Open Doors by Fair Means: a quasi-experimental controlled study on the effects of an open-door policy on acute psychiatric wards

Psychiatric wards with involuntarily hospitalized patients due to endangering themselves or others are usually locked to prevent absconding. Compared to other coercive measures such as seclusion or restraint, a closed door appears to be restricting personal freedom less. However, locked doors have considerable negative impact in terms of stigma and ward atmosphere [1]. Moreover, locked doors are often hard to justify as only a minority of patients is actually prone to absconding. For this reason, in Austria it is prohibited to treat voluntary patients on wards with locked doors. As a consequence, locked wards are nearly non-existent there [2]. In Germany, approximately 10% of the patients being treated in psychiatric acute care units, most of them locked, stay there involuntarily due to a court’s decision [3]. The number has even increased over the past years [4]. Although locked doors are often justified by the fact that involuntarily committed patients should be prevented from leaving the ward, doors in most countries usually remain locked even when no patient is being committed involuntarily [5].

In a Swedish study, the authors found that 19% of the wards were locked without a patient being committed involuntarily. Conversely, in case of 19% of the wards that were stated to be open at least one patient was committed involuntarily [5]. Similar to most other developed countries, the preconditions of involuntary commitment in Germany are the presence of a mental illness as well as the risk of imminent personal injury to self or others and mental aberration. Similar to most other developed countries, the preconditions of involuntary commitment in Germany are the presence of a mental illness as well as the risk of danger to oneself or others and causation by the respective mental disorder. All decisions on involuntary commitment must be made by a court. Only in acute cases patients may be detained preliminarily until the end of the day after admission. The Mental Health Law for the Federal State of Baden-Wuerttemberg provides detailed regulations for the use of any coercive measures. According to clinical judgement, unaccompanied leave is possible for patients who are treated involuntarily, too. Hence, the legal conclusion is drawn in practice that psychiatric wards must not be generally locked, if absconding can be prevented otherwise [6]. This development of legislations, that is in accordance with demands of service user organizations, psychiatric societies, and the UN Convention on the Rights of Persons with Disabilities, causes an urgent need for sound empirical research. Up to now, there is preliminary evidence from several cross-sectional and pre-post studies, predominantly from Switzerland [7, 8] and Germany [9, 10], that we had reviewed in a previous paper [11]. These studies suggest that suicide attempts, suicides, violent acts, and absconding are not more frequent in wards with an open-door policy.

Within recent years, observational evidence suggested that wards with an explicit open-door policy did not perform worse than locked doors in terms of preventing suicides and violent acts. However, fairness of comparisons and selection bias have been an issue of discussion [12]. Nevertheless, since 2015, several updated Mental Health Laws of Federal States in Germany basically enable treatment of involuntarily committed patients on open wards, if absconding is prevented by other appropriate means. For instance, the Mental Health Law of North Rhine-Westphalia states that”hospitals have to ensure by appropriate measures that the patients concerned do not withdraw themselves from the commitment” [13]. In a large observational study, data reported from 21 German hospitals with almost 350,000 cases over the course of 15 years found that suicides and suicide attempts, as well as absconding, did not occur more frequently in wards with open doors [7]. Limiting the conclusions which can be drawn, it has to be noted that data was collected retrospectively and the definition of open, semi-open and locked wards was fairly imprecise. Further doubts remain whether the definition of ‘open-door policy’ was precise enough, ranging from an informally expressed attitude to objectively open doors for 24 h on 365 days [11]. Prospective controlled trials with previously defined outcomes are missing.

The reason for the widespread use of locked wards may be due to treatment culture, control attempts and uncertainties [4, 14], but is by no means a therapeutic intervention. It relates primarily to the fear of involuntarily treated patients absconding and at worst committing suicide or harming themselves or others. However, between 50–85% of all suicides by psychiatric in-patients occur during a permitted leave [15]. Also, with regard to aggressive incidents and absconding, previous findings hint on the possibility that the opposite could be true: locked doors might increase the risk of absconding, of violent conflicts between patients or between patients and staff and further might reduce treatment compliance [9, 16].

Table 2 presents the mean number of patients present per day and the mean number of admissions per day. Regarding the total number of patients, no significant differences were found between baseline (Mdn = 19) and intervention period (Mdn = 19), U = 255,971.5, p = 0.160 nor between control (Mdn = 19) and intervention group (Mdn = 20), U = 436,127.5, p = 0.101. Regarding the number of daily admissions, no significant differences were found between baseline (Mdn = 1) and intervention period (Mdn = 1), U = 271,620.0, p = 0·671 nor between control (Mdn = 1) an intervention group (Mdn = 1), U = 421,167.0, p = 0.627. In summary, no significant differences between intervention and control wards were found, suggesting that the alternating allocation of admissions between the wards was successful in this aspect.

To our knowledge, this is the first prospective controlled trial with a quasi-experimental design for the implementation of an open-ward policy on acute care units. With over 3,000 treatment cases from two German psychiatric hospitals, our data is based on solid grounds. We managed to allocate the patients in a strictly alternating manner, ensuring the quasi-experimental design of the study. At both locations, mainly patients with psychotic or severe affective disorders were admitted to the acute care units. However, distributions of primary diagnoses differed significantly between each sites´ control and intervention wards. There is no clear explanation for the fact that particularly the number of patients with F2 diagnoses differ in this way. We assume that on the one hand the survey period was rather short and that the differences would have leveled out somewhat with a longer period of time. Even if we could not find any systematic evidence of this, it is possible that, especially outside the core working hours of the ward staff, i.e. in the evenings, at night and on weekends, or when several patients were admitted at the same time, patients were distributed more according to fit or custom. However, given the successful alternating allocation, this effect cannot yet be sufficient to explain this.

Regarding our primary outcomes, days with open doors increased 2.4-fold from 33.5% at baseline to 80.8% during intervention period in Friedrichshafen and 1.4-fold from 21.5 to 30.7% in Tuebingen with significant differences when compared to the control wards, respectively. The percentage of involuntary treatment days with open doors, which to our mind is the most robust indicator of an open-door policy, increased 3.9-fold from 23.2% at baseline to 90.7% during intervention period in Friedrichshafen and 1.9-fold from 16.1 to 30.1% in Tuebingen with significant differences when compared to the control wards. However, the control wards showed increased opening times, too, indicating potential spill-over effects between the wards. It might be possible that staff on control wards tried to demonstrate they could be as successful with open doors as their neighbored intervention wards.

Considerable differences between the sites at baseline make the interpretation of our findings more difficult but underscore the importance of conducting such a kind of study at more than one site. Baseline differences probably reflect longstanding attitudes and treatment practices within ward teams rather than differences between patient samples. As a matter of fact, the intervention was effective at both sites and there was no evidence of a ceiling effect, meaning that no further opening would have been possible in case of wards with a pre-existing open-ward policy.

However, even if the intervention was clearly effective, the intervention wards remained far from a 100% open-door practice which had been reported in the literature anecdotally [4, 11]. Reasons to keep doors locked were mostly patients with poor communication skills due to illness, disorientation, or acute suicidality. Establishing reliable relationships that allow agreements with those patients might be particularly difficult and staff for 1:1 supervision was not sufficiently available. Furthermore, some patients might consider continuous personal observation as more intrusive than a locked door. Consequently, we doubt that a 100% open-door policy can be realized on acute care wards with a considerable percentage of involuntary patients “by fair means”, i.e. without using other coercive interventions to prevent absconding. Differences between the two intervention wards show that criteria for locking the door in terms of safety reasons were differently specified, which cannot be explained sufficiently by differences of the patient sample. Probably not accidentally, the ward that was already more familiar with an open-door policy before the study achieved considerably higher rates of open doors. But, notably, this ward had received an additional nursing staff person for the time of the intervention, and for reasons beyond our control, no further nurse could be hired in Tuebingen. Instead, other assistants, of which the Tuebingen clinic as a teaching hospital has a larger number compared to Friedrichshafen, were involved. Although it is a topic of discussion that increasing the personnel ratio could contribute to reducing restrictive measures [18, 19], we cannot determine which single factors were the most relevant for a successful open-door policy.

One of the most important interventions at the patient level seems to be making individual arrangements that take into account the need for freedom and autonomy, including unaccompanied or accompanied leave in particular, but also more presence of staff in the entrance area and on the ward corridors [18, 19]. Furthermore, factors such as participation in decisions and acceptance of responsibility on the part of the patient are helpful to find a common understanding of the prerequisites of an open-door policy [20].

In terms of safety, suicides or violent attacks due to absconding did not occur during the data collection. We had registered aggressive incidents, suicide attempts, and absconding in the participating wards at baseline and during intervention period. Baseline data between wards differed significantly, which might be caused by single patients, considering the rather short period of three months. During intervention period, we did not observe a clear pattern of safety-relevant events. The number of adverse events mostly decreased on the intervention wards. Further, coercive measures, that were not allowed to be used to prevent absconding, occurred with roughly the same frequency in all participating wards during intervention period. That means that an open-door policy is not necessarily accompanied by more coercive interventions, as suggested by a recent study from four hospitals in Germany [21], but neither is there enough evidence to claim the opposite [11].

Our study has considerable strengths and some weaknesses. To our knowledge, this is the first study that investigated an open-door policy as an intervention in a prospective, controlled design on psychiatric wards treating involuntarily hospitalized patients. We did not only register the primary outcomes of open doors but also numerous other variables that can be considered as important control variables such as the percentage of treatment days with involuntary patients, absconding, suicide attempts, aggressive incidents, and the use of coercive measures. For the first time in the literature, we used a primary outcome, the percentage of involuntary treatment days with open doors, that we consider as the most conservative criterion of an open-door policy. Furthermore, a strength of this study is that it was conducted at two different sites. Like in many other studies with a small number of participating units, some of the results might be due to chance only and should not be interpreted as causal. For instance, absconding occurred more frequently in case of the intervention ward at one site but less frequent at the other. Moreover, the study at two sites allowed to compare the effects on wards with hospitals already applying open-door policies or not.

Our study has also several limitations. Whereas it included a detailed survey of some thousand admissions, the number of compared wards is relatively small. Robust evidence would require an RCT comprising dozens of hospitals which seems very difficult to achieve. Furthermore, our quasi-experimental design with alternating admissions on intervention wards and control wards had its limitations. From the presented data as well as from additional interviews, we observed considerable spill-over effects between the neighboring wards regarding the intervention. Staff on control wards started competing, resulting in an increase of open-door times also in case of the control wards. We found that the implementation of an open-door policy was successful, but we cannot determine which elements of the complex intervention were most effective. Notably, the ward that received an additional nursing staff position was particularly successful, but we do not know whether this was the reason. Eventually, we could not ascertain that secondary outcomes such as aggressive incidents were registered exactly in the same manner at both sites. Further studies on this issue are necessary.

In this prospective study with a quasi-experimental design, we could show that it is possible to manage psychiatric wards with open doors without taking inappropriate risks. The extent to which an open-door policy can be implemented appears to be related to factors such as staffing, patient characteristics and attitudes that require further investigation.