A retrospective comparison of inappropriate prescribing of psychotropics in three Norwegian nursing homes in 2000 and 2016 with prescribing quality indicators

One-day, point-prevalence of psychotropic prescribing in a convenience (non-randomised) sample was collected from the same three nursing homes in Bergen, Norway 16 years apart. No clinical or personal information other than patients’age and gender was collected. Nor was information pertaining to physicians, other staff, or details concerning organisation of the nursing homes. Only patients’aged ≥ 65 years were included at both sampling times. Researchers were blinded to patients’nursing home residence in 2016.

The three nursing homes in 2016 included regular units that provided long-term care, respite care, and rehabilitation as well as units for short-term care. They also included special care units that provide sheltered care for patients with dementia and psychogeriatric diseases. Thus, they reflect the transition in nursing homes from 2000 to 2016 with institutions consisting of mainly regular units to institutions that also provide various more specialized units. The two audits of prescribing were regarded as representative for identifying inappropriate prescribing of psychotropics.

The information on prescribing of psychotropics was manually collected from medical charts (cardex) by pharmacists in March 2000. Data included prescribed psychotropics, time of administration and respective doses. In March 2016, similar data were retrieved automatically from electronic charts. Psychotropics were classified according to the Anatomical Therapeutic Chemical (ATC) classification system [16], and included antipsychotics (ATC-code N05A), anxiolytics (N05B), hypnotics (N05C), antidepressants (N06A) and antiepileptics (N03A). Anxiolytics and hypnotics were collapsed into one category (anxiolytics-hypnotics) based on the assumption that the two categories have comparable adverse effects among nursing home patients. Antiepileptics are increasingly prescribed for psychiatric indications in Norway, and epilepsy increases with age [17]. Of relevance for inappropriate prescribing of psychotropics, data concerning all prescribed psychotropics (regularly prescribed and additional drugs as needed) were included [18]. Concurrent use of psychotropics from a similar ATC category is not uncommon in nursing homes, but resulted in a single registration for each patient (e.g. the patient uses ≥ 1 antidepressants).

Polypharmacy was defined as concurrent use of three or more psychotropics [12]. PICP were defined as combinations of two or more psychotropics from the same ATC-category. PIPS were defined as long-acting benzodiazepines and psychotropics with anticholinergic effects [12]. PICP were defined from the third level (pharmacological subgroup; e.g N05A, antipsychotics), and polypharmacy and PIPS were defined at the fifth level (chemical substance; e.g. N05A, risperidone) in the ATC-classification system [16]. The prescribing quality indicators above are recommended for patients aged 75 years or older, but could be relevant for younger patients (65 years or older) in nursing homes, and were used accordingly in the present study.

We also assessed the most important PIs in the prescription data of the respective years. The first ranked PI in a patient was selected by use of the subscription interaction database Stockley’s Interaction Alerts (SIA) [19]. SIA provides consistent albeit briefer information on drug interactions compared to Stockley’s Drug Interactions, and describes classification of the clinical relevance of a drug interaction. A clinically relevant PI in SIA is classified with the following three categories: recommended action, severity, and documentation. In SIA, the recommended action for a clinically relevant PI is either “informative”, “monitor”, “adjust dose”, or “avoid”. Recommended action for a clinically relevant PI in the present study was defined by collapsing “monitor” and “adjust dose” into the following three levels (from low to high); “informative”, “monitor or adjust dose”, or “avoid”. In the present study, severity was classified (from low to high) as “mild”, “moderate” or “high” in concordance with SIA. Documentation in SIA is classified with four levels; “theoretical”, “case”, “study” or “extensive”. “Extensive” is an option for documentation of interactions where the information provided is based on numerous small or medium size studies or several large studies usually supported by case reports [19]. Documentation of a clinically relevant PI in the present study was defined by collapsing “study” and “extensive” to aquire the following levels (from low to high); “theoretical”, “case”, or “study”. The first ranked PI in each patient was defined by the following hierarchy of categories and levels; recommended action (avoid > monitor or adjust dose > informative) > documentation (study > case > theoretical) > severity (severe > moderate > mild). The order of the categories were based on our experience that many PIs may be described as potentially severe, but recommended action (e.g. contraindicated) and documentation (study) is of importance when providing drug information. The number of concurrent PIs was registered for each patient. A clinical pharmacologist (JS) defined each PI as pharmacodynamic or pharmacokinetic based on the description in SIA. If a pharmacokinetic PI contained an additional description of a pharmacodynamic effect which could not be explained by a change in plasma level of one or both drugs, the PI was defined as pharmacokinetic.

The study investigated the appearance of different types of psychotropics as well as their combinations and interactions as outcomes. Predictors of having ≥ 1 PIs among patients with ≥ 2 concurrent psychotropics included age, gender, number of psychotropics, and use of antidepressants, antipsychotics, anxiolytics-hypnotics or antiepileptics. Dichotomous variables were coded as 0 or 1 for absence or presence of a quality (e.g. no use of antidepressants = 0; use of ≥ 1 antidepressants = 1). Age and number of psychotropics were introduced in the model as continuous variables.

In 2000 there were 386 patients in the three nursing homes increasing to 416 in 2016. The proportion of male residents increased from 26.4% to 33.9% in 2016. Mean age ± standard deviation (SD) among patients was 84.6 ± 7.2 years in 2000, and 84.3 ± 8.1 years in 2016 (p = 0.704). Women were older than men at both sampling times, but there were no significant age differences between the time-points within each gender. There were no significant differences with regard to age or gender among subgroups of patients who used psychotropics or had PIs. No patients were younger than 65 years in 2000. Twenty patients aged 23 to 64 years were excluded from the material in 2016.

Twenty-two different psychotropics were prescribed in both 2000 and 2016, while 14 and 13 psychotropics were unique for 2000 and 2016, respectively (Table 1). Relatively few patients (< 5%) were prescribed unique drugs in 2000. Three of the unique drugs in 2016 were prescribed to > 10% of the patients. The five most frequent psychotropics prescribed to patients in 2000 were citalopram, oxazepam, zopiclone, risperidone and carbamazepine. In 2016, the most frequent prescribed were oxazepam, zopiclone, mirtazapine, quetiapine and escitalopram. Among psychotropics prescribed in both 2000 and 2016, the increase of oxazepam and zopiclone was most striking. In 2000, 386 patients were prescribed 352 psychotropics compared to 416 patients and 765 psychotropics in 2016. On average, 0.91 psychotropics per person in 2000 increased to 1.84 in 2016 in the three nursing homes. The median number of concurrent psychotropics was two in both years. A range of two to four concurrent psychotropics in 2000 increased to two to six in 2016.

Comparison of prescribing psychotropics, including psychotropic polypharmacy, in the three nursing homes is shown in Table 2. Psychotropic polypharmacy increased significantly from 6.2% to 29.6% (p < 0.001). The use of antidepressants and anxiolytic-hypnotics increased in general, but also among patients with PIs. Table 3 shows use of combinations of psychotropics, PICP and PIPS. Prescribing of combinations of antidepressants and anxiolytic-hypnotics increased between the two years. In 2016, fourteen patients (3.4%) were prescribed combinations of two antidepressants and two anxiolytics-hypnotics. Among these, three patients were also prescribed two additional antipsychotics. Two patients were prescribed combinations of two antipsychotics and two anxiolytics-hypnotics, while one patient were prescribed a combination of two antidepressants and three anxiolytics-hypnotics. One patient was prescribed three concurrent anxiolytics-hypnotics. The increase in PICP was associated with anxiolytic-hypnotics, while a reduction in prescription of several concurrent antipsychotics or antiepileptics was observed. PIPS were reduced from 2000 to 2016, but an increase in prescription of diazepam was observed.

The median number of PIs was 1 in both 2000 and 2016. Sixty-three patients had 94 PIs in 2000 increasing to 146 patients with 236 PIs in 2016. A range of one to four concurrent PIs in 2000 increased to one to nine in 2016. The three psychotropics most frequently prescribed to patients with PIs in 2000 were citalopram (22 patients), risperidone (16 patients) and zopiclone (13 patients). The three psychotropics most frequently prescribed to patients with PIs in 2016 were oxazepam (90 patients), mirtazapine (77 patients) and zopiclone (51 patients). Table 4 shows classification of first ranked PIs with the most frequent psychotropics involved. Mechanisms of the interactions with the associated level of documentation are described. First ranked PIs constituted 67% of all interactions in 2000 and 62.7% in 2016. First ranked PIs were increasingly pharmacodynamic with a decreased level of recommended action and documentation, but not severity. First ranked PIs involved three PIPS in 2000 and none in 2016. Antipsychotics were frequently involved in first ranked PIs in 2000 while antidepressants were frequently involved in 2016. Pharmacodynamic PIs mainly concerned risk of sedation based on theoretical documentation. Twelve PIs that should be avoided concerned risk of QT-prolongation irrespective of psychotropics involved (not only those listed in Table 4). Other PIs ranked as severe concerned risk of the serotonin syndrome. Pharmacokinetic PIs involved as expected a change in plasma levels of one or both drugs, and pharmacokinetic PIs were frequently based on study.

The variance inflation factors were < 2.0 and tolerance values were > 0.5 for all predictors in multicollinearity analysis in both 2000 and 2016. Predictors for having PIs from the univariate logistic regression analysis with p < 0.25 that could be included in the multivariable logistic regression analysis was number of psychotropics and prescription of antidepressants, antipsychotics (only in 2016), and anxiolytics-hypnotics. In multivariable analysis only number of psychotropics was predictive of having PIs in 2000. In 2016, number of psychotropics and prescription of antidepressants were predictive of having PIs (Table 5).

Classification of PIs in both 2000 and 2016 was performed using a single subscription database updated in 2017. However, this database scores statistically high with regard to ownership, classification of interactions, primary information sources and staff qualification in comparison with open access drug interaction databases [14]. Prescribing indicators like PICP and PIPS have been used in several research studies [12, 43]. The nursing homes included in the present study had access to modern prescribing tools. From 2007/2008, two of the nursing homes had access to pilot a computerized decision support system called the Geriatric Basis Dataset (GBD), which included the interaction tool in the Norwegian Pharmaceutical Product Compendium, while the third nursing home gained access to pilot the system in 2012. Furthermore, GBD include The Norwegian General Practice (NORGEP) criteria for prescribing to elderly > 70 years [43].

Limitations include the theoretical approach of this small study, lack of clinical information, lack of information about other drugs, and information concerning organisation and staff within the institutions. The respective patient populations in 2000 and 2016 could be quite different since the nursing homes had more specialized units in 2016. Increased comorbidity, dementia and polypharmacy among the patients could warrant more or different psychotropics. Only one drug interaction database was used for classification of clinically relevant PIs while it is well known that drug interaction databases lacks consistency [20, 21]. Furthermore, the most important PI in each patient (about 63-67% of all PIs) was used to describe qualities of psychotropic interactions, while use of other databases could provide different ranking. Only potential PIs are described, and all prescriptions (regular and on demand) were included which may increase the risk of overestimating daily use of psychotropics. Anxiolytics and hypnotics as a single category could also overestimate PICP.