Are Primary Care Physicians More Likely to Record Diabetes on Death Certificates?

The Translating Research Into Action for Diabetes (TRIAD) study has been described in detail elsewhere (10). This study was limited to the Michigan study site. Patients aged ≥18 years were sampled (n = 1,650). The institutional review board approved the study, and all participants provided informed consent.

A baseline survey was administered to TRIAD participants either by computer-assisted telephone interview or by mail, and medical records were reviewed. TRIAD decedents were identified from phone calls from friends and relatives of the decedent, returned mailings, and internet searches using the Social Security Index Web site (11). Deaths were verified by matching first and last name, social security number, date of birth, and sex of the decedent with death certificates obtained from the State of Michigan Vital Statistics Office.

Vital status was determined for participants as of February 2007. One dichotomous dependent variable, the presence of the word “diabetes” in either part I or part II of the death certificate, was investigated. The primary independent variable was whether or not the certifying physician was the decedent's PCP. This was defined by matching the first and last name of the certifying physician with the first and last name of the PCP as recorded by the health plan. Covariates obtained from either the survey or medical record review included age, sex, race, education, income, age at diabetes diagnosis, duration of diabetes, type of diabetes, diabetes treatment, smoking status, and Charlson comorbidity index (12).

Bivariate analyses were performed using t tests for continuous variables and χ² tests for categorical variables. We constructed a multivariate model with all covariates and estimated adjusted conditional predicted probabilities for each variable. All analyses were performed using SAS version 9.1.3 (SAS Institute, Cary, NC).

Of 327 decedents, 123 (38%) had diabetes listed in either part I or part II of the death certificate. Table 1 shows the distribution of the characteristic of the entire population of decedents and those with diabetes listed and not listed on the death certificate. If the PCP was the certifying physician, diabetes was recorded 70% of the time versus 29% of the time if the certifying physician was not the PCP (P < 0.0001). Compared with those without diabetes recorded, those with diabetes recorded on the death certificate were younger (P = 0.06), were more likely to be male (P = 0.14), and had a longer duration of diabetes (P = 0.06). If a patient died as an inpatient (P = 0.14) or had an autopsy done (P = 0.70), he/she was less likely to have diabetes recorded on the death certificate.

In multivariable analyses, recording of diabetes was most strongly associated with the PCP as the certifying physician (70 [95% CI 68–72] vs. 30% [26–30], P < 0.0001). The conditional predicted probabilities of recording diabetes anywhere on the death certificate were younger age at death (aged 69 years at death 40% [37–44] vs. aged 79 years at death 31% [28–34], P = 0.07) and longer duration of diabetes (<5 years 21% [19–23] {referent}, 5–14 years 47% [44–50] {P = 0.01}, 15 years 37% [34–40] {P = 0.64}). No other covariates in the multivariable model were significant.

In subanalyses looking only at the characteristics of those decedents for whom the certifying physician was not the PCP, diabetes was not more likely to be listed on the death certificate if the patient died as an inpatient, died on a weekday between 8 A.M.–5 P.M., or had an autopsy.

Our results are consistent with previous studies showing that only 38–48% of decedents with diabetes have diabetes recorded anywhere on the death certificate (1–9). We found a very strong association between recording of diabetes on the death certificate and the certifying physician being the PCP after adjusting for all covariates. Our results confirm that younger age and longer duration of diabetes are associated with recording of diabetes on death certificates (1–3,13). Decedents having a postmortem examination and dying as an inpatient were less likely to have diabetes recorded on the death certificate, consistent with previous reports (6,14,15).

Our results extend the observation of Andresen et al. (2) to a larger, more representative population with diabetes and demonstrate a stronger association between PCPs as certifying physicians and the recording of diabetes on death certificates. Attribution of death to diabetes as the underlying cause, a contributing cause, or a significant condition contributing to death is often a subjective judgment on the part of the certifying physician. We hypothesize that non-PCP certifying physicians may not be as familiar with the decedent's medical history and that lack of awareness of diabetes as an underlying condition rather than lack of perception of diabetes as a step in the causal pathway leading to death may account for underreporting.

This study was jointly funded by the Centers for Disease Control and Prevention (Division of Diabetes Translation) and the National Institute of Diabetes and Digestive and Kidney Diseases.