Limits...
Determinants of per diem Hospital Costs in Mental Health.

Wolff J, McCrone P, Patel A, Normann C - PLoS ONE (2016)

Bottom Line: Mixed-effects maximum likelihood regression and an ensemble of conditional inference trees were used to analyse data.Although reliable cost drivers were identified, idiosyncrasies of mental health care complicated the identification of clear and consistent differences in hospital costs between patient groups.Further research could greatly inform current discussions about inpatient mental health reimbursement, in particular with multicentre studies that might find algorithms to split patients in more resource-homogeneous groups.

View Article: PubMed Central - PubMed

Affiliation: King's College London, Institute of Psychiatry, Psychology & Neuroscience, King's Health Economics, De Crespigny Park, London, United Kingdom.

ABSTRACT

Introduction: An understanding of differences in hospital costs between patient groups is relevant for the efficient organisation of inpatient care. The main aim of this study was to confirm the hypothesis that eight a priori identified cost drivers influence per diem hospital costs. A second aim was to explore further variables that might influence hospital costs.

Methods: The study included 667 inpatient episodes consecutively discharged in 2014 at the psychiatric hospital of the Medical Centre-University of Freiburg. Fifty-one patient characteristics were analysed. Per diem costs were calculated from the hospital perspective based on a detailed documentation of resource use. Mixed-effects maximum likelihood regression and an ensemble of conditional inference trees were used to analyse data.

Results: The study confirmed the a priori hypothesis that not being of middle age (33-64 years), danger to self, involuntary admission, problems in the activities of daily living, the presence of delusional symptoms, the presence of affective symptoms, short length of stay and the discharging ward affect per diem hospital costs. A patient classification system for prospective per diem payment was suggested with the highest per diem hospital costs in episodes having both delusional symptoms and involuntary admissions and the lowest hospital costs in episodes having neither delusional symptoms nor somatic comorbidities.

Conclusion: Although reliable cost drivers were identified, idiosyncrasies of mental health care complicated the identification of clear and consistent differences in hospital costs between patient groups. Further research could greatly inform current discussions about inpatient mental health reimbursement, in particular with multicentre studies that might find algorithms to split patients in more resource-homogeneous groups.

Show MeSH

Related in: MedlinePlus

Conditional variable importance of explored patient characteristics.AMDP = Association for Methodology and Documentation in Psychiatry, CGI = Clinical Global Impression, km = kilometres, ADL = Activities of Daily Living, F-groups were derived from the International Classification of Diseases German modification (ICD-10-GM).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4816317&req=5

pone.0152669.g003: Conditional variable importance of explored patient characteristics.AMDP = Association for Methodology and Documentation in Psychiatry, CGI = Clinical Global Impression, km = kilometres, ADL = Activities of Daily Living, F-groups were derived from the International Classification of Diseases German modification (ICD-10-GM).

Mentions: Fig 3 shows the relative conditional variable importance of the 25 most important variables in the prediction of per diem costs derived from an ensemble of 2,000 conditional inference trees. A figure of all 51 patient characteristics can be found in the supplementary material in S1 Fig. The absolute values of importance are neither informative by themselves nor comparable to other studies. Instead they should be used to compare the relative importance between variables under investigation [19]. The most important variable was the Global Assessment of Functioning (GAF) [20], followed by delusional symptoms, being referred from another hospital and neurotic disorders as main diagnosis. A rather steep decline in importance was found in the top ten variables. Thereafter, there were smaller differences in importance between variables. A prediction based on all 51 available variables yielded an explained variance of 19.4% and a Root-Mean-Squared-Error (RMSE) of 59.5 in out-of-bag internally cross-validated predictions [21]. Without cross-validation, an explained variance of 43.5% and a RMSE of 49.8 was found.


Determinants of per diem Hospital Costs in Mental Health.

Wolff J, McCrone P, Patel A, Normann C - PLoS ONE (2016)

Conditional variable importance of explored patient characteristics.AMDP = Association for Methodology and Documentation in Psychiatry, CGI = Clinical Global Impression, km = kilometres, ADL = Activities of Daily Living, F-groups were derived from the International Classification of Diseases German modification (ICD-10-GM).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4816317&req=5

pone.0152669.g003: Conditional variable importance of explored patient characteristics.AMDP = Association for Methodology and Documentation in Psychiatry, CGI = Clinical Global Impression, km = kilometres, ADL = Activities of Daily Living, F-groups were derived from the International Classification of Diseases German modification (ICD-10-GM).
Mentions: Fig 3 shows the relative conditional variable importance of the 25 most important variables in the prediction of per diem costs derived from an ensemble of 2,000 conditional inference trees. A figure of all 51 patient characteristics can be found in the supplementary material in S1 Fig. The absolute values of importance are neither informative by themselves nor comparable to other studies. Instead they should be used to compare the relative importance between variables under investigation [19]. The most important variable was the Global Assessment of Functioning (GAF) [20], followed by delusional symptoms, being referred from another hospital and neurotic disorders as main diagnosis. A rather steep decline in importance was found in the top ten variables. Thereafter, there were smaller differences in importance between variables. A prediction based on all 51 available variables yielded an explained variance of 19.4% and a Root-Mean-Squared-Error (RMSE) of 59.5 in out-of-bag internally cross-validated predictions [21]. Without cross-validation, an explained variance of 43.5% and a RMSE of 49.8 was found.

Bottom Line: Mixed-effects maximum likelihood regression and an ensemble of conditional inference trees were used to analyse data.Although reliable cost drivers were identified, idiosyncrasies of mental health care complicated the identification of clear and consistent differences in hospital costs between patient groups.Further research could greatly inform current discussions about inpatient mental health reimbursement, in particular with multicentre studies that might find algorithms to split patients in more resource-homogeneous groups.

View Article: PubMed Central - PubMed

Affiliation: King's College London, Institute of Psychiatry, Psychology & Neuroscience, King's Health Economics, De Crespigny Park, London, United Kingdom.

ABSTRACT

Introduction: An understanding of differences in hospital costs between patient groups is relevant for the efficient organisation of inpatient care. The main aim of this study was to confirm the hypothesis that eight a priori identified cost drivers influence per diem hospital costs. A second aim was to explore further variables that might influence hospital costs.

Methods: The study included 667 inpatient episodes consecutively discharged in 2014 at the psychiatric hospital of the Medical Centre-University of Freiburg. Fifty-one patient characteristics were analysed. Per diem costs were calculated from the hospital perspective based on a detailed documentation of resource use. Mixed-effects maximum likelihood regression and an ensemble of conditional inference trees were used to analyse data.

Results: The study confirmed the a priori hypothesis that not being of middle age (33-64 years), danger to self, involuntary admission, problems in the activities of daily living, the presence of delusional symptoms, the presence of affective symptoms, short length of stay and the discharging ward affect per diem hospital costs. A patient classification system for prospective per diem payment was suggested with the highest per diem hospital costs in episodes having both delusional symptoms and involuntary admissions and the lowest hospital costs in episodes having neither delusional symptoms nor somatic comorbidities.

Conclusion: Although reliable cost drivers were identified, idiosyncrasies of mental health care complicated the identification of clear and consistent differences in hospital costs between patient groups. Further research could greatly inform current discussions about inpatient mental health reimbursement, in particular with multicentre studies that might find algorithms to split patients in more resource-homogeneous groups.

Show MeSH
Related in: MedlinePlus