The novelty of the present study is its ability to provide the currently available most complete picture on the socio-economic burden and outcomes on young patients with recent diagnosis of schizophrenia or schizophreniform disorder, observed in a context of real clinical practice.
The study showed how these patients received several and complex treatments to manage their condition, using the many options available in the healthcare system. In particular, many antipsychotic drugs and concomitant therapies were used in different combinations. As regards the antipsychotic drug treatment, we found a higher persistence for the treatment with atypical antipsychotic drugs (more than 400 patient-days) than for typical drugs (263 patient-days). In this study we saw also how clinical and perceived health can change (improve) in nearly one year, according to the treatment strategies adopted. The full sample showed an overall improved trend in both clinical and perceived health (i.e., HRQoL): however, this improvement was on average higher among the naïve patients, i.e., those patients that were seen at the participating centres and started to receive health care starting from their enrolment in the study. Not only did clinical outcomes and HRQoL improve on average, but also the patients’ productivity. Furthermore, while no significantly different overall direct costs trends were found between naïve (i.e., patients that had started or changed antipsychotic treatment recently) and non naïve patients, naïve patients showed generally a significant mean higher improvement of clinical outcomes, HRQoL and indirect costs, compared to the others. However, the study sample was probably not suitable to obtain reliable results from these comparisons. Our results have a trend similar to that found in other studies, where treatment is associated with improvement in HRQoL [17, 18] and clinical outcomes. [19–21]. A similar trend was found also by Strakowsi et al., who showed that among patients with newly onset schizophrenic disorder, treatment is associated with improvements in most HRQoL domains, measured with the SF-36 instrument. Other studies show that outcomes in patients with a first-episode psychosis may be improved by an early intervention treatment and reduction of the untreated period [22, 23, 25–27] but long term effects are still not clear [26–28].
In our study sample, a quarter of medical costs are attributable to antipsychotic drug treatment, followed by costs for psychotherapy and for hospital admissions: these results are different from those obtained in previous research, where it has been shown that the highest proportion of direct costs is attributable to in-patient care, while drug treatment generates lower costs [7–10, 16]. However, it must be noted that the data available in the literature refer to different populations from different countries, where different treatment modalities could be applied, or different unit costs are applicable, or which include older patients than those involved in our study [6–11, 13–15, 51, 52]. In our study, medical costs during the follow up period remained stable overall. However, while some cost items increased (antipsychotic drug treatment, admission to health care units), others decreased (mainly hospital costs for symptom relapses), which could have depended also on the patients’ inclusion criteria at enrolment. Previous studies have demonstrated that loss of productivity is the main component of overall cost of schizophrenia [6–10, 13–15, 51, 52]. Our results confirm the high productivity loss among patients and their family caregivers, measured in terms of frequency of idle patients and in terms of days lost from productivity. However, we also observed a reduction of productivity loss during the follow up. This point is of great interest from the societal point of view, since schizophrenia is associated with poverty and homelessness, thus representing a significant amount of resource costs to deal with these problems. The National Institute for Health and Clinical Excellence (NICE) has recently recommended taking into account wider societal costs, including productivity losses of people with schizophrenia and of their family caregivers’, when cost-effectiveness of treatment is assessed . It must be noted that it is not easy to also include loss of productivity in the computation of costs , and this loss often remains excluded from calculations: for instance, Serretti and co-workers highlighted its importance, but did not include this parameter in their Italian simulation of the socio-economic burden of schizophrenia . Palazzolo et al.  pointed out that the combination of early diagnosis and the use of atypical medications would change the face of schizophrenia, allowing many patients to start working again. Our findings suggest that appropriate therapy reduces loss of productivity in a relatively short time period and that it may also promote an increment of the working/job inclusion of schizophrenia patients, improving in turn their social networks.
The treatment pattern observed in this study, which required several switches between drugs, demonstrates how treating patients with schizophrenia requires the availability of a wide and complex armamentarium of products (e.g., drugs) and services (e.g., psychosocial treatment, psychotherapy, etc.), likely involving different interacting factors, such as clinical severity, patient compliance, service accessibility . In this regard, the most recent NICE guidelines  report that choosing the most appropriate drug and formulation according to each patient’s needs and characteristics might be more important than only taking into account the main recognized properties of the different classes of drugs. In this way, NICE acknowledges the importance of a personalized treatment and highlights the importance of high levels of adherence to antipsychotic treatment to reduce the risk of relapse and further hospitalization costs.
Our study has some potential limitations: first, there was no control group in the present study, hence, we could not verify if the trends estimated with the different measures completed by the patients or the clinicians might actually represent a practice effect, i.e., an increase in the test scores from one administration to the next without any intervention. However, in a recent study aimed at investigating practice effects on a battery of scales, this was not found on the PANSS and on a visual analogue scale analysing the HRQoL . Second, although we could preliminarily estimate the possible incremental benefit attributable to the treatment applied on naïve patients, our sample size was too small to draw reliable results and conclusions. However, with regard to these two limits, this study was not designed for making comparisons between different patients, but rather to observe treatment strategies adopted in clinical practice and related consequences on patients’ health and costs to society. Third, the study may suffer from a selection bias, as patients who participated may be more likely to comply with medication than the overall target population. Fourth, our results on the higher persistence for atypical versus typical drugs classes could partially depend on the higher proportion of patients using atypical drugs, hence on a possible bias attributable to that these patients more probably switched to another atypical drug, which could not have been detected because we limited the attention on classes of drugs rather than on molecules. Nevertheless, the difference of persistence between the two classes of drugs appeared relevant. The approach we used to measure persistence in this study is not commonly applied and not comparable to those adopted in other studies conducted in this sector, where the observation was restricted to a limited number of drugs used (e.g., one out of 4 drugs in the CATIE randomized controlled study , one out of 7 drugs in the naturalistic study by Guo et al. ). However, because our study aimed at obtaining a real-world picture of our target population, we did not impose any criteria in regard to the treatment followed during the observational period. As a consequence, we enrolled patients that were using different combinations of many different antipsychotic drugs (18 in total), obtaining information that can actually be considered new for the community. Although not common, the approach of measuring persistence on classes of drugs can be useful to show the natural complexity of a treatment pattern, similarly to other areas, e.g., hypertension . However, because we did not observe persistence between molecules, or even dosages, within the same class of drugs, we obtained results that should be considered conservative. Nevertheless, our results already show how complex is the antipsychotic treatment even considering only classes of drugs, and suggest a much higher complexity that can be relevant for the treatment decisions and related consequences. Fifth, although we observed interesting and promising trends in a prospective1-year observation, this time horizon is however too short to know the long term consequences related to the strategy adopted in schizophrenic patients. Finally, some criticisms could arise with regard to the approach used to estimate direct and indirect costs. Regarding direct costs, we did not estimate non-medical costs (e.g. costs of transportation, housekeeping): we chose this approach because we considered the informative gain attributable to these costs, which we expected not to be relevant if compared to the other costs, not enough to justify the additional cognitive burden that would be caused by requesting more details from the patients. Indirect costs were not monetized: we chose this approach because of the high risk of underestimating indirect costs in a population where many patients are still students, many others are idle or have an unpaid job. We consider it more accurate and informative to describe and provide a picture of the study sample according to the patient’s educational and working status and to estimate the number of days that they and their caregivers miss work, school, or the possibility to do their usual activities.
As in the other studies focusing on costs e.g., [10, 59], our estimates are not totally applicable to other health care systems, because of the unavoidable country specificity of some data (e.g., unit costs) and methodologies used to conduct the study and perform the analyses. To expand applicability of our results, among the results we specify the mean consumption of specific categories of resources, which could be multiplied by different unit costs that are applicable in other health care systems. In any case, by keeping in mind the possible differences and adjustments to be made between the different healthcare sectors, these estimates remain valid to allow the community to understand the type and amount of the implications that are related to the management of subjects with the condition under study.