Introduction

Hip fracture among older adults has been associated with a 20% 1-year mortality rate and results in significant long-term functional limitations for many survivors [1, 2]. The national Irish Hip Fracture Database (IHFD), established in 2013, audits specific acute standards of care across 16 acute hospitals in Ireland for up to 3700 older patients annually [3]. In the UK, achievement of similar acute hip fracture care standards has been found to be associated with improved patient quality of life at four months [4]. The relationship between care quality and longer-term hip fracture outcomes after discharge from acute hospital is unknown in Ireland [3]. Furthermore, data monitoring in the Irish healthcare setting is limited by the absence of unique health identifiers and registry linkage [5]. The IHFD therefore aims to facilitate and encourage recording of these outcomes in the Irish setting [6].

The international Fragility Fracture Network (FFN) has recently recommended a minimum common dataset for hip fracture audit in order to support international comparisons. It includes the long-term outcomes of mortality, re-operation, mobility, residence, and bone protection medication [7]. They suggest that while 120 days post-fracture is the ideal time to collect these outcomes to show meaningful recovery, 30 days after fracture may be more feasible for many healthcare systems [7]. These outcomes have been informed by a systematically developed core outcome set for hip fracture trials, which additionally includes quality of life using the EQ-5D patient reported outcome measure (PROM) [8]. National hip fracture audits in Norway, the UK, Germany, Australia, and New Zealand have adopted the EQ-5D at 120 days, although completeness of data has been an ongoing challenge [9, 10]. A Core Outcome Set for orthogeriatric management also advocates outcome assessment at 1 year after hip fracture, and 1-year mortality is a commonly assessed outcome in much prognostic research [11,12,13].

As high-quality and internationally comparable data are required to inform decision-making, it is essential that the implementation of long-term outcome recording after hip fracture in the Irish setting receives adequate preparation and support. In advance of piloting any specific outcome fields in the IHFD, it is important that existing practices are described and understood. This includes identification of specific long-term hip fracture outcomes and collection methods that have been used in local settings, as well as determinants of successful collection within the Irish health system [14]. While many systematic literature reviews limit findings to full-length peer-reviewed publications, the inclusion of conference abstracts and academic theses has the potential to ensure that the maximum amount of relevant information is captured. Furthermore, a broad systematic review could help to identify authors who have experience conducting long-term outcome collection in this patient population within the country, enabling future exploration of implementation determinants [14].

The aim of this systematic review was to identify, describe, and appraise studies that have collected long-term outcomes after hip fracture in Ireland. Objectives include the following: (i) to describe outcomes, time points, and methods of outcome collection used in research after hip fracture in Ireland; (ii) to identify common sources of bias; and (iii) to summarise results for specific outcomes in studies generalisable to the broad older Irish hip fracture population.

Methods

Study design

A systematic review of studies that have collected long-term outcomes after hip fracture in Ireland was conducted. The systematic review was registered with PROSPERO (CRD42021265127), and the full protocol has been published [15].

Eligibility criteria

Studies were included if:

  • Data were collected in the Republic of Ireland and studies were published since 2005.

  • Patients were selected due to hip fracture diagnosis. Studies focussing on people under the age of 60 or only periprosthetic fractures were excluded.

  • One or more long-term outcomes were collected. This was defined as any data collected after discharge from an inpatient setting or at a fixed time point at least 30 days after fracture, admission to hospital, or surgery. Studies that collected only inpatient data were excluded.

  • They were published in Irish or English as a journal article, conference abstract, academic thesis, or report. Where the same study was duplicated in different publication types, the most detailed report was included.

Search strategy

Traditional and grey literature sources were searched in July 2021, and the search was updated in April 2022. Electronic databases searched included MEDLINE (Ovid), Embase, Scopus, Web of Science, CINAHL, and ProQuest Dissertations and Theses. Specific Irish research repositories were also searched (www.rian.ie and www.lenus.ie). This was supplemented by an advanced Google search of specific Irish governmental and health-system websites. Ongoing and unpublished research was sought through websites of national funding organisations and trial registries. Email contact was made with programme leads for 25 taught masters programmes across all major academic institutions that were likely to have supervised relevant research spanning the subjects of surgery, nursing, and rehabilitation. Finally, publicly available abstract books from key conferences in the areas of gerontology, orthopaedics, and rheumatology were hand-searched if not indexed in databases. Search terms used were related to hip fracture and locations in Ireland. The full search strategy for all sources has been previously published [16].

Study records

Selection process

All records from database searching were uploaded to Covidence Systematic Review Software [17], and duplicates were removed. Titles and abstracts were screened by two review authors (MW and CC or LB or BS). Full texts were retrieved for all potentially eligible studies. Two authors (MW and LB) reviewed each full text, and reasons for exclusion were recorded. Where disagreements arose, these were resolved through discussion.

Data items

The following data items were extracted from each study:

Study characteristics: first author, publication year, publication type, clinical site, number of participants, demographics of included sample (mean age, % female), dates of recruitment/patient hospital admission, and outcomes collected

For each outcome collected: method of data collection, specific outcome definition, time points collected, and results of outcome at each time point

Additional risk of bias items: sampling method and follow-up rate

Risk of bias in individual studies

The risk of bias assessment included five questions, four of which were used in a similar review of long-term disability outcomes after hip fracture covering international literature [18]. All included studies were appraised independently by two review authors (MW and LB or BS or MF). Where disagreements arose, these were resolved through discussion. Questions were as follows:

  1. 1.

    Is it a representative sample? (Yes if recruitment was consecutive or random)

  2. 2.

    Were patients followed from inception? (Yes if the baseline time was same for all patients and close to fracture time)

  3. 3.

    Is it a clearly defined sample? (Yes if inclusion defined by hip fracture hospital diagnosis with an age limit or age range provided in results)

  4. 4.

    Was there adequate follow-up? (Yes if long-term outcomes were collected for at least 80% of recruited/ identified participants)

  5. 5.

    Are outcomes clearly defined? (Yes if valid objective/broadly accepted measure was used)

A further question was asked to determine the applicability of the study to the Irish Hip Fracture Database cohort and thus the focus of the systematic review: Are inclusion criteria generalisable to a broad older hip fracture population? (Yes if the inclusion criteria were not narrowly limited to a specific age cohort, sex, fracture type, procedure type, or setting). Studies judged to meet this criterion will be referred to as ‘generalisable studies’.

Data synthesis

Narrative synthesis

Detailed characteristics of studies were narratively synthesised and presented in a table according to outcomes collected, the method of data collection, outcome definition used, and time points.

Meta-analysis

Meta-analysis was considered where the same outcome was reported at the same time point in multiple generalisable studies. To minimise the risk of biased estimates, studies where years of patient recruitment/hospitalisation were not reported, where patients were not followed from inception, or where sampling was not random or consecutive were excluded from the meta-analysis. Where more than one study was conducted at the same clinical site during overlapping time-periods, only one study (with maximum coverage) was included in the meta-analysis. As outcomes eligible for meta-analyses were dichotomous, the metaprop command in Stata version 17 was used [19]. A random effects model was specified using the DerSimonian and Laird method [20]. Pooled estimates and 95% confidence intervals (CIs) were calculated after Freeman-Tukey Double Arcsine transformation, as is recommended for meta-analyses of dichotomous outcomes in order to stabilise the variances [21, 22]. Results were presented with forest plots, and heterogeneity was assessed with the I2 statistic.

Results

Study identification

From electronic database searching, 72 records were deemed to be eligible. An additional 12 records were identified from other sources. A flow diagram of the search is presented in Fig. 1.

Fig. 1
figure 1

Flow diagram of included studies

Full size image

Study characteristics

The 84 studies were conducted in 20 clinical sites across Ireland. This includes 15 of 16 acute hospitals which operate on hip fracture and are represented in the national Irish Hip Fracture Database. Five of these hospitals accounted for 75% of the studies. In terms of design, n = 40 studies (48%) were retrospective cohort studies, 32 (38%) were prospective cohort studies, 7 (8%) were cross-sectional studies, and 5 (6%) were randomised trials. Studies included a median of 165 cases of hip fracture (range 9 to 2483 cases) and included patients admitted with hip fracture over a median of 24 months (IQR 12 to 48 months). While studies were published between 2005 and 2022, those that reported study dates (n = 69, 82%) included patients hospitalised with hip fracture between 1994 and 2020. More studies were from recent years however, with 54% (n = 45) published after 2016. A total of 51 studies (60.7%) were deemed to be ‘generalisable’ to a broad hip fracture population. The 33 studies (39%) deemed to be non-generalisable primarily explored specific fracture types treated with specific procedures (n = 12). Others were limited to patients in a specific age category, with specific co-morbidities or only those discharged to nursing home or inpatient rehabilitation. Where reported, the mean age of participants in generalisable studies ranged from 76 to 82 (n = 31 studies, 61% of generalisable), and the proportion of female participants ranged from 64% to 82% (n = 34 studies, 67% of generalisable). Table 1 shows outcomes collected, methods of data collection, specific outcome definitions used, and the time points after fracture at which outcomes were collected. The full list of included studies is presented in the Electronic Supplementary Material under outcomes measured and generalisability.

Table 1 Details of outcomes collected in included studies (n = 84 studies)
Full size table

Outcomes collected

Mortality

Over half of all studies included (n = 48, 57%) collected mortality as an outcome with 22 of these reporting mortality as the only long-term outcome. Mortality was most commonly reported at one year. Overall, 18 studies did not report the method used to verify death. Ten studies reported using multiple overlapping methods detailed in Table 1, while 20 others used one of these methods in isolation. While the actual mortality rate was reported in most studies, several reported only statistical associations with mortality.

Mobility and function

Over a third of studies (n = 32, 38%) reported collecting a measure of mobility or function at least 30 days after hip fracture. The most common fixed time point reported was 1 year after fracture for both mobility and function. For mobility, a description of the aid and assistance required to mobilise was given most commonly (n = 11 studies). Four additional studies reported using the New Mobility Score and one using the Parker Palmer score, both of which are based on similar mobility descriptions. There was much heterogeneity in other measures of function and mobility used (see Table 1). A total of 13 studies (40% of studies with these outcomes) did not report how they collected data. The majority that did report the method contacted patients or caregivers by telephone.

Residence

Residence was reported by 20 studies (24% of included), most commonly collected at 1 year after fracture by patient telephone contact. Most studies reported the number and/or percentage of patients who were living at home or in a nursing home before their hip fracture and at follow-up. In many cases, it was unclear if the denominator for reported percentages at follow-up included patients who had died since their hip fracture.

Bone health–related outcomes

While 20 studies (24%) reported outcomes related to bone health, 12 of these (60%) were from a single hospital and several were cross-sectional rather than longitudinal studies. Studies from that specific hospital reported a wide range of bone health-related measures collected during face-to-face clinics including prescription of bone protection medication, receipt of DXA scan, diagnosis of osteoporosis and osteopenia, vertebral fractures, subsequent fractures, levels of vitamin D, calcium and parathyroid hormone, T-scores, bone mass density, and other bone markers. Studies from the seven other clinical sites primarily reported prescription and persistence with bone protection medication at 1 year after fracture, receipt of DXA scanning, and new fractures, mostly identified based on health records.

Re-admission, re-operation, and complications

A total of 18 studies reported an outcome of hospital readmission (n = 9), re-operation (n = 5), or other complication (n = 5). In many cases, it was unclear if readmission included only those that returned to the operating hospital. Surgical complications and revision rates were mostly reported in non-generalisable studies that focussed on specific fracture types or surgical procedures and followed patients for time-periods longer than 1 year.

Other outcomes

Other common long-term hip fracture outcomes reported included pain (n = 5) and quality of life (n = 5). Outcomes collected by fewer than five studies are listed in Table 1.

Risk of bias in individual studies

Figure 2 shows a summary of the methodological quality of included studies. Most studies followed patients from inception (79%), were consecutive samples (66%), and clearly defined the outcomes that were collected (67%). Almost half of studies did not provide an age range to define the sample; however, all reported focussing on older adults. An area for concern is that only a third of studies (n = 28) reported follow-up rates of better than 80% and most studies did not clearly report their follow-up rate (n = 45, 54%).

Fig. 2
figure 2

Summary of risk of bias assessment (n = 84 studies)

Full size image

Meta-analysis of outcomes

Two meta-analyses were performed where multiple eligible studies reported the same outcome. A meta-analysis of 1-year mortality was conducted including 12 studies with a total of 4220 patients, and a meta-analysis of 30-day mortality was conducted including 7 studies with a total of 2092 patients. Table 2 shows the characteristics of each study included in the meta-analyses including reported follow-up rates. Figures 3 and 4 show forest plots with results of meta-analyses. One study reporting 1-year mortality contained two distinct cohorts from different time periods, and these were treated separately in the meta-analysis [23]. The pooled estimate for 1-year mortality was 24.2% (95% CI 19.1% to 29.8%); however, there was high heterogeneity (I2 = 93.8%), with earlier studies showing higher mortality. The pooled estimate for 30-day mortality was 4.7% (95% CI 3.6% to 5.9%, I2 = 31.3%). Studies reporting non-mortality outcomes were unable to be meta-analysed due to heterogeneity of measures or lack of clarity around definitions or descriptions of the patient cohort.

Table 2 Characteristics of studies included in meta-analyses (n = 16 cohorts in 15 studies)
Full size table
Fig. 3
figure 3

Forest plot for meta-analysis of 1-year mortality rate: Legend shows first author and years of patient hospitalisation with hip fracture

Full size image
Fig. 4
figure 4

Forest plot for meta-analysis of 30-day mortality rate: Legend shows first author and years of patient hospitalisation with hip fracture

Full size image

Discussion

This study has described and appraised a broad body of literature regarding research and audit in the Republic of Ireland published since 2005. It has revealed recent practices in long-term outcome collection and recording after acute hip fracture at a local level, covering most clinical sites that operate on hip fracture in Ireland. The study has several implications for future planning in national hip fracture audit within Ireland. It can also provide a template for similar countries or jurisdictions that currently lack infrastructure to link audit data to civil registers or that aspire to record patient reported outcomes after hip fracture in line with international recommendations [7].

Outcomes collected as part of recent hip fracture research in Ireland are broadly in line with those suggested by the international Fragility Fracture Network minimum common dataset for hip fracture audit, namely mortality, re-operation, mobility, residence, and bone protection medication [7]. There were varying time points of outcome collection in studies included in the current review. The FFN suggest 120 days post-fracture as ideal, but 30 days after fracture is likely to be more feasible [7]. Many studies in Ireland however collected outcomes at one year. This is in keeping with a published core outcome set for orthogeriatric management of hip fracture and is the time point at which mortality is recorded in the Danish national audit [11, 24]. While the 1-year time point appears to be important to Irish researchers in this area, its feasibility to be used as part of routine hip fracture care within the Irish healthcare setting is unknown.

All-cause mortality is undoubtedly an important outcome after hip fracture, and this view is reflected by the number of Irish studies collecting and reporting these data. Accurate ascertainment is however a challenge in a setting where automatic linkage to registries is not available [5]. Many studies in this review described using overlapping methods or manually searching data from the General Register’s Office to determine vitality status for each patient. This may be feasible but also burdensome in a routine clinical setting. Several studies did not report their methods of ascertainment or the rate of successful follow-up however, which limits our ability to assess the validity of estimates. The pooled estimate for one-year mortality (24.2%) was similar to rates estimated by a recent international systematic review [2] and those reported by the Danish national audit, which has the benefit of unambiguous registry linkage [24]. The pooled estimate for 30-day mortality (4.7%) was however low in comparison to robust sources in other similar European countries, which range from 5.5 to 9.5% [24]. It is unclear whether this represents a true difference or if it is an indication of under-estimation. Of note, 30-day mortality estimates were only obtained from abstracts and one thesis, with these figures not available in included full-length articles. More limited reporting in abstracts may make it difficult to understand potential sources of bias in detail.

The variation in methods of measuring mobility and function found by this review was notable. This heterogeneity has also been found in international work and has been described as a barrier to achieving successful comparisons in hip fracture research [18, 25]. In an international review of hip fracture outcomes, Dyer and colleagues estimated that 40–60% of people with hip fracture recover their pre=fracture mobility within 1 year. We were not able to estimate this proportion in the current study. Many studies used descriptions related to the level of assistance and walking aids needed to mobilise in different settings. The New Mobility Score (NMS) uses similar descriptions but applies numerical scores for indoor, outdoor, and shopping mobility [26]. It is a reliable and valid measure that has been recommended for use in post-acute and community settings by a recent Clinical Practice Guideline for Physical Therapy management of older people after hip fracture [27]. It has been included in the Irish Hip Fracture database since 2016 as a measure of pre-fracture mobility, and since 2018, completeness of the measure has been over 90% [1]. Repeating the NMS at fixed time points after hip fracture in Ireland would allow estimations of return to pre-fracture mobility, an outcome which has been reported as very important to patients [28].

Reoperation has traditionally been a common way of reporting outcome after hip fracture surgery, and it is recommended by the FFN minimum common dataset [7, 9]. More recent research and core outcome sets have however focussed more on functional and patient-reported outcomes [8, 9]. In the current systematic review, 10 studies were identified that recorded reoperation after hip fracture in Ireland; however, the majority of these studies focussed on specific fracture types and surgical procedures. Ascertainment methods reported also frequently involved access to and interpretation of surgical and radiological records. This highlights that accurate collection of reoperation as an outcome in the current Irish setting may require input from orthopaedic surgeons or specially trained professionals. The feasibility of this practice in more generalisable samples of patients with hip fracture requires further investigation. All-cause hospital re-admission may be easier to capture and for patients to self-report. It has been included in a core outcome set for orthogeriatric management of hip fractures and is collected as an outcome in several European hip fracture audits [11, 24].

A small minority of Irish studies collected quality of life, and none used the EQ-5D measure that is recommended by the hip fracture core outcome set and integrated into several hip fracture audits internationally [8, 10]. This is an important finding as quality of life has been suggested as an outcome that could capture what is important to patients [28]. A subjective report of pain, which is a component of the EQ-5D, was reported by a further three studies. The feasibility and acceptability of collecting quality of life as an outcome within the Irish setting require investigation.

Strengths and limitations

Including grey literature and conference abstracts in this systematic review was a deliberate decision and has allowed us to maximise learning from local studies. While it has also likely led to including more studies with poor reporting, we were able to conduct two meta-analyses of generalisable and representative studies. Despite the lack of clarity around follow-up rates in some studies, pooled estimates were largely in line with European findings based on robust sources. Our quality appraisal does highlight the need for an increased focus on reporting guidelines for authors submitting to conferences, especially considering the increased indexing and availability of these reports. To aid data extraction and usefulness, authors should clearly describe their participant samples, report numbers along with percentages, and report follow-up rates.

While many of the studies were inception cohorts, representative of all patients admitted with hip fracture and covered several years, it is unknown if data were originally collected specifically for a research study/audit activity or as part of routine follow-up care. Most studies contacted patients by telephone, sometimes as part of a ‘virtual clinic’. This may be a feasible way to follow-up patients after hip fracture in the Irish setting, especially as high non-attendance rates at routine orthopaedic and geriatric outpatient clinics among this population has been noted [29]. The staffing resources required to conduct telephone follow-up to a high standard is unknown and requires further investigation. While 16 acute hospitals operate on hip fracture in Ireland, three quarters of the studies in this review were based in only 5 hospitals. This study may therefore overrepresent centres with higher capacity and resources for research. It is also likely that this review did not capture other follow-up activity that is taking place as part of routine care in Ireland but has not been the subject of published audit or research. Further work is required to establish current routine activity across the country but also the challenges associated with the practice. An in-depth qualitative study is currently being conducted by the authors of this systematic review to explore healthcare professionals’ experiences of long-term outcome collection after hip fracture within the Irish setting. While current core outcome sets have been developed based on qualitative research with patients and family members in the UK [28], acceptability of follow-up and outcome collection methods should be explored with persons who have experienced hip fracture in Ireland.

Conclusion

Collection of long-term outcomes has been reported at most clinical sites that operate on hip fracture in Ireland since 2005. The choice of outcomes used within Irish research studies has been in line with international recommendations. Estimates of mortality rates after hip fracture in Ireland are also similar to European findings. Heterogeneity of specific measures for non-mortality outcomes and poor reporting of methods and findings hinders our ability to meaningfully synthesise results. Standardised recommendations around outcome definitions and time points are warranted as well as a mechanism to aggregate data nationally. Qualitative research is currently ongoing to identify barriers to conducting long-term outcome recording as part of routine hip fracture care across the country.