{
  "abstract": "Background Trainee-led UK research collaboratives often use ICD-10 codes for their patient identification, assuming sufficient diagnostic accuracy. For effective patient identification, codes must have a reasonable positive predictive value (PPV) to minimise unnecessary chart reviews of patients without the diagnosis (i.e. false positives), and adequate sensitivity to capture a meaningful proportion of those with the diagnosis. Using local data which contributed to the INSPIRE collaborative’s ERUPT and CLAMP studies, we aimed to assess the PPV of ICD-10 codes for pulmonary embolism (PE) and pneumothorax.Methods We identified patients with relevant diagnostic codes at two UK hospitals over defined timeframes, for 1) PE (I26.0, I26.9), 2) PE with cor pulmonale (I26.0), and 3) Pneumothorax (J93.0, J93.1, J93.8, J93.9). There were no exclusion criteria other than eligibility time period. The PE cohort spanned a 12-month period, and the pneumothorax cohort 30 months. Two clinicians extracted demographics and disease-specific variables, using a standardised template, blinded to the other’s assessments. We collected patient demographics (age, sex, index of multiple deprivation of home postcode), diagnoses, and disease-specific aspects (PE: provoked/unprovoked, malignancy, deep vein thrombosis. Pneumothorax: type, respiratory disease, smoking history, malignancy, pleural procedures). Conflicts were resolved by consensus.Results We reviewed 165 pneumothorax-coded and 604 PE-coded patients. There was a male predominance in pneumothorax group (pneumothorax: 36.4% female; PE: 53%), with similar ages (pneumothorax: mean 66 (SD 20); PE: 71 (SD = 17)). PPVs were: pneumothorax 80% (132/165), PE 85.3% (515/604), and PE with cor pulmonale 93.4% (155/166). Inter-reviewer reliability for decisions regarding diagnostic confirmation was generally high (Cohen’s kappa coefficients (κ): PE = 1, Presence of cor pulmonale = 0.53).Abstract P86 Table 1Positive predictive values of diagnostic codesConclusion ICD-10 codes for PE and pneumothorax showed moderate-high PPVs, supporting their use if accompanied by diagnostic confirmation. We recommend collaborative studies using ICD-10 codes should examine and report PPVs given the minimal associated workload and importance in establishing their validity.",
  "authors": [
    {
      "affiliations": [
        "Maidstone Hospital, Maidstone and Tunbridge Wells NHS Trust, Kent, UK"
      ],
      "name": "T Hamakarim"
    },
    {
      "affiliations": [
        "Maidstone Hospital, Maidstone and Tunbridge Wells NHS Trust, Kent, UK"
      ],
      "name": "P Bhutra"
    },
    {
      "affiliations": [
        "Deanery of Clinical Sciences, College of Medicine and Veterinary Medicine, The University of Edinburgh, Edinburgh, UK"
      ],
      "name": "D Henshall"
    }
  ],
  "title": "P86 The accuracy of ICD-10 diagnostic codes for the identification of pneumothorax and pulmonary embolism: a retrospective study in two UK centres",
  "uid": "2c310acd-cc78-5c0c-894e-0f3e50f8b78c"
}