If the title of this article had been something about the "epidemiology of stuttering", several readers would have been injured in the stampede to the escape key, and others would have said, "Ho-hum! Pardon me while I do my tatting". And that’s too bad.
Epidemiology, as a major science of public health, can and must become a complement to clinical and laboratory science as stuttering is studied in populations and in culture contexts (Lubker, 1997; Lubker & Tomblin, 1998). I became interested in epidemiology many years ago as a speech-language pathologist who served the "epidemic" of children with disabilities following the infectious disease epidemic of rubella. I did not know that my interest in population distributions of health events was called epidemiology. Many years later, I earned a doctorate in public health epidemiology. I have found it to be a fascinating science that approaches questions in ways that can, indeed, help to answer clinical questions.
What is epidemiology? How have tenacious theoretical biases about stuttering distorted epidemiological data? Why do we care about measures such as incidence and prevalence, perhaps the best known and yet, most frequently misapplied epidemiological concepts? How can these measures inform practice and professional preparation? The following paragraphs address a few of these questions.
It is notable that a couple of epidemiological facts about stuttering are so well established that they need no citations from the literatures. First, for example, the adult stuttering population is comprised of more males than females. And second, the age of stuttering onset is overwhelmingly among children during the preschool years. Why are these epidemiology facts in contrast to clinical facts? They are facts about the "population" classified as stutterers. Epidemiologists regard "the population" as the unit of study. This is sometimes a difficult conceptual transition for clinicians to make, because, clearly, populations are comprised of individuals. Let’s borrow an example from preterm babies. Clinicians often emphasize that many preterm babies turn out just fine and that not all babies born preterm have disabilities. A recent summary estimate from several sources is that as many as 60% of preterm infants ultimately have identifiable language disabilities (Goberman & Robb, 1999). The epidemiologist would take this a step further to ask also what proportion of babies with full-term gestation has language disabilities. Does every preterm child have a disability? Is every full-term baby disorder free? Obviously not, but, because the occurrence of disability is so much higher in the preterm group the expectation is that disability becomes a characteristic of the population, when, and this is important, the preterm population is compared with the full-term group. Then, from comparison of the two populations, we can quantify risk and make risk statements about the characteristic of gestational age (Lubker, 1997).
From epidemiological perspectives about stuttering, we can ask, "Is every adult person who stutters a male?" Of course not, but maleness is a characteristic of the population. Such differentiations are essential considerations in formulating questions about remission rates which are, in turn, essential to determining prevalence rates. Several investigators have studied the epidemiology of stuttering ( A few examples: Andrews, Harris, Garside & Kay, 1964; Gillespie & Cooper, 1973; Bloodstein, 1975; Andrews et al, 1983; Yairi & Ambrose, 1992; Ambrose & Yairi, 1999). Perusal of major speech-language journals with an "epidemiological eye" reveals that many publications contain more or less well-applied epidemiology concepts that are not identified as such. Only a few investigators seem to have used the rich, population-based accepted methodologies of epidemiology to inform theory and practice for stuttering.
What, then, is epidemiology? While incidence and prevalence rates are perhaps the most familiar epidemiological concepts, epidemiology is a great deal more than headcounting. It is actually a basic science essential to studies of cause; it is a science with research questions and answers based on populations comparisons. Epidemiology is defined in many basic texts (See, for example, Greenberg et al, 1996; Gordis, 1996).
Lubker and Tomblin (1998) defined epidemiology this way. "Epidemiology is the study of the distributions and determinants of diseases, disorders and disabilities and of desirable health events in populations" (p. 3 ). Gordis (1996; p.3) has defined epidemiology as, "...the study of the distribution and determinants of health related states or events in specified populations and the application of this study to control of health problems".
Application of the Lubker and Tomblin definition to stuttering reveals some interesting concepts.
Distributions: Who stutters? What sex, race and socioeconomic data identify stuttering distributions? (Gender is a social construct; sex is a biological designation) When? This does not mean "When I talk on the telephone". The question is whether stuttering in the population decreases or increases during identified time periods. Do birthdates of people who stutter tend to cluster in ways that change over time? (Birthdates of populations with congenital deafness have changed notably over time. Can readers guess why?) This is not a new concept (Bloodstein, 1975), but it is one to which epidemiological expertise in measuring populations outside treatment settings has rarely been applied. Where? Again, this does not mean, "I stutter in the drugstore." Do population clusters of people who stutter exist in geographic areas? There is tantalizing international data indicating that, in fact, such clusters do exist. Do the population clusters follow other known distributional differences in the northern and southern hemispheres? We’ll save this information for another time.
Determinants: Why? How? It is in variable groups of why and how that epidemiologists pose questions about that perennial favorite of discussion on stuttering: Causality.
Diseases, disorders and disabilities: What? What is it that we are studying? Epidemiologists insist that the topic of investigation be clearly defined. Research on stuttering has often not met this criterion. Further consideration of this appears below in the differentiation between incidence and prevalence.
Desirable health events: This is particularly of interest in stuttering with regard to rates of remission. Many of the data on this desirable health event seems to have been generated retrospectively, a strategy that causes epidemiologists to look askance for all sorts of reasons that are beyond this presentation. Issues of remission and recovery are elaborated briefly below in the section on prevalence.
Gordis’ definition of epidemiology has an important addition. Epidemiology is used for the "control of health problems" which indicates that knowledge and application of epidemiology are important to prevention. Prevention cannot be achieved merely by education and information but must be subjected to the same measures of outcome as those applied to intervention for health events (Lubker, 1999).
And finally in the definition, Populations: Careful definitions of criteria for group membership are a hallmark of good epidemiological methods.
Incidence Is Not the Same as Prevalence: Never in a Million Years
The differences between incidence and prevalence are not vocabulary, lexical differences. The differences are conceptual, and they underlie formulation of research questions and data interpretation.
Incidence: Let’s start with a shocker. From an epidemiological perspectives, no incidence data for stuttering exist. None. Zero. Nada. Data on occurrence of stuttering may or may not be accurate, appropriately collected or consistent, but one thing is sure: They are not incidence data. Some of our colleagues are coming close! (Ambrose, Cox & Yairi, 1997; Yairi & Ambrose, 1992; for example). The epidemiological caution is that clinical populations are often not representative of the larger population (Lubker & Tomblin, 1998).
The reader will want to take deep breath and read very carefully. First of all, incidence is not a number; it is a rate. Second of all...and this is the tricky part...incidence is a measure only of new cases . These are the numerator. They occur in a population of people who are disorder free (the denominator) during a specified period of time. The new cases are not those just newly identified; they are newly occurring. This means that incidence rates can be generated only from those people who do not stutter and then move from "healthy" (disorder free) to "unhealthy" (disordered). Furthermore, to generate incidence data, populations must be followed over time; time is another dimension of the rate. The old sayings that "Incidence is all the cases of stuttering that ever happened" and "Incidence of stuttering is always greater than prevalence of stuttering" are simply not true.
The term "birth incidence" is often used in epidemiology. This indicates that the numerator population, with, for example, congenital anomalies was never disorder free. It also carries the subtle connotation that the "true incidence" may be different from "birth incidence" because products of conception from spontaneously terminated pregnancies have a high prevalence of chromosomal and genetic anomalies.
Tenacious theoretical biases about stuttering have distorted epidemiological data. These biases have inhibited calculation of incidence rates, that is rates and proportions of new cases. The theoretical posture that dominated the profession for a quarter century, the belief that the disorder is in the ear of the listener and the eye of the beholder and that most dysfluencies are "normal", prohibits identification of those who comprise the incidence numerator, the new cases among child populations
Prevalence: The estimated prevalence rate for stuttering in adults is not the focus of these comments. Prevalence data are those that we read as "x" number of cases per 1,000 people. Prevalence is the total number of people who have a health condition (the numerator) in a carefully defined population (the denominator) at a specific point in time or during a carefully defined time. For several important epidemiological reasons, the prevalence data on stuttering are suspect. The reported overall prevalence rate hovers at something less than a ubiquitous 1%. Since sex disrtributions for stuttering and survival rates for males and females are different, who has attempted to calculate sex-specific prevalence rates in adults? What are the numerators? What are the denominators? Whether reported prevalence rates are correct or incorrect is a function of several variables. Prevalence is influenced by who dies, who survives, who goes into remission, who experiences recurrance, who recovers and methods of data collection and analysis. The estimated remission rate of about 80% for stuttering has been hanging around for decades. Epidemiological flaws in methods such as interviews with college students used to generate such proportions are subjects for another day. The point is that epidemiologists would exploit the principle of remission to study and compare in detail the characteristics not only of those who continue to stutter but also the characteristics of those who "recover". Epidemiologists would NOT use the strategy of issuing the disclaimer that those who "recovered" were not really stuttering to begin with. From an epidemiologic perspective, the fact that the speaking behaviors go away is not an incidence issue, but an influence on prevalence variability in age strata. Another complexity in deriving such rates is that speakers move in and out of the prevalence pool.
Ambrose and Yairi (1999) have written:
Stuttering is known to exhibit a very strong gender difference in its incidence. Early in the disorder’s course during the time when many children recover from stuttering, there are about twice as many males as females who stutter. But in older children and adults, there are four or five males who stutter for every female. Females, then, as explained in Ambrose, Cox, and Yairi (1997) are more likely to exhibit natural recovery from stuttering than males (p. 905).
Let’s look carefully at their comments from an epidemiological point of view. They seem to have used incidence as a global term to cover both gender-specific onset (incidence) and gender-specific prevalence in older populations where onset (new cases) is relatively rare. Their position actually seems to be that incident cases have less marked gender differences (about 2:1 males for females) than do gender ratios (4 or 5:1) in older groups where only prevalence is influenced by the variability of male and female recovery.
In addition to the above variables, at least two other powerful influences on prevalence estimations are in operation. The first is who is doing the estimating and reporting. The second is the continuum of stuttering severity. A clinical anecdote illustrates the complexities here. Edward, a 45-year-old professor of French, has observable speaking behaviors identified by speech-language pathologists as characteristics of stuttering. One day Edward said, "I used to stutter." A speech pathologist asked, "Do you still stutter?" Edward replied, "Oh, no. I haven’t stuttered in years." The speech pathologist would include all the Edwards in the prevalence numerator. The Edwards would not include themselves. Who is right and who is wrong is not the way to pose the epidemiological prevalence question. It is of greater interest to ask whether prevalence rates derived from speakers’ reports (men? women? race groups? ) are the same as those derived by observers (of men? of women? of race groups?). The clinical research on "the moment of stuttering" might be informative here.
Another major influence on prevalence estimates is the continuum of severity and again who is doing the estimating (Lubker, 1997). Clinicians are quite likely, and justifiably so, to report that stuttering in some individuals is very mild and not really clinically significant. Epidemiologists, on the other hand, will include everyone who stutters in the numerator and will conduct analyses in various combined configurations of severity, sex, age, and race strata.
The incidence/prevalence distinction is important in the HIV epidemic in the US and around the world. People who have been identified as being HIV positive comprise the prevalance rates. These are people who for whatever reason have found their way into clinics or are in treatment. How much of the disease prevails? True incidence rates are elusive. How many unidentified positive individuals are in the population? How many people die each year without having been diagnosed? All of these people, plus those diagnosed, form the elusive incidence numerator for HIV each year. Gross prevalence data show the magnitude of this public health problem. However, epidemiological analyses controlling for specific population variables (race, age, sex, geographic location, for example) have shown the epidemic’s inexorable march into the heterosexual female population.
Despite Americans’ devotion to how often something occurs in the United States, gross national prevalence rates of stuttering (or of other disorders or behaviors, for that matter) are not very useful in determining cause, in designing prevention programs or in measuring risk (Antoniadis & Lubker, 1997; Lubker, 1999; Lubker & Moscicki, 1993). As noted in the HIV example, national prevalence rates may give information on the magnitude of problems in public health. Only data derived from carefully defined groups with comparisons among groups according to selected population characteristics can permit interpretation of prevalence data in more sophisticated ways.
Strange Assertions about Stuttering that Defy Epidemiological Principles
It is hard to find citations for the exact origins of assertions about stuttering that need to be re-examined from epidemiological perspectives. (Despite careful searching, I have not found these assertions in print. Written citations and attributions for the following assertions are welcomed from readers. There is no intention to plagiarize. If I were going to steal somebody’s sayings, these would not be the ones.) The assertions seem to appear as "factoids" at speech-language pathology conventions, at dinner parties, and in line at hotel check-in when people are showing their knowledge and not listening to others. An example is the statement that "Genetics accounts for everything about stuttering but its severity." People who specialize in genetic epidemiology would ask several questions. What do we know about degrees of genetic penetrance in different populations? Have epidemiologic methods been applied to studies of the continuum of stuttering severity? Would we accept a similar assertion about genetics and the behaviors observed in the severity continua of autism or schizophrenia?
Another assertion making the rounds is, "When we know what causes it, it’s no longer stuttering." Does this assertion imply that dysfluencies in populations with Down syndrome are not to be classified as stuttering and that we know what causes stuttering in that population? An epidemiological question is whether the word "stuttering" is to be reserved only for idiopathic dysfluency. This decision has major consequences for estimating both incidence and prevalence rates in several well-defined populations. It seems to disregard the accepted principle of co-morbidity associated with other communication disorders (Lubker & Tomblin, 1998). Epidemiologists would suggest that we find other examples in the health sciences. "Skin spots" come to mind. We would never assert that once we are able to differentiate among chicken pox, scarlet fever, rubella and red measles that the major diagnostic signs of observable dermatological responses no longer exist. The epidemiological principle here is that the conditions observed as "skin eruptions" have multiple causal agents, and in fact, the patterns of eruption on the body and the nature of petechiae are different for each of these health conditions if we know where and when to look. These eruptions also have a continuum of severity, and we look for factors contributing to the severity. I suspect such epidemiological reasoning is applicable to stuttering. Fortunately, pursuit of different types and subtypes of stuttering and their variability according to, for example, gender and age of onset, continues apace (Ambrose &Yairi, 1999, for example).
Tools of Epidemiology
A sterotype is evident in the witicism adapted here that clinicians study matched pairs of ten people who stutter and ten who do not in a clinic in Butte, Montana; epidemiologists study Montana (Lubker, 1991a). Epidemiologists do tend to study larger populations and to control a broad range of variables through creative analyses. One problem is that large numbers of analyses on the same variables may produce statistical significance by chance. Many epidemiologists are experts in designing surveys and interview protocols. They utilize large data bases, census data, clinical records review from hospitals and other treatment settings, birth and death certificates. The Centers for Disease Control and Prevention employ some of the most sophisticated epidemiological research methodologies in the world. Their system of infectious disease reporting provides epidemiological data with powerful geographic distributions. Their surveillance of congenital anomalies could provide dissertation data for a host of prospective studies in speech-language pathology.
Epidemiologists are also respectful of good, informed clinical observations. The history of rubella as a teratogenic agent and consequent need for pediatric audiologists is a classic example of the interdependence of epidemiology, clinical activities, and professional education.
Summary
Some people say that epidemiologists aren’t very funny. They do tend to crunch numbers and to speak biostatistics. Many readers will have heard the old saw that epidemiologists are scientists who study people broken down by age and sex. However, epidemiology offers ways of thinking and posing questions and analyzing data that can enrich clinical practitioners' pursuit of truth. These methods seem to have, in some instances, been misapplied and only superficially understood. In other instances their utilitiy seems not to have been recognized. Practitioners are encouraged to search for evidence of epidemiology in their journals and for opportunities to apply practical epidemiology to their questions (Lubker, 1991b).
Works Cited
Ambrose, N.G., Cox, N., & Yairi, E. (1997). The genetic basis of persistence and recovery in stuttering. Journal of Speech, Language, and Hearing Research, 40, 567-580.
Ambrose, N. G., & Yairi, E. (1999). Normative dysfluency data for early childhood stuttering.Journal of Speech, Language, and Hearing Research, 42, 880-894.
Andrews, G., Harris, M.. Garside, R., & Kay, D. (1964). The syndrome of stuttering. (Clinics in Developmental Medicine., No. 17). London: Spastics Society Medical Education and Information Unit, in association with Wm. Heinemann Medical Books Ltd.:
Antoniadis, A. & Lubker, B. B. (1997). Uses of epidemiology in preventing
communication disorders. Journal of Communication Disorders. 30, 269-284.
Bloodstein, O. (1975). A handbook on stuttering. Chicago: The National Easter Seal Society for Crippled Children and Adults.
Goberman, A. M. & Robb (1999). Acoustic examination of preterm and full-term infant cries: The long-time average spectrum. Journal of Speech, Language, and Hearing Research, 42, 850-861.
Gordis, L. (1996) Epidemiology. Philadelphia: W. B. Saunders Company
Greenberg, R. S., Daniels, S. R., Flanders, W. D., Eley, J. W., & Boring, J. R. (1996). Medical epidemiology. Stamford, CT: Appleton and Lange.
Lubker, B. B. (1991a). Epidemiologic models for prevention, risk measurement
and causality. Seminars in Hearing. 12, 116-130.
Lubker, B. B. (1991b) Practical applications of epidemiology in programs for
infants and toddlers. Infant and Toddler Intervention. 1:1, 47-59.
Lubker, B. B. (1997). Epidemiology: An essential science for speech-language
pathology and audiology. Journal of Communication Disorders. 30, 251-268.
Lubker, B. B. (1999). Prevention science. In Prevention for speech-language pathologists and audiologists. Rockville, MD: American Speech-Language-Hearing Association. in press.
Lubker, B. B. & Moscicki, E. K. (1993). Practical uses of epidemiology in
prevention of communication disorders. In L. Cole (Ed.). Prevention of
Speech, Language, and Hearing disorders. (revision and update), American
Speech-Language-Hearing Association, Rockville, MD, 1-30.
Lubker, B. B. & Tomblin, J. B. (1998). Epidemiology principles and perspectives:
Informing clinical practice and research on language disorders of children. Topics in Communication Disorders. 19, 1-27.
Yairi, E., & Ambrose, N.G. (1992). A longitudinal study of stuttering in children: A preliminary report. Journal of Speech, Language and Hearing Research, 35, 755- 760.