Blood Testing in Inpatients – Computer-driven or Clinician Choice:

• To understand the proportion of blood tests, both clinician-ordered and system-ordered, which are collected successfully.
• To quantify the proportion of computer-driven blood orders that are manually rejected by clinicians.
• To estimate the expenditure on blood tests per patient between hospitals with and without automated ordering.
• To understand how levels of patient sickness and comorbidity influences blood-ordering behaviour.

Patients admitted to hospitals have high numbers of blood tests done during admissions. Electronic health records (EHRs), commonly utilise sophisticated clinical decision support systems (CDSS) that can, amongst other functions, propose new blood tests based on patient diagnoses, abnormal or changing test results, or other parameters such as observations, prescriptions, or planned procedures. This is perceived as improving safety as it helps ensure blood tests take place at critical timepoints in patient care and is widely employed in the version of PICS that is active on site.

However, as the complexity of any CDSS increases, so too does the likelihood that it will order unnecessary investigations. In this instance, this leads to unnecessary painful blood testing for patients, wastage of Trust resources with respect to equipment, laboratory capacity and staff time and contributes to problems associated with over-investigation such as iatrogenic anaemia, which itself may be associated with adverse patient outcomes.

In the same way that clinicians occasionally disagree over the value of a given investigation in a particular patient, no CDSS can avoid altogether the problem of ordering unnecessary investigations. However, based on anecdotal feedback both from patients and clinicians we hypothesise that there is scope to improve the existing system.

To assess this, we would like to use anonymised, individual-level patient and summary data to investigate the blood ordering behaviour of hospital departments in which computer-driven ordering is available and those in which it is entirely dependent on clinician input, to assess how both patient and system factors influence this behaviour, and to identify key areas where the CDSS may be refined to reduce unnecessary investigations whilst maintaining patient safety.