Routinely collected hospital data for prognosing surgical outcomes and costs: a diamond in the rough?

The accumulation of healthcare data has grown explosively over the past years. About 30% of all data that are collected today are generated by the healthcare sector. A substantial amount of these data are collected routinely from patients’ encounters with the healthcare system. These data might be a valuable source of information for other purposes, without additional costs of data collection.

In a recent study (currently under review), our research team (a collaboration between ESHPM, Amsterdam UMC and LOGEX Healthcare Analytics) investigated whether anonymised hospital benchmark data can be useful for predicting the outcomes and costs of four common surgical treatments. We found that these data can indeed provide clinically relevant insights into the drivers of outcomes and costs.

Value of care and prognostic factor research

Achieving excellent patient-relevant outcomes while at the same time minimising costs are the two main pillars in providing high-value healthcare. A fundamental first step in this is to define, measure and compare these outcomes across healthcare providers. A second step is to measure and compare the costs incurred in achieving those outcomes. Knowledge on which prognostic factors influence these outcomes and costs is paramount before providers can specifically target these factors and improve value of care.

Our study

In our study we investigated the associations between multiple patient factors and surgical outcomes and costs using routinely collected data in over 60 Dutch hospitals. The dataset comprised of anonymised hospital benchmark data with no traceability to individual hospitals or patients. We included patients undergoing surgical treatment for colon cancer, bladder cancer, myocardial infarction, or knee-osteoarthritis. The patient factors investigated were age, sex, comorbidity severity score, prior admission record, and socioeconomic status. We analysed possible associations with in-hospital mortality, intensive care unit admission, length of stay, readmission, reintervention and finally, in-hospital costs.

Findings

We found that routinely collected hospital data is a very promising source for prognostic factor identification and subsequent construction of accurate prognostic models. We found, for example, that male myocardial infarction patients are twice as likely to be readmitted to the hospital after discharge than female patients. In turn, patients’ prior admission record was found to be the strongest prognostic factor for in-hospital costs.

Conclusion

We conclude that it is possible to gain clinically relevant knowledge from using routinely collected hospital data for prognostic factor research, without any additional and expensive data-collection. We recommend clinicians and researchers to start using routinely collected data more efficiently for this purpose as this may support individualized prognosis and treatment and thereby further improvement of the value of care.

Nèwel Salet is a M.D. PhD candidate at Erasmus School of Health Policy & Management. Apart from prognostic factor research, his research focuses on alternative payment models and between-provider differences in outcome and costs. Given his medical background Nèwel has the goal of producing research findings that can aid clinical decision-making and improve healthcare delivery. You can contact him via email: salet@eshpm.eur.nl.

Compare @count study programme

  • @title

    • Duration: @duration
Compare study programmes