PPH is deemed a considerable health threat for older adults with hypertension across Asia LTCFs, underscoring the urgent need to determine its precise prevalence, influencing factors, and potential preprandial BP management targets in this vulnerable population to enhance their well-being.

This prospective multicenter cross-sectional study was approved by the Bioethics Committee of the West China Hospital, Sichuan University (2022 [771]) and registered with the Chinese Clinical Trial Registry (ChiCTR2200065020; https://www.chictr.org.cn).

We contacted 28 LTCFs in Chengdu, Sichuan Province, China, from June 2022 to August 2024. Among them, 25 institutions agreed to participate and provided a list of older residents with hypertension. These institutions were located in the Wuhou (9), Jinniu (5), Chenghua (4), Qingyang (3), Shuangliu (2), and Jinjiang (2) districts, and served adults who were able-minded and able-bodied, as well as those with mild, moderate, severe, and complete disabilities.

Residents who met the following criteria were included: (1) informed and agreed to participate; (2) living in an LTCF for ≥3 days; (3) diagnosed with hypertension, aged ≥65 years with an SBP of ≥140 mmHg or DBP of ≥90 mmHg, or currently receiving antihypertensive medications [17]; (4) able to maintain a sitting position [18]; and (5) without severe anxiety, depression, or visual or hearing impairments that could interfere with cooperation in conducting BP measurements and other data collection procedures. Participants were excluded if they were in the acute phase or had acute exacerbation of chronic diseases.

Before PPH evaluation for a confirmed diagnosis, participants fasted overnight and were prohibited from drinking tea, smoking, or alcohol consumption until BP measurements were completed [4, 19]. Additionally, delaying or discontinuing the use of antihypertensive medications would elevate BP, which poses a significant threat to the health of participants [14]. To adhere to research ethics while considering real-world scenarios, we instructed participants to continue taking their daily medications as usual [12, 20, 21].

In the morning, well-trained researchers used validated automatic BP monitors (OMRON; HEM-7121) with suitable cuffs to measure participants' seated BPs. The BP of both upper limbs of participants was measured to determine the dominant or healthy side 30 min before breakfast [18]. Subsequently, BP measurements were obtained at the start of meals and immediately and 30, 60, 90, and 120 min after eating, with the upper limb used for each measurement [4, 18, 22]. In summary, researchers measured the BP of each participant at seven different time points. At least two BP measurements were taken immediately before breakfast, and each participant underwent nine or more BP measurements around breakfast. Additionally, we required participants to remain seated for approximately 5 min before each BP measurement and instructed them to avoid strenuous exercise during observation [18]. Breakfast served at 7:00, 7:30, and 6:30 in 16, 8, and 1 institution (s), respectively, was selected as the test meal, and its weights and consumption times were meticulously recorded.

The BP and pulse rate (PR) measured immediately before eating were taken as the preprandial BP and PR. The minimum BP within 2 h after meals was defined as the postprandial BP. PPH was defined as a decrease of ≥20 mmHg in postprandial SBP or as a preprandial SBP of ≥100 mmHg and a postprandial SBP of ≤90 mmHg [4, 6]. Additionally, regarding a decrease in postprandial DBP, participants were categorized into four groups, involving (1) normal postprandial BP (a decrease in postprandial BP of <20/10 mmHg), (2) postprandial DBP decrease (a decrease in postprandial DBP of ≥10 mmHg and a decline in postprandial SBP of <20 mmHg), (3) type I PPH (a reduction in postprandial SBP of ≥20 mmHg and a decline in postprandial DBP of <10 mmHg), and (4) type II PPH (a decrease in postprandial BP of ≥20/10 mmHg) [4].

The Chinese version of the FRAIL-NH scale (C-FRAIL-NH), which comprises seven items (fatigue, resistance, ambulation, incontinence, weight loss, nutritional approach, and help with dressing), was used to assess the frailty status of participants, with a score of ≥2 points indicating frailty (scoring range: 0-14 points) [23]. The Chinese version of the Mini-Mental State Examination (C-MMSE) was used to evaluate the cognitive function of participants by assessing cognitive domains involving orientation, memory, attention and calculation, recall, and language ability. C-MMSE scores of <18, 21, and 25 points suggest cognitive dysfunction in participants with illiteracy, 1-6 years of education, and >6 years of education, respectively (scoring range: 0-30 points) [24]. The Chinese version of the modified Barthel index (C-MBI), a 10-item questionnaire on feeding, bathing, grooming, dressing, bowel and bladder control, toilet use, transfers, walking, and stair climbing, was employed to assess participants' basic activities of daily living (BADLs). A C-MBI score of ≤60 points indicates a moderate or higher level of impairment in BADLs (scoring range: 0-100 points) [25]. The comorbidities of participants were evaluated using the Charlson Comorbidity Index (CCI), which assesses 17 prevalent chronic diseases with a total score of 33 points. Higher scores indicate a greater burden of chronic diseases [26]. Polypharmacy is defined as the administration of five or more medications, including prescription and non-prescription drugs [27]. Additionally, standardized forms were employed to collect data on basic characteristics (e.g., age, sex, BMI, current smoker [yes or no], and current alcohol consumer [yes or no]), medical history, and routine daily medication of participants, which were well documented by their institutions of residence.

IBM-SPSS-Statistics 26.0 and R 4.2.2 were utilized for the statistical analyses. Statistical descriptions of variables are expressed as means (standard deviations [SDs]), medians (interquartile ranges [IQRs]), or proportions. Student's t-test, analysis of variance, the Mann-Whitney U test, the Kruskal-Wallis H test, and chi-squared tests were used for intergroup comparisons. Generalized estimating equations (GEEs) were used to assess whether differences existed in BPs and PRs measured repeatedly across different time points.

Logistic regression modeling was conducted to identify factors related to PPH. The Box-Tidwell test, variance inflation factor/tolerance, and cook distance were used to test the assumption of linearity, assess multicollinearity, and detect influential points, respectively. Model 1 was adjusted for age, sex, BMI, C-FRAIL-NH scores, average eating rate (food weight [g]/eating time [min]), and preprandial SBP, DBP, and PR. Model 2 was further adjusted for cognitive dysfunction, heart disease, history of CVD, diabetes mellitus, peripheral arteriosclerosis, and CCI scores. Model 3 was adjusted for the factors above and the use of calcium channel blockers (CCBs), angiotensin receptor blockers, β-adrenoceptor blockers, diuretics, benzodiazepines, α-glucosidase inhibitors, metformin, and polypharmacy.

Sensitivity analyses were performed with the corresponding method involving (1) excluding participants with residence <90 days to identify residents receiving long-term care [28], (2) defining preprandial BP as the mean of the BP measured 30 min before eating and at the start of meals [29], and (3) identifying PPH as a decline of ≥30 mmHg in postprandial SBP [30, 31]. Subgroup analyses were performed using grouping criteria, including (1) a decrease in postprandial DBP [4], (2) frailty (yes or no), (3) cognitive dysfunction (yes or no), and (4) a single or combined antihypertensive regimen. Restricted cubic spline (RCS) analyses with three knots were conducted to explore the relationship between preprandial BP and the risk of PPH in participants and their subgroups (i.e., a decrease in postprandial DBP, frailty [yes or no], and cognitive dysfunction [yes or no]). The adjusted covariates were aligned with Model 3. Furthermore, the inflection points were found by examining the confidence intervals of the estimated spline coefficients in the RCS [32].