How routine clinical data help with quality measurements in hospitals

The advancing digitalisation in hospitals means that clinical data are increasingly available electronically in the hospital information system. These data have so far been used for internal quality measurements, but hardly for external comparison with other hospitals. The potential for the latter was investigated within the framework of the “Vis(q)ual Data” project.

In Switzerland, healthcare providers are obliged to ensure and improve the quality of care. For this purpose, national quality measurements are carried out according to the Health Insurance Act (KVG) [1, 2]. Up to now, the data basis of quality measurements has been based on two data sources, the primary clinical and the administrative data [3]. Both data sources have their advantages and disadvantages.

  • Administrative data have the advantage that they are standardised and available with little effort. This includes billing data or data that must be submitted to the Federal Statistical Office. Disadvantages are that the data for certain topics are not very differentiated and important variables for risk adjustment (see explanation in the box) are missing [4].
  • Primary clinical data are collected specifically for the purpose of quality measurement, e.g. through direct observation at the patient’s bedside. These data are therefore highly differentiated and all necessary risk adjustment variables can be captured. The disadvantage of these data is that the collection of data is associated with high personnel costs [4].

Due to increasing digitisation in hospitals, clinical data are also increasingly available electronically in the hospital information system (HIS) [5]. These routine clinical data have so far been used for internal benchmarking (e.g. comparison of departments) and for internal quality improvement purposes (e.g. before and after the introduction of a new guideline). For external benchmarking, i.e. the comparison of hospitals (e.g. within the framework of a national quality measurement), clinical routine data have hardly been used so far. In Switzerland, to our knowledge, there is no national quality measurement based on routine clinical data. Even in the USA, according to the measurement portfolio of the National Quality Forum, only 2 of the 76 data collected for quality measurements in hospitals are based on routine clinical data [6]. Thus, in principle, it seems feasible to use routine clinical data for external comparisons, but the potential is unlikely to be fully exploited. For this reason, we have investigated whether routine clinical data in Switzerland can be used for national quality measurements.

The “Visual (Quality) Data” project

To investigate the feasibility in Switzerland, we used as a reference measurement. The national prevalence measurement of falls and pressure ulcers is a national quality measurement in which all hospitals in Switzerland that have joined it must participate. In this measurement, nurses collect defined data from all inpatients on one day per year with a questionnaire on falls or a skin inspection in order to be able to identify a pressure ulcer. This data, which is collected directly at the patient’s bedside, could also be supplemented by data from the HIS (e.g. age, diagnoses), if not recorded directly from the patient. Hospitals criticise the high personnel costs for this data collection because, according to their statements, all the necessary data are already available electronically in the HIS and could be used [5]. Based on this initial situation, we investigated with three hospitals in German-speaking Switzerland whether and how the necessary routine clinical data are available in the respective HIS [6]. Clinical routine data were defined as medical, nursing and other clinical records that contain information about the patient’s state of health or the results of (nursing) assessments. Specifically, the Vis(q)ual Data project sought to replicate prevalence measurement using routine clinical data. The results of the “Vis(q)ual Data” project were:

  • 20 of the 21 variables required for the prevalence measurement could be exported by all hospitals. The only variable that could not be exported was care dependency [7]. However, it became apparent that some of the variables were operationalised differently. Among other things, clinical data such as the assessment of pressure ulcer or fall risk were operationalised differently by the hospitals.
  • 18 of the 20 variables showed comparable descriptive results as in the reporting of the national prevalence measurement.
  • 6 of the 6 relevant and available variables for risk adjustment “pointed in the same direction”. That is, the variables were similarly correlated with outcome using the routine clinical data as in the reporting of the national prevalence measure.
  • from a technical point of view, a data extract took 5 to 1 working day according to the predefined data model.

Conclusions and visionary outlook

The results from the “Vis(q)ual Data” project show an overall positive picture. Most of the required data are available in the hospital information system, the technical feasibility for their export is given and the indications regarding data quality are positive in that the descriptive results as well as the risk adjustment model turned out to be similar to the reporting of the prevalence measurement. Nevertheless, there are some challenges that need to be addressed before routine clinical data can replace clinical measurement at the patient’s bedside. These are mainly in the area of operationalisation. In order to conduct fair external benchmarking, the data basis must be comparable. For the outcome, it must be ensured that it is recorded in the same way in all hospitals. For the risk adjustment and the variables used with it, it is either necessary to further explore whether and how the different operationalisation affects the results (e.g. whether a patient is assessed as being at risk of decubitus or not), or also to define a uniform standard. In addition, care dependency, which is an important risk adjustment variable in the prevalence measurement, could unfortunately not be included [7]. It must therefore be examined whether care dependency can be mapped by means of a proxy variable (e.g. a combination of diagnoses, medication and hours of care). The findings from German-speaking Switzerland are currently being validated in a follow-up project with other hospitals from French-speaking Switzerland and Ticino, in order to be able to formulate subsequent recommendations on the use of routine clinical data for national quality measurements in Switzerland. If it were possible to make clinical routine data usable, Switzerland’s measurement portfolio could be expanded in future to include further quality indicators with little personnel expenditure in the hospitals, up to and including longitudinal surveys or even live monitoring [8].


About risk adjustment

The patient mix in hospitals can vary greatly (due to the service mandate). It is therefore possible that one hospital treats many patients with an increased risk of falling, while another hospital treats a majority of patients with a lower risk. This circumstance must be taken into account when comparing the fall rates of hospitals with each other. For this purpose, so-called risk adjustments are made. These are statistical procedures in which the results of a hospital are adjusted for the risk of the patient group being cared for. Accordingly, in addition to the result (fall yes/no), the data basis must also reflect the risk factors (age, dependence on care, fall risk assessment, etc.) [4].


Bibliography

  1. Federal Office of Public Health [FOPH]. (2019). Evaluation der KVG-Revision im Bereich der Spitalfinanzierung – Schlussbericht des BAG an den Bundesrat. Bern: Federal Office of Public Health.
  2. Vincent, C. & Staines, A. (2019). Enhancing the Quality and Safety of Swiss Healthcare. Bern: Federal Office of Public Health.
  3. Busse, R., Klazinga, N., Panteli, D. & Quentin, W. (2019). Improving healthcare quality in Europe: Characteristics, effectiveness and implementation of different strategies. World Health Organization and OECD. Retrieved from: https://apps.who.int/iris/rest/bitstreams/1248308/retrieve
  4. Bernet, N., Everink, I., Schols, J., Halfens, R., Richter & D., Hahn, S. (2022). Hospital performance comparison of inpatient fall rates; the impact of risk adjusting for patient-related factors: a multicentre cross-sectional survey. BMC Health Services Research. https://doi.org/10.1186/s12913-022-07638-7.
  5. Bernet, N., Thomann, S., Kurpicz-Briki, M., Roos, L., Everink, I. H., Schols, J. M. & Hahn, S. (2022). Potential of Electronic Medical Record Data for National Quality Measurement. In Healthcare of the Future 2022 (pp. 51-56). IOS Press. https://doi. org/10.3233/SHTI220320
  6. National Quality Forum. National Quality Forum: Find Measures. Retrieved from: https://www.qualityforum.org/Qps/QpsTool.aspx.
  7. Dassen, T., Balzer, K., Bansemir, G., Kühne, P., Saborowski, R. & Dijkstra, A. (2001). The care dependency scale, a methodological study. Nursing, 14(2), 123-127.
  8. Swiss Personalized Health Network. About SPHN. Retrieved from: https://sphn.ch/organization/about-sphn/.

AUTOR/AUTORIN: Leonie Roos

Leonie Roos is a qualified nurse MScN and works as a research assistant in the innovation field of quality in healthcare at the Department of Health of the Bern University of Applied Sciences.

PDF erstellen

Related Posts

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *