Predicting visualization of hospital clinical reports using survival analysis of access logs from a virtual patient record

Abstract

The amount of data currently being produced, stored and used in hospital settings is stressing information technology infrastructure, making clinical reports to be stored in secondary memory devices. The aim of this work was to develop a model that predicts the probability of visualization, within a certain period after production, of each clinical report. We collected log data, from January 2013 till May 2011, from an existing virtual patient record, in a tertiary university hospital in Porto, Portugal, with information on report creation and report first-time visualization dates, along with contextual information. The main factors associated with visualization were defined using logistic regression. These factors were then used as explanatory variables for predicting the probability of a piece of information being accessed after production, using Kaplan-Meier analysis and the Weibull probability distribution. Clinical department, type of encounter and report type were found significantly associated with time-to-visualization and probability of visualization.

Predicting Within-24h Visualisation of Hospital Clinical Reports Using Bayesian Networks

“…The studied data was in accordance with previous works in terms of the relevance that some factors may have on the likelihood of visualisation of clinical reports, e.g. department and type of encounter that produced the report [21]. Additionally, patient data and time of report creation were also found relevant for the global model of predicting within 24-hour visualisations.…”

“…Globally, the results show that some clinical reports are still used one year after creation, regardless of the context in which they were created, although significant differences existed in reports created during distinct encounter types [8]. Other results show that half of all visualisations might be of reports more than 2 years-old [20], although this visualisation distribution also varies across clinical department and time of production [21]. Thus, usage of patients past information (data from previous hospital encounters), varied significantly according to the setting of health care and content, and is, therefore, not easy to prioritise.…”

“…emergency room, inpatient care, or outpatient consult), department where the report was generated (e.g. gynaecology or internal medicine) and even type of report in each department, but the possibility of modelling visualisations with survival analysis proved to be extremely difficult [21].…”

“…AgeCat (cat) discretised in decades; Sex (binary); Department (cat); EncType (cat) one of outpatient consult, inpatient care, emergency or other; Hour (cat) truncated from creation time; DoW (cat) one of Sun, Mon, Tue, Wed, Thu, Fri or Sat; Period (cat) one of morning (Hour=7-12), afternoon (13)(14)(15)(16)(17)(18), night (19)(20)(21)(22)(23)(24) or dawn (1-6); Visual24h (binary) target outcome, whether the report has been visualised in the first 24 hours after creation or not.…”