Background
During the last few years, an increasing number of care models have been established worldwide which offer home treatment, case management and multidisciplinary care to persons who are severely mentally ill. Available evidence suggests that home treatment provided to the severely mentally ill by multidisciplinary psychosocial intervention teams has the potential to be effective with regard to suicide prevention, promotion of patient satisfaction and the need for inpatient treatment [
1‐
3].
Also in Germany, models that offer some form of home treatment within the framework of integrated care are becoming more and more popular [
4]. Prior studies have shown that the effectiveness of home treatment programmes is related to the structures and processes of these programmes [
5]. Factors revealed to be relevant to patient outcomes in home treatment include the caseload per case manager, regular and frequent home visits, accountability of home treatment programmes for medical and social issues, multidisciplinary teams that include psychiatrists [
6], and burnout levels of staff [
7]. Evidence is still scarce as to what extent these and other structures and processes in home treatment contribute to its effectiveness. The research presented here aims to identify structures and processes in home treatment of severely mentally ill persons that probably impact on patient-related outcomes and that are beyond the central components of the intervention such as the existence of case management, availability of 24 h crisis hotlines and home visits.
Methods
The structures and processes of 17 regional home treatment networks that provide similar, but nonetheless, individually different types of home treatment were analysed. These regional care networks called the German Network for Mental Health (
das deutsche NetzWerk psychische Gesundheit (NWpG)) were initialized in 2009 by the Techniker Krankenkasse (TK), one of the largest statutory health insurance companies in Germany, with more than nine million insured persons. By providing home treatment as well as case management to the severely mentally ill, the NWpG’s aim was to reduce the need for mental health inpatient care. Only patients who are insured with the TK and who fulfil certain criteria in respect to their course of illness are eligible for being treated on this programme. The TK selects patients for the programme through an algorithm applied to health insurance claims data [
8]. This algorithm filters those patients carrying the highest risk for hospitalization due to mental illness. The risk for hospitalization is assessed by predictive modelling using among others data on use of inpatient care for mental disorders and psychotropic medication as well as receiving the diagnosis of mental disease in the past year(s). The TK contacts patients that score lies above a certain threshold in predictive modelling and proposes them to enrol in the programme. In 2013, NWpG networks existed in 11 of 16 federal states with a focus in northwest Germany; in 2015 they are present in more than 25 regions. Except for one network all networks operate in urban and rural areas. The time that networks were existing varied between 19 months prior to data collection (October 2013) and 4 years. In 2012 there were more than 9,000 TK insured persons that had signed up for the programme since it started in 2009.
All networks provide similar core services that include: home treatment and case management, sociotherapy, psychoeducation, a 24 h crisis hotline and crisis intervention apartments. Apart from these available services that are standard, the networks are free to choose how they organize themselves, e.g., which professionals they employ (nurses, social workers, psychologists, patient experts), who steers the case management (psychiatrist or social worker), with what other local services they cooperate and to whom they propose specific interventions such as sociotherapy.
Study design and data basis
For this comparison, network data was derived from three different sources:
1.
Routine assessments of psychosocial functioning
2.
Questionnaires on structures and processes applied to the networks
3.
Health care claims data for the patients enrolled on the NWpGs.
1.
Routine assessment of psychosocial functioning was performed under contractual obligation at admission and every 6 months thereafter. The assessment was done by the patient’s case manager, who was unaware of the purpose of this study. Psychosocial functioning was assessed by means of the German version of the Health of the Nation Outcome Scales (HoNOS) [
9]. It is a third party assessment tool, with scores ranging from 0 to 48 points; thereby higher scores indicate more impairment in psychosocial functioning. HoNOS data was linked to the patient’s claims data via pseudonymized patient IDs. Relative change in the HoNOS over time was the outcome.
2.
A questionnaire on non-standardized structures and processes of the networks was completed by the network managers (for network questionnaire see Additional file
1). Thereby, overall characteristics of the networks were assessed, such as the number of staff, the staff / patient ratio, the staff professions and outside cooperation partners of the networks. In addition, a further questionnaire was completed by the staff involved in direct patient care (for staff questionnaire see Additional file
2). It addressed issues like time and frequency of home visits, time spent with patients but also job satisfaction and psychosocial stress in the work place. This questionnaire referred generally to the status of the last 3 months prior to completion of the questionnaire. Both questionnaires were completed between November 2013 and January 2014. We developed the questionnaires based on a systematic literature review and a Delphi like discussion process with each network on the structures and processes they believed to be relevant for the quality of care that the network provided [
10].
3.
Health care claims data was finally used to include information on patient demographics, use of inpatient treatment, medication, somatic and psychiatric diagnoses and use of outpatient care. The data was available for each patient starting from the year prior to enrolment in the NWpGs up until June 2013. Since not all the patients enrolled at the same time, patient information was restructured according to the individual date of enrolment (t0). Patients were anonymized by the TK making it impossible to trace their identity.
For data analysis the results of the staff questionnaires and also patient information through claims data were aggregated at network level. Thus, they became one of the “characteristics” of the networks. Linking this data to the respective networks was possible via network ID.
In summary, the resulting data collection included information on the following:
Network names are not published. They are numbered consecutively in a descending order according to the number of enrolled patients from 1 to 17. The ethics commission of the State Medical Chamber of Lower Saxony (Ethikkommission bei der Ärztekammer Niedersachsen) ruled that an ethics approval was not required since patients were not directly involved in the study.
Patients
All patients aged 18 years or older who were treated from 2009 to July 2013 by one of the 17 networks for more than 6 months were eligible subjects to this study (n = 7,243). Of these patients about half had to be excluded due to incomplete HoNOS data: 2,174 patients lacked a HoNOS score at either t0 or t1; most likely due to delays in data transfer from the networks to the health insurance and further on to the external managed data base where data was stored. In addition, 1,502 patients had to be excluded because of too many missings in one of their HoNOS scores. 3,567 patients remained; who’s HoNOS scores at both t0 and t1 were complete and could be included in the study. Patients included in the study and those excluded, did not differ significantly by gender or diagnoses. Patients included were slightly older (46.1 years) as compared to those excluded (45.3 years).
Outcome
The intended outcome of the analyses was the clinical improvement of patients over time, as represented by the relative change in the HoNOS between t0 and t1.
Statistical analysis
To determine whether and which network structures and processes induce a more positive patient- related improvement, a univariate linear regression analysis and multilevel regression was applied [
14]. In both analyses patients’ improvement in the HoNOS is presumed to originate from structures and processes of the network to which the patient is associated. With the univariate linear regression each independent variable was measured individually against the dependent variable.
In the multi-level analysis the relative change in the HoNOS between t0 and t1 served as the patient dependent variable that was explained by variables such as age, sex and comorbidity at patient level (1st level) and network characteristics at network level (2nd level). For all statistical procedures SPSS 21 advanced statistics were used. Statistical significance was defined as p < .05 for all analyses.
Discussion
This study focuses on the impact of structures and processes, other than the central standardised components of the home treatment model under investigation (case management, 24 h crisis hotline and home visits), on patient outcome. Using a naturalistic approach, 17 home treatment networks were studied in routine care.
In essence, beyond certain characteristics of the staff (experience in mental health care and the amount of effort put into their work), none of the non-standardised components in home treatment seemed to matter much in respect to patient outcome after 6 months of treatment. Probably the standardised components of the home treatment networks, in particular the fact that home treatment was provided at all, have so much more effect during these first months, that it is of minor importance as to whether the networks contract with a lot of outside service providers, or whether their personnel spends a greater share of its daily working time with patients. However, some of the non-standardised interventions, such as psychoeducation, providing a treatment plan and sociotherapy, which is a specific intervention for supporting the participation in social and professional life, showed a trend, albeit non-significant, towards being correlated with better patient outcome. If conclusions should be drawn from these results, it should be that it is probably better to offer these interventions, than not to have them at all.
The fact that the following were associated with less improvement to patient outcome: more patients per case manager, a higher number of home visits per patient, a higher share of patients whose family was contacted, more face-to-face contact and also more contact in general to patients, should be interpreted by considering that the applied regression analysis does not show causal relation. Therefore, these results probably show that sicker patients are receiving more attention and interventions by staff, e.g., more home visits, more family intervention, and more contact and care in general. Thus, the home treatment networks do in fact what they are supposed to do. This is, given the fact that it is a naturalistic study, a promising result.
If we had to make cautious recommendations on the basis of this research, it would be that home treatment overall, as a visiting mental health service might be recommendable because also in a routine care it succeeds in providing a higher intensity of care to those that are sicker and after 6 months of treatment patients are usually better off in respect of psychosocial functioning. For better patient outcomes it would be advisable to have staff in networks who are highly experienced in mental health care and who are ready to put a lot of effort into their work. Moreover, it would probably also be better if staff were satisfied with their income (a slight, however non-significant trend towards this finding). Furthermore, when considering other interventions beyond the mere home visits, it could be useful if networks provide sociotherapy, psychoeducation and treatment plans. Evidence from this study is not strong enough to state the absolute necessity of these three interventions. However, these interventions are also considered part of the treatment model by other home treatment models, such as the Functional Assertive Community Treatment (“FACT”), a Dutch version of assertive community treatment [
15,
16]. It is recommended by treatment guidelines [
17‐
20]. Furthermore, sociotherapy, psychoeducation and providing treatment plans (implying that a treatment plan is in fact present and developed in cooperation with the patient) are all measures to assist patients. This highlights their potential to contribute to better patient outcomes [
17].
Strengths and limitations
It should be considered as one of the strengths of this study that it is a naturalistic study exploring the effect of components of home treatment in a real world setting. A further strength relates to the outcome that considers the psychosocial functioning of the patient instead of using the use of inpatient services as an outcome only. The use of psychiatric inpatient services that is often taken as a surrogate for the wellbeing of patients is not only determined by the health of the patients but also by the availability of in- and outpatient services in a region [
21].
Overall, the psychosocial status of the patients improved across networks, which was measured by an average decrease in the HoNOS of .84 points. This improvement is in line with findings of controlled studies indicating the general effectiveness of home treatment [
22]. However, from a clinical point of view, this change in the HoNOS cannot be considered as being particularly strong [
23]. Clinically, a change in the HoNOS of 4 points during a whole year is regarded as minimal improvement.
Unfortunately, we could include fewer patients with available HoNOS data than desired: Firstly, the low number of HoNOS data is mostly caused by the network’s short period of operating. Networks that started their services in 2012 did not have a large number of patients who were enrolled long enough to have to complete HoNOS data both for t0 and t1 (after 6 months). Secondly, about half the patients that were treated long enough to have their data considered in the study could in fact be included. Most time a delay in data transfer is assumed as the reason for missing data. But also the high number of incomplete filled HoNOS questionnaires points to that networks – despite being contractually obliged to assess HoNOS – were not well prepared to do so. The decision to assess HoNOS routinely was taken by the TK long before this research project was even thought of. However, it shows that only obliging service providers to fill an assessment tool without an evaluation plan and feedback system produces poor assessment compliance. However, these technical problems with HoNOS collections seem not to have caused a considerable bias, since included patients do not differ from excluded patients and the study population can still be regarded as representative.
The fact that the networks were similar on the central and standardised components of care and differed only slightly in structures and processes beyond central components, may explain why multivariate analysis was not successful in explaining differences in outcome between the networks. In addition, the observation period of only 6 months was probably too short for factors less prominent than the mere fact of providing home treatment.
Our findings need to be interpreted by also considering the fact that the applied method only presents correlations but no causal relationship. Finally, we analysed the data at patient level by disaggregating the network level information. This might have biased our results. It would have been better if we could have used more information related to the individual patient, such as how many hours of face-to-face contact were received. However, such data was not available. A further problem relates to structures and processes in the networks being measured by self-measurement of the professionals. While items such as “number of patients per case manager” relies on clear data that is available in the administration of the networks, item such as “average number of hours of face-to-face contacts with patients during the last month” should be based on the own documentation of the case managers and thus might be subject to bias.
Conclusion
The home treatment networks observed in this study succeed in providing a higher intensity of care, including family intervention to the more psychosocially impaired patients and real life routine conditions. The fact that home treatment is provided at all seems to be more important in the first 6 months of treatment than details of what treatment was provided. However, there are signs indicating that being treated in a network that provides sociotherapy, psychoeducation, and provides patients with a treatment plan might be of some relevance to patient outcome. These findings might become more evident if treatment were observed for a longer follow-up period. Future research should consider this. Finally, this research suggests that for improving home treatment networks, it is advisable to invest in staff and employ highly experienced staff who are satisfied with their pay and who are ready to put a lot of effort into their work.
Acknowledgements
We would like to thank all networks for participating in this study and the Techniker Krankenkasse for providing the data. We are thankful to Sharon Janicke for editing the English.