INVEstigating Safety Tipping points in Swiss hospitals: assessing the causal effect of capacity utilization on in-hospital mortality using routine data [INVEST]

Sharma, Narayan. INVEstigating Safety Tipping points in Swiss hospitals: assessing the causal effect of capacity utilization on in-hospital mortality using routine data [INVEST]. 2021, Doctoral Thesis, University of Basel, Faculty of Medicine.

Available under License CC BY-NC-ND (Attribution-NonCommercial-NoDerivatives).


Official URL: https://edoc.unibas.ch/88110/

Downloads: Statistics Overview


Measuring patient safety is challenged by the differences in methods for measurement, variation in services provided, and care demands in hospitals. To minimize them health service providers, policymakers and researchers investigate potential causes of safety issues. Capacity utilization, (i.e., bed-occupancy rate), is one aspect of safety tipping points i.e., high capacity utilization levels in which patient safety incidents like in-hospital mortalities are more likely to occur. To see whether the entire hospital system can supply sufficient resources to match the care demand, several aspects of care like the volume of patients, turnover, and severity of patient’s illnesses need to be considered and their link with outcomes such as mortality should be measured.
Previous studies have shown the association, particularly capacity utilization and in-hospital mortality. Yet, the time-varying nature of capacity utilization in hospitals with its time-varying confounders was not investigated at all. Even though, studies included other individual factors and comorbidities in the model of capacity utilization and mortality. The risk adjustment of both individual and time-varying covariates with a causal inference methodology in health service research could enhance our understanding based on observational data. Applying the causal perspectives in the routine data helps to understand patient safety in the absence of randomized experiments.
Therefore, the overall aim of this dissertation was to investigate different aspects of the safety tipping phenomenon. We first explored hospital care demand from a longitudinal perspective, mainly daily capacity utilization, patient turnover, and patient clinical complexity level (PCCL) of 102 Swiss general hospitals. Secondly, we derived Swiss comorbidity weights and compared these with existing weights to predict in-hospital mortality. Finally, a time-varying causal model with individual factors was explored via directed acyclic graph (DAG) and the potential causal effect of daily capacity utilization on in-hospital mortality was estimated.
This dissertation is embedded in the INVEST (INVEstigating Safety Tipping points in Swiss hospitals) study, focusing on in-hospital mortality as a patient safety indicator, which is all-cause mortality from the routinely collected data of all Swiss hospitals from 2012 to 2017. Overall, this dissertation is structured in six chapters.
Chapter 1 provides a general introduction in the field of patient safety, in-hospital mortality, system thinking, and the tipping point phenomenon in hospital care. Furthermore, a brief outline of comorbidity and weighting systems to predict in-hospital mortality is provided. Additionally, an introduction of causal inference with time-varying variables and the causal effect estimations are offered for capacity utilization and in-hospital mortality. Chapter 2 states the aims of this dissertation.
The article presented in Chapter 3 constitutes the findings of care demand variation in Swiss hospitals from the longitudinal perspective. We utilized one-year patients’ data to explore the variation of each hospital care demand. Hospital care demand was longitudinally explored as a percentage of daily capacity utilization, percentage of daily patient turnover, and average daily patient clinical complexity level values per general hospital. Patient clinical complexity level (PCCL) was measured as a cumulative effect of a patient’s clinical complexities or comorbidities (CC) for each episode of care and it ranged from 0 to 4, i.e., no CC to very severe CC. We used Swiss diagnosis-related groups version 6 for compiling clinical conditions reported in ICD 10 codes for generating each patient’s PCCL value. The results indicate the average daily capacity utilization (87.7%) and average daily PCCL value (2.06) were highest in university hospitals but the patient turnover (22.5%) was lowest. Similarly, patient turnover (34.5%) was highest in medium basic hospitals but the lowest daily PCCL value (1.26). Moreover, the capacity utilization (57.8%) was lowest in small basic hospitals. The findings suggest there was pronounced variability of all three measures of care demand in Swiss hospitals and they distinctly varied between days, weeks, and seasons throughout the year. Variation and daily trends in care demand required a balanced care supply in Swiss hospitals.
Chapter 4 describes different comorbidities indices in Swiss hospitals, one of the measures of patient complexity and the predictive performance of different comorbidity weighting systems with in-hospital mortality. We used six years of Swiss general hospital data with a total of 6.09 million patients’ cases for the analysis. The data were randomly split into two halves to derive and validate Swiss comorbidity weights from the Elixhauser comorbidity index. Both sets were used for validation of the new weights and to compare new weights with existing comorbidity weights to predict in-hospital mortality. Derivation and validation of weightings were conducted with generalized additive models adjusted for age, gender, and hospital types, and the comparison was done using c-statistic and net reclassification improvement (NRI). Overall, c-statistic with Swiss weights (0.867, 95% CI, 0.865–0.868) was slightly higher than van Walraven’s weights (0.863, 95% CI, 0.862–0.864) and Charlson’s weights (0.850, 95% CI, 0.849–0.851). The NRI of new Swiss weights improved the predictive performance by 1.6% on the Elixhauser-van Walraven and 4.9% on the Charlson weights. The patient population-based Swiss weights support the analysis of in-hospital mortality and other health outcomes. Comorbidities/comorbidity scores are important individual factors for risk adjustment for hospital exposure and health outcomes.
Chapter 5 analyzed the causal effect of capacity utilization on in-hospital mortality in Swiss hospitals. Grounded on the variation of the time-varying measures of care demand (Chapter 3), we hypothesized exposure to daily capacity utilization has a causal effect on 14-days in-hospital mortality. Time-varying confounders and time-fix variables of capacity utilization and in-hospital mortality were linked along with an unmeasured variable using DAGs. Daily capacity utilization in Swiss hospitals was explored to observe safety tipping points at the 85th percentile of the distribution on the hospital level. To test our hypothesis, we used one-year patient data, adding 14-days exposure to capacity utilization, the time-varying confounders, patient turnover, and PCCL and time-fixed variables weekdays, hospital types, Elixhauser comorbidity index weighting score, age, and sex. Inverse probability of treatment weights (IPTW) was computed to balance the weight between high capacity utilization and low capacity utilization from the safety tipping points cutoff of 85th percentile and to eliminate treatment-confounder feedback. The computed IPTW was incorporated with a marginal structural model (MSM) to estimate the causal effect of capacity utilization on in-hospital mortality. The finding shows that daily exposure to high capacity utilization yielded a 2% increase in 14-days in-hospital mortality for each additional day of high capacity utilization (Odds Ratio (OR) 1.02, 95% CI: 1.01 to 1.03). Additionally, we found weekend effects and increased mortality risk with increasing comorbidity scores in the Swiss patient population. The variation of safety tipping points between hospitals suggests Swiss hospitals get strained at different points ranging between 42.1% to 95.9%. Thus, hospitals require the planning of resources based on the intensity of the care demand.
Chapter 6 of this dissertation summarizes the key findings of all studies and discusses them in the context of the literature. Furthermore, the strengths and limitations of the studies are discussed, and implications of causal inference in an observational study and the application of evidence in hospitals are presented. This dissertation contributes to the current literature in the field of patient safety and the causal relationship between capacity utilization and in-hospital mortality in an observational study. It identifies safety tipping points variation in terms of capacity utilization in Swiss general hospitals. After reaching tipping points, hospitals are under strain, and matching hospital resources with care demands is required to improve patient safety and quality of care.
Advisors:Simon, Michael
Committee Members:Schwendimann, René and Ausserhofer, Dietmar and Endrich, Olga and Needleman, Jack
Faculties and Departments:03 Faculty of Medicine
UniBasel Contributors:Sharma, Narayan and Simon, Michael and Schwendimann, René and Ausserhofer, Dietmar
Item Type:Thesis
Thesis Subtype:Doctoral Thesis
Thesis no:14682
Thesis status:Complete
Number of Pages:vii, 158
Identification Number:
  • urn: urn:nbn:ch:bel-bau-diss146827
edoc DOI:
Last Modified:19 May 2022 04:30
Deposited On:18 May 2022 08:36

Repository Staff Only: item control page