2019 年 42 巻 5 号 p. 778-785
Fragility fractures associated with age-related bone loss are of urgent concern worldwide because they reduce QOL and pose financial burdens for health care services. Currently, national data in Japan are limited. This study provides quantitative data for older patients throughout Japan who, although otherwise relatively healthy, sustained fragility fractures and were hospitalized for them. The National Database of Health Insurance Claims and Specific Health Checkups of Japan was accessed to target patients aged 65 years or older who sustained fractures between May 2013 and September 2014 and were not hospitalized for at least 13 months prior to fracture. We investigated whether the first fracture sustained was fragility related at any of four locations (proximal humerus, distal radius, vertebra, or femoral neck) and whether it necessitated hospitalization. Fragility fractures were identified in 490138 of 1188754 patients (41.2%, 345980 patients/year; 1 : 4 male-to-female ratio). Regardless of gender, vertebral fractures were most common across the age cohorts studied (43286 males and 162767 females/year), and femoral neck fractures increased markedly with increased patient age. Approximately 80% of patients with femoral neck fractures were hospitalized (62.3% of males, 71.1% of females) compared with up to 10.4% of patients with other fragility fractures. Data provided in this study can be used as a baseline for evaluating the health economy and establishing health policy in Japan.
Incapacitating fractures in older people cause their QOL to deteriorate and represent a considerable financial burden on health care services1,2) due to reduced mobility, hospitalizations, and the requirement for nursing assistance. The Japanese government estimates that the proportion of the population aged 65 years or older in Japan will increase from 26.6% in 2015 to 32.8% by 2035.3) Furthermore, the number of patients hospitalized with fractures is likely to rise 1.4-fold within the next two decades.4)
Fragility fractures caused by a slight external force, such as a fall from standing height or lower, due to age-related loss of bone density,1,2,5,6) occur more frequently among older people. However, to the best of our knowledge, previous epidemiologic studies that have evaluated fragility fractures in the Japanese population7–10) were conducted in discrete regions rather than nationally, and have not elucidated the proportion of patients hospitalized for fractures. To minimize medical costs in a nation whose older population is continuing to expand and live longer than previous generations, it is imperative to appropriately evaluate the health economy associated with fragility fractures and adapt the health policy in Japan accordingly.
This study uses the National Database of Health Insurance Claims and Specific Health Checkups of Japan (NDB Japan) and was designed to identify the number of older patients throughout Japan who developed any fragility fractures located at the proximal humerus, distal radius, vertebra, or femoral neck, and determine who among them were treated for pathological fractures during hospitalization or ambulatory care.
This study was undertaken as a part of the Polypharmacy and Fracture in Older People Study, with a case-crossover design, to focus on the harmful effects of concomitant use of central nervous system agents on fragility fractures. Within the larger study, this study aimed to identify patients who experienced fragility fractures that were likely to have been caused by falls. The studies were approved by the ethics committee at Tokushima Bunri University in September 2015 (No. H27-8).
A claims dataset derived from the NDB Japan was used for this study. The NDB Japan was generated by the Ministry of Health, Labour and Welfare (MHLW) from the records of all Japanese citizens (130 million people) except those with cases documented with workers’ accident compensation insurance, automobile accident insurance, public assistance, or cases which represent private expenses.
The dataset provided by the MHLW was designed to capture the requisite study population and relevant information but also to remain in compliance with the directive to preserve the privacy of citizens. Two patient identifiers were used from the NDB Japan data to assign a unique hash value to each patient for the purpose of researcher blinding. Inpatient- or outpatient-medical claims and Diagnosis Procedure Combination/Pre-Diem Payment System (DPC) claims from April 2012 to September 2014 were retrieved for our working dataset. It included the following information: two identifiers for each patient, gender, age group based on cohorts spanning 5-year intervals, domestic diagnosis codes mapped to the 10th Revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10), modifier codes added to diagnoses, domestic diagnosis codes designating the reason for hospitalization if the claims were DPC claims (original form), the date on which domestic diagnosis codes were assigned in each hospital or clinic if the claims were inpatient- or outpatient-medical claims or DPC claims (medical claims form), hospitalization date, and radiography codes.
The study population was defined as follows: patients aged 65 years or older who sustained any fracture during the follow-up period from May 2013 to September 2014, and were not hospitalized at least 13 months prior to their fracture. The follow-up ended when patients sustained any fracture, were hospitalized, or the September 30 2014, date was reached, whichever came first.
First fractures sustained by individuals at the age of 65 years or older during the follow-up period were reviewed to ascertain whether the fractures met our defined fragility criteria, and to determine whether ambulatory care or hospitalization were necessary. After tallying the number of people identified with a fracture, dividing by 17 (i.e., the number of months within the follow-up period) and multiplying by 12, an annual count was achieved.
A fracture was attributed when a plain radiograph was taken and dated at the institution where the fracture was first diagnosed. If the first fracture during the follow-up period was not radiographed, but a second fracture was, the second fracture sustained was documented for those patients who were then considered to have a fracture history. We ascertained the diagnostic designation of fractures using both disease codes and the modifier codes added to them. Fragility fractures were identified at four locations: proximal humerus, distal radius, vertebra, or femoral neck. Fractures sustained at locations other than these were not considered fragility fractures because they likely would not have been caused by a slight external force such as fall, but rather by excessive external forces such as falls from more substantial heights or traffic accidents for example.
Fractures of the proximal humerus included those of the neck of the humerus, the greater or lesser tubercle, and/or the humeral head. Fractures of the distal radius included the radial styloid process, ulnar notch, or dorsal tubercle, as well as distal ulna fractures associated with distal radius fractures. Vertebral fractures included compression fractures of the lower thoracic spine (T9–T12), or lumbar spine. Femoral neck fractures included those of the greater or lesser trochanter, and/or fractures of the proximal femur. We did not regard cervical vertebral compression fractures, upper or middle thoracic spine compression fractures (T1–T8), and transverse or spinous process fractures as vertebral fractures, because they were more likely to be caused by trauma other than falls. Femoral head or subtrochanteric fractures, hip fractures, and pelvic fractures resulting from femoral neck fractures, were excluded for a similar reason.
Fractures diagnosed with modifiers that were unlikely to represent fragility fractures, such as axillary and lower abdominal fractures, were excluded from this study. Two physicians independently judged whether the diagnostic designation of a fracture met the study definition of fragility fractures by ascertaining the patterns and/or combinations of fracture diagnoses and the modifiers. If their judgments were different, they discussed the case and came to a consensus.
To determine whether a fracture necessitated ambulatory care or hospitalization, we compared the date on which the fracture occurred with the earliest hospitalization date. Patients who received ambulatory care for a fracture were identified if the date on which the fracture was sustained preceded the earliest hospitalization date. Patients hospitalized for a fracture were identified if the hospitalization date matched the date of fracture in outpatient-medical claims or when the reason for hospitalization in DPC claims (original forms) was a fracture. We did not consider hospitalizations in cases when the date on which the fracture was sustained was shown in inpatient-medical claims or DPC claims (medical claims forms), because those fractures could have occurred after hospitalization. Given that DPC claims (original forms) do not include the date of diagnoses, hospitalization dates were used as an alternative to the date of fracture diagnosis if the reason for hospitalization in the DPC claims (original form) included fracture.
We identified any history of a fracture and complications using diagnosis codes prior to the date the fracture was sustained and calculated the Charlson comorbidity score.11) This score was estimated as the sum of the weighted score allocated for each of 17 different disease comorbidity categories. Two well-validated ICD-10 coding algorithms (the version by Quan et al.12) for malignancy and the version by Sundararajan et al.13) for the remaining categories) were used. No distinction was made between diabetes with and without complications (2 and 1 of their original weights, respectively) and we used a weight of 2 for diabetes,14) as the majority of ICD-10 codes that were mapped to diagnosis codes of diabetes did not have 4 characters.
When identifying disease, including fractures, definitive rather than suspicious disease codes were used and we referred to the maps between the disease name and ICD-10 codes used for Japanese government morbidity and mortality surveys.15,16)
The agreement between the two physicians regarding what constituted fragility fractures was estimated using the kappa statistic.17) The kappa statistic shows the proportion of agreement that remains after chance has been excluded18) and is interpreted as follows: −1.00 (perfect disagreement), 0.00 (chance agreement), 0.01–0.40 (poor agreement), 0.41–0.74 (fair to good agreement), 0.75–0.99 (excellent agreement), and 1.00 (perfect agreement).19) The kappa statistic was tested based on the null hypothesis that the agreement is purely by chance.20)
A two-tailed p-value with statistical significance was defined apriori as p < 0.05. Descriptive statistics were used to summarize all other data, because our large sample size increased the likelihood of finding numerous significant differences. SAS Enterprise Guide 7.13 (SAS Institute Inc., Cary, NC, U.S.A.) was used for data processing and the estimation and test of the kappa statistic.
The MHLW dataset used in this study originally included 2493550 patients. After excluding 1304796 patients based on hospitalizations before the follow-up period, no fracture by 65 years or older during the follow-up period, fractures after hospitalization, or the absence of a plain radiograph on the date of presentation at the institution where the fracture was diagnosed, we identified a total of 1188754 patients (839120 patients/year) who sustained any fracture at the age of 65 years or older during the follow-up period and made up the study population. Figure 1 displays a flowchart of the study population.
MHLW; Ministry of Health, Labour and Welfare, DPC; Diagnosis Procedure Combination/Pre-Diem Payment System, [A]; Ambulatory care, [H]; Hospitalized. a Actual identified numbers of patients not yet converted to year totals. b Modifiers such as after effect, chronic phase, and prophylaxis. c Proportion of 1188754 patients. d Proportion of 490138 patients. e Proportion of 487938 patients.
Of the 1188754 patients further evaluated, 490138 (41.2%, 345980 patients/year, 65442 males/year and 280537 females/year, 1 : 4 male-to-female ratio) were identified as having a fragility fracture with almost all patients sustaining a fracture in a single location. Specifically, the proportions for patients with a single fragility fracture included 59.8% of patients (206053 patients/year) with a vertebral fracture, 17.7% of patients (60945 patients/year) with a femoral neck fracture, 17.1% of patients (58874 patients/year) with a distal radius fracture, and 5.4% of patients (18555 patients/year) with a fracture of the proximal humerus.
The two physicians determining which fractures were fragility-related, in accordance with our definition, ascertained 11356 patterns and/or combinations of fracture diagnoses and the modifiers. The agreement between them was 92.7% with a kappa of 0.73 (95% confidence interval 0.71–0.75, p < 0.001), indicative of fair to good agreement.
Table 1 summarizes the characteristics of patients who sustained any fracture and fragility fractures. The proportion of females was numerically higher among the patients with fragility fractures than among patients with any fracture (79–86% and 76%, respectively). The proportion of patients within each age group was numerically different based on the location of the fragility fracture; distal radius fractures were sustained in more patients aged between 65 and 74 years than in patients aged 75 years or older, whereas at all the other locations assessed, fragility fractures increased with age. Femoral neck was the most frequent fracture location in patients with Alzheimer’s disease. Vertebral fractures were the most frequent type of fracture among patients with osteoporosis.
a) Numbers are actual identified values, not the converted number/year. b) Patients with a fragility fracture at a single location are shown. c) People without claims data between April 2012 and April 2013. d) This was defined as any disease code for fracture predating the fracture sustained per the claims data from April 2012; therefore, these data do not reveal fracture history from birth. e) ICD-10 codes. f) ICD-10 codes used for estimation of the Charlson comorbidity score.
Overall, for both genders, and across all age groups, there were more fractures sustained in the vertebra than at any other location assessed (Figs. 2a–c); proportions of patients with a fragility fracture at this one location during 1 year included 66.3% (43286/65247) of males and 58.3% (162767/279180) of females. Fragility fractures of the distal radius were next most prolific among the younger age cohorts assessed, but this trend was overtaken by the increase in femoral neck fragility fractures in both genders which surpassed distal radius fractures occurring earlier in males (between 75–79 years) than in females (between 80–84 years) (Figs. 2b, c). Among patients from all age cohorts who sustained a single fracture, 16.7% (10872/65247) of males and 17.9% (50073/279180) of females sustained a femoral neck fracture and 12.4% (8093/65247) of males and 18.2% (50781/279180) of females sustained fractures of the distal radius.
(a) Overall. (b) Male. (c) Female.
A total of 69935 (20.3%) patients/year were hospitalized for a fragility fracture at one location as opposed to 104948 (12.5%) patients/year with any fracture (Table 2). Similar results were observed among patients hospitalized for a fragility fracture, regardless of gender or age cohort. Of the patients with femoral neck fractures, 48498 (79.6%) patients/year were hospitalized and the percentage was relatively consistent regardless of gender or age cohort evaluated (Table 3). However, for patients with fragility fractures at the other locations assessed, hospitalization occurred for up to 10.4% of patients. A higher percentage of males with vertebral fractures were hospitalized than females across all age cohorts, but this gender difference was not observed at the other three locations assessed for fragility fractures.
a) The numbers are the converted number/year; therefore, the values shown in this table do not always compute to total sum numbers.
a) The numbers are the converted number/year; therefore, the values shown in this table do not always compute to total sum numbers.
Figure 3 depicts the number of patients by gender who sustained a single fragility fracture and received either ambulatory care or were hospitalized. Across all age cohorts for both genders, most of the patients who received ambulatory care had vertebral fractures (Figs. 3a, b); among them, 75.3% (38516/51172) were males and 67.2% (150045/223320) were females over a one-year period. The second most frequent fragility fracture location that required ambulatory care was the distal radius across all age cohorts in 15.4% (7856/51172) of males and 21.9% (49011/223320) of females over 1 year. Across all age cohorts for both genders, most of the patients who were hospitalized had sustained a single fragility fracture of the femoral neck (Figs. 3c, d); among them, 62.3% (8766/14075) were males and 71.1% (39731/55860) were females. Numerically, the next most common single fragility fracture that required hospitalization across all age cohorts during 1 year was in the vertebra and included 33.9% (4770/14075) of males and 22.8% (12721/55860) of females.
(a) Ambulatory care and male. (b) Ambulatory care and female. (c) Hospitalized and male. (d) Hospitalized and female.
Based on the NDB Japan, a profile was created for people aged 65 years or older throughout Japan who had been relatively healthy apart from sustaining fractures, including fragility fractures. It was determined from among our study population that almost 840000 people/year sustained a fracture, and approximately 40%, or an estimated 350000 people/year incurred a fragility fracture in any of the four locations assessed.
Notably, based on our knowledge, this study is the first to provide data about the proportion of patients in Japan who were hospitalized for a fragility fracture. Approximately 80% of older patients who were diagnosed with a fragility fracture of the femoral neck, regardless of gender or the age cohort assessed in this study, were hospitalized, as opposed to up to 10.4% of older patients with fragility fractures at the other three locations who were hospitalized. This study revealed that among the patients hospitalized for fragility fractures, femoral neck fractures accounted for 62% of admissions in males and 71% of admissions in females.
The findings of this study suggest that decreasing the incidence of femoral neck fractures is important in terms of minimizing medical costs. This is a particularly important endeavor considering the severe disability associated with femoral neck fractures; 10–25% of patients with femoral neck fractures have died within 1 year after surgery2,5,21,22) and 25–50% of patients experience a loss of mobility.2,5)
Interestingly, this study determined that a higher proportion of males than females sustained a vertebral fracture for which they were hospitalized. The reason for this is unclear, but we surmise that it could be because physicians are less likely to attribute male complaints about back pain to vertebral fractures and more likely to incidentally identify vertebral fractures in routine radiographs of female patients. Eventually, males with vertebral fractures may be diagnosed, but it is likely to be when the fracture becomes more severe.
The results of this study, which encompassed the entire Japanese population aged 65 years and older, closely match the results of the previous studies performed in restricted regions of Japan,7–10) by prefecture or city unit that assessed data trends pertaining to: (1) the highest number of fractures, regardless of age cohort for older people, occurring in patients with vertebral fractures, and (2) the second highest number of fractures occurring in patients with distal radius fractures for the earlier older age cohorts and patients with femoral neck fractures for the later older age cohorts, which were exponentially increased. These similarities strengthen the validity of our study. Although imprecision of diagnostic designations is a known limitation of claims database studies, the precision of fracture diagnoses in this study was able to be enhanced via: (1) the use of modifier codes in addition to diagnosis codes for fractures, (2) the expertise of two physicians making independent judgments and reaching a consensus when necessary, and (3) confirmation of radiographic examinations at the fracture-diagnosis-assigned institution on the date of the diagnosis.
There are some limitations associated with this study. First, validity of the methodology used for fracture diagnoses requires further verification in additional validation studies, although implementation of validation studies is difficult in Japan.23) Second, for comorbidities, no attempt was made to improve the precision of the information by assessing modifier codes and any other examinations which could potentially introduce information bias. Third, we were not able to estimate the incidence rate of fractures or compare the incident rates in the previous regional studies performed inside7–10) and outside1,2,22) of Japan, because our data are derived from a polypharmacy study with a case-crossover design. In this study, fractures were recorded but there was no cohort of all older people in Japan who were relatively healthy, although that cohort is needed to calculate the incidence rate.
Regardless of these study limitations, the figures presented in this study provide important baseline data that can be utilized for evaluating the health economy and establishing health policy in Japan. In addition, the methodology used in this study to enhance validity should be useful for future studies using claims data in Japan.
We thank the MHLW for generating and making the dataset from the NDB Japan available for our study. This work was supported by JSPS KAKENHI Grant Number JP15K08121.
The authors declare no conflicts of interest.