Patients are spending less time in hospital and are being managed by a greater number and diversity of health professionals. This means that better coordination of care is required. Donabedian described coordination of care as the "process by which the elements and relationships of medical care during any one sequence of care are fitted together in an overall design" [1]. This definition covers many aspects of coordination. Some of these relate to information technology, e.g. the development of the electronic medical record [2, 3] and others to work organization, i.e. to pre-specified programs and mutual adjustment or feed-back [4–6]. Programming sets responsibilities and activities for a known and predictable task, and involves standardizing the information or the skills required. It tends to work well for routine procedures but not for highly uncertain procedures [7]. When circumstances or events cannot be foreseen, there is a need for feedback mechanisms such as process supervision or peer interaction [8].

Practically, this coordination of the work organization involves, for instance, scheduling, communicating, and responding to unexpected situations. Although these functions are sometimes forgotten by health professionals, they can impact on quality of care. An example of a key function is the transmission of written information on actions taken [9]. Another is setting up a longitudinal relationship with a single identifiable provider with optimal cooperation as goal [10].

The question thus arises how to assess through measurement, functions such as making available useful written information and cooperation among healthcare providers. Thus far, most coordination indicators have focused on primary care and the hospital-ambulatory sector interface [11, 12] with coordination being linked to the notion of continuity of care as given by successive related sequences of medical care [13]. This article describes the development of three quality indicators (QIs) that measure either the availability of written information or staff cooperation in hospitals. The 3 QIs are the completeness and quality of the content of medical records, the completeness of the order for a radiology exam, and the holding of multidisciplinary team meetings (MDTM) on cancer patient management. The aim of the article is to describe the design of the QIs and discuss how they can be used to assess coordination within hospitals.

The study was part of the COMPAQH project (COordination for Measuring Performance and Assuring Quality in Hospitals). The project is managed by INSERM (the French national Institute for Health and Medical Research) and is sponsored by the Ministry of Health and the Haute Autorité de Santé (HAS - National Authority for Health). Its objective is to develop QIs in order to monitor quality in French hospitals and to design ranking methods and pay-for-quality programs.

We have developed 3 acceptable QIs covering two aspects of coordination (transfer of written information and adapting to the needs of others). But before concluding that they measure differences in coordination quality among hospitals, we need to consider several points.

First, the relationship between QI score and coordination quality may be subject to bias and erroneous interpretation. Selection bias may arise from incorrect sampling of medical records. To minimize such bias, we standardized the sampling method, gave confidence intervals to take sampling variability into account, used a standard grid, and checked for consistency during data collection. In the case of written information, differences in QI score between hospitals could have been due to different documentation processes rather than to genuine lack of information and/or cooperation between healthcare providers. However, whatever the cause of poor written information may be, it results in lack of coordination. In the case of MDTM, what was recorded may not have matched what actually happened. We can nevertheless reasonably assume that "false positives" (something recorded but not done) are far less frequent than "false negatives" (something done but not recorded) even if this has not been definitively established.

Second, coordination depends on many qualitative factors such as trust among staff members, experience in working together, the resilience of individuals and of the system, and the level of uncertainty encountered [20]. How meaningful are quantitative scores in such a context. Information can be transferred in ways that do not use written material as in the medical record or radiology exam order. Examples of information transfer are morning reports, verbal handovers, and informal conversations. All of these may be of high quality, but it is the written material that is somehow considered to provide the highest guarantee of coordination for understanding and sharing the meaning of previous actions.

We mentioned in the introduction part that coordination stresses the need to include both a standardized approach - programming - and a personal approach - feedback. Quantitative assessments like our set of 3 QIs seem to more cover aspects of standardization. Medical record content and the order for a radiology exam are standardized in order to create a common data core useful to all. We can say that feedback on cancer treatment plans is given to each person attending the MDTM. However, in this last case, true coordination really requires iterative feedback among health professionals as each decision is singular and its outcome difficult to predict. Iterative feedback should thus be captured by quality indicators and also by qualitative assessments.

The strength of the standards or guidelines underpinning QIs also needs to be considered. Guidelines are supported by an evidence base and, despite the wealth of literature, guidelines are not as common for evidence-based management as for evidence-based medicine. In the absence of strong evidence, QI use may not guarantee better coordination. We consider that there is a need to develop guidelines by applying conceptual frameworks to coordination, whilst at the same time complying with legal requirements for records and MDTMs.

A third point to be considered is that our 3 QIs explore aspects of coordination but not overall coordination within hospitals. It would be nice to know whether each hospital had similar results for all three QIs but, unfortunately, each QI was assessed in a different sample of hospitals. Nevertheless, wanting to tackle failures is already a step towards coming to grips with quality issues [21]. A QI can be used in a learning process to help understand the causes of failure, set improvement goals, and see whether the changes made have taught something [22, 23]. To institute such a learning process based on results for QIs requires acceptance by health professionals. A QI may be rejected because it generates conflict. For instance, the order for a radiology exam is written by one set of professionals - clinicians - but evaluated by another set - radiologists. Such QIs may not be universally acceptable if there is a gap between their practical and theoretical use [24]. This is one reason why we worked in close collaboration with the healthcare professionals involved.

A final important point is that we need more evidence to be able to relate the quality of work coordination to outcome of care in hospitals [25, 26].

Our QIs are a first step towards measuring work coordination and resolving management failures. The QI for medical record content was included in 2008 in the national accreditation procedure for healthcare organizations (HCOs) run by HAS. Results for over 1300 HCOs reveal high variability among hospitals. The QIs for radiology orders and MDTMs are currently being assessed for inclusion in the accreditation procedure. Further research in management and social sciences is needed to ensure that our QIs are able (i) to capture a sufficient number of aspects of coordination, (ii) encourage a more in-depth analysis of work coordination by a wide range of health professionals, (iii) and help explain variability in clinical outcomes.