Sampling

All 21 Italian regions, with the exception of Sicily, participated in the study. Each region designated a coordinator to organize and supervise data collection. Phase 1 surveyed the physical characteristics, staffing arrangements, admission rules, and activities of all public and private acute inpatient facilities [5]. In phase 2, a 20% random sample of GHPUs (stratified by Region) and all remaining public and private facilities surveyed in phase 1 were initially selected for a more in-depth study. Severe financial constraints at the time in the Lazio Region limited its participation to a survey of facilities available to provide data on a voluntary basis. At the same time, however, eight, not originally sampled GHPUs asked to participate and were included in the study.

Furthermore, organizational problems prevented the participation of six 24-h CMHCs, three UPCs, and 18 private facilities. Phase 2 therefore examined 130 public facilities and 36 private facilities.

The study sample comprised all inpatients with a length of stay of more than 90 days and no planned discharge ("long-stay inpatients"), and all inpatients staying for 90 days or less with a planned discharge ("short-stay inpatients"), in public and private facilities surveyed during an index period of 12 and 3 days, respectively. The shorter, 3-day index period was used for private facilities, because the National Association of Private Hospitals consented to patient recruitment and evaluation for only a limited number of days, due to time and work constraints in these facilities. Given that the study focused on factors associated with long- vs. short-term inpatient stays, patients admitted and discharged during these time periods were not considered for inclusion. The final sample therefore included 127 long-stay and 870 discharged (or planned discharge) inpatients.

Assessment

All regional coordinators were trained in the administration of the study instruments. In larger regions, additional research assistants participated in data collection under the coordinators' close supervision. A researcher visited each facility and completed standardized assessment for each patient included in the study. Socio-demographic information was retrieved from patient records. An ICD-10 primary diagnosis was assigned by the patients' treating physician. Detailed information on patients' behavior and psychosocial interventions received was also obtained. Assessment was based on the 24-item BPRS [7] and on the Personal and Social Performance scale (PSP). The PSP is a modified validated version of the Social and Occupational Functioning Assessment Scale (SOFAS) [8]. As with the SOFAS scale, PSP scale scores range from 100 (excellent functioning) to 1 (extremely severe impairment with risk for survival). With reference to long-stay inpatients only, treatment teams were asked to express their opinions on the reasons for delayed discharge; the responses were then coded into the following non mutually exclusive categories: presence of serious physical illness, lack of independent or staffed accommodation (placement not available after discharge approval by hospital staff), lack of economic resources, lack of close relationships and social support, risk of antisocial behavior, and lack of treatment response.

Data and statistical analysis

The first analysis used descriptive statistics (Student's t-test or Chi-square test, as appropriate) to examine the characteristics of long-stay vs. short-stay inpatients and to summarize the information about reasons contributing to the prolonged stay.

In the second stage, a multiple logistic regression model was developed to investigate relations among inpatients' socio-demographic and clinical characteristics, violent behaviour, psychosocial interventions received occurring during hospitalization, and long-stay status. The following criteria were used to classify psychosocial interventions: (1) 'psychotherapy' was considered any individual supportive, dynamic, or cognitive-behavioral psychotherapy; or any group dynamic, cognitive-behavioral, or psycho-educational therapy/intervention; (2) 'rehabilitative treatment' was any form of individual or group social skills training; and (3) 'family education treatment' referred to any type of information and counseling intervention aimed at providing patients and their family members with information on mental disorders and available treatments. 'Violent behavior' was considered any incident occurring during hospitalization, in which a patient had attempted to physically harm others (e.g., hospital staff members, other patients, or visitors) or to damage property.

A set of sociodemographic, clinical, and treatment-related variables were entered into the regression model as independent variables, and long-stay in-patient status (no vs. yes) was entered as the dependent variable. The Odds Ratio (OR) with 95% Confidence Interval (CI) was used as a measure of association.

Given that dichotomous variables yield limited information on psychosocial interventions, the proportion of inpatients receiving more or less than 5 sessions of any type of psychosocial intervention was also calculated and examined in further analyses.

Ethical issues

Ethical approval is granted for this study from the Italian Ministry of Health which approved and financed the PROGRES-Acute Project.

Reasons contributing to prolonged hospital stay and treatments received

Multivariate analysis

Variables associated with hospital long-stay

Several studies have previously reported length-of-stay associations with patient age [9] and gender characteristics [10]. Our results are consistent with these findings in terms of age, but not gender. It is possible that gender per se is not a key long-stay risk factor and that other confounding factors, such as the specific clinical characteristics of study participants and situational variables, have a greater impact on hospital stay duration [11].

Our study also did not replicate previous findings showing an association between long-stay status and diagnosis: based on ICD-10 diagnostic categories, and controlling for other variables, no significant differences between long-stay and short-stay inpatients emerged. We also observed no association with patients' severity in terms of psychopathology. Although severity of psychopathology has been found to influence length of stay [3], the issue of how best to develop valid and reliable ways to measure illness severity has always represented a challenge for the field [12]. Various serious proposals in this regard have emerged in recent years; our study used the 24-item BPRS, an expanded version of the 18-item BPRS, with defined scale points and probe questions, which has shown improved inter-rater reliability over previous versions.

An high percentage of short-stay patients was found among compulsorily admitted subjects. This is consistent with data found by other authors [11], and is probably due to the fact that a "short" duration of admission (up to 90 days in our study) was sufficient for controlling acute psychiatric symptoms, while long-stay status is determined by variables other than symptomatology (as shown also by the results of the multivariate analysis).

The other variables examined herein have also been investigated by other authors, and our finding that unmarried status is associated with long hospital stay is consistent with results from several previous studies [3, 4, 11].

Although we found that poor psychosocial functioning was positively associated with long-stay status, it is generally unclear whether more impaired patients have longer lengths of stay due to difficulties in planning discharges and in community resettlement efforts, or whether hospital stay leads to a loss of basic daily living skills and impairment in ability to function in social environments [13]. Of course, it is possible that both factors influence long-stay: patients with poor psychosocial functioning tend to remain in hospital longer, and this situation in turn facilitates the further loss of daily living skills, increases dependence, and weakens social networks.

An intriguing finding was that patients with higher education level (more than 8 years) were hospitalized for longer than those with lower education level. However, an high school degree (or beyond) was also found in the majority (60%) of long-stay patients in a North-American study [14], while in a German sample of inpatients subjects who had failed to attain a grammar school certificate were more readily discharged as compared to inpatients with an high school degree [11]. An hypothetic explanation may be that treatment psychiatrists had higher expectations of improvement about patients with higher education, and a longer stay of those patients was related to the tendency of these physicians to not discharge patients who still did not show a clinically significant improvement.

Length of stay and hospital setting

The strongest predictor of long hospital stay was admission to a private inpatient facility. It should be noted that this is the first national survey ever conducted in Italy to include patients admitted to private inpatient facilities. Indeed, very few investigations at an international level have examined inpatients admitted to private facilities. It can be argued that the tendency to longer stay in private inpatient facilities is due to specific differences in their functioning and staffing arrangements. In fact, private Italian hospital facilities, vs. public facilities, present several characteristics found to be associated with longer length of stay [5], such as lower staff-patient ratio, poorer coordination with community mental health services (although referral to community services may be associated with higher rehospitalization risk, or does not necessarily reduce the risk of rehospitalization [2, 15]), and a higher number of beds [16].

Risk of violent behavior and long-stay

Our finding that long-stay inpatients were more likely to display violent behaviors during hospitalization has also been reported previously [17]. Various explanations can account for this association: duration of hospital stay may be determined by the violence a patient exhibits. Alternatively, long-term hospitalization itself may contribute to the manifestation of violent behavior – perhaps due to patient frustration about a delayed discharge. Studies conducted in other countries have shown that violent behavior during hospitalization is most likely linked to factors other than long-term hospitalization.

Specifically, acute symptom severity and prior history of dangerousness appear to be the strongest predictors for disruptive behavior in psychiatric units [18, 19]. Patients in the present study frequently had their stay prolonged because they were judged to be at risk of antisocial behavior. Professionals tend to be cautious when placing patients who have exhibited violent behavior in less-supervised settings, in an attempt to avoid triggering events with undesirable consequences. Perhaps the team members in our study tended to carefully weight the risks of discharging patients into family environments, and these patients therefore remain hospitalized for longer periods. This finding has direct implications for the complex ward management of patients showing violent behavior, because they should receive more intensive management during their hospital stay. Yet, as observed elsewhere [20], Italian psychiatric units rarely provide intensive case management and direct behavioural forms of intervention – such as social skills training -which is specific to the post-discharge environment and helps patients better cope with stress and anger. This general scenario persists, in spite of wide-scale acknowledgment of the importance of the inclusion of these programmes in individual inpatient care plans [21].

Treatments and long-stay

The findings about process of care were more complex and difficult to interpret. On one hand, psychotherapies that can require prolonged treatments did not predict prolonged hospital stay. On the other, long-stay inpatients (vs. discharged patients), as expected, were more likely to have received rehabilitative intervention that can require longer hospitalization. In any event, it is worth noting that rehabilitative interventions were delivered to only 24% of these inpatients – a finding that is reason for concern, because 50% of them stayed in hospital for approximately 5 months, and 25% stayed for more than 11 months. Moreover, these findings raise the question of why these patients were not moved from hospital- to community-based settings, including residential facilities.

Staff members noted that many patients continued to occupy acute hospital beds due to the unavailability of appropriate alternative placements – a problem found to be linked to inappropriate bed use in acute settings, in other studies conducted in European countries with similar financing and health care delivery systems. The issue is pressing in the UK, where high proportions of 'new long-stay patients' tend to receive inappropriate, continued acute care, following the country's historical shift to community care [22]. Some of these patients merely require home-based community support, such as group homes. The situation is similar in Italy, where poorly suitable post-discharge accommodation inevitably places greater pressure on acute psychiatric beds. For example, just one out of six residential facilities occasionally admits patients with acute illness episodes [2].

Residential rehabilitative facilities should therefore be made available to patients requiring more intense levels of care. Yet, in Italy, beds in these facilities are not easily made vacant, as shown in a recent nationwide study on Italian psychiatric residential facilities [23], which found that resident turnover was extremely low. These findings point to the need for the development of sufficient and appropriate alternative care settings for patients admitted to acute psychiatric units.