Shortening the SWAL-QOL: Is it possible to create a valid short form?

Abstract

Purpose:

Dysphagia is highly prevalent in movement disorders resulting in marked decrements to quality of life. The Swallowing-related Quality of Life Questionnaire (SWAL-QOL) provides critical information about the impact of dysphagia on quality of life (QOL) but is long and cumbersome. We examined whether the SWAL-QOL could be shortened, while maintaining the properties of the original scale.

Methods:

We used a combinatoric approach to reduce the 30 items that inform the QOL domains to 10 items and the 14 items that inform the symptom-profile to five items in Parkinson’s disease (PD). We tested the concurrent validity of the short-form in Progressive Supranuclear Palsy (PSP) and Cerebellar Ataxia (CA), and over time.

Results:

Our sample included individuals with PD (n=304), PSP (n=33), and CA (n=26) (age 65.7 [27-89]; 34% female). Correlations between the candidate short forms and the original QOL items ranged from 0.95 to 0.98 and for the symptom-profile from 0.91 to 0.95. Cronbach’s alpha for the final 10-item short-form was 0.85 and for the symptom-profile was 0.74. Correlations between the short-form and original QOL domains were 0.98 in PSP and 0.98 in CA, and between the short and original symptom-profile were 0.98 in PSP and 0.96 in CA. Correlations between the short and original QOL items over time was 0.97 in PD, 0.99 in PSP, and 0.98 in CA and for the symptom-profile was 0.94 in PD, 0.95 in PSP, and 0.96 in CA.

Conclusions:

We identified a 15-item SWAL-QOL with strong concurrent validity in movement disorder populations.

Background

Swallowing disorders (dysphagia) are highly prevalent in neurogenic movement disorders including Parkinson’s disease (PD), atypical parkinsonism such as Progressive Supranuclear Palsy (PSP), and spinocerebellar ataxia (SCA).^1–3 These swallowing disorders can result in malnutrition and dehydration⁴ and contribute to the development of aspiration pneumonia – a leading cause of death in these populations.⁵ In addition to the devastating health consequences, swallowing disorders also have a marked impact on quality of life, resulting in anxiety about choking, embarrassment when eating with others, the need for extra vigilance when eating/drinking, reduced ability to participate in family meals, and feelings of isolation.^6–8

Easy-to-use patient-reported outcome measures (PROMs) are essential to obtain standardized information about patient experiences.⁹ One of the most widely used PROMs to measure swallowing-related quality of life is the Swallowing-related Quality of Life survey (SWAL-QOL).^10–12 The SWAL-QOL was developed in three phases: (1) focus groups with patients (n=52) and caregivers (n=15) to identify constructs of interest,¹⁰ (2) a preliminary scale of 199 items disseminated to patients with dysphagia and reduced to 93 items via item response theory,¹¹ and (3) field testing with another 386 individuals with dysphagia (12.7% with neurogenic disease). This process resulted in the validated 44-item SWAL-QOL.¹² The specific items from the original SWAL-QOL as seen in Mchorney et al. (2002) Table 4, are included in Supplemental Table 1.¹²

Studies consistently show reduced SWAL-QOL scores in people with PD and atypical Parkinsonism,^6,13 with mean scores ranging between 60-80, as compared to healthy adults who typically score above 90.¹³ A recent study highlighted concordance between qualitative report and SWAL-QOL scores in individuals with PD, providing further support for the SWAL-QOL as a PROM in this population.⁶ Although the SWAL-QOL is a valuable and widely used tool, it is long and cumbersome, with estimates suggesting that it takes 14 minutes to complete¹² and clinical anecdote suggesting it may take even longer for patients with movement disorders. Long questionnaires result in increased respondent burden, increased cost and time in clinic visits, and may reduce data integrity.¹⁴ Shortened scales that yield scores that align with the original full-length versions may reduce patient burden and enhance utility in research and clinical practice.¹⁴

The SWAL-QOL was developed using advanced statistical methods, including item response theory and factor analytic methods, whereby several items per domain were included to increase precision of the estimated scores. Thus, the SWAL-QOL was specifically designed to include several items per domain, even if those items were repetitive or did not add much additional information to an overall score estimate. Although the SWAL-QOL was designed in this way,^10–12 it is generally scored and reported using an average of domain scores, rescaled to a percent out of 100, where each domain score is the average of the item scores that comprise that domain. The total SWAL-QOL score is based on the average of domain scores, where each item on the SWAL-QOL contributes equally to the overall summary score. It is unclear whether multiple items per domain are necessary, because person-level measurement precision (e.g., standard error or confidence intervals for each respondent) is not typically reported. Therefore, it is of interest to understand whether it is possible to create a shortened version of the SWAL-QOL that continues to cover each of the domains, but with fewer questions, and if so, whether there is a loss of information when a shortened form is used.

We aimed to determine whether the SWAL-QOL could be shortened for individuals with movement disorders, while maintaining the desirable properties of the original scale – assessed by correlations between their scores. Our primary aim was to identify a SWAL-QOL-short form in PD that maintained excellent concurrent validity with the original version. Specifically, we sought to reduce the SWAL-QOL, with one question representing each of the ten domains and one question representing each symptom category. Our second aim was to examine whether the shortened SWAL-QOL from PD maintained high concurrent validity in similar neurodegenerative movement disorders – PSP and cerebellar ataxia (CA), including both Multiple System Atrophy-Cerebellar and SCA. Third, we examined whether the shortened SWAL-QOL maintained strong concurrent validity with the original version using data from repeated administrations over time in all three populations (PD, PSP, and CA) as a validation sample.

Methods

Study Design & Participants

This study utilized SWAL-QOL data that were obtained as part of comprehensive swallowing evaluations in a clinical research laboratory at Teachers College, Columbia University via several IRB-approved protocols. All participants provided written informed consent. Outpatients with neurogenic movement disorders including PD, PSP, and CA who completed the SWAL-QOL between 2015 and 2025 were included. Diagnoses were made by fellowship-trained movement disorders neurologists. The SWAL-QOL was completed by-hand during an evaluation visit or online (via a secure REDCap survey link) prior to the evaluation. Participants either completed the SWAL-QOL independently or with the help of a care-partner. SWAL-QOLs with missing data were excluded from analysis. A subset of participants returned for repeat evaluations over time, resulting in multiple administrations of the SWAL-QOL. Analysis for aims 1 and 2 include data from initial SWAL-QOL completion in PD (aim 1) and in PSP and CA (aim 2). Analysis for aim 3 includes data from a subset of participants who completed the SWAL-QOL at follow-up visits in all three groups (PD, PSP, CA).

SWAL-QOL Instrument

The SWAL-QOL contains 44 items, 30 of which measure quality of life across ten domains and 14 of which measure symptom experience across oral, pharyngeal, and secretion symptoms, as well as two additional symptoms (oral residue and nasal regurgitation) that are not included in the categories but are included in the total symptom score. Each domain and symptom category is represented by ≥2 questions. A score ranging from 0–100 is computed for each domain and is averaged to yield a total score (0-100), where lower scores indicate worse QOL. Symptom scores also range from 0-100, where lower scores indicate worse symptom experience.

Statistical Analysis

Aim 1. Validation of a short form in PD.

We included the first administration of the SWAL-QOL for all participants with PD and used a combinatoric concurrent validity approach to create a shortened SWAL-QOL . Because of the content restriction of including exactly one item from each domain, there were 28,800 possible unique shortened forms. We explored all potential shortened forms to select the one that maintained the highest concurrent validity with the original SWAL-QOL by calculating the Pearson’s correlation between the score on each of the 10-item candidate shortened forms and the full-length form. We selected the shortened form with the maximal correlation and assessed internal consistency of this ‘best performing’ shortened form with Cronbach’s alpha. Similarly, we wanted to shorten the 14-item symptom scale to five items that represented the symptom categories. We tested the 42 possible 5-item shortened symptom scales, selected the one with the highest correlation with the full 14-item symptom scale, and computed Cronbach’s alpha to assess its internal consistency.

Aim 2. Validation of the short form in PSP and CA.

We tested the concurrent validity of the shortened forms developed in PD (aim 1) in related populations – PSP and CA. To do this, we examined the Pearson’s correlations between the full-length SWAL-QOL and the new shortened version (developed in aim 1) in patients with PSP and CA. We only included the first administration of the SWAL-QOL in PSP and CA.

Aim 3. Validation of the short form over time.

We validated the new shortened form (developed in aim 1) in subsequent administrations of the SWAL-QOL to a subset of the same participants. To do this, we calculated the Pearson’s correlations between the original full-length SWAL-QOL and the new shortened version (developed in aim 1) when repeatedly administered over time in all three patient groups (PD, PSP, and CA).

Results

Our sample included 304 participants with PD, 33 with PSP, and 26 with CA (age 65.7 [27-89] years; 34% female) (Table 1).

Table 1.

Demographics

Patient Characteristics	Total Sample (n=363)	PD (n=304)	PSP (n=33)	CA (n=26)
Age, years mean ±SD (range)	68.6 ± 10.1 (27 – 89)	69.3 ± 9.6 (41 – 89)	70.6 ± 7.1 (51 – 84)	57.0 ± 12.9 (27 – 74)
Sex, n (%) Female Male Unknown	122 (33.6%) 238 (65.6%) 3 (0.8%)	97 (31.9%) 204 (67.1%) 3 (1%)	12 (36.4%) 21 (63.6%) n/a	13 (50.0%) 13 (50.0%) n/a
Disease duration from symptom onset	8.9 ± 7.0 (0.02 – 38.5)	9.4 ± 6.8 (0.02 – 36.1)	4.5 ± 2.0 (1.0 – 11.4)	10.9 ± 11.5 (1.9 – 38.5)
Disease duration from diagnosis	7.1 ± 6.3 (0.02 – 34.1)	7.8 ± 6.3 (0.02 – 34.1)	2.2 ± 1.5 (0.13 – 5.9)	7.7 ± 8.3 (0.04 – 31.5)
SWAL-QOL administration, total n (%)
1	363 (100%)	304 (100%)	33 (100%)	26 (100%)
2	136 (37.5%)	116 (38.2%)	11 (33.3%)	9 (34.6%)
3	67 (18.5%)	62 (20.4%)	1 (3.0%)	4 (15.4%)
4	39 (10.7%)	37 (12.2%)	n/a	2 (7.7%)
5	11 (3.0%)	9 (3.0%)	n/a	2 (7.7%)
6	7 (1.9%)	6 (2.0%)	n/a	1 (3.8%)
7	6 (1.7%)	5 (1.6%)	n/a	1 (3.8%)
8	3 (0.8%)	2 (0.7%)	n/a	1 (3.8%)
9	2 (0.6%)	1 (0.3%)	n/a	1 (3.8%)

PD = Parkinson’s disease. PSP = Progressive Supranuclear Palsy. CA = Cerebellar Ataxia. SWAL-QOL = Swallowing-related Quality of Life Questionnaire.

Aim 1. Creating a SWAL-QOL short form for patients with PD

Pearson’s correlations between the scores from the 28,800 candidate 10-item short forms and the scores from the original 30 quality of life items ranged from 0.954 (95% CI 0.943, 0.964) to 0.982 (95% CI 0.978, 0.986) (Supplemental Figure 1). Cronbach’s alpha for the best performing (i.e., highest correlation) 10-item shortened form was 0.853 (95% CI 0.827, 0.876). Correlations between the 42 candidate 5-item shortened symptom profile forms and the full-length form ranged from 0.910 (95% CI 0.888, 0.927) to 0.951 (95% CI 0.939, 0.961). Cronbach’s alpha for the best performing 5-item shortened symptom profile was 0.743 (95% CI 0.694, 0.786). The items that achieved these maximal correlations are displayed in Table 2.

Table 2:

Items to be included in the shortened SWAL-QOL

QUALITY OF LIFE QUESTIONS	DOMAIN
1. My swallowing problem is a major distraction in my life	Burden
2. It takes me forever to eat a meal	Eating Duration
3. I don’ enjoy eating anymore	Eating Desire
4. It is difficult to find foods that I both like and can eat	Food selection
5. People have a hard time understanding me	Communication
6. I fear I may start choking when I eat food	Fear
7. Having to be so careful when I eat or drink annoys me	Mental health
8. My swallowing problem makes is hard to have a social life	Social
9. Feel weak	Fatigue
10. Have trouble falling asleep	Sleep
SYMPTOM PROFILE	CATEGORY
1. Choking when you eat food	Pharyngeal
2. Having thick saliva or phlegm	Secretions
3. Drooling	Oral
4. Food sticking in your mouth	--
5. Food or liquid coming out of your nose	--

Aim 2. Testing the short form for patients with PSP and CA

In patients with PSP, the Pearson’s correlation between the shortened 10-item QOL form developed in aim 1 and the original QOL domain score was 0.975 (95% CI 0.951, 0.988), and between the 5-item shortened symptom profile developed in aim 1 and the original symptom profile was 0.980 (95% CI 0.960, 0.990). In patients with CA, the correlation between the short form and the original QOL domain score was 0.982 (95% CI 0.961, 0.992), and between the shortened symptom profile and the original symptom profile was 0.963 (95% CI 0.912, 0.983).

Aim 3. Validation of the short form over time in PD, PSP, and CA

In PD, the correlation between the shortened and the original QOL items over time was 0.967 (95% CI 0.957, 0.974) and between the shortened and the original symptom profile over time was 0.939 (95% CI 0.921, 0.952). In PSP, the correlation between the shortened and the original QOL items over time was 0.993 (95% CI 0.973, 0.998) and between the shortened and the original symptom profiles was 0.947 (95% CI 0.804, 0.986). In CA, the correlation between the shortened and the original QOL items over time was 0.978 (95% CI 0.945, 0.991) and between the original and shortened symptom profiles was 0.960 (95% CI 0.903, 0.984).

Discussion

Findings from this study suggest that a shortened SWAL-QOL that maintains the desirable psychometric properties of the original scale is achievable and valid for individuals with movement disorders. Specifically, a shortened form where each domain is represented by one question resulted in a 10-item burden questionnaire and a 5-item symptom scale that maintains excellent concurrent validity with the original SWAL-QOL in PD, PSP, and CA. The substantial reduction in the number of questions is expected to dramatically reduce the time taken and respondent burden associated with completing the SWAL-QOL. Especially in neurodegenerative movement disorders, where patients and care partners already experience high disease and appointment burden, an easier-to-complete PROM that yields similar insight into patient experience is of high clinical significance. A shortened SWAL-QOL can directly inform tailored treatment planning and goal setting that aligns with patient needs and addresses the QOL impact of dysphagia.

However, the tradeoff between original full-length questionnaires and their shortened forms must be considered. Despite the high concurrent validity found between the shortened SWAL-QOL and the original full-length SWAL-QOL we must still consider balancing the efficiency of a shorter form versus the depth of data that asking multiple questions to represent a single construct may offer. Traditional theory has suggested the superiority of longer measures because of greater content coverage and high internal reliability.¹⁵ However, long surveys have also been criticized for redundancy, which increases boredom and irritation, reduces patient engagement and attention, and may yield poor data quality.¹⁵ While shortened scales may obviate these challenges, they may also reduce measurement precision. A single item may not adequately represent complex, multidimensional constructs, such as domain-specific quality of life.^14,16 To minimize data loss and ensure all aspects of swallowing-related QOL were included, we preserved the ten content domains of the original SWAL-QOL, each representing a specific aspect of quality of life.

Nonetheless, the question remains: Are the removed items truly redundant or do they perhaps capture relevant nuance that distinguishes subtle changes in swallowing-related QOL and may represent clinically meaningful variation? This both depends on the intended use of the questionnaire as well as poses a question to be addressed in future studies. If the SWAL-QOL is used to monitor specific symptoms, then a shortened form is not advised. However, in many cases, including in research, only the total SWAL-QOL score is used to summarize the respondent’s experiences or determine the degree of impact that swallowing has on QOL. In this use case, as long as the score on the shortened form is highly correlated with the score on the full-length form – as identified in the present study – the utility of the SWAL-QOL will not be reduced by using a shortened form.

Nonetheless, this study is data-driven and further empirical testing is a critical next step. The present analysis used a correlation-based validation approach and did not include external prospective validation or patient-centered validation methods. Examining the face validity of the shortened SWAL-QOL as well as considering additional validation approaches that include qualitative investigation such as expert review, focus groups, and cognitive testing with the target populations are important future steps to confirm the validity of the shortened SWAL-QOL.¹⁷ Cognitive interviewing, specifically, is a necessary next step to confirm that the shortened SWAL-QOL is appropriate, comprehensive, and understandable in the target population, and to establish content validity, prior to widespread clinical implementation¹⁸. Additional validation approaches and testing are also needed to inform best use cases and sensitivity to change. Because the shortened scale maintained such high concurrent validity with the original scale and has potential to markedly reduce response time and burden, we hypothesize it may be more useful in clinical practice and in larger group-level epidemiological studies, given the importance of efficiency in these contexts. However, to fully guide clinical decision making, further probing into the various aspects of swallowing-related QOL will be necessary, and similarly, a research study where swallowing-related QOL is the primary outcome would require additional in-depth analysis.

One additional limitation of the shortened SWAL-QOL is the uncertainty regarding the ideal response categories, given the lack of uniformity in the original SWAL-QOL and the potential impact of wording differences on responses. While all questions have a 5-item Likert scale response choice in the original SWAL-QOL, the scales are variable. To reduce respondent burden and improve interpretability and ease of completion, standardizing the scales across items on the shortened SWAL-QOL will be important.¹⁹ We propose using the terminology that is used on the greatest number of SWAL-QOL domains as well as for the symptom profile, which is: “How often do you experience the following?” with response choices of (1) almost always, (2) often, (3) sometimes, (4) hardly ever, and (5) never. This would allow for one clear scale across the domain and symptom questions, reducing cognitive demands and measurement error.¹⁹ This should be validated in future studies via cognitive interview pre-testing and/or a randomized trial design.

Conclusion

We identified a 15-item SWAL-QOL short-form that maintained high concurrent validity with the original SWAL-QOL. Although additional validation testing is needed, the shortened version of the SWAL-QOL may efficiently generate data that can be used to enhance person-centered dysphagia management.

Supplementary Material

Supplemental Table 1 - Original SWAL-QOL items

Supplemental Table 1. Original SWAL-QOL Items

Supplemental Figure 1

Supplemental Figure 1: (Left) Correlations between the 10-item shortened quality of life questions and the original quality of life questions across the 28,800 possible candidate forms. (Right) Correlations between the 5-item shortened symptom questions and the original symptom questions across the 42 possible candidate forms.