Applied example: Complaints analysis

Complaints analysis: Health
sector Colombia - 2016

The analysis of information is a tool applicable to any sector. The purpose of this mini project is to demonstrate the power of data analysis in an environment outside the industry.
For this case we used a database that has 671,471 rows and 9 columns, looks something like this:

  PQR_   CANAL     FECHA_CREACION        PET_DPTO          PQR_TIPOPETICION    ...
10000    Telefonico  12/01/2016 06:13:00 AM     BOGOTA D.C        Reclamo                      ...
10001    Telefonico  12/01/2016 06:34:00 AM     BOGOTA D.C        Reclamo                      ...
10002    Telefonico  12/01/2016 06:43:00 AM     CALDAS                Reclamo                      ...
10003    Telefonico  12/01/2016 06:48:00 AM     ANTIOQUIA          Reclamo                      ...
10004    Telefonico  12/01/2016 06:51:00 AM     ATLANTICO          Reclamo                      ...

An important relationship includes knowing the frequency of complaints by department and we obtain the following:

The city of Bogota has the largest number of complaints, followed by the departments of Antioquia and Valle.

The second relation is due to the frequency of complaints against EPS, which is of great importance for the object of study.

It is observed that Cafe salud, Nueva EPS and Comeva lead the number of complaints from all 135 organizations registered in the database.

An interesting pattern that was found is that women lead the number of complaints against EPS and the range of ages that show a higher frequency are those older than 63 years as shown below:

Then it is evident that the complaints represent the largest volume of complaints and the health, coomeva and new EPS entities disperse the largest number of complaints, in general, it is shown that the other EPSs have low volumes of PQR. Another element that is observed is that citizens' follow-ups to tutelas are very low, which would lead to a great impact of the tutelas to assert the right to health (more information is required to obtain a Conclusion):

Finally, the EPS can be categorized considering the complaints associated with them. We performed an analysis of the 89 types of complaints for each EPS and using a machine learning technique to categorize (k-means), in this way it was possible to obtain 4 types of categories, where it is observed that Cafe salud is itself a category , This may be due to the discrepancy between the volumes of complaints by EPS.

Coomeva, Nueva EPS and Salud Total fall into one category, which relates their similar volumes of complaints, the last category groups approximately 89% of the entities, and are those whose individual volumes are less than 3000 complaints. As shown in the following graph: