Factors Associated with the Severity and Progression of Self-Reported Hand Pain and Functional Difficulty in Community-Dwelling Older Adults: A Systematic Review

Abstract

Background

Hand problems are common in older adults and cause significant pain and disruption to everyday living. The aim of this systematic review was to summarize evidence on the factors associated with the severity and progression of self-reported hand pain and functional difficulty in population-based studies of older adults.

Methods

MEDLINE, EMBASE, CINAL, BNI, AMED, HMIC, PsycINFO and ISI Web of Knowledge were searched up to January 2011 for relevant articles. The search strategy combined text words for hand, pain, function and epidemiological study. Inclusion criteria were applied and articles in the review assessed for quality using the QUality In Prognosis Studies (QUIPS) assessment tool. Data extraction included: author, year of publication, study location, participant inclusion criteria, risk factor and outcome measurement, and association with hand pain and/or function.

Results

Seven articles from five studies met the inclusion criteria from 5,679 citations. All studies were cross-sectional and provided no information on progression of hand pain and function over time. Factors associated with limited hand function were older age, female gender, manual occupation, neck or shoulder pain, clinical and radiographic osteoarthritis, weaker hand strength, hand pain, history of Parkinson's disease, stroke, diabetes or rheumatoid arthritis, and illness perceptions (namely, frustration, impact and symptom count). Key factors associated with hand pain severity were age, impact, frustration, patient expectation of a long disease time course and self-reported diagnosis of the cause of the hand problem.

Conclusions

Both demographic and clinical factors were found to be related to self-reported hand pain severity and functional difficulty in older adults; however, the results were derived from a small number of studies, with no information on progression of hand pain and functional difficulty over time. Copyright © 2012 John Wiley & Sons, Ltd.

Background

Hand and wrist problems in the general population are common, with prevalence rates varying from 5% to 26%, depending on how severity and duration of symptoms are defined (Palmer, 2003). In older adults, prevalence rates are higher for both pain and functional difficulty (Palmer, 2003). For adults aged over 55 and 50 years, respectively, the one-month and one-year period prevalence of hand pain was estimated at 17% and 30% (Dahaghin et al., 2005a; Dziedzic et al., 2007), with reports of loss of hand function and difficulty in completing everyday tasks.

Many hand problems are assessed and managed in primary care. A recent systematic review (Mallen et al., 2007a) highlighted that knowledge of the clinical course of musculoskeletal pain problems over time may assist clinical management and identify those at risk of poor outcome. However, most studies in the review focused on back or spinal pain, and no relevant studies of hand pain patients were found, leading the authors to highlight the need for further studies of hand problems in primary care (Mallen et al., 2007a).

People presenting to general practice may not represent the full spectrum of adults with a hand condition, as not all will consult their general practitioner (Dziedzic et al., 2007; Spies-Dorgelo et al., 2007), and consultation may be motivated by more severe symptoms. Hence, studies based in the general population (i.e. from an unselected population) will provide information across a wider spectrum of hand symptoms and conditions.

Older adults represent a subgroup of the general population with a greater risk of hand pain (Palmer, 2003). Risk factors for progression of hand pain and functional difficulty in older adults may differ from those in younger adults (Gagliese, 2009) and from people defined by clinical diagnosis alone. For example, the role of occupational risk factors may differ in populations of retired workers, and the influence of any co-existing conditions may also impact on outcome in older adults. The objective of this systematic review was to summarize the available evidence on factors associated with self-reported severity and progression of hand pain and functional difficulty in population-based studies of older adults.

Methods

Selection criteria

Publications were included in the review if they had explored factors associated with severity or progression of self-reported hand pain or functional difficulty in older adults selected from the general population. Studies were excluded if they: (1) measured only the presence of hand pain or functional difficulty (yes/no); (2) were based in subgroups of the general population (such as those with specific medical complaints – e.g. Parkinson's disease – or specific hand conditions – e.g. rheumatoid arthritis); (3) were not written in English; (4) did not involve original research published in a peer-reviewed journal; (5) reported only measures of hand stiffness or numbness; (6) were studies of hand injury, treatment or surgery; (7) were validation studies of questionnaire tools or diagnostic tests (e.g. x-ray or magnetic resonance imaging); or (8) were clinical case studies, case series, or qualitative studies.

Search strategy

NHS Healthcare Databases (search 2.0) was used to search the following databases for relevant review articles: MEDLINE, EMBASE, CINAL, British Nursing Index (BNI), Allied and Complementary Medicine (AMED), Health Management Information Consortium (HMIC), PsycINFO (psychology and allied fields). The search included any publication in any of the databases prior to January 2011. ISI Web of Knowledge was searched to identify any further key articles not included in the main search.

Three components were included in the search strategy and were combined using Boolean logic: (hand) AND (pain or function) AND (epidemiological study). Subject Headings (e.g. MESH headings) were used to describe the concepts of ‘hand’, ‘pain’ or ‘function’, if available; otherwise, text words were used. Titles and abstracts were searched in all databases, with the exception of Web of Knowledge, where a title search only was completed, as an abstract search was not available. Published filters were used to identify epidemiological studies (Scottish Intercollegiate Guidelines Network, http://www.sign.ac.uk/methodology/filters.html). Full details of the search strategy are available from the authors on request.

The titles and abstracts of publications generated by the search strategy were screened for possible inclusion in the review. In the first stage, two authors (E.N. and E.T.) independently reviewed the first 100 abstracts, to ensure that the inclusion and exclusion criteria were appropriately and consistently applied. One reviewer (E.N.) then continued to review the abstracts of the remaining papers, with a second independent reviewer (E.T.) being consulted where any ambiguity arose. A third reviewer (D.vdW.) was involved when consensus was not achieved between the first two reviewers. Where inclusion or exclusion could not be determined from the abstract alone, full text articles were obtained and screened using the same consensus process that was applied at the abstract selection stage.

The reference lists of all articles included in the review were hand searched to identify any further relevant publications. In addition, experts within the systematic review team were asked if any further articles could be identified for inclusion in the review.

Quality assessment and data extraction

Articles included in the review were assessed using a quality assessment tool designed specifically for use in systematic reviews of prognostic studies (QUality In Prognosis Studies; QUIPS) (Hayden et al., 2006). The quality assessment tool included six major headings, each addressing a possible bias that could occur in a prognostic study (study participation, study attrition, prognostic factor measurement, outcome measurement, confounding and analysis). Before the quality assessment tool was used, minor modifications were made, to ensure that it applied both to cross-sectional and longitudinal studies – that is, the heading ‘prognostic factor measurement’ was simplified to ‘factor measurement’ and the section on study attrition was not scored for cross-sectional studies. The tool was applied to each paper in the review (using guidelines published by the developers of the tool) by two independent reviewers (E.N. and either E.T., D.vdW. or J.J.). Each heading was rated as ‘Low’, ‘Moderate’ or ‘High’ risk of bias. Any disagreements were resolved by consensus.

Data extraction was completed for each article and included the following information: author, year of publication, study location, participant inclusion criteria, and measure of hand pain and function. Factors explored for association with hand pain and function were listed and their strength of association recorded (e.g. odds ratio, mean difference, correlation). When more than one adjusted analysis was presented (from several multiple regression models), data were only extracted for the model with the highest number of adjusting factors. All data extraction was completed by one reviewer (E.N.) and was checked by two independent reviewers (E.T. or D.vdW.) for completeness and accuracy.

A meta-analysis to pool estimates of association was planned if data collection methods and statistical methodology were similar across studies. The meta-analysis would assess the heterogeneity of study results (including a test for homogeneity and computation of I²) (Higgins et al., 2003) and pooling of estimates by random effects modelling if appropriate (Kirkwood and Sterne, 2003). A sensitivity analysis would test the stability of associations after excluding any studies of poor quality (i.e. those scoring high risk from bias on any quality assessment domain).

Results

The search strategy identified 6,363 citations (MEDLINE 2,074; EMBASE 1,287; CINAL 220; BNI 129; AMED 525; HMIC 140; PsychINFO 561; Web of Knowledge 1,427). After removal of duplicate citations in more than one database (duplicate citations identified by electronic filters), 5,679 citations were considered for inclusion in the review. Screening of article titles and abstracts excluded 5,207 articles from the review. Common reasons for exclusion were study samples not selected from the general population (e.g. studies evaluating the effectiveness of surgery or based in a group of patients with a specific clinical condition not directly related to the hand – e.g. stroke) or studies not focused on older adults.

The remaining 472 abstracts were screened by a second reviewer (E.T.) and, after a consensus meeting, 315 were excluded. Articles were mainly removed because they focused on a particular hand condition requiring specific treatment or specialist care (e.g. rheumatoid arthritis or carpal tunnel syndrome). Papers on hand osteoarthritis, however, were kept in at this stage, to ensure that no studies using a clinical diagnosis of ‘hand osteoarthritis’ based on ‘hand pain’ were missed and also because this condition is likely to specifically affect our population of interest (i.e. older adults in the community).

A total of 157 full text articles were reviewed and, after applying the exclusion criteria, six articles remained in the review (Baron et al., 1987; Dahaghin et al., 2005a,2005b; Hill et al., 2007; Marshall et al., 2009; Niu et al., 2003). An additional article (Dziedzic et al., 2007) was identified by contact with experts in the systematic review team, giving a total of seven articles in the review. A search of the reference lists of these seven articles did not yield any further articles for inclusion in the review. Further details of article selection are given in Figure 1.

Figure 1. Flow chart of article selection procedure

Description of articles included

Details regarding the seven articles included in the review are shown in Table 1. The articles in the review were based on five independent studies of older adults, which varied in size from 32 to 7,983 participants. Response rates varied across studies (16–79%). Data extracted from all studies were cross-sectional in nature.

Table 1. Summary of study populations, design and outcome measures in the review

Author (year) and country	Study design	Study population	Response rate	Outcome	Outcome
				Hand function	Hand pain
HAQ, Health Assessment Questionnaire; AIMS2, Arthritis Impact Measurement Scales 2; KL, Kellgren–Lawrence; AUSCAN, Australian/Canadian Hand Osteoarthritis index. † Data from Dahaghin et al., 2005a,2005b were derived from the same cohort study, although Dahaghin et al., 2005b was based on a sub-sample of participants with radiographic data. ‡ Data from Hill et al., 2007 and Dziedzic et al., 2007 were derived from the same cohort study.
Dahaghin et al., 2005a†	Population-based cohort study	Inhabitants of Ommoord aged 55 years and over	N = 7,983.	Eight questions from HAQ (Fries et al., 1982) – higher score (≥0.5), more functional difficulty	Not included: measured as pain presence not severity
Netherlands			Response rate = 78%
Dahaghin et al., 2005b†	Population-based cohort study	Inhabitants of Ommoord aged 55 years and over	N = 3,906.	Eight questions from HAQ (Fries et al., 1982) – higher score (≥0.5), more functional difficulty	Not included: measured as pain presence not severity
Netherlands			Response rate = 35% (only cases with radiographic data at time of analysis are included)
Hill et al., 2007‡	Population-based cohort study	Participants aged 50 years and over, reporting hand pain on a health survey	N = 2,113	Hand and finger function sub-scale of AIMS2 (Meenan et al., 1992) – higher score (>1.5), more functional difficulty	Hand pain sub-scale of AIMS2 (Meenan et al., 1992) – higher score (>3.5), more pain
UK			Two-stage survey:
			Survey 1 response rate = 71%
			Survey 2 response rate = 79%
Baron et al., 1987	Cohort study	Tenants of a senior citizens apartment building aged 60 years and over	N = 32	Questions from HAQ (Fries et al., 1982) on upper extremity activities – higher score, more functional difficulty	Self-reported, data not given
Canada			Response rate = 16%
Dziedzic et al., 2007‡	Population-based cohort study	Participants aged 50 years and over, reporting hand pain on a health survey	N = 2113	Hand and finger function sub-scale of AIMS2 (Meenan et al., 1992) – higher score, more functional difficulty. Severe functional difficulty = top 25% of observed sub-scale	Hand pain sub-scale of AIMS2 (Meenan et al., 1992) – higher score, more pain. Severe pain = top 25% of observed sub-scale
UK			Two-stage survey:
			Survey 1 response rate = 71%
			Survey 2 response rate = 79%
Niu et al., 2003	Cohort study (Framingham Osteoarthritis Study)	Participants aged 70 years or over reporting pain, aching or stiffness on most days in any joint	N = 976	Self-reported, data not given	Number of self-reported painful joints with radiographic OA (defined as a joint with KL grade ≥2)
USA			Response rate unclear
Marshall et al., 2009	Cohort study	Participants aged 50 years and over, reporting hand pain on a health survey and attending a hand assessment clinic	N = 623	AUSCAN function subscale (Bellamy et al., 2002) – higher score, more functional difficulty	AUSCAN pain subscale (Bellamy et al., 2002) – higher score, more pain
UK			Response rate to attend clinical assessment = 46%

Three self-reported measures were used to measure hand function: the Arthritis Impact Measurement Scales 2 (AIMS2) hand and finger function subscale (Meenan et al., 1992), the Australian/Canadian Hand Osteoarthritis Index (AUSCAN) hand and finger function subscale (Bellamy et al., 2002) and the upper limb components of the Stanford Health Assessment Questionnaire (HAQ) (Fries et al., 1982). The AIMS2 and AUSCAN pain subscales were also used to measure hand pain severity, along with the number of painful joints with radiographic osteoarthritis (OA) (Kellgren–Lawrence (KL) grade ≥2) (Kellgren and Lawrence, 1957).

Quality assessment

Complete agreement between reviewers on the quality assessment scores was not obtained initially, but easily resolved. There was no one category on the quality checklist where disagreements between reviewers were more common. In many instances, the reviewer had scored the item as ‘Low to Moderate’, or ‘Moderate to High’ risk of bias, so the consensus process was to aid a clear decision on category allocation, or to resolve issues where aspects of the study had been overlooked or misinterpreted.

Of the six categories on the quality assessment checklist, two were frequently rated as ‘Moderate’ risk of bias: ‘Study participation’ and ‘Measuring and controlling for confounders’ (Table 2). For ‘Study participation’, this was mainly because of the lack of information describing the target population or insufficient evidence that the sample was representative of the target population; for ‘Measuring and controlling for confounding’, this was mainly because of a lack of explanation of why specific confounding variables were chosen, how they fitted into the conceptual model and whether confounders were assessed using validated measures. Reasons for deviating from ‘Low risk’ of bias in the remaining domains included use of non-validated assessment methods and presentation of estimates without quantifying their statistical precision (i.e. lack of confidence intervals). Overall, the quality of the papers included in the review was satisfactory.

Table 2. Quality assessment of review articles

	Study participation	Factor measurement †	Outcome measurement	Measurement and controlling for confounding variables	Statistical analysis
The text in the table indicates the level of risk from bias. Study attrition not evaluated, as studies were not longitudinal in design. † Includes any factor tested for association with level of hand pain or functional difficulty. ‡ Note that Dziedzic et al., 2007 scored ‘Low risk’ of bias for ‘Measurement and controlling for confounding variables’. This study aimed to describe a population of interest, rather than look at association, so the issue of confounding was not relevant.
Dahaghin et al., 2005a	Moderate	Moderate	Moderate	Moderate	Low
Dahaghin et al., 2005b	Moderate	Low	Moderate	Moderate	Low
Dziedzic et al., 2007‡	Moderate	Low	Low	Low	Low
Hill et al., 2007	Moderate	Low	Low	Moderate	Low
Niu et al., 2003	High	Low	Low	Moderate	Low
Baron et al., 1987	Moderate	Low	Moderate	Moderate	Moderate
Marshall et al., 2009	Moderate	Low	Low	Moderate	Low

Main results

Factors tested for association with self-reported hand pain and function could be broadly categorised under six headings: demographic factors, history of previous health conditions, radiographic/clinical evidence of hand osteoarthritis, illness perceptions, self-reported diagnosis, and performance-based measures of hand function (Table 3). Most factors were assessed in a single study, with the exception of age, gender and presence of OA, which were assessed in two, four and five independent studies, respectively. A smaller number of studies assessed associations with self-reported hand pain than with hand function.

Table 3. Factor associations with hand pain and hand function

Factor measurement	Association with hand function									Association with hand pain
Unadjusted OR (95% CI)				Adjusted † OR (95% CI)
† Adjusted for all other factors measured. ‡ Adjusted for age and gender. § Adjusted for age, gender and all other factors measured. ¶ Adjusted for gender. †† Adjusted for age. ‡‡ Adjusted for number of hand joints with radiographic OA. §§ Adjusted for the presence of moderate to severe radiographic OA. ¶¶ Data derived from smaller subset of cases with radiographic data (N = 3,906). ††† Data from Dahaghin et al., 2005a,2005b were derived from the same cohort study, although Dahaghin et al., 2005b was based on a sub-sample of participants with radiographic data. ‡‡‡ Data from Hill et al., 2007 and Dziedzic et al., 2007 were derived from the same cohort study. –, term not entered in the model. NA, data not available for review. DIP, distal interphalangeal joints; IP, first interphalangeal joint; PIP, proximal interphalangeal joints; MCP, metacarpophalangeal joints; CMC1, first carpometacarpal joint; TS, trapezio-scaphoid joint; SD, standard deviation; OR, odds ratio; CI, confidence interval, c.f., compared with; p, p-value; R, Pearson's correlation; t, t-statistic.
Dahaghin et al., 2005a†††
Age 70+ years (c.f. 55–69 years)	6.4 (5.4, 7.6)				4.5 (3.6, 5.6)					NA				NA
Female gender	2.8 (2.4, 3.3)				2.2 (1.7, 2.8)					NA				NA
Manual occupation	2.0 (1.8, 2.3)				1.5 (1.2, 1.8)					NA				NA
Body mass index ≥30 kg/m^2	1.3 (1.0, 1.5)				0.8 (0.6, 1.1)					NA				NA
Self-reported history of:
Rheumatoid arthritis (RA)	6.3 (4.9, 8.1)				3.3 (2.3, 4.7)					NA				NA
Osteoarthritis (OA) in any joint	1.6 (1.4, 1.9)				1.1 (0.9, 1.4)					NA				NA
Diabetes	2.4 (2.0, 3.0)				1.6 (1.1, 2.2)					NA				NA
Stroke	5.2 (4.1, 6.5)				4.8 (3.4, 6.8)					NA				NA
Thyroid disease	2.0 (1.7, 2.3)				1.2 (0.9, 1.6)					NA				NA
Neck & shoulder pain (past month)	2.2 (1.9, 2.5)				1.8 (1.4, 2.2)					NA				NA
Gout	0.9 (0.4, 2.0)				–					NA				NA
Hand/wrist fracture past 5 years	1.8 (1.5, 2.1)				0.9 (0.6, 1.3)					NA				NA
Parkinson's disease	18.4 (10.9, 30.8)				23.8 (11.4, 49.5)					NA				NA
Hand pain (last month)	2.6 (2.3, 3.1)				2.4 (1.9, 3.0)					NA				NA
Radiographic OA	2.1 (1.5, 2.9) ¶¶				1.4 (0.9, 2.0) ¶¶					NA				NA
	Adjusted ‡				Adjusted §
Dahaghin et al., 2005b†††	OR (95% CI)				OR (95% CI)
Radiographic hand OA
KL≥2 in any DIP or IP joint	1.3 (0.9, 1.8)				1.2 (0.8, 1.7)					NA				NA
KL≥2 in any PIP joint	1.1 (0.8, 1.7)				0.9 (0.6, 1.4)					NA				NA
KL≥2 in any MCP joint	2.0 (1.3, 3.0)				1.8 (1.2, 2.9)					NA				NA
KL≥2 at the CMC1 or TS joint	1.3 (1.0, 1.9)				1.2 (0.8, 1.7)					NA				NA
KL ≥ 2 in two hand joint groups	1.5 (1.1, 2.1)				–					NA				NA
KL ≥ 3 in two hand joint groups	1.6 (1.1, 2.5)				–					NA				NA
KL ≥ 4 in two hand joint groups	1.6 (0.9, 2.9)				–					NA				NA
Number of joints with KL ≥ 2	Borderline significant (data not given)				–					NA				NA
Number of joints with KL ≥ 2	1.1 (1.0, 1.2)				–					NA				NA
(Dominant hand only)
KL ≥ 2 in all four hand joint groups	2.7 (1.3, 6.0)				–					NA				NA
Hill et al., 2007‡‡‡	Unadjusted OR (95% CI)				Adjusted † OR (95% CI)					Unadjusted OR (95% CI)				Adjusted† OR (95% CI)
Age
60–69 years (c.f. 50–59)	1.27 (1.03, 1.57)				1.37 (0.98, 1.91)					1.21 (0.98, 1.50)				1.01 (0.80, 1.45)
70 + years (c.f. 50–59)	1.84 (1.49, 2.28)				2.04 (1.44, 2.90)					1.49 (1.20, 1.84)				1.63 (1.20, 2.21)
Female gender	1.88 (1.57, 2.26)				2.02 (1.50, 2.73)					1.18 (0.99, 1.41)				0.89 (0.68, 1.15)
Self-reported diagnosis
RA (c.f. OA)	1.06 (0.82, 1.38)				1.24 (0.84, 1.85)					0.92 (0.71, 1.20)				0.90 (0.64, 1.27)
Other (c.f. OA)	0.76 (0.59, 0.97)				1.28 (0.88, 1.89)					0.57 (0.43, 0.73)				0.59 (0.42, 0.83)
Do not know (c.f. OA)	0.47 (0.36, 0.61)				0.92 (0.61, 1.39)					0.38 (0.29, 0.49)				0.53 (0.37, 0.76)
Frustration with hand problem	8.45 (6.85, 10.44)				4.31 (3.17, 5.86)					9.10 (7.36, 11.26)				4.84 (3.70, 6.34)
Illness perception subscales (Moss-Morris et al., 2002)
Timeline cyclical	p > 0.05				–					p > 0.05				-
Timeline acute chronic	p > 0.05				–					2.51 (2.07, 3.04)				1.41 (1.06, 1.87)
Consequences	1.26 (1.23, 1.29)				1.18 (1.14, 1.23)					1.29 (1.25, 1.32)				1.18 (1.13, 1.22)
Personal control	p > 0.05				–					p > 0.05				–
Treatment control	p > 0.05				–					p > 0.05				–
Emotional representations	p > 0.05				–					p > 0.05				–
Illness coherence	p > 0.05				–					p > 0.05				–
Psychological attribution	p > 0.05				–					p > 0.05				–
Identity	5.34 (4.29, 6.64)				2.32 (1.73, 3.12)					–				–
Baron et al., 1987
Gender	Females more functional difficulty than males (t = 2.35, p = 0.026)									NA
Hand function index [Smith hand function test (Smith, 1973)]	Uncorrelated, but no estimates given									NA
Hand strength index	R = −0.56 (p = 0.001)									NA
Hand pain	R = 0.67 (p < 0.001)									NA
Adduction deformity of the CMC1 joint	R = 0.28 (p = 0.057)									NA
Tenderness on motion	R = 0.33 (p = 0.034)									NA
Clinical OA index	R adjusted for gender = 0.37 (p = 0.03)									NA
Dziedzic et al., 2007‡‡‡	Males				Females					Males				Females
Age	Mean (SD)		Severe function N (%)		Mean (SD)		Severe function N (%)			Mean (SD)		Severe function N (%)		Mean (SD)		Severe function N (%)
50–59 years	1.5 (2.1)		37 (15)		2.3 (2.4)		100 (23)			3.7 (2.4)		47 (19)		3.8 (2.5)		102 (23)
60–69 years	2.0 (2.6)		50 (19)		2.5 (2.3)		109 (25)			4.2 (2.5)		75 (28)		4.1 (2.4)		107 (26)
70–79 years	2.0 (2.6)		40 (20)		3.0 (2.5)		109 (35)			3.9 (2.4)		45 (23)		4.3 (2.4)		86 (28)
80+ years	2.5 (3.0)		13 (27)		4.0 (2.8)		65 (48)			4.2 (2.1)		13 (25)		4.6 (2.7)		49 (38)
Niu et al., 2003
Number (%) of painful hand joints with radiographic OA										Males				Females
0	NA									309 (88)				NA	464 (74)
1	NA									13 (4)				NA	35 (6)
2	NA									7 (2)				NA	29 (5)
3	NA									5 (1)				NA	11 (2)
4	NA									5 (1)				NA	16 (3)
5+	NA									12 (3)				NA	70 (11)
Marshall et al., 2009
Radiographic subgroup	Un-adj mean (95% CI)	Adj ¶ mean (95% CI)		Adj †† mean (95% CI)		Adj ‡‡ mean (95% CI)		Adj §§ mean (95% CI)	Un-adj mean (95% CI)		Adj ¶ mean (95% CI)		Ad j†† mean (95% CI)		Adj ‡‡ mean (95% CI)		Adj §§ mean (95% CI)
No OA	8.3 (6.7–9.8)	8.1 (6.6–9.7)		8.7 (7.1–10.3)		8.9 (7.1–10.7)		8.9 (7.1–10.6)	5.4 (4.6–6.2)		5.4 (4.6–6.2)		5.4 (4.5–6.2)		5.7 (4.7–6.6)		5.8 (4.9–6.7)
Finger only OA	8.2 (6.3–10.1)	8.2 (6.4–10.1)		8.3 (6.4–10.2)		8.6 (6.6–10.6)		8.6 (6.6–10.5)	5.7 (4.7–6.8)		5.7 (4.7–6.7)		5.7 (4.7–6.8)		5.9 (4.8–6.9)		6.0 (4.9–7.0)
Thumb only OA	8.6 (6.9–10.3)	8.1 (6.5–9.8)		8.7 (7.0–10.4)		9.0 (7.2–10.8)		8.8 (7.1–10.5)	5.8 (4.9–6.7)		5.7 (4.8–6.6)		5.8 (4.9–6.7)		5.9 (5.0–6.9)		5.9 (5.0–6.8)
Combined thumb and finger OA	10.5 (9.6–11.4)	9.9 (9.0–10.8)		10.3 (9.4–11.2)		10.1 (9.0–11.2)		10.4 (9.4–11.3)	6.5 (6.1–7.0)		6.5 (6.0–7.0)		6.5 (6.1–7.0)		6.4 (5.8–6.9)		6.4 (5.9–6.9)
p-value for overall association	0.018	0.084		0.093		0.601		0.206	0.077		0.095		0.091		0.698		0.573

Factors significantly associated with limited hand function were older age; female gender; manual occupation; neck or shoulder pain; clinical and radiographic OA (although evidence depended on the definition of OA); weaker hand strength; hand pain; history of Parkinson's disease, stroke, diabetes or rheumatoid arthritis; and illness perceptions (namely frustration, impact and symptom count). Key factors associated with hand pain severity were age, impact, frustration, patient expectation of a long disease time course, and self-reported diagnosis of the cause of the hand problem (Table 3).

Meta-analysis

A meta-analysis was not conducted to pool estimates across studies because the factors measured in each study differed greatly and the statistical methods used to describe associations were not consistent (e.g. odds ratio versus mean difference versus correlation).

Discussion

The present review has summarized the results of five cross-sectional studies (seven published articles) that have investigated factors associated with the severity of hand pain and functional difficulty in general population samples of older adults. Factors associated with hand pain and/or function included generic factors (e.g. age, gender), those related to previous clinical conditions (e.g. stroke) and hand-specific factors (e.g. illness perceptions and radiographic/clinical hand osteoarthritis).

Age was the only factor for which a (cross-sectional) ‘dose–response’ (Woodward, 1999) relationship with outcome was tested (i.e. the severity of hand pain and function progressively worsened with increasing age). This observation is in line with findings from both population and clinical studies that the prevalence of hand pain, pain interference and functional difficulty increases with age (Jones et al., 2001; Palmer, 2003; Thomas et al., 2004).

Some of the strongest predictors of hand function found in this review relate to disease history (e.g. history of Parkinson's disease or stroke). The prevalence of such conditions in the population is likely to be low, which may lead to unreliable estimates of association, although it has been shown in several clinical studies that such conditions are related to impaired hand function (e.g. Hunter and Crome, 2002; Cano-de-la-Cuerda et al., 2010).

Only three (Dziedzic et al., 2007; Hill et al., 2007; Marshall et al., 2009) out of the seven papers in the review examined both hand pain and function, thus allowing direct comparisons to be made. In these studies, the findings were similar for pain and function. There were some notable exceptions to this: hand pain but not hand function was related to self-reported diagnosis of the cause of the hand problem and to patient expectation of a long disease time course (Hill et al., 2007). This may reflect patients' focus on the absence of pain as the main sign of recovery from their hand condition, or that pain may encourage consultation to receive a clinical diagnosis. It may also be that receiving a medical diagnosis may be associated with more pain perceptions, or that some diagnoses reflect more painful conditions.

Moreover, Hill et al. (2007) also showed gender differences in hand function severity that were not observed for hand pain. The prevalence of upper limb musculoskeletal pain has been shown to be higher for women (Walker-Bone et al., 2003), and female gender is a risk factor for many common hand conditions (Hart and Spector, 2000; Walker-Bone et al., 2003). This may suggest that a more complex relationship between hand pain and function and gender may exist which could be explained by other external factors, such as ability to cope and adapt to limited hand function (Myers et al., 2008).

A key aim of this review was to identify factors that predict the clinical course of hand pain and hand function over time in population-based studies of older adults. Despite a comprehensive and inclusive search strategy, no studies on the progression of hand problems were identified. Of the cross-sectional factors identified, many cannot be modified by treatment (e.g. age, gender and occupation), or relate to disease history that cannot be altered at the point of consultation. Illness perceptions, however, have the potential to be modified and have been identified as important predictors of outcome in studies of consulters with hand pain (Spies-Dorgelo et al., 2008) and back pain (Foster et al., 2008; Macfarlane, 2008).

Strengths and limitations

A major strength of this review was the comprehensive and inclusive search strategy that was developed to minimize the risk of missing key articles. This was achieved by searching in multiple health care databases and tailoring searches to apply directly to the particular databases' indexing method. At each stage in the review, methods were piloted and key decisions on abstract inclusion, quality assessment and data extraction were derived by consensus, improving the quality of the data reported.

Selected databases included conference abstracts and other non-journal articles [HMIC and Web of Knowledge (Centre for Reviews and Dissemination, 2009)]; however, only full journal articles were included in the review. Searching of grey literature or unpublished studies was not undertaken. It might be speculated that as the number of published studies in the review is low, the number of extra studies identified by this method would also be small. Articles written in English were selected electronically and included in the review. Only a small percentage of all articles found in the search were written in other languages, so it is unlikely that this number would bias the results of the review (Centre for Reviews and Dissemination, 2009).

The search was focused to include only self-report measures of pain and function, so studies measuring hand function using clinical tests alone, such as grip strength and timed performance tests, were not included. Self-report measures, although potentially prone to recall bias, were chosen, as they are frequently used in population-based surveys to fully capture the range of limitations experienced during everyday activities (Jordan et al., 2009).

A further limitation of the study was that three out of the seven papers in the review were co-authored by one or more of the research team involved in this review. However, to ensure that the quality assessment process was impartial and unbiased, quality assessment was only completed by researchers not named as co-authors on the papers that they scored. When constructing the review, we anticipated that more studies would be identified from other research teams, but this was not found.

Implications

All associations described in this review have used cross-sectional data, so further evidence is needed to determine if similar factors would also predict change in hand pain and functional difficulty over time. In addition, several studies in the review presented unadjusted associations of predictors with outcome, which may not remain clinically important or statistically significant after adjustment for other theoretically plausible variables. As many of the associations described in this review are derived from individual cohort studies, there is the potential that such findings will not be replicated in other studies. This implies that further cohort studies are needed, so that the strength of association can be determined from meta-analysis to pool estimates across multiple studies, rather than relying on a single study as the evidence base.

Future work is also needed using data from cohorts that meet our inclusion criteria, so that prognostic indicators found important in other clinical populations or pain sites can be tested. For example, Spies-Dorgelo et al. (2008) identified female gender, older age, longer symptom duration, low baseline pain intensity, low coping score, lower personal control and high somatization as predictors of the short- and long-term clinical course of hand and wrist problems presenting to primary care in adults 18 years and over. These factors, along with anxiety, depression and obesity, were also shown to be key prognostic factors for musculoskeletal pain in primary care and for older adults with knee pain (Mallen et al., 2007a,2007b) and so are plausible factors to explain the clinical course of hand problems in older adults in the general population and, as such, require further exploration.

Conclusions

Both demographic and clinical factors have been shown to be associated with level of hand pain or functional difficulty in older adults selected from the general population. However, evidence for such associations is based on a small number of cross-sectional studies, making conclusions tentative until further replication has been achieved. No studies on the progression of hand pain and functional difficulty were identified in this review, so the ability of such factors to predict future outcome in this population is yet to be determined.

Acknowledgements

We would like to thank Rachel Gick (Keele University health librarian) for help with the search strategy. This work was supported by the Arthritis Research UK (Program Grant Number 18174).