• length of in-hospital stay
• quality of care

The relationship between length of in-hospital stay (LOS) and quality of care is difficult. LOS is determined by a complex interweaving network of multiple supply and demand factors which operate at macro-, meso-, and micro-levels. These factors range from organisational culture and hospital bed availability, through availability of “step down” or intermediate care services, to the customs and cultures of the local populace.^1–3 On top of these many factors there is also usually an underlying downward trend in LOS for any one particular condition over time.³

In health policy terms, LOS remains an easily measurable index of “efficiency” and is quoted as such in one of the most recent publications of the UK Department of Health NHS performance indicators.⁴ In this publication the percentage “improvement” or percentage reduction in LOS compared with the previous year is plotted for each local area. The clear message from the UK Department of Health is that reductions in LOS are expected to be achieved year on year and represent “efficiency” of local health services.

Much of the literature in this area would support or certainly not refute this policy drive by the UK Department of Health. Many of the studies that have been undertaken show that quality of care or health outcomes do not appear to be compromised by reductions in LOS,^5–7 and for a long time there have been suggestions that LOS could itself be a cause of increased morbidity resulting, for example, from increased risks of hospital acquired infection or thromboembolic disease.^8,9

In contrast to this view, Kossovsky and colleagues have produced some interesting findings in their study of the relationship between LOS and quality of care in congestive heart failure.¹⁰ In their paper published in this issue of QSHC they looked at the relationship between LOS and three validated indices of quality of care—an admission score, a treatment score, and a discharge score. Within each index there were a number of items which were obviously directly related to quality—for example, taking an adequate history was an item in the admission score, daily weight measurements were included in the treatment score, and improvements in clinical signs in the discharge score. In one institution (their own) the authors found a statistically significant association between longer LOS and treatment and discharge scores, having adjusted for relevant confounding factors such as age, comorbidity and severity.

This finding of an association does not, however, tell us about the causal nature of the relationship found. Bradford-Hill originally described the basic criteria for assessing causality when an association has been found.¹¹ One of the most important is the strength of any relationship, and consistency describes the repeatability of the finding. A dose-response relationship or biological gradient indicates how the dependent variable—in this case, quality—varies in line with the independent variable—LOS (“longer LOS, more quality” or, conversely, “shorter LOS, less quality”). Coherence, biological plausibility, and analogy all relate to whether a plausible mechanism for a causal chain of events is possible. However, the sine qua non for causality is temporality. The cause must precede the effect.¹²

How can these criteria be used in assessing whether a relationship found between quality and LOS is causal? The strength of the relationship is not at all clear. Studies have been published which suggest an increase in quality with both a shorter and a longer LOS, and this finding does not easily comply with the consistency or biological gradient criteria. Plausible reasons for the relationship between LOS and quality of care can be put forward to support either a longer or shorter LOS—for example, a longer LOS might be thought of as allowing more time for appropriate investigation and treatment while a shorter LOS may be consistent with a rapid, ordered and systematic care pathway. The criteria of coherence, biological plausibility, and analogy are therefore not particularly useful in this context. The criterion of temporality is one of the most important, however, for considering the causal relationship between LOS and the quality of care. In the study by Kossovsky et al¹⁰ it is possible that the patients with a longer LOS had increased time available. This increased time would not only allow time for the investigations included in the treatment score such as an echocardiogram to be undertaken, but it might also allow for the patients' health to start to improve over time thus improving the discharge score as well. In this case the effects (improvement in indices of quality) may not precede the cause (LOS). It is possible that the indices of quality may not be independent of LOS.

“The problematic nature of the relationship between LOS and quality needs to be acknowledged”

Where does this leave us? It appears that a longer LOS does not (and cannot) “cause” an increase in quality on its own. Both very good and very poor quality of care can be provided with the same LOS for the same condition. This problematic nature of the relationship between LOS and quality needs to be acknowledged. LOS is most likely to have an inverted “U” shaped relationship to quality of care. Above and below a certain optimum LOS, quality may deteriorate. The optimum LOS for any one condition will have a range which depends on local supply and demand factors such as the individual patient's needs or the availability of the relevant community services.

Current patterns in health care—including the increasing role of intermediate, primary, and community care in many industrialised countries—point to a decreasing role for the hospital. We need to move away from an obsession with LOS. The “right” care needs to be provided in the “right” place.¹³ This study by Kossovsky et al is a useful addition to the literature on LOS and quality.¹⁰ I would concur with the authors' conclusions that LOS should not be reduced without consideration of care pathways and appropriate treatment patterns, but I would go further—ensuring the delivery of appropriate care pathways and treatment patterns is crucial for quality of care; LOS itself may be irrelevant to this process.