Blood pressure variability associated with falls in nursing home residents

Abstract

Background

High variations in serially measured blood pressures (BPs) portend a variety of adverse clinical events including dementia, cardiovascular sequelae and frailty. In this study, systolic blood pressure variability (BPV) was examined for its association with fall frequency and time to next fall among older adults living in nursing homes.

Methods

BP values and falls over time were extracted from medical records of nursing home residents aged ≥65 years over a 10‐month period. BPV was measured as the standard deviation of 17 to 20 systolic values, and its correlation with falls and time to next fall were evaluated according to quartile values.

Results

One hundred patient charts were analyzed with nearly 2000 BP data points. All older adults had at least one fall incident. Higher BPV was related to more falls, shorter time between the first and second fall and fewer average days between falls. Subgroups of high BP and different diagnoses affected this association between BPV and falls.

Conclusions

People who fall often show a high variability in BP; as the number of falls increases, the BPV also increases. This study suggests that BPV may be marker for patients who might benefit from more aggressive application of fall reduction strategies. Geriatr Gerontol Int 2024; 24: 1315–1319.

High blood pressure variability is associated with increased risk of frequent falling. The higher the blood pressure variability, the higher the likelihood of multiple falls.

Introduction

Roughly 50% of nursing homes residents fall at least once each year, with an average of 1.5 falls per year. ¹ , ² , ³ Falls result in fractures, loss of function, hospitalization, immobilization, soft tissue damage and even death. Indeed, the most prevalent external cause of deaths within nursing homes are fall related, and this has been increasing over time. ⁴ The estimated average cost per patient for hospitalization after a fall in a nursing home is $23 723–$31 507. ⁵ , ⁶ Thus, falls in the nursing home are consequential and costly.

Several studies have examined physiological parameters that predict falls. For instance, slow gait speed, postural sway and decreased grip strength are associated with falls in older adults. ⁷ , ⁸ , ⁹ More recently, cardiovascular metrics such as heart rate variability have been studied as fall risk indicators. ¹⁰

Less well understood are predictors of frequent fall. Clinicians are all too familiar with older adults who repetitively fall despite known measures to prevent future falls. Frequent fallers may have various characteristics of frailty, cognitive impairment and sarcopenia. The role of long COVID has yet to be determined as an additional contributor to frequent falls. However, no physiological measure has yet been used as a predictor of frequent falls, especially in long‐term care.

Blood pressure variability (BPV) is an independent risk factor for cardiovascular disease, dementia, frailty, morbidity and mortality. ¹¹ , ¹² , ¹³ , ¹⁴ We now hypothesize that BPV is associated with fall frequency and time to next fall in the nursing home setting. ¹⁵ , ¹⁶ The rationale for understanding the association of BPV is multifold: its ease of access as a metric for identification of frequent fallers, the potential it has for better understanding fall mechanisms that may be based on muscle and connective tissue stiffness and autonomic nervous system dysregulation. ¹⁷

Methods

A database was created by extrapolating information from clinical medical records of 100 residents who had at least one previous fall at one of two teaching, not‐for‐profit, urban‐based nursing homes. Blood pressure (BP) values were typically recorded at least once weekly in the nursing home charts for residents; thus, we selected values from the first and third week of each month over a 10‐month period. If only one BP value was recorded in a particular month, we would carry that value forward as the next recorded value in our bimonthly survey. ¹⁸ Both nursing home and clinic‐based medical records were perused for falls as well as emergency department visits and hospitalization for falls. All residents had from 17 to 20 BP values.

BPV was measured by the standard deviation (SD) of the systolic value. Diastolic BP values were not used because systolic values are more sensitive to BPV (data not shown). ¹¹

As outcomes, we calculated the total number of falls in a 10‐month time period, the time from the first fall to the second fall and the average number of days between falls. These outcomes were also divided into quartiles. Covariates obtained from the record included resident age, sex, use of BP medication, BP stage (normal, elevated, stage 1 and stage 2, based on average systolic and diastolic values over the 10‐month time period) and diagnoses of one or more chronic conditions (hypertension; diabetes mellitus; heart, kidney, neurologic and lung disease). BP was defined as high if the resident's value was in the range of stage I or stage II hypertension. The presence of four or more aforementioned chronic conditions was defined as high chronic disease burden.

Analyses were conducted with SAS version 9.4 (SAS Institute Inc., Cary, NC). Mean BPV was compared among the quartiles for the number of falls, days from first to second fall (97 residents), and average number of days between all falls (97 residents) using one‐way analysis of variance with Tukey multiple comparison. Association of covariates with the three fall measures was tested using chi‐square statistics. Negative binomial regression was used to predict the three outcomes from BPV, and the covariates as the outcomes were all positively skewed. Subgroups defined by covariates were analyzed for specific BPV‐to‐outcome relationships. An α of 0.05 was used for the analyses.

Results

BP values approximately ≈2 weeks apart were gathered for 100 residents over 10 months. Ninety‐four percent of the residents had 20 BP recordings over the 10‐month time period, while five people had 18 and one person had 17. The average resident's age was 85.39 years (SD = 9.9), and 40% were men. All values were obtained prior to the COVID‐19 pandemic.

All residents had experienced at least one fall over the 10‐month period. Table 1 describes BPV, number of fall incidents and time to next fall. The average BPV was 18.0 (SD = 5.0) and ranged from 8.6 to 31.1 mm Hg. The average number of falls was 6.64 (SD = 4.8) and ranged from 1 to 24 per resident. The average number of days between consecutive falls was 53.8 (range, 9–266) days and between the first recorded and second fall was 57.7 (range, 0–266) days, noting that two falls happened on the same day.

Table 1.

Systolic blood pressure as blood pressure variability and falls in 10‐month period

	Mean	SD	Min	25%	Median	75%	Max
Systolic SD	17.96	5.01	8.57	14.2	17.50	20.98	31.12
Total falls	6.64	4.79	1	3	5	9	24
Average days between falls*	53.84	44.60	9	24	39	71	266
Days between first and second fall*	57.68	60.80	0	15	41	79	266

^*Only 97 observations, as three people had only one fall in the 10‐month period.

SD, standard deviation.

The BPV was compared between the quartiles of total falls in a 10‐month period (n = 100), days from first fall to second fall (n = 97) and average number of days between falls (n = 97). BPV was significantly different between total falls quartiles, specifically between the first quartile (1–3 falls, mean BPV = 15.4) and the third (6–9, mean BPV = 19.0) and fourth (10–24, mean BPV = 20.2) quartiles (F = 4.4, P = 0.006) (Fig. 1a). Residents with the highest number of falls overall had the greatest mean SD of their systolic BP (Fig. 1a).

Figure 1.

Systolic blood pressure variability (BPV) with falls. Average systolic BPV compared to (a) total falls, (b) days until the next fall and (c) the number of days between fall incidents.

When the time between falls was evaluated, the SD for systolic BPV had no significant difference between residents with high and low BPV. Average BPV ranges from 19.2 in the first quartile (0–15 days) to 17.5 in the fourth quartile (80–266 days) (F = 0.60, P = 0.615) (Fig. 1b). Similar results were found for average number of days between falls. BPV ranged from 17.4 (fourth quartile) to 18.7 (third quartile) (F = 0.38, P = 0.766) (Fig. 1c).

Covariates of age, sex, BP medication, hypertension stage, chronic conditions (diabetes mellitus; dementia; heart, kidney, neurologic and lung disease) and total number of diagnoses were tested for an association with each quartile of total falls, time to second fall and average days between falls using chi‐square statistics (Tables S1–S3). Association of covariates was minimal. High BP was associated with total falls (chi‐square = 9.9, P = 0.019). Being older was associated with fewer days between first and second fall (chi‐square = 17.5, P = 0.008). The diagnoses of dementia (chi‐square = 9.2, P = 0.027) and kidney disease (chi‐square = 7.9, P = 0.048) was associated with a reduced number of days between falls (average).

Negative binomial regression was used to test relationships between BPV and fall outcomes while adjusting for covariates. Table 2 shows the coefficients for BPV predicting falls, and all models had a good fit. For total number of falls, no covariates were significant in predicting when BPV was in the model. The coefficient for BPV was significant (β = 0.03, P = 0.007). This observation suggests that for every increase in BPV by one unit, the expected number of falls changes by 3.6%. When BPV was examined in subgroups based on sex, high BP and number of diagnoses, the percentage of falls increased by 3.9% for women (P = 0.025) for every increase in BPV. No such statistical relationship was observed in men (P = 0.132). Falls also increased by 3.7% for residents with high BP (P = 0.039) and 4.4% for having one to three diagnoses (P = 0.011).

Table 2.

Negative binomial regression predicting falls from BPV

	β	95% CI lower limit	95% CI upper limit	P‐value	Change in fall outcome, %	Deviance/DF
Total falls
BPV	0.035	0.010	0.060	0.007	3.57	1.003
Women	0.036	0.005	0.068	0.025	3.69	0.995
Men	0.031	−0.009	0.072	0.132	n.s.	1.061
BP low	0.011	−0.035	0.057	0.648	n.s.	1.000
BP high	0.036	0.002	0.070	0.039	3.67	1.036
1–3 diagnoses	0.043	0.010	0.077	0.011	4.44	1.025
4–6 diagnoses	0.015	−0.021	0.050	0.419	n.s.	1.041
Days between first and second falls
BPV	−0.047	−0.095	0.001	0.055	−4.57	1.202
65–79	−0.096	−0.204	0.012	0.083	n.s.	1.224
80–89	−0.008	−0.124	0.109	0.898	n.s.	1.277
90–104	−0.063	−0.117	−0.010	0.020	−6.14	1.178
BP low	−0.105	−0.187	−0.023	0.012	−9.94	1.211
BP high	−0.012	−0.092	0.068	0.772	n.s.	1.232
Diagnosis low	−0.057	−0.114	−0.001	0.047	−5.58	1.227
Diagnosis high	−0.018	−0.108	0.072	0.697	n.s.	1.223
Average days between falls
BPV	−0.029	−0.056	−0.002	0.034	−2.86	1.109
Dem	0.367	0.094	0.640	0.009	44.33
65–79	−0.019	−0.075	0.037	0.512	n.s.	1.140
80–89	0.013	−0.039	0.065	0.635	n.s.	1.126
90–104	−0.045	−0.082	−0.008	0.018	−4.36	1.117
No dementia	−0.046	−0.084	−0.008	0.017	−4.50	1.090
Dementia	−0.017	−0.054	0.020	0.371	n.s.	1.121
Diagnosis low	−0.047	−0.081	−0.014	0.005	−4.63	1.102
Diagnosis high	0.020	−0.021	0.061	0.347	n.s.	1.108

Deviance/DF: Ratio of deviance criterion to degrees of freedom. A value near 1 indicates a good fit of the model to negative binomial distribution.

n.s.: Not a significant percentage change in fall outcome from BPV for that subgroup. BP, blood pressure; BPV, blood pressure variability; CI, confidence interval; DF, degrees of freedom.

BPV by itself was not significant in affecting the number of days between the first and second falls (P = 0.055). Subgroup analyses show an increase in BPV is related to a decrease in days between falls of 6.1% for the 90‐ to 104‐year‐old residents (P = 0.036), 9.9% for those with low BP (P = 0.012) and 5.6% for residents with fewer than four chronic conditions (P = 0.047). BPV contributes to a similar decrease of 2.9% in average number of days between all falls (P = 0.034) along with dementia (P = 0.009). The BPV relationship with days between falls was strongest again for the age 90–104 subgroup (4.4% decrease, P = 0.018), 4.5% those with no dementia (P = 0.017), and 4.6% for residents with fewer diagnoses (P = 0.005).

Discussion

This study correlates fall frequency with BPV, thus adding to the connection of this metric to other geriatric issues such as postural orthostasis. ¹⁹ , ²⁰ The greatest risk for frequent falls, in terms of both number and time between falls, is proportionate to rising BPV values. BPV most significantly predicts fall number among nursing home residents who are women with high BP and few diagnoses (Table 2). Curiously, fall frequency does not correlate with fall injury rates. While there have been studies linking BPV to “frailty,” to the best of our knowledge, this is the first study to link BPV to falls in this special patient population (nursing home residents). ²¹ , ²² , ²³ The relationship of multiple falls with a continuum of BPV as well as the relationship with the time of the next fall has not been reported. In short, we find that the higher the BPV, the higher the likelihood of multiple falls, which is a novel finding.

BPV does not predict future fall dates; nonetheless, second falls occur ≈6 weeks apart, which represents a time period commonly associated with muscle performance changes. ¹⁹

While this report identifies BPV as a metric for frequent fallers, others have linked this value as a frailty and dementia risk. ¹⁴ , ²² , ²⁴ Thus, BPV has concordance with risk factors for older adults vulnerable to cognitive and functional decline. BPV may represent a mechanistic thread among these geriatric syndromes that have varying degrees of autonomic dysfunction, connective tissue stiffness and other physiological issues linked to multicomplexity.

A key question is whether BPV has physiological relevance to fall pathogenesis. Elevated BPV is linked to orthostasis and postprandial hypotension. ²⁵ , ²⁶ Thus, high BPV may suggest a cardiovascular cause for at least some recurrent falls. Because elevated BPV is associated with increased arterial stiffness, one might question whether such stiffness is a proxy measure for connective tissue stiffness that promulgates falls. ²⁷

Some limitations are noted with this study. We did not record fall locations, noting that 75% of all falls occur in nursing home residents' rooms or in bathrooms. ²⁸ Additionally, we do not know how many frequent fallers had orthostasis, which has a positive association with falls, with an odds ratio of 1.7. ²⁹ Because we were interested in real‐world data, BP measurements were not standardized according to position, time of day, or other protocols. BPV was not correlated with polypharmacy or medications linked to falls. ²² We also note that the study is limited to older adults with functional disabilities living in nursing homes and may not extrapolate to community‐dwelling older adults. We also did not determine whether the older adults had falls prior to nursing home entry, nor did we determine their functional trajectory to understand if further deconditioning and loss of lean body mass contributed to fall frequency. Indeed, the relationship to BPV and deconditioning has yet to be established. Another study limitation is the large number of frequent fallers in our cohort, noting that 26% of nursing home residents experience repeat falls. ²⁴ , ³⁰

Although initial studies of BPV were often performed using secondary data analyses of clinical trials, including those in the Veterans Affairs system, in which blood pressures were measured strictly using standard protocols; subsequently, numerous studies have documented that BPV has clinical significance even when the BP readings are taken from medical chart extraction. ¹¹ , ¹² , ¹⁶ , ²¹ , ³¹ , ³² These have been measured off protocol; despite the possible errors engendered by variability in time before measurement, interval of measurement, body position, technique and device of measurement, they still provide meaningful data. Indeed, one might argue that since patients in the clinical setting rarely have their BP measured by a strict protocol, such “real‐world” data are more clinically relevant.

Falls in nursing home residents can be catastrophic and cost not only the resident's quality of life but also increase health care costs. A study investigating health care cost of falls in 2015 estimated that nonfatal falls cost Medicare $28.9 billion, Medicaid $8.7 billion and private pay $12 billion, while fatal falls cost $754 million. ³³ It has been estimated that half of all nursing home patients fall each year, but fatal and nonfatal injuries remain largely underreported so the annual cost is likely much higher. ³⁴ While some risk factors of falls in this patient population have previously been described, such as increased age, impaired balance and vision, these are imperfect, and the search for additional predictors is therefore important, so we can intervene to attempt to reduce fall risk in such a high‐risk population. ³⁵

Conclusion

This study supports the hypothesis that high BPV is associated with increased risk of frequent falling. This metric may complement other assessments for high fall risk in older adults. Since this association was found over a 10‐month period, additional questions exist whether shorter intervals of BPV measurements, such as over a 1‐month period, can equally predict frequent fallers. Because high BPV comes with arterial stiffness, strategies to address either arterial or musculoskeletal stiffness may be a mechanistic approach to reduce fall frequency.

Disclosure statement

The authors declare no conflicts of interest.

Author contributions

DAJ, MDB, LD, RM and GDM helped with study design, literature review, and manuscript writing. EHS, AH, PK and EH helped with data collections and analyses. MK helped with statistical analyses. All authors have read and agreed to the published version of the article.

Ethics statement

The protocol is approved by Institutional Review Board for Sanford Health (IRB ID MOD00009713).

Supporting information

Table S1. Total falls by patient attributes.

Table S2. Days between first and second falls by patient attributes.

Table S3. Average days between falls by patient attributes.

Soultan EH, Hara A, Knutson P, et al. Blood pressure variability associated with falls in nursing home residents. Geriatr. Gerontol. Int. 2024;24:1315–1319. 10.1111/ggi.15014

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.