Predicting hospitalization following psychiatric crisis care using machine learning.

BACKGROUND: Accurate prediction models for whether patients on the verge of a psychiatric criseis need hospitalization are lacking and machine learning methods may help improve the accuracy of psychiatric hospitalization prediction models. In this paper we evaluate the accuracy of ten machine learning algorithms, including the generalized linear model (GLM/logistic regression) to predict psychiatric hospitalization in the first 12 months after a psychiatric crisis care contact. We also evaluate an ensemble model to optimize the accuracy and we explore individual predictors of hospitalization. METHODS: Data from 2084 patients included in the longitudinal Amsterdam Study of Acute Psychiatry with at least one reported psychiatric crisis care contact were included. Target variable for the prediction models was whether the patient was hospitalized in the 12 months following inclusion. The predictive power of 39 variables related to patients' socio-demographics, clinical characteristics and previous mental health care contacts was evaluated. The accuracy and area under the receiver operating characteristic curve (AUC) of the machine learning algorithms were compared and we also estimated the relative importance of each predictor variable. The best and least performing algorithms were compared with GLM/logistic regression using net reclassification improvement analysis and the five best performing algorithms were combined in an ensemble model using stacking. RESULTS: All models performed above chance level. We found Gradient Boosting to be the best performing algorithm (AUC = 0.774) and K-Nearest Neighbors to be the least performing (AUC = 0.702). The performance of GLM/logistic regression (AUC = 0.76) was slightly above average among the tested algorithms. In a Net Reclassification Improvement analysis Gradient Boosting outperformed GLM/logistic regression by 2.9% and K-Nearest Neighbors by 11.3%. GLM/logistic regression outperformed K-Nearest Neighbors by 8.7%. Nine of the top-10 most important predictor variables were related to previous mental health care use. CONCLUSIONS: Gradient Boosting led to the highest predictive accuracy and AUC while GLM/logistic regression performed average among the tested algorithms. Although statistically significant, the magnitude of the differences between the machine learning algorithms was in most cases modest. The results show that a predictive accuracy similar to the best performing model can be achieved when combining multiple algorithms in an ensemble model.

A prediction model for the incidence of civil detention for crisis patients with psychiatric illnesses; the Amsterdam study of acute psychiatry VII.

OBJECTIVE: Given raised numbers of civil detentions in the Netherlands and other European countries, it is important to assess the patient risk profile with respect to the incidence of those far-reaching treatment decisions. The aim of the ASAP study is to develop a comprehensive prediction model that considers all possible patient-related predictors known from earlier research. METHODS: We took a random sample of 252 from the 2,682 patients coming into contact with two psychiatric emergency teams in Amsterdam between September 2004 and September 2006. We recorded socio-demographic and clinical characteristics, aspects of social support and psychiatric history. We interviewed the patients using the Verona Service Satisfaction Scale (Verona-EU) and the Birchwood Insight Scale. During a two-year follow-up period we noted their use of mental health care facilities. RESULTS: Stepwise logistic regression analyses with resulted in a final prediction model (P ≤ 0.001) including preceding involuntary admission (OR 9.4, 95% CI 3.6-24.7, P ≤ 0.001), domestic situation alone (OR 4.5, 95% CI 1.9-11.0, P = 0.001) and VSSS score satisfactory (OR 0.2, 95% CI 0.0-0.8, P = 0.030) as predictors of civil detention during 2 years of follow-up. CONCLUSION: With the presented prediction model it will be possible to identify patients at a high risk of civil detention: patients with a history of previous involuntary admissions who live alone and are not satisfied with the mental health care they got before. This suggests the possibility that timely preventive measures can be taken comprising the adjustment or intensification of the treatment plan for this specific group of patients.

Police Encounters, Agitation, Diagnosis, and Employment Predict Psychiatric Hospitalisation of Intensive Home Treatment Patients During a Psychiatric Crisis.

Objective: This study aims to determine factors associated with psychiatric hospitalisation of patients treated for an acute psychiatric crisis who had access to intensive home treatment (IHT). Methods: This study was performed using data from a randomised controlled trial. Interviews, digital health records and eight internationally validated questionnaires were used to collect data from patients on the verge of an acute psychiatric crisis enrolled from two mental health organisations. Thirty-eight factors were assigned to seven risk domains. The seven domains are "sociodemographic", "social engagement", "diagnosis and psychopathology", "aggression", "substance use", "mental health services" and "quality of life". Multiple logistic regression analysis (MLRA) was conducted to assess how much pseudo variance in hospitalisation these seven domains explained. Forward MLRA was used to identify individual risk factors associated with hospitalisation. Risks were expressed in terms of relative risk (RR) and absolute risk difference (ARD). Results: Data from 183 participants were used. The mean age of the participants was 40.03 (SD 12.71), 57.4% was female, 78.9% was born in the Netherlands and 51.4% was employed. The range of explained variance for the domains related to "psychopathology and care" was between 0.34 and 0.08. The "aggression" domain explained the highest proportion (R 2 = 0.34) of the variance in hospitalisation. "Quality of life" had the lowest explained proportion of variance (R 2 = 0.05). The forward MLRA identified four predictive factors for hospitalisation: previous contact with the police or judiciary (OR = 7.55, 95% CI = 1.10-51.63; ARD = 0.24; RR = 1.47), agitation (OR = 2.80, 95% CI = 1.02-7.72; ARD = 0.22; RR = 1.36), schizophrenia spectrum and other psychotic disorders (OR = 22.22, 95% CI = 1.74-284.54; ARD = 0.31; RR = 1.50) and employment status (OR = 0.10, 95% CI = 0.01-0.63; ARD = -0.28; RR = 0.66). Conclusion: IHT teams should be aware of patients who have histories of encounters with the police/judiciary or were agitated at outset of treatment. As those patients benefit less from IHT due to the higher risk of hospitalisation. Moreover, type of diagnoses and employment status play an important role in predicting hospitalisation.

The Amsterdam Studies of Acute Psychiatry I (ASAP-I); a prospective cohort study of determinants and outcome of coercive versus voluntary treatment interventions in a metropolitan area.

BACKGROUND: The overall number of involuntary admissions is increasing in many European countries. Patients with severe mental illnesses more often progress to stages in which acute, coercive treatment is warranted. The number of studies that have examined this development and possible consequences in terms of optimizing health care delivery in emergency psychiatry is small and have a number of methodological shortcomings. The current study seeks to examine factors associated with compulsory admissions in the Amsterdam region, taking into account a comprehensive model with four groups of predictors: patient vulnerability, social support, responsiveness of the health care system and treatment adherence. METHODS/DESIGN: This paper describes the design of the Amsterdam Study of Acute Psychiatry-I (ASAP-I). The study is a prospective cohort study, with one and two-year follow-up, comparing patients with and without forced admission by means of a selected nested case-control design. An estimated total number of 4,600 patients, aged 18 years and over, consecutively coming into contact with the Psychiatric Emergency Service Amsterdam (PESA) are included in the study. From this cohort, a randomly selected group of 125 involuntary admitted subjects and 125 subjects receiving non-coercive treatment are selected for further evaluation and comparison.First, socio-demographic, psychopathological and network characteristics, and prior use of health services will be described for all patients who come into contact with PESA. Second, the in-depth study of compulsory versus voluntary patients will examine which patient characteristics are associated with acute compulsory admission, also taking into account social network and healthcare variables. The third focus of the study is on the associations between patient vulnerability, social support, healthcare characteristics and treatment adherence in a two-year follow-up for patients with or without involuntarily admittance at the index consultation. DISCUSSION: The current study seeks to establish a picture of the determinants of acute compulsory admissions in the Netherlands and tries to gain a better understanding of the association with the course of illness and patient's perception of services and treatment adherence. The final aim is to find specific patient and health care factors that can be influenced by adjusting treatment programs in order to reduce the number of involuntary admissions.

Changing patient perspectives after compulsory admission and the risk of re-admission during 5 years of follow-up: The Amsterdam Study of Acute Psychiatry IX.

OBJECTIVE: Given increasing numbers of compulsory admissions, it is important to assess patient risk profiles and identify factors that are appropriate for intervention. METHODS: A sample of 116 patients who were compulsorily admitted was studied. Socio-demographic and clinical characteristics were recorded. Patients were interviewed twice using the Verona Service Satisfaction Scale and the Birchwood Insight Scale. Changes in insight and satisfaction during 2 years were linked to the incidence of involuntary re-admissions in the next three follow-up years. RESULTS: A higher mean score for patient satisfaction was found (mean=3.77, standard deviation (SD)=0.56; p⩽.001) in the second interview than in the baseline interview (mean=3.26, SD=0.65). There was also an improvement in insight (in the second interview: mean=7.22, SD=2.86 and in the baseline interview: mean=6.34, SD=3.18; p=.027). There was an inverse correlation between increasing satisfaction (in years 1-2) and the incidence of involuntary admission in years 3-5 (odds ratio (OR)=0.445, 95% confidence interval (CI)=0.252-0.793; p=.006). This association proved to be dependent on a history of involuntary admission. CONCLUSION: Increasing patient satisfaction in the first 2 years was associated with a lower risk of compulsory re-admission in the subsequent follow-up period, but this association proved to be dependent on a history of involuntary admissions in these first two follow-up years. Increase in insight during the same period did not show any effect whatsoever.

Patient perspectives and the risk of compulsory admission: the Amsterdam Study of Acute Psychiatry V.

OBJECTIVE: Compulsory admission to a psychiatric hospital is associated with a three- to fourfold increase in the risk of another compulsory admission. Given the rising numbers of civil detentions in The Netherlands and other European countries, it is important to understand the mechanism behind this association. Our aim is to study the links between opinions about prior psychiatric treatment, insight, service engagement and the risk of (new) civil detentions. METHODS: We took a random sample of 252 from the 2,682 patients consecutively coming into contact with two psychiatric emergency teams in Amsterdam between September 2004 and September 2006. We recorded socio-demographic and clinical characteristics, and information about prior involuntary admissions. We interviewed the patients using the Verona Service Satisfaction Scale (Verona-EU), the Birchwood Insight Scale and the Service Engagement Scale. During a two-year follow-up period we noted their use of mental health care facilities. RESULTS: Patients with a satisfactory score on the Verona-EU had significantly lower odds for civil detentions during follow-up compared to patients with a dissatisfactory score on this scale (OR = 0.3). Level of insight did not influence the risk of detention during follow-up. Furthermore, of the 131 patients admitted involuntarily the year before, one-third looked back on their involuntary admission with unambiguous satisfaction. CONCLUSION: More satisfaction with prior treatment seems to reduce the risk of civil detention remarkably. Low levels of satisfaction seem to be mainly dependent on a history of previous involuntary admission. These findings seem to open up a new perspective for diminishing the risk of (new) civil detention by trying to enhance satisfaction with treatment, especially for patients under detention.

Factors associated with higher risks of emergency compulsory admission for immigrants: a report from the ASAP study.

OBJECTIVE: Specific ethnic groups appear to be disproportionally represented in emergency compulsory admissions. This may be due to a parallel higher risk of psychopathology, but different pathways to care in patients from ethnic minorities may also be an explanatory factor. In this article we concentrate on the influence of ethnic background, pathways to psychiatric emergency services and the amount of past psychiatric treatment as predictors of emergency compulsory admission. METHODS: All the patients coming into contact consecutively with the Psychiatric Emergency Service Amsterdam (PESA) from September 2004 to September 2006 were included in the study. We collected socio-demographic and clinical characteristics, and data about prior use of mental health services and referral to the emergency service. RESULTS: The risk of compulsory admission for immigrants from Surinam and the Dutch Antilles (OR 2.6), sub-Saharan Africa (OR 3.1), Morocco and other non-Western countries (each OR 1.7) was significantly higher than for native Dutch people. After controlling for socio-demographic characteristics, diagnosis, referral pattern and psychiatric treatment history, this correlation was no longer found. CONCLUSION: We found no evidence to substantiate the hypothesis that ethnic background plays an independent role in emergency compulsory admission. Police referral rather than referral by a GP, and being diagnosed with a psychosis, seemed to be explanatory factors for the high risk of compulsory admission for non-Western immigrants. Infrequent contact with secondary mental healthcare during the previous five years was not found to be consistently associated with a higher risk.

Social support and risk of compulsory admission: part IV of the Amsterdam Study of Acute Psychiatry.

OBJECTIVE: Social support for patients with a mental illness has been associated with a lower rate of hospitalization. It is important to clarify the role played by a lack of social support as a possible predictor of emergency compulsory admission. METHODS: A random sample of 252 patients who were evaluated by two psychiatric emergency teams in Amsterdam from September 2004 to September 2006 were interviewed approximately one month later about their social networks and social interactions. The number of emergency compulsory admissions was recorded for 244 patients during a two-year follow-up period after the interviews. RESULTS: Patients who lived alone had a higher risk of compulsory admission (p≤.05) and had fewer people in their social network (4.6 versus 6.1, p≤.001) compared with patients who lived with others. Among patients who lived alone, the percentage of patients with a compulsory admission was significantly higher among the patients with a high score for negative interactions than among patients with a low score (34% versus 13%, p≤.05). CONCLUSIONS: Of the social support variables, living alone proved to be the only predictor of emergency compulsory admission and readmission, and patients who lived alone had a smaller social network. A high level of negative social interactions increased the risk of compulsory admission among patients who lived alone.

Involuntary admission of emergency psychiatric patients: report from the Amsterdam Study of Acute Psychiatry.

OBJECTIVE: This brief report presents initial data from the Amsterdam Study of Acute Psychiatry (ASAP-I) about factors associated with the decision to admit patients compulsorily (involuntarily) to emergency psychiatric services in the Amsterdam region of the Netherlands. METHODS: The study was a prospective cohort study of 1,970 consecutive patients who came into contact with the Psychiatric Emergency Service Amsterdam. RESULTS: A history of more than 14 outpatient contacts the previous year was associated with a low risk of compulsory admission (OR=.3). An involuntary admission in the previous five years was associated with a higher risk (OR=3.7). Referral by a general practitioner was associated with a low risk compared with referral by police (OR=2.4) or by mental health services (OR=2.3). CONCLUSIONS: The hypothesis that outpatient treatment may help to prevent compulsory admission found some support in this study. More research is needed to understand the mechanisms of the associations so that an intervention study can be developed to test this hypothesis.