Please log in

RiskCede KNOW login



Forgot your password

Powered by RiskCede



RiskCede KNOW platform

KNOW provides a seamless analysis and reporting plat- form, integrating healthcare and related data in real time. The result is a holistic view on all factors that impact on deci- sions and strategy.


Exploratory data analysis

This section provies an overview of the schemes data and performs datamining on claims and membership data. Reports are available to follow monthly trends.

An additional section on Employer group analysis is also available.


Fraud, Waste and Abuse

Various visualisations and machine learning models are used to identify possible areas of fraud, waste and abuse

Users can log actions and investigations to draw reports on progress. The systems will load cases for investigation automatically on each monthly data import


Surveys

The survey mosule includes the designer to design new surveys and all the data analytics and reporting. The results are integratd with the rest of the database.


Product development

A section to simulate future results based on different contribution benefit combinations.


Managed healthcare

Results on the RiskCede MHC application and mobile application.


Admin

Administrators can perform certain tasks themselves.





logo Powered by RiskCede




Scheme overview

Below is some of the key statistics for the scheme.

The table below contains summary data per employer (sorted by number of principal members) for the past 12 months. The province column indicates the province with the most member for the relevant employer.

The plot below illustrates the change in loss ratio relative to the other groups.

Cluster analysis

PCA clustering of employers into groups of similar characteristics.


Client reports

Use the parameters below to generate a client report.

Summary report


Download report

Detail report

to








Download report

Malaria incidents in Africa:


Based on World Health data

Cluster analysis


Group reports

Use the parameters below to generate a client report

Overview of claims data

The information below is used to verify claims data and illustrate areas of highest claim volume.

Service dates

Disciplines (Paid amounts)

Products

Rejections


Similar claims submitted by same provider with different PR number

Click on row to view detail data.

Only claims with Paid amount != 0 included.

Consultations, 0145-0147 in combination with 0190-1093


Case summary

This section gives an overview of FWA cases loaded and feedback given. It also gives the option to doanload a report.

Download report

to

Summary information



Detail data




Upload new case

Complete the fields below to upload a new case for investigation.





Select a case and complete the relevant fields to provide feedback on the investigations and actions taken.




Edit case or feedback

Select a case and edit the text of any of the description fields.



Special projects



Capture information on special projects launched.





Protocol parameters:



to








Case summary

Below is the aggregated cost and counts of all cases

Individual cases

The two plots below split the cases into accommodation (hospital) and treatment (other providers). The plots give a distribution of total cost per case.

brush over plot to view detail.

TREATMENT portion of the case


Protocols

Only providers that are present more than the selected percentage of times are included in the protocol
The provided percentile will be used as the cut off amount for the sugested protocol

Download all protocols

Simulations (Find cases outside of the new norm)

Based on the protocols the following cases would have been identified
Download data


List of outliers

Below is the aggregated lits of outliers per category.

Hospital networks

The graph below illustrates the differences between networks for in-hospital procedures.

Input:




Below is a summary of weighting per provider within each discipline

The peer comparison model compares claiming behaviours of providers within their respective disciplines.

GP visit counts

Click in tables to view details.
Below is a count of daily visits per provider.

Provider visit counts


Click in tables to view details.
Below is a count of daily visits per provider.

Anaesthetist time and tariffs

Download report

Below is the time comparison between hospital and anaethetist.

Next is a list of anaethetist with the biggest percentage cases with high time differences.

Below is a summary of the number of tariffs used per procedure by anaethetist.

Next is the providers who are outliers most of the time.

Clinical technologists rules and tariffs


Rule based analyis.
Below is a summary of the number of tariffs used per procedure by clinical technologists

Next is the providers who are outliers most of the time.

Physiotherapists


Below is the rule tariff codes most used by physiotherapists

Psychologists


Summary table

Below is the rule tariff codes most used by psychologists

Summary table

Detail

Use the link below to view detail data for the selected provider:
Claim detail

Claim pattern

Select a provider in the Summary table



Follow up visits

Select a provider in the Summary table



Psyciatry


Summary table

Below is the rule tariff codes most used by psychologists

Tariff combinations

This table contains records of specific tariff rules and apply it to tariff rules.

Summary table

Detail

Use the link below to view detail data for the selected provider:
Claim detail

Claim pattern

Select a provider in the Summary table



Follow up visits

Select a provider in the Summary table










Markers based on provider postal code, not physical address.

Utilisation parameters:









The objective of the medicine utilisation model is to identify cases of addiction and reselling of drugs. The model identifies the top users of each medicine category.

The table below only looks at individual nappi codes and summarise the number of claims per nappi.

Membership anti selection

Below is a table of all members with significant claims within the frist three months of joining.

Knowledgeshare:



Search existing pdfs











Survey design


Question design





The data used in the analysis gets updated once per day.

During this data preperation phase some analyses are already done to improve the speed of this interactive application



Download report
Download

Participants complete surveys on their mobile devices and the results are analised in this app.

The following tabs aim to transform this data into information that can deliver insights into a scheme's business.


Survey activity

Below is a plot to illustrate the participation in the survey.


Survey results (more info)




Correlations (more info)


Response clusters (more info)


Individual questions (more info)



Sentiment analysis (more info)

Ignoring unknown and missing statements




Total contributions:


Rand amounts adjusted to new benefit year and outliers removed.



Beneficiaries per disease


Acumen parameters





Download the Acumen report
Download



Project summary

Below is a table with all relevant projects and the number of participants in each project.
Performance gauges:

Assessments

Below is a classification of all the different assessment models.



Self assessed modules

The following data is plotted over the selcted time.

Interventions

Stats on interventions

High risk analysis





Population segmentation

Based on analysis performed, the following segments are identified in the pouation







Comorbidity

The graph below gives a summary of the chronic diseases and their comorbidities

Correlation

Below is a graph illustrating the correlations between variables

Regression

This section predicts the annual claim amount of a beneficiary. The model can be used to identify beneficiaries whos claims are significantly higher than predicted.
The features below can be used to estimate future claims for a beneficiary.






Model details

The training set used for this model is 5000 randomly selected members data. The model used for prediction is a general linear model. Claim values have been transformed with the log function to make up for skewness.

Model validation


Comments

The first graph shows that the model predictions is not accurate for extreme values. The Rsquared value for this model is 0.203, meaning the model only explain 20% of the variance in the claim amount. Additional features like smoking, diet, stress, glucose and cholesterol levels, etc will improve the model output.
The second graph shows there is no bias in the residuals.
The last graph gives an indication of the importance of features in predicting the outcome.

Event prediction



This section predicts a class outcome. In this case whether a member will end up in hospital in the next six months. The features used in this model is Age, Gender, Number of chronic conditions, and whether the member went to see a physician in the past two months.
This model again doesnt predict the outcome with too much accuracy, features on smoking, cholesterol, blood pressure, etc would help improve accuracy.
This model is only implemented for Scheme 1.

You can also select a sample list of members to see what their predicted chances are of going to hospital in the next six months. This output does not take the product into account. The column rpPred contains the prediction for the specific member (benid)
Download Data


Below is the tree plot with all the nodes for the classification model.
The last node, the bar, indicates probability of ending up in hospital.

And below is a summary of the model predictive power

                    

System users


Loaded user data

The table below contains key info on all users


User logs

The table below contains the login times of all users


Next is a heatmap showing time of day users log on.

Read more about the: KNOW platform