State Differences in the Reporting of Diabetes-Related Incorrect Cause-of-Death Causal Sequences on Death Certificates

Abstract

OBJECTIVE

To examine state differences in the reporting of diabetes-related incorrect cause-of-death (COD) causal sequences on death certificates in the U.S.

RESEARCH DESIGN AND METHODS

We conducted a cross-sectional descriptive study to determine the prevalence of two types of incorrect COD causal sequences with data from the Multiple Cause Mortality File of the year 2004.

RESULTS

Among deaths in which diabetes was reported as the first diagnosis on line a, b, c, or d in Part I of the death certificate in the U.S., 21% had below diabetes placement error (ranged from 30% in Maryland to 7% in Hawaii) and 11% had above diabetes placement error (ranged from 18% in Kentucky to 5% in California). The net effects of the two types of error ranged from −0.7% in Nevada to 19.6% in the District of Columbia.

CONCLUSIONS

Because the rates of incorrect reporting of diabetes-related COD causal sequence varied across states, the comparability of the diabetes death rate between states may have been compromised.

Mortality from diabetes is one of the important indicators in the state diabetes surveillance system initiated by the Centers for Disease Control and Prevention, Division of Diabetes Translation (1). If, however, certifying physicians in different states were found to have different practices in reporting cause-of-death (COD) causal sequences in Part I of the death certificate, the comparability of diabetes death rates across states would be in doubt (2). A recent study indicated an increase in the reporting of diabetes-related incorrect COD causal sequences in the U.S. (3). Little is known, however, regarding the prevalence of this COD certification problem at the state level. We aimed in this study to examine state differences in the reporting of diabetes-related incorrect COD causal sequences and to evaluate the possible effects on the reported diabetes death rate at the state level in the U.S.

RESEARCH DESIGN AND METHODS

We extracted all diabetes-related deaths occurring in the year 2004 from the Multiple-Cause Mortality File compiled by the National Center for Health Statistics of the National Centers for Disease Control and Prevention (4). All cases in which diabetes was mentioned anywhere on the death certificate were included for analysis. ICD-10 codes E10–E14 were used to identify diabetes-related death (5).

We determined two types of errors in the reporting of an incorrect COD causal sequence, the “below diabetes error” and the “above diabetes error,” according to the algorithm developed by Lu et al. (3). A below diabetes error occurred when diagnoses were incorrectly reported as the cause of diabetes (on the line below) in Part I of the death certificate. For example, see the following report:

1. Acute myocardial infarction

2. Diabetes mellitus

3. Hypertension

In this example, hypertension was incorrectly reported (on the line below) as a cause of diabetes mellitus. In contrast, an above diabetes error occurred when diagnoses were incorrectly reported as a consequence of diabetes (on the line above). In the following example, lung cancer was incorrectly reported as a consequence of diabetes mellitus (on the line above):

1. Respiratory failure

2. Lung cancer

3. Diabetes mellitus

With regard to the analysis, we first calculated the proportions of the two types of errors in reporting of incorrect COD causal sequences among deaths in which diabetes was reported as the first diagnosis on line a, b, c, or d in Part I of the death certificate in each state. We did not include cases in which diabetes was reported in Part II of the death certificate, because diabetes was very unlikely to be selected as the underlying COD in these cases according to the International Selection Rules set by ICD-10. All percentages were age adjusted according to the age structure of the U.S. as a whole. To estimate the possible effects of incorrect reporting on the state diabetes death rates, we calculated the net effect of the two types of error for each state.

RESULTS

Among deaths in which diabetes was reported as the first diagnosis on either line a, b, c, or d in Part I of the death certificate in the U.S., 32% had an incorrect COD causal sequence reported. There were 13 states with a percentage of improper COD statement significantly higher than the U.S. average (32%) and 12 states with a percentage lower than the U.S. average (Table 1).

Table 1.

Age-adjusted diabetes death rate (deaths per 100,000 population) and numbers and percentages of reporting of incorrect COD causal sequences on the death certificate by medical certifiers in each state in the U.S., 2004

For below diabetes error, the percentage ranged from 30% in Maryland to 7% in Hawaii. For above diabetes error, the percentage ranged from 18% in Kentucky to 5% in California. For the net effect of the two types of error, the percentage ranged from −0.7% in Nevada and 0.6% in Hawaii to 19.6% (24.2% − 4.6%) in the District of Columbia. We found a significant positive correlation (r = 0.53, P < 0.001) between the net effect of the two types of error in each state and the state diabetes death rate.

CONCLUSIONS

The findings of this study indicate a more than threefold interstate difference in the reporting of diabetes-related incorrect COD causal sequences across states. In most states, the proportion of below diabetes error was larger than the proportion of above diabetes error. The net effect of these two types of incorrect reporting therefore resulted in overreporting of the state diabetes death rate.

There are three possible explanations for large variations in rates of incorrect reporting across states. The first is that states with a higher percentage of specialists (cardiologists, endocrinologists, and nephrologists) would have a greater likelihood of valid COD assignment, as suggested by Murray et al. (2) and Lu et al. (6). The second is that the states with more aggressive practices in querying certifiers to clarify the incorrect COD causal sequence statements would have a lower percentage of improper COD statements (7). The third is that the states with more diabetic patients with cardiovascular complications or comorbidities would have a higher percentage of improper COD statements, because a previous study has indicated higher error rates among decedents with both diabetes and cardiovascular diseases (3).

In conclusion, the rates of incorrect reporting of diabetes-related COD causal sequence vary across states. Thus the comparability of the diabetes death rate between states is in question. Efforts (such as education, training, and querying certifying physicians for improper COD statements) are needed to improve the comparability. Further studies are needed to verify the possible explanations of large variation in rates of improper COD statements across states and to compare multiple-year error rates, especially in some states with high error rates.