Abstract
In a recent study of 980 Australians with a lifetime diagnosis of a psychotic disorder, 62.3% described impairment in their daily life due to side-effects of medications. Of this sample 51.6% had been admitted to an inpatient unit in the previous year [1]. Episodes in hospital are critical periods when management plans are generated (or revised), therapeutic alliances are established (or re-established), insight is fostered, and new links to treatment and support services are established. To allow the potential therapeutic gains of this period of intensive treatment to be realized, it is important to minimize the development of drug-induced impairment or disability during the acute episode.
Pro re nata (abbreviated as PRN and variously defined as ‘as needed’, ‘whenever necessary’ or ‘as the occasion arises’) medications are commonly used during inpatient psychiatric care [2–5]. Pro re nata medication is directed at alleviation of symptoms rather than treatment of the underlying syndrome. Despite the ubiquity of PRN medication it is probably the least well studied aspect of pharmacotherapy for psychosis. Earlier studies examining orders for PRN medications indicate they were written for approximately 75% of psychiatric inpatients and administered to approximately 50% (ranging between 22.9% [6] and 88% [7], according to the type of unit). Once prescribed, most PRN medications were given on the initiative of nursing staff [3]. In this paper we review the evidence for best practice regarding PRN medication and then examine current clinical practice in two major Australian metropolitan hospitals by retrospective chart audit.
Antipsychotic treatment
Modern approaches to the pharmacotherapy of psychotic disorders have two key themes, the use of lower doses of typical antipsychotics and increased use of atypical antipsychotic medication. Most Australian psychiatrists have altered their practice to reflect these changes in the prescribing guidelines for antipsychotic medication [8].
Atypical antipsychotics have a wider therapeutic window between onset of antipsychotic activity and extrapyramidal symptoms (EPS) than typical antipsychotics [9, 10]. They are better tolerated, produce less secondary morbidity, and are at least as effective as typical antipsychotics. While the efficacy of atypical antipsychotics against some core negative symptoms is still the subject of debate, little doubt exists about their capacity to reduce the contribution of EPS to negative symptoms [11]. Since the development of EPS is strongly associated with subsequent non-adherence to treatment [12, 13], the lesser propensity of atypical antipsychotics to cause EPS means that many patients formerly non-compliant with oral medication are now prepared to take them [14, 15].
Pro re nata use of antipsychotics and benzodiazepines
There are no randomized controlled trials of the relative efficacy of various types of oral PRN medications for acute management of psychotic patients. However, retrospective analyses of oral PRN psychotropic medication outcomes based on non-blinded staff ratings suggest that anxiolytics are more likely to be judged effective than antipsychotics [5].
Rapid tranquillization (RT) is a practice in emergency psychiatry that has similar therapeutic goals to PRN treatment. As such, controlled studies of RT may provide some evidence regarding ‘best practice’ in the management of acute psychotic symptoms. Rapid tranquillization was originally defined as ‘a circumscribed procedure of giving varying amounts of antipsychotic medication over brief intervals of time to control agitated, threatening, and potentially destructive patients’ [16]. However, a 1988 review [17] concluded that both antipsychotics and benzodiazepines were effective for this purpose, but that antipsychotics produced more adverse effects (‘toxicity’). Since 1989, four controlled trials of RT have been published [18–21]. Mostly, they compared intramuscular lorazepam to haloperidol. These studies were consistent in concluding that benzodiazepines and antipsychotics were equally effective against both agitation and psychotic symptoms over 24 hours. Benzodiazepines, however, were more likely to induce sleep, were better tolerated and produced markedly less EPS than antipsychotics. Two of these trials examined the combination of a benzodiazepine and antipsychotic, which they found produced more rapid onset of action and a reduced need for multiple administrations compared to monotherapy with either agent [18, 19].
An important issue in using typical antipsychotics PRN is that they may interfere with the therapeutic action of regularly prescribed atypical antipsychotics by modifying the profile of receptor occupancy. The alternative monotherapy approach of prescribing atypical antipsychotics on both a regular and PRN basis was not supported by a recent review [22]. This remains an open question deserving of further investigation.
The use of benzodiazepine as first-line PRN agents to control agitation has received endorsement from the RANZCP Committee on Psychotropic Drugs and Other Physical Treatments [23]. However, use of benzodiazepines is not without problems. The development of tolerance and/or withdrawal effects may arise with prolonged use. Moreover, individuals with substance use disorders may already be tolerant and thus require higher doses. In less tolerant individuals, dose dependent sideeffects of sedation, confusion and ataxia may occur. Although rare, paradoxical behavioural reactions or respiratory depression are possible in vulnerable individuals.
In summary, the sparse existing evidence suggests that best practice in the acute pharmacotherapy of psychosis is to prescribe benzodiazepines as first-line PRN agents in conjunction with atypical or low doses of typical antipsychotics as regular medication. It is necessary also to consider the role of both doctors in ordering the medication and nurses in its administration. Ideally, PRN medication orders should specify indications for use and a maximum frequency or maximum daily dose [7]. Administration should be in accordance with the specified parameters, and accompanied by documentation of the specific reason for medication use and outcome [24, 25].
Method
We sought to obtain records of consecutive admissions from each of two inner-Brisbane general hospitals with large psychiatric units. Two hospitals were used to improve the generalizability of findings and to provide a baseline data set for a further study that will allow one hospital to act as a control site while the other receives specific interventions aimed at optimizing PRN practices. All admissions were drawn from a three-month period from December 1998–February 1999.
For inclusion, a primary diagnosis of a psychotic disorder was required (schizophrenia, schizoaffective disorder, schizophreniform disorder, bipolar disorder, severe depression, brief psychotic episode, psychotic disorder secondary to substance use, psychotic disorder secondary to a medical condition and psychotic disorder not otherwise specified). Where the medical record coding diagnosis did not correlate with the clinical team's diagnosis, the latter was used. Depressed patients were only included if their illness was at the severe end of the depressive spectrum, (i.e. characterized by melancholia/agitation/psychotic features).
We audited 85 consecutive admissions from Hospital 1 and 99 from Hospital 2. Each audit involved a thorough examination of clinical records including admission notes, medication sheets and emergency department notes relevant to that admission. JG interpreted clinical notes commensurate with his psychiatric training, and all other aspects of the audit were within LS's pharmacy expertise. The data collection tool 1 designed for the audit included sections on demographics, treatment settings, diagnoses, regular medication (if altered during the admission the most consistently used doses were recorded) and morbidity associated with treatment. The section on PRN medication recorded the treatment orders (drug, dose, indications for use, maximum daily dose), administrations (mean number and dose) and outcomes (indications for use and response). As most PRN psychotropic medication is given early in treatment, audit of PRN medications was limited to the first two weeks of each admission.
The data were analysed using Access© and SPSS© version 9.0 software packages. Dose equivalence estimates were obtained from the relevant literature [26–28]. Continuous variables were examined using one-way ANOVA, t-test or Pearson's correlation, with logarithmic transformation applied to variables with non-Gaussian distribution. Higher–order interactions were evaluated using four-way ANOVAS that included the factors Axis I diagnosis, comorbid substance abuse/dependence, gender and age (three bands: ≥ 29, 30–45, 46 ≤). Post hoc analyses were conducted using Tukey's Honestly Significant Differences test. Categorical data were examined using χ2, Odds Ratio (OR) and Relative Risk tests.
Results
The demographic and diagnostic features of the sample are shown in Table 1. There were no significant differences between the samples drawn from the two hospitals for gender, age, primary diagnosis, length of stay or comorbidity. All subsequent data are therefore presented in aggregated form. Schizophrenia related disorders (60%) and manic states (18%) were the major psychotic diagnoses. Substance use disorders (27%) were the most common comorbid disorders.
Demographics and diagnoses in the two groups
PRN medication
PRN medications were administered during 82% of admissions, on a total of 1786 occasions (mean 11.8 administrations per admission). The total number and the mean daily doses of each drug averaged over length of stay are shown in Table 2. Of the antipsychotics, haloperidol was given most often and in the highest mean doses when expressed as chlorpromazine equivalents. Chlorpromazine was the next most commonly given PRN medication at a mean dose of one-third that of haloperidol. Atypical antipsychotics were rarely prescribed PRN. Of the benzodiazepines, diazepam was given most often followed by temazepam and clonazepam. Clonazepam was given in the highest diazepam equivalent doses. Some patients received more than one drug in each class.
PRN psychotropics administered during the first 14 days
Axis I diagnosis had a significant influence on whether PRN antipsychotics were administered (χ2 = 10.76, p < 0.05). Of patients with the diagnosis of mania/mixed affective psychosis, 82.35% received antipsychotics PRN. For patients with schizophrenia related disorders, major depression and substance-induced psychosis, the percentage receiving PRN antipsychotics was 66.67%, 46.67% and 45.83%, respectively. In contrast, diagnostic group had no impact on whether PRN benzodiazepines were administered.
Axis I diagnosis also significantly affected the daily dose of antipsychotic medication administered (chlorpromazine equivalents) (F = 5.73, df = 3, 116, p < 0.001). Post-hoc examination revealed that patients with mania/mixed affective psychosis (Mean Difference = 1.77, 5% CI = 0.44–3.09, p < 0.01, Tukey's HSD) and schizophrenia related disorders (Mean Difference =1.71, 95% CI = 0.47–2.96, p < 0.01, Tukey's HSD) received higher doses of antipsychotics than did patients with depression.
The daily dose of benzodiazepines (diazepam equivalents) was significantly influenced by Axis I diagnosis (F = 2.69, df = 3, 135, p < 0.05). Those with a diagnosis of mania/mixed affective psychosis received higher doses of benzodiazepines than did those with schizophrenia (Mean Difference = 0.66, 95% CI = 0.01–1.31, p < 0.05, Tukey's HSD). Patients with major depression and substance-induced psychosis received intermediate doses.
No differences in frequency of PRN antipsychotic medication administration were found for gender, age, or comorbid substance abuse. In contrast, older patients (r = 0.19, p < 0.05) and those with comorbid substance abuse (χ2 = 6.93; OR = 3.58, 95% CI = 1.28– 12.37, p < 0.01, Fisher's exact test) were more likely to receive PRN benzodiazepines.
Higher daily doses of antipsychotics were given to males (F = 8.58, df = 1, 118, p < 0.01), younger patients (r = −0.23, p < 0.05) and those with comorbid substance abuse/dependence (F = 5.32, df = 1, 118, p < 0.05). There was an interaction between gender and substance abuse for dosage of PRN benzodiazepines (F = 4.97, df = 1, 138, p < 0.05). Thus, female patients with comorbid substance abuse received greater doses of benzodiazepines than males with this diagnosis (median dosage: females = 11.07 mEq vs males = 5.54 mEq), whereas patients without comorbid substance abuse received similar doses regardless of gender (median dosage: females = 4.10 vs males = 4.75 mEq).
Most PRN prescriptions (87%) specified a maximum frequency or maximum total daily dose specification but very few (6%) specified indications for administration. Reasons for PRN administration were recorded for only 59% of administrations. The main reasons reported for giving PRN medication were agitation (49%), insomnia (17%) or psychotic symptoms (15%). Nearly two-thirds (64%) of administrations had no outcome recorded. Of the recorded outcomes, threequarters (76%) were reported as being partially or completely effective and the remainder as ineffective.
Regular medication
Regularly prescribed medication for psychoses are shown in Table 3. Antipsychotics (predominantly atypical) were most frequently prescribed, followed by benzodiazepines and mood stabilizers, followed by antidepressants and anticholinergics. When regular and PRN medications were combined, 76% of admissions involved prescription of two or more antipsychotics and 38% two or more benzodiazepines. Of those on regular atypical antipsychotics, 64% also received PRN typical antipsychotics. Pro re nata medications accounted for 31% of the total antipsychotic dose and 28% of the total benzodiazepine dose in patients who also received them by schedule.
Regular medication administered during the first 14 days
Medication related morbidity
Medication related morbidity was recorded in the clinical notes during 31% (57/184) of admissions, being noted in 37% (56/151) of those who received PRN medication and in only 3% (1/33) of those only receiving regular medication. Extra pyramidal symptoms were the most frequently recorded type of morbidity (28/57), followed by excess sedation/confusion (18/57) and autonomic disturbance (17/57). There was a significant association between the receipt of PRN medication and morbidity (χ2 = 14.69; OR = 18.69, 95% CI = 2.95–781.49, p < 0.001, Fisher's exact test). The association was apparent for both antipsychotics (χ2 = 28.07; OR = 11.87, 95% CI = 3.97–47.29, p < 0.001, Fisher's exact test) and benzodiazepines (χ2 = 11.00; OR = 4.78, 95% CI = 1.72–16.39, p < 0.001, Fisher's exact test). Morbidity was also associated with greater frequency of PRN administration (t = −3.55, df = 149, p < 0.001; Mean Difference = −0.57, 95% CI = −0.25 to −0.89). Patients who were administered haloperidol PRN were 5.61 times more likely (95% CI = 2.36–13.73) to have EPS (23/68) than those administered other PRN agents (5/83) (χ2 = 19.12, df = 1, p < 0.001).
Discussion
The results of the present study need to be interpreted with caution. While the retrospective nature of an unannounced audit has the advantage of not biasing the practices being audited (thus avoiding the Hawthorne effect), such clinical audits are dependent on clinicians accurately recording their activity. Other problems include reliability of diagnoses and outcome measurements that are not assessed with standardized instruments and by multiple raters, and the low recording rates of comorbid disorders and medication-induced sideeffects [29, 30].
In the present study, the sample was drawn from two inner-urban public hospitals in the same city and thus may not reflect practice in other inpatient settings. Another potential source of bias may have been the use of one researcher to interpret the clinical notes, and another to audit the medications. However, the audit protocols were designed to minimize such bias. The use of equivalencies for drug comparisons within a class is useful, but potency equivalencies do not take into account pharmacokinetic differences between drugs. Furthermore, side-effect profiles may vary in a way that is not well captured by potency equivalence [26].
Most findings were concordant with clinical experience. Schizophrenia and mania are common in public urban general psychiatry units. The mean length of stay for acutely psychotic patients was about three weeks (excluding outliers). Manic patients were more likely to receive antipsychotics than any other diagnostic group. They also received the highest doses of antipsychotics and benzodiazepines. The fact that manic patients received more PRN psychotropic medication than other groups reflects their disruptiveness, disinhibition, affective lability, impaired judgement and increased levels of activity.
A relatively low rate (27%) of comorbid substance use disorder diagnoses was recorded compared to 30–61% found in screening studies of similar populations [29]. This disparity may be due to poor recording, but probably reflects the underrecognition of comorbid substance use disorders in clinical settings [31]. Underlying drug withdrawal states (causing greater levels of agitation) and cross-tolerance to sedative drugs probably contributed to the higher doses of PRN antipsychotics and benzodiazepines given to patients with substance use disorders.
Pro re nata prescribing of antipsychotics was primarily of typical antipsychotics in contrast to the predominance of regular prescriptions for atypical antipsychotics. Benzodiazepines (especially temazepam) were more likely to be prescribed PRN than as a regular treatment. Although anticholinergics were prescribed widely on a PRN basis, mean doses were low, reasons for administration were rarely cited, and administration was often in conjunction with antipsychotics. This suggests anticholinergic administration was largely prophylactic.
Pro re nata prescriptions in the two study sites usually specified a maximum daily dose or maximum frequency, but as was observed in earlier studies [5, 7], very few prescriptions included indications for administration. With so few orders specifying indications for use, much was left to the discretion of nursing staff whose knowledge of pharmacotherapy and views on appropriate indications for giving medication can differ considerably from those of prescribers [32].
Of the one-third of administrations of PRN psychotropic medication where outcome was recorded, 76% were judged to be partially or completely effective. However, clinician bias, the use of multiple interventions (many of them non-pharmacological) and the frequent recording of brief and uninformative notes such as ‘given PRN with good effect’ mean that the reliability of this reported success rate is questionable.
It is important to consider the contribution of PRN antipsychotics to any observed medication-related morbidity. Typical antipsychotic doses equivalent to between 300 and 600 mg of chlorpromazine are considered effective in the management of acute psychosis, whereas higher doses of typical antipsychotics produce little benefit and escalating toxicity [26, 33]. In the present study, the mean daily dose of PRN haloperidol (at 217 mg chlorpromazine equivalent) represented a substantial contribution to the total antipsychotic dose, particularly as it was averaged over two weeks and most administration occurs in the first few days.
The relatively low rate of EPS (15%) observed could reflect the success of changes to scheduled treatment for psychoses. However, since prevalence rates of EPS are higher in studies that actively screen for EPS [36], it is more likely that a retrospective chart review simply underestimates the true prevalence. The more than fivefold increase in risk of EPS with PRN haloperidol compared with other agents is particularly disturbing. Common side-effects of low potency antipsychotics, such as, excess sedation, postural hypotension, dry mouth and blurred vision often occur at lower doses than are associated with either EPS or antipsychotic effect. However, the reported rates of these autonomic symptoms were too low to meaningfully examine the contribution of low potency antipsychotics to the development of this type of morbidity.
Less common side-effects were excess sedation (18 subjects) and falls/ataxia (5 subjects). Benzodiazepines may have contributed to the development of these effects although these types of morbidity are typically multigenic in origin.
The results of this audit, in conjunction with the evidence for best practice in the management of psychosis reviewed above, suggest that despite recent improvements in clinical practice, it is possible to further reduce medication related morbidity. One approach to reducing the need for PRN psychotropic medication is to anticipate that agitation is likely to occur in psychotic patients. Since benzodiazepines are an effective and relatively safe way of managing acute psychotic agitation, and many patients will also have sedative cross-tolerance and/or symptoms of substance withdrawal, it may be prudent to use moderate doses of scheduled and PRN benzodiazepines early in treatment that can be tapered after a few days. There is evidence to suggest that this approach is ‘antipsychotic drug sparing’ and reduces EPS [34]. It also ensures that doctors are more directly responsible for choosing medication and dosage. Multidisciplinary education about rational pharmacotherapy and psychosocial approaches to managing agitation may also help reduce the burden of medication-related morbidity.
Conclusions
Pro re nata psychotropic medications are a significant component of the treatment that psychotic patients receive during the acute phase of illness. Polypharmacy and a failure to specify indications for use or record outcome of administration are common problems with PRN medications. Medication related morbidity is strongly associated with PRN administration.
Psychiatrists have embraced the use of atypical antipsychotic medications for treatment of psychoses. However, the usefulness of atypical antipsychotics as PRN medications remains an open question. If the potential of atypical antipsychotics to improve long-term outcomes is to be realized, more attention needs to be given to what is added to them during acute treatment. The production of EPS by PRN use of typical antipsychotics during the acute phase has the potential to exacerbate subsequent non-compliance. On the other hand, benzodiazepines are relatively safe and effective first-line PRN agents, which avoid the problems of antipsychotic polypharmacy and EPS.
Footnotes
Acknowledgements
We thank the Royal Australian and New Zealand College of Psychiatrists, the Princess Alexandra Hospital Research & Development Foundation, the Lions Medical Research Foundation, the Australian National Health & Medical Research Council and the Danish Research Academy, the Royal Brisbane Hospital, Princess Alexandra Hospital Divisions of Mental Health for their cooperation and participants in the study.