Historically, additional critical care capacity in a trust has usually occurred in response to repeated crises. This reactive approach is clearly less than ideal and a better solution involves planned or staged alterations in bed capacity to anticipate changes within the trust.
Judging the correct critical care capacity for a hospital's size or service is an important but very difficult process. Critical care is expensive so over-provision is not financially realistic.
Under-provision, on the other hand, can have serious consequences for patients with life-threatening illnesses or awaiting major surgery. Because of their cost, there is keen political interest in critical care beds and, along with day beds and total hospital beds, they are subject to a bi-annual bed census.
Historically, additional critical care capacity has usually occurred in response to repeated crises. This reactive approach is clearly less than ideal and a better solution involves planned or staged alterations in bed capacity to anticipate changes in the trust.
Key to good planning is a detailed understanding of the existing demands on critical care beds and accurate prediction of the likely effects of changing capacity. This requires appropriate classification of the patients and the use of detailed mathematical models at the level of individual patients..
Various classification tools are available, but one of the most helpful is classification and regression tree (CART) analysis. With CART analysis, the necessary classification can be derived using criteria such as type of patient (elective or emergency), age, source of admission, speciality and organ failures.
Variability in patients' lengths of stay is taken into account and the arrival patterns of the patients, by month, by day of the week, and by time of day can capture the uncertainties involved, when patients are likely to present.
Models of capacity
The combination of CART and historical data analysis, together with the appropriate information about the critical care unit and its operating rules, can be used to develop detailed models for critical care capacity.
In a large teaching hospital in England, there were repeated proposals by the departments of medicine and surgery to increase their separate four-bedded high dependency units (HDUs) to 14 and 10 beds respectively.
CART and the other statistical analysis of the data provided the necessary inputs for models for the separate units. A variety of scenarios were evaluated and these provided quantitative information for making robust decisions about HDU organisation and capacities...
Table 1 shows the effects of increasing the capacities in the separate HDUs on the number of patients that can be admitted and on the percentage of time when no beds would be available. Decisions about the capacities required can be taken on the basis of the likely demand for high-dependency units (as measured by the number of annual patients to be admitted) and on the level of availability of the service to patients (as measured by the percentage of the time when the beds will not be available)......
Mathematically, and intuitively, it is known that small units can be inefficient in terms of patient throughput. The mathematical expectation is that for a given level of patient service, larger units can admit more patients.
Table 2 demonstrates the penalties of organising critical care in small units by comparing separate surgical and medical high dependency units.of.eight beds each with a combined HDU of 16 beds. With a level of patient service that has no bed available 5-10 per cent of the time, the combined unit can admit 2,254 patients in comparison with a total of 1,935 patients for the separate units with the same total of 16 beds.
This is an increase of 16.5 per cent in the number of patients able to be admitted. With the higher level of service that would have no bed available for only 0.5-1 per cent of the time, the combined HDU would allow an increase in admissions of 26.5 per cent...
This type of work allows improved strategic planning and offers better predictions of the results of investment in service improvement. Even if the decision is not to alter capacity immediately, this type of modelling may produce, as incidental findings, results that could improve the delivery of care.
This kind of data analysis and modelling can be carried out in any critical care unit that has reliable historical data. This approach quantifies the effects of various step changes in capacity and allows more informed decision-making. Such planning will also allow performance review since actual improvements can be compared to the predicted changes.
The models do require careful preparation and statistical analysis of the data, and skills for tuning the models to represent a particular critical care unit.
After data preparation, statistical analysis and tuning of the model, the bed capacity software can evaluate several scenarios in a few minutes. This approach to better understanding capacity and demand issues is being actively promoted via critical care networks and the Intensive Care Society...
This work is being undertaken by the society's working group on patient flows in critical care and funded by a grant from the Department of Health.
Dr Mick Nielsen is an intensive care consultant at Southampton University Hospitals trust. Dr Saxon Ridley is an intensive care consultant at Conwy and Denbighshire trust. Dr Arjan K Shahani is director of the health data analysis and modelling group at the GeoData Institute, Southampton University.