Journal of the New Zealand Medical Association, 07-November-2003, Vol 116 No 1185
Upward trends in the incidence of neck of femur fractures in the elderly
Shaun Stephenson, John Langley, John Campbell and William Gillespie
Fielden and colleagues recently reported that between 1988 and 1999 there had been only a minor increase for women and a modest increase for men in the incidence of neck of femur fractures among those 65 years of age and over.1 They also reported that for both women and men age-group-specific incidence rates had decreased over the period, though these declines were significant only for women. Their study was based on patients discharged from public hospitals.
The rates reported by Fielden and colleagues were significantly lower than those predicted in an earlier report.2 They suggested that it was not immediately apparent why this was the case. Possible explanations included the impact of a number of prevention strategies, such as improved diet and increased exercise, fall-prevention programmes, and hormone replacement therapy. They concluded that, in the absence of nationally agreed, implemented, and evaluated strategies, it was not possible to draw definitive conclusions about the impact these strategies may have had. The question as to why the rates had not increased as predicted remains unanswered.
When observed effects are not as predicted, consideration should be given to the validity of the predictive model and whether there was measurement error. There is a strong case, for example, for considering the effect on public hospital discharge-based injury rates of whether one includes readmissions in the estimates.3 Fielden and colleagues did not provide such detail on their case selection procedures.1 Given the absolute size of the neck of femur fracture problem in New Zealand, and the implications of changes in incidence rates for prevention and health service delivery, we sought to re-examine the incidence of this problem paying particular attention to case selection.
Electronic data files were obtained from the New Zealand Health Information Service (NZHIS) for all injury discharges from public hospitals in New Zealand for the period from 1988 to 1999. We did not include discharges from private hospitals as data were unavailable for the period from 1996 onwards and few neck of femur fractures in the elderly are treated in the acute phase in private hospitals. For example, in 1995 only 31 cases of neck of femur fracture involving patients aged 65 years and over were reported by private hospitals, not all of whom would have been acute admissions.4
In this data set, diagnoses were coded according to two derivatives of the ninth edition of the International Classification of Diseases (ICD-9),5 namely ICD-9-CM6 and ICD-9-CM-A.7 Discharges from the second half of 1999 were originally coded according to ICD-10-AM8 and then mapped back to ICD-9-CM-A by NZHIS to generate the equivalent ICD-9-CM-A diagnosis. The NZHIS data set allows for the recording of multiple diagnostic codes on discharge. Each diagnostic code has an accompanying short-text description (up to 50 characters) that may provide extra detail on the nature of the diagnosis. Our analysis was restricted to examining a maximum of four diagnostic codes for each case. It was considered highly unlikely that any case would have a diagnosis of neck of femur fracture that lay outside the first four diagnoses.
From this data set, all persons 65 years of age and older who were discharged alive or dead and who had a diagnosis of ‘820: Fracture of the neck of the femur’ were selected. It has been previously recommended, when using the NZHIS data for injury epidemiology, that cases should be selected on the basis of their principal diagnosis only (ie, primary reason for admission).9 This recommendation was made on the grounds that, for cases that had a disease as a principal diagnosis and injury as the second or subsequent diagnosis, it would be difficult to determine, in many cases, whether they would have been admitted to hospital if they only had the injury. This problem should not exist for neck of femur fractures, as all persons with this condition would require and receive inpatient treatment. Grimley Evans and colleagues have suggested including cases with a diagnostic code of ‘821: Fracture of other and unspecified parts of femur’ in estimates of hospitalised neck of femur fracture. They reviewed cases with this code and found many had neck of femur fractures.10 We took a more conservative approach, including only those cases with a diagnostic code of 821 where the electronic text description of the diagnosis mentioned neck of femur. A small number of cases with diagnoses of ‘804: Multiple fractures involving skull or face with other bones’, ‘827: Other, multiple, and ill-defined fractures of lower limb’ or ‘828: Multiple fractures involving both lower limbs, lower with upper limb, and lower limb(s) with rib(s) and sternum’, where the text description of the diagnosis included mention of a neck of femur fracture, were also selected.
Given that people can be admitted to hospital for the treatment of injury in both the acute and rehabilitative phases, it is important to be able to differentiate between the two. Failure to do so could produce a substantial error in the estimate of person-based injury incidence, if the data set being examined contained individuals who have had a series of readmissions for ongoing rehabilitation.9
Readmission status was determined using four data elements available in the NZHIS data since 1989. These data elements were: a unique personal identifier (National Health Index (NHI) number), date of injury, date of admission, and date of discharge. It has been shown previously that reasonably accurate estimates of readmission status can be derived by coding all cases with the same NHI number and date of injury as a case with an earlier date of admission as readmissions.11 Nevertheless, to allow for incorrect and missing dates of injury, where two cases were identified with the same NHI number and one case had a date of admission within one day of the date of discharge of the other case, the former case was coded as a readmission. All readmissions have been excluded.
Day patients, that is cases who were discharged the same day they were admitted, have been inconsistently recorded over time in the NZHIS data.9 Hence, in order to produce a consistent data set, all day patients who did not die in hospital have been excluded.
Annual mid-year population estimates based on the 1986, 1991 and 1996 Censuses were obtained from Statistics New Zealand for the purpose of calculating incidence rates. Poisson regression was used to test for changes in the incidence and age-group-specific incidence rates of neck of femur fractures over time. A scale factor was fitted to adjust for possible overdispersion. Trends are reported as (RR = a, χ2 = b, p = c), which indicates a relative rate of ‘a’, a χ2 value for the significance of the trend of ‘b’, and a p value of ‘c’. The relative rate is the estimated rate for a year relative to the previous year. For example, a relative rate of 0.986 indicates each year’s estimated rate was 98.6% of the previous year’s rate. Fielden and colleagues provided the data used in their earlier study.1
We identified 45 297 potential cases for the study period, of whom 12 695 (28%) were excluded as either readmissions or day patients. Hence, we estimate the incidence of neck of femur fractures resulting in hospitalisation occurring to New Zealanders aged 65 years and over during the period 1988 to 1999 was 32 602 cases. The diagnostic codes for all potential patients are shown in Table 1.
Table 1. Neck of femur fracture hospitalisations for those aged 65 and older by diagnostic code and admission status
Figures 1 and 2 show yearly estimated incidence of neck of femur fracture cases identified for the period 1988 to 1999 using our selection criteria and the data supplied by Fielden and colleagues for women and men respectively. For both women and men our selection criteria generally resulted in a lower estimate of the incidence in each year but a stronger upward trend over the period than found by Fielden and colleagues.1 The upward trends in incidence for both women and men using our selection criteria were highly significant (women: RR = 1.018, χ2 = 53.4, p <0.0001; men: RR = 1.034, χ2 = 113, p <0.0001). Those for the data from Fielden and colleagues1 were also significant (women: RR = 1.004, χ2 = 5.89, p = 0.0152; men: RR = 1.018, χ2 = 11.0, p = 0.0009).
The trends in age-group-specific incidence rates of neck of femur fracture cases for women are shown in Figure 3. The rates for all age groups declined over the period, with the strongest declining trends being in the 70–74 and 75–79 age groups. The trends for all age groups except 65–69 were significant.
Figure 1. Incidence of neck of femur fracture hospitalisations for women aged 65 years and over, 1988–1999
Figure 2. Incidence of neck of femur fracture hospitalisations for men aged 65 years and over, 1988–1999
Figure 3. Incidence rates of neck of femur fracture hospitalisations by age group for women, 1988–1999
Figure 4. Incidence rates of neck of femur fracture hospitalisations by age group for men, 1988–1999
The trends in age-group-specific incidence rates of neck of femur fracture cases for men are shown in Figure 4. The rates for the younger age groups, 65–69, 70–74, and 75–79, have declined slightly, while those for the older age groups, 80–84 and 85+, have increased slightly. None of these trends were significant. The magnitude and significance of trends for women and men are summarised in Table 2. All of the fitted models appeared to fit the data reasonably well.
Table 2. Trends in incidence rates of neck of femur fracture hospitalisations by age group and gender, 1988–1999
Though both our selection criteria and those of Fielden and colleagues1 result in an upward trend in the incidence of neck of femur fractures over the period, the trends observed, for both women and men, using our criteria were considerably stronger. Similarly, our estimates of the trends in age-specific rates amongst women showed the decline in these rates to be much less significant than that reported by Fielden and colleagues.1
There were three differences in the selection criteria used in this paper compared with those of Fielden and colleagues.1 These were our inclusion of diagnostic codes apart from 820, our inclusion of cases with second or subsequent diagnoses of neck of femur fracture, and our exclusion of readmission and day patients. There were 512 cases in our data set with diagnostic codes apart from 820, 1763 cases with non-principal neck of femur diagnoses, and 12 695 cases excluded as readmission or day patients. Hence, most of the differences, both in incidence and incidence rates, between our results and those of Fielden and colleagues1 were due to our exclusion of readmissions and day patients. As our results show, failure to consider these cases can produce significant errors in estimates, which may in turn have policy implications for prevention and health service delivery.
This study relies on neck of femur fractures being coded accurately in hospital records throughout the study period. No published studies have assessed the accuracy of this coding in New Zealand.
It should be noted that ideally neck of femur fracture incidence estimates should include all people who suffer the injury but die without being admitted to hospital. For New Zealand, it is difficult to produce an accurate estimate of the size of this group, as injury diagnoses have traditionally not been recorded for deaths. Most neck of femur fractures in those aged 65 and over are a result of falls. There were 1805 cases of falls resulting in death for 1988 to 1998, some of whom would not have been hospitalised.12
In order to make predictions of future neck of femur fracture incidence it is important to exclude readmissions and day patients, identify deaths outside hospital and consider the potential for differing incidence rates between birth cohorts. Birth cohort effects may substantially influence predictions.13
In conclusion, New Zealand has faced an increasing morbidity burden throughout the 1990s due to neck of femur fractures amongst those aged 65 years and older. This is attributable to increasing numbers of elderly people in the population rather than increased risk.
Author information: Shaun Stephenson, Biostatistician; John Langley, Director, Injury Prevention Research Unit, Department of Preventive and Social Medicine; John Campbell, Dean, Faculty of Medicine, Dunedin School of Medicine, University of Otago, Dunedin; William Gillespie, Dean, Hull York Medical School, United Kingdom
Acknowledgements: The Injury Prevention Research Unit (IPRU) is jointly funded by the Health Research Council of New Zealand and the Accident Compensation Corporation (ACC). The views expressed in this paper are those of the authors and do not necessarily reflect those of the above organisations. New Zealand hospital data were sourced from the New Zealand Health Information Service. We thank David Chalmers, Colin Cryer and Geoffrey Horne for providing comments on an earlier draft of the paper, and Jann Fielden and Gordon Purdie for providing the data from their study.
Correspondence: Shaun Stephenson, Injury Prevention Research Unit, Department of Preventive and Social Medicine, Dunedin School of Medicine, University of Otago, P O Box 913, Dunedin. Fax: (03) 479 8337; email: email@example.com
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals