• 沒有找到結果。

Assessing the Quality of Tuberculosis-related Underlying Cause of Death Assignment in Taiwan, 2001–2005

N/A
N/A
Protected

Academic year: 2021

Share "Assessing the Quality of Tuberculosis-related Underlying Cause of Death Assignment in Taiwan, 2001–2005"

Copied!
7
0
0

加載中.... (立即查看全文)

全文

(1)

The underlying cause of death (UCOD) is defined by the World Health Organization as “the dis-ease or injury which initiated the train of morbid events leading directly to death” and is the cause of death used for tabulating official cause of death statistics in most countries.1 Ideally, the attend-ing physician of the deceased should determine the UCOD for each death and properly report the causal relationship between cause of death on death certificates.2,3In reality however, many physicians do not properly report the cause of

death, which makes the assignment of UCOD inconsistent among coders.4–7

Cause of death statistics is an important refer-ence for national tuberculosis (TB) control policy planning and evaluation. For example, of 55 sta-tistical tables in the appendix of the Tuberculosis Annual Report 2001 published by the Center for Disease Control (CDC) of Taiwan, 18 were based on mortality data.8Many scholars also use mor-tality data to examine the epidemiologic profiles of TB problems in Taiwan.9–12If the quality of TB

Assessing the Quality of Tuberculosis-related

Underlying Cause of Death Assignment in

Taiwan, 2001–2005

Yi-Chun Wu,1,2Ruey-Shiung Lin,2Shiang-Lin Yang,3Tsung-Hsueh Lu4*

Background/Purpose: Assignment of underlying cause of death (UCOD) might be inconsistent among coders if physicians do not properly record cause of death on death certificates. This study aimed to assess the changes in the quality of tuberculosis-related UCOD assignment in Taiwan after interventions by the Center for Disease Control (CDC).

Methods: The reference (gold standard) we used to assess the quality of UCOD assignment by coders was the UCOD selected by the Automated Classification of Medical Entities (ACME) computer program. The agreement, over- and under-coding rates between coders and the reference were calculated by years before and after the CDC interventions.

Results: An abrupt decrease in tuberculosis death rates according to the UCOD assigned by coders was noted from 2003 to 2004, but no such decrease was noted according to the reference. The agreement in UCOD assignments between coders and ACME decreased from 0.75 in 2001 to 0.67 in 2005. We found a significant decrease in the over-coding rate from 0.21 in 2003 to 0.11 in 2004, and a prominent increase in under-coding rates from 0.08 in 2003 to 0.24 in 2004.

Conclusion: The abrupt decrease in the official published tuberculosis mortality rate from 2003 to 2004 was due to significant changes in the practice of UCOD assignment of official coders, which might have been a response to interventions initiated by the CDC. [J Formos Med Assoc 2008;107(1):30–36]

(2)

mortality data cannot be ensured, then the validity of the conclusions derived is threatened.

The determination of TB-related UCOD has become progressively more difficult because many elderly people have recurrent TB and comorbid diseases.13–15For example, when an elderly per-son has chronic obstructive pulmonary disease and recurrent TB that is under treatment, and then dies from acute myocardial infarction, it is some-times very difficult for the attending physician to determine whether the elderly person had died from TB or with TB. If the physician thought the elderly person had died from TB, then the UCOD would be recorded as TB. In contrast, if the physi-cian thought the elderly person had died with TB, the UCOD would not be recorded as TB. A previ-ous study indicated that half of the 2129 death certificates that mentioned TB was inappropri-ately certified by physicians and that 16% had an inaccurately assigned UCOD by official coders in Taiwan.16

To improve the quality of the TB-related cause of death certification behavior of physicians and TB-related UCOD assignment of official coders, the Taiwan CDC initiated several interventions in 2002. The aim of this study was to assess the changes in the quality of TB-related UCOD assign-ment of official coders before and after the CDC interventions.

Methods

Interventions

Since 2002, the Taiwan CDC has asked the Office of Statistics of the Department of Health to mail all TB-related death certificates to them on a monthly basis for quality assurance before the official cause of death statistics are published. The CDC can thus assess the accuracy of certifica-tion by certifiers, identify the quescertifica-tionable cases and mail them back to the Office of Statistics to consider modifying the UCOD assignment. About 10–15% of death certificates in which TB was originally assigned as the UCOD by official coders were queried by the CDC from 2002

through 2005. The Office of Statistics organized a panel to improve the reliability of UCOD assign-ment between coders, and did not respond to the queried cases until 2004, when the official coders then changed half of the original UCOD assign-ments in the queried cases. The change rate de-creased to 12% in 2005 because the coders had by then changed their habits with regard to assigning TB-related UCOD.

Reference for assessment

The reference (gold standard) we used to assess the accuracy of TB-related UCOD assignment by official coders was the UCOD assigned by the Automated Classification of Medical Entities (ACME) computer program. ACME was devel-oped by the National Center for Health Statistics of the United States to standardize the processes of UCOD assignment. ACME is now used by more than 20 countries and was introduced to Taiwan in 2000.17

To prepare for the implementation of the ICD-10 version for cause of death statistics, the Office of Statistics initiated a double coding sys-tem in 2001. Each death certificate was assigned the ICD-9 UCOD code manually by official coders and the ICD-10 UCOD code by ACME, which provided us with a good opportunity to assess the quality of UCOD assignment of official coders from 2001 through 2005.

Analysis

All the death certificates of people who died be-tween 2001 and 2005 on which TB was assigned as the UCOD either by official coders (ICD-9 codes 010–018) or by ACME (ICD-10 codes A16–A19) were included in this study. We first calculated the age-standardized death rates and age-specific death rates for TB according to the UCOD assigned by official coders and by ACME. The 95% confidence intervals (CI) for death rates were also computed to examine the differences between the official coders and ACME.

Using the UCOD assigned by ACME as the reference (gold standard), we calculated the rates of agreement, over-coding and under-coding by

(3)

year, by sex and by age group for analyses. Because of the extremely large number of deaths in cell d in the 2 by 2 table (more than 130,000 deaths in this study), we did not include the number in cell d in the denominator when cal-culating the agreement rate and over-coding rate as suggested by Gordis18(Table 1).

Results

TB death rates decreased from 2001 to 2005 according to both official coders and ACME. An abrupt decrease in TB death rates according to the UCOD assigned by official coders was noted from 2003 to 2004, but no such change was found according to the UCOD assigned by the refer-ence (Figure). The differrefer-ence in age-standardized TB death rate between coders and ACME was sig-nificant only in 2003 and mainly among the de-ceased aged ≥ 75 years (Table 2). We also found that the differences in TB death rates increased with increasing age of the deceased.

The agreement in TB-related UCOD assign-ment between official coders and reference de-creased from 0.75 in 2001 and 2002 to 0.67 in 2004 and 2005 (Table 3). We found a significant decrease in the rate of over-coding from 2003 (0.21; 95% CI, 0.19–0.23) to 2004 (0.11; 95% CI,

pneumonia, followed by diabetes mellitus and liver cirrhosis. On the other hand, the diagnosis that was most often replaced by TB as the UCOD (over-coding) by official coders was old TB followed by sepsis and chronic obstructive pul-monary disease (Table 4).

Discussion

Using the UCOD assigned by ACME as the refer-ence (gold standard), our findings indicated that the abrupt decrease in the official published TB mortality rate from 2003 to 2004 was due to significant changes in the practices of UCOD as-signment of official coders. In 2003 and before, a relatively high percentage of over-coding of TB as the UCOD by official coders was noted. On the contrary, a relatively high percentage of under-coding of TB as the UCOD by official coders was noted from 2004 onwards. The official published TB mortality rates are the result of net

compensa-Table 1. 2 by 2 table

UCOD assigned UCOD assigned by ACME by official coders Tuberculosis Non-tuberculosis

Tuberculosis a c

Non-tuberculosis b d

Agreement rate = a/(a + b + c) Over-coding rate = c/(a + b + c) Under-coding rate = b/(a + b)

4.0 4.5 5.0 5.5 6.0 2001 2002 2003 2004 2005 Death per 100,000 Official coders Reference

Figure. Estimated tuberculosis death rates according to

the underlying cause of death assigned by official coders and by the ACME computer system (reference) in Taiwan, 2001–2005.

(4)

Most previous studies used only sampled cases and were cross-sectional.19–21In addition, many studies used the UCOD assigned by senior coders as the reference, which still has problems of reli-ability.19–21Another strength of this study was the use of UCOD assigned by ACME as the reference, which was more objective and robust.16,17

One of the limitations of this study was that we did not interview the official coders to deter-mine their logic behind changing their coding practices from 2003 to 2004. The official coders might have had other information provided on the death certificate that was not accessed by ACME. In addition, the information input to ACME might

Table 2. Age-standardized death rate (ASDR) and age-specific death rate and 95% confidence intervals (CI) for tuberculosis

according to the underlying cause of death assigned by official coders or by the ACME computer system (reference) in Taiwan, 2001–2005 2001 2002 2003 2004 2005 ASDR Coders (95% CI) 5.80 (5.48–6.11) 5.67 (5.36–5.98) 5.79 (5.48–6.10) 4.22 (3.95–4.49) 4.26 (3.99–4.53) Reference (95% CI) 5.75 (5.43–6.06) 5.12 (4.83–5.42) 4.86 (4.57–5.15) 4.73 (4.45–5.02) 4.72 (4.43–5.00) ≤ 44 Coders (95% CI) 0.52 (0.40–0.63) 0.36 (0.27–0.45) 0.47 (0.37–0.58) 0.32 (0.23–0.41) 0.26 (0.17–0.34) Reference (95% CI) 0.60 (0.48–0.72) 0.34 (0.25–0.43) 0.42 (0.31–0.52) 0.34 (0.25–0.43) 0.35 (0.26–0.45) 45–64 Coders (95% CI) 4.71 (4.08–5.35) 4.16 (3.57–4.74) 4.18 (3.61–4.76) 3.00 (2.52–3.47) 2.72 (2.27–3.16) Reference (95% CI) 4.98 (4.33–5.63) 3.99 (3.41–4.56) 3.65 (3.11–4.18) 3.39 (2.88–3.90) 3.10 (2.62–3.57) 65–74 Coders (95% CI) 25.01 24.14 22.96 15.75 13.44 (22.24–27.79) (21.43–26.86) (20.33–25.60) (13.58–17.92) (11.45–15.44) Reference (95% CI) 23.97 22.55 18.62 17.31 16.29 (21.25–26.69) (19.93–25.18) (16.25–21.00) (15.03–19.59) (14.10–18.48) ≥ 75 Coders (95% CI) 95.32 93.38 90.18 63.83 66.98 (88.22–102.42) (86.57–100.19) (83.69–96.68) (58.52–69.14) (61.68–72.28) Reference (95% CI) 92.15 81.46 75.56 72.24 70.48 (85.17–99.13) (75.10–87.83) (69.61–81.50) (66.59–77.89) (65.04–75.91)

Table 3. Number of deaths where tuberculosis (TB) was assigned as the underlying cause of death (UCOD) either by

official coders or by the ACME computer system (reference) and indicators of quality of UCOD assignment in Taiwan, 2001–2005

Coders TB Non-TB TB

Agreement (95% CI) Over-coding (95% CI) Under-coding (95% CI)

Reference TB TB Non-TB

a b c a/(a+ b + c) c/(a+ b + c) b/(a+ b)

Total 2001 1109 179 190 0.75 (0.73–0.77) 0.13 (0.11–0.15) 0.14 (0.12–0.16) 2002 1038 116 239 0.75 (0.72–0.77) 0.17 (0.15–0.19) 0.10 (0.08–0.12) 2003 1012 87 297 0.72 (0.70–0.75) 0.21 (0.19–0.23) 0.08 (0.06–0.10) 2004 818 256 139 0.67 (0.65–0.70) 0.11 (0.10–0.13) 0.24 (0.21–0.26) 2005 823 251 147 0.67 (0.65–0.70) 0.12 (0.10–0.14) 0.23 (0.21–0.26)

(5)

have sustained typos and coding errors, as found in a previous study.16

Why over-coding?

Many “old” TB cases were over-coded by official coders and TB assigned as the UCOD. As indi-cated in Table 3, there were 168 deaths assigned as old TB by ACME, which were coded as “active” TB by official coders. The reason for the discrep-ancy was that some of the official coders did not take into account the length of time between di-agnosis and death that was reported by physi-cians. All deaths with duration of time between diagnosis and death that was longer than 1 year were classified by ACME as old TB.

The second reason for over-coding of TB as the UCOD by official coders was the reluctance in accepting mechanisms of death (i.e. sepsis, heart failure, respiratory failure, arrhythmia, and many ill-defined symptoms and signs) as the UCOD. Because the mechanisms of death were terminal events of the deaths which could not provide useful information for disease prevention, guide-lines suggest that physicians should not report

World Health Organization.1 Variations in the interpretation of Selection Rule 3 comprised the main reason for the discrepancy in UCOD assignments across countries.19–21

Why under-coding?

The main reason for under-coding of TB as the UCOD by official coders since 2004 might be the reactionary responses to the aggressive queries from the CDC. The CDC checked only those death certificates “with” tuberculosis assigned as the UCOD by official coders and did not check death certificates of notified TB cases “without” reporting TB as the UCOD. The biased selection for assessment would certainly result in under-coding of TB by official coders.

The second reason for under-coding of TB as the UCOD by official coders since 2004 was the changes in the International UCOD Selection Rule 3 in ICD-10 which was used by ACME. The most affected diagnosis was pneumonia as indicated in Table 2 and in other countries.22–24For many cases with pneumonia reported in Part I of the death certificate and TB in Part II of the death certificate,

Table 4. Diagnoses that most often replaced tuberculosis (under-coding) or that were replaced by tuberculosis (over-coding)

as the underlying cause of death by official coders in Taiwan, 2001–2005

Diagnosis that replaced tuberculosis n (%) Diagnosis replaced by tuberculosis n (%)

Pneumonia 105 (11.8) Old tuberculosis 168 (16.6)

Diabetes mellitus 93 (10.5) Sepsis 122 (12.1)

Liver cirrhosis 73 (8.2) Chronic obstructive pulmonary disease 103 (10.2)

Chronic obstructive pulmonary disease 42 (4.7) Acute myocardial infarction 49 (4.8)

Old stroke 31 (3.5) Cerebral infarction 42 (4.2)

Lung cancer 30 (3.4) Heart failure 35 (3.5)

Stroke 29 (3.3) Diabetes mellitus 31 (3.1)

Chronic renal failure 23 (2.6) Urinary tract infection 25 (2.5)

Bronchiectasis 22 (2.5) Stroke 24 (2.4)

Others 441 (49.6) Others 413 (40.8)

(6)

evaluating healthcare and disease control pro-grams, the quality of mortality data should be ensured. Our findings indicate that the interpre-tation of changes in TB mortality from 2003 to 2004 in Taiwan should be cautious. The TB death rates calculated according to the UCOD assigned by ACME would be a better estimation of real epidemiological profiles. The Office of Statistics has scheduled to use ICD-10 and UCOD accord-ing to ACME for the official published cause of death statistics in 2008. We believe that the qual-ity of UCOD assignment will be improved in the near future.

With regard to the reactionary responses of official coders from over-coding bias to under-coding bias in face of the CDC interventions, the Office of Statistics should provide more system-atic education and training programs for better assignment of UCOD. As indicated by a previous study, the logic in ACME could be used as a teaching aid to train coders on how to correctly assign UCOD.17

If the CDC plans to continually assess the qual-ity of TB-related UCOD assignment, we suggest that the death certificates of notified TB cases on which official coders did not assign TB as the UCOD should also be sampled to estimate the extent of under-coding bias.

To re-stress then, if physicians can report an appropriate cause of death, i.e. correct causal re-lationships between diseases on death certificates, then UCOD assignment can be very straightfor-ward and simple. Both the CDC and the Office of Statistics should educate and urge physicians to correctly report the cause of death on death certificates, which is the most important deter-minant of the quality of cause of death statistics. In conclusion, the abrupt decrease in official published TB mortality rates from 2003 to 2004 was due to significant changes in the practices of UCOD assignment of official coders, which might have been their response to the interventions ini-tiated by the CDC. The over-coding and under-coding of TB death remains a problem in Taiwan. Using ACME to improve the quality of UCOD assignment is warranted.

Acknowledgments

This study was supported by a grant (DOH92-DC-113) from the Center for Disease Control, Taiwan.

References

1. World Health Organization. International Statistical

Classi-fication of Diseases and Related Health Problems, Tenth Revision. Geneva: World Health Organization, 1992.

2. Kircher T, Anderson RE. Cause of death: proper completion of the death certificate. JAMA 1987;258:349–52. 3. National Center for Health Statistics. Physicians’ Handbook

on Medical Certification of Death. Hyattsville, MD: US

Department of Health and Human Services, 2003. 4. Maudsley G, Williams EMI. ‘Inaccuracy’ in death

certification—where are we now? J Public Health Med 1996;18:59–66.

5. Messite J, Stellman SD. Accuracy of death certificate com-pletion: the need for formalized physician training. JAMA 1996;275:794–6.

6. Smith Sehdev AE, Hutchins GM. Problems with proper completion and accuracy of the cause-of-death statement.

Arch Intern Med 2001;161:277–84.

7. Lu TH, Shou WY, Shih TP, et al. Factors associated with errors in death certificate completion: a national study in Taiwan. J Clin Epidemiol 2001;54:232–8.

8. Center for Disease Control. Tuberculosis Annual Report

2001. Taipei: Center for Disease Control, Department of

Health, 2002.

9. Lee LT, Chen CJ, Tsai SF, et al. Morbidity and mortality trends of pulmonary tuberculosis in Taiwan. J Formos

Med Assoc 1992;91:867–72.

10. Lee LT, Chen CJ, Lee WC, et al. Age-period-cohort analysis of pulmonary tuberculosis mortality in Taiwan. J Formos

Med Assoc 1994;93:657–62.

11. Yu MC, Bai KJ, Chang JH, et al. Tuberculosis incidence and mortality in aboriginal areas of Taiwan, 1997–2001.

J Formos Med Assoc 2004;103:817–23.

12. Lu TH, Huang RM, Chang TD, et al. Tuberculosis mortality trends in Taiwan: a resurgence of non-respiratory tubercu-losis. Int J Tuberc Lung Dis 2005;9:105–10.

13. Anonymous. Tuberculosis mortality under scrutiny. Lancet 1971;i:1167–8.

14. Heng BH, Tan KK, Chan KW, et al. An evaluation of 1987 tuberculosis deaths in Singapore. Singapore Med J 1990; 31:418–21.

15. McKeown PJ. Tuberculosis mortality—deaths with, rather than from tuberculosis? Irish Med J 1997;90:17.

16. Lu TH, Tsau SM, Wu TC. ACME (Automated Classification of Medical Entities) system could help objectively assess

(7)

the appropriateness of underlying cause of death certifica-tion and assignment. J Clin Epidemiol 2005;58:1277–81. 17. Lu TH. Using ACME (Automatic Classification of Medical

Entry) software to monitor and improve the quality of cause of death statistics. J Epidemiol Community Health 2003;57:470–1.

18. Gordis L. Chapter 5: Assessing the validity and reliability of diagnostic and screening tests. In: Gordis L, ed.

Epi-demiology, 3rd edition. Philadelphia: Elsevier Saunders, 2004:89.

19. Kelson M, Farebrother M. The effect of inaccuracies in death certification and coding practices in the European Economic Community (EEC) on international cancer mortality statistics. Int J Epidemiol 1987;16:411–4. 20. Kelson MC, Heller R. The effect of death certification and

coding practices on observed differences in respiratory

disease mortality in 8 EEC countries. Rev Epidemiol Sante

Publique 1983;31:423–32.

21. Lu TH, Lee MC, Chou MC. Accuracy of cause of death coding in Taiwan: types of miscoding and effects on mortality statistics. Int J Epidemiol 2000;29:336–43. 22. Anderson RN, Rosenberg HM. Disease classification:

measuring the effect of the Tenth Revision of the Interna-tional Classification of Diseases on cause-of-death data in the United States. Stat Med 2003;22:1551–70.

23. Rooney C, Griffiths C, Cook L. The implementation of ICD-10 for cause of death coding—some preliminary results from the bridge coding study. Health Stat Q 2002; 13:31–41.

24. Office for National Statistics. Report: Results of the ICD-10 bridge coding study, England and Wales 1999.

數據

Table 3. Number of deaths where tuberculosis (TB) was assigned as the underlying cause of death (UCOD) either by  official coders or by the ACME computer system (reference) and indicators of quality of UCOD assignment in  Taiwan, 2001–2005

參考文獻

相關文件

• A put gives its holder the right to sell a number of the underlying asset for the strike price.. • An embedded option has to be traded along with the

6 《中論·觀因緣品》,《佛藏要籍選刊》第 9 冊,上海古籍出版社 1994 年版,第 1

The first row shows the eyespot with white inner ring, black middle ring, and yellow outer ring in Bicyclus anynana.. The second row provides the eyespot with black inner ring

You are given the wavelength and total energy of a light pulse and asked to find the number of photons it

After students have had ample practice with developing characters, describing a setting and writing realistic dialogue, they will need to go back to the Short Story Writing Task

A statistically significant decrease was noted in the percentages of P6 students reported using digital resources assigned by teachers (from 60% to 54%) beyond school hours and

Robinson Crusoe is an Englishman from the 1) t_______ of York in the seventeenth century, the youngest son of a merchant of German origin. This trip is financially successful,

fostering independent application of reading strategies Strategy 7: Provide opportunities for students to track, reflect on, and share their learning progress (destination). •