Language-group differences in medication adherence disappeared after controlling for patient-provider language concordance

Description #

When the analysis controlled for patient–provider language concordance, the differences in medication adherence between language groups largely vanished: groups remained significantly different only for regular-insulin actual days of use, and for all other medications and outcomes there were no significant differences once concordance was controlled (Table 3). The authors conclude language concordance — not preferred language per se — accounted for most of the observed differences.

"After controlling for language concordance between patient and provider, language groups remained significantly different only for regular insulin actual days of use. For the medication ratio and maximum days of medication adherence for regular insulin, and for all other medication and all other outcomes, there were no significant differences between language groups once language concordance was controlled." (Kahler, 2022, p. 33)

"Thus, language concordance accounted for most of the above findings." (Kahler, 2022, p. 34)

Methods Context #

What? #ⓘ

The observable: per-medication Adherence Ratio, Actual Days of Medication Adherence, and Maximum Days Non-Adherent, compared between language-concordant and non-concordant clinical visits.

"mean and standard deviation values for each medication and outcome by language and concordance group were calculated, using one-way analysis of variance (Table 3)." (Kahler, 2022, p. 33)

How? #ⓘ

OLS regression models re-examined for whether language-discordant visits affected results; a per-participant concordance ratio (proportion of visits with a language-concordant provider; median 67%) was used to stratify, and means were compared above vs below the median concordance level. (Language-discordant encounters used a professional medical interpreter, live or by phone.)

"Finally, all models were examined for evidence that language discordant visits impacted the models. The proportion of visits were calculated for each participant that was provided by a language concordant provider (concordance ratio)." (Kahler, 2022, p. 32)

Who? #ⓘ

The same 59 type 2 diabetes primary-care patients (English/Spanish/Nepali preference); concordance achieved either by a language-concordant provider or a professional interpreter. English-preference encounters were concordant by design.

"All language discordant encounters were conducted with the use of a professional medical interpreter, either live or by phone." (Kahler, 2022, p. 32)

Other Notes #

The single significant residual (regular-insulin actual days) appeared only for patients above the median level of language-concordant visits.

Caveats #

• Small underpowered sample yields wide confidence intervals in the language-concordance adherence analysis The study analyzed only 59 patients, split across three language groups and further across medication subsamples (as small as n=2 in some Table 3 cells). The authors state the language-concordance analysis was underpowered, producing wide confidence intervals — so the concordance findings, while suggestive, are statistically fragile.
• Pharmacy-refill Adherence Ratio does not capture whether medication was actually taken correctly Adherence was measured indirectly from pharmacy refill data (the Adherence Ratio), a surrogate chosen because no validated direct metric was available. This captures whether medication was obtained, not whether it was actually taken correctly — a construct-validity gap. The authors also note patients may have filled prescriptions at unlisted pharmacies, undercounting fills.