{
  "abstract": "Background Falls in older adults are a major public health concern, particularly when individuals remain on the floor for extended periods—a ‘long lie.’ However, there is a lack of evidence about the impact of lie duration on patient outcomes, and the use of routine data to understand this impact, is inhibited by lack of recording of the time that the fall occurred in fields that can be easily extracted for analysis. We aimed to develop an algorithm to extract time of fall from free text entries in ambulance electronic patient records (ePRs) to address this.Methods We randomly sampled 1,000 electronic patient records (ePRs) from a UK ambulance service where incidents were triaged as AMPDS Card 17 (falls) or NHS Pathways code DX013 (assistance needed due to inability to get up). Two research paramedics independently reviewed each record to confirm whether a fall had occurred and the time of fall, if recorded. Free-text fields were reviewed by a research paramedic to identify where and how time of fall was recorded, and the algorithm iteratively developed by creating regular expressions that matched these observations. Inter-rater agreement was calculated, and algorithm performance was assessed, using common diagnostic measures.Results After removing 83 non-fall cases, 337/917 (37%) contained a documented time of fall. The algorithm correctly identified 255/337 (sensitivity 76%) of cases with a time of fall and 514/580 (specificity 89%) without. Inter-rater reliability was high for both fall identification and time of fall.Conclusion By harnessing clinician’s knowledge of documentation styles, it is feasible to develop an algorithm to extract time of fall from routine ambulance free-text data. Future work utilising machine learning methods may further enhance performance.",
  "authors": [
    {
      "affiliations": [
        "Yorkshire Ambulance Service NHS Trust, UK"
      ],
      "name": "Richard Pilbery"
    },
    {
      "affiliations": [
        "University of Sheffield, UK"
      ],
      "name": "Jen Lewis"
    },
    {
      "affiliations": [
        "University of Sheffield, UK"
      ],
      "name": "Fiona C Sampson"
    }
  ],
  "title": "EP08 Algorithmic extraction of fall time from the free text fields of ambulance service electronic patient records",
  "uid": "a5d56ad6-a1f5-54aa-b665-ee48b69a405f"
}