This is a retrospective, population-based analysis using 1 year of hospital data extracted from the 6-year (2012-2017) dataset obtained from the Swiss Federal Statistics Office (FSO). Based on a data protection contract (as stipulated by article 22 of the Swiss Federal Act on Data Protection), the FSO provided anonymized data from all Swiss hospital inpatients hospitalized over the study period. The data covered general as well as specialized care facilities such as pediatric, gynecological, psychiatric, and rehabilitative hospitals [34]. The FSO divides general hospitals into 5 classifications: university hospitals, tertiary care hospitals, large basic hospitals, medium basic hospitals, and small basic hospitals. Classification is based on the number of cases treated per year and a weighted sum of service points (based on a combination of the number of hospital units and the levels of care delivered) assigned by the Swiss Medical Association [17,34]. For instance, based on the Swiss Medical Association classification, a university hospital requires the weighted sum of service points to be >100 units and >30,000 cases per year [34]. For this study, we included a 1-year patient population dataset from general hospitals to limit interhospital heterogeneity. Due to the Swiss Data Protection Act’s stipulations regarding patient anonymity, we were unable to trace patients across years. An overview of the process of selecting inpatient cases for analysis is included in the flow diagram in Figure 1.

We extracted 1-year datasets, including (1) all inpatients discharged during each study year, regardless of admission date, and (2) all inpatients admitted during each study year, regardless of discharge date. From the full FSO hospital dataset, we also extracted data for all relevant routine administrative and clinical variables for the study period (Multimedia Appendix 1).

All statistical analyses were conducted using R, version 3.6.3 for Mac OS [35]. The following statistical packages were used: (1) dplyr [36] and tidyr [37] for data preparation; (2) lubridate [38] and stringr [39] for handling time and date; and (3) ggplot2 [40], patchwork [41], and scales [42] for plotting.

During the study year, 1,214,875 inpatients stayed in the 102 Swiss general hospitals, of which 16,379 cases (1.35%) continued to the following year. Of the 1,214,875 inpatients, the 5 university hospitals covered 19.50% (n=236,649), the 39 tertiary care hospitals covered 59.70% (n=725,317), and the 58 basic hospitals covered 20.81% (n=252,909) of the patient population. Overall, of the 1,214,875 inpatients, there were approximately 6.87% (n=91,078) more female than male patients. The average patient LOS across all general hospitals was 6.43 (95% CI 6.40-6.46) days; the median LOS was 3.7 (IQR 2.0-7.0) days. The general characteristics of the study population by hospital type are presented in Multimedia Appendix 1.

Average daily capacity utilization ranged from 527-2340 patients in university hospitals, 87-1099 patients in tertiary care hospitals, 16-304 patients in large basic hospitals, 7-179 patients in medium basic hospitals, and 1-93 patients in small basic hospitals. Notably, 3 small basic hospitals had average daily capacity utilization numbers below 10 patients.

Across the study period, the average daily capacity utilization was highest in university hospitals and the lowest in small basic hospitals (Table 1). However, the range (ie, the variation between the lowest and the highest daily capacity utilizations for a study year) was almost 98% (eg, 1.7-100.0) for small basic hospitals, 92% for medium basic hospitals, 87% for large basic hospitals, 73% for tertiary care hospitals, and 44% for university hospitals.

As indicated by the smooth curves and line charts by hospital categories, university hospitals’ daily capacity utilizations were high throughout the year. Among all hospital types, capacity utilization was lower throughout the summer months (June-August; Figure 2).

There was a gradual increase in capacity utilization through the early days of the week (Mondays-Wednesdays; Figure 3), followed by a gradual decrease from Fridays to Sundays, across all hospital types. There was roughly a 10% difference in capacity utilization during weekdays than on the weekend, in all hospital types (Table 2).

Comparing weekdays with weekends, variations in capacity utilization for each week are shown in Figure 4. With very few exceptions (eg, the 2nd, the 17th, and the final week of the year), the weekly capacity utilization for weekdays was higher than for weekends. These weekly graphs also show lower capacity utilization during the summer months (weeks 20-35).

Throughout the year, daily patient turnover ranged from 72-468 patients for university hospitals, 5-318 patients for tertiary care hospitals, 0-91 patients for large basic hospitals, 0-78 patients for medium basic hospitals, and 0-33 patients for small basic hospitals. The minimum value of 0 indicates that some hospitals saw neither admissions nor discharges on some days.

During the study year, the mean daily patient turnover percentage was highest in medium-sized basic hospitals and lowest in university hospitals (Table 1). The difference in daily patient turnover range was highest for small basic hospitals, decreasing with each increase in hospital size class. And, as illustrated in the smooth and line chart, daily turnover varied the least in university hospitals and the most in medium basic hospitals (Multimedia Appendix 2).

Exploring the frequency of change in patient movement by day of the week (Multimedia Appendix 2), we found that turnover was highest on Mondays and lowest during weekends. Differences in patient turnover during weekdays and weekends across the 5 hospital types are shown in Table 2. Across the 5 hospital types, daily patient turnover was almost 10% higher on weekdays than on weekends.

The variation of patient turnover for each day of the week and for weekdays vs weekends is shown in Multimedia Appendix 2. In small basic hospitals, with few exceptions (eg, the 31st and 52nd weeks), the mean daily patient turnover was higher for weekdays than for weekends, for all hospital types. Overall patient turnover was lower during the holiday seasons (ie, weeks 17-18 and the final 2 weeks of the year).

Overall, of the total discharged patients (1,198,496), 10.97% (n=131,442) of patients had severe clinical complexity (patient clinical complexity level 4), while roughly 60.97% (n=730,893) had no clinical complexity (patient clinical complexity level 0; Table 3).

Throughout the year, mean daily patient clinical complexity level varied across the 5 hospital types. It was highest in university hospitals (2.06, 95% CI 2.05-2.07) and lowest in medium basic hospitals (1.26, 95% CI 1.25-1.27; Table 1). This is depicted in the smooth and line chart for the 5 general hospital types (Multimedia Appendix 3). Mean patient clinical complexity level gradually decreased from Monday until midweek but remained highest during the weekend—the opposite of the usual pattern of capacity utilization and patient turnover explored by the days of the week (Multimedia Appendix 3).

Weekday and weekend differences in patient clinical complexity level for the 5 general hospital types are shown in Table 2. Across all hospital types, patient clinical complexity level was almost 0.07 points higher during the weekend than on weekdays. During weekdays, university hospitals’ average daily patient clinical complexity level was 2.04 (95% CI 2.03-2.05); during the weekend, it was 2.1 (95% CI 2.08-2.13). Weekday vs weekend patient clinical complexity level variation over 1 year is shown in Multimedia Appendix 3. Except for a small number of weeks (eg, the 7th and the 31st weeks in small basic hospitals), across all hospital types, the patient clinical complexity level for weekends was higher than for weekdays. Moreover, except for small basic hospitals, patient clinical complexity level dropped in December. This was partly because patient clinical complexity levels could only be calculated for patients discharged during the study year (ie, ICD diagnostic codes were unavailable for patients not discharged during the year, and anonymity considerations made it impossible to track patients across years). However, mean values both for patient clinical complexity level and for LOS were also lower for patients discharged in November and December, with higher patient clinical complexity level values assigned to patients who remained in the hospital across the years’ end (Multimedia Appendix 4).

We examined 1 year of routine patient data from all 102 general hospitals in Switzerland. Average daily capacity utilization varied widely, from 57.8% in small basic hospitals to 87.7% in university hospitals. However, patient turnover was highest, at 34.5%, in medium basic hospitals and lowest, at 22.5%, in university hospitals. Moreover, average daily patient clinical complexity level was highest in university hospitals, at 2.06, and lowest in medium basic hospitals, at 1.26. Surprisingly, in small basic hospitals, patient turnover was lower than in tertiary hospitals or either of the 2 other basic hospital types, both of which also had higher mean patient clinical complexity levels throughout the year. Another interesting finding was that the average daily patient clinical complexity level was highest on weekends. Additionally, all hospital types showed distinct weekday, weekend, and seasonal effects regarding capacity utilization, patient turnover, and patient clinical complexity level.

Concerns have been raised that capacity utilization alone does not explain hospitals’ total care demand [6,18]. Therefore, we viewed this alongside daily volumes of admitted and discharged patients and complexity [18]. This study explored all 3 measures, showing that capacity utilizations and patient clinical complexity levels were highest but patient turnovers were lowest in university hospitals. Even with a large proportion of inpatients in tertiary care institutions, university hospitals generally operate at close to full capacity and with the most complex patient cases. Thus, more care resources need to be allocated to university hospitals [3]. On the other hand, in small basic hospitals, where capacity utilization and patient turnover were relatively low, patient clinical complexity level was above those of the other basic hospitals. This indicates that complex cases are still treated in small basic hospitals, possibly, due to geographic proximity, which may also relate to older patients’ preference for them: across all hospital types, these hospitals have the highest mean patient age. In light of these small basic hospitals’ continued relevance (as they still treat complex cases), they may warrant greater resource allocation.

We measured daily demand for Swiss general hospital care longitudinally for 1 year. As it has also been observed in other studies regarding days of the week, Saturdays and Sundays had the lowest capacity utilization and patient turnover [3,46]. Moreover, we observed that patient clinical complexity level was highest during weekends, possibly, because more complex patients remain in the hospital through the weekend. Comparing weekly demand throughout the year, a clear distinction between weekends and weekdays was shown, with the highest variability occurring in small basic hospitals, possibly indicating suboptimal patient flow. Concurrently, seasonal variations were also seen. Capacity utilization was mainly highest in the winter and relatively low in the summer months, whereas patient turnover was constant throughout the year, dropping off toward the end of the year. However, patient clinical complexity level remained quite constant, with a slight drop during the summer months and a marked reduction during December. These changes tended to correspond with holidays, possibly, including higher patient discharge rates and fewer admissions before the holidays and at the end of the year. The capture of daily patient complexity during the end of the year was also reduced because ICD-10 codes were unavailable for patients who were admitted but not discharged during the study year. Furthermore, patients with higher patient clinical complexity levels were more likely to have longer LOSs, particularly, across the Christmas and end-of-year period.

Based on capacity utilization, patient turnover, and patient clinical complexity level, the variability of daily care demand in general hospitals directly impacts resource use. From the perspective of a single hospital, the extent of that impact depends on the degree of variation in care need, as well as on the hospital administrators’ ability to adapt or otherwise respond to changes either in resource supply or demand. Our analysis on the demand dynamics of the Swiss health care system indicates that monitoring of care demand is useful to create surge capacity during disasters or the COVID-19 pandemic [20,47], by offering alternative solutions such as smoothing workloads and coordinating early discharges. It also has the potential to help health system planners and hospital managers to tailor their staffing and other resources to match care demand and the early planning of admissions (eg, surgeries or follow-up treatment) [48], to control patient flow for smoother service use.

What this analysis cannot describe is the human resources and other resources needed to meet care demands (ie, balancing care demands or any other supply-demand chain). To do so would require a full exploration of the relevant human resources (eg, physician and nurse staffing), in the light of each hospital’s care demand. Furthermore, application of time-driven activity-based costing methods could provide a framework to identify process improvements for health care delivery [49,50]. However, we were not able to consider time-driven activity-based costing, as we do not have sufficient data regarding resource consumption (eg, personnel, equipment, and supplies) during the patients’ journey along the clinical pathway [49].

Some studies have linked higher capacity utilization and patient turnover with adverse patient outcomes [5,21]. In addition to these results, acknowledging the effect of clinical complexity alongside capacity utilization and patient turnover might bring us closer to understanding the factors that stress hospital systems and the effects that a stressed system have on patient outcomes such as in-hospital mortality. Describing daily care demand to identify meaningful variation will require further studies (eg, examining patients’ time-varying exposure to hospitals or units under pressure and the impact on the quality-of-care indicators and patient outcomes in causal models). Particularly, extending previous research on capacity utilization and in-hospital mortality [5,51] and using daily capacity utilization as time-varying exposure (ie, systemic stress factor), it would be of interest to explore the daily patient turnover and patient clinical complexity level as time-varying confounders. In a practical sense, this might also allow monitoring of pressure zones (eg, to manage care demand, where possible) in hospitals, which could reduce avoidable adverse events or death [5].

To our knowledge, this was the first study to explore hospital care demand dynamics via daily measurements of capacity utilization, patient turnover, and patient clinical complexity level on a national health system level. Furthermore, applying the standard methodology, programming, and software for large datasets allowed a longitudinal perspective by computing and visualizing demand dynamics per day of the year, day of the week, and weekdays vs weekends.

This study also had notable limitations. While we explored demand dynamics in detail, we could not do so with supply dynamics (eg, staffing, resources, etc)—an entire category of critical information in the demand-supply equation. Due to the large sample and FSO data composition (ie, aggregated data), it was also not possible to explore demand dynamics at the unit level—the interface between patients’ care demand and health professionals’ provision of care. Also, as we used codes assigned in routine data, the patient diagnoses and other variables could be biased by factors such as the accuracy of physicians’ and nurses’ documentation, lack of availability of ICD-10 codes for patients who were not discharged, and intentional upcoding of diagnoses to more expensive Swiss DRG categories [52,53].

This study illustrates daily care demand based on capacity utilization, patient turnover, patient clinical complexity level, and the variability of these factors between the 5 classes of Swiss general hospitals. For all 5 types, our analyses indicated distinct differences in capacity utilization, patient turnover, and patient clinical complexity level between days of the week, weekdays vs weekends, and seasons. This longitudinal study is a step toward detecting possible variables to be considered for time-varying exposure (eg, capacity utilization) and confounders (eg, patient clinical complexity level) in developing a casual model of tipping points and their links with quality of care or patient outcomes. Essentially, the variability of care demand provides a new perspective for gauging when hospitals are under strain, and this could help avoid pressure zones with a combination of appropriate resource allocation and care-demand planning in general hospitals.