bs_bs_banner

Pain Medicine 2014; 15: 1115–1128 Wiley Periodicals, Inc.

PAIN & AGING SECTION Review Article Pain Is Associated with Recurrent Falls in Community-Dwelling Older Adults: Evidence from a Systematic Review and Meta-Analysis Brendon Stubbs, BSc (Hons), MSc, MCSP,* Pat Schofield, PhD,* Tarik Binnekade, MSc,‡ Sandhi Patchay, PhD,* Amir Sepehry, BSc, MSc† and Laura Eggermont, PhD,‡ *School of Health and Social Care, University of Greenwich, London, UK; †College for Interdisciplinary Studies, Graduate Program in Neuroscience, University of British Columbia, Vancouver, BC, Canada; ‡Department of Clinical Neuropsychology, VU University Amsterdam, Amsterdam, The Netherlands Reprint requests to: Brendon Stubbs, BSc (Hons), MSc, MCSP, University of Greenwich—School of Health and Social Care Avery Hill site, Gray Building London, SE9 2UG, UK. Tel: +4420 8331 8000; Fax: +441604696126; E-mail: [email protected] Disclosure: BS is supported by a Vice Chancellors scholarship at the University of Greenwich, but this did not affect the research in any way or the decision to submit for publication. None of the other authors have any disclosures of conflicts of interest to declare.

Abstract Background. Pain and recurrent falls are highly problematic in community-dwelling older adults, yet the association remains elusive. Objective. The objective of this study was to investigate the association between pain and recurrent falls in community-dwelling older adults. Design. Two independent reviewers conducted searches of major electronic databases, completed methodological assessment, and extracted the data of all included articles. Articles that were included

are those that 1) involved community-dwelling older adults; 2) recorded recurrent falls; and 3) assessed pain. Articles that were excluded are those that included participants with dementia, any neurological conditions, or those with orthopedic trauma/ surgery in the past 6 months. Results. Out of a potential of 71 articles, 11 met the inclusion criteria and 7 (N = 9,581) were eligible for the meta-analysis. The annual prevalence of recurrent falls in those reporting pain (12.9%) was higher than the pain-free control group (7.2%, P < 0.001). A global meta-analysis established that pain was associated with recurrent falls (odds ratio [OR]: 2.04, confidence interval [CI]: 1.75–2.39; N = 3,950 with pain and N = 5,631 controls), and this was decreased in a subgroup meta-analysis utilizing prospective studies only (OR: 1.79, CI: 1.44–2.21, P < 0.001, I2 = 0%; N = 3, N = 2,646). A subgroup analysis comparing recurrent fallers vs non-fallers only (OR: 2.18, CI: 1.82–2.60, N = 6,320, I 2 = 0%) established the odds were particularly higher than single fallers vs non-fallers (OR:1.44, CI: 1.26–1.64, N = 6,903, I 2 = 0%). Conclusion. Older adults with pain are at particularly increased risk of recurrent falls. Clinicians working with recurrent fallers should routinely assess pain while pain specialists should inquire about older adults’ falls history. Key Words: Falls; Activities of Daily Living; Older Adult; Chronic Pain; Musculoskeletal Pain; Recurrent Falls; Falls Prevention; Falls Risk Factors. Introduction Falls in community-dwelling older adults are a serious global public health concern [1]. The consequences of an older adult experiencing a fall can be profound and may result in functional decline, admission to long-term care facilities, and increased mortality [2–5]. It has been demonstrated that about 5% of falls lead to a fracture, 1115

Stubbs et al. another 5% lead to other serious consequences [6], but many who fall experience psychological concerns, such as fear of falling, that increase their risk of future falls [7]. In addition, one in four people who fall and half of those acquiring an injury as a consequence will seek treatment from an emergency department or general practitioner [6,8]. Falling is common in community-dwelling older adults, and each year about 30% and 15% will experience a single or recurrent falls [9–11]. Naturally, the likelihood of experiencing adverse consequences from a fall is increased in those who fall more often, and outcomes are considerably worse in recurrent fallers [8]. For instance, compared with those who do not fall or experience single falls, recurrent falls are associated with a more pronounced loss of confidence, greater physician contact, social isolation, greater functional decline, increased likelihood of nursing home admission, and mortality [10,12,13]. Naturally, there is an urgent need to prevent falls, and an essential strategy to achieve this is the identification and management of important risk factors [13,14]. Within the last few years, a number of studies have established that pain is independently associated with falls in communitydwelling older adults [3,15,16]. This finding is important, as up to 50% of community-dwelling older adults experience pain [17]. A recent meta-analysis [18] involving over 17,000 older adults summarized the evidence and established that older adults with pain were more likely to experience any fall (1>) in the past 12 months and fall again in the future, but crucially this review did not specifically explore the relationship between pain and recurrent falls. However, the investigation of a possible relationship between pain and recurrent falls is essential as their prevention is an international public health priority, and recurrent fallers are recognized as a distinct “atrisk” group [8,12,13]. Both the American and British Geriatrics Society [13] and National Institute of Health and Clinical Excellence (NICE) [14] stipulate the need to provide comprehensive assessment for those at risk of recurrent falls; however, neither currently recognizes that pain may be a risk factor for recurrent falls. This is despite the fact that a number of studies [9,19–22] have demonstrated that older adults with pain are at a particularly pronounced risk of recurrent falls over single falls. In addition, a recent systematic review [5] investigating over 30 falls risk factors established that pain was particularly associated with recurrent falls. However, the results were overshadowed by its generic focus on multiple risk factors, and it offered minimal information regarding pain, leaving the influence of pain as a risk factor uncertain. Because pain is common in community-dwelling older adults and a number of individual studies have reported a particular increased risk with recurrent falls, a unique meta-analysis specifically investigating this relationship is required to quantify this risk. Therefore, in order to address this gap within the literature, the primary aim of this article is to establish if community-dwelling older adults with pain are at increased odds of recurrent falls compared with asymptomatic controls. 1116

Method The study was conducted and reported in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement [23]. Eligibility Criteria Studies were eligible for the review upon meeting the following criteria: 1) The study was conducted in community-dwelling older adults with a mean age of 60 years and older [5]; 2) The authors recorded recurrent falls as an outcome, defined as two or more falls over a monitoring period of at least 12 months [24]. Falls could be ascertained through either prospective or retrospective measurement [18]; and 3) The study assessed pain (through a validated outcome measure, questionnaire, or clinical assessment), and there was a sample with and without pain [18]. If the study included participants whose pain was identified as being caused by a previous fall, the study was excluded in order to decrease the likelihood of encountering reverse causality. We also excluded studies conducted on people with dementia, due to the substantially increased risk of falls in this population and difficulty obtaining accurate records of falls in this group [25]. We also excluded studies reporting falls in people with other neurological conditions such as stroke or Parkinson’s disease in order to reduce the influence of major comorbidity on falls risk [26]. In addition, we excluded studies conducted on individuals with orthopedic surgery or trauma in the past 6 months in order to reduce heterogeneity in our results. Information Sources Two reviewers (BS/TB) independently conducted searches on major electronic databases from inception until May 2013, including Cochrane Library, CINAHL, EBSCO, EMBASE, PubMed, and PsycINFO. Search Strategy The search terms used were “older adults” or “aged” or “elderly” or “old age” AND “pain*” or “chronic pain” or “persistent pain” or “musculoskeletal pain” AND “fall” and “recurrent fall.” In addition, all corresponding authors of studies identified from a previous systematic review [18] were contacted to provide the raw data for participants from their study who experienced recurrent falls in the samples with and without pain in a 2 × 2 table or incorporating single fallers also in a 2 × 3 design. We also requested the mean age and gender of participants in both groups (pain and no pain) in addition to information on other important falls risk factors. Study Selection Two authors (BS/TB) were involved in the study selection, and a third reviewer was available if required. The two reviewers screened the titles and abstracts before compiling a list of possible studies that were considered in a full

Pain and Recurrent Falls text review. Two authors reviewed the full texts and agreed upon the list of included studies.

(chronic vs non-chronic), and location of pain, we reported these results within the narrative synthesis.

Data Collection

Due to the heterogeneity of the data acquired, the DerSimonian and Laird random effects model was utilized [32]. In order to measure heterogeneity, the I2 statistic was used and scores of 25%, 50%, and 75% were considered low, moderate, and high heterogeneity, respectively [33]. In an exploratory analysis, wherever possible, we attempted to assess the influence of age and the percentage of females on any observed results and conducted meta-regression analyses using a mixed random effects model examining whether the observed variance was significantly explained by age and the percentage of females. All analyses were conducted using the Comprehensive Meta-Analysis software (Version 2.0; Biostat, Englewood, NJ, USA). In order to assess publication bias, we undertook an inspection of a funnel plot of all of the included studies in the meta-analysis [34].

Two authors (BS/TB) were involved in the data collection and extraction process. The data extracted from each study included year of publication, study design, sample size, participant information (age, percentage of females), method of pain assessment, and the number of participants experiencing recurrent falls in the sample identified as having pain and without pain. If an article reported an association statistic to quantify the relationship between pain and recurrent falls, these data were extracted together with the 95% confidence interval (CI) and P value. Wherever possible, we calculated the unadjusted odds ratio (OR, together with 95% CI and P value) from the data that authors provided us with or from data available within the article. Wherever possible, we used the falls data collected over 12 months as this time frame is commonly used in clinical practice and research [13,27,28]. This also improved homogeneity in the reporting of results.

Results Study Selection

Methodological Assessment of Included Studies Two authors (BS/TB) conducted the methodological assessment of all included articles using the Newcastle Ottawa Scale (NOS [29]). The NOS provides an assessment of the quality of nonrandomized controlled trials and has been used in a recent similar systematic review [18]. Each article received a methodological quality score out of 9, and all articles were judged across three key areas: selection, comparability, and outcomes. The NOS validity and reliability have been established [29], and scores of 5 out of 9 were considered reasonable quality [30]. The NOS can be adopted and we adapted the scores to give one star to account for age and another for gender or comorbidity. In addition, we updated the requirements for a star when considering the ascertainment of falls in the exposure category, and only allocated a star when a valid measure was used to collect falls data (either retrospective or prospective ascertainment).

Out of a potential 795 articles, 294 titles and abstracts were considered by two reviewers. Seventy-one full texts were subsequently reviewed and upon applying the eligibility criteria, 60 articles were excluded with reasons. At the full text screening stage, we contacted 21 authors up to three times over a 6-week period requesting additional information, and four authors provided additional information for the meta-analysis (see Acknowledgements section). Of the 21 authors contacted, a further 10 were unable to provide additional data on recurrent fallers from their sample, and we identified a further two studies that were overlapping with studies already included. Altogether, 11 studies [9,16,19,20,22,31,35–39] were included within the narrative synthesis and 7 (total N = 9,581: 3,950 with pain and 5,631 without pain, 7 [16,31,35–39] of these were eligible for inclusion in the meta-analysis. For full details of the search strategy, see Figure 1.

Summary Measures and Data Synthesis

Study and Participant Characteristics

Wherever possible, the raw data from each study was pooled to establish the relationship between pain and recurrent falls in a 2 × 2 table and an unadjusted OR and 95% CI and P value calculated. In accordance with previous research [31], recurrent falls were compared with non-fallers and single fallers as the comparison group. In order to establish the 12-month prevalence of recurrent falls in older adults with and without pain, we calculated a point estimate. Furthermore, a subgroup analysis was conducted to assess the influence of the method of falls ascertainment upon any observed outcomes (prospective vs retrospective falls data collection). Wherever possible, we also investigated the relationship between pain and 1) recurrent fallers vs non-fallers and 2) single falls vs nonfallers separately in a 2 × 3 table. In order to investigate the influence of the type (e.g., musculoskeletal), duration

The summary of the included articles is presented in Table 1. One study employed a cross-sectional design [38] and ten were cohort studies [9,16,19,20,22,31,35– 37,39]. The sample size varied; the largest study was Arden et al. [20] with 5,552 participants, while Mickle et al. [37] was the smallest with 303 participants. The mean age across the studies differed, although Arden et al. [20] included the youngest participants (mean age 71.4 years) and Sturnieks et al. [38] included the oldest (mean age 80.2 years). The percentage of females also varied considerably, and one study [36] only provided data for women only. The method of assessing pain was varied, and the duration, type, and location of pain also varied considerably (see Table 1 for more details). Three studies assessed pain 1117

IdenƟficaƟon

Stubbs et al.

Records iden fied through database searching (N = 1,334)

Addi onal records iden fied through other sources 9

Eligibility

Screening

Records a er duplicates removed (N = 795)

Records screened (N = 294)

Full-text ar cles assessed for eligibility (N = 71)

Included

Studies included in narra ve synthesis (N = 11)

Studies included in quan ta ve synthesis (meta-analysis) (N = 7)

Records excluded (N = 223)

Full-text ar cles excluded (N = 60), with reasons: N = 17 no measure of falls N = 9 no measure of recurrent falls N = 8 no control group N = 4 unable to differen ate between those with and without pain N = 3 par cipants within study met exclusion criteria N = 7 overlap with other studies N = 1 other reasons N = 1 recorded falls less than 6 months N = 6 excluded a er contact with authors as did not meet inclusion criteria N = 2 unable to contact primary author N =1 nursing home sample N = 1 author unable to provide recurrent falls data

Figure 1 PRISMA 2009 flow diagram for search strategy. PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses. over the past 12 months [19,20,39], five inquired about pain over the past month [16,22,31,36,38], and three assessed current pain or pain of an unknown duration [9,35,37] . The location of pain was not clear in five studies [19,22,31,35,38], while one looked at chronic hip pain [20], two inquired about lower limb pain in the past month (hip or knee [36]; hip [16]), one assessed current hip pain [9], and another one foot pain [37].

Methodological Quality Across Studies All of the included studies had an acceptable methodological quality rating score. The mean NOS score across the 11 studies was 6.8 ± 0.9. Therefore, there were not any major concerns about including any of the studies based on the methodological quality score. The Newcastle Ottawa scores are represented in online supplementary tables (Supporting Information Tables S1–2).

Prevalence of Recurrent Falls per Annum Narrative Review Results It was possible to calculate the annual prevalence of recurrent falls from six studies [16,35–39], and this established that 12.9% (463/3,573) older adults with pain reported recurrent falls compared with 7.2% (335/4,603) older adults without pain (P < 0.001). 1118

All 11 of the included studies reported that pain increased the odds of recurrent falls (see Table 2). Seven studies reported an association statistic within their article, establishing an increased risk of falls [9,16,19,20,22,31,36];

Design and Settings

Cohort study Community (United States)

Cohort study (cross-sectional analysis of falls) Community (AUS)

Cohort Community (TW, CN, and AUS)

Cohort Community (United States)

Cross-sectional (baseline data) Community (AUS)

Cohort Community (Japan)

Arden et al. [20]

Blyth et al. [22]

Kwan et al.* [35]

Leveille et al. [36]

Morris et al. [19]

Muraki et al. [16]

N = 5,552 71.4 ± 5.1 years 100% female 60.6% confirmed they had self-report physician diagnosed OA. 11.6% had definite radiographic hip OA. Cases matched for both groups N = 3,181 65.1% female N = 2,227 pain in last 4 weeks (with or without interfering with activity) N = 710 slight pain causing interference N = 711 moderate–severe pain causing interference N = 784 no pain N = 3,035 for the analysis of recurrent falls N = 978* 76.0 years 297 had pain 681 had no pain N = 940 100% female N = 233 no pain N = 189 other pain N = 293 moderate/severe lower extremity pain N = 225 widespread pain Total N with pain = 707 Age (years) No pain 80.2 ± 8.1 Other pain 78.8 ± 7.7 Lower extremity pain 77.3 ± 8.4; widespread pain 76.5 ± 7.3 (P < 0.001) N = 940 participated in 6/12 falls follow-up and included in analysis N = 1,000 73.4 (65–94 range) 53.3% female Unclear number of participants who had chronic pain (>12months) N = 1,348 with baseline and follow-up data N = 452 males, 64.9 ± 11.7 years N = 896 females, 63.3 ± 11.8 years 69/453 males had knee pain 230/896 females had knee pain 85/452 males had LBP 193/896 had LBP Data on LBP and falls not available

Participant Information

Summary of included articles

Study

Table 1

Asked if had pain in 1) knee and 2) LBP for most days in past month

Pain frequency measured 5-point Likert scale (never to everyday) over past 12 months

Questionnaire on current pain interfering with activity No details on location and duration of pain NRS for hip and knee pain over past month

SF 36—bodily pain and pain interfering with activities; last 4 weeks

Self-report chronic hip pain over 12 months. Chronic hip pain N = 1,914 (34.5%) sample

Pain Ascertainment Location Severity

3 years (R) over follow-up Recurrent falls classed as ≥2 falls in 3-year follow up

12 months history of falls (R) Recurrent falls classed as ≥2 falls over 12/12

Baseline data 12 months (R); recurrent falls classed as ≥2 falls. 3-year follow-up (R) falls history classed as ≥2 falls

12–24 months (P) Recurrent falls classed as ≥2 falls

12 months history of falls (R) Recurrent falls classed as ≥2 falls over 12/12

12 months history of falls (R) Asked about falls every 4/12

Recurrent Falls Ascertainment and Details

Pain and Recurrent Falls

1119

1120

Cohort Community (AUS)

Cohort Community (United States)

Cross-sectional Community (AUS)

Cohort Community (NL)

Cohort Community (HK)

Mickle et al.* [37]

Nevitt et al. [9]

Sturnieks et al.* [38]

Stel et al. [31]

Woo et al. [39]

N = 283 arthritis (41.3%): 80.2 ± 4.3 years 74.6% female N = 401 no arthritis: 80.0 ± 4.6 years (ns) 58.6% female (P < 0.05) N = 231 had pain N = 416 no pain N = 1,365 74.8 ± 6.2 years non/single fallers 76.8 ± 6.8 years recurrent fallers (P < 0.01) 51.0% female non/single fallers 51.6% females non/single fallers 27.3% non/single fallers had pain 39.5% recurrent fallers had pain N = 4,000* 50.0% female N = 1,927 had back pain 73.2 ± 5.3 years N = 2,073 no back pain 72.4 ± 5.1 years 72.1 ± 4.9 years

Manchester Foot Pain and Disability Index Duration and severity unknown

N = 303* 49.3% female 50% had foot pain 50% no foot pain, comorbid problems not mentioned N = 325 83.1% female >60 years, mean ages not available. All had reported at least one fall in past 12 months. N = 32 had hip or knee pain N = 647 participants*

Participants asked about back pain over last 12 months

Asked if had pain in past 4 weeks

Used SF 12 question asking if had pain interfering with activity in last 4 weeks N = 106 a little pain N = 71 moderate N = 51 quite a lot N = 3 unclear if those with pain had arthritis or not

Physician examination and classed as having hip/knee pain upon passive movement

Pain Ascertainment Location Severity

Participant Information

12 months history of falls (R) Recurrent falls classed as ≥2 falls over 12/12

3 years (P) falls calendars Recurrent faller classed as person who had ≥2 falls over 6/12

12 months history of falls (R) Recurrent falls classed as ≥2 falls over 12/12

12 months (P) weekly postcards Recurrent falls classed as ≥2 falls over 12/12

12 months (P) monthly calendars Recurrent falls classed as ≥2 falls over 12/12

Recurrent Falls Ascertainment and Details

*Author provided additional data. AUS = Australia; CN = China; HK = Hong Kong; LBP = low back pain; NL = The Netherlands; NRS = numerical rating scale; OA = osteoarthritis; P = prospective falls ascertainment; R = retrospective falls ascertainment; SF 12 = short form 12; SF 36 = long form SF 36; TW = Taiwan.

Design and Settings

Continued

Study

Table 1

Stubbs et al.

Chronic hip pain

Pain over last 4 weeks: No interference with activities Slight interference with activities Moderate to severe interference with activities Current pain

Arden et al. [20]

Blyth et al. [22]

Foot pain

Frequency of pain over 12 months: Sometimes Frequent Pain over past month in knee Male Female

Mickle et al. [37]

Morris et al. [19]

Pain over past 4 weeks Back pain over past 12 months

Stel et al. [31] Woo et al. [39]

3 year (P) 12 months (R)

12 months (P) 12 months (R)

3 year (R)

12 months (R)

12 months (P)

12 months (R)

3 year (R)

12 months (P)

12 months (R)

12 months (R)

Falls Ascertainment

1.54 (CI: 1.01–2.35) 1.38 (CI: 0.93–2.03) 1.66 (CI:1.10–2.50) 1.66 (CI: 1.03–2.68) recurrent vs non/single fallers 1.967 (CI: 1.20–3.21) recurrent vs non-fallers only 2.50 (CI: 1.18–5.29) recurrent vs non/single fallers 2.67 (1.24–6.73) recurrent vs non-fallers only

OR: 2.05 (CI:0.99–4.00) OR: 2.22 (CI:1.44–3.37) OR: 1.98 (CI: 1.39–2.8252) recurrent vs non/single fallers OR: 2.05 (CI: 1.44–2.93) recurrent vs non-fallers only RR: 1.9 (CI:1.3–3.7) OR: 3.02 (CI: 2.04–4.48) recurrent vs non/single fallers OR: 18.42 (CI: 10.2–33.20) recurrent vs non-fallers only OR: 1.73 (CI: 1.33–2.24) OR: 2.26 (CI: 1.71–2.99) recurrent vs non/single fallers OR: 2.38 (CI: 1.79–3.15) recurrent vs non-fallers only

OR: 2.52 (CI:1.41–4.51) OR: 2.86 (CI:1.74–4.71)

OR: OR: OR: OR: OR: OR: OR:

PR: 1.31 (CI: 0.92–1.86) PR: 1.66 (CI: 1.19–2.33) PR: 2.29 (CI: 1.67–3.13) OR: 1.75 (CI:1.14–2.68) recurrent vs non/single fallers OR: 1.96 (1.27–3.04) recurrent vs non-fallers only

RR: 1.5 (CI: 1.3–1.8)

Association Statistic Recurrent Falls

Calculated from raw data Calculated from raw data

Unadjusted Calculated from raw data

Calculated from raw data

Crude

Unadjusted

Calculated from raw data

Calculated from raw data

*

Calculated from raw data

Age, knee height, weight, clinic Age and gender

Adjusted for

*Leveille et al. 2002—Adjusted from discrete time survival analysis (using logistic regression), updating pain level to most recent follow-up interview before event. Covariates included age, race, education, body mass index, confirmed diseases (hip fracture, angina pectoris, diabetes mellitus, peripheral arterial disease, stroke, Parkinson’s disease), walking disability, fell in 12 months before baseline, Mini-Mental State Examination score, daily use of psychoactive medications, daily use of analgesic medications, gait speed, balance test score, proxy respondent, and follow-up round. Calculated from raw data—unadjusted OR calculated from raw data. OR = odds ratio; P = prospective ascertainment of falls; PR = prevalence ratio; R = retrospective ascertainment of falls; RR = relative risk.

Current hip/knee pain Pooled pain data over past month

Nevitt et al. [9] Sturnieks et al. [38]

Muraki et al. [16]

Pain over last 4 weeks: Other pain Moderate/severe pain in lower extremities Widespread pain Pooled pain data

Leveille et al. [36]

Kwan et al. [35]

Pain Details

Association between pain and recurrent falls

Study

Table 2

Pain and Recurrent Falls

1121

Stubbs et al. although four did not, it was still possible to calculate the unadjusted OR after the authors provided additional data [35,37–39]. Regardless of the site or duration of pain, the odds of recurrent falls were consistently increased in the individual studies. In total, it was possible to calculate the unadjusted OR from seven studies, establishing the relationship between pain and recurrent falls (using non-fallers and single fallers as the reference group), and all demonstrated an increased falls risk. Of these seven studies, it was possible to calculate the relationship between pain and recurrent falls, using only non-fallers as the comparison group in six studies [16,35– 39]. This additional analysis demonstrated that the increased odds of recurrent falls were increased in each study compared with the analysis when the reference group was single and non-fallers combined. This was a relatively small increase in five studies, while a sixth study demonstrated a substantial increased risk of recurrent falls [38]. Meta-Analysis The Overall Odds of Older Adults with Pain Experiencing Recurrent Falls A meta-analysis was conducted with seven studies incorporating 9,581 older adults (3,950 with pain and 5,631 without pain). Comparing recurrent fallers vs non-fallers and single fallers together established pain increased the odds of recurrent falls (OR: 2.04, CI: 1.75–2.39, P < 0.001). There was a small amount of heterogeneity in this analysis (I2 = 19.7%); the forest plot is displayed in Figure 2A. In order to establish publication bias, we undertook a visual inspection of the funnel plot for all studies included in the global analysis and subsequently, no studies appeared to be an outlier (see Figure 2B). Next, we analyzed the results for the studies where the falls data were collected prospectively (N = 3, [31,36,38]) and retrospectively (N = 4, [16,37,39,40]). This established that the odds of recurrent falls were higher for studies measuring falls retrospectively (OR: 2.21, CI: 1.79–2.75, P < 0.0001, I2 = 27.8%, total N = 6,935) compared with prospectively (OR: 1.79, CI: 1.44–2.21, P < 0.001, I2 = 0%; N = 2,646), although both were significant (see Figure 3). From the available data of six studies [16,35–39], it was possible to develop 2 × 3 tables to investigate the odds of falling comparing 1) recurrent and non-fallers and b) single vs non-fallers only. This established that pain was more strongly associated with recurrent falls (OR: 3.05, CI: 1.75–5.31, N = 7,418, I2 = 93%) although pain was still significantly associated with single falls (OR: 2.15, CI: 1.20–3.83, N = 7,778, I2 = 93%). This analysis is presented in Figures 4A and B. Both of these analyses were heterogeneous and we had concerns about one study [38] being an anomaly when the results from the 2 × 2 tables were broken down into a 1122

2 × 3 table, and it was deemed this study likely increased heterogeneity. We therefore conducted the analysis with this study removed in a sensitivity analysis measure, and this established a more moderate relationship between pain and recurrent falls (OR: 2.18, CI: 1.82–2.60, N = 6,320, I2 = 0%) and single falls (OR: 1.44, CI: 1.26– 1.64, N = 6,903, I2 = 0%) and both were nonheterogeneous (see Figures 4C and D). The Influence of the Type, Duration, and Location of Pain on Recurrent Falls Due to the limited number of studies and heterogeneity in the type, duration, and location of pain in each of the studies, it was not possible to conduct any subgroup analysis to determine if these had a specific influence upon the relationship with recurrent falls. Meta-Regression In an exploratory meta-regression analysis, it was possible to use the data from three studies [35,37,39] to investigate the influence of age and percentage of females upon the observed outcomes of the meta-analysis. For the effect of age on the pain group that experienced recurrent falls, on the effect-size estimate, mixed effect regression slope was –0.043 (standard error [SE]: 0.041; P = 0.287); for the pain but no recurrent falls group, the slope was −0.081 (SE: 0.076; P = 0.282); for the no pain and recurrent falls group, the slope was −0.0906 (SE: 0.0872; P = 0.298); and for the no pain and no recurrent falls group, the slope was −0.0885 (SE: 0.0819, P = 0.279). In addition, female gender was not related to recurrent falls in the pain group that experienced recurrent falls (slope: 0.0124, SE: 0.0134; P = 0.355), the non-pain group that fell (slope: −0.0137, SE: 0.0127; P = 0.278), the pain group that did not fall (slope: −0.0468, SE: 0.0468; P = 0.317) nor in the non-pain group that did not fall (slope: −0.0107, SE: 0.0238; P = 0.652). Discussion To our knowledge, this is the first systematic review and meta-analysis specifically investigating the association between pain and recurrent falls in community-dwelling older adults. Our meta-analysis incorporating 9,581 unique older adults established that pain was associated with approximately a 100% increased odds of recurrent falls (OR: 2.04, CI: 1.75–2.39). We conducted a subgroup meta-analysis to establish if the monitoring period of falls (prospective vs retrospective) influenced the outcome and found that both reported pain increased the odds of falls, although this was higher when falls were measured retrospectively. It was possible to conduct a subgroup metaanalysis directly comparing recurrent fallers and nonfallers only, and this established that pain was associated with approximately a twofold increased odds of recurrent falls (OR: 2.18, CI: 1.82–2.60, N = 6,320, I2 = 0%). This was more pronounced than the available data comparing single fallers vs non -fallers only (OR: 1.44, CI: 1.26–1.64, N = 6,903, I2 = 0%). The annual prevalence of recurrent

Pain and Recurrent Falls (A)

Funnel Plot of Standard Error by Log odds ratio

(B) 0.0

Figure 2 (A) Forest plot for poling of all studies comparing recurrent fallers vs single/non-fallers. N = 9,581 (3,950 with pain, 5,631 with no pain); heterogeneity = I2 = 19.7%, P = 0.27. (B) Funnel plot of standard error for all included studies in main analysis. CI = confidence interval.

Standard Error

0.1

0.2

0.3

0.4 -2.0

-1.5

falls was significantly higher in those with pain (12.9 %) than those without pain (7.2%, P < 0.001). However, the number of recurrent falls reported in both our samples is below the 15% reported in the literature [40]. Within the regression analysis, we found that age and gender had no significant effect on the observed results. A previous meta-analysis demonstrated a more moderate association between pain and falls (1>), but this looked at all falls (single and recurrent combined) and did not make the distinction between single and recurrent falls [18]. Our results also indicate a stronger relationship between pain and recurrent falls than a previous meta-analysis [5] that investigated over 30 falls risk factors, which reported an OR of 1.60 (CI: 1.44–1.78). However, it is unclear if the

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Log odds ratio

authors of this review [5] used non-fallers only as their comparison group or if they pooled single and non-fallers together. Thus, the association provided in their study is uncertain and did not receive the necessary attention due to its wide scope. All 11 studies included within our review established that there was a significantly increased risk of recurrent falls for older adults with pain. When we compared the risk of older adults with pain experiencing single and recurrent falls separately, the risk for recurrent falls was higher on each occasion. Our sensitivity analysis comparing recurrent falls vs non-fallers only and with one study removed [38] established an OR of 2.18 (CI: 1.82–2.60) with no heterogeneity (I2 = 0%). This analysis with 6,320 older 1123

Stubbs et al.

Figure 3 Subgroup analysis comparing recurrent vs single/non-fallers separated by the design of falls collection. Prospective N total = 2,646 (786 with pain and 1,860 without pain). Retrospective N total = 6,935 (3,164 with pain and 3,771 without pain). Heterogeneity: prospective = I2 = 0%, P = 0.6558; retrospective I2 = 27.8%, P = 0.2396. CI = confidence interval. adults (N = 2,694 with pain and 3,626 without pain) may therefore represent the most accurate representation of the relationship between pain and recurrent falls. A number of studies within the literature have reported that pain is more strongly associated with recurrent falls compared with single falls [15,19,22]. For instance, Morris et al. [19] established that recurrent fallers were more likely to be older females and commonly reported extrinsic risk factors for falls including loss of balance, lack of attention, and dizziness. Out of 1,671 responses in their study, no older adults attributed their fall to pain. However, the authors report that recurrent falls tended to occur indoors rather than outdoors. Previous research [41] has demonstrated that older adults with chronic musculoskeletal pain are less active than asymptomatic controls, and Morris et al. [19] specifically investigated chronic pain in their study (over 12 months). A possible explanation for the increased risk of indoor falls is that older adults experiencing chronic pain reduce their activity and stay indoors more due to their pain, and their risk of indoor falls is subsequently increased. However, a recent meta-analysis [42] demonstrated that physical activity is essential to maintain mobility and independence in activities in daily living. Previous research has also demonstrated that exercise is effective in reducing falls [43] and also injurious falls [44]. Therefore, encouraging older adults with pain to be physically active is likely to be important to maintain independence and reduce falls. Blyth et al. [22] also established that pain was associated more strongly with recurrent falls but did not offer an explanation. It may be that depressive symptoms contribute to the increased falls risk seen in those older adults with pain. Previous research has identified that depressive symptoms are strongly related to recurrent falls [45], and depression is common in older adults with chronic pain [46]. However, no study included in our review provided data on depressive symptoms for those with and without pain and so we could not conduct analysis to elucidate the influence of depressive symptoms on recurrent falls in our review. It appears that 1124

no other study has yet established what is different about recurrent fallers with pain and non-fallers and single fallers. However, Leveille et al. [3] postulate that the association between chronic musculoskeletal pain and falls may be due to local joint pathology, neuromuscular effects of pain, or central mechanisms whereby pain affects cognition. The mechanisms underlying the relationship between pain and the increased risk of recurrent falls found in older adults warrant further investigation. Clinical Implications Our finding that older adults with pain are over 5% more likely to fall per annum is of high clinical relevance for several reasons. First, the prevention of recurrent falls is an international priority as people who experience recurrent falls are at seriously increased risk of experiencing the multiple adverse health consequences and even death following a fall [13]. In addition, older adults who experience recurrent falls are associated with profound economic burden on health care systems [1]. Therefore, identifying factors that may contribute to recurrent falls is crucial and importantly, pain is not considered as a risk factor for recurrent falls at present. When one considers that up to 50% of community-dwelling older adults experience pain [17], this increased risk of falls observed in this group is likely to be highly clinically relevant and represents a substantial proportion of our aging worldwide population. The findings of the current review may contribute to key clinical assessment guidelines for clinicians working with people who are at risk of falling or have fallen (e.g., American Geriatrics Society and British Geriatrics Society [13] and NICE [14]). The presence of pain is an important risk factor that clinicians should routinely assess, specifically so because the strength of the association between pain and recurrent falls (OR: 2.18, CI: 1.82–2.6) is similar to other commonly well-established risk factors such as

Pain and Recurrent Falls (A)

(B)

(C)

(D)

Figure 4 (A) Forest plot comparing recurrent fallers vs non-fallers only. Total N = 7,418 (3,150 with pain and 4,268 without pain); heterogeneity I2 = 93.7%, P < 0.001. (B) Forest plot comparing single vs non-fallers only. N = 7,778 (3,150 with pain and 4,268 without pain); heterogeneity I2 = 93%, P < 0.001. (C) Forest plot comparing recurrent fallers vs non-fallers only (Sturnieks et al. 2004) [38]. Total N = 6,320 (2,694 with pain and 3,626 with pain); heterogeneity I2 = 0%, P = 0.8. (D) Forest plot comparing single vs non-fallers only (Sturnieks et al. 2004) [38]. Total N = 6,903 (2,994 with pain and 3,909 without pain); heterogeneity I2 = 0%, P = 0.86. CI = confidence interval. increasing age (OR: 1.12, CI: 1.07–1.18 [5]), physical disability (OR: 2.42, CI: 1.80–3.26 [5]), cognitive impairment (OR: 1.56, CI: 1.26–1.94 [5]), depression (OR: 1.86, CI: 1.26–2.38 [5]), and fear of falling (OR: 2.51, CI: 1.78–3.54 [5];). Clinicians should note that these risk factors were adjusted for age and sex while ours were not. However, our exploratory meta-regression analysis demonstrated that age and sex had no significant effect on the observed outcomes and that our results were consistent that pain is associated with an increased risk of recurrent falls. Our results therefore have important implications for clinicians working in general older adult services. We recommend that if an older adult presents at risk of, or has fallen already, that the clinician assesses pain and if present, seek appropriate treatment that may include pain management strategies. Previous research [36] has demonstrated that analgesic medication actually lowered falls in older adults with pain, but a more recent study found no

such influence [3]. This highlights the important role of the pain clinician in the management of falls in older adults. Our review demonstrates that older adults presenting with pain are more likely to have fallen in the past 12 months (OR: 2.21, CI: 1.79–2.75) and fall again in the future (OR: 1.79, CI: 1.44–2.21). Pain medicine clinicians working with older adults with pain should routinely inquire about their falls history and link in with their local falls service. In general medicine, a single question inquiring about a history of falls over the past 12 months is commonly used as an indicator to identify the risk of future falls [13,27,28] and if pain clinicians establish this, a referral to a falls service should be made. Unfortunately, due to the limited number of studies and the heterogeneity in the assessment of pain, it was not possible to establish whether certain types (e.g., musculoskeletal pain), sites (e.g., back pain), or duration of pain (e.g., chronic) are particularly associated with an increased falls risk. 1125

Stubbs et al. Limitations When considering the results of this review, it is important that a number of limitations are noted. First, it is not possible to rule out reverse causality for the studies that measured falls retrospectively. However, we attempted to negate this by excluding studies whereby pain was identified from a previous fall. In order to investigate this, we conducted a separate analysis comparing falls data measured retrospectively and prospectively and established a small difference. Regardless, retrospective history of falls is strongly associated with future falls and is routinely used to identify those at risk of future falls [13]. Second, there was considerable heterogeneity in the assessment and classification of pain within the studies included, making it not possible to conduct separate subgroup metaanalyses. Third, we used unadjusted OR for the metaanalysis as it was not possible to consistently adjust for other known falls risk factors. In reality, falls are often multifactorial [1,13,14] and it is likely that if we were able to consistently adjust for known risk factors, we would have a more accurate association between pain and recurrent falls. In addition, because most of the studies’ primary aim was not to investigate the association between pain and recurrent falls, we could not consistently adjust for age and gender in our analysis. However, meta-regression analysis demonstrated from the available data that these had no influence on our observed results. Therefore, using unadjusted OR may have slightly inflated the results found in our article. Another limitation is that due to the paucity of available data on mean age and gender, we could not stratify our results according to age groupings (e.g.,

Pain is associated with recurrent falls in community-dwelling older adults: evidence from a systematic review and meta-analysis.

Pain and recurrent falls are highly problematic in community-dwelling older adults, yet the association remains elusive...
705KB Sizes 1 Downloads 3 Views