Machine Learning Model to Predict HOLS and Mortality After Discharge in Hospitalized Oncologic Patients

BACKGROUND:

Cancer is the second leading cause of death worldwide and is responsible for about 18.1 million new cases and 9.6 million deaths in 2018 alone according to the International Agency for Research on Cancer. Cancer is anticipated to rank as the leading cause of death and the most important barrier to increasing life expectancy in every country of the world in the mid-21st century1. The economic impact of cancer is significant. The annual economic cost of cancer in 2010 was estimated at approximately US$ 1.16 trillion. The reasons are complex but both cancer incidence and mortality are increasing worldwide due to aging and increasing risk factors for cancer, several of which are associated with socioeconomic development. Cancer will probably soon reach the top leading cause of death due to the rapid population growth and the declines in mortality rates by stroke or coronary heart disease in many developed countries.

Cancer patients often require inpatient care due to treatment toxicities, complications from cancer such as thrombosis, illness not related to the disease itself or terminally ill patients. Among these individuals, their treatment should balance prolongation of survival and maximization of the quality of remaining life. However, hospitalization is a stressful event for individuals with advanced cancer and their caregivers. Hospitalization often antagonizes these goals, contributing to the high cost of cancer care, worsens survival, and is increasingly recognized as poor-quality cancer care. Thus, interventions that reduce unnecessary hospitalizations, or shorten them, will likely improve quality of life and reduce costs.

Some studies relate malnutrition, which presents a marked sarcopenia and loss of lean mass, with prolonged hospitalization, reduced response to treatment, a worse overall survival and impaired quality of life. A study published in 2007 found that lung cancer patients had a longer hospitalization and required inpatient hospital treatment more frequently than any other type of tumor. Moreover, in the surgical setting there have been studies linking preoperative opioid usage and increased opioid doses with increased length of stay. Based on this data, there have been protocols developed like the ERAS (Enhanced recovery after surgery) applied first to colorectal cancer and now being tested in other settings like head and neck and gynecologic tumors, showing that it is possible to reduce opioid use with good pain control and a statistically significant shorter average length of stay.

Prognostic factors for oncologic patients after surgery or curative systemic treatment have been described, but there is no solid evidence on which combination of parameters predict mortality after hospitalization of metastatic cancer patients under active treatment. A potential solution to improve this scenario might be nutritional support to malnourished cancer patients that also has proven to be effective in shorten hospital stay and improve survival, or community based palliative care interventions that are proven to improve quality of life and reduce costs of terminally ill patients. Thus, a prognostic tool would be useful to help physicians adjust medical interventions for hospitalized cancer patients.

To the best of our knowledge, this is the first study that examines independent clinical, psychological, nutritional status, and laboratory characteristics of oncologic patients in order to grasp a comprehensive picture of what factors play a role in the length of stay, mortality, and quality of life.

MEANING The investigators pretend with this work to fill a gap of knowledge in the oncology field through a prospective study. The investigators would like to measure the effect of hospitalization on oncologic patients after discharge and how clinical and laboratory parameters at admission may be able to predict HOLS and 30-day mortality after discharge. The investigators would also like to validate the different scales already published to assess nutritional status, psychological status, quality of life or prediction of rehospitalization for oncologic patients in all-in-one study.

This study will hopefully be able to develop a predictive tool at admission to help physicians adjust medical interventions and detect possible actions that will need to be implemented during hospitalization in order to improve the overall survival and quality of life of our patients.