Chapter 11 Health Services Research in Radiation Oncology
What Is Health Services Research?
What is the Scope of Health Services Research?
Figure 11-1 shows a general framework for a program of HSR aimed at improving some specific aspect of health system performance. The first step is to select, define, and validate appropriate indicators of the aspect of performance that has been targeted for investigation. The next two steps are to develop methods of measuring system performance in terms of the chosen indicator(s) and to prescribe standards or targets for system performance in terms of the chosen indicators. These two steps, which often involve the use of very different methods, can sometimes be undertaken in parallel. Once standards have been set and methods for measuring performance have been established and validated, it becomes possible to evaluate the performance of the system against the standards. This, in turn, permits further explanatory studies aimed at identifying factors that are associated with better or worse performance. This information can be used to design interventions aimed at improving performance. The interventions may then be implemented and systematically evaluated. Interventions may be refined through further cycles of improvement, before they are suitable for dissemination and incorporation into routine practice.
Studies on the Accessibility of Radiotherapy
Concept of Health Care Accessibility
The term accessibility was originally used narrowly to describe the ability of patients to obtain entry into the health system.1 It is now used more broadly to represent the overall “degree of fit between the clients and the system.”2 Accessibility can be seen as having a number of dimensions that determine that overall degree of fit (Table 11-1). Availability describes the total volume of the service available in relation to the total number of clients that would benefit from it. Availability depends on the adequacy of supply of health care workers and on the adequacy of facilities and equipment. For any given level of resources, availability also depends on the degree of efficiency in production of services. Spatial accessibility describes the geographic relationships between the places where services are provided and the places where potential clients reside. Accommodation describes the extent to which the system is designed and operated to facilitate clients’ access to service, for example, by operating at convenient hours or by providing transportation for patients who may need it. Affordability describes the relationship between the cost of health services and clients’ ability and willingness to pay. It depends not only on the direct cost of services but also on indirect costs, for example, loss of earnings during a protracted course of treatment. Awareness describes the extent to which those who need the service know that it is available and that they might benefit from it. In the context of a specialized service such as RT, patients’ awareness of the potential benefits of RT depends largely on their attending physician’s awareness of the indications for RT.
| Availability |
Need for Studies on Access to Radiotherapy
There are compelling reasons for doing research aimed at optimizing the accessibility of RT. To achieve optimal cancer outcomes at the population level it is necessary to make effective treatments accessible to every patient who needs them. RT is known to be effective in many clinical situations and the World Health Organization recognizes RT as a key component of any overall program of cancer control. In its 2005 declaration on Cancer Control, the World Health Organization stated that “recognizing that the technology for treatment of cancer is mature and that many cases of cancer can be cured…., all nations should improve access to appropriate technologies.”3 Many nations aspire to providing adequate and equitable access to health care for all of their citizens, but there is remarkably little information available about how successful they are in achieving this laudable goal with respect to RT. In reality, the widespread reports of waiting lists for RT in the medical literature and the news media suggest that access to RT remains suboptimal in many parts of the world.
Waiting Lists for Radiotherapy
Long waiting times for RT were first identified as a cause for concern in the medical literature in a report from Norway more than 20 years ago.4 Waiting lists for RT have since been reported in many other countries, including Australia,5 the United Kingdom,6 New Zealand,7 Canada,8 Denmark,9 Germany,10 Spain,11 and Italy.12 In the countries affected by waiting lists they have been a major concern for both patients and the providers of RT. The problem of waiting lists for RT is an ongoing challenge for health services researchers in radiation oncology, but the first step in dealing with the problem is to learn how to measure waiting times for RT.
Measuring Waiting Times for Radiotherapy
Mail surveys and email surveys can provide a lot of information about waiting times from multiple institutions at modest expense. They also can be used to compare waiting times between different centers within one country or can compare waiting times between different countries. Surveys can be structured to elicit information about delays in RT in a number of specific clinical situations. In the 1990s, a survey of heads of radiation oncology at comprehensive cancer centers in the United States and Canada showed that waiting lists for RT were widespread throughout Canada but revealed no evidence of similar problems anywhere in the United States. Median waiting times for a range of indications for RT were two to three times longer in Canada than in the United States.13 Figure 11-2 shows, for example, that at almost every Canadian center, patients with laryngeal cancer waited longer for RT than they did at almost any U.S. center. However, the validity of such surveys may be questioned because they rely on the veracity of self reports and because the primary information on which each report is based may differ from center to center.
Retrospective analysis of data that have been gathered for other purposes is also relatively inexpensive and can provide more objective information about waiting times for RT. This may be an important first step in addressing this type of problem. At the beginning of the 1990s, reports of long waiting lists for RT in Ontario were frequently in Canadian news media. Health system managers believed that these reports were unduly alarmist and at first denied that there was any systemic problem.7 To clarify the situation, an analysis of waiting times for RT was undertaken based on computerized electronic records of all visits to the province’s radiotherapy centers over the preceding decade. Once these administrative records had been linked to the province’s cancer registry, waiting times for RT for various specific conditions were able to be described.7 For example, Figure 11-3A shows that waiting times from diagnosis to start of radical RT for laryngeal cancer increased dramatically through the late 1980s and early 1990s. Similar large increases in waiting times were found in many other clinical situations. The administrative data archived at the cancer centers also allowed us also to calculate waiting times between various milestones along the pathway from diagnosis to start of treatment. Figure 11-3A shows that the observed increases in overall waiting time between diagnosis and treatment were entirely the result of increases in the waiting time between the first visit to a radiation oncologist and the start of RT. There was no increase in the interval between diagnosis and referral to radiation oncology or between referral and consultation. These findings pointed to rate-limiting problems in access to planning and/or treatment machines. It is useful, whenever possible, to report observed waiting times in relation to standards or guidelines. At the time of this first report the Canadian Association of Radiation Oncologists (CARO) had already set standards for acceptable waiting times for RT: the maximum acceptable delay between referral to, and consultation by, a radiation oncologist was deemed to be 2 weeks, and the maximum acceptable delay between consultation and the start of RT was deemed to be 2 weeks.7 Although these standards were based only on expert opinion, they provided a useful framework for comparison. Figure 11-3B shows trends in compliance with these standards over time. Most patients met the CARO standard for prompt consultation throughout the study period, but the proportion of patients meeting the CARO standard for prompt start of RT fell from 90% to 10%. This simple study, which merely quantified the magnitude of the problem in our community, was useful because it led to public recognition of the seriousness of the problem.7 This proved to be an important first step in promoting the reinvestment in the infrastructure of the provincial RT system.
Causes of Waiting Lists for Radiotherapy
Kinetics of Waiting Lists
When demand for RT exceeds supply, waiting times inevitably increase and a waiting list for RT starts to grow. In theory, the waiting list will then continue to grow for as long as demand continues to exceed supply. In reality, waiting lists for RT do not grow indefinitely. When waiting times for RT become longer than the referring physicians believe is acceptable, they may begin to offer their patients alternative treatments, in circumstances in which RT would normally have been their first choice. For example, when long waiting lists for RT developed in Ontario in the early 1990s, there was a significant decline in the use of primary RT in the management of head and neck cancer, followed by a rebound when waiting lists decreased after a major reinvestment in facilities.14,15 It has been shown that there is a significant negative association between the prevailing waiting time for RT and the proportion of patients receiving postoperative RT after a partial mastectomy for breast cancer.16 Furthermore, tumor progression or deterioration in a patient’s general condition during the delay may render the patient ineligible for RT that would initially have been appropriate, and these cases drop off the list. Decreasing referrals and increasing dropoffs from the waiting list serve to reduce demand for RT. As demand declines, the balance between supply and demand is eventually restored; the waiting list then ceases to grow, waiting times stabilize at a higher level, and RT utilization rates stabilize at a lower level. This phenomenon has been referred to as implicit rationing, because it limits utilization without explicitly limiting access to care.17
Even when average supply is equal to average demand for RT, random fluctuations in referral rates may produce transient peaks in demand that exceed supply, and this may be sufficient to cause a substantial waiting list.18 This risk can be reduced by forward planning that provides a buffer of reserve capacity or by building flexibility of capacity into the system. The smaller the functional unit, the greater is the impact of random fluctuations, and the more reserve capacity is required to avoid a waiting list.18
A Canadian Case Study
Why did waiting lists for RT become such a widespread problem around the world in the 1990s? Was there an increase in demand or a decrease in supply or both? Ontario’s experience serves as a useful case study. Analysis of historical data showed that three different factors conspired to cause a huge increase in demand for RT over the critical period.15 First, the incidence of cancer increased inexorably by approximately 3% per year, owing primarily to the aging of the Ontario population.15 Second, there was a dramatic increase in the proportion of patients referred for RT in three very common cancers. The proportion of new breast cancer cases referred for RT increased, consistent with the evidence-based trend toward breast conservation surgery and postoperative RT.15 The proportion of new cases of rectal cancer referred for RT increased, consistent with the evidence-based adoption of postoperative RT and chemotherapy. The proportion of prostate cancer patients treated with RT increased owing to an increase in the proportion of early-stage cases detected after the widespread adoption of screening for prostate-specific antigen.15 Third, there was a significant increase in the average number of fractions prescribed per course of RT. This was driven by an increase in the number of fractions per curative or adjuvant course, which outweighed a concomitant but smaller decrease in the number of fractions per palliative course of RT.15 There was no decrease in treatment capacity. In fact, the number of treatment machines in the province increased faster than the incidence of cancer.15
Consequences of Waiting Lists for Radiotherapy
Delays in starting RT are a source of great concern both to the patients and to those involved in their care. The potential adverse effects of a waiting list for RT are summarized in Table 11-2. Delays have both direct and indirect effects on the well-being of patients, and waiting lists also have broader economic and social consequences. It is useful to classify the direct effects of delay on the well-being of individual patients as nonstochastic or stochastic.19 These terms are used here as they have been used in the field of radiation protection, where they provide a useful distinction between the effects of radiation that depend on chance and those that do not. The nonstochastic effects of delay include the psychological distress caused by the delay and the physical symptoms resulting from the untreated cancer. They occur in most cases and often increase in intensity with time, although they may not occur at all until some initial threshold period has been exceeded. The stochastic effects of treatment delay include the development of metastases and failure to achieve local control with radiation. These are all-or-nothing phenomena. Their probability increases as a function of time, but their severity is independent of time, and there is no lower limit of waiting time below which they will not occur. Waiting lists may also have indirect adverse effects on patient care, mediated by changes in medical practice. In addition to their effects on health outcomes, waiting lists have important economic and societal implications.19
TABLE 11-2 Effects of Waiting Lists for Radiotherapy on Well-Being of Patients
| Direct Effects |
RT, radiotherapy.
Measuring the Direct Effects of Delays in Radiotherapy
The risk of carcinogenesis has been estimated by mathematical modeling using the best available clinical data combined with principles derived from laboratory studies. Given the very scant direct evidence that was available about the impact of delay on outcomes in the early 1990s, a similar approach was initially used to estimate the risks of treatment delay.19 Figure 11-4 shows the predicted effects of delay in RT on local control in cancer of the oropharynx, derived from a mathematical model.19 The model was based on radiobiologic principles that had been validated in experimental systems, and it incorporated the best available clinical information about tumor doubling times and the relationship between tumor volume and local control.19 The model predicted a decrease in local control rates of between 5% and 10% per month of delay in the start of RT. Others have since made similar predictions.20,21 Although this approach was credible to radiation oncologists, it did not prove to be any more persuasive than expert opinion from the perspective of health system managers in our community. When decisions about the allocation of scarce resources are at stake, there is no substitute for direct clinical evidence.
Measuring the magnitude of the stochastic effects of treatment delay is not straightforward. It is inherently difficult to measure the risk of treatment failure resulting from delay, because local failures caused by delay are absolutely indistinguishable from treatment failures resulting from other causes. The problem is analogous to that of defining the risk of carcinogenesis associated with low-dose radiation. One cannot simply count the cancers caused by radiation because they are usually indistinguishable from the many other cancers that may occur in the population as the result of causes other than radiation. Rates of failure must, therefore, be compared in groups of patients who have been exposed to longer and shorter delays, and the challenge is to ensure that those groups are comparable with respect to all other relevant prognostic factors. A randomized trial would be the best way of creating truly comparable groups, but it would be unethical to randomize patients to timely RT versus delayed RT because there is no conceivable benefit in delay. Comparisons of the outcomes of RT in nonrandomized groups of patients who have waited longer or shorter periods of time are subject to all the biases that may affect any retrospective observational study. However, in this context, such studies are very important because they represent the best available direct source of information.22,23
Recent systematic reviews have identified a growing number of observational studies that have investigated the association between treatment delay and the outcome of radiotherapy in certain clinical situations.24 Figure 11-5 summarizes the results of the 20 high-quality studies included in the most recent published meta-analysis.24 Most of these studies had been done in the context of head and neck cancer and breast cancer. In these two disease groups, meta-analysis has shown a significant increase in the risk of local recurrence in patients who waited longer for RT.24 A very large population-based outcomes study, not included in this meta-analysis, recently confirmed that delay was associated with a higher risk of local failure after postlumpectomy RT for breast cancer.25 No evidence of a threshold below which delay was free of risk was found. Moreover, although there was less evidence of an association between delay and local control in sites other than breast and head and neck, there were insufficient data available to conclude that delay in RT is free of this risk in any situation.
No significant association was found between delay in RT and the risk of distant metastasis, although there was less information available about this outcome.24 There was a small but significant decrease in survival with increasing delay in head and neck cancer.24
The relative risk of local recurrence of 1.1 per month of delay in starting postoperative RT for breast cancer translates into an absolute increase in recurrence rate of about 1% per month of delay in a population with a baseline rate of local recurrence of 10%.24 Although this is a small risk for any individual, it has the potential to cause an important increase in the number of recurrences at the population level. The increase in the relative risk of recurrence by 1.15 per month of delay in patients undergoing definitive RT for head and neck cancer translates into an absolute increase in risk of recurrence of 3% in a population with a baseline risk of failure of 20% or an absolute increase of 6% in a population with a baseline risk of failure of 40%.24 Interestingly, these findings are consistent with those predicted by the mathematical models described earlier.19,21 Thus a few weeks of delay in RT may have an adverse effect on outcome that is sufficient to cancel out all the improvements in outcome achieved by advances in the practice of RT over the past 20 years.21,22 Given that there is no theoretical reason to think that there is a threshold below which delay is safe, it would be prudent to adopt the principle that delays in RT should be as short as reasonably achievable (ASARA), modeled on the as low as reasonable achievable (ALARA) principle, which guides risk management in the field of radiation protection.22
Indirect Effects of Waiting Lists for Radiotherapy on Patient Care
The indirect effects of waiting lists on patient care and population health are summarized in Table 11-2. The phenomenon of implicit rationing by which waiting lists reduce the use of RT has already been described (see Kinetics of Waiting Lists). Waiting lists may also increase the use of alternative treatments that may be less effective, more morbid, and more expensive than RT. There is evidence that long waiting lists may cause radiation oncologists to modify the way they prescribe RT. A study from the Queensland Radium Institute showed a significant negative correlation between waiting times and the number of fractions prescribed per course.26 This was due primarily to decreases in palliative fractionation as waiting times increased.20 A similar association between prevailing waiting time and the choice of fractionation for bone metastases was found in Ontario.27 There are obviously serious risks in deviating from accepted practice in radiation oncology for the sole purpose of getting more patients treated. However, in circumstances in which randomized trials have demonstrated that shorter courses of RT are equivalent to longer courses of treatment, adoption of the more parsimonious approach has the potential to reduce overall workload and greatly increase the availability of RT, without adversely affecting outcomes.28 The challenge is to ensure that shorter-than-standard courses of RT are used only in circumstances in which they have been shown to be medically appropriate. Under conditions of scarcity of resources, it is particularly important to have explicit standards of care to prevent deterioration of quality in the attempt to maintain accessibility.
Societal Effects of Waiting Lists for Radiotherapy
Waiting lists for RT are potentially costly (see Table 11-2). Patients require both care and counseling during delays, and the costs of alternative treatments may be considerably higher than those of RT. Waiting lists have sometimes caused patients to be referred to distant centers for RT with loss of continuity of care and support for patients and added costs to the health system. The inability to provide timely RT may also be frustrating and distressing for the staff of RT programs. Waiting lists also expose RT providers to legal liability. In Quebec, a class action suit was launched against the hospitals responsible for providing RT, on behalf of several thousand women who had to wait long periods for adjuvant RT after surgery for breast cancer. The judge accepted the evidence that delay was associated with an increase in the risk of local failure, and the case was ultimately settled with financial compensation for approximately 10,000 women who had waited for longer than 12 weeks to begin postoperative RT after lumpectomy.29 Chronic waiting lists for RT and other important medical services eventually became an important political issue. Waiting lists are often used as evidence of the need for change, both by advocates of privatization of the health system and by those who favor reinvestment in the public system. By the early 2000s, public opinion polls showed that “wait times” for medical care had become the greatest concern of most Canadian voters and there were increasing demands that government set waiting time standards.
Measuring Access to Radiotherapy
Limitations of Waiting Times as an Indicator of Access to Radiotherapy
The existence of a long waiting list for RT is a symptom of inadequate access to RT. The duration of the waiting time for RT may be directly related to the probability of an adverse outcome, and waiting times may therefore serve as a quantitative measure of the quality of care. The length of a waiting list, however, provides no information about the magnitude of the shortfall between supply and demand, and waiting times cannot, therefore, serve as a quantitative measure of accessibility. The absence of a waiting list does not mean that access is optimal. Waiting lists develop only in response to supply-side problems in the availability of services. Waiting times are entirely insensitive to demand-side problems with respect to spatial accessibility, accommodation, affordability, or awareness (see Table 11-1). Problems in those dimensions of accessibility, in fact, reduce demand and may serve to reduce or avoid waiting lists. To ensure appropriate access to RT, it is necessary to monitor RT utilization rates as well as waiting times.
Defining Accessibility of Radiotherapy
The best quantitative measure of the accessibility of any service is the rate of its appropriate utilization, that is, the proportion of patients who need a service who actually receive it. The term need is used here as defined by Cuyler,30 who states that “the need for medical care exists when an individual has an illness for which there is effective and acceptable treatment.”
Measuring Accessibility of Radiotherapy
The incidence of cancer (i.e., number of new cases diagnosed in the population of interest, over the period of interest) may be used as the denominator for describing the rate use of RT in the initial management of the disease. The best way to establish the proportion of cases that are treated with RT is to follow all of them forward in time from the date of diagnosis and find out if and when the patient received RT. The approach was first used in the Netherlands31 and subsequently was used to describe the use of RT in Ontario.14 The estimated rate of use of RT in the initial management of cancer depends on the cutoff point in time used to define initial RT. If a short cutoff point is chosen to define initial RT (e.g., RT within 3 months of diagnosis), the indicator will miss some patients who receive adjuvant RT after surgery. If a longer cutoff point is chosen (e.g., RT within 1 year), the indicator will include almost all patients who receive RT as part of their initial management, but it will also wrongly include some patients who are actually receiving RT for an early recurrence after primary surgery. The best cutoff point depends on the specific disease under consideration. For practical purposes, the proportion of incident cases treated within 1 year of diagnosis has been used to describe the initial use of RT in the general cancer population.14 Figure 11-6A describes variations in the use of RT in the initial management of cancer in Ontario in terms of this indicator (R1 year).
The incidence of cancer is a less suitable denominator for describing the utilization of palliative RT, because a high proportion of incident cases will never develop indications for palliative RT and many of those who ultimately do need palliative RT will not require it until years after the diagnosis. It is preferable to describe the use of palliative RT among patients who die of their cancer. This can be accomplished by identifying patients who died of their disease in a population-based cancer registry and following them back in time to identify those who received RT within a defined interval before death.32 Figure 11-6B describes variations in the use of palliative RT in the past 2 years of life among patients who died of their disease in Ontario. The same approach lends itself well to the description of the rates of use of other types of care in the terminal phase of the illness.
What Factors Affect the Rate of Radiotherapy in the General Cancer Population?
Figure 11-6 illustrates geographic variations in the rate of use of RT in Ontario, with the highest rates being observed in the counties where RT facilities are located.14
