Advanced

Prediction of Caregiver’s Next Action in Digital Healthcare

Andersson, Elin LU and Sager, Paulina LU (2020) In Master's Theses in Mathematical Sciences FMAM05 20201
Mathematics (Faculty of Engineering)
Abstract
Digital healthcare has proven to be increasingly important as the pressure on primary healthcare is enhanced, with a growing elderly population that lives longer. The importance of an effective digital healthcare with high quality is consequently increasing. By creating a reliable model, using machine learning algorithms, that can predict the next action for a caregiver in digital healthcare, not only the patient experience is improved but also the efficiency of digital healthcare.

To narrow down the project the specific action studied was if a case handled by the Swedish digital healthcare provider Min Doktor was going to get closed without action or closed normally. The data set consisted of patient data from Min Doktor collected from... (More)
Digital healthcare has proven to be increasingly important as the pressure on primary healthcare is enhanced, with a growing elderly population that lives longer. The importance of an effective digital healthcare with high quality is consequently increasing. By creating a reliable model, using machine learning algorithms, that can predict the next action for a caregiver in digital healthcare, not only the patient experience is improved but also the efficiency of digital healthcare.

To narrow down the project the specific action studied was if a case handled by the Swedish digital healthcare provider Min Doktor was going to get closed without action or closed normally. The data set consisted of patient data from Min Doktor collected from the questions that the patients are given when reaching out for a digital consultation. The data, 44 416 cases in total, was a mixture of structured and unstructured data. The structured data included multiple choice questions while the unstructured data was data on free text form.

Four machine learning models were used; SVM, random forest, logistic regression and ANN. All of them produced very similar results. The best performing classifiers, SVM and logistic regression, received an AUC score of 0.72. This indicates that the models can find patterns in the data, but need to be further developed. Additionally, the input data needs to be more processed. The free text data, that was analysed in the later stage of the project, holds important information and would be beneficial to process further. An addition of new features may give a more complete picture of each case, and thereby a more reliable model. (Less)
Popular Abstract
Digital healthcare has proven to be increasingly important as the pressure on primary healthcare is enhanced, with a growing elderly population that lives longer. The importance of an effective digital healthcare with high quality is consequently increasing. By creating a reliable model, using machine learning algorithms, that can predict the next action for a caregiver in digital healthcare, not only the patient experience is improved but also the efficiency of digital healthcare.

To narrow down the project the specific action studied was if a case handled by the Swedish digital healthcare provider Min Doktor was going to get closed without action or closed normally. The data set consisted of patient data from Min Doktor collected from... (More)
Digital healthcare has proven to be increasingly important as the pressure on primary healthcare is enhanced, with a growing elderly population that lives longer. The importance of an effective digital healthcare with high quality is consequently increasing. By creating a reliable model, using machine learning algorithms, that can predict the next action for a caregiver in digital healthcare, not only the patient experience is improved but also the efficiency of digital healthcare.

To narrow down the project the specific action studied was if a case handled by the Swedish digital healthcare provider Min Doktor was going to get closed without action or closed normally. The data set consisted of patient data from Min Doktor collected from the questions that the patients are given when reaching out for a digital consultation. The data, 44 416 cases in total, was a mixture of structured and unstructured data. The structured data included multiple choice questions while the unstructured data was data on free text form.

Four machine learning models were used; SVM, random forest, logistic regression and ANN. All of them produced very similar results. The best performing classifiers, SVM and logistic regression, received an AUC score of 0.72. This indicates that the models can find patterns in the data, but need to be further developed. Additionally, the input data needs to be more processed. The free text data, that was analysed in the later stage of the project, holds important information and would be beneficial to process further. An addition of new features may give a more complete picture of each case, and thereby a more reliable model. (Less)
Please use this url to cite or link to this publication:
author
Andersson, Elin LU and Sager, Paulina LU
supervisor
organization
course
FMAM05 20201
year
type
H2 - Master's Degree (Two Years)
subject
publication/series
Master's Theses in Mathematical Sciences
report number
LUTFMA-3412-2020
ISSN
1404-6342
other publication id
2020:E32
language
English
id
9020249
date added to LUP
2020-06-24 13:52:48
date last changed
2020-06-24 13:52:48
@misc{9020249,
  abstract     = {Digital healthcare has proven to be increasingly important as the pressure on primary healthcare is enhanced, with a growing elderly population that lives longer. The importance of an effective digital healthcare with high quality is consequently increasing. By creating a reliable model, using machine learning algorithms, that can predict the next action for a caregiver in digital healthcare, not only the patient experience is improved but also the efficiency of digital healthcare.

To narrow down the project the specific action studied was if a case handled by the Swedish digital healthcare provider Min Doktor was going to get closed without action or closed normally. The data set consisted of patient data from Min Doktor collected from the questions that the patients are given when reaching out for a digital consultation. The data, 44 416 cases in total, was a mixture of structured and unstructured data. The structured data included multiple choice questions while the unstructured data was data on free text form. 

Four machine learning models were used; SVM, random forest, logistic regression and ANN. All of them produced very similar results. The best performing classifiers, SVM and logistic regression, received an AUC score of 0.72. This indicates that the models can find patterns in the data, but need to be further developed. Additionally, the input data needs to be more processed. The free text data, that was analysed in the later stage of the project, holds important information and would be beneficial to process further. An addition of new features may give a more complete picture of each case, and thereby a more reliable model.},
  author       = {Andersson, Elin and Sager, Paulina},
  issn         = {1404-6342},
  language     = {eng},
  note         = {Student Paper},
  series       = {Master's Theses in Mathematical Sciences},
  title        = {Prediction of Caregiver’s Next Action in Digital Healthcare},
  year         = {2020},
}