Review of the performance of methods to identify diabetes cases among vital statistics, administrative, and survey data

Reviewing performance of birth certificate and hospital discharge data to identify births complicated by maternal diabetes

OBJECTIVES

Public health surveillance of diabetes during pregnancy is needed. Birth certificate and hospital discharge data are population-based, routinely available and economical to obtain and analyze, but their quality has been criticized. It is important to understand the usefulness and limitations of these data sources for surveillance of diabetes during pregnancy.

METHODS

We conducted a comprehensive literature review to summarize the validity of birth certificate and hospital discharge data for identifying diabetes-complicated births.

RESULTS

Sensitivities for birth certificate data identifying prepregnancy diabetes mellitus (PDM) ranged from 47% to 52%, median 50% (kappas: min = 0.210, med = 0.497, max = 0.523). Sensitivities for birth certificate data identifying gestational diabetes mellitus (GDM) ranged from 46% to 83%, median 65% (kappas: min = 0.545, med = 0.667, max = 0.828). Sensitivities for the two studies using hospital discharge data for identifying PDM were 78% and 95% (kappas: 0.839 and 0.964), and for GDM were 71% and 81% (kappas: 0.584 and 0.840). Specificities were consistently above 98% for both data sources.

CONCLUSIONS

Overall, hospital discharge data performed better than birth certificates, marginally so for identifying GDM but substantially so for identifying PDM. Reports based on either source alone should focus on trends and disparities and include the caveat that results under represent the problem. Linking the two data sources may improve identification of both GDM and PDM cases.

Mitigating case mix factors by choice of glycemic control performance measure threshold

OBJECTIVE

Performance measures are tools for assessing quality of care but may be influenced by patient factors. We investigated how currently endorsed performance measures for glycemic control in diabetes may be influenced by case mix composition. We assessed differences in A1C performance measure threshold attainment by case mix factors for A1C >9% and examined how lowering the threshold to A1C >8% or >7% changed these differences.

RESEARCH DESIGN AND METHODS

Using data from the 1999-2002 National Health and Nutrition Examination Survey for 843 adults self-reporting diabetes, we computed the mean difference in A1C threshold attainment of >9, >8, and >7% by various case mix factors. The mean difference is the average percentage point difference in threshold attainment for population groups compared with that for the overall population.

RESULTS

Diabetes medication was the only factor for which the difference in threshold attainment increased at lower thresholds, with mean differences of 5.7 percentage points at A1C >9% (reference), 10.1 percentage points at A1C >8% (P < 0.05), and 14.1 percentage points at A1C >7% (P < 0.001).

CONCLUSIONS

As 87% of U.S. adults have A1C <9%, a performance measure threshold of >9% will not drive major improvements in glycemic control. Lower thresholds do not exacerbate differences in threshold attainment for most factors. Reporting by diabetes medication use may compensate for heterogeneous case mix when a performance measure threshold of A1C >8% or lower is used.

Diabetes mellitus and birth defects

OBJECTIVE

The purpose of this study was to examine associations between diabetes mellitus and 39 birth defects.

STUDY DESIGN

This was a multicenter case-control study of mothers of infants who were born with (n = 13,030) and without (n = 4895) birth defects in the National Birth Defects Prevention Study (1997-2003).

RESULTS

Pregestational diabetes mellitus (PGDM) was associated significantly with noncardiac defects (isolated, 7/23 defects; multiples, 13/23 defects) and cardiac defects (isolated, 11/16 defects; multiples, 8/16 defects). Adjusted odds ratios for PGDM and all isolated and multiple defects were 3.17 (95% CI, 2.20-4.99) and 8.62 (95% CI, 5.27-14.10), respectively. Gestational diabetes mellitus (GDM) was associated with fewer noncardiac defects (isolated, 3/23 defects; multiples, 3/23 defects) and cardiac defects (isolated, 3/16 defects; multiples, 2/16 defects). Odds ratios between GDM and all isolated and multiple defects were 1.42 (95% CI, 1.17-1.73) and 1.50 (95% CI, 1.13-2.00), respectively. These associations were limited generally to offspring of women with prepregnancy body mass index > or =25 kg/m(2).

CONCLUSION

PGDM was associated with a wide range of birth defects; GDM was associated with a limited group of birth defects.

Using multiple data features improved the validity of osteoporosis case ascertainment from administrative databases

OBJECTIVES

The aim was to construct and validate algorithms for osteoporosis case ascertainment from administrative databases and to estimate the population prevalence of osteoporosis for these algorithms.

STUDY DESIGN AND SETTING

Artificial neural networks, classification trees, and logistic regression were applied to hospital, physician, and pharmacy data from Manitoba, Canada. Discriminative performance and calibration (i.e., error) were compared for algorithms defined from different sets of diagnosis, prescription drug, comorbidity, and demographic variables. Algorithms were validated against a regional bone mineral density testing program.

RESULTS

Discriminative performance and calibration were poorer and sensitivity was generally lower for algorithms based on diagnosis codes alone than for algorithms based on an expanded set of data features that included osteoporosis prescriptions and age. Validation measures were similar for neural networks and classification trees, but prevalence estimates were lower for the former model.

CONCLUSION

Multiple features of administrative data generally resulted in improved sensitivity of osteoporosis case-detection algorithm without loss of specificity. However, prevalence estimates using an expanded set of features were still slightly lower than estimates from a population-based study with primary data collection. The classification methods developed in this study can be extended to other chronic diseases for which there may be multiple markers in administrative data.

Estimating disease prevalence using a population-based administrative healthcare database

AIMS

In Ostergötland County, Sweden, all data on hospital care and primary healthcare (PHC) have been entered in a diagnosis-related administrative database since 1999. This database was used to estimate the prevalence of four chronic diseases and to examine the capture of data in PHC, outpatient hospital care, and inpatient hospital care, considered in different time frames.

METHODS

A case-finding algorithm identified patients with at least one healthcare contact involving a diagnosis of diabetes, hypertension, asthma, or chronic obstructive pulmonary disease (COPD) in 1999-2003. Prevalence rates were calculated as the ratio of the number of identified patients alive to the total number of inhabitants on 31 December 2003 (n approximately 415,000).

RESULTS

Prevalence rates were 4.4% for diabetes, 10.3% for hypertension, 4.5% for asthma, and 1.2% for COPD. For all four diagnoses, the proportions of patients identified on only one healthcare level were greatest for PHC, reaching rates of 23%, 68%, 53%, and 48%, respectively. The cases identified solely in PHC comprised larger proportions of women and patients over the age of 65 years. Considering the proportion of patients identified in 2003 in relation to the total five-year period gave values of 71%, 50%, 38%, and 58%, respectively, for the four diagnoses.

CONCLUSIONS

The administrative healthcare databases in Sweden today can be important tools in epidemiological research. However, data on several consecutive years and both PHC and hospital data are needed to achieve valid prevalence estimates.

Predictive Values of Self-Reported Periodontal Need: National Health and Nutrition Examination Survey III

BACKGROUND

This study assessed predictive values of self-reported periodontal need to identify periodontal conditions using clinical examinations as the gold standard.

METHODS

We identified 12,370 adults > or = 18 years of age in the third National Health and Nutrition Examination Survey. Self-reported periodontal need was based on participants responding that gum treatment and/or cleaning was needed when asked: "What type of dental care do you need now?" Two periodontal conditions were at least two sites with pockets (pockets > or = 3 mm or pockets > or = 4 mm) and at least two sites with calculus. Main outcomes were: 1) positive predictive value (PPV(Clean)): proportion of those who self-reported the need for cleaning who had calculus; and PPV(Gum): proportion who self-reported the need for gum treatment who had pockets; 2) negative predictive value (NPV(Clean)): proportion of those who self-reported no need for cleaning who did not have calculus; and NPV(Gum): proportion who self-reported no need for gum treatment who did not have pockets; 3) association between predictive values and sociodemographic and behavioral characteristics; and 4) proportion of individuals with specific sociodemographic and behavioral characteristics whose self-reported periodontal need predicted periodontal conditions.

RESULTS

The prevalence of periodontal conditions influenced predictive values. Calculus prevalence = 85%: corresponding PPV(Clean) = 88% and NPV(Clean) = 16%. Prevalence of pockets > or = 3 mm = 47%: corresponding PPV(Gum) = 62% and NPV(Gum) = 54%. Prevalence of pockets > or = 4 mm = 11%: corresponding PPV(Gum) = 25% and NPV(Gum) = 90%. Ninety percent of 30- to 44-year-old minority female smokers who did not visit the dentist in the past year and reported the need for gum treatment had pockets > or = 3 mm (PPV(Gum) = 90%).

CONCLUSIONS

Self-reported periodontal need (cleaning/gum treatment) predicted the presence of the prevalent conditions (calculus/pockets > or = 3 mm). Not reporting a need for periodontal treatment predicted the absence of the less common condition (pockets > or = 4 mm) but not the more prevalent condition (calculus).

Back to the future? Diabetes, HIV, and the boundaries of public health

The control of infectious diseases has traditionally fallen to public health and the clinical care of chronic diseases to private medicine. In New York City, however, the Department of Health and Mental Hygiene (DOHMH) has recently sought to expand its responsibilities in the oversight and management of chronic-disease care. In December 2005, in an effort to control epidemic rates of diabetes, the DOHMH began implementing a bold new plan for increased disease surveillance through electronic, laboratory-based reporting of A1C test results (a robust measure of blood-sugar levels). The controversy A1C reporting produced was relatively contained, but when Dr. Thomas Frieden, New York City health commissioner, called for the state to begin tracking viral loads and drug resistance among patients with HIV, both the medical community and a wider public took notice and have started to grapple with the meaning of expanded surveillance. In the context of the past century of medical surveillance in America, we analyze the current debates, focusing first on diabetes and then HIV. We identify the points of contention that arise from the city's proposed blend of public health surveillance, disease management, and quality improvement and suggest an approach to balancing the measures' perils and promises.

Relationship of physician volume with process measures and outcomes in diabetes

OBJECTIVE

The volume of patients cared for by an individual physician (physician volume) has been linked to improved outcomes for a number of conditions. It is not known whether a similar association exists for treatment of diabetes. In this study we aimed to determine whether physician volume is associated with improved process measures and outcomes in diabetes care.

RESEARCH DESIGN AND METHODS

This retrospective cohort study analyzed electronic medical records data for 7,120 patients with diabetes treated by 368 primary care physicians at practices affiliated with two large academic hospitals. The associations between physician volume of diabetic patients (diabetes volume) and annual A1C and LDL testing, as well as blood pressure, A1C, and LDL levels, were evaluated.

RESULTS

In multivariable analysis, absolute diabetes volume was linked to decreased odds of A1C testing (4% less for each additional patient seen; P = 0.05), and relative diabetes volume (fraction of the total patients seen who had diabetes) was associated with decreased odds of both A1C (25% less for every 10% increase in the number of diabetic patients seen annually; P = 0.03) and LDL testing (20% less for every 10% increase in the number of diabetic patients; P < 0.001). Physician volume was not significantly associated with the odds of blood pressure, A1C, or LDL control at the end of the study.

CONCLUSIONS

Higher physician volume in care of diabetic patients is associated with decreased adherence to surveillance guidelines and no measurable difference in treatment outcomes.

Effect of BMI on lifetime risk for diabetes in the U.S

OBJECTIVE

At birth, the lifetime risk of developing diabetes is one in three, but lifetime risks across BMI categories are unknown. We estimated BMI-specific lifetime diabetes risk in the U.S. for age-, sex-, and ethnicity-specific subgroups.

RESEARCH DESIGN AND METHODS

National Health Interview Survey data (n = 780,694, 1997-2004) were used to estimate age-, race-, sex-, and BMI-specific prevalence and incidence of diabetes in 2004. U.S. Census Bureau age-, race-, and sex-specific population and mortality rate estimates for 2004 were combined with two previous studies of mortality to estimate diabetes- and BMI-specific mortality rates. These estimates were used in a Markov model to project lifetime risk of diagnosed diabetes by baseline age, race, sex, and BMI.

RESULTS

Lifetime diabetes risk at 18 years of age increased from 7.6 to 70.3% between underweight and very obese men and from 12.2 to 74.4% for women. The lifetime risk difference was lower at older ages. At 65 years of age, compared with normal-weight male subjects, lifetime risk differences (percent) increased from 3.7 to 23.9 percentage points between overweight and very obese men and from 8.7 to 26.7 percentage points for women. The impact of BMI on diabetes duration also decreased with age.

CONCLUSIONS

Overweight and especially obesity, particularly at younger ages, substantially increases lifetime risk of diagnosed diabetes, while their impact on diabetes risk, life expectancy, and diabetes duration diminishes with age.

Monitoring the occurrence of diabetes mellitus and its major complications: the combined use of different administrative databases

OBJECTIVE

Diabetes mellitus is a growing public health problem, for which efficient and timely surveillance is a key policy. Administrative databases offer relevant opportunities for this purpose. We aim to monitor the incidence of diabetes and its major complications using administrative data.

STUDY DESIGN AND METHODS

We study a population of about 850,000 inhabitants in the Veneto Region (Italy) from the end of year 2001 to the end of year 2004. We use four administrative databases with record linkage. Databases of drug prescriptions and of exemptions from medical charge were linked to identify diabetic subjects; hospital discharge records and mortality data were used for the assessment of macrovascular and renal complications and vital status.

RESULTS

We identified 30,230 and 34,620 diabetic subjects at the start and at the end of the study respectively. The row prevalence increased from 38.3/1000 (95% CI 37.2-39.5) to 43.2/1000 (95% CI 42.3-44) for males and from 34.7/1000 (95% CI 33.9-35.5) to 38.1/1000 (95% CI 37.4-39) for females. The mean row incidence is 5.3/1000 (95% CI 5-5.6) person years for males and 4.8/1000 (95% CI 4.4-5.2) person years for females. The rate of hospitalisations for cardiovascular or kidney diseases is greatly increased in diabetic people with respect to non diabetics for both genders. The mortality relative risk is particularly important in younger age classes: diabetic males and females aged 45-64 years present relative risk for death of 1.7 (95% CI 1.58-1.88) and 2.6 (95% CI 2.29-2.97) respectively.

CONCLUSION

This study provides a feasible and efficient method to determine and monitor the incidence and prevalence of diabetes and the occurrence of its complications along with indexes of morbidity and mortality.

Trends in the prevalence and ratio of diagnosed to undiagnosed diabetes according to obesity levels in the U.S

OBJECTIVE

To examine trends in the prevalence of diagnosed and undiagnosed diabetes and the proportion of total cases previously diagnosed, according to obesity status in the U.S. over the past 40 years.

RESEARCH DESIGN AND METHODS

We assembled data from five consecutive cross-sectional national surveys: National Health Examination Survey I (1960-1962), National Health and Nutrition Examination Survey (NHANES) I (1971-1974), NHANES II (1976-1980), NHANES III (1988-1994), and NHANES 1999-2000. Diagnosed diabetes was ascertained, and height and weight were measured in adults aged 20-74 years in all surveys. In NHANES II, NHANES III, and NHANES 1999-2000, a fasting glucose level > or =126 mg/dl was used to identify cases among individuals not reporting diabetes. Design-based analyses and Bayesian models estimate the probability that prevalence of diabetes increased within four BMI groups (<25, 25-29, 30-34, and > or =35 kg/m2).

RESULTS

In the U.S. population aged 20-74 years between 1976-1980 and 1999-2000, significant increases in the prevalence of diagnosed diabetes (3.3-5.8%, probability >99.9%) were accompanied by nonsignificant increases in undiagnosed diabetes (2.0-2.4%, 66.6%). This resulted in an increase in total diabetes (5.3-8.2%, >99.9%) and a modest nonsignificant increase in the proportion of cases that were diagnosed (62-70%, 62.4%). However, these trends varied considerably by BMI level. In individuals with BMI > or =35 kg/m2, diagnosed diabetes increased markedly (from 4.9% in 1960, to 8.6% during 1976-1980, to 15.1% in 1999-2000; probability >99.9%), whereas undiagnosed diabetes declined considerably (12.5% during 1976-1980 to 3.2% in 1999-2000, probability of increase 4.5%) Therefore, the proportion of total diabetes cases that were diagnosed increased from 41 to 83% (probability 99.9%) among individuals with BMI > or =35 kg/m2. By comparison, changes in prevalence within BMI strata <35 kg/m2 were modest and there was no increase in the percent of total cases that were diagnosed.

CONCLUSIONS

National surveys over the last several decades have found large increases in diagnosed diabetes, particularly in overweight and obese individuals, but this has been accompanied by large decreases in undiagnosed diabetes only among individuals with BMI > or =35 kg/m2. This suggests that improvements in diabetes awareness and detection are most prominent among this subgroup.