Comparison of different strategies of referral to a fall clinic: How to achieve an optimal casemix?

Abstract

Objective

To study the potential differences in patient characteristics between two referral methods to a fall clinic, specifically: case-finding of patients admitted to an emergency department because of a fall, compared to direct referral to the fall clinic via the general practitioner.

Design

Cross-sectional study.

Setting

Fall clinics in two university teaching hospitals in the Netherlands.

Participants

Three hundred community-dwelling older people aged 65 years or over currently attending the fall clinics in Nijmegen (Group 1, n=154) and in Amsterdam (Group 2, n=146).

Measurements

Patients were referred by a general practitioner (Group 1) or were selected using the Carefall Triage Instrument (CTI) after visiting the emergency department (Group 2). In all patients, modifiable risk factors for recurrent falls were assessed.

Results

Group 1 had less modifiable risk factors for falling (a mean of 4 (SD 1.6) vs. a mean of 5 (SD 1.5) in Group 2, p<0.001). Compared to Group 2, Group 1 had more prevalent ‘recurrent falling (≥2 falls)' (p=0.001) and ‘assisted living in homes for the aged' (p=0.037). ‘Fear of falling', ‘mobility and balance problems', ‘home hazards' and ‘osteoporosis' were significantly less prevalent in Group 1.

Conclusion

This study suggests that patients referred to a multidisciplinary fall prevention clinic by their general practitioner have a different risk profile than those selected by case finding using the CTI. These differences have consequences for the reach of secondary care for fall-preventive interventions and will probably influence the effectiveness and efficiency of a fall prevention program.

Introduction

Falling in the elderly is an important health problem that has a high prevalence and can potentially result in functional decline. Between 22 and 40% of all community-dwelling people aged 65 years or older fall at least once a year (1, 2, 3). With an aging population, the absolute number of people who sustain a fall and concomitant injury will increase. In the Netherlands, 3% of all yearly visitors to the emergency department (ED) above 65 years of age visit because of fall-related injuries, which equals, 66,000 people out of over 2 million in total (4).

The causal pathway of falling among older people is dominantly multifactorial (5). Thus, the risk of falling is associated with many risk factors and increases as the number of risk factors increases (6, 7). Risk factors are commonly divided in intrinsic (patient-related) and extrinsic (situated in the environment) factors of falls. This distinction is important for evaluating the recurrence risk and to develop strategies for secondary prevention. Both general practitioners (GPs) and geriatricians are involved in case finding and (secondary) prevention of falls and fall-related injuries in older individuals. Although several guidelines have been developed for the prevention of falls in older people (8), most general practitioners currently do not provide standardized multidisciplinary risk factor assessment. Additionally, (modifiable) determinants of falls are mostly not assessed directly in the ED.

In the 23 fall prevention clinics in the Netherlands, referrals are usually from GPs or the patients are selected after attending an ED. Both recruitment strategies may be efficient because multiple risk factor assessment and subsequent interventions may reduce monthly falls by 30% (9, 10). However, trials show conflicting outcomes when the effect of multifactorial prevention strategies is studied (11, 12, 13, 14, 15). These differences are probably directly related to the casemix included in these trials and thus are dependent on the system of recruitment and referral. In the UK, many different methods of entry into a fall service are used and only a minority of services (22%) is using a validated screening tool (16). In the successful PROFET study community living older people presenting to a hospital emergency department after a fall were recruited by systematic identification by a geriatrician, excluding patients with cognitive impairment (11). An unsuccessful Dutch trial was different in the selection method by including participants attending the GP-Cooperative instead of only the ED and lacking the systematic identification by a geriatrician (14). For instance, the percentage of recurrent falls was only 27% in the PROFET study versus 49% in the Dutch trial, pointing out the difference in the target population. Controversial outcomes of fall-clinics suggest that selection of the optimal target population is essential for the realization of a reduction of falling after a visit to the fall-clinic. This population should probably have a considerable risk (not too high or low) of recurrent falling and sufficient modifiable risk factors in order to realize a cost-effective intervention (17, 18). In contrast, triage on specific single risk factors (e.g. balance problems causing falls) may allow for selection of a population that might benefit most from a single component intervention, such as Tai Chi training (10, 19, 20). Therefore, the translation of effective recruitment strategies in clinical trials to efficient referral methods in clinical practice probably holds the key to cost-effectiveness of health services for prevention of recurrent falling and fall-related injuries in the elderly.

Therefore, we decided to study whether there are differences between patient characteristics of two groups visiting an academic multidisciplinary fall prevention clinic. The visits were based on two regularly applied but different methods of referral: direct referral by general practitioners (GPs) versus referral by selection by hospital professionals after attending the emergency department because of a fall.

Methods

Study design and participants

We used a cross-sectional study design to compare two methods of referral to an academic fall clinic. The study population consisted of a convenience sample of 300 community-dwelling older people = 65 years of age, who were referred to the fall clinic of the Radboud University Medical Center in Nijmegen (Group 1, consecutive referrals from March 2006 - August 2008, n=154) and to the Academic Medical Center in Amsterdam (Group 2, consecutive (self)-referrals January 2005 - July 2008, n=146; see table 1).

Table 1.

Baseline Characteristics of Group 1 (Nijmegen, referral by general practitioner) and Group 2 (Amsterdam, emergency ward triage)

Variable	Group 1 (n=154)	General population Nijmegen	Group 2 (n=146)	General population Amsterdam	P-Value
Number of persons		13.1		12.4
above 65 years (%)
Female, n (%)	112 (72.7)	59.5	112 (76.7)	59.7	0.51
Male, n (%)	42 (27.3)	40.5	34 (23.3)	40.3
Age, mean ±SD	79.3 (8.6)		78.5 (8.5)		0.33
BMI, mean ±SD	26.1 (5.2)		26.5 (5.1)		0.95
Education, lower level,%	85.8		77.4		0.10
Marital status, %					0.08
-married	32.5	50.2	40.4	42.9
-single	15.5	49.8	19.2	57.1
-widowed	52.0	31.3	40.4	31.3
Institutionalised, %	18.2	27.2	8.2	26.8	0.037
		(80+year)		(80+year)
Katz ratio ±SD	0.32 (0.2)		0.32 (0.2)		0.87
Polypharmacy, %	85.7		79.3		0.14
Experienced falls <2, %	9.7		24.8		0.001
Physical activity daily, %	29.4		34.2		0.51
Cognitive impairment, %	31.2		0-

In Group 1, the patients were referred based on the clinical judgment of their GP about whether treatment of modifiable risk factors for falling would be feasible. The GPs received no education or training. The GPs were only informed about the clinical work-up to be expected from the fall clinic and its referral indication: age 65 and over and a minimum of one unexplained fall in the past year. Before visiting the fall clinic, patients completed the CAREFALL Triage Instrument (CTI), a self-administered fall-history questionnaire used to determine modifiable risk factors for falling and fractures (21). Patients with cognitive impairment were not excluded from visiting the fall clinic in this group. All patients were included regardless the number of risk factors by the CTI.

In Group 2, the patients were sent the CTI within two weeks after visiting the emergency department (ED) for a fall or fall-related injury. Response rate of the CTI rate was 60%, and 70% after an extra reminder by telephone. Patients with sufficient modifiable risk factors were identified, meaning that when a person had three or more risk factors, he was invited to the multidisciplinary fall prevention clinic. Patients with known dementia were not invited because fall prevention interventions were not considered to be evidence-based in this group (22).

The CTI

The CTI is a 44-item fall history questionnaire proven to be reliable and valid in assessing (modifiable) risk factors associated with recurrent falls (21), though not specifically validated in subjects with cognitive decline. Modifiable risk factors were defined as risk factors that can be improved or removed by an intervention. However, international definitions and consensus on which risk factors should be regarded as modifiable are currently lacking. The CTI specifically identifies nine modifiable risk factors for falls or fall-related injuries: polypharmacy, osteoporosis, balance and mobility disorders, mood disorders, fear of falling, impaired vision, urinary incontinence, orthostatic hypotension and reporting hazards at home. The definitions of the nine modifiable risk factors are listed below.

1. Polypharmacy: using 3 or more drugs, independent of the type of medication. 2. Orthostatic hypotension: one or more of the 9 CTI questions on orthostatic hypotension was positive. 3. Mobility and balance problems: difficulties walking and/or use of an aid for walking and/or a lack of balance and/or pain in feet or legs and/or reduced feeling in feet or legs and/or reduced strength in one or both feet and/or stiffness of the joints. 4. Fear of falling: a score of 5 or more on a scale from 1 (no fear of falling) to 10 (severe fear of falling) on the CTI question: „Are you afraid to fall?“. 5.Visual disturbance: unable to read the newspaper, even with (magnifying) glasses or a loupe and/or substantially reduced eyesight since the last 6 months. 6.Urinary incontinence: daily problems with urinary continence and/or need to get out of bed twice or more per night to visit the toilet. 7.Home hazards: positive when fall was due to tripping at home and one or more of the three CTI questions on home hazards was positive. 8. Mood disorder: feeling down or depressed and/or reporting loss of interest. 9.Osteoporosis: high risk of osteoporosis; patients with a fracture after the age of 50 years and/or a fracture of the vertebra and/or having 2 or more of the following risk factors: a mother with a hip fracture, low body weight (men <67 kg; women <60 kg) and severe immobility.

Multidisciplinary fall risk assessment

In both groups the geriatric assessment was performed according to the Dutch National Fall Clinics guideline; all of the patients visiting the fall prevention clinics underwent a comprehensive geriatric assessment (CGA), including an assessment by a geriatrician (or a registrar in geriatrics supervised by a consultant geriatrician), an experienced nurse specialist and a physical therapist. Global cognitive assessment was standard for and performed in Group 1, but was optional in Group 2, as determined by a Mini Mental State Examination (MMSE). A MMSE score of less than 24 out of 30 was considered as confirming the presence of cognitive impairment.

Functional status was measured differently in the two groups: the Groningen Activity Restriction Scale (GARS) was used in Group 1 and the modified Katz ADL index score was used in Group 2 (23). To compare the groups, the GARS scores were converted into KATZ scores. Out of all 15 items of the Katz ADL index score, a total of nine matched with the GARS. The six non-matching items were completed by reviewing the

patient files. Because of some missing items in both groups, we computed a Katz ADL index ratio, the total score divided by the total number of items completed (range 0-1; 0= totally independent and 1=totally dependent), to make comparison of both groups feasible. Furthermore, between group comparisons were made for: mean age, gender, level of education, marital status, place of residence (institutionalized is defined as living in a home for the elderly), body mass index (BMI), presence of polypharmacy (using 3 or more drugs), experienced number of falls < 2 (defined as having only one fall sustained in the past 12 months).

Comparability of the two regions

The fall clinics in Amsterdam and Nijmegen are known to have similar aims: both sites serve GPs as well as other medical specialists, and primarily act as a secondary referral center for fall patients. To be able to compare elderly demographics in the regions surrounding both university hospitals, key characteristics of the entire older population in Regions 1 and 2 were acquired from Statistics Netherlands (24).

Statistical Analysis

Both populations were compared using a Chi-square test for categorical variables and an independent two-sample t-test for continuous variables. Differences in the characteristics between the two groups were considered significant at a p-level below 0.05. SPSS (version 16.0.01 for Windows, SPSS Inc., Chicago, IL, USA) was used to analyze the data.

Results

Of 204 consecutive patients visiting the fall clinic of the Radboud University Nijmegen Medical Center, 154 patients (75%) were included in this study (Group 1). Twenty-four patients were excluded because they attended the fall clinic without a fall history (but with syncope or dizziness) and 26 patients were excluded because too many data were missing. Missing data were primarily resulting from patient refusal or the absence of assistance of proxies in the data collection.

Of a total 2,720 patients aged 65 years and over, visiting the emergency department of the Academic Medical Center Amsterdam following a fall, all 2,149 patients living in Amsterdam (79%) received the CTI, of which 68% (n=1,461) was returned. The non-responders group was significantly older (mean age 79.7 (SD 9.5 yr) years versus 77.4 (SD 8.1 yr) (p=0.001), with slightly more females in the non-responders group (68.5% versus 66.4%) (p=0.06), without differences place of residence. Half of the responders (n=788, 53%) met the inclusion criteria (=3 modifiable risk factors) and were invited to the fall prevention clinic. Of these 788 patients, 338 refused the optional full geriatric assessment for various reasons. Exact data about the frequency of these reasons are not collected. Of the 450 patients willing to have a CGA, one third (n=150) finally attended the multidisciplinary fall prevention clinic and could be included. From these 150 patients, four patients had to be excluded because of missing data.

The mean age of the 154 participants in Group 1 was 79.3 (range 65-96; SD 8.6 yr) years; 73% was female, and this was similar to the 146 patients in Group 2, with a mean age of 78.5 (range 65-95; SD 8.5 yr) years and 77% female. Other characteristics are shown in Table 1.

In Group 1 ‘recurrent falling (=2 falls)’ was more prevalent when compared to Group 2 (90.3% vs. 75.2%, p=0.001). Additionally, the participants of Group 1 were more likely to live in ‘homes for the aged’ (18.2% vs. 8.2%, p=0.037) compared to Group 2.

In Group 1, the mean number of modifiable risk factors per patient was 4 (SD 1.6), compared to 5 (SD 1.5) in Group 2 (p<0.001). Figure 1 presents the distribution of identified modifiable risk factors in both groups. The prevalence of distinctive modifiable risk factors in each group is shown in Table 2.

Figure 1.

The distribution of modifiable risk factors in Group 1 referred by general practitioners and Group 2 identified by selection after attending an emergency department

Table 2.

Comparison of modifiable risk factors in two different referral groups for an academic falls clinic, based on referral by general practitioner (1) vs. selected by triage of the population that visited the emergency department (2)

	Group 1 (n=154)	Group 2 (n=146)	P-value
Polypharmacy, n (%)	132 (85.7)	112 (76.7)	0.14
Orthostatic hypotension, n (%)	77 (50)	74(51)	0.59
Mobility and balance problems, n (%)	103 (66.9)	128 (88.3)	0.000
Fear of falling, n (%)	91 (59.1)	116 (80)	0.000
Visual disturbance, n (%)	40 (26)	36 (24.8)	0.57
Urinary incontinence, n (%)	85 (55.2)	69 (47.6)	0.25
Home hazards, n (%)	50 (32.5)	77 (53.1)	0.002
Mood disorder, n (%)	45 (29.2)	51 (35.2)	0.38
Osteoporosis, n (%)	52 (33.8)	90 (62.1)	0.000

Overall, four modifiable risk factors were less prevalent in Group 1: ‘fear of falling’ (59.1% vs. 80%, p<0.001), ‘mobility and balance problems’ (66.9% vs. 88.3%, p<0.001), ‘home hazards’ (32.5% vs. 53.1%, p=0.002) and ‘osteoporosis’ (33.8% vs.62.1%, p<0.001).

Importantly, re-analysis after exclusion of the cognitively impaired people (N=49, 31.2%) in Group 1 did not change the comparison between the groups. In particular, the mean number of risk factors in Group 1 did not change and remained 4 (SD 1.7). The difference between the two groups in number of people living in ‘homes for the aged’ was no longer significant (11.4% in Group 1 vs. 8.2% Group 2; p=0.33). All other characteristics remained comparable, for instance the Katz ratio was 0.26 (SD 0.2) in group 1, excluding dementia patients (p=0.87). The mean age of the 105 non-cognitively impaired participants in Group 1 was 77.8 (range 65-95, SD 7.7 yr) years; 77% was female. Recurrently falling (=2 falls) remained more prevalent in Group 1 compared to Group 2 (88.6% vs. 75.2%, p=0.01).

Discussion

This study demonstrates that the risk profile and the number of modifiable risk factors in older patients visiting a fall clinic are associated with the method of referral to the fall clinic.

The effect of the referral method on patient characteristics of recurrent fallers at a fall clinic has not been described before. We only found one study using different recruitment strategies in primary care practices in relation to patient characteristics in a falls prevention program (25).

In our study, we assume that more accidental fallers were included in group 2 explaining the result of a lower rate of recurrent fallers in this group. A GP referral results in a population with less (mean 4) modifiable risk factors and a higher rate of recurrent fallers (90.3% of those referred by a GP sustained at least 2 falls in the past year versus 75.2% in those who visited the ED). An explanation for this difference might be that GPs are not well trained in recognizing the risk profile of recurrent fallers (26). Alternatively, it may be also possible that the GP has already managed some modifiable risk factors before the fall-clinic visit. Otherwise, case-finding based on fall risk factor scores among older patients at the ED results in a rather lower rate of enrollment in our fall prevention program (± 7%). Other studies also confirm that the source of referral has a significant effect on the rate of enrollment in fall prevention programs (27). The differences in the number and type of modifiable risk factors for falling between both groups suggest that there may also be differences in the prevalence of frailty. Group 1 might be considered to be frailest, because there are more frequent fallers and more people living in ‘homes for the aged’. The cognitive impaired subgroup of group 1 largely explained this significant difference in living arrangements. However, because ‘mobility and balance problems’ as well as ‘osteoporosis’ are actually less frequent in the GP referral, frailty characteristics are not homogeneously spread over the two groups. Differences in frailty between both groups most likely have consequences for the effectiveness of fall preventive interventions. Interventions appear to be less effective when the target population becomes too frail (i.e. having very low physical activity, extreme muscle weakness or nearly having lost physical activities) (28, 29).

An inevitable limitation of the study is caused by the fact that the two referral routes could not be compared within one center. Theoretically, the difference in the number of modifiable fall risk factors may be attributed to selection bias, because two centers in different regions are compared, and different professionals were involved in the referral process. However, both fall clinics described are expertise centers and work strictly according to the Dutch National Fall Clinics Guideline. Therefore results of casemix differences probably still are highly dependent on referral strategy and other sources of selection bias are unlikely to explain the differences found.

The cross-sectional design of this study is another limitation, and conclusions on causality with respect to the association between method of referral and risk profile in elderly patients at a fall clinic in essence are not possible. However, we cannot think of other major intermediary or confounding variables explaining the differences.

The data collection had to be limited because of its inclusion in daily practice. Therefore, we did not have complete information on demographics, socioeconomic status and medical consumption, for which we used data from Statistics Netherlands to compare the two regions (www.cbs.nl). We also could not collect detailed process variables on the referral techniques. The CTI was used in both groups: it was used in group 1 as part of the assessment, and as part of the referral strategy in group 2.

Part of the differences between groups is explained by the inclusion of cognitive impaired persons in group 1. Any cognitive impairment influences behavior and increases fall risk. Probably, there were also patients with cognitive impairment in group 2, but another study limitation is that this group did not have routine cognitive assessment. Subgroup analysis confirmed that the effects of referral strategy remained significant, when patients with cognitive impairments were excluded. Only, the number of people living in homes for the aged was not different anymore. This is not surprising, because presence of cognitive impairment is a major risk factor for institutionalization.

As far as we know, this is the first study that focuses specifically on subject selection, which surprisingly still is an undervalued element of designing an effective and efficient fall clinic. In fact, the best way to evaluate efficiency of a fall clinic would be to regularly evaluate the responder ratio of the casemix in reducing fall-related injuries. Unfortunately, this is probably difficult to realize in clinical practice, especially due to methodological problems of continuously monitoring fall rate. Reliable electronic fall detectors probably will solve this problem in the near future. Looking at the results of our study, we advocate regular casemix reviews. In auditing casemixes, other fall-clinics may confirm the differences between GP referral versus ED-selection. Efficiency of fall clinics may be improved by focusing more on case-mix characteristics. On the one hand, ED-risk profile criteria that are used to invite people for a fall clinic assessment can be used to directly influence the casemix. On the other hand, a different casemix may result from training GPs on referring particular patients or developing specific referral criteria with panels of GPs. In such a way, one can actively influence the efficiency of a fall clinic when the responder characteristics have been carefully described.

Depending on the rest of the network of geriatric services and collaboration between health care providers, both GP and ED referral may best fit a region. However, as the numbers of fall clinic referrals will increase, the necessity of an adequate entrance policy will become important everywhere. Moreover, recurrent falling is associated with fall-related injuries concomitant with health care costs and a functional decline (30). Intervention during the process of decline by fall risk assessment and by fall preventive measures at an early stage may result in a considerable gain in health, well-being of individual patients and a reduction of health care costs, in general (31). Furthermore, early (in a pre-frail population) fall preventive interventions may be more effective (29). Therefore, the GP can play a key role in selecting older persons for secondary fall prevention programs and interventions. Next to GP referral streamlining, the ED referral method may be improved. ED selection proved to select a fall population with first, accidental falls in persons high at risk for recurrent falling. The postal CTI method, however, proved to be inefficient in recruiting people. ED recruitment in the ideal situation exists of an immediate screening following or even at presentation at the ED, and guaranteeing follow up for fall risk assessment.

In conclusion, this study shows that the method of referral to the fall clinic determines the risk profile and number of modifiable risk factors in older patients presenting at a fall clinic. In clinical practice, the selection of a specific referral method, and subsequent evaluation of the selection process will highly contribute to efficiency of the fall clinic. Ongoing study of the effect of triage should have a high priority and should be based on simple screening tools fit for either the general practice or ED settings, to objectively determine which patients should be referred to a fall clinic. Several tools for fall risk assessment are available (26, 32, 33). However, none of them is easy to use in the setting of general practice, and those developed for the setting of ED are not appropriate for bedside use. We recommend that defined diagnostic procedures to determine the risk for recurrent falling become an integrated part of clinical practice, not only at ED, but also in general practice.

As case finding is an integral part of the complex intervention performed in fall clinics, and safeguarding the right casemix probably is the key issue for sufficient efficiency, both should be an integral part evaluating quality of care delivered by fall clinics.

Conflict of Interest: There are no conflicts of interest for any of the authors.

Author Contributions: YS was principally responsible for conception of the research idea, project management, data analysis and writing of the manuscript. YS supervised the collection of the data of the fall clinic patients from the Radboud University Nijmegen Medical Center. MH was involved in data analysis; she compared the socio-demographic data and critically reviewed the article. AS was involved in recruiting the patients and collecting the data of the fall clinic patients in the Academic Medical Center in Amsterdam and critically reviewed the article. FR collected the data of the fall clinic patients in the Radboud University Nijmegen Medical Center. MOR was involved in the design and in critically reviewing the article. SR was the coordinator of the study and supervised the collection of the data of the fall clinic patients from the Academic Medical Center in Amsterdam, and SR was also involved in writing of the article.

Sponsor's Role: This project received no sponsorship.

Financial disclosure: None of the authors had any financial interest or support for this paper.