A Mathematical Model For Claims Adjudication


A Mathematical Model for Claims Adjudication - By Dr. Singh

By: Dr. Pariksith Singh, MD, June 01, 2017

Medical Claims Adjudication is an arduous process. Creating a platform to facilitate it via automation, or even manually, is even more arduous. There are innumerable kinds of claims, more than 11, 000 possible edits, approximately 80,000 ICD 10 codes and countless variations in patient benefit packages, provider contracts with plans, nuances of Medicare DRG payments and pricers among other things.

Looking at the whole complex weft of algorithms and interactive situations, one feels that a simpler approach might be possible. One of the first step, in the author’s opinion, is to classify all the diverse data flowing in. This segregation and organization of data into coherent files is the first step, where all the incoming claims are tagged and ready tracked. The whole vast tree of organization of all the possible scenarios can be drawn out with each branch breaking off into further specific possibilities.

For examples, the claims can be divided for managed care into hospital and non-hospital claims. The hospital claims can be further subdivided into inpatient, outpatient, observation, emergency room visits, radiological procedures, lab tests, etc. Each of these subdivisions can be further assigned to various contracted groups. Say, one branch could be Memorial Hospital group and another could be Community Hospitals due to the uniqueness of contract between the specific hospital chains and the insurance companies. Another split could be for all hospital claims with non-contracted hospitals.

Similarly, all physician claims could be subdivided into capitated and fee for service contracts. Fee for service branch into interventional and non-interventional physician specialties. The former would include gastroenterology, cardiology, surgery, orthopedics, neurosurgery, urology, neurosurgery, cardio-thoracic surgery, etc. The non-interventional specialties would include infectious disease, renal, psychiatry, etc.

Once the claims are all filed appropriately, classified and the files named, then they can be tracked. For each scenario, there will be an appropriate rule corresponding to the situation. Thus, a Rule Tree overlapping the Organization Tree can be created. Every situation can thus be addressed appropriately as per contract, Medicare rules and patient benefits.

The two overlapping trees create a decision tree thus helping create an algorithm for each situation.

Another mathematical model is possible by projecting the Organization Tree into a matrix. The details and specifics of each scenario can be defined in the rows and columns. This matrix will thus hold several dimensions together in its simple two-dimensional view and it will be connected to the patient and provider enterprise data warehouse for timely and verified data. The corresponding rule tree can be projected on a matching Rules Matrix which overlaps each situation in the Organization Matrix precisely. This rule matrix can be fed from a rule engine which contains the contracts, benefits, Medicare rates, pricers, etc.

Thus the two interacting matrices can in a snapshot capture several dimensions of a complex array of data comprising of several dimensions. The third matrix or set of matrices would be the Result Matrix. It is the author’s opinion that other complex modules can be similarly mapped on multiple simple and elegant matrices thereby assisting in software development. Such a model gives granular details while giving an overview of the complete picture at the same time.

Thus, multiple matrices that give real time data are possible. The question is can one create one algorithm to connect all these matrices into one simple and elegant software. For that software logic would need to used with matrix mathematics.

The multidimensional matrix may offer the ability to visually represent multiple combinations through a simple 2- or 3- D model. And then additional dimensions can be added as needed.

The classification and segregation described, allows tagging and creating views that can drive analytics, taking into account different perspective for different stakeholders. Further development of such a model might be needed if it is already not being used in esoteric circles.

To know more or schedule an appointment with Dr. Singh, visit: https://theaccesshealthcare.com/providers/pariksith-singh