A model was developed to explain the demand for care based on readily available patient characteristics. Number of medications during hospitalisation, number of co-morbidities, number of complications, age, surgical specialty, as well as undergoing a surgical intervention and length of stay significantly contributed to an increased demand for care. It is likely that these results are generalizable to other specialties because these are blanket factors, applicable to a broad patient population.
No significant associations were found between the patient’s ASA class, nutritional status, admission type and their demand for care. This is partially in agreement with the results from other investigators , a weak but significant correlation (r=0.35 p<0.0001) between admission type and nursing workload. For ASA class and nutritional status, no comparable evidence is available. ASA class appeared to be a promising influencing factor in the univariable analysis, but was found not significant in the multivariable analysis. Probably too few patients belonged to ASA class 3, because we found significant associations between ASA classes 1 and 2, and between 1 and 3, but not between classes 2 and 3. Also delirium, pressure ulcers, patient isolation, and in-hospital mortality were not significantly associated with demand for care. This is likely because their incidence was quite low in our study, but not unusual for these wards. Furthermore, these factors are less useful as factors predicting the demand for care because they occur during hospitalisation and are not known beforehand. If they would contribute significantly to the model, they can still be useful as a managerial tool to monitor amount of care on a more aggregate level on wards to detect trends in time as to patients’ demand for care.
Some nursing care models have found the case-mix groups (CMG) or Diagnosis Related Groups (DRGs) to be explanatory factors for the demand for nursing care [6, 9, 11–13]. In this study the investigators categorised the medical diagnoses at a more abstract level, i.e., surgical specialty, because of the large variety in diagnoses present. This specialty appeared relevant as it showed to be an important significant factor, explaining 49% of the variance in the demand for care in terms of costs.
The number of complications during hospitalisation also had a large influence on the demand for care. This number is likely to be related to co-morbidity and medication. Therefore, this number seems a sensitive indicator for the complexity of care and the following demand for care. Complexity is an important concept in research as to the demand for nursing care [12, 14]. In the nursing realm, complexity has been measured by parameters like severity of illness  nature of nursing tasks [12, 18] and nursing diagnoses [9, 12]. These variables had similar predicting values. The impact of complications on the demand for care was mainly due to the costs for diagnostic or therapeutic interventions, such as (redo) surgery to treat complications, and mostly occurred in patients undergoing gastro-intestinal surgery. This may be exemplary for the tertiary referral hospital in which this study was conducted.
The number of medications used during hospitalisation had less influence on the demand for care. No comparable evidence is available but this limited influence is possibly caused by the fact that medication is principally given to cure, and therefore associated with an increase in the demand for care. Also ‘age’ had less influence on the demand for care. This parameter nearly reached statistical significance (P=0.072) in the multivariable model and was added because of its clinical relevance. Such poor associations were also found by other researchers [8, 9, 12].
The negative association found between co-morbidity and demand for care may be because the severity of the various co-morbidities was not weighed in this study. Less severe co-morbidities may have been managed through medication, while patients with more severe co-morbidities were less likely to undergo surgery. This is confirmed by the study of Gijsen et al. . They proposed the Charlson Comorbidity Index (CCI) to operationalise the severity of the co-morbidity.
Also, undergoing a surgical intervention and length of hospital stay were significant factors associated with the demand for care. This seems obvious, given the additional costs of surgery and of each extra day spent in the hospital. Previous studies have shown this is likely to be related to the severity of the patient’s illness and therefore their demand for care [2, 10, 12].
Some limitations of this study should be discussed. First, the investigators calculated and modelled the care the patients received, which may not be commensurate with what they needed. We did check that the results of our study represented demand for care rather than the mere usage of personnel and resources. The delivered care was independent of bed occupancy and available personnel. This suggests that indeed the demand for care was measured instead of offered resources. In retrospect, the investigators might also have appreciated whether the care given had met the patient’s expectations and had cured or relieved their disorder.
Second, the investigators used a diversity of input, structure, process, and outcome variables in the model. As mentioned earlier, variables occurring during hospitalisation are unknown beforehand and therefore not useful as predictive factors. It seems plausible to use input variables for the explanatory model and use process and structure variables as specialty-specific or centre-specific characteristics, e.g. undergoing a surgical intervention, level of education [7, 8], or organisational factors [12, 13], in an additional model. Furthermore, the success of the care given could also have been estimated, e.g. by measuring outcome variables as the number of complications or readmissions within 30 days after dismissal or by appreciating the quality of care . This was beyond the possibilities of the present study, but will be incorporated in a recently started follow-up study among Dutch top-clinical hospitals.
Third, the investigators took for this study an innovative approach to measure the demand for care by time and motion research. This method was performed with rigour to collect data on individual patient contacts by professionals. Otherwise, continuous time and motion research provides precise results only if the professionals involved are willing to accurately record the time spent. The investigators found under-recording of time data, predominantly among doctors, resulting in an under-reporting of the total costs involved. Although this will have weakened the power of our model to predict demand for care, there was no reason to suspect selective under-recording that would have influenced the ability to detect predictive characteristics. It may explain, however, that the demand for care in our model appeared determined by the costs of the surgical and diagnostic interventions rather than the costs of personnel outside the operating theatre. As the investigators could not incorporate all costs at the same level of detail (e.g. overhead cost on wards or surgical interventions were not taken into account), a representative estimate was used of the costs for (para)medical and nursing care during admission. However, the overhead costs are likely to be proportional to the personnel costs we measured and therefore not influencing the outcome of our model.
Fourth, by expressing the demand for care as costs, the contribution of unpaid medical trainees to the patient care was not taken into account, although they deliver a substantial contribution to patient care in university clinics and affiliated hospitals. In addition, costs for overhead, patient transport, medication, material costs for surgical procedures in the operating theatre or on the nursing ward were not taken into account, while costs for ICU- and recovery stays were entered as fixed costs. Finally, no additional charges were included for surgical interventions during weekends, evenings and nights. Further detailing of these costs was beyond our possibilities but it is doubtful whether this would have had a major impact on the general outcome of our study.