International Classification of Diseases 10th edition-based disability adjusted life years for measuring of burden of specific injury.

OBJECTIVE: We aimed to develop an International Classification of Diseases (ICD) 10th edition injury code-based disability-adjusted life year (DALY) to measure the burden of specific injuries.
METHODS: Three independent panels used novel methods to score disability weights (DWs) of 130 indicator codes sampled from 1,284 ICD injury codes. The DWs were interpolated into the remaining injury codes (n=1,154) to estimate DWs for all ICD injury codes. The reliability of the estimated DWs was evaluated using the test-retest method. We calculated ICD-DALYs for individual injury episodes using the DWs from the Korean National Hospital Discharge Injury Survey (HDIS, n=23,160 of 2004) database and compared them with DALY based on a global burden of disease study (GBD-DALY) regarding validation, correlation, and agreement for 32 injury categories.
RESULTS: Using 130 ICD 10th edition injury indicator codes, three panels determined the DWs using the highest reliability (person trade-off 1, Spearman r=0.724, 0.788, and 0.875 for the three panel groups). The test-retest results for the reliability were excellent (Spearman r=0.932) (P<0.001). The HDIS database revealed injury burden (years) as follows: GBD-DALY (138,548), GBD-years of life disabled (130,481), and GBD-years of life lost (8,117) versus ICD-DALY (262,246), ICD-years of life disabled (255,710), and ICD-years of life lost (6,537), respectively. Spearman's correlation coefficient of the DALYs between the two methods was 0.759 (P<0.001), and the Bland-Altman test displayed an acceptable agreement, with exception of two categories among 32 injury groups.
CONCLUSION: The ICD-DALY was developed to calculate the burden of injury for all injury codes and was validated with the GBD-DALY. The ICD-DALY was higher than the GBD-DALY but showed acceptable agreement.

INTRODUCTION

Injury is the leading cause of death among children and young people, and it is the leading cause of years of potential life lost in most countries [1]. Injury may also result in disability, which diminishes the subject’s quality of life [2]. The disability-adjusted life year (DALY) was created to measure the burden of disease after adjusting for both mortality and disability, and is a useful method for health policy decision-making [3-6]. The DALY has been used to evaluate the global burden of disease (GBD) for major diseases, including injury [7-10]. However, injury is not a single disease entity but a group of very complex processes consisting of multiple injury mechanisms and natures of injury. Previous GBD studies have categorized injury into 32 distinct injury groups [11]. Although the burden of injury may be measured using these simple categories, the GBD of injury may be too crude to reveal the burden of each specific injury. For example, the GBD group classified the burden of poisoning as a single category. However, poisoning involves various materials, ranging from mild substances (i.e., sedatives) to extremely fatal substances (i.e., paraquat) [12]. A more specific method for measuring the disability associated with specific injuries would allow researchers and administrators in more precisely assessing subject disability in large databases. The International Classification of Diseases (ICD) includes all injuries, injury subgroups, and adverse effects. The study hypothesis is that developing a method to measure the disability of each injury diagnostic code (ICD 10th edition S/T codes) would allow the calculation of the entire burden of specific injuries. The objectives of the current study were to develop an ICD 10th edition-based disability-adjusted life year (DALY) (ICD-DALY) for injury and to test its reliability and validity.

METHODS

This study was supported by the Ministry of Health, Welfare, and Family Affairs of Republic of the Korea in 2008 (Health Promotion Fund A0104208A00). The Seoul National University Hospital institutional review board approved the study with waiver of informed consent because the study did not require the enrollment of human participants. Patient records/information were anonymized and de-identified prior to analysis. The study flow diagram is presented in Fig. 1.

Fig. 1.

Study flow diagram. DW, disability weight; ICD-10, International Classification of Diseases 10th edition; VAS, visual analogue scale; PTO1, person trade-off 1; PTO2, person trade-off 2; TTO, time trade-off; DALY, disability-adjusted life years; GBD, Global Burden of Disease Study.

Determining the disability weight for injury codes

To determine the disability weight (DW) for each injury S/T code (n=1,284 codes), we randomly sampled 10% (n=130 codes) of the ICD 10th edition (ICD-10) injury codes from an existing injury database (National Injury Database, NIDB) using a stratified random sampling. The NIDB (n=29,285,528) included total injury data for all patients who had utilized medical services between 2001 and 2003 in Korea; 93.7% for outpatients, 6.0% for hospital admissions, and 0.3% for death after treatment [13]. Using the NIDB, we calculated the admission rate ratio (ARR) as the number of deaths and hospital admissions associated with a specific ICD injury code over the total number of patients corresponding to the same ICD injury code. Stratified by the ARR distribution per 10 percentiles, we randomly sampled the indicator injury codes in even order (n=130) (Appendix 1). As the ARR reflects morbidity and mortality associated with specific injury codes, the 10th percentile of the ARR represents a stratum for which no valid parameter exists.

The DW for each injury code was determined using the same method applied by the GBD research group [11]. We established three panel groups, each composed of six experts (five emergency and trauma care physicians and one preventive medicine physician) and one trained coordinator. There was no panel turnover and there were no panel dropouts during the one full-day survey workshop for DW measurement. Each panel group was assigned to separate rooms to avoid any bias. The 130 indicator injury codes were randomly assigned to one of the three groups. An additional 16 codes were shared by the groups to calculate the level of agreement between the groups via correlation analysis. Ideally, all three panels should evaluate all 130 indicator diagnoses independently. However, cost and time limitations existed. Therefore, we assigned a fixed number of codes (16 common codes and 38 different codes) per panel with the assumption that approximately 10 minutes were required to determine the DW of each injury code. We benchmarked this method on the basis of previous studies [14,15]. To determine the DW, we utilized four separate methods, as follows: visual analogue scale (VAS), person trade-off 1 (PTO1), person trade-off 2 (PTO2), and time trade-off (TTO). These were used to determine the DW in prior studies [16,17]. For example, in the VAS method, one panel member discloses their VAS result regarding a specific code to the other members in their group and explains their reasoning. After discussion, each panel reconsiders the VAS for each injury code and documents the score on a designed questionnaire. This process was performed and repeated for all injury codes (54 codes) assigned to each panel. The median value for every code assessed using each method was calculated and considered to be the DW score. After calculating the DW score, we selected the most reliable of the four methods (VAS, PTO1, PTO2, or TTO) by calculating Spearman’s correlation for the 16 common codes assigned to the three panel groups.

Upon determining the DW for each indicator injury code (n=130), we estimated the DW for the remaining injury codes using the interpolation method [18]. Each panel scored the DW for the remaining 1,154 S/T codes (i.e., 385 codes for each of the three panel groups). The panels referenced the DW values of the 130 indicator injury codes. For this procedure, the panels used the VAS method, as it required the least amount of time to generate an agreement. After scoring the DW for all injury codes, the median VAS among the panels was calculated to determine the DW for each injury code. Finally, we estimated the DWs for all 1,284 ICD-10 injury codes.

To determine the reliability of the DW of each code, we used the test-retest method for the 16 common indicator injury codes. Two months after completing the initial panel study, we repeated the same panel study using the same procedural methods for VAS, PTO1, PTO2, and TTO. Pearson's correlation analysis was used to assess the reliability of the test-retest method.

Calculating the ICD-DALY using the estimated DW of injury codes

To calculate the DALY, we used the same formula, which has been proposed in several previous studies [2,3]. The ICD-DALY was calculated using the same formula as the conventional DALY (GBD-DALY), but the specific DW of the ICD codes and assumptions regarding duration of morbidity and life expectancy differed. We considered the discount rate (γ=0.03), age weight parameter (β=0.04), modulation factor (K=0 or 1) according to age weight, and constant (C=0.1658) to calculate the DALY for each injury episode. The estimated life expectancy (“L” in years of life lost [YLL]) or duration of morbidity due to a specific injury (“L” in years of life disabled [YLD]) was calculated from the age at the time of the injury event to the life expectancy, based on the life tables of the National Statistical Office of Korea. According to previous GBD studies, the cure rate of injury was 0% for the YLD measurement, and we surrogated the remaining life expectancy for morbidity due to a specific injury [19,20]. A comparison of essential variables of the ICD-DALY and GBD-DALY calculations are shown in Appendix 2.

Validation of the ICD-DALY

To validate the ICD-DALY, we compared it with the conventional GBD-DALY. We assessed and compared both DALY results (ICD-DALY vs. GBD-DALY) using a pre-existing injury database, the Korean National Hospital Discharge Injury Survey. The Hospital Discharge Injury Survey (HDIS) is a nationwide, stratified sampled, and abstract survey data from 170 general hospitals with more than 100 beds in Korea since 2004 by Korea Centers for Disease Control and Prevention, including all patients with diagnosis code (ICD S or T codes) at discharge hospital. We recategorized 1,284 injury-related ICD codes into 32 categories using the “GBD 2000 nature of injury categories and ICD codes” (http://www.who.int/healthinfo/statistics/bod_injuries.pdf ). We calculated the GBD-DALY by using the mean age of injury event, mean duration of injury prevalence (YLD) or mean life expectancy (YLL) using DisMod II (http://www.who.int/healthinfo/global_burden_disease/tools_software/en/). The DisMod II is a software tool that can be used to verify the consistency of estimates of incidence, prevalence, duration, and case fatality for diseases. We estimated the incidence and case fatality for 32 categories of injury, by sex and age group, and replaced the remission rate with 0%, as previously reported [21,22]. We calculated the ICD-DALY using individual age at the time of injury, individual age at death, and the estimated life expectancy based on the life tables of the National Statistical Office of Korea and ICD-10 code–based DW. Spearman’s rank correlation test and the Bland-Altman test for agreement were used to compare 32 injury categories between ICD-DALY and GBD-DALY.

RESULTS

DW estimations and the calculation of the ICD-DALY

We sampled 130 indicator ICD-10 injury codes based on the distribution of the ARR of each ICD-10 code in the NIDB. Appendix 1 lists the 130 selected indicator ICD-10 injury codes. Table 1 displays the distribution injury codes according to the ARR. The 130 codes were distributed similarly according to the strata of the ARR percentile. Table 2 displays Spearman’s rank correlation coefficients among the panel groups for each of the four methods (i.e., VAS, PTO1, PTO2, and TTO). The PTO1 method demonstrated the highest correlation coefficient among the three panel groups (0.788, 0.685, and 0.875). Fig. 2 shows the distribution of the median DWs for the 16 common indicator injury codes derived from a panel study applying four different methods. The values derived using the VAS and PTO1 methods were distributed evenly throughout the total range of 0.0 to 1.0, while the values derived using the PTO2 and TTO methods were densely concentrated below 0.2. Therefore, we selected the DW for 160 indicator ICD injury codes derived using the PTO1 method, which was considered the most reliable and discriminative method for determining DW. We determined the DW of the remaining ICD-10 injury codes using the interpolation method. Fig. 3 shows the markedly high correlation between the VAS- and PTO1-based DWs for the 130 indicator ICD-10 injury codes (ρ=0.721, P<0.001). Using the VAS-based interpolation method, we calculated the median values of the DWs for each ICD-10 code. Appendix 3 displays the DWs of all injury codes.

Table 1.

Indicator injury ICD-10 codes sampled from total injury ICD-10 codes using the admission rate ratio

Admission rate ratio	Injury ICD-10 codes, total	Indicator injury ICD-10 codes sampled
Total	1,284 (100.0)	130 (100.0)
0.00–0.05	181 (14.1)	18 (13.8)
0.05–0.15	295 (23.0)	30 (23.1)
0.15–0.25	208 (16.2)	21 (16.2)
0.25–0.35	146 (11.4)	15 (11.5)
0.35–0.45	113 (8.8)	11 (8.5)
0.45–0.55	108 (8.4)	11 (8.5)
0.55–0.65	73 (5.7)	7 (5.4)
0.65–0.75	75 (5.8)	8 (6.2)
0.75–0.85	47 (3.7)	5 (3.8)
0.85–0.95	32 (2.5)	3 (2.3)
0.95–1.00	6 (0.5)	1 (0.8)

Values are presented as number (%).

ICD-10, International Classification of Diseases 10th edition.

Table 2.

Spearman rank correlation coefficients between two panels according to valuation methods

Valuation method	Panel 1–2	Panel 2–3	Panel 1–3
VAS	0.686 (P=0.003)	0.708 (P=0.002)	0.690 (P=0.003)
PTO1	0.788 (P=0.003)	0.685 (P=0.003)	0.875 (P=0.001)
PTO2	0.549 (P=0.028)	0.685 (P=0.003)	0.754 (P=0.007)
TTO	0.403 (P=0.014)	0.510 (P=0.052)	0.723 (P=0.023)

VAS, visual analogue scale; PTO1, person trade-off 1; PTO2, person trade-off 2; TTO, time trade-off.

Fig. 2.

Distributions of median disability weight (DW) calculated by each panel method. VAS, visual analogue scale; PTO1, person trade-off 1; PTO2, person trade-off 2; TTO, time trade-off.

Fig. 3.

Correlation between person trade-off 1 (PTO1)-based and visual analogue scale (VAS)-based disability weights (DWs) for 130 indicator codes.

We performed a test-retest study on the 16 common indicator ICD-10 injury codes. Pearson's correlation coefficient and Spearman’s rank correlation coefficient between the test-retest of DW valuation via the PTO1 method were 0.932 (P<0.001) and 0.740 (P<0.001), respectively, thereby demonstrating good reliability (Table 3). Pearson's correlation coefficients between the test-retest results of the 5 panelists participating in the study were as follows: 0.728, 0.852, 0.891, 0.937, and 0.962 (all P<0.001). The formula was completed to calculate the ICD-DALY using estimated DWs and variables in an existing injury database.

Table 3.

Test-retest results of 16 common indicator ICD-10 injury codes using person trade-off 1 (median)

ICD-10 code	Test	Retest
S00.5	0.005	0.004
S02.0	0.200	0.167
S15.7	0.375	0.375
S23.3	0.074	0.034
S46.1	0.167	0.200
S65.9	0.167	0.153
S66.3	0.167	0.167
S93.3	0.167	0.153
S94.9	0.167	0.138
S95.7	0.231	0.167
T20.1	0.029	0.015
T45.5	0.200	0.130
T71.0	0.714	0.714
T75.1	0.800	0.833
T92.9	0.310	0.091
T98.3	0.444	0.167

Pearson correlation coefficient=0.932 (P<0.001); Spearman correlation coefficients=0.740 (P<0.001).

ICD-10, International Classification of Diseases 10th edition.

Most codes in the HDIS (n=23,160, collected in 2004, male subjects 61.5%, adults (15 to 64 years old) 77.3% and elderly (≥65 years old) 17.9% were automatically converted into 32 nature of injury categories using the GBD-DALY method, including 271 codes by manual conversion by the investigators (conversion rate 1,166/1,284=91%). However, 118 codes were not reclassified into 32 injury categories due to: extremely rare nature of the injury, non-traumatic injury, post-injury complication, environmental injury, or side effects of medical treatment. The excluded codes are displayed in Appendix 4.

Table 4 displays the GBD-DALY and ICD-DALY results according to 32 categories derived from the HDIS database. The GBD-DALY, GBD-YLD, and GBD-YLL were 138,548, 130,481, and 8,117 years, respectively. The ICD-DALY, ICD-YLD, and ICD-YLL were 262,246, 255,710, and 6,537 years, respectively. The mean YLD/YLL proportions were 16.1 for the GBD-DALY and 39.1 for the ICD-DALY.

Table 4.

Comparison of GBD-DALY and ICD-DALY according to 32 nature of injury categories

Nature of injury category	GBD-YLD (a)	GBD- YLL (b)	(a)/(b)	GBD- DALY	ICD-YLD (c)	ICD- YLL (d)	(c)/(d)	ICD- DALY
Fracture-skull	3,990	603	6.6	4,593	6,530	361	18.1	6,891
Fracture-face bones	11,367	281	40.5	11,647	10,854	63	172.3	10,917
Fracture-vertebral column	5,151	190	27.1	5,342	10,797	108	100.0	10,905
Fracture-rib or sternum	2,955	341	8.7	3,296	6,974	215	32.4	7,189
Fracture-pelvis	1,901	180	10.6	2,081	4,174	112	37.3	4,286
Fracture-clavicle, scapula, or humerus	3,579	199	18.0	3,778	8,149	55	148.2	8,204
Fracture-radius or ulna	3,929	40	98.2	3,970	8,451	17	497.1	8,469
Fracture-hand bones	1,987	10	198.7	1,997	4,873	3	1,624.3	4,876
Fracture-femur	4,577	280	16.3	4,857	6,824	159	42.9	6,983
Fracture-patella, tibia, or fibula	8,158	126	64.7	8,284	13,213	54	244.7	13,267
Fracture-ankle	1,101	0	NA	1,101	2,096	0	NA	2,096
Fracture-foot bones	1,112	0	NA	1,112	4,734	0	NA	4,734
Injured spinal cord	3,658	72	50.8	3,730	2,519	36	70.0	2,555
Dislocations-shoulder, elbow, or hip	302	27	11.2	329	1,123	7	160.4	1,130
Dislocations-other dislocation	4,923	49	100.5	4,972	21,551	22	979.6	21,573
Sprains	18,812	299	62.9	19,111	42,943	30	1,431.4	42,973
Intracranial injuries	24,177	2,342	10.3	26,519	36,465	2,252	16.2	38,716
Internal injuries	3,285	1,259	2.6	4,543	16,176	1,373	11.8	17,549
Open wound	8,808	459	19.2	9,267	13,157	89	147.8	13,246
Injury to eyes	2,558	57	44.9	2,615	5,194	12	432.8	5,206
Amputations-thumb	240	0	NA	240	780	0	NA	780
Amputations-finger	615	0	NA	615	3,021	0	NA	3,021
Amputations-arm	328	0	NA	328	486	0	NA	486
Amputations-toe	10	0	NA	10	46	0	NA	46
Amputations-foot	18	0	NA	18	37	0	NA	37
Amputations-leg	157	0	NA	157	224	0	NA	224
Crushing	1,867	41	45.5	1,908	3,683	43	85.7	3,725
Burns-less than 20%	17	167	0.1	184	3,243	37	87.6	3,279
Burns-20% to 60%	3,887	251	15.5	4,137	5,482	113	48.5	5,594
Burns-greater than 60%	164	183	0.9	348	1,506	615	2.4	2,122
Injured nerves	613	22	27.9	635	3,100	7	442.9	3,107
Poisoning	6,186	641	9.7	6,826	7,301	755	9.7	8,063
Total	130,431	8,117	16.1	138,548	255,710	6,537	39.1	262,246

GBD, Global Burden of Disease Study; DALY, disability-adjusted life year; ICD, International Classification of Diseases 10th edition; YLD, years of life disabled; YLL, years of life lost; NA, not applicable.

Fig. 4 shows the comparison between the ICD-DALY and the GBD-DALY based on 32 nature of injury categories. The injury distribution differed between the GBD-DALY and the ICD-DALY (e.g., intracranial injuries>sprains>fracture-face bones>open wound>fracture-patella, tibia, or fibula>poisoning for GBD-DALY and sprains>intracranial injuries>dislocation-other dislocation> internal injuries>fracture-patella, tibia, or fibula>open wound for ICD-DALY). The ICD-DALY data exceeded the GBD-DALY results in most injury categories. The ICD-DALY results displayed an increased burden compared with the GBD-DALY results by 18-fold, for the burns less than 20% category, and by 6-fold, for the burns greater than 60% category. Meanwhile, the ICD-DALY was lower for the fracture-face bones and spinal cord injury categories compared with the GBD-DALY.

Fig. 4.

Comparison of global burden of disease (GBD)–disability-adjusted life year (DALY) and International Classification of Disease 10th edition (ICD)-DALY by 32 injury nature categories.

We observed a high correlation between the GBD-DALY and ICD-DALY. Spearman’s rank correlation coefficients for YLL, YLD, and DALY were 0.988 (P<0.001), 0.738 (P<0.001), and 0.759 (P<0.001), respectively (Fig. 5A–C). To test the agreement between the two methods, we analyzed the results using the Bland-Altman test (Fig. 5D). Two categories were in disagreement, thereby demonstrating the limits of the agreement range (e.g., dislocation-other dislocation and sprains).

Fig. 5.

Comparison of global burden of disease (GBD)–disability-adjusted life year (DALY) and International Classification of Disease 10th edition (ICD)-DALY. (A) Comparison for years of life disabled (YLD), (B) comparison for years of life lost (YLL), (C) comparison for DALY, and (D) Bland-Altman plot. SD, standard deviation.

DISCUSSION

We developed the ICD-DALY in the current study, which represents the first attempt to describe an injury measurement for specific injury codes regarding the GBD. Most GBD studies have focused on the community-based burden of disease [2,3,15,18]. The GBD-DALY has potential for comparing the burden of disease but not individuals with specific injuries. However, the current study focused on individual data with specific individual injury diagnosis codes. In many countries, injury surveillance and nation-wide injury datasets exist, which include exact diagnosis codes and injury mechanisms. Most injury datasets include subject age, diagnosis, time of injury, and mortality outcome. These variables are included for analyzing risk factors, developing injury prevention strategies, and evaluating the effect of such interventions. Using this tool and these three variables, the ICD-DALY can be calculated for every injured individual.

The current technique described in our study is a cost-saving method designed for calculating the burden of injury. Resource intensive methods, including the Disability Rating Scale and the Glasgow Outcome Scale, assess the individual disability of an injured victim, but these tools require follow-up interviews with the subject [21,22]. Although these clinical tools provide a reliable assessment of patient disability, their limitations primarily involve the amount of effort required to perform the calculations. In contrast, the ICD-DALY provides a risk-based DALY, which requires fewer resources and is applicable across large databases. This tool now allows for these new outcome parameters in injury research, incidence, mortality, and DALY per individual patient. Hospital-based injury data are also useful for calculating disability and are comparable with other conditions [23].

This study used a complex methodology to determine the DW derived from injury diagnostic codes. Injury codes vary in terms of severity, mechanism, and outcome, but an injury group cannot be considered one disease entity. Poisonings display a broad spectrum of severity; therefore, a DW for poisoning must account for specific subcategories. Therefore, the determination of a disability for each individual ICD-10 injury code is ideal, but there are too many injury codes (n=1,284) to be decided via expert-based consensus (panel study). Therefore, indicator injury codes (n=130, 10%) based on the strata by morbidity/mortality (like ARR) were chosen. To interpolate the DW using the value of the indicator ICD-10 codes, it is essential to have evenly distributed DW values for all ICD-10 codes. We successfully selected indicator injury codes and then interpolated the corresponding DWs into the remaining ICD-10 injury codes. Previous panel studies have demonstrated varying methodological results [16,17,24]. In accordance with previous studies, the current study demonstrated that the PTO1 method displays the best correlation among panel groups. We accepted the PTO1 method as the ideal tool and used it to determine the DWs for the 130 codes, which were then interpolated in the next step. The PTO1 and VAS methods demonstrated a high correlation for common standard injury codes. We used the VAS method for interpolation to all other codes, which demonstrated a good correlation performance with the PTO1 method. Similarly, other studies have utilized the VAS method due to this benefit [15,18]. This study tested the reliability of the DW of each code using the test-retest method [25]. We found excellent reliability among the median DWs derived from the PTO1 method.

Validation was performed by comparing the ICD-DALY and GBD-DALY values of 32 nature of injury categories, as a direct comparison for all injury codes was impossible. Both measurements correlated strongly (ρ=0.759, P<0.001), although the ICD-DALY were higher than the GBD-DALY results across most categories. The Bland-Altman test showed an acceptable agreement in most categories, with the exception of two categories. Overall, higher agreements were observed between the ICD-DALY and GBD-DALY results in homogeneous categories with similar anatomic areas, injury depth, and severity. Meanwhile, for the remaining heterogeneous categories with different anatomical injuries, the agreement was poor between ICD-DALY and GBD-DALY (e.g., dislocation-shoulder/elbow/hip, dislocation-other dislocation, sprain, intracranial injury, internal injury, open wound, injury of the eyes, crushing, burns-less than 20%, burns-20% to 60%, burns-greater than 60%, injured nerves and poisoning). We assume that the differences between ICD-DALY and GBD-DALY are more marked in the heterogeneous injury categories because they have an increased variety of ICD injury codes compared with the homogeneous category. The GBD-DALY may have excessively reclassified all injuries into 32 injury categories. For example, neurologic deficits, soft tissue injuries, burns, and poisoning would be overlooked in 32 injury categories despite the wide range of disabilities that these injuries incur. The GBD-YLL and ICD-YLL showed very good agreement but GBD-YLD and ICD-YLD showed poorer agreement (Fig. 5A, B). The YLD can be calculated from death, which is clearly defined, for calculation but YLD might be incorrect due to different DWs of injury categories.

This study has certain limitations. First, the interpolation method may be used to determine the DALY. Only 10% of the indicator injury codes were reviewed by the panel groups, while the other codes were estimated via interpolation. Although we tested the reliability of the DW as an indicator for the ICD code (n=16), the entire ICD-10 code was not tested for reliability. Second, we calculated DALYs using the maximum DW among the multiple injury codes that were diagnosed during the same episode (multiple injuries). Therefore, the DALYs might be underestimated because the remaining injury-associated disability was not included. For patients with multiple injuries, the DALY would be minimally calculated. Third, our validation method in which two DALY results were compared is limited in terms of the injury codes that were not compared due to the difficulty of re-categorizing the injury codes into 32 injury categories. Finally, the ICD-DALY method was not validated with clinical disability parameters or the quality-adjusted life year [26].

To efficiently calculate the burden of injury for individual victims, the ICD-10-based DALY was developed and validated using a nationwide database. We found that this new method was easy and feasible for estimating the disability of each individual injury victim and comparable to the GBD-DALY. The ICD-DALY should be extensively validated to apply it for injury epidemiology and prevention.