Underestimation of Disability Occurrence in Epidemiological Studies of Older People: Is Research on Disability Still Alive?
Year: 2002
Type: article
Abstract: The high prevalence of disability in the rapidly growing, oldest segment of the population is one of the most critical public health problems the healthcare systems of developed countries are facing. Disability in older persons is associated with increased use of healthcare services and need for formal and informal care.1,2 On a personal level, maintenance of good physical function and autonomy in activities of daily living are extraordinarily important for high quality of life.3,4 Measures of disability are now included in national health surveys in many countries. The development and implementation of public health policies aimed at preventing and reducing disability in the older segment of the population would be greatly improved if highly reliable and valid national and regional data on disability were available. Unfortunately, as has been pointed out in the recent literature, the lack of a clear-cut definition of disability and a number of methodological issues limit efforts to define precisely the magnitude of this problem.5 Each survey defines disability somewhat differently. There are also differences in methods used to address difficult methodological issues, such as whether to consider individuals nondisabled if they say they could perform a task or only if they actually do perform the task and how to address individuals who cannot perform a specific task, such as meal preparation, because they never performed that task. Clinical trials in older populations are just beginning to assess functional outcomes,6–8 and it is likely that, in the future, even drug trials focused on specific diseases of older persons will use disability measures as major outcomes because of their value in reflecting the overall health status of the individual. There are thus a wide variety of reasons why the accurate measurement of disability is an important goal. Irrespective of how precise our tools are, how we apply these tools affects what we learn from them. In the important paper in this issue by Gill et al.,9 we are reminded that a prevalence rate is nothing more than a simple snapshot at a single point in time and does not tell the whole story of the occurrence of a medical condition. A simple relationship taught to every epidemiology student is that prevalence equals incidence times duration, with the implication that a short-duration condition can have a low prevalence rate even if it has a high incidence. The classic example of this is the common cold, which has a high 1-year incidence but a low prevalence, because of its short duration. In their paper, Gill et al. do not imply that our prevalence estimates for disability are wrong but that they do not reflect the disability that occurs and lasts only a short time. They effectively address the reasons for short-duration disability, showing data on recovery, death, and loss to follow-up in the study participants who reported disability at one of their monthly interviews. In the past, we conceptualized disability as a progressive condition, assuming that, after the onset of disability, any change that takes place is downward. Many studies have considered disability as an absorbing state in their analyses.10,11 It has been methodologically much simpler in longitudinal epidemiological research on predictors of disability to consider the first report of disability to be a definitive outcome, in the same way that myocardial infarction or cancer occurrence are considered outcomes in more traditional epidemiological studies. Nevertheless, the approach that considers disability as an enduring condition has turned out to be wrong. It conflicts with clinical experience, and a variety of longitudinal studies that document the highly dynamic nature of disability, with transitions over time following every conceivable pattern, have proven it untrue.12–14 Thus, current estimates of disability mix together transient (short-term) and chronic (long-term) disability. There is growing interest in better understanding the association of specific diseases with disability. The findings of Gill et al. suggest that we may underestimate the importance of conditions that are associated with transient disability or that cause sudden severe disability followed by high rates of recovery. For example, in a study with annual follow-ups, reports of conditions such as back pain or even conditions as catastrophic as hip fracture may be weakly associated with disability at the follow-up interview because people with these conditions have recovered during the year. Alternatively, the contribution to disability of diseases that cause progressive reduction of physical function with little or no recovery, such as Parkinson's disease, is fully captured. It is conceivable that the risk factors and causes of transient disability are quite different from those involved in permanent disability and that the kind of underascertainment demonstrated in the Gill et al. study will mean that most research will not be effective in demonstrating this. Because certain diseases causing short-duration disability15 have different incidence rates in men and women, the magnitude of this systematic error is probably sex-related. In general, if we assume that episodes of temporary disability are distributed homogeneously over the year, then the size of underestimation is directly proportional to the length of the interval between successive follow-ups and inversely proportional to the length of the episode. Unfortunately, some conditions that cause transient disability, such as pneumonia, allergy, and even stroke,16 tend to occur more frequently at specific times of the year. Therefore, the assumption of homogeneous distribution is unlikely to be valid, and the true percentage of transient episodes that are missed depends on the season when the assessment is performed. As shown in the data reported by Gill et al., once the follow-up interval in a study is 1 year or longer, part of the underascertainment of disability is related to loss to death and loss to follow-up in addition to the effect of short-duration disability. A high proportion of persons have disability in the months before death, with the rates very high in people dying at advanced old age, so missing these occurrences of disability would tend to reduce the differences in rates of disability observed across age groups. Men are two to four times more likely than women to develop sudden death, often with no disability before death.17 This may introduce a further sex-related bias because the higher rates of disability before death in women, who have less sudden death, may not be fully appreciated because of underascertainment of short-duration disability before death. The observation that there is a substantial amount of short-duration disability and that a good portion of it is missed when we perform widely spaced surveys does not answer the question of how important this kind of disability is. What is clear from the work of Gill et al. is that there were many 1-month intervals of disability. Adding up all these short intervals could result in quite a substantial increase in person-years of disability that our usual way of assessing disability just does not capture. This additional time spent disabled will certainly have a negative effect on quality of life. It is also likely that people recovering from disability have demonstrated a vulnerability that is an indication of much higher risk of future adverse events and will continue to be at high risk for a prolonged period. We are unaware of research that has demonstrated that people who report disability one year but are not disabled the next year do worse in the long run than people who do not report disability at either assessment. Further research is necessary to understand fully the effect of short-duration disability. The extremely low attrition in Gill et al.'s study should be highlighted because it makes the findings of this study even more striking. Most other studies examining incidence and prevalence of disability report participation rates that are substantially lower. There is evidence that nonparticipation is a nonrandom phenomenon that is much higher in persons who are older, depressed, suffering from acute illnesses, affected by cognitive and physical impairments, and lacking adequate social support.18,19 These very factors that are related to loss to follow-up are also related to disability, but, because these individuals are lost in studies with widely spaced assessments, there will be a general underestimation of population disability and falsely low estimates of disability in subsets of the population with these characteristics. The results of the Gill et al. paper are relevant for clinical trials with disability outcomes, which are expected to become more common in the future. Differential ascertainment of disability (whether due to transient disability, differential loss to follow-up, or differential mortality) in the different arms of the trial can have a powerful affect on the results. Strategies aimed at increasing the identification of disability episodes will increase the validity of these trials. Despite the difficulties encountered in measuring disability, the answer to the question posed in the title of this editorial is an emphatic yes; research on disability in the older population is most definitely alive. The issues raised in the paper by Gill et al. do not challenge our disability measurement tools' reliability and validity. The work of Gill et al. and many others provide reliable and valid data. The problem lies in the discovery that the natural course of disability is such that its up-and-down trajectory means that, when we take a snapshot, no matter how accurate that snapshot is, we may be observing an individual after or even between episodes of disability. Several alternative solutions are possible. First, we can perform more frequent assessments. The differences between cumulative incidence (determined from monthly evaluations) and prevalence at 6 months were not large, ranging from about 4% to 12% in the three functional status groups studied. The differences get much larger when considering the 1-year interval, so assessments done every 6 months seem to have some advantage in capturing more of the disability experience of a cohort. Second, we must consider whether study participants can validly report short-duration disability that occurred during the interval being evaluated. Most epidemiological research on specific diseases assesses whether a discrete disease event has occurred at any time since the last contact with the participant. In conditions such as myocardial infarction, stroke, and hip fracture, the participant is asked whether the condition occurred, and medical records may be searched to confirm this occurrence. This is essentially a cumulative incidence of the condition and is aimed at identifying all occurrences during the follow-up interval. Questions used in national surveys that assess restricted activity days and bed disability days over the past year attempt to capture a measure of health status over an entire interval and have been found to be valid.20 Similar questions that evaluate physical disability have had little developmental work but offer the potential to address the deficiency in capturing disability occurrence demonstrated in Gill et al.'s report. Additionally, information on functional status before death can be included in an interview with a proxy done after death, such as in the National Mortality Follow-Back Survey.21 Eventually, a new nosology of disability could be developed that takes into account time course, recovery, severity, and modality of onset. The demand for data on disability in the older population has increased greatly in recent years. Despite the potential for misreporting and underestimating disability, national surveys and large cohort studies remain useful in assessing population trends and factors associated with disability. An improved national surveillance system or a national registry will be essential in providing better information on prevalence and incidence and will help determine the success of public health interventions aimed at preventing or delaying disability in the older population, but the quality of the information collected in these registries may profit from advances made in disability assessment. The authors wish to thank Chiara Cavazzini for her help in the preparation of this manuscript.
Authors: Jack M. Guralnik, Luigi Ferrucci
Language: English
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2094785783