Analysis of health care data with HyperCube

Client Business Challenges

• Our client provides analytics services for 28 of the top 30 German sickness funds
  • 7 M records in their research database
  • Complete, representative, approved by authorities
• However there are many cost driving rules and subgroups are still undiscovered by commonly used analytic tools
• Typical examples of interesting analytical questions in the German sickness fund sector are:
  • Identification and analysis of non-standard morbidity pattern
  • Analysis of the potential in regard to root causes of coding errors and management solutions
  • Setup of typical patient and customer profiles for the forecast of cost developments
  • Creating measures for securing data quality in regard to the MRSA system
  • Status analyses and setup of morbidity dependent risk groups
  • Identification, analysis and validation of care strategies and concepts
  • Measurement approaches on the results of healthcare programmes and schemes

BearingPoint’s Contribution

• As a starting point in using the data with HyperCube we designed a specific research question to identify subgroups and significant rules with at least 90% and 95% probability of hospitalisation within one year among patients with congestive heart failure diagnosis
• The research question was addressed with the following sample:
  • Sample of 10,000 cases analysed
  • Output / dependent variable: Occurrence of a hospital stay among patients with Congestive Heart Failure
  • Independent variables = age, gender, > 200 variables with binary classifications of reported illnesses, > 200 variables with binary classification of medical drug treatments, some aggregates variables

Client Business Outcomes

• All HyperCube findings are supported by advanced clinical and medical studies, verified in the aggregated data of the client
• Subgroup discovery and rules identification with HyperCube can bring interesting results on sickness fund data, that go beyond the outcomes of GKV standard tools (e.g. SPSS, Risk KV etc.) due to real service gaps
• The identified patterns show huge potential for further analysis in two regards
  • Application of the analytic approach to further specific diagnoses and drug treatment patterns
  • Incorporation of further variables to the models – e.g. geographical information, socio-demographic data
• Rules identified using the project approach can potentially be used by the GKV, for example, to
  • strengthen disease and health management
  • optimise care management
  • strengthen their position in the dialogue with health service providers