A Brief:

A significant opportunity exists to enhance the administration of care for chronic diseases, which are the main cause of death and disability. There could be a huge influence on both financial and health consequences. To lessen the burden of chronic illnesses on healthcare organisations, RPM’s capacity to gather and send health data outside of a traditional care environment is optimal for Health Care Organisations (HCO). It offers the information required to regularly monitor health changes and respond to them, enabling better healthcare resource allocation and fostering better patient participation.

Importance:

RPM allows for routine evaluation without the need for many appointments and may help to avert acute clinical occurrences. RPM offers continuous collection of important data and behaviours, such as changes in blood pressure, heart rate, or activity levels, eliminating the need to postpone exams and treatment until a scheduled appointment. With this knowledge, care teams and patients can take preventative measures before chronic illnesses deteriorate to the point where hospitalisation or a trip to the emergency medical support are necessary.

Goal and Opportunity:

RPM as a data source has enormous potential to enhance outcomes, but in order to fully realise this potential, HCOs must combine it with the appropriate analytical tools. These continuing data streams can be ingested by AI-based models, which can then reliably identify which patients are most likely to suffer from adverse events, uncover the precise risk variables considered in making predictions, and improve proactive engagement and therapies. AI can provide these findings to care teams in current clinical workflows to speed up outreach. By giving patients much more knowledge about their own health and advice that encourages healthy behaviours, it can also increase patient involvement.

Our Approach:

We have downloaded an extensive dataset from Kaggle which consists of medical history of more than 5 lacs people on more than 900 parameters. Among those, ‘’disposition’’ is our target variable

(Ref: https://www.kaggle.com/datasets/maalona/hospital-triage-and-patient-history-data)

We are to build a model, to predict which person is in critical health condition and required hospitalisation. We shall make an end-to-end study on the dataset and prepare the best fitted model/s.  From visual inspection and descriptive statistics, we

have noticed some possible anomalies in the dataset. Necessary steps have been initiated to clean, manipulate and arrange the dataset properly for further analysis.

We also have a deep and detailed analysis to identify prime factors influencing health indicators. Finally, following 49 parameters have been considered for model creation.

meds_gastrointestinal, n_admissions, age, n_edvisits, vital_hr, vital_o2, vital_rr, pregtestur_count, vital_temp, vital_sbp, vital_dbp, bloodculture_routine_count, glucoseua_count, glucoseua_npos, otherxr_count, sbp_median, spo2_min, proteinua_npos, spo2_max, temp_median, n_surgeries, resp_median, sbp_max, resp_max, dbp_median, resp_min, sbp_min, pulse_min, temp_min, temp_max, dbp_min, otherimg_count, ekg_count, dbp_max, sbp_last, dbp_last, pulse_median, cxr_count, pulse_max, urineculture_routine_count, religion, pulse_last, resp_last, nitriteua_count, arrivalmonth, bloodua_npos, otherus_count, bloodua_count, spo2_median, headct_count, otherct_count, ketonesua_npos, leukocytesua_npos, proteinua_count, ketonesua_count, leukocytesua_count.

We are able to build the best & consistent model with 96 % accuracy. With this supreme accuracy, we can completely rely on the prediction of hospitalisation possibilities forecasted by the model.    

Customer Segmentation & Recommendation:

The individuals will be segregated on the basis of their hospitalization possibilities on the ground of their current health parameters. The segments are:

• Critical: They require immediate admission and medical support.  

• Require Attention: Not critical but require constant support and monitoring. They should have controlled life style, food intake & proper medicine support.

• Stable Condition: Indications of fluctuations from normal range of measures, but not in extreme level. Regular life style, proper food habits & medical precautions to follow.  

• Healthy: Complete control over different parameters.  

On implementing our findings and recommendations, any HCO will be able to monitor their patients remotely. The segmented group of patients will help to formulate future strategies and arrangements smoothly. They can also prioritise & personalise their planning with the help of real-time status update of patients. For a new patient also, with the required input, the hospitalisation possibility can be predicted instantly.