Quality Standards for Diabetes Care Toolkit

Implementation examples / innovations

Reducing length of stay and improving quality of care for inpatients with diabetes (DIPLoS)

Three parallel prospective controlled studies examined three different process models of delivering inpatient care in Auckland, Lakes and Mid Central Health. Each ran over three consecutive periods: A baseline run-in phase (two months), the intervention period (four months), a return to baseline run-out phase (two months). Data have also been collected from two additional non-intervention sites (Waikato and Wellington) to control for external/seasonal factors.
The REACTIVE model (Lakes) followed the current usual pattern of ward-initiated referrals with subsequent intervention as above but, for the intervention period, prioritised this activity above other routine tasks with little change in overall inpatient staffing.
The PROACTIVE intervention (MidCentral) employed daily (weekday) ward visits/phone calls and electronic notification seeking to identify a person with diabetes early during an admission. Nursing intervention was then promptly arranged to optimise care and minimise length of stay. This method used increased dedicated nurse inpatient staffing for the period of the study.
In the ELECTRONIC group (Auckland), all admissions of people with known diabetes from previous admission ICD-10 coding were notified by email to the diabetes inpatient team. Central triaging then prioritised patients to arrange intervention as above where needed, but with no change in inpatient staffing.
Headline results included:

the REACTIVE intervention produced no significant change in time to ward visit or length of stay

the PROACTIVE method led to an approximate doubling in the number of patients seen but no significant changes in time to ward visit or length of stay

the ELECTRONIC method led to a 35% increase in patients seen, a >50% reduction in time to ward visit (3.9 to 1.8 days; p<0.001) and an apparent 30% reduction in length of stay (8.6 to 5.9 days; p<0.005). Part of the LoS change could potentially be an artefact caused by earlier visits and further analyses are under way to address this (Drury 2011).