---
title: "The DePaul Symptom Questionnaire-2: A Validation Study"
authors: "Bedree H, Sunnquist M, Jason LA"
journal: "Fatigue: Biomedicine, Health & Behavior"
year: 2019
doi: "10.1080/21641846.2019.1653471"
conditions: ["ME/CFS"]
studyType: "Cross-sectional online survey, n=399, patients only"
design: observational
topics: ["diagnosis-biomarkers", "exercise-pem"]
kind: paper
reviewedOn: 2026-10-29
guideVersion: "0.10.1"
model: "claude-opus-5-5"
effort: "medium"
humanInput: moderate
read:
  - "Full text, author manuscript from PubMed Central (PMC7367506), fetched as XML through the PubMed Central OAI service and converted to plain text"
  - "Tables 1 to 5, read from the same XML; the paper has no figures"
  - "No supplementary files: none are listed in the manuscript"
  - "The DSQ-2 questionnaire itself (linked in the paper) was not opened"
  - "Cited papers were not opened; claims about them rest on the paper's own descriptions and reference titles"
  - "Added after the first draft, to check the existing post-exertional malaise items: Jason and Sunnquist 2018, The development of the DePaul Symptom Questionnaire, Frontiers in Pediatrics (PMC6232226), main text from PubMed Central XML; its questionnaire data sheets were not opened"
  - "Added after the first draft: Cotler et al. 2018, A brief questionnaire to assess post-exertional malaise, Diagnostics (PMC6165517), full text including Appendix A (the item list) and Tables 1 to 3, from PubMed Central XML; Figure 1 not viewed"
score: 4.5
verdict: "A sensible revision of a patient questionnaire, honestly described, but with no comparison group, no reference diagnosis and no repeat testing it cannot show the validity and reliability its title and conclusion claim."
confidence: "high"
importance: "moderate"
strengths:
  - "Splitting two-in-one questions shows clear differences"
  - "Better alcohol question, with a clear reason"
  - "Simple analyses, results set out in tables"
weaknesses:
  - "No comparison group or reference diagnosis"
  - "More people meeting criteria is built into the design"
  - "Symptom grouping chosen by judgement, not checked"
  - "New post-exertional malaise items not tested against others"
guideNotes:
  - "Questionnaire validation: check for controls, a reference standard and repeat testing; endorsement rates in patients alone say nothing about accuracy"
  - "A revised scoring rule that only adds ways to qualify will raise case rates by construction; that is not evidence of validity"
  - "The guide names the DSQ-2 as best for post-exertional malaise; cite the source, likely Cotler et al. 2018 (the DSQ-PEM subscale, duration item tested against multiple sclerosis and post-polio syndrome), since the DSQ-2's own validation paper does not test it"
---

## Study design

This is a cross-sectional online survey. Adults with a diagnosis of myalgic encephalomyelitis (ME) or chronic fatigue syndrome (CFS) filled in the DePaul Symptom Questionnaire-2 (DSQ-2) once. The DSQ-2 is a revised version of the DePaul Symptom Questionnaire (DSQ-1), a self-report questionnaire that the same research group developed in 2010. It asks how often and how severely each symptom was present over the past six months, each on a 0 to 4 scale. The answers are then run through scoring rules ("algorithms") that say whether a person meets each of several case definitions.

The DSQ-2 keeps every DSQ-1 item and adds new ones. The paper says 36 new items were added. They come from the ME International Consensus Criteria (ME-ICC), the Institute of Medicine (IOM) criteria, Ramsay's 1988 description of ME, and patient feedback. Two DSQ-1 items that asked about two things at once were split in two: "unable to focus vision or attention" and "losing or gaining weight without trying". The alcohol intolerance item got two follow-up questions, because people who avoid alcohol because of intolerance could not answer the original item sensibly.

The paper does three things with the data:

1. It reports, for the split and new items, the mean symptom score and the share of people at the "2/2 threshold" (frequency and severity both at least 2, meaning about half the time or more, and moderate or worse).
2. It applies both the DSQ-1 and the DSQ-2 scoring rules to the same answers and compares how many people meet each case definition.
3. It runs an exploratory factor analysis on the symptom scores and reports Cronbach's alpha (a measure of how well items in a group correlate with each other) for each factor.

There is no comparison group, no second time point, and no comparison with any outside reference such as a clinician's diagnosis. The paper is titled a validation study, but none of the three analyses tests whether the DSQ-2 measures what it is meant to measure against something outside the questionnaire. That gap shapes most of what follows.

## Sample selection

The 399 participants were drawn from a larger online study of people with various chronic illnesses (the paper's reference 14, a comparison of ME and CFS with multiple sclerosis). Recruitment was by email to national foundations and posts to support groups, research forums and social media. Only those "with a diagnosis of ME or CFS" were analysed. The paper does not say who made the diagnosis or how it was checked. With online recruitment, it was presumably self-reported.

The sample is 88.0% female and 96.7% white, with a mean age of 48.7. Of the 399, 69.1% had a college degree or higher, and 48.6% were on disability. The paper does not report illness duration, onset type, or severity on any scale. It does not say whether housebound or bedbound people took part, or in what numbers. It does not report how many people started the survey and how many finished it, or how much data was missing per item.

Almost everyone met the broadest case definition: 97.3% met the Fukuda (1994) CFS criteria, 84.5% to 85.2% the IOM criteria, and 53.9% to 66.3% the ME-ICC, depending on the scoring rules. So by the questionnaire's own reckoning, this is mostly a group with substantial symptoms. That is fine for describing how patients answer new items. It is not enough for validation, which needs people who do not have the illness or who have something that looks like it.

The paper says the larger study included people with other chronic illnesses. None of them were used as a comparison group here. Nor were healthy controls, or people with depression, fibromyalgia or other fatiguing conditions. Without them, no analysis can show whether the new items tell ME/CFS apart from anything else.

The guide asks every study to note status on postural orthostatic tachycardia syndrome or orthostatic intolerance (POTS/OI) and mast cell activation syndrome (MCAS). The paper reports neither as a diagnosis. Its new orthostatic items describe symptoms, not a diagnosis.

## Outcome measures

Everything here is self-report, from one sitting, about the past six months. That is unavoidable for a questionnaire study, and it is not a flaw in itself. The issue is what the numbers can show.

### The split items

The split of double-barrelled items works as intended, and the result is useful. On the DSQ-1, 61.2% met the threshold for "unable to focus vision/attention". On the DSQ-2, 62.4% met it for focusing attention and 34.6% for focusing vision. For weight, 45.1% met it for "losing/gaining weight without trying". When split, 40.9% met it for gaining weight and 6.5% for losing weight. These show that the old items were mostly picking up one half of each pair. That is a sound, if modest, finding. It matters because the ME-ICC and Canadian criteria count some of these symptoms in specific domains, so a person could have met a criterion through the "wrong" half of an item.

### The alcohol items

On the DSQ-1 item, 35.3% met the threshold for alcohol intolerance. On the new question, 53.1% said they had avoided alcohol in the past six months because of intolerance, and 23.7% avoided it for other reasons. The paper reads this as showing many people answered "none of the time" on the old item because they had not been drinking. That is plausible, and the revision is sensible. But the paper does not cross-tabulate the two items, so it does not show directly that the people who avoided alcohol were the ones who scored low on the old item. The discussion says "many participants skipped this question or indicated that this symptom was not present". The results report no figures on skipping.

### The new post-exertional malaise items

The three new Ramsay items on post-exertional malaise (PEM) were the most endorsed of all the new items. "Worsening of symptoms after mild physical activity" was at threshold for 90.2%, "muscle fatigue after mild physical activity" for 86.2%, and "worsening of symptoms after mild mental activity" for 74.7%. The paper reads high endorsement as a sign of the items' value.

High endorsement in a patient-only sample does not show that. An item that 90% of patients endorse tells you nothing about how well it separates patients from other people, unless you know how often other people endorse it. People with depression, deconditioning or other fatiguing illnesses may also report feeling worse after activity. The paper has no data on that. Very high endorsement also means the item cannot tell much apart within this sample.

The three new items also do not capture what the guide treats as defining PEM: a delay in onset and a prolonged recovery. They ask how often and how badly symptoms worsen after mild activity. That is part of PEM, but not the part that best separates it from ordinary post-exercise fatigue.

The questionnaire as a whole does ask about these things, through items it already had. The DSQ-2 keeps every DSQ-1 item, and the DSQ-1 includes "Next day soreness or fatigue after non-strenuous, everyday activities", which touches on delay; "If you were to become exhausted after ... outings with friends, would you recover within an hour or two after the activity ended?"; and "If you feel worse after activities, how long does this last?", with answers from under an hour to 24 hours or more. No item asks how soon after exertion symptoms begin. These item wordings come from a separate 2018 paper by the same group (Cotler et al., in Diagnostics), not from this one.

That 2018 paper tested the duration item against other illnesses. Among people reporting ME or CFS, 73.4% said worsening lasted 24 hours or more, against 8.9% of people reporting multiple sclerosis and 13.8% reporting post-polio syndrome. The item's area under the curve (a measure of how well it separates the groups, where 1 is perfect and 0.5 is chance) was 0.88. That study has its own limits: diagnoses were self-reported, recruitment was online, the comparison groups were other illnesses rather than healthy people or people with depression, and the authors were testing their own instrument. But it is evidence that the questionnaire can capture prolonged PEM, and it is the likeliest basis for the guide naming the DSQ-2 as the current best questionnaire for PEM.

The present paper does not report the duration item at all. Its "Next day soreness" item appears only as a factor loading (.78). So the questionnaire's most discriminating PEM question goes unmentioned, while three new and much less specific items are presented as the main PEM finding.

### Missing items from the tables

The paper says 36 items were added. Table 2 lists fewer than that by name, and Table 4 gives results for 23 new items. Some listed additions, such as "Since the onset of your fatigue/energy related illness" and "Have you stopped getting sick with colds or flus", have no results anywhere. "Graying or blacking out after standing" is described in the text as new and appears in Table 4, but not in Table 2. A reader cannot match the 36 items to results. The full questionnaire is online (the paper gives a link), but which items count as new is not set out.

## Bias

The authors developed the questionnaire. The senior author, Leonard Jason, leads the DePaul group that made the DSQ-1 and DSQ-2. Developers evaluating their own instrument is common in questionnaire research. It is not a financial conflict, and the paper declares none. But it is the kind of allegiance the guide warns about, and it shows in the framing: every result is read as support for the revision, including results that could not have come out otherwise (see the next section).

Selection bias from online recruitment through support groups and foundations is likely. Such samples tend to be more educated, more engaged and less severely ill than the full patient population. The 69.1% with a college degree fits that. This matters less for describing item endorsement, and more for the claim that the DSQ-2 "may prove useful in medical and research settings" across the patient population.

## Statistical and methodological issues

### More people meeting the ME-ICC is not evidence of validity

Under the DSQ-1 scoring rules, 53.9% met the ME-ICC. Under the DSQ-2 rules, 66.3% did. The paper presents this as the revision's largest effect. But the DSQ-2 rules give people more items through which to meet each domain, and the old items are all still there. So the share meeting criteria could only go up or stay the same, assuming the rules were otherwise unchanged. The paper does not set out the scoring rules, so this cannot be checked. A rise is therefore built in. Whether the extra 12.4 percentage points are people who truly meet the ME-ICC, or people who now meet it more easily, cannot be told from these data. With no controls, the paper also cannot say whether the change raises the share of non-cases who meet criteria. A looser instrument would show exactly the same pattern.

The discussion then says that the results, apart from the ME-ICC, "were consistent across DSQ-1 and DSQ-2 algorithms, suggesting that both measures reliably provide diagnostic assessments for this illness." This does not follow. Both sets of rules were applied to the same answers, from the same people, at the same time, and the DSQ-2 contains every DSQ-1 item. Agreement between them is close to certain. It says nothing about reliability, which needs repeat measurement, or about diagnostic accuracy, which needs a reference standard and people without the illness.

### The factor analysis

The exploratory factor analysis used principal axis factoring with promax rotation on the symptom scores. Two highly correlated eye items (r = .797) were excluded. The paper reports 72 symptoms in total, without saying whether that count is before or after this exclusion. The method raises several problems.

The number of factors was not chosen by a set rule. The scree plot "showed multiple points of inflection", so the authors examined "each solution" and picked "the most interpretable and parsimonious" one. No parallel analysis or similar objective method is reported. This leaves a lot of room for judgement, and the eight-factor result may not be stable.

Items were then dropped in rounds. Any item loading below 0.4 on every factor was removed and the analysis re-run, "until all symptoms loaded onto a factor". In the end 39 of 72 symptoms were kept. So 33 symptoms, nearly half, fit no factor. The paper does not list them or discuss what their exclusion means for the instrument's structure.

There was no confirmatory factor analysis and no split-sample check. With 399 people and about 70 items, there are between 5 and 6 people per item, which is on the low side for a stable exploratory solution.

Cronbach's alphas of .73 to .91 are reported as showing "strong internal consistency". But the items in each factor were picked because they correlate with each other. High alpha within factors built that way is expected, and it does not add independent support.

Some factor labels do not match their items. The "Pain" factor is made up of abdominal or stomach pain, irritable bowel and bloating, plus muscle pain at a loading of .42. It is mostly a gastrointestinal factor. The "PEM" factor includes "fatigue/extreme tiredness" (.74) and "unrefreshing sleep" (.48). So it may be measuring general fatigue severity as much as PEM.

### The PEM factor's "30% of variance"

The paper says the PEM factor explained 29.95% of the variance, compared with 4.9% and 3.8% in two earlier DSQ-1 analyses. It reads this as "further evidence of [the new items'] utility". This reading is weak. In factor analysis, the first factor extracted usually takes the largest share of variance, and the share depends heavily on how many items there are and which ones were kept. Adding several items that ask nearly the same thing (here, two new items on fatigue or worsening after mild physical activity that loaded on this factor) makes a factor bigger by redundancy. Dropping 33 other items raises every remaining factor's share. A larger share of variance shows the items overlap, not that they are useful.

### "Content validity"

The discussion says the eight-factor solution "demonstrated that the instrument has high content validity". Content validity is about whether the items cover the intended domains. It is judged by comparing items with a definition, usually with expert or patient input. A factor analysis does not show it. The paper does have a reasonable content argument, since the new items were taken from the wording of the ME-ICC, the IOM criteria and Ramsay. But that argument comes from how the items were chosen, not from the factor analysis.

### Reliability

The abstract says the study evaluated "the utility and reliability of the new and revised items". The only reliability measure is internal consistency within factors, discussed above. There is no test-retest data for the new items. The paper cites an earlier test-retest study of the DSQ-1, but that does not cover the 36 new items.

## Interpretation and spin

The title calls this "A Validation Study". It is closer to an item-development and descriptive report. It shows how a patient sample answers revised and new items, and it shows that the new scoring rules let more people meet the ME-ICC. It does not test validity against anything outside the questionnaire.

The paper's claims compared with what the data show:

- "These analyses indicate that the DSQ-2 offers a more thorough and precise understanding ME and CFS symptomology and case definition fulfillment." More thorough, yes: there are more items. More precise is not tested. Precision would need repeat measurement or comparison with a reference.
- "item revisions provided a more accurate depiction of participants' symptomatology." For the split items, this is a fair reading. For accuracy in general, there is no standard to compare against.
- "suggesting that both measures reliably provide diagnostic assessments for this illness." Not supported, as explained under statistical issues.
- "The eight-factor solution ... demonstrated that the instrument has high content validity." Not something factor analysis can show.
- "This increase likely resulted from the addition of new PEM items that loaded onto this factor ... providing further evidence of their utility." The increase in variance is better explained by redundancy and item removal.

The results section itself is mostly plain and accurate. The overreach is in the title, the abstract's conclusion and the discussion.

## What this demonstrates

It shows that:

- the two double-barrelled DSQ-1 items were mostly picking up one half each (attention more than vision, weight gain far more than weight loss);
- many patients avoid alcohol, so the old alcohol item probably undercounted intolerance;
- people with a reported ME or CFS diagnosis, recruited online, very often endorse the new items on worsening after mild activity;
- the DSQ-2 scoring rules classify more of the same people as meeting the ME-ICC than the DSQ-1 rules did;
- in this sample, the symptom items group into recognisable clusters, after nearly half the items are dropped.

It does not show:

- that the DSQ-2 tells ME/CFS apart from other conditions or from health;
- that DSQ-2 case definition classifications are accurate, or more accurate than the DSQ-1's;
- that the new items give the same answers on repeat testing;
- that the new items capture PEM as the guide defines it, with delayed onset and prolonged recovery;
- that the eight-factor structure would hold in another sample.

## Critical missing elements

- A comparison group: healthy controls, and people with other fatiguing or chronic illnesses. The paper says the larger study had people with other chronic illnesses, so this data may have existed.
- A reference standard for diagnosis, such as a clinician's assessment, to test the scoring rules.
- Test-retest data for the new items.
- How the diagnosis was established for inclusion.
- Severity, illness duration and how many housebound or bedbound people took part.
- Response and completion rates, and missing data per item.
- The scoring rules themselves, or a reference to where they are published, for both versions.
- A full list of the 36 new items and results for all of them.
- Results for the existing PEM duration and recovery items, which the paper had but does not report.
- An objective rule for the number of factors, and a confirmatory check in a separate sample.

## Recommendations

For researchers using the DSQ-2: this paper supports using the split attention, vision and weight items and the revised alcohol items, since these are clear improvements in wording. It does not support citing the DSQ-2 as "validated" for case definition fulfillment or for PEM ascertainment. If those claims are needed, look for studies that test the DSQ-2 against controls, a clinical reference or repeat measurement. For PEM, the item on how long worsening lasts has been tested against people with multiple sclerosis and post-polio syndrome in a separate study (Cotler et al., 2018); use it, not the three new items, when PEM needs to be told apart from other illnesses.

For researchers reading papers that use the DSQ-2 to classify people as meeting the ME-ICC: the DSQ-2 rules classify almost a quarter more people as meeting it than the DSQ-1 rules (66.3% against 53.9% here). Rates from the two versions are not directly comparable.

For the developers: run the DSQ-2 in a sample with controls, report sensitivity and specificity for each scoring rule against a clinical reference, and test-retest the new items. Publish the scoring rules alongside the paper.

For patients and advocates: this paper is a reasonable record of how the questionnaire was revised, and the revisions make sense. It is not evidence that the questionnaire can pick out ME/CFS from other illnesses.

## Final assessment

The paper does what a careful item-revision report should do in parts. It fixes obvious problems in the old wording, and it shows the effect of the fixes clearly. The alcohol and double-barrelled item findings are sound and useful. The analyses are simple and their results are set out in tables, so most of the reported numbers can be checked, though some items have no results shown.

But the paper's main claim, in its title and conclusion, is validation. The design cannot support that. There is no comparison group, no reference standard and no repeat measurement. Two of the headline results are built in: more people meet the ME-ICC because there are more ways to meet it, and the DSQ-1 and DSQ-2 rules agree because they share items and were applied to the same answers. Both are then read as evidence of value and reliability. The factor analysis picked its solution by judgement, dropped nearly half the items, and was not checked in a second sample. Its high alphas and large PEM factor follow from how it was built, and the claim that it shows content validity misnames what factor analysis does.

What prevents a higher score: the validation and reliability claims rest on analyses that cannot test them, and the factor analysis is too flexible to carry weight. What prevents a lower score: the descriptive findings are correct and honestly reported, the item revisions are well reasoned, and the results section itself mostly says what the data show. The overreach is in interpretation, not in the numbers.

This places it at the top of the 3–4.9 band. The design cannot separate a better instrument from a looser one, which is the central question for a validation study. The descriptive content is real, which keeps it at the upper end.

## Executive summary

This paper describes how the DePaul group revised its own symptom questionnaire for ME and CFS, creating the DSQ-2. They asked 399 adults with a reported ME or CFS diagnosis, recruited online, to fill it in once.

Some results are useful. Two old questions asked about two things at once. When split, far more people had trouble focusing attention than focusing vision, and far more were gaining weight than losing it. A new question showed that about half the sample avoids alcohol because of intolerance, so the old alcohol question probably undercounted it.

But the paper calls itself a validation study, and it does not test validity. There is no comparison group of healthy people or people with other illnesses. There is no check against a doctor's diagnosis, and no repeat testing. Without these, no analysis can show whether the questionnaire tells ME/CFS apart from anything else.

Two headline results are built into the design. More people met the strict international consensus criteria for ME with the new version (66.3% against 53.9%), but the new version simply gives more ways to meet them. The paper also says the old and new scoring agree, so both are reliable. But they were applied to the same answers and share most questions, so they had to agree. The statistical grouping of symptoms was chosen by judgement, dropped 33 of 72 symptoms, and was not checked in a second sample.

The new questions on feeling worse after mild activity were endorsed by up to 90% of patients. Without controls, that says nothing about whether they pick out post-exertional malaise. They also do not ask about the delay and slow recovery that define it. The questionnaire already had a question on how long the worsening lasts. A separate study by the same group found it separates ME/CFS from multiple sclerosis and post-polio syndrome fairly well. This paper does not report it.

The revisions are sensible and the numbers are honestly reported. The claims of validity and reliability go well beyond what the study can show.

Grade: 4.5 / 10.