Summation, Saliency, and Saturation (3S): a continuum for a practical approach to assessing sample adequacy and analytic completeness in qualitative implementation research

Abstract

Background

Saturation is a common criterion for determining qualitative sample size adequacy and analytic completeness. The dynamic and fast-paced implementation research environment poses unique challenges for investigators conducting qualitative studies that seek to reach saturation. Saturated studies require an iterative, often lengthy and labor-intensive process of data collection and analysis, which is frequently at odds with implementation science’s focus on rapid turnaround times for translating knowledge into practice. Moreover, despite its common usage, uncertainty around saturation’s meaning and application remains. To date, there has been no systematic attempt to understand how the concept of saturation is defined and deployed specifically in the context of qualitative implementation research, or guidance on how to adapt the saturation concept in response to field-specific needs.

Methods

A concept synthesis was conducted to establish baseline knowledge that would inform field-specific guidance for assessing sample adequacy and analytic completeness in qualitative implementation research. Three leading implementation science journals were searched. Eligible studies (a) described empirical research, (b) discussed the saturation concept in the context of qualitative methodology, and (c) mentioned saturation in the body of the manuscript. Articles were systematically read and coded to identify meaningful content and patterns of interpretation.

Results

Of 207 studies identified, 158 met eligibility for full-text review, and 146 were included in the final analysis. Findings show cursory treatment of the saturation concept. Various saturation-related terms and definitions were identified, as were prevailing interview sample sizes and citation patterns. Studies rarely explained how analytic completeness was determined, and discussion of saturation leading to theory or concept generation was sparse. These findings informed development of the 3S Continuum as an alternative approach for assessing qualitative sample adequacy and analytic completeness.

Conclusions

In implementation research, saturation as an analytic benchmark is seldom explained and difficult to attain. We propose a practical approach for reconceptualizing saturation as part of a larger continuum for assessing sample adequacy and analytic completeness. We aim to help implementation researchers navigate decisions about qualitative sample adequacy and analytic completeness in pragmatic and transparent ways.

Supplementary Information

The online version contains supplementary material available at 10.1186/s43058-025-00833-7.

Contributions to the literature.

• Saturation is a common criterion for justifying qualitative sample sizes in implementation research; however, the implementation research environment presents challenges for saturation attainment.

• The field lacks systematic understanding of how saturation is utilized in qualitative implementation research and guidance on how to adapt the saturation concept in response to field-specific needs.

• Drawing upon findings from a concept synthesis, social science methods literature, and the field’s pragmatic needs, the proposed 3S Continuum fills a gap by offering nuanced criteria for assessing an adequate qualitative sample and analytic completeness in implementation research. This enhances methodological rigor in qualitative implementation research.

Background

A challenging aspect of designing qualitative implementation studies is determining adequate sample sizes, particularly for interview-based studies, a popular method in implementation science [1, 2]. Often, researchers who employ qualitative methods to study and inform implementation processes—whether as qualitative methodologists working within multidisciplinary implementation teams or as implementation scientists incorporating qualitative research into their studies—draw on the concept of saturation to rationalize their qualitative sample size. Originally introduced by Glaser and Strauss as a method to construct sociological theories from qualitative data, saturation (or, what they term “theoretical saturation”) refers to the period during data analysis when the analytic categories identified by the researcher become fully developed or “saturated,” and any additionally collected data fail to yield further development of a category’s properties [3]. This analytic turning point indicates the analysis is sufficiently complete, data collection can stop, and the given sample is adequate for fleshing out the conceptual categories and properties—and the relationships among them—that can generate a theory. Glaser and Strauss’s theoretical saturation has evolved to accommodate looser and broader notions of saturation, generally understood as information redundancy [4, 5]. In practice, saturation attainment involves substantial effort and skill, requiring deep engagement with data, protracted time in the field, and sharp understanding of epistemological assumptions that underpin qualitative research questions to build empirically valid theory or robust conceptual models.

Implementation researchers face a dynamic, fast-paced research environment, which poses unique challenges for qualitative investigations seeking to reach saturation. The time and skills needed to achieve empirical certainty of saturation are frequently at odds with implementation science’s focus on rapid turnaround times for translating knowledge into practice [e.g., rapid analysis [6], rapid ethnography [1, 7, 8], as well as timely research-clinical operations alignment [9–11]. Further, ethics boards, granting agencies, and other forms of soft funding require investigators to pre-specify research parameters and scope. But inherently, saturation is an iterative concept that relies on engagement with data rather than pre-selection. Efforts to quantify a priori qualitative sample sizes needed for saturation [12–14] remain controversial among qualitative methodologists [5, 15].

While some have questioned saturation’s utility in qualitative health research [16–18], saturation is, nonetheless, common criterion for determining sample size adequacy and analytic rigor [16, 19]. Yet, uncertainty around saturation’s meaning and application persists, and efforts to explain and clarify the concept abound [19–26]. This has involved the development of “hybrid” forms of saturation [22] and surrogate terms [5, 22, 23, 25, 27, 28]. Alternative concepts—like information power [18], conceptual depth [21], and sufficiency [29, 30]—represent additional efforts to grapple with saturation’s challenges and ambiguity. This dizzying array of options may leave less experienced qualitative implementation researchers confused about which concept to use.

To date, there has been no systematic attempt to understand how the concept of saturation is defined and deployed specifically in the context of qualitative implementation research. This is a necessary starting point for meaningful engagement with and innovation of the saturation concept in response to field-specific needs. The aims of this paper are twofold: (1) conduct a concept synthesis to understand how investigators use the saturation concept in qualitative implementation research and (2) drawing on these findings, social science methods literature, and the pragmatic needs of implementation research, propose a practical approach for implementation researchers to conceptualize saturation as one option on a larger continuum for assessing sample adequacy and analytic completeness. This expanded conceptual toolkit provides implementation investigators with a pragmatic, rigorous, and transparent way of navigating decisions about sample adequacy and analytic completeness.

Methods

To understand how saturation is invoked in qualitative implementation research, we conducted a concept synthesis focused on the saturation concept [31, 32]. A concept synthesis is an exploratory method used to identify attributes that comprise a concept’s “essence” to refine conceptual clarity [33]. Consistent with the method’s “fit for purpose” approach [31], or strategic selection of sources most likely to yield information needed for concept refinement, we selected papers published in three leading implementation journals—Implementation Science (IS), BMC Health Services Research (BMC-HSR), and Implementation Research and Practice (IRP)—to ensure publications had explicit implementation science focus. International implementation science experts have characterized these journals as most relevant for identifying implementation science articles [34]. The term “saturat*” was searched, which captures iterations of the concept’s deployment (e.g., saturation, saturated, saturate, etc.). Given its broader scope, “implementation” was added to the BMC-HSR search. Web of Science for IS and BMC-HSR queries were utilized; not indexed in Web of Science, PubMed, or Scopus at the time of the search, IRP was searched through the journal’s website. No date range was included. The search was conducted on July 6, 2022, thus capturing articles “first published online” by this date. Eligible studies (a) described empirical research, (b) discussed the saturation concept in the context of qualitative methodology, and (c) mentioned saturation in the body of the manuscript (not only in the bibliography). We used a deductive content analytic strategy, focusing on descriptive analysis [35]. All articles were read and coded in their entirety using predetermined codes to identify meaningful content and patterns of interpretation. Codes were based on research aims, scholarly literature, expert knowledge, and discussions among the research team about saturation’s prevalence and use in qualitative implementation literature (see Additional File 1 for codes and corresponding descriptions). To strengthen validity and allow for discussion and reconciliation of differing interpretations, two coders were responsible for each coding category. All authors conducted the analysis, each qualitatively trained health services researchers with PhDs in sociology; two work explicitly on implementation research. We used Google Sheets to organize and facilitate analysis.

Results

The term saturat* was invoked in 207 papers across IS (n = 73), BMC-HSR (n = 111), and IRP (n = 23). Given our primary interest in saturation’s actual deployment in applied implementation research environments, rather than hypothetical, anticipatory, theoretical, or instructive applications of the concept, articles were ineligible if the title or article type indicated a study protocol, method/methodology article, debate piece, review article, or technical advice (n = 49). The remaining 158 articles were subject to full article review, resulting in the removal of 12 articles due to irrelevant use of the saturation concept (n = 5), ineligible article type (n = 1), or the term only appearing in the bibliography (n = 6). The final dataset comprised 146 articles (Fig. 1). This analysis presents findings on how saturation is defined, utilized, determined, and justified in qualitative implementation research. Through a dialogue between our findings and the wider qualitative methodological literature, we develop a practical approach that reconceptualizes saturation one point along a larger continuum of analytic completeness, providing implementation researchers with a transparent and pragmatic approach for evaluating analytic sufficiency and, in turn, sample adequacy.

Fig. 1.

Study flow diagram

Frequency of use & article section where saturation appears

On average, saturat* was invoked two times per article (range 1 to 9). The term was predominantly found in the article’s Methods section, which was the case for 127 (87%) of the articles. Here, the term was often used to quickly validate the study’s methods with little additional discussion provided. Twenty-eight articles (19%) referenced saturation in the Discussion section, often when addressing “strengths and limitations;” here, saturation attainment was almost exclusively described as a study strength or justification for a small or non-diverse sample (e.g., although sample size was small, saturation was achieved).

Terms used

Across all articles, 15 different saturation-related terms were used (Table 1). “Data saturation” was most common, used in 73 (50%) of the articles, followed by “saturation” (without a modifier), used in 58 (40%) of the articles; taken together, these two terms appeared in 90% (n = 131) of the articles. One third of the articles (n = 49) used two or more terms, often not in meaningfully distinct ways, sometimes interchangeably. It was generally not clear why one term was chosen over another.

Table 1.

Summary of saturation terms used

Term Used	Frequency (number of articles that used the term)
Data saturation*	73
Saturation	58
Thematic saturation†	35
Code saturation	7
Theoretical saturation‡	6
Saturation point§	4
Meaning saturation	3
Information saturation/saturation of information	2
Participant saturation	2
Construct saturation	1
Category saturation	1
Saturation method	1
Content saturation	1
Topic saturation	1
Saturation grid	1

^*Alongside “data saturation,” one article also mentioned “data analysis saturation,” another mentioned “saturated with explanatory data,” another mentioned “saturating the data,” and two others mentioned “saturation of data.” In all of these cases, “data saturation” was also used.

^†Alongside “thematic saturation,” one article also mentioned “themes became saturated.” Three articles used the term “theme saturation” but only one of these appeared alongside the term “thematic saturation.” Thus, two articles used the term “theme saturation” and not “thematic saturation;” however, these were included in the “thematic saturation” count.

^‡Includes one article that used “theoretical data saturation.”

^§Includes one article that used “saturation point criterion.”

Definition

Ninety-one articles (62%) provided a definition of saturation. The most common definition was “no new themes,” cited in 34 of the articles, followed by “no new information” (n = 16) and “no new ideas” (n = 8). This notion of saturation being the point when data collection yielded no new “something” was pervasive, cited in 79 (87%) of the 91 articles with saturation definitions. However, definitional terminology varied widely, with the following 18 variations appearing at least once in the dataset: no new themes, information, beliefs, categories, codes, concepts, conceptual insights, data, elements, ideas, input, insights, issues, material, narratives, opinions, patterns. Other cited definitions of saturation included verbiage such as “an adequate sample size [that] allows sufficiently answering the research questions and includes a range of opinions” and “when further data collection would produce diminishing returns to the richness and completeness of our analysis.” Some definitions simply referenced a citation. Overall, 29 unique definitions for the saturation concept were identified.

Eighteen articles (12%) referenced two or more definitions, often used interchangeably, even when terms were not identical (e.g., no new categories AND no new themes, or no new codes AND no new themes). Some articles conflated concepts and/or listed a series of concepts (e.g., recruitment continued until “no new data or no new patterns or themes emerged.”). Fifty-five articles (38%) did not provide a definition of saturation. Fourteen articles (9.6%) did not include a definition of or citation about saturation.

A priori saturation claims

We examined whether articles specified a threshold for meeting saturation prior to data collection and analysis. Most often, this was defined as the number of interviews needed to reach saturation. Twenty-five articles (17%) contained a priori claims. This ranged from 6 to 30 interviews, with most predicting saturation would occur by the twelfth interview. The majority of these articles cited a study by Guest et al. [16] that suggests twelve interviews are sufficient for reaching saturation.

Data collection methods

Nearly all articles (96%; n = 140) relied on interviews (n = 133) and/or focus groups (n = 26). Overwhelmingly, articles were based on interview studies; 93 articles (64%) reported only using interviews and no other data collection method. Six articles reported only using focus groups. Other qualitative methods represented in the dataset, often in conjunction with interviews, included observations (n = 11) and document review or textual analysis (n = 6). Thirteen articles (9%) employed surveys or questionnaires. Only six articles (4%) reported results based exclusively on a method or methods other than interviews or focus groups.

Number of interviews conducted

For studies that included interviews in their study design (whether single or multi-method), the average interview sample size was 30 participants (median: 20; mode:15; min: 1; max: 320). Sixteen articles (11%) were based on 10 or fewer interviews and 55 articles (38%) were based on 11–20 interviews. Given only three articles explicitly claimed to have NOT reached saturation, these findings illustrate that nearly half of the studies in the dataset reported saturation with 20 interviews or less (Fig. 2).

Fig. 2.

Summary of interview sample sizes

Works cited

We examined the references authors used to support saturation claims. There were 158 total instances in which a citation was provided for the saturation concept, with 46 unique sources. The most commonly cited reference, accounting for almost a quarter of all citations (23%; n = 36), was Guest et al. [16]; in 31 instances, this article was the only reference used to explain and justify the authors’ use of saturation. Two other frequently cited articles were Francis et al. [24] with 22 citations, and Saunders et al. [22] with 20 citations. These three sources accounted for nearly half (49%; n = 78) of all citations in the dataset. Five additional sources were cited anywhere from 5 to 7 times each. Twenty-nine sources appeared only once in the dataset. Twenty articles included no citation in reference to the saturation concept (Fig. 3). Only two articles cited grounded theory literature when discussing saturation.

Fig. 3.

Summary of saturation citation patterns

Discussion

Synthesizing these findings, we shed light on saturation’s use in implementation literature. While different terms were deployed, “data saturation” or simply “saturation” were overwhelmingly used. Saturation was often given cursory treatment, appearing only once or twice in most articles, suggesting limited discussion. At times, the concept was left undefined. When a definition was provided, saturation was overwhelmingly understood to describe the point when data collection could stop because the analysis ceased to yield additional information; however, the verbiage around the type of information reportedly saturated varied widely (e.g., themes, codes, beliefs, patterns, material, and narratives). These words were sometimes used interchangeably, despite representing distinct concepts in qualitative research. Little to no discussion of saturation leading to theory generation or concept development was advanced.

While saturation’s dual-purpose of justifying a sample and demonstrating analytic completeness are equally important and necessarily interrelated, in this concept synthesis, saturation was discussed more frequently in the context of sample size, with less attention devoted to analytic completeness. Mostly, saturation was used in the Methods section to justify sampling strategy and/or analytic processes. Occasionally, saturation was mentioned in the Discussion, with authors reporting having reached saturation as a study strength. Yet, rarely did authors discuss how analytic completeness was determined; in general, readers were expected to take researchers’ saturation claims at face value. In this regard, implementation research is not unique, with others noting a tendency to omit details pertaining to saturation’s determination [17, 26, 36]. Saturation was discussed almost entirely in the context of interview-based or focus group studies, versus other qualitative methods, perhaps further highlighting the field’s reliance on saturation as justification for sample size, particularly the pragmatic task of figuring out how many people to interview and for whom to budget. A priori claims about saturation were not unusual, most asserting twelve interviews being sufficient and citing Guest et al.’s study as rationale.

The larger structure of the implementation research field may help explain these patterns of saturation usage in the literature [37]. Given the field’s reliance on mostly soft funding, the applied nature of the research, and roots in more positivist quantitative methodologies [38, 39], heightened attention on enumerating sample size estimates during and after research planning is not surprising. To assess feasibility, grant applications require a description of the proposed qualitative work, more easily expressed through sample size and likely to be more familiar to quantitatively-oriented reviewers. Although nearly all studies claimed saturation, the limited discussion around how analytic completeness was established makes these saturation claims difficult to assess. Journal word count limits, often premised on the presentation of quantitative results in charts and graphs rather than words, undoubtedly constrain authors’ ability to offer more details. Yet, experienced qualitative methodologists recognize saturation is a high bar to achieve [5, 17, 21, 40]; nonetheless, this concept synthesis suggests qualitative implementation researchers report to consistently achieve it despite relatively small sample sizes, quick turnaround times, and research environments where resources (e.g., funding, time, staff) are constrained [41] and demand for expertise frequently exceeds availability (a product of the field’s rapid growth) [42].

We suspect researchers may claim saturation, not only due to larger structural realities of the implementation research environment, but also due to a lack of options. Indeed, without compelling alternatives to explain the adequacy and rigor of their qualitative sample and analysis, practitioners may “cite methodological classics in the absence of better guidance” [43]. Concepts like information power [18], conceptual depth [21], and sufficiency [29, 30] have helpfully moved saturation conversations beyond narrow debates about sample size to more meaningfully focus on sample adequacy—that is, the quality (not just quantity) of data collected. We advocate a similar shift in focus, building upon work of other qualitative experts. High quality data comes from a sample (e.g., selected events, cases, participants, documents) that is “information-rich” [44, 45] or “dense” [18], meaning the sample holds specific knowledge or relevancy to study aims. A good qualitative dataset (e.g., fieldnotes, interview transcripts) should have sufficient breadth [29] or range [21], that is, multiple instances or patterns that support analytic findings. Good qualitative data should also have adequate depth [29, 46], attained from interviews with strong dialogue [18] or fieldnotes containing thick descriptions [47]. The dataset should exhibit a high degree of exposure: “a depth of contact with the social world or its people” [48]. Exposure is achieved not by increasing sample size (e.g., more interviews or field sites), but rather by spending more time talking to people or observing social contexts; here, even small sample sizes can result in deep exposure. Although enumerative means of achieving exposure is emphasized (i.e., time spent), a depth of contact can also be achieved through non-enumerative means, including skillful interview probes, astute observational training, scrupulous documentation techniques, and deeply cultivated trust and rapport with participants.

Throughout these discussions, authors underscore the close relationship between sample adequacy and research aims, yet fewer focus on the relationship between research aims and analytic completeness. For example, although information power includes “analytic strategy” (case vs. cross-case) as one of its five dimensions for sample appraisal, the model does not evaluate analytic completeness. Nelson’s five criteria for establishing “sufficient conceptual depth” offers assessment of analytic completeness but is limited to grounded theory studies [21].

Distinct from sample adequacy, analytic completeness considers the degree of analytic “complexity” [21], that is, a demonstrably rich analytic network of conceptual and thematic connections. The degree of analytic complexity needed for a complete analysis is determined, in part, by study aims: some research questions require more complex analyses. As such, analytic completeness can reflect varying degrees of analytic sophistication, depending on the study. Notably, qualitative analyses with limited complexity can still be rigorous (i.e., transparent, systematic, credible, reflexive) and complete, an important distinction for applied research environments with constrained resources for carrying out elaborate analyses.

Our concept synthesis highlights how qualitative implementation literature exhibits ongoing claims of “reaching saturation,” heightened attention on “ideal” sample sizes (vs. analytic completeness), particularly prefixed, a priori sizes deemed sufficient for saturation (demonstrated by the high reliance on Guest et al.). Yet, the notion that a single threshold can signal the completion of data collection and analysis (whether that’s saturation, sufficiency, information power, or conceptual depth), overlooks the reality of how these processes unfold. There is no absolute, fixed tipping point at which completeness has been achieved [18, 28], an insight captured by saturation’s characterization as “an unfortunate metaphor” [30]. Rather, there are degrees of completeness, and always opportunities to engage in more discovery. The aforementioned concepts wrestle with these challenges but may still leave researchers with the same problem they confront when relying on saturation to determine an adequate sample—that it really depends on the degree of analytic completeness needed to achieve project goals. This becomes particularly relevant for qualitative implementation researchers who must strike a balance between upholding methodological rigor while producing meaningful results on short timelines with limited resources. To juggle these competing demands, implementation researchers often attempt to “fit” [41] a qualitative project’s needs, accounting for available personnel and budget, data collection expectations, and feasibility of analytic approach in a given timeframe. At this point, expectations around analytic completeness for achieving project goals should also be considered. Not all qualitative implementation projects may need—or are much less able—to reach saturation to produce impactful results. Often, implementation research goals are pragmatic and local, focused on implementation success rather than theory/concept development. For example, from our concept synthesis, Schilling et al. explicitly sought not to reach saturation but rather inform “in a pragmatic manner” the selection of implementation strategies [49]. Here, a saturated analysis, extensive sampling, and/or new theory/concept development was deemed unnecessary.

Failure to critically and honestly engage the concept of saturation runs several risks. First, without more conceptual tools at their disposal, qualitative implementation researchers lack alternative mechanisms that may more accurately reflect their research process. Further, implementation researchers may feel pressured to claim saturation when a different concept would more appropriately describe their qualitative approach and characterize the adequacy of their sample and analysis. Transparency, reflexivity, and credibility are hallmarks of rigorous qualitative research [4, 50], but demonstrating these principles requires an alignment between available methodological concepts and the practice of real-world research. Existing concepts for explaining and justifying data adequacy may not be sufficient for the realities experienced by implementation researchers.

Saturation revisited

Based on our concept synthesis, and for the reasons outlined above, we believe implementation researchers would benefit from an expanded conceptual toolkit when thinking about sample adequacy and analytic completeness.

A continuum model—rather than tipping point model—which posits a range of ways to both pre-assess sample adequacy and evaluate data completeness—provides more options for qualitative researchers working in applied (and often resource-constrained) implementation settings. Specifically, a continuum model anchored by three benchmarks—summation, saliency, and saturation—offers a flexible, yet more precise approach to designing qualitative implementation studies and reporting on research processes and analytic findings. We call this the 3S Continuum, and suggest that core cognitive processes, each requiring increasingly complex and comprehensive thought [51] and deeper data engagement, map onto continuum benchmarks (Fig. 4).

Fig. 4.

3S Continuum

Saturation is at one end of the 3S Continuum, representing the point at which conceptual categories and properties—and the relationships among them—have been fleshed out to the point that theory building or novel concept generation is possible; such a process tends to rely on some element of inductive analysis and produces generalizable or transferrable knowledge explanatory in nature. To carry out this analysis, practitioners must engage in knowledge creation involving cognitive tasks of designing, constructing, generating, and building. This definition of saturation preserves Glaser and Strauss’s original conceptualization, and builds upon it to include non-grounded theory approaches, which may be ill-fitted for implementation study features, such as coding by teams and mixed-methods analysis [43]. Thus, beyond theory development, a saturated analysis can also generate a “robust conceptual model” poised to answer explanatory questions about “how” and “why,” as advocated by “pragmatic saturation” [28]. Nelson’s criteria for conceptual depth also aligns with this end of the 3S Continuum [21].

At the 3S Continuum’s opposite end, summation signals a concise overview of main analytic points, focused on key information with limited additional details. Summation studies typically do not aim to establish relationships between conceptual categories; analytic points tend to serve as stand-alone findings, and attempts to connect findings to generate larger theoretical or conceptual claims are rarely pursued. Summation frequently (but not necessarily) employs deductive analysis, relies heavily on pre-established frameworks, and findings tend to be descriptive accounts of “what” questions. Cognitive processes that characterize summation studies include identifying, reporting, classifying, defining, summarizing, and describing.

Saliency occupies the midpoint of the 3S Continuum, comprised of moderated elements, to varying degrees, of the Continuum’s two endpoints. In saliency studies, some relationships may be established between analytic themes or main points, but a comprehensive theory is typically not proposed, although some of the knowledge gleaned could be transferred to other similar contexts. Findings can be descriptive or, based on the extent of the analysis, conclusions may extend beyond description to include an evaluative component. For example, perhaps an analysis identifies and describes implementation barriers (i.e., summation), but also produces an assessment of those barriers (perhaps by situating them in the social context in which they were identified) that results in proposed solutions; the cognitive tasks associated with this latter analysis characterize saliency. Alternatively, findings might be exclusively descriptive but highly elaborative and detailed; for example, implementation barriers might be meticulously described, as well as compared, contrasted, and categorized in some organizational schema. Cognitive processes that typify saliency studies include solving, differentiating, applying, testing, and evaluating.

3S Continuum: methodological precedent, rationale, and utilization

The 3S Continuum is grounded in epistemological assumptions emerging out of the long and rich tradition of qualitative social science methodologies, where researchers have developed techniques that capture the dynamism, contingency, and complexity of the social world. These well-established qualitative approaches are rooted in logical rather than statistical inference [52], in which sampling procedures aim not for statistical representativeness, but rather occur sequentially until a (more) complete understanding of a social phenomenon is obtained. These approaches align with a continuum (vs. tipping point) sensibility, emphasizing data quality as a means to understanding social phenomena, while saying relatively little about quantity or sample size. Consider the technique of thick description, which highlights the importance of documenting, in rich detail, the social context in which human acts take place as a way to achieve sufficient understanding [47]. Or the concept of verstehen, tapping into the excellence of collected data, rather than amount, as a way to understand the meaning of social actions from an insider’s point-of-view [53]. The aforementioned concept of exposure also refers to data excellence, exhibited by a depth of contact with the social world or people under investigation [48]. There is no “tipping point” for determining when thick description, verstehen, or exposure has been reached. Rather, these qualitative concepts are understood as continuums, each captured to varying degrees. The same is true of saturation.

While we advance definitions of and distinctions between summation, saliency, and saturation, rarely will they appear in these pure forms in the reality of an actual study. Summation, saliency, and saturation each serve as an ideal type, a social scientific heuristic device that is an intentional exaggeration and simplification [53, 54]. As ideal types, summation, saliency, and saturation serve as yardsticks for determining the degree of analytic completeness necessary to achieve project goals and, in turn, an adequate sample. However, in reality, implementation studies will undoubtedly reside at various “in-between” points on the 3S Continuum, exhibiting varying elements of each of the “S’s”.

To put the 3S Continuum into practice, we propose implementation researchers adopt a backwards design approach to study planning. This involves first specifying the project goals and parameters outlined in Table 2, then working backwards to make an informed sample estimation using summation, saliency, and saturation as conceptual benchmarks. Pragmatic parameters of time, budget, and personnel should not drive core study design decisions, such as choice of method or hypothesis generation, as these elements should still be motivated by the scientific method’s forward design logic. Rather, pragmatic considerations—which also entail attention to dissemination product type including those extending beyond scientific publications (e.g., concise reports, implementation toolkits, playbooks, policy briefs)—have heightened significance in the implementation research environment [41]. The 3S Continuum can encourage investigators to reflect systematically and transparently, accounting for these pragmatic considerations when assessing elements like scope, exposure, or degree of analytic completeness. A study that aims to reach saturation may be impractical and/or unnecessary given project goals. Some parameters may change as research gets underway: perhaps additional funding, personnel, or expertise is secured (or eliminated), recruitment proves particularly challenging, or maybe an evaluation project expands to a theory generation project as findings emerge and resources shift. In these instances, researchers can adjust their application of the 3S Continuum accordingly, which may prove especially helpful for maintaining project integrity and methodological rigor in instances of project scale-back.

Table 2.

3S Continuum with project goals and parameters

Project Goals and Parameters	Summation	Saliency	Saturation
Budget and Timeline for Project Completion*	Limited budget and/or short timeline for project completion	Moderate budget and/or moderate timeline for project completion	Considerable budget and protracted timeline for project completion
Personnel Available	Limited personnel available with limited percent effort and/or limited qualitative expertise	Some personnel available with some percent effort and/or some qualitative expertise	Personnel available with dedicated percent, qualitative expertise, and/or subject area knowledge
Depth of Exposure†	Compact exposure period (i.e., limited contact hours) characterized by a dataset with limited breadth, depth, and/or rapport with participants	Moderate exposure period (i.e., middling contact hours); characterized by a dataset with moderate breadth, depth, and/or rapport with participants	Extensive exposure period (i.e., protracted contact hours) characterized by a dataset with extensive breadth, depth, and/or rapport with participants
Analytic Approach	Deductive; abductive; rapid analysis; framework-driven	Deductive; abductive; inductive; rapid analysis; framework-driven, thematic analysis	Abductive; inductive; thematic analysis; grounded theory
Project Scope	Research question is narrow and aims are limited; project is focused on local context	Research question and/or aims are targeted yet contain some complexity; project is focused on local context with some applicability across multiple contexts	Research question and/or aims are expansive, ambitious, comprehensive; project is focused on multiple sites, diverse contexts, or dynamic processes across health care settings
Contributions to Knowledge	Descriptive; local quality improvement; problem identification (e.g., barriers and facilitators)	Evaluative; quality improvement; problem identification; problem solving; assessment; concept/theory testing	Explanatory; theory building; novel concept generation; generalizable or transferable; mechanistic identification; process elucidations
Dissemination Product (examples)	Operational report; policy brief; community correspondence‡; real-time or rapid feedback report; journal article	White/gray paper; community correspondence; journal article; book chapter; book	Journal article; book chapter; book

^*Timeline for project completion should include time for study design, IRB approval, recruitment, data collection, analysis, and write-up of findings.

^†Exposure refers to the depth of direct contact with the people or social world under investigation. Distinct from sample size, exposure is defined as the number of hours researchers spend interviewing individuals or conducting observations. Depth of contact can also be achieved through non-enumerative means (e.g., skillful interview probes, astute observational training and documentation techniques, and deeply cultivated trust and rapport with participants). Summation, saliency, and saturation studies can all be sufficiently exposed; however, greater exposure is typically required as one moves across the 3S Continuum.

^‡Examples of community correspondence can include short reports or briefs, presentations, toolkits, infographics, and websites.

Table 2 illustrates common study characteristics that typically align with idealized conceptualizations of 3S benchmarks but should not be read as an overly rigid, prescriptive map to follow. Instead, it should be interpreted with flexibility. Project parameters can cut across Table 2 columns. For example, a highly targeted study narrow in scope may have access to ample resources; here, saturation might be a plausible pursuit if the research team intends to engage in a deep inductive analysis and explanatory knowledge production. In this scenario, a protracted exposure period (and possibly, but not necessarily, larger sample size) would be required. However, if the intended dissemination product is a brief operational report and/or researchers are operating on a short timeline, saturation would likely be imprudent, unfeasible, and unnecessary. Here, summation or saliency could suffice, and resources could be more efficiently (re)allocated. By mapping project goals and parameters alongside 3S Continuum characteristics, implementation researchers can engage in more systematic, reflexive, and transparent assessments of qualitative sample adequacy and analytic completeness, thereby helping teams scope projects, manage resources, and enhance research rigor and validity.

There is methodological precedent for tethering qualitative sample adequacy to the pragmatic task of delineating project goals and parameters, including “the uses to which [the data] will be put, the particular method and purposeful sampling strategy employed, and the research product intended” [55]. There is also precedent for aligning analytic approaches to project goals and parameters, including pragmatic considerations that characterize implementation research [41].

The 3S Continuum is not beholden to any singular analytic approach and can be used with inductive, abductive, and deductive approaches (e.g., rapid, thematic, grounded theory, etc.). The Continuum is also compatible with different implementation theories, models, and frameworks, which include descriptive, evaluative, and explanatory contributions to knowledge production [56]. Importantly, summation, saliency, and saturation—and all points on the Continuum—have the potential to rigorously signify an adequate sample and analytic completeness, depending on an implementation project’s goals and parameters. We encourage Continuum users and evaluators (including funders and peer reviewers) to resist the urge of assigning “better” or “worse” value or rigor to different points on the Continuum: a project claiming to have reached summation can be just as good, impactful, and rigorous as a project claiming to have reached saturation (especially if findings can more quickly translate into positive changes in care delivery). Rather, these concepts should be viewed as helpful tools for implementation researchers to pragmatically, transparently, and systematically reflect on qualitative sample adequacy and analytic sufficiency within the context of implementation’s unique research environment.

Enhancing qualitative rigor in implementation research

The 3S Continuum can enhance methodological rigor, transparency, and credibility by offering new concepts to demonstrate sample adequacy and analytic completeness while taking resources and other project parameters into account. The 3S Continuum’s range of options may encourage more intentional and responsive project planning and reporting, rather than defaulting to a singular and perhaps inappropriate measure. Providing investigators with more nuanced language for communicating research processes and decision making in study write-ups—including to audiences with different (and perhaps non-qualitative) expertise—encourages methodological rigor, and helps readers evaluate research findings. The 3S Continuum is an attempt to advance implementation research and we invite others to build upon/edit/refine this effort, all in pursuit of enhancing qualitative rigor while recognizing the uniqueness of the implementation research environment.

The landscape of implementation research is changing rapidly, and methodological innovations like the 3S Continuum can help to adeptly position the field to meet future challenges. For example, the introduction of artificial intelligence (AI) approaches may allow for quick digestion of large amounts of textual “qualitative” data. But qualitative researchers have already cautioned against too much reliance on AI for analysis, particularly when context and richness are important [57–61]. The 3S Continuum might be helpful for navigating this new environment, where AI may optimize the pace of qualitative analysis for implementation studies. The 3S Continuum helps make the case that, at least in its current capabilities, AI might be a tool to help execute well-defined, deductive summation studies, albeit under the supervision of an experienced and trained qualitative researcher. However, AI is likely unable to carry out saturation or even saliency studies, which rely on deeper engagement with contextualized data interpretations. As the field evolves, we anticipate the Continuum may evolve as well. For instance, depending on AI’s development, there may be opportunity to incorporate human versus artificial intelligence tools as other criteria for delineating summation, saliency, and saturation studies.

This paper has several limitations. Our concept synthesis, though intentionally exploratory, focused on only three journals (albeit important ones to the field) with our search ending in 2022, and thus may not fully capture saturation’s usage in qualitative implementation literature. Expanding the search to include other article types could add relevant information about saturation’s deployment. Some of our arguments may rest on incorrect assumptions about how implementation scientists and practitioners employ saturation and carry out their qualitative inquiries. Because we relied on authors’ reports of saturation, we are unable to verify saturation attainment or lack thereof in study reports. Additionally, some concepts we draw upon in the discussion of the 3S Continuum, while integral to qualitative, interpretive social science methodology, may have less relevance to qualitative methods with different epistemological assumptions. Nevertheless, we still believe the field can benefit from additional attention paid to conceptual elucidation of saturation and attempts to align the realities of the practice environment with methods and concepts at our disposal.

Conclusion

We understand saturation’s appeal and utility. But saturation is difficult to fully realize in the context of an actual implementation study. Building on existing concepts and considering the field’s pragmatic needs, the 3S Continuum offers a practical approach for assessing qualitative sample adequacy and analytic completeness, providing investigators with a tool with which to systematically reflect on and transparently communicate research decisions. We hope the 3S Continuum will give experienced and novice implementation researchers alike a more detailed way to think about designing and carrying out qualitative data collection and analysis. We consider this a starting point for a conversation about sample adequacy and analytic completeness in qualitative implementation research and encourage more debate and discussion about the ideas we have presented here.

Authors’ contributions

VC conceptualized and designed the project. VC and CG were the main developers of the Continuum framework. VC and CG drafted and revised the manuscript, to which all authors provided critical feedback for important intellectual content. VC, MS, and CG prepared and cleaned the dataset, conducted the analysis, and contributed to the final interpretation of results. All authors provided content expertise. All authors read and approved the final manuscript.

Funding

Not applicable.

Data availability

The dataset generated and analyzed during the current study is available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.