Based on the 3213 prescriptions from the 308 enrolled older adults, 55.52% had at least 1 pDDI. The most frequent pDDI that should be avoided was the coadministration of lorazepam and olanzapine. Constipation was the most common ADR induced by pDDIs. The number of medical diagnoses, the number of drugs used and a length of hospitalization of 18–28 days were risk factors for the number of ADR types; however, readmission was a protective factor for the number of ADR types. The length of hospitalization was a risk factor for abnormal liver function, and the number of drugs was a risk factor for gastric distress and dizziness and fainting. The length of hospitalization and the number of ADR types were risk factors for the clinical significance of the severity of ADRs. The number of drugs was a protective factor for the clinical significance of the severity of ADRs. Coadministrations of lorazepam and olanzapine, quetiapine and potassium chloride, quetiapine and escitalopram, and olanzapine and clonazepam were not risk factors for the clinical significance of the severity of ADRs.

pDDIs are prevalent in older adults with psychiatric disorders, and the rate is increasing. This finding is similar to those of other studies. Ocana-Zurita [22] performed a retrospective and cross-sectional study and found that 68.25% of schizophrenic patients were at risk of pDDIs. Ruangritchankul [23] found that 76% of older adults diagnosed with dementia experienced at least 1 pDDI.

Independent factors of ADRs vary in studies. de Vries [24] performed a cohort study and found that polypharmacy was a risk factor for ADRs. O′Mahony [25] identified that 8 factors, including female sex and having 4 or more multimorbidities were independent risk factors for ADRs. However, Lee [26] performed a meta-analysis and found that sex did not significantly influence the incidence of ADRs. Sun [27] conducted a study and found that 96.8% of ADRs occurred within 14 days of hospitalization, and length of stay, the number of drugs used in the hospital and underlying basic diseases were independent risk factors for ADRs. However, Lavan [28] did not identify associations between ADRs and age, sex, the number of daily medication or length of stay. The difference in results may be due to different populations and different types of drugs.

Many pDDIs may not be clinically significant in terms of the severity of ADRs caused by pDDIs. Madhusoodanan [29] conducted a pilot study and found that coadministration of lorazepam and olanzapine caused no adverse consequences. Bergemann [30] discovered that after the coadministration of olanzapine and lorazepam, the dose-corrected olanzapine plasma concentration was no different from the plasma levels under olanzapine monotherapy. However, a case report showed that IM olanzapine and IM lorazepam lowered blood pressure and caused dizziness [31].

The probable reasons are as follows. First, pDDIs are mainly speculated based on drug pharmacokinetic features, and psychotropic drugs are usually metabolized by several enzymes that reduce the risk of pDDIs [16]. Second, the route of administration may affect pDDIs and ADRs. The occurrence of pDDIs may be lower for the oral administration of olanzapine and lorazepam according to the database. In this study, antipsychotics, antidepressants and benzodiazepines were all orally administered and the incidence of ADRs was lowered. Third, the data are biased. For example, there were no ADRs among any older adults who used quetiapine in combination with escitalopram. Fourth, the majority of ADRs were mild [32] and preventable [33]. Psychiatrists have taken pDDIs seriously, and have prevented ADRs by controlling drug doses, strengthening monitoring, and taking measures in advance.

The strength of this study is that we discussed pDDIs and ADRs not only from a statistical perspective but also from a clinical significance perspective. Considering clinical medication safety, we analysed the relationship between the coadministration of drugs and ADRs.

Our work has five limitations. First, there is no standardization for clinical significance, and none of the approaches to set the MCID are ideal. This was a retrospective study, and other more objective calculation methods for establishing the MCID, such as anchor-based methods and triangulation of methods, could not be applied. Another limitation would be since this was a retrospective study and the causality of ADRs was not routinely assessed, the causal relationship between ADRs and the coadministration of drugs may not be accurate. Moreover, we focused on older adults with a major diagnosis of a psychiatric disorder who were taking central nervous system drugs. The fourth limitation is that we extracted information from the long-term physicians’ orders and ignored temporary physicians’ orders. Lastly, only four combinations of drugs were analysed in this study, and other combinations were ignored due to the low number of cases.