620 (23.1%) |
| Health care fragmentation: 3 months before admission | |||
| Distinct prescribers, mean (SD) | 2.8 (2.2) | 3.1 (2.4) | 2.8 (2.1) |
| Distinct pharmacies visited, mean (SD) | 1.1 (0.6) | 1.1 (0.7) | 1.1 (0.6) |
| Medication burden: 3 months before admission | |||
| Distinct drugs dispensed (CODEGEN), mean (SD) | 8.2 (6.0) | 9.0 (6.4) | 8.0 (5.9) |
| High-risk medications: 3 months before admission | |||
| High-risk medication count, mean (SD) | 3.2 (2.6) | 3.1 (2.7) | 3.2 (2.6) |
| Baseline psychotropic medications (categories) | |||
| 1 psychotropic | 739 (20.7%) | 225 (25.7%) | 514 (19.1%) |
| 2–4 psychotropics | 460 (12.9%) | 238 (27.1%) | 222 (8.3%) |
| 5+ psychotropics | 10 (0.3%) | 6 (0.7%) | 4 (0.1%) |
| Trial arm | |||
| Control | 1877 (52.6%) | 497 (56.7%) | 1380 (51.3%) |
| Intervention | 1690 (47.4%) | 380 (43.3%) | 1310 (48.7%) |
| Hospital unit | |||
| Cardiac surgery | 1687 (47.3%) | 284 (32.4%) | 1403 (52.2%) |
| Internal medicine | 1880 (52.7%) | 593 (67.6%) | 1287 (47.8%) |
| Index hospitalization | |||
| Length of stay (days), mean (SD) | 11.3 (12.7) | 14.3 (18.9) | 10.3 (14.9) |
| Number of discharge medications, mean (SD) | 11.6 (6.5) | 11.9 (6.8) | 11.5 (6.4) |
| Baseline high-risk medication categories | |||
| 0 high-risk meds | 682 (19.1%) | 180 (20.5%) | 502 (18.7%) |
| 1 high-risk med | 452 (12.7%) | 116 (13.2%) | 336 (12.5%) |
| 2–4 high-risk meds | 1403 (39.3%) | 335 (38.2%) | 1068 (39.7%) |
| 5+ high-risk meds | 1030 (28.9%) | 246 (28.1%) | 784 (29.1%) |
Values are mean (SD) or n (%). Percentages are within column. Mental health status was defined using qualifying ICD-9/ICD-10 diagnostic codes recorded from 12 months before admission through 7 days after discharge. Discrepancies were identified by comparing the validated community medication list compiled at admission with the discharge prescription. Baseline refers to measurements taken in the 3 months prior to admission.
Table 2.
Characteristics among patients with mental health conditions: SMI vs. other mental health conditions. Cohort: Mental health subgroup only (N = 877): SMI n = 346; Other MH n = 531.
Table 2.
Characteristics among patients with mental health conditions: SMI vs. other mental health conditions. Cohort: Mental health subgroup only (N = 877): SMI n = 346; Other MH n = 531.
| Characteristic | SMI (n = 346) | Other MH (n = 531) |
|---|---|---|
| Patient gender | ||
| Female | 159 (46.0%) | 252 (47.5%) |
| Male | 187 (54.1%) | 279 (52.5%) |
| Age at admission | ||
| Age (years), mean (SD) | 73.5 (14.3) | 65.7 (15.9) |
| Comorbidity | ||
| Elixhauser comorbidity count, mean (SD) | 3.0 (1.9) | 2.6 (1.8) |
| Health care use: 3 months before admission | ||
| Any ED visit | 229 (66.2%) | 307 (57.8%) |
| Any hospitalization | 84 (24.3%) | 127 (23.9%) |
| Health care fragmentation: 3 months before admission | ||
| Distinct prescribers, mean (SD) | 3.2 (2.4) | 3.0 (2.3) |
| Distinct pharmacies visited, mean (SD) | 1.1 (0.6) | 1.1 (0.7) |
| Medication burden: 3 months before admission | ||
| Distinct drugs dispensed (CODEGEN), mean (SD) | 9.9 (6.6) | 8.4 (6.2) |
| High-risk medications: 3 months before admission | ||
| High-risk medication count, mean (SD) | 3.3 (2.8) | 3.0 (2.6) |
| Baseline psychotropic medications (categories) | ||
| 1 psychotropic | 88 (25.4%) | 137 (25.8%) |
| 2–4 psychotropics | 108 (31.2%) | 130 (24.5%) |
| 5+ psychotropics | 3 (0.9%) | 3 (0.6%) |
| Trial arm | ||
| Control | 187 (54.1%) | 310 (58.4%) |
| Intervention | 159 (46.0%) | 221 (41.6%) |
| Hospital unit | ||
| Cardiac surgery | 60 (17.3%) | 224 (42.2%) |
| Internal medicine | 286 (82.7%) | 307 (57.8%) |
| Index hospitalization | ||
| Length of stay (days), mean (SD) | 20.1 (24.0) | 10.5 (11.6) |
| Number of discharge medications, mean (SD) | 12.5 (7.1) | 11.5 (6.6) |
| Baseline high-risk medication categories | ||
| 0 high-risk meds | 64 (18.5%) | 116 (21.8%) |
| 1 high-risk med | 46 (13.3%) | 70 (13.2%) |
| 2–4 high-risk meds | 130 (37.6%) | 205 (38.6%) |
| 5+ high-risk meds | 106 (30.6%) | 140 (26.4%) |
Values are n (%) or mean (SD). SMI = psychotic spectrum disorders or bipolar disorder (as defined in the protocol); Other MH = other mental health conditions. Values are descriptive.
Table 3.
Diagnostic composition of the mental health subgroup (n = 877).
| Diagnostic Category * | n (%) |
|---|---|
| Anxiety disorders | 277 (31.6) |
| Delirium/confusion | 271 (30.9) |
| Depressive disorders | 193 (22.0) |
| Psychotic disorders | 176 (20.1) |
| Addictive disorders | 130 (14.8) |
| Adjustment disorders | 94 (10.7) |
| Two or more mental health categories | 286 (32.6) |
* Categories are not mutually exclusive; therefore, percentages do not sum to 100%. Full diagnostic detail is provided in Table S18.
Table 4.
Descriptive prevalence of unintended medication discrepancies at discharge by binary mental health status. Cohort: No MH (n = 2690); MH (n = 877). Values are n (%), percentages within column.
Table 4.
Descriptive prevalence of unintended medication discrepancies at discharge by binary mental health status. Cohort: No MH (n = 2690); MH (n = 877). Values are n (%), percentages within column.
| Outcome | No MH (n = 2690) | MH (n = 877) |
|---|---|---|
| Any discrepancy | 1101 (40.9%) | 369 (42.1%) |
| Omission | 700 (26.0%) | 222 (25.3%) |
| Therapeutic duplication | 165 (6.1%) | 60 (6.8%) |
| Unintended dose change | 537 (20.0%) | 206 (23.5%) |
Percentages are within-column. Discrepancies were defined by comparing the validated community medication list at admission with the discharge prescription.
Table 5.
Adjusted associations between mental health status and unintended medication discrepancies at discharge. Panel (A). Full cohort: any mental health condition vs. no mental health condition. Panel (B). Mental health subgroup only: serious mental illness vs. other mental health conditions.
Table 5.
Adjusted associations between mental health status and unintended medication discrepancies at discharge. Panel (A). Full cohort: any mental health condition vs. no mental health condition. Panel (B). Mental health subgroup only: serious mental illness vs. other mental health conditions.
| (A) | ||
|---|---|---|
| Outcome | aOR (95% CI) | p-Value |
| Any unintended discrepancy | 0.81 (0.67–0.98) | 0.027 |
| Omission | 0.77 (0.62–0.95) | 0.016 |
| Therapeutic duplication | 0.97 (0.70–1.36) | 0.863 |
| Unintended dose change | 1.04 (0.84–1.28) | 0.747 |
| (B) | ||
| Outcome | aOR (95% CI) | p-Value |
| Any unintended discrepancy | 0.55 (0.37–0.82) | 0.003 |
| Omission | 0.64 (0.41–1.00) | 0.049 |
| Therapeutic duplication | 0.72 (0.36–1.45) | 0.359 |
| Unintended dose change | 0.66 (0.43–1.02) | 0.060 |
aOR = adjusted odds ratio. Panel A compares patients with any documented mental health condition with those without a documented mental health condition. Panel B compares patients with serious mental illness (SMI; psychotic spectrum disorder or bipolar disorder) with patients with other mental health conditions. Models adjusted for age, sex, Elixhauser comorbidity count, baseline emergency department visit and hospitalization in the prior 3 months, number of distinct prescribers and pharmacies, number of distinct baseline drugs dispensed, baseline psychotropic medication category, trial arm, baseline high-risk medication count, and discharge medication count. Full covariate estimates are provided in Tables S2–S5. The association shown in Panel A for any unintended discrepancy was the prespecified primary confirmatory analysis. All other analyses shown here were secondary or supplementary and were evaluated in sensitivity analyses using Holm correction (Table S21).
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.
Share and Cite
MDPI and ACS Style
Nassar, N.; Tamblyn, R. Medication Discrepancies at Hospital Discharge Among Adults with and Without Mental Health Conditions: A Retrospective Cohort Study. Pharmacoepidemiology 2026, 5, 17. https://doi.org/10.3390/pharma5020017
AMA Style
Nassar N, Tamblyn R. Medication Discrepancies at Hospital Discharge Among Adults with and Without Mental Health Conditions: A Retrospective Cohort Study. Pharmacoepidemiology. 2026; 5(2):17. https://doi.org/10.3390/pharma5020017
Chicago/Turabian StyleNassar, Nabil, and Robyn Tamblyn. 2026. "Medication Discrepancies at Hospital Discharge Among Adults with and Without Mental Health Conditions: A Retrospective Cohort Study" Pharmacoepidemiology 5, no. 2: 17. https://doi.org/10.3390/pharma5020017
APA StyleNassar, N., & Tamblyn, R. (2026). Medication Discrepancies at Hospital Discharge Among Adults with and Without Mental Health Conditions: A Retrospective Cohort Study. Pharmacoepidemiology, 5(2), 17. https://doi.org/10.3390/pharma5020017
