ABSTRACT
BACKGROUND
It is unknown which medical diagnoses are strongly associated with long-term care needs certification.
METHODS
We conducted a case-control study using linked medical and long-term care data from two Japanese cities. The participants were aged ≥75 years, without any previous long-term care needs certification, and had at least one medical insurance claim record during a period between April 2013 and March 2015 in City A and between April 2013 and November 2016 in City B. Cases were newly certified people for long-term care needs during the study period, whereas controls (matched on age category, sex, city, and calendar date) were randomly selected in a 1:4 ratio. We conducted multivariable conditional logistic regression analyses to estimate the association between 22 categories of medical diagnoses recorded in the past six months and new (i.e., first ever) long-term care needs certification.
RESULTS
Among 38,338 eligible people, 5,434 (14.2%) newly received long-term care needs certification. The adjusted odds ratio (95% confidence interval) was largest for femur fractures, 8.80 (6.35–12.20), followed by dementia, 6.70 (5.96–7.53), pneumonia, 3.72 (3.19–4.32), hemorrhagic stroke, 3.31 (2.53–4.34), Parkinson’s disease, 2.74 (2.07–3.63), and other fractures, 2.68 (2.38–3.02). A restricted analysis to more severe outcome (care need levels 2 to 5), sensitivity analysis to use different periods for exposure definition, and separate analysis by city showed consistent results.
CONCLUSIONS
Among a range of recorded medical diagnoses, fractures (especially femur fractures), dementia, pneumonia, hemorrhagic stroke, and Parkinson’s disease were strongly associated with long-term care needs certification.
INTRODUCTION
Long-term care or nursing care generally indicates in-home and facility-based services for older or disabled people to conduct their daily lives outside of acute care hospitals. In Japan, the public long-term care insurance system was introduced in 2000 [1]. People aged ≥65 or those aged 40–64 with specified diseases (e.g., rheumatoid arthritis, cancer) who satisfied eligibility criteria can receive long-term care services under the insurance coverage. Eligibility is assessed mainly by a 74-item questionnaire (mostly related to activities of daily living) recorded by certified investigators, as well as a doctor’s opinion paper on long-term care needs from a medical point of view [2]. The number of people with long-term care needs certification has been increasing annually in Japan [3].
Apart from aging and associated senility, medical conditions or illnesses seem to lay behind the initiation of long-term care. According to the 2016 Comprehensive Survey of Living Conditions in Japan [4], in which randomly sampled people with long-term care needs certification or their family members answered the questions, the most common primary reason for the initiation of long-term care was dementia, followed by cerebrovascular disease, senility, fractures or falls, and joint diseases [3]. Another study based on linked medical and long-term care insurance claims data suggested that the most common medical diagnosis recorded in the past six months of long-term care needs certification was cerebral vascular disorders, followed by cancer, fractures, arthropathy, and dementia [5]. However, due to a lack of a comparison group of people without long-term care needs certification in these studies [3, 5], the relative risk of each medical condition on long-term care initiation remained unknown.
The identification of medical diagnoses strongly associated with long-term care needs certification is warranted to potentially prevent or extend the initiation of long-term care. The national or local government may be able to conduct an efficient intervention by focusing on people at risk of such medical conditions. Previous studies estimated the relative risks of some medical conditions (including strokes, joint pain, osteoporosis, cancer, diabetes, and cognitive dysfunction) on long-term care [6–8]. However, to our knowledge, there has been no study that systematically assessed the associations between a variety of medical diagnoses and long-term care needs certification. Therefore, using linked medial and long-term care datasets in two Japanese cities, we investigated the associations between recorded medical diagnoses and the incidence of long-term care needs certification.
METHODS
DATA SOURCE
We obtained linked medical and long-term care insurance claims data from the municipal governments of two cities (named Cities A and B) in Japan. The medical claims data here include data of people with citizens’ health insurance for municipalities and unions and Late Elders’ Health insurance for individual prefectures, while data of people with other medical insurance certifications (society-managed health insurance for large companies, Mutual Aid Associations, the public sector, and the National Health Insurance Association for medium to small companies [9]) were not included. Cities A and B had a population of around 405,000 and 55,000 (of which the numbers of people aged ≥75 years were around 39,000 [9.6%] and 7,500 [13.6%]), respectively, in 2014. Linkage was performed in each municipal government using personally identifiable information. In the data we received, dummy ID numbers were assigned to individuals in both medical and long-term care insurance claim datasets. At the time of this research, the data were available since April 2012 in both cities, up to March 2015 in City A and November 2016 in City B.
Details of Japanese medical and long-term care insurance claims data are described elsewhere [10–13]. In brief, the medical insurance claims datasets include recoded diagnoses both at outpatient or inpatient facilities and prescription information at outpatient facilities on a monthly basis. Recorded diagnoses are based on Japanese original disease codes, which are linked to the International Classification of Diseases 10th Revision (ICD-10) codes [14]. The long-term care datasets include the date of new (i.e., first ever) and updated long-term care needs certification and certified care needs levels, including support levels 1 and 2 and care need levels 1 to 5 (most disabled).
This study was approved by the ethics committee of the University of Tsukuba (approval numbers: 1178-3 and 1184-1). Consent to individual participants was waived because of the anonymous nature of the data. Data from one of the two cities included age information using five-year age categories (instead of the exact age or birth year) to preserve privacy.
STUDY DESIGN
We conducted a case-control study to investigate the association between a variety of recorded medical diagnoses (i.e., exposures of interest) and the incidence of new long-term care needs certifications (i.e., outcome of interest) [15–18].
We first identified people aged ≥75 years who had at least one medical insurance claim record during a period between April 2013 and March 2015 in City A and between April 2013 to November 2016 in City B. Data in the first year (i.e., from April 2012 to March 2013) in the datasets were not used for the definition of the study population and outcome, while the data could be used to define the exposures of interest. The choice of age 75 as a lower limit was made because (i) the vast majority (over 80%) of long-term care needs certification is made for this age group [3], and (ii) all people aged ≥75 years are qualified for Late Elders’ Health insurance (while people aged <75 years may be qualified for different types of medical insurance systems, but only data on beneficiaries for the National Health Insurance system were available in our datasets); thus, population representativeness is more likely to be ensured by restricting the study participants to those aged ≥75 years. We excluded people with any previous long-term care needs certification before April 2013, with or without service use.
Among these eligible people, we identified cases who were newly certified for long-term care needs at any care need level, from support level 1 to care need level 5. Next, we randomly selected four controls among those who were alive and not yet having long-term care needs certification at each case’s certification month, matched on age category (75–79, 80–84, 85–89, and ≥90), sex, city, and calendar date. We used an incidence-density sampling method (meaning that cases could be selected as controls until they became a case) allowing for replacement (meaning that an individual could be selected as controls several times) [15–18].
LIST OF MEDICAL DIAGNOSIS CATEGORIES AND ICD-10 CODES
For the exposures of interest, we established a list of medical diagnosis categories, as well as a list of ICD-10 codes suggesting each medical diagnosis category. First, we listed medical diagnosis categories potentially associated with the initiation of long-term care, referring to a questionnaire on long-term care in the Comprehensive Survey of Living Conditions in Japan [19]: cerebrovascular diseases (further divided by the authors into hemorrhagic stroke, ischemic stroke, and other cerebrovascular diseases), cardiac diseases (divided into ischemic heart disease, arrhythmia, heart failure, and other cardiac diseases), cancer, lower respiratory tract diseases (divided into chronic obstructive pulmonary disease, pneumonia, and other lower respiratory tract diseases), joint diseases (divided into rheumatoid arthritis, other arthropathies, and dorsopathies [disorders of the back or spine]), dementia, Parkinson’s disease, diabetes (divided into insulin-dependent diabetes and non-insulin-dependent diabetes), visual or hearing impairment (divided into visual impairment and hearing impairment), and fractures (divided into femur fractures and other fractures) (Table 1). Notably, among the disease list in the questionnaire on long-term care in the Comprehensive Survey of Living Conditions in Japan [19], we interpreted “respiratory disease” as lower respiratory tract diseases (because upper respiratory tract diseases are unlikely to result in long-term care), and “fractures or falls” as fractures (because ICD-10 codes [W00-W19] indicating falls were not recorded in our datasets). In addition, we did not include “spinal code injury” (suggested by ICD-10 codes S14, S24, S34) and “senility” (suggested by ICD-10 code R54) due to the extremely small number of people with these ICD-10 codes in our datasets.
Table 1
List of medical diagnosis categories and International Classification of Diseases 10th Revision codes
| Medical diagnosis category |
ICD-10 codes* |
| 1. Cerebrovascular diseases: |
|
| – Hemorrhagic stroke |
I60–I62 |
| – Ischemic stroke |
I63 |
| – Other cerebrovascular diseases** |
I64–I69 |
| 2. Cardiac diseases: |
|
| – Ischemic heart disease |
I20–I25 |
| – Arrhythmia |
I44, I45, I47–I49 |
| – Heart failure |
I50 |
| – Other cardiac diseases |
I01, I05–I09, I11, I13, I30–I43, I51, I52 |
| 3. Cancer |
C00–C97 |
| 4. Lower respiratory tract diseases |
|
| – Chronic obstructive pulmonary disease |
J43, J44 |
| – Pneumonia |
J12–18, J69 |
| – Other lower respiratory tract diseases |
A15, A16, J20–J22, J40–J42, J45–J47, J60–J68, J70, J80–J86, J90–J94 |
| 5. Joint diseases |
|
| – Rheumatoid arthritis |
M05, M06 |
| – Other arthropathies |
M00–03, M07, M10–M25 |
| – Dorsopathies (disorders of the back or spine) |
M40–M54 |
| 6. Dementia |
F00–F03, G30 |
| 7. Parkinson’s disease |
G20 |
| 8. Diabetes |
|
| – Insulin-dependent diabetes |
E10–E14 with prescription records of insulin products |
| – Non-insulin-dependent diabetes |
E10–E14 with prescription records of oral antidiabetic drugs (without insulin products) |
| 9. Visual or hearing impairment |
|
| – Visual impairment |
H53, H54 |
| – Hearing impairment |
H90, H91 |
| 10. Fractures |
|
| – Femur fractures |
S72 |
| – Other fractures |
S02, S12, S22, S32, S42, S52, S62, S82, S92, T02, T08, T10, T12 |
ICD-10 = International Classification of Diseases 10th Revision.
*Prescription records were additionally used to define diabetes.
**Including unspecified stroke and sequelae of cerebrovascular disease.
Next, two physicians (M.I. and Y.T.) among the authors independently and blindly chose corresponding ICD-10 codes for each medical diagnosis category. If there was discrepancy between the two physicians, another physician (T.M.) made decisions to produce the final ICD-10 code list. For this study, we used the three-character category level of ICD-10 codes (e.g., I50 for heart failure) because the four-character category level (e.g., I50.1 for left ventricular failure) was not always available in our datasets. To define and classify insulin-dependent diabetes and non-insulin-dependent diabetes, we additionally required prescription records of antidiabetic drugs, including oral antidiabetic drugs and insulin products. The finally established list of 22 medical diagnosis categories and corresponding ICD-10 codes are shown in Table 1.
We were interested in medical diagnoses recorded in the past six months, including the month when cases were certified for long-term care needs and the same month in matched controls. For example, if a case and controls were matched in September 2014, we defined the exposures of interest during the six months between April and September 2014. As sensitivity analyses, we examined medical diagnoses recorded (i) in the past twelve months, and (ii) in the past six months, but not including the calendar month when the cases were certified for long-term care needs (and the same month in the matched controls). For example, if a case and controls were matched in September 2014, we defined the exposures of interest during the six months between March and August 2014. We excluded “suspected” diagnosis codes from the datasets.
STATISTICAL ANALYSIS
First, we compared the proportion of people with ICD-10 codes in each medical diagnosis category between cases and controls using Wald tests (as part of univariable conditional logistic regression analyses) to take matched pairs into account.
Next, we conducted multivariable conditional logistic regression analyses to include all medical diagnoses at the same time (after we confirmed that any two diseases were not strongly correlated, with Pearson’s correlation coefficients <0.4, in the pooled data of matched cases and controls) for the incidence of long-term care needs certification, taking matched pairs into account.
As an additional analysis, to focus on more severe outcomes, we restricted the multivariable conditional logistic regression analysis to cases certified as care need levels 2 to 5 and their matched controls.
In addition, to investigate whether the patterns in the associations between recorded medical diagnoses and long-term care needs certification were similar or different between the two cities, we repeated the analyses separately by city.
All analyses were conducted using STATA version 15 (Stata Corp, Texas). P-values <0.05 were considered statistically significant.
RESULTS
There were 38,338 eligible people aged ≥75 years without any previous long-term care needs certification, of which 5,434 (14.2%) were newly certified for long-term care needs during the study period (Fig. 1), including people with support levels 1 (n = 1,392) and 2 (n = 801), and care need levels 1 (n = 1,258), 2 (n = 809), 3 (n = 465), 4 (n = 416), and 5 (n = 293). Through case-control matching, we selected 5,434 cases and 21,736 controls.
Fig. 1 Flow chart to select cases and controls.
Table 2 shows the number and proportion of people with ICD-10 codes in each medical diagnosis category recorded in the past six months among matched cases and controls. The recorded diagnoses of all medical conditions except for non-insulin-dependent diabetes and visual and hearing impairment were significantly more common in the cases than in the controls.
Table 2
Distribution of demography and medical diagnoses recorded in the past six months among matched cases and controls
|
Cases
N = 5,434
n (%) |
Controls
N = 21,736
n (%) |
P-value |
| Age category (years): |
|
|
— |
| 75–79 |
1,399 (25.8) |
5,596 (25.8) |
|
| 80–84 |
1,813 (33.4) |
7,252 (33.4) |
|
| 85–89 |
1,513 (27.8) |
6,052 (27.8) |
|
| 90– |
709 (13.1) |
2,836 (13.1) |
|
| Sex: |
|
|
— |
| Men |
2,175 (40.0) |
8,700 (40.0) |
|
| Women |
3,259 (60.0) |
13,036 (60.0) |
|
| Region: |
|
|
— |
| City A |
4,526 (83.3) |
18,104 (83.3) |
|
| City B |
908 (16.7) |
3,632 (16.7) |
|
| Recorded medical diagnosis: |
|
|
|
| Hemorrhagic stroke |
148 (2.7) |
134 (0.6) |
<0.001 |
| Ischemic stroke |
1,082 (19.9) |
2,335 (10.7) |
<0.001 |
| Other cerebrovascular diseases |
859 (15.8) |
1,989 (9.2) |
<0.001 |
| Ischemic heart disease |
1,398 (25.7) |
4,332 (19.9) |
<0.001 |
| Arrhythmia |
1,077 (19.8) |
3,149 (14.5) |
<0.001 |
| Heart failure |
1,438 (26.5) |
3,389 (15.6) |
<0.001 |
| Other cardiac diseases |
703 (12.9) |
1,878 (8.6) |
<0.001 |
| Cancer |
1,083 (19.9) |
2,452 (11.3) |
<0.001 |
| Chronic obstructive pulmonary disease |
348 (6.4) |
778 (3.6) |
<0.001 |
| Pneumonia |
548 (10.1) |
426 (2.0) |
<0.001 |
| Other lower respiratory tract diseases |
1,698 (31.3) |
4,819 (22.2) |
<0.001 |
| Rheumatoid arthritis |
206 (3.8) |
474 (2.2) |
<0.001 |
| Other arthropathies |
1,760 (32.4) |
5,712 (26.3) |
<0.001 |
| Dorsopathies (disorders of the back or spine) |
2,963 (54.5) |
9,456 (43.5) |
<0.001 |
| Dementia |
962 (17.7) |
681 (3.1) |
<0.001 |
| Parkinson’s disease |
135 (2.5) |
134 (0.6) |
<0.001 |
| Insulin-dependent-diabetes |
245 (4.5) |
331 (1.5) |
<0.001 |
| Non-insulin-dependent diabetes |
552 (10.2) |
2,037 (9.4) |
0.077 |
| Visual impairment |
297 (5.5) |
1,261 (5.8) |
0.34 |
| Hearing impairment |
291 (5.4) |
1,113 (5.1) |
0.483 |
| Femur fractures |
157 (2.9) |
64 (0.3) |
<0.001 |
| Other fractures |
658 (12.1) |
936 (4.3) |
<0.001 |
Table 3 shows the results of multivariable conditional logistic regression analysis. The adjusted odds ratio (95% confidence interval [CI]) was largest for femur fractures at 8.80 (6.35–12.20), followed by dementia at 6.70 (5.96–7.53), pneumonia at 3.72 (3.19–4.32), hemorrhagic stroke at 3.31 (2.53–4.34), Parkinson’s disease at 2.74 (2.07–3.63), other fractures at 2.68 (2.38–3.02), and insulin-dependent diabetes at 2.36 (1.94–2.86). The adjusted odds ratios (point estimates) of other medical diagnoses were below 2.
Table 3
Adjusted odds ratios of medical diagnoses recorded in the past six months on long-term care needs certification
|
Adjusted odds ratio (95% confidence interval) |
| Hemorrhagic stroke |
3.31 (2.53–4.34) |
| Ischemic stroke |
1.52 (1.39–1.67) |
| Other cerebrovascular diseases |
1.17 (1.05–1.29) |
| Ischemic heart disease |
1.02 (0.94–1.11) |
| Arrhythmia |
1.05 (0.96–1.16) |
| Heart failure |
1.40 (1.28–1.53) |
| Other cardiac diseases |
1.16 (1.04–1.30) |
| Cancer |
1.86 (1.70–2.04) |
| Chronic obstructive pulmonary disease |
1.25 (1.07–1.46) |
| Pneumonia |
3.72 (3.19–4.32) |
| Other lower respiratory tract diseases |
1.15 (1.07–1.25) |
| Rheumatoid arthritis |
1.34 (1.11–1.62) |
| Other arthropathies |
1.15 (1.07–1.24) |
| Dorsopathies (disorders of the back or spine) |
1.24 (1.16–1.33) |
| Dementia |
6.70 (5.96–7.53) |
| Parkinson’s disease |
2.74 (2.07–3.63) |
| Insulin-dependent diabetes |
2.36 (1.94–2.86) |
| Non-insulin dependent diabetes |
1.11 (1.00–1.24) |
| Visual impairment |
0.86 (0.74–0.99) |
| Hearing impairment |
0.88 (0.76–1.03) |
| Femur fractures |
8.80 (6.35–12.20) |
| Other fractures |
2.68 (2.38–3.02) |
In sensitivity analyses (Supplementary Appendix 1), the number and proportion of people with ICD-10 codes in each medical diagnosis category tended to be slightly increased by using the exposure definition of medical diagnoses recorded in the past twelve months, and slightly decreased by the definition of the past six months, but not including the calendar month when cases were certified for long-term care needs. The adjusted odds ratios were generally smaller than those estimated in the main analysis. However, the overall ranking in the size of the adjusted odds ratios was similar to the main analysis: fractures (especially femur fractures), dementia, pneumonia, hemorrhagic stroke, Parkinson’s disease, and insulin-dependent diabetes were identified as relatively strong risk factors for long-term care needs certification among the 22 categories.
In additional analysis, we focused on 1,983 cases certified as care need levels 2 to 5 and 7,932 matched controls. Characteristics of these patients are shown in Table 4. The recorded diagnoses of all medical conditions except for other arthropathies, non-insulin-dependent diabetes, and hearing impairment were significantly more common in the cases than in the controls. The adjusted odds ratios were generally larger than those estimated in the main analysis (Table 5). However, similar to the main analysis, fractures (especially femur fractures), dementia, pneumonia, hemorrhagic stroke, Parkinson’s disease, and insulin-dependent diabetes were identified as relatively strong risk factors for the certification of long-term care need levels 2 to 5. In addition, recorded diagnoses of cancer were strongly associated with the certification of long-term care need levels 2 to 5, at the adjusted odds ratio of 3.48 (3.00–4.04).
Table 4
Distribution of demography and medical diagnoses recorded in the past six months among cases with certification of care need levels 2 to 5 and their matched controls
|
Cases
N = 1,983
n (%) |
Controls
N = 7,932
n (%) |
P-value |
| Age category (years): |
|
|
— |
| 75–79 |
514 (25.9) |
2,056 (25.9) |
|
| 80–84 |
599 (30.2) |
2,396 (30.2) |
|
| 85–89 |
520 (26.2) |
2,080 (26.2) |
|
| 90– |
350 (17.7) |
1,400 (17.7) |
|
| Sex: |
|
|
— |
| Men |
903 (45.5) |
3,612 (45.5) |
|
| Women |
1,080 (54.5) |
4,320 (54.5) |
|
| Region: |
|
|
— |
| City A |
1,560 (78.7) |
6,240 (78.7) |
|
| City B |
423 (21.3) |
1,692 (21.3) |
|
| Recorded medical diagnosis: |
|
|
|
| Hemorrhagic stroke |
87 (4.4) |
55 (0.7) |
<0.001 |
| Ischemic stroke |
418 (21.1) |
849 (10.7) |
<0.001 |
| Other cerebrovascular diseases |
350 (17.7) |
735 (9.3) |
<0.001 |
| Ischemic heart disease |
481 (24.3) |
1,594 (20.1) |
<0.001 |
| Arrhythmia |
430 (21.7) |
1,164 (14.7) |
<0.001 |
| Heart failure |
606 (30.6) |
1,248 (15.7) |
<0.001 |
| Other cardiac diseases |
283 (14.3) |
660 (8.3) |
<0.001 |
| Cancer |
574 (29.0) |
905 (11.4) |
<0.001 |
| Chronic obstructive pulmonary disease |
158 (8.0) |
316 (4.0) |
<0.001 |
| Pneumonia |
356 (18.0) |
173 (2.2) |
<0.001 |
| Other lower respiratory tract diseases |
702 (35.4) |
1,738 (21.9) |
<0.001 |
| Rheumatoid arthritis |
66 (3.3) |
137 (1.7) |
<0.001 |
| Other arthropathies |
521 (26.3) |
2,025 (25.5) |
0.494 |
| Dorsopathies (disorders of the back or spine) |
1,033 (52.1) |
3,433 (43.3) |
<0.001 |
| Dementia |
377 (19.0) |
248 (3.1) |
<0.001 |
| Parkinson’s disease |
60 (3.0) |
63 (0.8) |
<0.001 |
| Insulin-dependent diabetes |
125 (6.3) |
119 (1.5) |
<0.001 |
| Non-insulin-dependent diabetes |
178 (9.0) |
779 (9.8) |
0.252 |
| Visual impairment |
72 (3.6) |
442 (5.6) |
0.001 |
| Hearing impairment |
84 (4.2) |
410 (5.2) |
0.089 |
| Femur fractures |
93 (4.7) |
26 (0.3) |
<0.001 |
| Other fractures |
265 (13.4) |
335 (4.2) |
<0.001 |
Table 5
Adjusted odds ratios of medical diagnoses recorded in the past six months on long-term care needs certification of care need levels 2 to 5 by restricting the analysis to cases with certification of care need levels 2 to 5 and their matched controls
|
Adjusted odds ratio (95% confidence interval) |
| Hemorrhagic stroke |
6.21 (4.06–9.50) |
| Ischemic stroke |
1.62 (1.37–1.92) |
| Other cerebrovascular diseases |
1.27 (1.05–1.52) |
| Ischemic heart disease |
0.84 (0.72–0.98) |
| Arrhythmia |
1.06 (0.90–1.25) |
| Heart failure |
1.64 (1.40–1.92) |
| Other cardiac diseases |
1.23 (1.01–1.51) |
| Cancer |
3.48 (3.00–4.04) |
| Chronic obstructive pulmonary disease. |
1.19 (0.92–1.54) |
| Pneumonia |
5.71 (4.51–7.21) |
| Other lower respiratory tract diseases |
1.28 (1.12–1.47) |
| Rheumatoid arthritis |
1.35 (0.92–1.97) |
| Other arthropathies |
0.82 (0.71–0.95) |
| Dorsopathies (disorders of the back or spine) |
1.13 (1.00–1.29) |
| Dementia |
7.64 (6.21–9.39) |
| Parkinson’s disease |
3.22 (2.05–5.05) |
| Insulin-dependent diabetes |
3.29 (2.38–4.56) |
| Non-insulin dependent diabetes |
0.98 (0.80–1.21) |
| Visual impairment |
0.54 (0.40–0.74) |
| Hearing impairment |
0.62 (0.46–0.83) |
| Femur fractures |
14.10 (8.32–23.88) |
| Other fractures |
3.34 (2.69–4.13) |
The results of analyses separately by city are shown in Supplementary Appendix 2. The patterns in the associations between recorded medical diagnoses and long-term care needs certification were roughly similar between the two cities: fractures (especially femur fractures), dementia, pneumonia, hemorrhagic stroke, and Parkinson’s disease were identified as relatively strong risk factors for long-term care needs certification in both cities. Insulin-dependent diabetes was significantly associated with the incidence of long-term care needs certification in City A, but not in City B, with the adjusted odds ratios (95%CI) of 2.54 (2.08–3.11) and 1.01 (0.50–2.06), respectively.
DISCUSSION
The number of people with long-term care needs certification has been increasing annually in Japan [3]. In this aging society with increasing life expectancy [20], how to reduce the burden of long-term care is a topic of debate. In the long term, interventions on upstream determinants of health (e.g., health literacy and social inequality) may be important to reduce the overall and regional variations of long-term care burdens. In the short term, interventions on more downstream determinants, including medical conditions resulting in long-term care, may be similarly or even more important. However, thus far there has been limited knowledge on the association between medical diagnoses and long-term care needs certification.
A recent study in a Japanese prefecture examined the distribution of nine medical diagnoses recorded in medical insurance claims data in people with long-term care needs certification [5]. Among 3,715 people aged ≥75 newly certified for long-term care needs, the most common medical diagnosis recorded in the past six months was cerebral vascular disorders (13.6%), followed by cancer (11.0%), fractures (9.2%), arthropathy (8.2%), and dementia (7.6%). However, the study did not establish a comparison group of people without long-term care needs certification [5], similarly to the 2016 Comprehensive Survey of Living Conditions which demonstrated that the most common primary reason for long-term care initiation was dementia (18.0%), followed by cerebrovascular disease (16.6%), senility (13.3%), fractures or falls (12.1%), and joint diseases (10.2%) [4]. Information on the disease prevalence among people receiving long-term care is insufficient to discuss the priority for intervention in the prevention of long-term care. Indeed, in our study, the prevalence of joint diseases was high among cases (i.e., people with long-term care needs certification), but its relative risk (i.e., adjusted odds ratios) was not large because the controls (i.e., people without long-term care needs certification) also had a high prevalence of joint diseases.
Some previous studies estimated the relative risks of medical conditions on long-term care [6–8]. A cross-sectional study including 1,004 community-dwelling older people found that strokes, joint pain, osteoporosis, and cancer showed statistically significant associations with long-term care [6]. A cohort study including 784 older people identified diabetes as a risk factor for long-term care initiation [7], while another cohort study reported that cognitive dysfunction was significantly associated with the incidence of long-term care needs certification [8]. However, these studies were not systematic enough to understand which medical conditions are more (or less) strongly associated with long-term care. To our knowledge, the current study was the first to examine a range of medical diagnoses and long-term care needs certification in a systematic and comprehensive manner. We found that fractures (especially femur fractures), dementia, pneumonia, hemorrhagic stroke, and Parkinson’s disease were relatively strong risk factors for the incidence of long-term care needs certification among 22 medical diagnosis categories.
The strengths of our study include (i) the use of a case-control design (instead of a cross-sectional design) to ensure the temporal relationship between medical diagnoses and long-term care needs certification, (ii) the best available method to prepare for a list of medical diagnosis categories and ICD-10 codes, and (iii) a range of additional analyses to ensure the robustness of the study findings.
However, we also need to acknowledge several limitations of the study. First, although the temporal relationship was ensured by the case-control design, it does not necessarily ensure causality between medical diagnoses and long-term care certification. Unlike the Comprehensive Survey of Living Conditions in Japan [3], this study could not investigate which medical diagnoses were direct reasons for long-term care needs. Therefore, caution may be needed to interpret our results; some diseases showing relatively high odds ratios may simply be predictive markers for long-term care needs rather than modifiable risk factors. For example, the large adjusted odds ratio of pneumonia might in part reflect the frailty of older people [21]. Second, the misclassification of recorded medical diagnoses (against “true” medical conditions as the gold-standard) is possible in medical claims data. Misclassification is generally classified as differential (meaning that the extent of the misclassification differs between the cases and controls) and non-differential. We tried to minimize differential misclassification by restricting cases and controls to people with at least one medical claim record (i.e., evidence of visiting a clinic or hospital) during the study period, so that both cases and controls had the chance to receive medical diagnoses (if they truly had the diseases). However, there may be non-differential misclassification, which could have diluted the true association between medical diagnoses and long-term care needs certification. Therefore, some medical conditions strongly associated with long-term care needs certification might have been masked due to non-differential misclassification. We excluded “suspected” diagnosis codes from the datasets to increase specificity and positive predictive value of recorded medical diagnoses. However, it is still possible that medical diagnoses were recorded for reimbursement of examination, such as a gastrointestinal cancer diagnosis for endoscopic screening and a diabetes diagnosis for HbA1c testing. To define and classify insulin-dependent diabetes and non-insulin-dependent diabetes, we additionally required prescription records of antidiabetic drugs. We believe that this strategy was effective to increase specificity and positive predictive value of diabetes definition (against “true” diabetes status as the gold-standard), while sensitivity could be reduced (because untreated diabetes is excluded by our definition). However, this strategy may not be always applicable to other medical conditions due to various patterns of treatment and examination. It is difficult to accurately estimate the influence of misclassification on the current study findings without information on the validity (e.g., sensitivity, specificity, and positive predictive value) of individual medical diagnoses or disease definitions. Validation studies of medical diagnosis codes or disease definition are thus needed in Japanese claims data [22]. Finally, previous studies adjusted for potential confounding factors including lifestyle and social factors [6–8], but the current study was unable to account for this information in the claims data.
From a viewpoint of high-risk approach (i.e., strategy for prevention upon people with the strongest likelihood of developing outcome [23]), fractures, dementia, pneumonia, hemorrhagic stroke, and Parkinson’s disease could be targets for reducing the long-term care burden in the society. Further research is warranted to investigate causality between these conditions and long-term care initiation. If the association we found reflects causality, early identification and appropriate treatment of the conditions would help prevent or delay the initiation of long-term care. Even without causality, recorded diagnoses of these conditions would be useful in identifying people at risk for long-term care needs.
CONCLUSION
In this study, by using linked medical and long-term insurance claims data from two Japanese cities, we found that among 22 medical diagnosis categories, recorded diagnoses of fractures (especially femur fractures), dementia, pneumonia, hemorrhagic stroke, and Parkinson’s disease were strongly associated with the incidence of new (i.e., first ever) long-term care needs certification. More public attention and research efforts should be focused on these conditions to reduce the burden of long-term care in the society.
ACKNOWLEDGMENTS
This study was supported by a grant-in-aid from the Ministry of Health, Labour and Welfare; Health and Labour Sciences Research Grant, Japan (H28-junkankitou-ippan-009, H30-choju-ippan-007, and H30-iryou-ippan-011). We would like to thank Editage (www.editage.jp) for its English language editing.
CONFLICT OF INTERESTS
Takahiro Mori’s joint appointment as an associate professor at the University of Tsukuba was sponsored by JMDC Inc. in the 2018 financial year (i.e., April 2018 to March 2019), and by SMS CO., LTD. in the 2019 financial year (i.e., April 2019 to the present). JMDC Inc. or SMS CO., LTD had no role in conducting this study.
Supplementary Appendix 1
Distribution of medical diagnoses recorded (i) in the past twelve months, and (ii) in the past six months, but not including the calendar month when cases were certified for long-term care needs and the same month in the matched controls, and adjusted odds ratios
|
(i) 12 months |
(ii) 6 months, but not including the matched month |
Cases
N = 5,434
n (%) |
Controls
N = 21,736
n (%) |
P-value |
Adjusted OR
(95% CI) |
Cases
N = 5,434
n (%) |
Controls
N = 21,736
n (%) |
P-value |
Adjusted OR
(95% CI) |
| Hemorrhagic stroke |
160 (2.9) |
159 (0.7) |
<0.001 |
3.03 (2.35–3.89) |
115 (2.1) |
130 (0.6) |
<0.001 |
2.60 (1.97–3.42) |
| Ischemic stroke |
1,141 (21.0) |
2,510 (11.6) |
<0.001 |
1.50 (1.37–1.64) |
954 (17.6) |
2,300 (10.6) |
<0.001 |
1.42 (1.30–1.56) |
| Other cerebrovascular diseases |
920 (16.9) |
2,146 (9.9) |
<0.001 |
1.18 (1.07–1.30) |
759 (14.0) |
1,964 (9.0) |
<0.001 |
1.13 (1.02–1.25) |
| Ischemic heart disease |
1,453 (26.7) |
4,530 (20.8) |
<0.001 |
1.02 (0.94–1.11) |
1,307 (24.1) |
4,284 (19.7) |
<0.001 |
0.99 (0.92–1.08) |
| Arrhythmia |
1,157 (21.3) |
3,450 (15.9) |
<0.001 |
1.06 (0.97–1.15) |
991 (18.2) |
3,116 (14.3) |
<0.001 |
1.01 (0.92–1.10) |
| Heart failure |
1,517 (27.9) |
3,652 (16.8) |
<0.001 |
1.40 (1.28–1.53) |
1,273 (23.4) |
3,323 (15.3) |
<0.001 |
1.35 (1.24–1.47) |
| Other cardiac diseases |
768 (14.1) |
2,117 (9.7) |
<0.001 |
1.12 (1.01–1.25) |
646 (11.9) |
1,854 (8.5) |
<0.001 |
1.14 (1.02–1.27) |
| Cancer |
1,135 (20.9) |
2,650 (12.2) |
<0.001 |
1.83 (1.67–2.00) |
979 (18.0) |
2,432 (11.2) |
<0.001 |
1.67 (1.53–1.83) |
| COPD |
379 (7.0) |
871 (4.0) |
<0.001 |
1.27 (1.10–1.47) |
305 (5.6) |
775 (3.6) |
<0.001 |
1.21 (1.04–1.41) |
| Pneumonia |
616 (11.3) |
637 (2.9) |
<0.001 |
2.95 (2.58–3.38) |
373 (6.9) |
397 (1.8) |
<0.001 |
2.80 (2.38–3.29) |
| Other LRTDs |
1,977 (36.4) |
6,084 (28.0) |
<0.001 |
1.09 (1.01–1.18) |
1,535 (28.3) |
4,723 (21.7) |
<0.001 |
1.11 (1.03–1.19) |
| Rheumatoid arthritis |
221 (4.1) |
536 (2.5) |
<0.001 |
1.34 (1.12–1.60) |
196 (3.6) |
474 (2.2) |
<0.001 |
1.39 (1.16–1.67) |
| Other arthropathies |
1,947 (35.8) |
6,476 (29.8) |
<0.001 |
1.11 (1.03–1.19) |
1,638 (30.1) |
5,678 (26.1) |
<0.001 |
1.05 (0.98–1.13) |
| Dorsopathies |
3,148 (57.9) |
10,333 (47.5) |
<0.001 |
1.21 (1.13–1.30) |
2,791 (51.4) |
9,356 (43.0) |
<0.001 |
1.17 (1.09–1.25) |
| Dementia |
986 (18.2) |
725 (3.3) |
<0.001 |
6.34 (5.66–7.10) |
711 (13.1) |
633 (2.9) |
<0.001 |
4.74 (4.20–5.35) |
| Parkinson’s disease |
137 (2.5) |
143 (0.7) |
<0.001 |
2.55 (1.94–3.35) |
124 (2.3) |
128 (0.6) |
<0.001 |
2.76 (2.10–3.62) |
| Insulin-dependent diabetes |
256 (4.7) |
353 (1.6) |
<0.001 |
2.32 (1.92–2.79) |
201 (3.7) |
318 (1.5) |
<0.001 |
2.14 (1.76–2.60) |
| Non-insulin-dependent diabetes |
560 (10.3) |
2,064 (9.5) |
0.07 |
1.11 (0.99–1.24) |
545 (10.0) |
2,015 (9.3) |
0.086 |
1.07 (0.96–1.19) |
| Visual impairment |
375 (6.9) |
1,452 (6.7) |
0.561 |
0.96 (0.84–1.09) |
299 (5.5) |
1,259 (5.8) |
0.41 |
0.87 (0.76–1.00) |
| Hearing impairment |
395 (7.3) |
1,530 (7.0) |
0.553 |
0.88 (0.77–1.00) |
286 (5.3) |
1,109 (5.1) |
0.63 |
0.91 (0.79–1.05) |
| Femur fracture |
162 (3.0) |
85 (0.4) |
<0.001 |
6.72 (5.01–9.00) |
95 (1.8) |
55 (0.3) |
<0.001 |
5.61 (3.91–8.05) |
| Other fractures |
754 (13.9) |
1,210 (5.6) |
<0.001 |
2.36 (2.12–2.63) |
491 (9.0) |
906 (4.2) |
<0.001 |
1.96 (1.73–2.21) |
CI = confidence interval, COPD = chronic obstructive pulmonary disease, LRTDs = lower respiratory tract diseases, OR = odds ratio.
Supplementary Appendix 2
Distribution of demography, care need levels, and medical diagnoses recorded in the past six months among matched cases and controls and adjusted odds ratios estimated separately by city
|
City A |
City B |
Cases
N = 4,526
n (%) |
Controls
N = 18,104
n (%) |
P-value |
Adjusted OR
(95% CI) |
Cases
N = 908
n (%) |
Controls
N = 3,632
n (%) |
P-value |
Adjusted OR
(95% CI) |
| Age category (years): |
|
|
— |
|
|
|
— |
|
| 75–79 |
1,209 (26.7) |
4,836 (26.7) |
|
— |
190 (20.9) |
760 (20.9) |
|
— |
| 80–84 |
1,539 (34.0) |
6,156 (34.0) |
|
— |
274 (30.2) |
1,096 (30.2) |
|
— |
| 85–89 |
1,234 (27.3) |
4,936 (27.3) |
|
— |
279 (30.7) |
1,116 (30.7) |
|
— |
| >90 |
544 (12.0) |
2,176 (12.0) |
|
— |
165 (18.2) |
660 (18.2) |
|
— |
| Sex: |
|
|
— |
|
|
|
— |
|
| Men |
1,834 (40.5) |
7,336 (40.5) |
|
— |
341 (37.6) |
1,364 (37.6) |
|
— |
| Women |
2,692 (59.5) |
10,768 (59.5) |
|
— |
567 (62.4) |
2,268 (62.4) |
|
— |
| Care need levels: |
|
|
— |
|
|
|
— |
|
| Support level 1 |
1,265 (28.0) |
— |
|
— |
127 (14.0) |
— |
|
— |
| Support level 2 |
669 (14.8) |
— |
|
— |
132 (14.5) |
— |
|
— |
| Care need level 1 |
1,032 (22.8) |
— |
|
— |
226 (24.9) |
— |
|
— |
| Care need level 2 |
658 (14.5) |
— |
|
— |
151 (16.6) |
— |
|
— |
| Care need level 3 |
369 (8.2) |
— |
|
— |
96 (10.6) |
— |
|
— |
| Care need level 4 |
308 (6.8) |
— |
|
— |
108 (11.9) |
— |
|
— |
| Care need level 5 |
225 (5.0) |
— |
|
— |
68 (7.5) |
— |
|
— |
| Recorded medical diagnosis: |
|
|
|
|
|
|
|
|
| Hemorrhagic stroke |
123 (2.7) |
124 (0.7) |
<0.001 |
2.89 (2.17–3.84) |
25 (2.8) |
10 (0.3) |
<0.001 |
10.82 (4.53–25.83) |
| Ischemic stroke |
932 (20.6) |
2,096 (11.6) |
<0.001 |
1.44 (1.30–1.59) |
150 (16.5) |
239 (6.6) |
<0.001 |
2.21 (1.69–2.88) |
| Other cerebrovascular diseases |
757 (16.7) |
1,794 (9.9) |
<0.001 |
1.19 (1.06–1.32) |
102 (11.2) |
195 (5.4) |
<0.001 |
1.06 (0.77–1.46) |
| Ischemic heart disease |
1,213 (26.8) |
3,758 (20.8) |
<0.001 |
1.02 (0.93–1.12) |
185 (20.4) |
574 (15.8) |
0.001 |
1.02 (0.81–1.28) |
| Arrhythmia |
928 (20.5) |
2,737 (15.1) |
<0.001 |
1.04 (0.94–1.15) |
149 (16.4) |
412 (11.3) |
<0.001 |
1.14 (0.88–1.48) |
| Heart failure |
1,200 (26.5) |
2,834 (15.7) |
<0.001 |
1.42 (1.29–1.57) |
238 (26.2) |
555 (15.3) |
<0.001 |
1.33 (1.06–1.68) |
| Other cardiac diseases |
583 (12.9) |
1,584 (8.8) |
<0.001 |
1.15 (1.02–1.30) |
120 (13.2) |
294 (8.1) |
<0.001 |
1.17 (0.87–1.56) |
| Cancer |
922 (20.4) |
2,125 (11.7) |
<0.001 |
1.83 (1.66–2.02) |
161 (17.7) |
327 (9.0) |
<0.001 |
2.10 (1.64–2.68) |
| COPD |
284 (6.3) |
657 (3.6) |
<0.001 |
1.27 (1.07–1.50) |
64 (7.1) |
121 (3.3) |
<0.001 |
1.04 (0.69–1.58) |
| Pneumonia |
419 (9.3) |
367 (2.0) |
<0.001 |
3.29 (2.78–3.89) |
129 (14.2) |
59 (1.6) |
<0.001 |
6.25 (4.31–9.07) |
| Other LRTDs |
1,378 (30.5) |
4,042 (22.3) |
<0.001 |
1.11 (1.02–1.21) |
320 (35.2) |
777 (21.4) |
<0.001 |
1.43 (1.18–1.75) |
| Rheumatoid arthritis |
164 (3.6) |
395 (2.2) |
<0.001 |
1.32 (1.07–1.62) |
42 (4.6) |
79 (2.2) |
<0.001 |
1.57 (0.97–2.53) |
| Other arthropathies |
1,498 (33.1) |
4,789 (26.5) |
<0.001 |
1.20 (1.10–1.30) |
262 (28.9) |
923 (25.4) |
0.033 |
0.92 (0.76–1.13) |
| Dorsopathies |
2,482 (54.8) |
7,943 (43.9) |
<0.001 |
1.26 (1.16–1.36) |
481 (53.0) |
1,513 (41.7) |
<0.001 |
1.16 (0.97–1.40) |
| Dementia |
817 (18.1) |
583 (3.2) |
<0.001 |
6.73 (5.93–7.63) |
145 (16.0) |
98 (2.7) |
<0.001 |
6.92 (5.07–9.46) |
| Parkinson’s disease |
112 (2.5) |
105 (0.6) |
<0.001 |
2.82 (2.07–3.85) |
23 (2.5) |
29 (0.8) |
<0.001 |
2.55 (1.28–5.07) |
| Insulin-dependent diabetes |
227 (5.0) |
297 (1.6) |
<0.001 |
2.54 (2.08–3.11) |
18 (2.0) |
34 (0.9) |
0.01 |
1.01 (0.50–2.06) |
| Non-insulin-dependent diabetes |
471 (10.4) |
1,806 (10.0) |
0.388 |
1.08 (0.96–1.21) |
81 (8.9) |
231 (6.4) |
0.006 |
1.41 (1.03–1.92) |
| Visual impairment |
263 (5.8) |
1,107 (6.1) |
0.443 |
0.85 (0.73–0.99) |
34 (3.7) |
154 (4.2) |
0.502 |
1.03 (0.67–1.59) |
| Hearing impairment |
271 (6.0) |
1,033 (5.7) |
0.466 |
0.89 (0.76–1.04) |
20 (2.2) |
80 (2.2) |
1 |
0.92 (0.52–1.60) |
| Femur fractures |
130 (2.9) |
63 (0.4) |
<0.001 |
7.43 (5.30–10.42) |
27 (3.0) |
<5 (<0.1) |
<0.001 |
81.70 (10.70–624.09) |
| Other fractures |
544 (12.0) |
806 (4.5) |
<0.001 |
2.52 (2.21–2.87) |
114 (12.6) |
130 (3.6) |
<0.001 |
3.79 (2.79–5.14) |
CI = confidence interval, COPD = chronic obstructive pulmonary disease, LRTDs = lower respiratory tract diseases, OR = odds ratio.
REFERENCES
- 1. Tamiya N, Noguchi H, Nishi A, Reich MR, Ikegami N, Hashimoto H, et al. Population Ageing and Wellbeing: Lessons from Japan’s Long-term Care Insurance Policy. Lancet 2011;378:1183–1192.
- 2. Moriyama Y, Tamiya N, Kamimura A, Sandoval F, Luptak M. Doctors’ Opinion Papers in Long-term Care Need Certification in Japan: Comparison between Clinic and Advanced Treatment Hospital Settings. Public Policy and Administration Research 2014;4:31–37.
- 3. Cabinet Office of Japan. Annual Report on Aging Society 2018 (in Japanese). Available from: https://www8.cao.go.jp/kourei/whitepaper/w-2018/html/zenbun/s1_2_2.html, Accessed 2019 May 14.
- 4. Ministry of Health, Labour and Welfare of Japan. Comprehensive Survey of Living Conditions. Available from: http://www.mhlw.go.jp/english/database/db-hss/cslc.html, Accessed 2019 May 14.
- 5. Naruse T, Sakai M, Matsumoto H, Nagata S. Diseases that Precede Disability Among Latter-stage Elderly Individuals in Japan. Biosci Trends 2015;9:270–274.
- 6. Kiyoshige E, Kabayama M, Gondo Y, Masui Y, Ryuno H, Sawayama Y, et al. Association Between Long-term Care and Chronic and Lifestyle-related Disease Modified by Social Profiles in Community-dwelling People Aged 80 and 90; SONIC Study. Arch Gerontol Geriatr 2019;81:176–181.
- 7. Moriya S, Murata A, Kimura S, Inoue N, Miura H. Predictors of Eligibility for Long-term care Funding for Older People in Japan. Australas J Ageing 2013;32:79–85.
- 8. Fujiwara Y, Amano H, Kumagai S, Yoshida H, Fujita K, Naito T, et al. Physical and Psychological Predictors for the Onset of Certification of Long-term Care Insurance Among Older Adults Living Independently in a Community: A 40-month Follow-up Study (in Japanese). Nihon Koshu Eisei Zasshi 2006;53:77–91.
- 9. Ikegami N, Yoo BK, Hashimoto H, Matsumoto M, Ogata H, Babazono A, et al. Japanese Universal Health Coverage: Evolution, Achievements, And Challenges. Lancet 2011;378:1106–1115.
- 10. Mori T, Hamada S, Yoshie S, Jeon B, Jin X, Takahashi H, et al. The Associations of Multimorbidity with the Sum of Annual Medical and Long-term Care Expenditures in Japan. BMC Geriatr 2019;19:69.
- 11. Kuroda N, Hamada S, Sakata N, Jeon B, Iijima K, Yoshie S, et al. Antipsychotic Use and Related Factors Among People with Dementia Aged 75 Years or Older in Japan: A Comprehensive Population-based Estimation Using Medical and Long-term Care Data. Int J Geriatr Psychiatry 2019;34:472–479.
- 12. Hamada S, Takahashi H, Sakata N, Jeon B, Mori T, Iijima K, et al. Household Income Relationship with Health Services Utilization and Healthcare Expenditures in People Aged 75 Years or Older in Japan: A Population-based Study Using Medical and Long-term Care Insurance Claims Data. J Epidemiol 2018 (in press).
- 13. Mori T, Tamiya N, Jin X, Jeon B, Yoshie S, Iijima K, Ishizaki T. Estimated Expenditures for Hip Fractures Using Merged Healthcare Insurance Data for Individuals Aged >/= 75 years and Long-term Care Insurance Claims Data in Japan. Arch Osteoporos 2018;13:37.
- 14. ICD-10 Corresponding Diagnosis Code Master (in Japanese). Available from: http://www2.medis.or.jp/stdcd/byomei/index.html, Accessed 2019 May 14.
- 15. Wacholder S, McLaughlin JK, Silverman DT, Mandel JS. Selection of controls in case-control studies. I. Principles. Am J Epidemiol 1992;135:1019-1028.
- 16. Wacholder S, Silverman DT, McLaughlin JK, Mandel JS. Selection of controls in case-control studies. II. Types of controls. Am J Epidemiol 1992;135:1029-1041.
- 17. Wacholder S, Silverman DT, McLaughlin JK, Mandel JS. Selection of controls in case-control studies. III. Design options. Am J Epidemiol 1992;135:1042-1050.
- 18. Vandenbroucke JP, Pearce N. Case-control studies: basic concepts. Int J Epidemiol 2012;41:1480-1489.
- 19. A Questionnaire on Long-term Care in the Comprehensive Survey of Living Conditions in Japan (in Japanese). Available from: https://www.mhlw.go.jp/toukei/chousahyo/koku28ka.pdf, Accessed 2019 May 14.
- 20. Nomura S, Sakamoto H, Glenn S, Tsugawa Y, Abe SK, Rahman MM, et al. Population Health and Regional Variations of Disease Burden in Japan, 1990–2015: A Systematic Subnational Analysis for the Global Burden of Disease Study 2015. Lancet 2017;390:1521–1538.
- 21. Falcone M, Blasi F, Menichetti F, Pea F, Violi F. Pneumonia in Frail Older Patients: An Up to Date. Intern Emerg Med 2012;7:415–424.
- 22. Iwagami M, Aoki K, Akazawa M, Ishiguro C, Imai S, Ooba N, et al. Task Force Report on the Validation of Diagnosis Codes and Other Outcome Definitions in the Japanese Receipt Data (in Japanese). Jpn J Pharmacoepidemiol 2018;23:95–123.
- 23. Platt JM, Keyes KM, Galea S. Efficiency or equity? Simulating the impact of high-risk and population intervention strategies for the prevention of disease. SSM Popul Health 2016;3:1-8.
