Lower Costs and Improve Care

The Healthcare Experimental Learning Platform is designed to support healthcare providers with a 360-degree view of each individual, and to leverage this information to analyze, asses, and predict health-related incidents. Predictive capabilities and hypothesis testing allow providers to intervene at critical times, and to better define and assess individual care plans.

HELP combines data from the Electronic Health Record (EHR), claims data, RPM vitals, and social and community determinants, physical needs, cognitive needs, medication compliance, and lifestyle data. With this consolidated view of the individual and the aggregated population, modeling allows the provider to identify previously unknown correlations, and to run a series of hypotheses based on proposed care plan interventions.

Combined Data Sources Into a Single Analytics Platform

Rapid Experiment Hypothesis Testing

The seven steps of hypothesis testing:

  1. Problem Definition – definition of the goal of the hypothesis test. Often related to the elimination of specific medical condition or conditions (e.g. a set of CPT codes), the problem definition might also include important life events (such as suicide).  As an example, we may wish to identify precursors to depression. As part of the problem definitions, Clinicians may also define the relative cost for the average population, for later use in defining the gain to be achieved relative to a specified care plan.
  2. Care Plan Test Condition Definition – determination of the alteration to the current care plan that represents the test condition(s). This is used as a comparison to the precondition status, and against a control group to verify the veracity of the care plan modification. Using the example above, a hypothesis might be that increased activity reduces depression, so the condition definition might be to increase activity. Alternatively, a condition might be to engage in more social activities, so this too could be a defined test condition.
  3. Overall Test Population Selection – the test population is derived by querying the overall population for those who meet the group of individuals to be used for the test and control groups. It allows for the selection of individuals that have specific medical conditions, cognitive challenges, social conditions, gender, or virtually any data element available. The overall test population defines the segment of the overall population to be studied.
  4. Control and Study Group(s) Selection – once the overall test population is selected, a subset of the population must be derived to apply the test condition(s), and others that act as the control group, used to compare if the test condition(s) provided statistical changes, and the extent to which these modifications vary from the control group.
  5. Correlation and Propensity Modeling – in order to derive greater commonalities of the population selected by the test population and past history related to the condition definition defined, HELP performs a data correlation exercise which identifies individual sub-segments of people based on the full set of known data about those individuals. These correlations act as potential predictors that may be considered for further hypothesis modeling, and for additional care plan initiatives. For example, a correlation between pet size and broken hips could be derived, allowing for a future care plan that provides a different service to those of a particular age with large dogs.  Propensity modeling then defines those individuals that are more prone to the problem defined in the problem definition phase.
  6. Hypothesis Time/Score Definition – the care provider then sets the time definition for the hypothesis (i.e. run until proven or disproven, or run for three months), and acceptable score level that would be used to derive a valid proof for a given hypothesis model.
  7. Hypothesis Test – this phase is simply running of the test and monitoring the results. It provides a number of statistical measures that demonstrate the distribution across the population, and that of the test group, to provide relevant percent change metrics, and to ascertain possible health improvement or cost saving estimates if applied to the overall population.