Edited by: Jimmy Thomas Efird, East Carolina University, USA
Reviewed by: Eugenia M. Bastos, Bastos Consulting, USA; Brian Godman, Karolinska Institutet, Sweden
Specialty section: This article was submitted to Epidemiology, a section of the journal Frontiers in Public Health
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
To examine the application of continuum models to tuberculosis, HIV, and other conditions; to theorize the concept of continua; and to learn lessons that could inform the development of improved care and prevention continua as public health metrics.
An analytic review of literature drawn from several fields of health care.
The continuum construct is now part of public health evaluation systems for HIV, and is increasingly used in public health and the medical literature. Issues with the comparability and optimal design of care continuum models have been raised, and their methodologic and theoretic underpinnings and scope of focus have been under-addressed. Review of relevant publications suggests that a key limitation of current models is their lack of measures reflecting incidence and mortality. Issues relating to continua data being longitudinal or cross-sectional, definition of numerators and denominators for each step, data sources, measures of timeliness of step completion, theoretic models to facilitate inferences of causes of care continuum gaps, how measures of prevention efforts, reinfection/relapses, and interactions of continua for co-occurring comorbidities should be reflected, and how analyses of differences in retention over time, across geographic regions, and in response to interventions should be conducted are critical to the development of sound care and prevention continuum models.
Lessons learned from the application of continuum models to HIV and other conditions suggest that the application of well-formulated constructs of care and prevention continua, that depict, in well defined, standardized steps, incidence and mortality, along with degrees of and time to screening, engagement in care and prevention, treatment and treatment outcomes, including relapse or reinfection, may be vital tools in evaluating intervention and program outcomes, and in improving population health and population health metrics for a wide range conditions.
The construct of “care continua” has become an important tool in the evaluation and improvement of the overall care for certain conditions (
The construct of care continua is also being increasingly used in other clinical and public health settings, such as in evaluations of care systems for other infections, such as hepatitis C virus (HCV), sexually transmitted infections (STIs) (
Our goal is to more fully theorize the continuum construct and to help develop understandings, definitions, and applications of care continua so as to improve their use as valuable tools for scientific, programmatic and public health evaluations and interventions generally. To achieve this goal, we will examine some examples of the valuable application of continuum models and explore issues with and limitations of existing models and make suggestion for their improved use.
Examined literature comes from several fields of social science, health care and public health over several decades. In order to explore the continuum construct through a variety of lenses, we adopted an analytic and synthetic approach to draw lessons from these diverse sources. This approach relied on traditional literature review search methods and on the application of the case study methodology and the comparative method (
The concept of care continua derive from the Piot and Piot-Fransen models for tuberculosis (TB) and STIs, respectively, where models focused on operational considerations that arose and that reduced the overall effectiveness of clinical and public health efforts already demonstrated by studies to be efficacious in idealized settings (
Continuum models are usually graphically represented as a bar graph where each bar represents the proportion of persons completing each step (
HIV likely represents the most established and successful application of a continuum model. More recent applications of continuum models for the prevention of maternal-to-child transmission of HIV and for HIV care generally (
One model, “the prevalence-based HIV care continuum,” examines each specified step as a proportion of the total number living with HIV including those diagnosed and those undiagnosed. The other model, “the diagnosis-based HIV care continuum,” uses as a denominator the total number of those diagnosed with HIV excluding those who have not been diagnosed. The prevalence-based model can be used to assess outcomes for broad populations such as young women, but not subgroups of these populations, e.g., low-income young women. In contrast, the diagnosis-based model allows examination of more detailed population subgroups referred to as stratified continuum models.
Similarly, continuum models are also proving valuable in identifying gaps and focusing resources for HCV, TB, and other services (
Despite the contributions already made through the use of continuum models in HIV and other fields, there are a number of key issues central to their optimal use as evaluation tools generally that require further consideration. These include issues of theory; the delineation of specific steps standardization; reflecting incidence, time, and disease-specific morbidity and mortality; and statistical analysis of the models that will be addressed in the sections that follow.
An identification of gaps does not in and of itself provide an understanding of the reasons for such gaps. An understanding of reasons for gaps in or barriers to the progress of individuals or populations through the steps of care requires the application of an appropriate theoretic framework. Where continuum analyses are guided by an appropriate theoretic model, factors affecting progress through sequential steps can be more fully examined; for example, some studies demonstrate the importance of structural- and individual-level factors as determinants of progress through continua (
There are frequently large time gaps between demonstrations of efficacy and implementation in practice (
The steps of any specific continuum model should reflect the specific clinical features of that condition and actual processes of prevention and care and should be chosen to facilitate elucidation of potential barriers to progress through the continuum so that discrete barriers can be addressed. Many continua models begin with an initial step of awareness of risk or a condition and whether individuals may seek attention or of active testing, case finding, or screening of either high-risk or general populations to identify those with the condition in question (
One key limitation of beginning a continuum model with the steps of seeking or testing is that the model then may fail to reflect a key public health indicator of disease: incidence. Continuum models often begin with the proportion of a general or known-to-be-positive population who are screened (
Continua constructs may be valuably applied to prevention as well as to care where initial testing or screening are crucial both in identifying those affected by the specified condition and those who may be at risk for but not yet have the condition (
The issue of the connection between screening and acting on the results of screening was highlighted by John Sbarbaro in an editorial entitled “To seek, find, and yet fail” written in response to a novel TB skin testing program, which identified a high prevalence of latently infected persons and yet included no efforts to link such persons to evaluations to exclude active TB or initiate treatment of latent TB infection (
Further, diagnostic evaluations are often individualized by providers influenced by hidden cognitive processes related to the providers’ estimation of a patient’s resources, often within constraints imposed by patients, organizations, and insurers (
Lourenco et al. have noted that HIV continuum models used different countries vary both in enumerated steps and definitions of steps and argue for the need for standardization of the HIV continuum construct to allow continuum comparisons (
A recent systematic review of publications examining HIV continuum models focusing on data sources, methodology, and study comparability with respect to these parameters (
Another recent paper examined an 8-step HIV continuum model and conducted sensitivity analyses based on variations in continua definitions (
Overall, these issues highlight the potential impact of variations in definitions on measured outcomes and the need for definitional clarity and sensitivity analyses.
In current continuum models, the depiction of sequential steps along an axis does not represent the time required for transitions from one step to the next. For example, the HIV continuum does not convey the fact that many people are diagnosed late in the course of their infection (
Another key aspect of prevention and care relates both to the defined population and to the recognition that populations are not stable overtime (
Another issue is whether continuum analyses examine cross-sectional or longitudinal data, a critical distinction where achieving outcomes takes variable amounts of time and where outcomes must be sustained overtime. Colasanti et al. examined continuous retention in care and continuous viral load suppression over 36 rather than 12- or 24-month periods (
Use of continuum models that reflect longitudinal rather than cross-sectional data may be particularly important for understanding chronic conditions. For example, if a gonorrhea or syphilis case is diagnosed in any given year (where treatment can be a single dose or a brief course), treatment and cure should be obtained within that same year; a cross-section analysis would reflect this. Alternatively, for conditions requiring longer treatment (e.g., HIV), cross-sectional analyses may overestimate continuum progress.
For both HIV and HCV, a key continuum “endpoint” of viral suppression will remain relevant and vital. The rationale for the importance of this endpoint is the excellent data that viral load suppression translates both into individual-level quality health outcomes and population-level conditions that result in lower likelihoods of transmission with subsequent reductions in incidence which might then lead to epidemic control, elimination, and theoretically eradication (
Further, virologic suppression or any other biologic or behavioral outcome measures are at best surrogates of the fundamentally more critical endpoint of mortality. The nineteenth century physician and epidemiologist William Farr noted that “the death rate is a fact. Everything else is an inference” (
Another issue to be addressed relates to the reality that individuals may have more than one health condition, so that an individual may in fact be moving through multiple continua which may be interrelated to varying degrees. For example, for individuals with HIV/HCV, coinfection will be considered as part of both HIV and HCV continuum, respectively, and at an individual-level will need to move through the steps of both as part of optimal health care.
Movement through continuum for two or more conditions are likely to be impacted by the specifics of the service delivery systems, which would include whether care occurs through an integrated system that addresses all of the conditions or whether care is delivered through separate systems. Broadly speaking, this relates to the issues of vertical versus horizontal models of care and public health funding streams (
Movement through one specific continuum may also be directly tied to movement through another continuum. Providers may make the decision to prioritize achieving HIV viral suppression prior to the initiation or consideration of HCV treatment; in fact, some treatment guidelines and insurance policies suggest or “mandate” this (
Understanding that conditions identified as potential barriers to progress through a continuum for one condition may in fact be disorders requiring intervention, may allow improved understanding of the interaction of care systems and of how conditions may act as barriers, and may allow the development of more refined variables.
Central to the development of any continuum model is the issue of identifying appropriate data sources for each of the identified steps. For HIV and TB, such data sources are reasonably well developed (
In formal analyses, continuum progress could be viewed a sequential ordinal variable where earlier stages are prerequisites for later stages (e.g., there is no HCV treatment initiation without HCV medical evaluation completion). Several formal quantitative analytic methods may be particularly valuable. In formal analyses, achieving sequential continuum steps may be viewed as a count variable. One approach could be to use the continuous ratio model (CRM) which is well suited to sequential outcomes of this type (
In analyzing progress through sequential steps is the consideration of whether each step requires comparable effort or results in comparable public health impact. Steps then may need to be weighed based on the varying on these considerations. One consideration is also whether the steps of a continuum should (or do) represent critical individual-level or population-level milestones, or whether they primarily reflect measurable milestones, and these may not be the same thing. It might be that shifting a continuum curve in which an improvement of some proportion at one step may not translate to relevant gains in population health, while an improvement of the same proportion at another step might do so (
This leads to the issue of whether continuum steps are of necessity sequential, or should always be viewed so representationally or analytically. Qualitatively, it has been abundantly noted that individuals may complete any given step of a continuum model and for a range of reasons, not proceed to the next clinically logical or desired step, but then at some subsequent point in time become re-engaged at the same or even re-enter at an earlier step in the continuum model (
An underutilized potential of continuum analyses is their application to evaluating the impact of interventions to improve the steps of care. Hayes et al. used graphical representations of STI continuum outcomes to estimate the potential impact of different public health strategies on STI outcomes (
Review of the literature suggests several key implications for the improved use of continua models as clinical and public health tools. Models would optimally reflect incidence and distinguish incident and prevalent cases. They should reflect disease-specific and all-cause morbidity and mortality. Optimal models would reflect relapses and reinfections, as well as measures of primary and secondary prevention. Models should also reflect the timeliness between steps and have explicit and appropriate definitions, data sources, and means for handling those who move or die. Models should be understood to require the use of theoretic frameworks that consider structural as well as individual causes of identified gaps.
Numerous issues in the delivery of care and prevention for many conditions resemble those identified in the HIV, HCV, and TB continua, including issues of underdiagnoses, gaps in linkages between screening and initial diagnosis and engagement in treatment, issues in treatment retention, adherence, and relapse, and the interdigitation of continua for relevant comorbidities. Systems of care and prevention for many conditions can appropriately and usefully be viewed as consisting of a care and prevention continuum including steps of incidence, screening/identification, medical/psychosocial evaluation for treatment, engagement in evidence-based treatment, retention in treatment through to well-defined measures of treatment success, as well as degrees of engagement in evidence-based interventions to prevent relapse, and measures of overall and substance-related-specific mortality. It would be critical to define the denominator most relevant to each specified step. As with HIV and other continua, the use of various population denominators will be important in addressing different questions. It would be essential to identify appropriate data sources for each step, relevant and valid measures of treatment success, standardized definitions of numerators and denominators, and standard methods to account for those who move or die, and handling missing data. It would also be appropriate to reflect relationships between a given continuum and continua for key comorbidities (
In reviewing the literature from several fields using traditional literature review search methods, some important contributions may have been missed and selection bias could have introduced. However, the review of literature from multiple fields serves as a form of triangulation which may ameliorate this risk (
Well-constructed and standardized continua models are proving to be invaluable for program development, evaluation and policy, for public and private health systems in standardizing and evaluating outcomes, informing study design, modeling, resource allocation, and for facilitating standardized comparisons of an expanding range of health outcomes across programs, states, and countries. Review of lessons learned from the valuable application of continuum constructs suggests that steps of the awareness, of screening, of linkage to evidence-based treatment and retention in such treatment, and of monitoring timely movement between steps, incidence, relapse/reinfection, and mortality. How best to reflect some of these factors will require more consideration and development. Identifying optimal data sources for continuum steps and standardizing definitions for these steps and of relevant numerators and denominators will be needed. Similarly, optimizing methods for quantitative analysis of progress through continua and of the impact of interventions on such progress is also needed.
In conclusion, the application of well-formulated constructs of care and prevention continua, that depict, in well defined, standardized steps, incidence and mortality, along with degrees of and time to screening, engagement in care and prevention, treatment, and treatment outcomes including relapse or reinfection, may be vital tools in evaluating intervention and program outcomes and in improving population health and population health metrics for a wide range conditions.
DP, AJ, and DN analyzed and interpreted the data. AJ wrote the first draft of the manuscript. DP, AJ, and DN read and contributed to multiple versions of the manuscript. All the authors read and approved the final manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This study was supported in part by P30 DA011041. There was no other funding to support this work.