STOPP/START to prevent adverse drug events: HM17 update

STOPP and START give hospitalists a repeatable way to examine medicines in adults aged 65 and over. STOPP screens for potentially inappropriate prescriptions, while START looks for treatments that may be indicated but missing. Used alongside medication reconciliation, renal function, falls history and patient goals, they can expose risks such as anticoagulant doses that are too high, sulfonylurea-related hypoglycaemia or polypharmacy after discharge. The criteria do not diagnose a problem on their own; they create prompts for a documented clinical decision.

Hospital Medicine 2017, held from 1–4 May 2017 at Mandalay Bay Resort and Casino in Las Vegas, brought hospitalists together around research, practice innovation, continuing medical education credits and professional networking. That setting is especially relevant to STOPP/START because adverse drug events often emerge when a fragmented medication history meets a complex inpatient system. For Australian teams, the update becomes useful when it is adapted to local formularies, pharmacy systems and transitions of care.

Why structured review matters at HM17

Older inpatients have more diagnoses, more medicines and a narrower physiological reserve. A drug that was appropriate at discharge may become inappropriate when kidney function deteriorates, oral intake falls or mobility changes. Prescribing cascades can add another medicine to treat an adverse effect, and medicines supplied by different family members may not appear on the admission list. These are routine realities on Australian wards, where patients often arrive with medicines from several community pharmacies.

The HM17 update is best understood as a move from passive review to structured screening. Instead of waiting for a near miss or relying on memory, teams can apply a defined set of prompts at admission, transfer and discharge. This creates a common language among doctors, pharmacists and nurses, and makes discrepancies visible in a shared record. It also supports audit: teams can determine whether high-risk prompts were checked, whether actions were taken and whether monitoring was arranged. Conference information and learning resources for teams reviewing the update are available through the official Hospital Medicine 2017 site.

Yet systematic screening is not a substitute for judgement. A STOPP prompt may be wrong when there is a clear indication, and a START prompt may not suit a person with limited life expectancy, frailty or a history of adverse effects. The safety value comes from pairing criteria with a best possible medication history and shared decisions. Every suggestion should end with accept, modify, defer or decline, plus a reason and a follow-up date where monitoring is needed.

Reading STOPP and START signals safely

STOPP questions focus on medicines that are potentially inappropriate for the person’s current circumstances. Examples include a benzodiazepine in someone prone to falls, an anticholinergic drug that can worsen delirium or constipation, an NSAID in severe renal impairment, and a sulfonylurea such as gliclazide when recurrent hypoglycaemia makes it hazardous. Criteria also prompt checks of anticoagulant indication and dose, digoxin toxicity, prolonged proton-pump inhibitor therapy without an indication, and combinations that increase bleeding or bradycardia.

START questions work in the opposite direction by looking for omitted therapy. They may flag anticoagulation for atrial fibrillation, a statin for established atherosclerotic cardiovascular disease, blood-pressure treatment at markedly elevated readings, renal protection in appropriate chronic kidney disease, or osteoporosis treatment after a fragility fracture. These prompts work best when the indication is already known. A structured record of diagnoses, eGFR, weight, blood pressure, fracture history and relevant allergies gives the reviewer enough context to avoid treating every alert as an instruction.

Australian teams should use the published criteria set without paraphrasing individual rules from memory. Check the version in use, document local implementation notes, and link prompts to the hospital formulary and approved monitoring guidance. Names and brands differ between the PBS and inpatient formularies, so a medicine should be matched by ingredient, strength and indication. Renal dosing may require creatinine clearance rather than eGFR alone, particularly in frail patients or when the medicine’s product information specifies it.

Deprescribing needs a withdrawal-aware plan. Long-term benzodiazepines, corticosteroids, antidepressants and opioids should not be stopped suddenly without a taper and a monitoring plan. For a high-risk medicine, the prescriber should also specify who will check the result, such as an international normalised ratio, haemoglobin, renal function or blood glucose, and how the patient or carer will be contacted.

Comparing review methods in practice

A routine medication review and a STOPP/START-supported review solve different problems. Routine review is patient-led and flexible, which makes it well suited to distress, adherence concerns and complex goals. Criteria-supported review is population-based and repeatable, making it useful for identifying overlooked risks in busy wards. The strongest model combines both: criteria create the queue and structure, while clinical assessment determines the response.

The comparison also clarifies where automation helps and where it can mislead. Electronic prompts can exclude contraindicated combinations, flag duplicate ingredients and calculate renal dose adjustments, but they depend on accurate diagnoses and medication lists. A criteria result should therefore be reviewed by a pharmacist or prescriber, not accepted automatically. When an alert is overridden, recording the indication and plan prevents another team from repeating the same debate and turns the override into useful safety data.

Building an Australian medication safety workflow

Implementation works differently in a metropolitan Melbourne or Sydney hospital than in a regional service such as Bendigo or Darwin. Large hospitals may have specialist medicine teams, electronic medication reconciliation and dedicated clinical pharmacists; smaller services may rely more heavily on visiting pharmacy support and paper discharge summaries. A workable Australian model starts with a standard best possible medication history completed within 24 hours of admission, including the patient or carer, the community pharmacy and the general practice where possible. Medication changes should then flow through one reconciled chart rather than parallel inpatient and community lists.

The same discipline applies to transfers and discharge. Prioritise anticoagulants, insulin, opioids, antibiotics and medicines with narrow therapeutic ranges. Confirm who will supply the medicine, whether the dose is appropriate and what monitoring is due. The conference also captured operational debates such as new insulin protocols, where safe escalation, documentation and follow-up offer useful parallels for high-alert medicines. These are process lessons rather than prescribing mandates: every local protocol still needs medicine review, staff training and a clear escalation route.

Australian discharge planning must bridge the PBS and the hospital formulary. Subsidised community medicines may have different brands, strengths or pack sizes from inpatient products, and a patient may use an over-the-counter analgesic or complementary product that changes the risk profile. Many older people organise tablets in dose boxes but keep repeats at more than one pharmacy, so the stated regimen may not match what is actually being taken. Ask directly about recent changes, missed doses, affordability and medicines supplied by family, then record the answer rather than merely copying the last discharge summary.

Clinical governance should name the standards that matter in each jurisdiction. The Therapeutic Goods Act 1989 and TGA safety monitoring processes help identify concerns with marketed medicines, while the National Health Act 2012 and PBS rules govern subsidised access; neither replaces local prescribing governance. State health services and private hospitals also maintain medication management and incident-reporting requirements. Teams should document how criteria alerts are escalated, who is accountable for changes, how clinical pharmacy reviews are prioritised and how privacy is protected when information moves between the hospital, general practice and community pharmacy.

A ward checklist for safer prescribing

A short, repeatable checklist turns criteria into routine practice without turning every pharmacist into a data-entry service. The following prompts are suitable for ward rounds, discharge preparation or structured medication reviews.

Before the review

• Build a verified medicine list with the patient or carer and community pharmacy.
• Record diagnoses, age, weight, eGFR, falls, allergies and adherence.
• Match each medicine to an indication and current formulary strength.
• Prioritise high-risk drugs, recent changes and unresolved discrepancies.

During review and follow-up

• Check dose, interactions, monitoring and treatment duration for each STOPP/START prompt.
• Agree accept, modify, defer or decline with the patient and prescriber.
• Provide taper, monitoring and communication details for every medicine change.
• Record the action, responsible clinician, review date and patient instructions.

A useful audit combines process and clinical outcomes: percentage of eligible patients screened, prompts resolved, changes made, unintended effects detected and information correctly transferred at discharge. Review a small sample every month and use local falls, hypoglycaemia, medication-related admissions and unplanned readmissions to test whether the workflow changes care. On the next medication round in your ward, select one patient aged 65 or older with polypharmacy, complete a STOPP/START review with the ward pharmacist, and record one agreed action in the medication chart before the round ends.