Definitions and Measurement of Chronic Health Conditions in Childhood:  A Systematic Review FREE

As a result of improved efficacy of treatments and care for life-threatening pediatric conditions, the survival of children with serious congenital or acquired diseases in developed countries has increased during the last decades.^1- 2 Yet, many of these children have not been cured (eg, those with cystic fibrosis) or they have disabling sequelae of their disease (eg, cerebral palsy and juvenile rheumatoid arthritis) or its treatment (eg, retinopathy of prematurity). The incidences of some other conditions, such as asthma and obesity, are rising. Therefore, the prevalence of chronic diseases in children and in young adults, which is a function of incidence and duration, has increased since the 1980s and will likely increase further.^3- 6

In addition to the added disease burden of children and their families, this increase in prevalence has considerable financial and organizational consequences for health care planning and for employment. To enhance the societal participation of these children as they approach adulthood and to be able to plan for sufficient and adequate facilities, policy makers and politicians need to rely on valid prevalence data. Such prevalence data on chronic conditions in children and young adults are not only useful for planning for health care and provisions for the welfare of young adults with chronic conditions, they can also be considered, in addition to mortality statistics, as an outcome measure in comparisons of determinants of youth health between countries and over time.⁵ So far, many different approaches have been used to measure the prevalence and consequences of chronic diseases and health conditions in children, resulting in a wide variability of prevalence estimates that cannot be readily compared. Research on prevalence of chronic diseases in children is often focused on 1 or more specific diagnoses.⁷ An unequivocal definition of the concept of chronic health conditions in childhood has been lacking, and, as a consequence, operationalizations to measure the prevalence are diverse.

The aim of this article is to provide a systematic overview of all definitions and operationalizations that have been used to measure the prevalence of chronic health conditions in childhood. By operationalization, we mean all aspects of the methods of a prevalence study, including (1) the source of the information (eg, parents, medical files, or insurance data); (2) the instrument used to obtain the information (eg, questionnaire, interview, list of diagnosis codes, or computer query); and (3) the study population (eg, a representative sample of all noninstitutionalized children aged 0-18 years or all children registered in a hospital database).

Two literature searches were performed in MEDLINE using PubMed (complete database through December 2006) on the topic of definitions of chronic conditions in children, and 1 in Web of Science (1988 through December 2006). The search strategy is available from the corresponding author. In addition, we searched references by hand for related articles.

Articles that fulfilled the following inclusion criteria were included: (1) the main focus of the article was on children (ie, aged 0-18 years); (2) the definition concerned chronic conditions, excluding articles focusing on specified chronic conditions and on disability; and (3) the aim of the article was either to present a definition or to estimate the prevalence of chronic conditions in children. We excluded non-English and non-Dutch literature. Two of the authors (J.H.L. and L.B.M.) independently performed the selection of articles based on titles and abstracts retrieved in the 3 searches. When in doubt, the complete article was screened using the same criteria. The decision to include articles in this review was reached by consensus.

Information was extracted from the articles on (1) the conceptual definition; (2) the operationalization in terms of source of information, method of information retrieval, and study population; and (3) the resulting prevalence rate. Data were collected by J.H.L. and L.B.M. Only minor differences occurred, which were solved by discussion. All definitions and operationalizations were assessed in light of our aim of contributing to the development of a definition that could be used to make comparisons of the prevalence of chronic conditions in children between countries and at different points in time. We also tried to evaluate the influence of the following possible sources of variance on the reported prevalence rates: (1) the concept of chronic disease underlying the definition; (2) the operationalization of the definition; (3) the country; and (4) the year in which the study was performed.

The search yielded 7252 articles, of which 64 articles were included. All of the articles were in English, from 14 countries. Articles that did not comply with at least 1 of the inclusion criteria were excluded. Table 1 describes the most frequently cited definitions and Table 2 presents the prevalence studies with the operationalizations used. Table 3 describes articles in which a definition is presented with a different goal than to measure prevalence.

Many different terms are used to denote the central concept, the most frequently used terms being chronic conditions, chronic health conditions, chronic illness, and (children with) special health care needs (Tables 1, 2, and 3). A conceptual definition was presented in 61 of the 64 articles; the clarity and theoretical basis of these definitions varies considerably. Many authors referred to 1 of the 4 definitions presented in Table 1, which were the most frequently measured concepts.^8- 10,12

The earliest definition concerned duration of disease and its impact on daily activities.⁸ The type and extent of impact of the condition on the child's functioning and health care needs is emphasized more in the later definitions of Perrin et al⁹ and Stein et al,¹⁰ while in the most recent definition of children with special health care needs, the types of conditions included are more extensive, even including “increased risk for a chronic condition,” whereas the impact is limited to “requiring health and related services.”¹²

An important issue for debate in the United States in the 1990s was the distinction between a conceptual definition based on a list of diagnoses and a noncategorical definition based on the duration and consequences of the condition of the child.⁹ The noncategorical approach was adopted as a reaction to the diagnosis-based approaches that had been used before to assign health care facilities and financing to individual children.¹⁰ Instead of a list of diagnoses, the basis of the noncategorical approach was the functional limitations resulting from the health condition, including dependency on medication, assistive devices, or medical care. In this way, the similarity of specific consequences of various diseases and conditions was taken into account.

Table 2 includes the 32 articles presenting operationalizations of the definitions (ie, the methods to measure prevalence) and the prevalence rates. The conceptual definitions vary considerably with regard to their clarity and comprehensiveness; sometimes they were not stated in the article.^16,31 In 8 of the 32 articles in Table 2, data generated by the National Health Interview Survey were presented.^{11,13,18,21,23,26,30,34} The concepts and methods of the National Health Interview Survey have been adapted every 10 to 15 years since the first survey in 1957.⁷⁶ Both the noncategorical approach and lists of diagnoses were used. Information was retrieved from interviews with parents^{11,13,16- 18,21- 23,26,30,32,34,40- 41,44,46} or with the patients themselves,³⁶ from medical examination of children,^25,27 from claims files,^33,35,45 or from medical records.^{19- 20,24,32,39} Sometimes health care professionals were asked to report all designated children known to them, either based on medical records³¹ or without a specified method.^14- 15 One study combined prevalence data from several registries and earlier prevalence studies,²⁹ and in another article an overview of surveys based on school samples in several countries was presented.⁴²

The study population described in each article depended on the purpose of the study. General population samples were drawn in a systematic way in studies aiming at generating prevalence data^{13,17- 18,21,23,26,30,49- 50}; other samples were used to answer different questions—eg, when comparing the lifestyles of adolescents with chronic conditions with those of their healthy peers.³⁶

The prevalence rates varied from 0.22% for “medically fragile children” aged 0 to 19 years, a small subset of all children with special health care needs identified from North Carolina's Medicaid paid claims database,⁴⁵ to 44% of children aged 0 to 18 years visiting an urban health center identified by a diagnosis list.³² Restriction to the terms chronic conditions, chronic health conditions, and chronic illness in the general population yielded a range in estimated prevalence rates from 3.49% of children younger than 16 years identified by public health nurses¹⁵ to 35.3% of youth 12 to 17 years old identified by a combination of parental history and physical examination.²⁷

Comparison of different operationalizations in the study by Rolnick et al³⁹ showed variation in the resulting prevalence rates, from 11% when using the “clinical risk grouper” on administrative data to 37% when using the Questionnaire for Identifying Children With Chronic Conditions in telephone interviews in the same population. Prevalence differences due to source of information were found by Youngblade and Shenkman⁴³ by comparing interviews with parents (17.8%) and adolescents (12.6%). Jessop and Stein²⁷ found little overlap between children identified by physical examination and parental history. Not surprisingly, generally higher prevalence rates were reported in older children; eg, in the article by Vuorinen et al,²² a prevalence rate of 8.8% in 0- to 6-year-olds and 15% in 7- to 14-year-olds. There also appeared to be a time effect indicated by higher prevalence rates in more recent studies (Table 2).

Our review summarizes the current definitions and operationalizations of the concept of chronic conditions in childhood. The included articles contain a wide range of concepts, from “medically fragile children” to “children with special health care needs.” The large number of excluded articles (7188/7252) illustrates the difficulty of performing a specific literature search on such a broad concept as chronic health conditions in childhood. Because of the complexity of the concept and the diversity in methods used in studies on this subject, there is, at present, no such thing as a “proper way” to combine prevalence rates quantitatively. We attempted to use transparent and reproducible methods. However, a fair amount of judgment in selecting and appraising individual articles proved unavoidable because of the considerable variation in terminology, types of articles, and study designs that we found. This makes a systematic review of definitions, measurement methods, and prevalence rates of chronic health conditions in childhood less convenient than a systematic review of, for instance, randomized controlled intervention studies.

Only articles in English were included because none of the Dutch articles met all inclusion criteria. It is possible that the language restriction to English and Dutch resulted in omission of 1 or more existing definitions or operationalizations. Also, it is likely that national and international public health organizations assemble and report data based on diverse definitions that are not reported in the scientific journals indexed in MEDLINE or Web of Science. Most of the included articles (41/64) originate from the United States, raising the possibility of selection bias. Nevertheless, a clear conclusion of a current wide variability in concepts, operationalizations, and, as a consequence, prevalence rates can be drawn. This variability appeared not only due to variations in definitions (eg, whether noncategorical definitions or diagnosis lists are used) but also to diversity in operationalization of similar definitions, including the source of information (eg, parents or adolescents), whether interviews or medical records were used, the age of the study population, and the year in which a study was performed.

Newacheck et al¹⁸ suggested several possible causes for the apparent increase in prevalence rate in the National Health Interview Survey between 1960 and 1981, including the following: (1) prior screening using a checklist of conditions in the early years, thereby ignoring conditions not listed; (2) increased survival of neonates and older children with chronic illnesses; (3) deinstitutionalization of disabled children; and (4) increased availability of medical services for children. This suggests that at least part of the increase can be ascribed to changes in the definition and operationalization of the concept of chronic conditions in childhood. Whether the prevalence rates also vary between countries is unknown, since no studies were found in which the same definition and operationalization were applied to assess prevalence rates in different countries.

One would expect that the choice between a noncategorical definition and a list of conditions or diagnoses would have consequences for the operationalization. For example, medical records, except in rehabilitation settings, do not generally contain comprehensive information on children's functional limitations. Parents are not always able to state all medical diagnoses given to their child, while they do know his/her functional limitations well. However, surprisingly, several authors who adopted a noncategorical definition report use of medical files,⁶² and, vice versa, some authors use diagnosis lists in interviews.⁷³ However, in several studies it seemed that the concept of what was measured was dictated by the methods by which data were gathered, rather than the other way around.^18,50

When considering the variety of studies included in this review, one can conclude that the various concepts that authors have aimed to measure cannot be denoted by a single term. However, differences or similarities among the various concepts are not always transparent. For example, it is not difficult to understand that the group “medically fragile children” is a subsample of children with special health care needs, but even authors who used similar terms presented very diverse operationalizations and results. The variety in terms and operationalizations that we found shows that there is no consensus among researchers about concepts and definitions and that, in most articles, the only source of information on what exactly was measured is the methods section concerning the operationalization.

All conceptual definitions that are summarized in Table 1 incorporate the consequences of the chronic condition in terms of functional limitations and/or the need for medical or other care services. Theoretically, this leads to a lower measured prevalence compared with the prevalence of all children with 1 or more conditions that last longer than a certain period of time (usually 3 or 12 months). If the prevalence is measured with the aim of establishing the need for services, this is justified. However, comparisons between countries and within a country over time are hampered because the consequences of a chronic condition in terms of care needs or functional limitations vary according to cultural, educational, and financial circumstances and availability of treatment and care. For example, behavioral problems that might have been labeled as “bad behavior” in the past may now be diagnosed as attention-deficit/hyperactivity disorder, a chronic condition needing treatment. Or a young child in rural Africa may be considered lazy, whereas he or she might be diagnosed as having sickle cell disease if his/her parents had the time and money to go to an outpatient clinic.

Differences among studies in definitions and operationalizations can in part be ascribed to different goals with which these studies were performed. However, these differences are most often obscured by the use of an unstandardized terminology. Comparable with the International Classification of Functioning, Disability and Health, researchers in this field need a comprehensive framework of terminology. The International Classification of Functioning, Disability and Health is a revision of the International Classification of Impairments, Disabilities and Handicaps under the auspices of the World Health Organization (http://www3.who.int/icf/icftemplate.cfm). In this model, the components of functioning and disability can be classified in 3 different dimensions: body structures and functions, performance of activities, and participation in communal life.⁷⁷ A comprehensive framework for chronic conditions should allow for differences in definitions related to different research goals. For instance, evaluation of the outcome of determinants of youth health, including pediatric medical care, calls for a different definition than planning for financing of social security and labor availabilities for young disabled people. The framework might consist of the following concepts:

More levels can be added when addressing subsets of conditions, severity levels, or specific limitations or needs.

Many prevalence studies or registries are concerned with a single disease entity. Simply adding up these prevalence rates disregards possible biases in methods and the possibility of comorbidity, since the units are diagnoses instead of individuals. National population-based surveys are the most useful data source for health state information, including information on chronic conditions in childhood. Since 1957, the National Health Interview Survey has been performed regularly in the United States. Data about chronic conditions in children have been collected by different types of questions, often in the absence of an underlying definition, thereby causing bias in longitudinal comparisons.¹⁴ To avoid this bias in future longitudinal and international comparisons, standardization is important, first of concepts and definitions and subsequently of operationalizations.

To evaluate the effects of determinants of child health, including perinatal and pediatric health care factors, and for the efficient planning of health care facilities and services for children and young adults with chronic conditions, valid and reliable prevalence estimates are needed. The wide variability in reported prevalence rates results from the considerable diversity in concepts and operationalizations in terms of source of information, method of information retrieval, and study population. To make valid international comparisons of this prevalence and comparisons over time, a first requirement is to achieve international consensus about the conceptual definition of chronic health conditions in childhood.