In recent decades, measuring quality of care in hospitals has become a major challenge for many countries. Indeed, measuring is crucial for assessing and improving internal quality, informing health policies and justifying patients’ choices. It is also requested by payers for performance assessment and value-based purchasing [
1]. Three types of measures are commonly used to assess the quality of care in hospitals: structural, process and outcome indicators [
2‐
5]. Structural measures relate to the characteristics of the health-care setting [
3] such as the number of units and equipment, number and qualifications of medical and nursing staff, etc. Assessing an association between structural measures and both outcomes and process indicators may sometimes be very tricky. Indeed, increasing health resources does not necessarily lead either to a reduction in mortality or an improvement in processes [
5]. Process indicators (PIs) aim to assess the quality of clinical processes and answer the following question: do patients receive the best care possible according to current knowledge? This implies that achieving the best care processes leads to better health [
6]. Thus, PIs have to be strongly associated with related health outcomes to be used for quality assessment. Apart from this requirement, PIs present many advantages and are commonly used in quality improvement, public reporting (e.g. Hospital Inpatient Quality Reporting Program in the United States or the Canadian Institute for Health Information’s hospital performance program), or pay-for-performance (e.g. Medicare Hospital Quality Alliance Program [
7]) programs [
8]. Furthermore, they may be used for hospital accreditation [
9,
10]. Outcome measures of hospital care quality generally refer to patient health status as a result of health-care processes or to patient experience with hospital care. Among outcome indicators, mortality rates are the most widespread measures. They answer the question about inpatients’ survival or death during a fixed or variable period of time. Indeed, mortality is typically the type of information that the public and patients are interested in [
3,
5]. It is easily measurable, understandable to everyone, and supposedly cheaper to produce than other types of indicators since it is regularly collected in countries with health information systems. Moreover, it is frequently used for comparing performance between hospitals [
11]. However, like other outcome indicators, mortality rates depend on various factors including patient case-mix [
12] (i.e. patient characteristics, comorbidity and severity at admission) and data accuracy [
13], which could be confounding factors for measuring quality of care. Hence, they have to be accurately measured and adjusted on case-mix before comparing mortality across providers or being used for hospital profiling. Owing to the above mentioned limitations, they act as signals or flags to identify structures where further investigations have to be conducted [
13].
Many studies aiming to assess the relationship between process and mortality indicators have been conducted mainly on US and UK data [
7,
11,
14‐
17]. Various statistical methods have been used to assess this relationship ranging from simple correlations to hierarchical models. The underlying hypothesis is that hospitals with high mortality rates are very likely to have poor PI results [
8]. However, some studies have failed to show this association. A systematic review published in 2007 [
17] which included 36 studies examining 51 relationships between PIs and risk-adjusted mortality found a positive correlation in only half of the relationships (51%). There was no association between the two types of indicators in 31% of the relationships and paradoxical associations in 18%. The authors concluded that there is neither consistency nor reliability to assert that high risk-adjusted mortality is related to poor quality of care in hospitals. In this context, providing new evidence based on French data would be of significant interest [
18‐
20]. Moreover, our study should fuel the quality-of-care debate in France and help inform French policy-makers’ decision regarding the inclusion of quality indicators into quality improvement, public reporting, pay-for-performance and accreditation programs.
The objective of our study was to explore whether optimal care delivery in French hospitals as assessed by their PIs was associated with low HSMRs measured for different conditions and timeframes.