Drinking liquids with dysphagia: the plot thickens (sorry)

You may find this study hard to swallow

Sometimes it seems like half the ICU revolves around aspiration: accepting patients who’ve just done it, helping others not to, considering it in still others who are deteriorating, etc.

Thickening a dysphasic patient’s oral liquids is among the measures commonly employed to prevent aspiration.

The rationale: liquids are harder to control than solids because they slip down the throat so quickly and easily. Thickening slows them down, making them easier to manipulate and eventually swallow, rather than aspirate.

Liquids may be thickened to “nectar-thick” (easily pourable) or “honey-thick” (viscous, not easily pourable).

Speech pathologists determine proper liquid consistency for an individual patient based on their bedside evaluation, often augmented by radiographic tests like a modified barium swallow.

Like innumerable other medical interventions, thickened liquids have a solid rationale in common sense and physiology, but have gone for decades without being rigorously tested.

(Spoiler: they still haven’t.)

A new clinical study has put thickened liquids under the astigmatic eye of the semi-scientific method, and was covered in the New York Times as “a therapy older adults should question.” What can we learn?

Mmmm… Nectar-Thickened Data

Makhnevich et al performed a retrospective matched cohort study of 8,916 patients with Alzheimer-type dementia admitted at 11 hospitals in New York (2017-2022) who had documented clinical suspicion of dysphagia.

Cohorts were created based on whether the patients were permitted regular (“thin”) liquids or were prescribed thickened liquids in their diets.

It was not randomized.

Stop there for a second. Might there be some substantial differences between patients already prescribed thickened liquids, and those who haven’t? Like the degree of their dysphagia, dementia, or both?

Propensity matching was used to control for this and other sources of confounding —which seems impossible for this population of patients, but here we go.

The natural language processing model they used resulted in inclusion of many patients without dysphagia (82% positive predictive value). There was no method of adjusting for stage of dementia, past history of aspiration or pneumonia.

The primary outcome was death in the hospital. Secondary outcomes included pneumonia, intubation and hospital length of stay (LOS).

• There was no significant difference in the primary outcome of in-hospital mortality (hazard ratio 0.92, favoring thick liquids but with 95% confidence intervals 0.75 to 1.14).

• Intubation occurred significantly less often with thick liquids (odds ratio 0.66 with 95% CI, 0.54-0.80),

• Pneumonia (or aspiration or choking) was more likely with thick liquids (odds ratio 1.73; 95% CI, 1.56-1.91).

Authors then deployed a technique called inverse probability of treatment weighting, which is another round of processing after propensity matching. (Patients in the analysis are assigned weights based on the inverse of their propensity score, and then re-analyzed as a new artificial population in which some “count” more based on certain observed differences.)

In this ultraprocessed data confection, mortality was 5% higher in the thick liquid group (the opposite direction of the first analysis), not reaching statistical significance. In other words: more noise.

The NYT Gets It Wrong

The New York Times reported: “Now, a large-scale study has found that liquid thickening doesn’t actually help.” Nonsense; the study didn’t show that at all.