Health care customer satisfaction survey analysis using self-organizing maps and “exponentially smeared” data vectors

The Tanimoto coefficient self-organizing map (TCSOM) [Garavaglia (1998)] is combined with a data transformation that “smears” binary vectors with an exponential decay transformation according to a subjective assessment of how close in meaning adjacent vector elements are. The methodology is demonstrated using managed health care plan satisfaction survey data originally in ordinal integer values. Five different SOMs were trained and tested using ordinal, binary, and “smeared” vectors. The result is that the TCSOM provides a useful visualization of the complexity of responses and highlights specific areas for health plan quality and service improvements that might be missed if only simple average total scores were considered