Skip navigation

When things go wrong

Randomised controlled trials are not the only source of evidence. Cohort studies ( Bandolier 23 , Bandolier 24 ) can provide valuable information, especially when things go wrong - with rare but serious adverse events, for instance.

An interesting example of such a cohort study was published in JAMA last year [1]. This Boston investigation set out to assess the incidence and preventability of adverse drug events (ADEs) and potential ADEs and to analyse these events in order to develop prevention strategies.

Not as rare as you may think

Answer the following question. Do you think that ADEs in hospitals in the USA, compared with road traffic accidents in the USA, are the cause of:

A: many fewer deaths?
B: about the same number?
C: many more?

According to the introduction to the paper [1], which summarises some of the information about this, the correct answer is C. Roads are safer in the UK, so the same answer to the same question is even more likely in the UK.

Adverse drug event

An ADE is defined as an injury resulting from medical intervention relating to a drug. This definition is different from that of an adverse drug reaction (ADR), defined by WHO as "noxious and unintended, and which occurs at doses used in man for prophylaxis, diagnosis or therapy". The ADR definition is likely to be idiosyncratic and rare, and would, for instance, not include an unintended misprescription of a larger than therapeutic dose of a drug which caused harm because it was an overdose. Such an event would be an ADE. It should not occur, and is likely to be preventable.


All adults at the Brigham and Women's Hospital (726 beds) and Massachusetts General Hospital (846 beds) admitted to any of 11 units over six months were included in the study. Obstetric units were not included. The 61 non-obstetric adult units were stratified between hospital, whether medical or surgical and whether intensive or general care. Study units were then selected randomly using a random number generator.


Three methods were used to identify incidents:

  1. Nurses and pharmacists were asked to report incidents to nurse investigators.
  2. A nurse investigator visited each unit twice daily on weekdays to solicit information.
  3. The nurse investigator reviewed all charts at least daily.

The primary outcome was an ADE, and a secondary outcome a potential ADE (an example would be a patient with known sensitivity to penicillin who was given penicillin, but did not react). To discover the cause of preventable ADEs, persons involved were interviewed and the results investigated by a multidisciplinary team.

All incidents were evaluated independently by two physicians who classified them according to the following criteria: whether or not an ADE or potential ADE had occurred, severity, preventability, and, if an error had occurred, the type of error and the stage in the process at which it occurred. Reviewers were asked to consider ADEs as preventable if they were due to an error or were preventable by any means currently available.


In the study period, there were 4,031 admissions to the study units, comprising 21,412 patient-days (about 10% of the 214,000 patient days in adult, nonobstetric units at the two hospitals).

The study found 247 ADEs and 194 potential ADEs. Extrapolated, this amounted to 1900 ADEs per hospital per year, with, for every 100 admissions, 6.5 ADEs and 5.5 potential ADEs. Of all ADEs, 1% were fatal, 12% life-threatening, 30% serious and 57% significant.

The rate of ADEs was highest in medical intensive care units (19 per 1,000 patient-days) and relatively similar among surgical intensive care and medical and surgical general care units (9 - 11 per 1,000 patient-days).

Over 50% of all ADEs were associated with the use of analgesics (30%) or antibiotics (24%). No single drug accounted for more than 9% of ADEs. Analgesics were the leading drug class associated with preventable ADEs, and half of these involved misuse or malfunction of infusion pumps or devices (epidural catheters or patient-controlled analgesia).

The Table shows the data by actual numbers in the study, by the rate per 100 admissions, and by extrapolation to events per hospital per year. It is instructive that even the lowest of these event rates adds up to a significant number of patients.
Event Number Rate per 100 admissions Number per hospital per year
All ADEs 247 6.5 1923
Due to analgesics 73 1.9 568
Due to opiates 57 1.5 444
Preventable ADEs due to pump or device malfunction 9 0.2 70

Preventable ADEs

Seventy (28%) of 247 ADEs were preventable and 83 (43%) of 194 potential ADEs were intercepted before the drug was given. Errors resulting in preventable ADEs occurred most often at the stages of ordering (56%) and administration (34%); transcription and dispensing errors were uncommon. Errors were more likely to be intercepted if they occurred early in the process - 48% at the ordering stage but none at administration of the drug.


ADEs generate costs to the patient, costs to the hospital though treating the effects of the ADE, and costs (at least potentially) of medical negligence claims. The hospital cost alone was estimated to cost $2,000 per event - making about $3.8 million per hospital per year. $1 million of this was preventable today, notwithstanding the human cost.


How might a hospital improve the quality of its drug delivery process? To get the full flavour of the lessons from this study, read the paper and some of the many useful references. In précis the lessons are:

  1. Have an effective mechanism for systematically collecting and feeding back data about ADEs.
  2. The organisation must look for preventable ADEs, not just ADRs.
  3. System change must be directed at specific parts of the process where error can occur, or where errors can be minimised.

Total Quality Management

The total quality improvement perspective assumes that most providers are doing their best in the current system. Major improvements in system performance require redesign of systems rather than pushing people harder with the current system. Better systems should promote fewer errors and include effective mechanisms for catching those that do occur. They also cost less.

Bandolier is curious to know the equivalent ADE numbers in primary care.


  1. DW Bates, DJ Cullen, N Laird et al (for the ADE Prevention Study Group). Incidence of adverse drug events and potential adverse drug events. Implications for prevention. Journal of the American Medical Association 1995 274:29-34.

next story in this issue