The main finding of this study is that the prevalence of antipsychotic combination treatment increased with number of hospital admissions, severity of the disease as measured with PANSS, and level of dysfunction, as measured with GAF. The current finding that previous hospital admissions were related to antipsychotic combination treatment is in line with Kroken el. al  who found that in-patient treatment in the previous 12 months predicted polypharmacy. As seen from table 5, increase in number of hospital admissions beyond four did not seem to increase to probability of receiving two or more antipsychotics. We dichotomized the sample and chose a cut-off between one and two previous admissions. The reason for choosing this cut-off value was primarily the clinical relevant distinction between patients who were readmitted and those who were not. In a number of the patients first admittance to hospital was not necessarily due to problems with ongoing treatment, but due to a more acute or dramatic onset of symptoms of schizophrenia. Choosing a cut-off between one and two admissions therefore reflects to a greater extend patients with poor compliance or lack of response to ongoing treatment.
The cut-off points could just as well have been set between two and three, or between three and four previous admissions, providing slightly higher odds ratios. However, more than five previous admissions did not further increase the probability of receiving an antipsychotic combination treatment. Our results seem to be in line with studies that suggest a combination of antipsychotics as an option in non-responders with a higher degree of relapse, or in patients with more severe schizophrenia [5, 12, 19, 20]. The finding that the use of antipsychotics were mainly in accordance with guidelines up to the second admission to hospital, supports the hypothesis that antipsychotic combination therapy is more likely to be prescribed when treatment according to guidelines has not achieved an adequate therapeutic response.
Previous studies have not reported any significant correlation between prescription pattern and decline in global- or daily functioning, measured with GAF [9, 21]. This could be due to a type II error, at least in one of the studies, since this only included inpatients . The current finding of PANSS score versus combination treatment, has not been reported earlier. A higher level of current psychotic symptoms, as measured with total PANSS scores, further supports the hypothesis that antipsychotic combination therapy is more likely to be prescribed when guideline treatments have not achieved an adequate therapeutic response.
Our naturalistic sample of patients consisting of both inpatients and outpatients at the time of the examination, might be more representative for the population of patients with schizophrenia at various stages of the illness, providing a relatively wide spectrum in symptom levels and functioning and thus GAF and PANSS scores. This probably enabled us to detect important associations that are difficult to find in more selected groups of patients e.g. inpatients only.
Duration of untreated psychosis (DUP) was verified in only a portion of the patients but did not show any significant relationship to combination treatment with antipsychotics. This may be in line with our finding that age did not show any significant relationship with such treatment either. Future studies should further explore the role of DUP with regard to medication regimens.
The overall rate of antipsychotic combination treatment among our patients was comparable to other naturalistic studies [9, 12]. In our study SGAs were used more frequently as the preferred antipsychotic drug than reported from some European studies performed during the same time period [9, 11, 22, 23], but was in line with other study reports . The use of FGA as a primary therapeutic agent was relatively infrequent in our study. The variation in prescription patterns of SGAs may be attributed to both guideline adherence and how the public health systems work in different countries, including to what extent prescriptions of all antipsychotic medications are reimbursed by the social security program, as well as differences in the hospitals' financial schemes which influence the choice of low-versus high-cost medications.
Evidence-based guidelines for the psychopharmacological treatment of schizophrenia are important for securing a high quality of clinical practice including rational strategies to minimize adverse effects. However, the knowledge is rather scarce on how to guide treatment decisions in non-responders to antipsychotic monotherapy, which may be reflected by the lack of evidence-based recommendations for this group of schizophrenia patients. A better discrimination between subgroups of patients with different clinical courses of the illness is therefore needed when proposing new recommendations, moving today's guidelines with their "one size fits all" approach to antipsychotic medications closer to clinical practice.
The current body of evidence to support a combination of two or more antipsychotics in schizophrenia is not conclusive [20, 24–26], even though antipsychotic combination treatment may be superior to monotherapy in a limited number of patients [12, 26, 27]. A few randomized controlled trials have reported treatment with clozapine in combination with a second antipsychotic, to be superior to clozapine in monotherapy in subgroups of patients .
The current study involved all psychiatric hospitals in Oslo and included both in- and outpatients. The public health care service in Norway is good and provides adequate treatment for all psychiatric patients. There is no privately financed health care that offers long-term treatment for patients with schizophrenia, which enabled us to collect representative data on current treatment with a rather low degree of selection bias.