Application of Prescribing Safety Indicators to the Clinical Practice Research Database: a retrospective cohort study

The problem

Prescribing errors in primary care cause significant harm; adverse drug events account for around 7% of hospital admissions and half of these are considered preventable. Over half of these admissions are associated with the following groups of drugs: anti-platelets, diuretics, NSAIDs and anticoagulants. Eight prescribing safety indicators (PSI, listed below) to electronically identify patients at increased risk of a prescribing error have been applied in 72 English general practices. In Scotland the prevalence of a set of indicators of high risk prescribing has also been measured in 315 practices. However there is a need to systematically identify which PSI are the most useful for identifying patients at increased risk of a prescribing error, and how frequently these events occur in a large nationally representative dataset.

The approach

The prevalence for each PSI listed below is being measured within the Clinical Practice Research Database (over 600 practices)

• Diagnosis of peptic ulcer and prescribed a NSAID without PPI protection F Diagnosis of asthma and prescribed a betablocker
• Aged ≥75 or over prescribed ACEI or loop diuretics and have not had urea and electrolyte monitoring
• Diagnosis of thrombosis and prescribed a combined oral contraceptive F Prescribed methotrexate and not had a recent full blood count or liver function test
• Prescribed warfarin and not had an INR
• Prescribed lithium and not had a lithium level check
• Prescribed amiodarone and not had a thyroid function test

For each PSI the prevalence is measured relative to an audit date using a rolling time window. For long term conditions, such as asthma or peptic ulcer, the cumulative number of patients at risk since their first recorded diagnosis is included in the denominator. If being at risk is reversible such as requiring monitoring when prescribed a medication the denominator is the patients prescribed the medication within a fixed time prior to the audit date. Multilevel regression models will be used to explore the associations 

between prevalence and trends and practice attributes such as list size and patient attributes such as age, gender and co-morbidities.

Findings

...will be reported as follows:

• The mean annual prevalence of patients with a PSI recorded from 2000–2010
• The variation between practices in recorded PSI and associations with practice attributes e.g. list size
• The associations between recorded PSI and patient demographics e.g. age, gender, multimorbidity
• Changes over time in the prevalence of PSI

Consequences

This is the first report of the prevalence of PSI in the Clinical Practice Research Database. It will show whether these PSI are reliable in terms of identifying groups at increased risk of a prescribing error.