Retrospective outpatient study links SSRI interaction flags to polypharmacy

A retrospective analysis of outpatient prescriptions at Beijing Tongren Hospital found that nearly one in five medication regimens containing a selective serotonin reuptake inhibitor (SSRI) triggered at least one potential drug-drug interaction when checked with two commercial screening databases. The study included 3,476 distinct prescription combinations for 1,186 adults, covering five SSRIs: escitalopram, sertraline, citalopram, paroxetine and fluoxetine. Overall, 684 prescriptions, or 19.68%, contained at least one flagged interaction, with 1,404 potential interactions detected across those regimens.

These were database-defined risks rather than confirmed adverse events, and the single-hospital design limits how widely the findings can be generalized. Researchers applied Micromedex and Lexi-Interact to historical prescriptions containing an SSRI and at least one other systemic medication between January 2023 and December 2024. The prescriptions represented distinct medication-combination instances per patient rather than individual dispensing or visit events, and an identical combination recurring across visits was counted only once.

This approach was intended to prevent repeat dispensing of an unchanged regimen from inflating either the prevalence denominator or the interaction tally. Two pharmacists independently reviewed the screening results, with a senior pharmacist resolving discrepancies or ambiguities. The investigators excluded topical, ophthalmic and other non-systemic formulations, as well as prescriptions consisting only of an SSRI and traditional Chinese medicine.

When traditional medicines appeared alongside Western medicines, only Western drug–SSRI pairs were screened. Among regimens containing two or three concomitant drugs, 19.93% had at least one potential interaction, compared with 51.50% of those containing four to six drugs and 61.90% of those containing seven or more. In multivariable logistic regression, prescriptions with four to six drugs had higher odds of a flagged interaction than those with two or three, with an adjusted odds ratio of 4.31 and a 95% confidence interval of 3.46 to 5.37.

For seven or more drugs, the adjusted odds ratio was 6.52, with a 95% confidence interval of 2.73 to 15.58. The model had a C-statistic of 0.707 and an acceptable calibration result from the Hosmer–Lemeshow test, with a p value of 0.396. These estimates describe associations in the analyzed prescriptions; they do not show that adding a particular medicine caused an interaction or clinical harm.

Patient characteristics also differed between prescriptions with and without at least one flag. Potential interactions occurred in 21.56% of prescriptions for women and 16.94% for men. After adjustment for age and medication count, female sex was associated with higher odds of a flagged interaction, with an adjusted odds ratio of 1.48 and a 95% confidence interval of 1.23 to 1.78.

Patients aged 40 to 64 had higher odds than those aged 18 to 39, with an adjusted odds ratio of 1.33 and a 95% confidence interval of 1.02 to 1.73. By contrast, patients aged 65 or older had lower odds than the youngest group, with an adjusted odds ratio of 0.73 and a 95% confidence interval of 0.56 to 0.96. The cohort was relatively old, with a mean age of 62.25 years, and 59.23% of prescriptions were associated with women, so these estimates may not represent younger or more male-dominated outpatient populations.

Antipsychotic medications formed the largest class of interacting drugs, appearing in 292 interaction events, followed by other antidepressants and blood glucose-lowering agents, with 207 events each. Antithrombotic agents accounted for 193 events and calcium channel blockers for 157. The most frequently listed potential clinical consequences were QT interval prolongation in 569 instances, serotonin syndrome in 390, bleeding in 229 and hypoglycemia in 207.

These figures count database entries and their listed consequences, not patient outcomes. The results were concentrated in recognizable combinations: antipsychotic combinations with SSRIs were associated with QT interval prolongation or serotonin syndrome, while combinations with antithrombotics were associated with bleeding. Among the 207 blood glucose-lowering interaction events, the listed consequence was hypoglycemia.

The study therefore captured several different clinical hazards in the same outpatient medication data, even though it could not determine whether any of the possible consequences occurred. Escitalopram accounted for 2,559 of the 3,476 prescriptions examined, followed by sertraline with 492, citalopram with 243, paroxetine with 149 and fluoxetine with 33. The number of interaction events relative to prescriptions containing each SSRI was highest for fluoxetine, at 78.79%, followed by escitalopram at 44.70%, sertraline at 35.98%, paroxetine at 30.87% and citalopram at 4.53%.

This measure reflects the average interaction-event burden per prescription within each SSRI subgroup rather than the percentage of individual prescriptions with at least one flag. The authors emphasized that the comparatively low value for citalopram was descriptive and had not been established through a formal comparison among SSRIs adjusted for age, medication count and other potential confounders. For the same reason, the higher event counts for escitalopram and fluoxetine should not be interpreted as a definitive ranking of their real-world interaction risk.

The two databases produced markedly different severity profiles. Micromedex detected 939 interactions in 94 drug combinations and classified 98.83% of them as major, 0.43% as severe and 0.75% as moderate. Lexi-Interact detected 1,335 interactions in 137 combinations, with 68.76% rated moderate, 21.50% minor, 9.21% major and 0.52% severe.

The databases identified 856 interactions in common. Among the drug pairs rated by both systems in the subset analyzed for concordance, 15.6% were concordant, with a Cohen’s kappa of 0.034, indicating essentially no agreement beyond chance. All 27 discordant pairs followed the same directional pattern: Micromedex assigned a major rating while Lexi-Interact assigned a moderate or minor rating.

When weighted by the number of prescriptions affected rather than the number of unique pairs, concordant classifications covered only 16.5% of overlapping interaction events. Severe classifications were uncommon, but several pairs were treated as contraindicated by one database or the other. Escitalopram with domperidone, escitalopram with rasagiline, sertraline with fluconazole and paroxetine with intravenous potassium chloride appeared in the table of severe interactions.

The databases also differed in the clinical consequences attached to identical combinations. For example, Lexi-Interact listed both serotonin syndrome and QT interval prolongation for several olanzapine–SSRI pairs, whereas Micromedex listed QT interval prolongation alone. The discrepancy means that a single tool may produce materially different impressions of severity and of the adverse outcome attached to a combination.

The study did not test which database classification better predicted clinical harm, and the retrospective analysis did not record whether clinicians saw, accepted or acted on alerts generated while prescribing. Several limitations constrain the findings. The sample was older and more likely to be female than many general outpatient populations, while the exclusion of traditional Chinese medicines limits applicability to prescribing environments in which herb–drug combinations are common.

The analysis measured interactions detected after the fact by applying two commercial databases to historical prescriptions rather than through prospective monitoring of clinical decision support alerts. It therefore could not determine which warnings clinicians actually received, whether they were acted upon or how alert fatigue influenced prescribing. It also lacked patient-reported adverse reactions, chart-confirmed events and formal outcome-validation tools such as the Drug Interaction Probability Scale.

Prospective, multicenter studies incorporating real-time alert logs, adverse-event tracking and pharmacogenomic information would be needed to determine how database-defined interaction risks translate into patient outcomes and safer prescribing decisions.

Subject of Research: Retrospective screening of potential drug-drug interactions involving SSRIs in hospital outpatients

Article Title: Potential drug-drug interaction risks of selective serotonin reuptake inhibitors in outpatients: a retrospective analysis using two screening databases

Article References: Xu, S., Song, Z., Li, Y., Bai, J., Wang, D., Wang, E., & Wang, J. (2026). Potential drug-drug interaction risks of selective serotonin reuptake inhibitors in outpatients: a retrospective analysis using two screening databases. Frontiers in Pharmacology, 17, Article 1965339. https://doi.org/10.3389/fphar.2026.1965339

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1965339

Keywords: SSRIs, drug-drug interactions, outpatients, polypharmacy, Micromedex, Lexi-Interact, QT prolongation, serotonin syndrome, bleeding, hypoglycemia, electronic prescribing, Potential