Comparison of survey and physician claims data for detecting hypertension

Ascertaining the Francophone population in Ontario: validating the language variable in health data

BACKGROUND

Language barriers can impact health care and outcomes. Valid and reliable language data is central to studying health inequalities in linguistic minorities. In Canada, language variables are available in administrative health databases; however, the validity of these variables has not been studied. This study assessed concordance between language variables from administrative health databases and language variables from the Canadian Community Health Survey (CCHS) to identify Francophones in Ontario.

METHODS

An Ontario combined sample of CCHS cycles from 2000 to 2012 (from participants who consented to link their data) was individually linked to three administrative databases (home care, long-term care [LTC], and mental health admissions). In total, 27,111 respondents had at least one encounter in one of the three databases. Language spoken at home (LOSH) and first official language spoken (FOLS) from CCHS were used as reference standards to assess their concordance with the language variables in administrative health databases, using the Cohen kappa, sensitivity, specificity, positive predictive value (PPV), and negative predictive values (NPV).

RESULTS

Language variables from home care and LTC databases had the highest agreement with LOSH (kappa = 0.76 [95%CI, 0.735-0.793] and 0.75 [95%CI, 0.70-0.80], respectively) and FOLS (kappa = 0.66 for both). Sensitivity was higher with LOSH as the reference standard (75.5% [95%CI, 71.6-79.0] and 74.2% [95%CI, 67.3-80.1] for home care and LTC, respectively). With FOLS as the reference standard, the language variables in both data sources had modest sensitivity (53.1% [95%CI, 49.8-56.4] and 54.1% [95%CI, 48.3-59.7] in home care and LTC, respectively) but very high specificity (99.8% [95%CI, 99.7-99.9] and 99.6% [95%CI, 99.4-99.8]) and predictive values. The language variable from mental health admissions had poor agreement with all language variables in the CCHS.

CONCLUSIONS

Language variables in home care and LTC health databases were most consistent with the language often spoken at home. Studies using language variables from administrative data can use the sensitivity and specificity reported from this study to gauge the level of mis-ascertainment error and the resulting bias.

Validation of self-reported cardiovascular problems in childhood cancer survivors by contacting general practitioners: feasibility and results

BACKGROUND

Epidemiological studies often rely on self-reported health problems and validation greatly improves study quality. In a study of late effects after childhood cancer, we validated self-reported cardiovascular problems by contacting general practitioners (GPs). This paper describes: (a) the feasibility of this approach; and (b) the agreement between survivor-reports and reports from their GP.

METHODS

The Swiss Childhood Cancer Survivor Study (SCCSS) contacts all childhood cancer survivors registered in the Swiss Childhood Cancer Registry since 1976 who survived at least 5 years from cancer diagnosis. We validated answers of all survivors who reported a cardiovascular problem in the questionnaire. Reported cardiovascular problems were hypertension, arrhythmia, congestive heart failure, myocardial infarction, angina pectoris, stroke, thrombosis, and valvular problems. In the questionnaire, we further asked survivors to provide a valid address of their GP and a consent for contact. We sent case-report forms to survivors' GPs and requested information on cardiovascular diagnoses of their patients. To determine agreement between information reported by survivors and GPs, we calculated Cohen's kappa (κ) coefficients for each category of cardiovascular problems.

RESULTS

We used questionnaires from 2172 respondents of the SCCSS. Of 290 survivors (13% of 2172) who reported cardiovascular problems, 166 gave consent to contact their GP and provided a valid address. Of those, 135 GPs (81%) replied, and 128 returned the completed case-report form. Survivor-reports were confirmed by 54/128 GPs (42%). Of the 54 GPs, 36 (28% of 128) confirmed the problems as reported by the survivors; 11 (9% of 128) confirmed the reported problem(s) and gave additional information on more cardiovascular outcomes; and seven GPs (5% of 128) confirmed some, but not all cardiovascular problems. Agreement between GPs and survivors was good for stroke (κ = 0.79), moderate for hypertension (κ = 0.51), arrhythmias (κ = 0.41), valvular problems (κ = 0.41) and thrombosis (κ = 0.56), and poor for coronary heart disease (κ = 0.15) and heart failure (κ = 0.32).

CONCLUSIONS

Despite excellent GP compliance, it was found unfeasible to validate self-reported cardiovascular problems via GPs because they do not serve as gatekeepers in the Swiss health care system. It is thus necessary to develop other validation methods to improve the quality of patient-reported outcomes.

Cardiovascular disease risk factors in autistic adults: The impact of sleep quality and antipsychotic medication use

Approximately 40% of American adults are affected by cardiovascular disease (CVD) risk factors (e.g., high blood pressure, high cholesterol, diabetes, and overweight or obesity), and risk among autistic adults may be even higher. Mechanisms underlying the high prevalence of CVD risk factors in autistic people may include known correlates of CVD risk factors in other groups, including high levels of perceived stress, poor sleep quality, and antipsychotic medication use. A sample of 545 autistic adults without intellectual disability aged 18+ were recruited through the Simons Foundation Powering Autism Research, Research Match. Multiple linear regression models examined the association between key independent variables (self-reported perceived stress, sleep quality, and antipsychotic medication use) and CVD risk factors, controlling for demographic variables (age, sex assigned at birth, race, low-income status, autistic traits). Overall, 73.2% of autistic adults in our sample had an overweight/obesity classification, 45.3% had high cholesterol, 39.4% had high blood pressure, and 10.3% had diabetes. Older age, male sex assigned at birth, and poorer sleep quality were associated with a higher number of CVD risk factors. Using antipsychotic medications was associated with an increased likelihood of having diabetes. Poorer sleep quality was associated with an increased likelihood of having an overweight/obesity classification. Self-reported CVD risk factors are highly prevalent among autistic adults. Both improving sleep quality and closely monitoring CVD risk factors among autistic adults who use antipsychotic medications have the potential to reduce risk for CVD.

Claims-based algorithms for common chronic conditions were efficiently constructed using machine learning methods

Identification of medical conditions using claims data is generally conducted with algorithms based on subject-matter knowledge. However, these claims-based algorithms (CBAs) are highly dependent on the knowledge level and not necessarily optimized for target conditions. We investigated whether machine learning methods can supplement researchers' knowledge of target conditions in building CBAs. Retrospective cohort study using a claims database combined with annual health check-up results of employees' health insurance programs for fiscal year 2016-17 in Japan (study population for hypertension, N = 631,289; diabetes, N = 152,368; dyslipidemia, N = 614,434). We constructed CBAs with logistic regression, k-nearest neighbor, support vector machine, penalized logistic regression, tree-based model, and neural network for identifying patients with three common chronic conditions: hypertension, diabetes, and dyslipidemia. We then compared their association measures using a completely hold-out test set (25% of the study population). Among the test cohorts of 157,822, 38,092, and 153,608 enrollees for hypertension, diabetes, and dyslipidemia, 25.4%, 8.4%, and 38.7% of them had a diagnosis of the corresponding condition. The areas under the receiver operating characteristic curve (AUCs) of the logistic regression with/without subject-matter knowledge about the target condition were .923/.921 for hypertension, .957/.938 for diabetes, and .739/.747 for dyslipidemia. The logistic lasso, logistic elastic-net, and tree-based methods yielded AUCs comparable to those of the logistic regression with subject-matter knowledge: .923-.931 for hypertension; .958-.966 for diabetes; .747-.773 for dyslipidemia. We found that machine learning methods can attain AUCs comparable to the conventional knowledge-based method in building CBAs.

Agreement Between Administrative Database and Medical Chart Review for the Prediction of Chronic Kidney Disease G category

Background

Chronic kidney disease (CKD) is a major health issue and cardiovascular risk factor. Validity assessment of administrative data for the detection of CKD in research for drug benefit and risk using real-world data is important. Existing algorithms have limitations and we need to develop new algorithms using administrative data, giving the importance of drug benefit/risk ratio in real world.

Objective

The aim of this study was to validate a predictive algorithm for CKD GFR category 4-5 (eGFR < 30 mL/min/1.73 m² but not receiving dialysis or CKD G4-5ND) using the administrative databases of the province of Quebec relative to estimated glomerular filtration rate (eGFR) as a reference standard.

Design

This is a retrospective cohort study using chart collection and administrative databases.

Setting

The study was conducted in a community outpatient medical clinic and pre-dialysis outpatient clinic in downtown Montreal and rural area.

Patients

Patient medical files with at least 2 serum creatinine measures (up to 1 year apart) between September 1, 2013, and June 30, 2015, were reviewed consecutively (going back in time from the day we started the study). We excluded patients with end-stage renal disease on dialysis. The study was started in September 2013.

Measurement

Glomerular filtration rate was estimated using the CKD Epidemiological Collaboration (CKD-EPI) from each patient's file. Several algorithms were developed using 3 administrative databases with different combinations of physician claims (diagnostics and number of visits) and hospital discharge data in the 5 years prior to the cohort entry, as well as specific drug use and medical intervention in preparation for dialysis in the 2 years prior to the cohort entry.

Methods

Chart data were used to assess eGFR. The validity of various algorithms for detection of CKD groups was assessed with sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Results

A total of 434 medical files were reviewed; mean age of patients was 74.2 ± 10.6 years, and 83% were older than 65 years. Sensitivity of algorithm #3 (diagnosis within 2-5 years and/or specific drug use within 2 years and nephrologist visit ≥4 within 2-5 years) in identification of CKD G4-5ND ranged from 82.5% to 89.0%, specificity from 97.1% to 98.9% with PPV and NPV ranging from 94.5% to 97.7% and 91.1% to 94.2%, respectively. The subsequent subgroup analysis (diabetes, hypertension, and <65 and ≥65 years) and also the comparisons of predicted prevalence in a cohort of older adults relative to published data emphasized the accuracy of our algorithm for patients with severe CKD (CKD G4-5ND).

Limitations

Our cohort comprised mostly older adults, and results may not be generalizable to all adults. Participants with CKD without 2 serum creatinine measurements up to 1 year apart were excluded.

Conclusions

The case definition of severe CKD G4-5ND derived from an algorithm using diagnosis code, drug use, and nephrologist visits from administrative databases is a valid algorithm compared with medical chart reviews in older adults.

Acute kidney injury in critically ill children and 5-year hypertension

BACKGROUND

To develop a pediatric-specific hypertension algorithm using administrative data and use it to evaluate the association between acute kidney injury (AKI) in the intensive care unit (ICU) and hypertension diagnosis 5 years post-discharge.

METHODS

Two-center retrospective cohort study of children (≤ 18 years old) admitted to the pediatric ICU in Montreal, Canada, between 2003 and 2005 and followed until 2010. Patients with a valid healthcare number and without end-stage renal disease were included. Patients who could not be merged with the provincial database, did not survive admission, underwent cardiac surgery, had pre-existing renal disease associated with hypertension or a prior diagnosis of hypertension were excluded. AKI defined using the Kidney Disease: Improving Global Outcomes (KDIGO) definition. Using diagnostic codes and medications from administrative data, novel pediatric-specific hypertension definitions were designed. Both the evaluation of the prevalence of hypertension diagnosis and the association between AKI and hypertension occurred.

RESULTS

Nineteen hundred and seventy eight patients were included (median age at admission [interquartile range] 4.3 years [1.1-11.8], 44% female, 325 (16.4%) developed AKI). Of these patients, 130 (7%) had a hypertension diagnosis 5 years after discharge. Patients with AKI had a higher prevalence of hypertension diagnosis [non-AKI: 84/1653 (5.1%) vs. AKI: 46/325 (14.2%), p < .001]. Children with AKI had a higher adjusted risk of hypertension diagnosis (hazard ratio [95% confidence interval] 2.19 [1.47-3.26]).

CONCLUSIONS

Children admitted to the ICU have a high prevalence of hypertension post-discharge and children with AKI have over two times higher risk of hypertension compared to those with no AKI.

Developing a case definition for type 1 diabetes mellitus in a primary care electronic medical record database: an exploratory study

BACKGROUND

Identifying cases of disease in primary care electronic medical records (EMRs) is important for surveillance, research, quality improvement and clinical care. We aimed to develop and validate a case definition for type 1 diabetes mellitus using EMRs.

METHODS

For this exploratory study, we used EMR data from the Southern Alberta Primary Care Network within the Canadian Primary Care Sentinel Surveillance Network (CPCSSN), for the period 2008 to 2016. For patients identified as having diabetes mellitus according to the existing CPCSSN case definition, we asked family physicians to confirm the diabetes subtype, to create the reference standard. We used 3 decision-tree classification algorithms and least absolute shrinkage and selection operator logistic regression to identify variables that correctly distinguished between type 1 and type 2 diabetes cases.

RESULTS

We identified a total of 1309 people with type 1 or type 2 diabetes, 110 of whom were confirmed by their physicians as having type 1 diabetes. Two machine learning algorithms were useful in identifying these cases in the EMRs. The first algorithm used "type 1" text words or age less than 22 years at time of initial diabetes diagnosis; this algorithm had sensitivity 42.7% (95% confidence interval [CI] 33.5%-52.5%), specificity 99.3% (95% CI 98.6%-99.7%), positive predictive value 85.5% (95% CI 72.8%-93.1%) and negative predictive value 94.9% (95% CI 93.5%-96.1%). The second algorithm used a combination of free-text terms, insulin prescriptions and age; it had sensitivity 87.3% (95% CI 79.2%-92.6%), specificity 85.4% (95% CI 83.2%-87.3%), positive predictive value 35.6% (95% CI 29.9%-41.6%) and negative predictive value 98.6% (95% CI 97.7%-99.2%).

INTERPRETATION

We used machine learning to develop and validate 2 case definitions that achieve different goals in distinguishing between type 1 and type 2 diabetes in CPCSSN data. Further validation and testing with a larger and more diverse sample are recommended.

Validity of Canadian discharge abstract data for hypertension and diabetes from 2002 to 2013

BACKGROUND

Surveillance using coded administrative health data has shown that the prevalence of hypertension and diabetes in Canada increased substantially between 1998 to 2008. These findings require an assumption that the validity of hypertension and diabetes coding is stable over time. We tested this assumption by examining temporal trends in the validity of coding for hypertension and diabetes in the Canadian hospital Discharge Abstract Database.

METHODS

We used the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) database, a clinical registry, as the reference standard to evaluate the validity of the Discharge Abstract Database in recording hypertension and diabetes in Alberta. The APPROACH database contains data for all Alberta residents who have undergone cardiac catheterization and includes prospective ascertainment of comorbid conditions before each procedure. We linked patient data between the 2 databases for 2002 to 2013 using patient provincial health number. Temporal trends in sensitivity, specificity, positive predictive value, negative predictive value and Cohen κ were calculated for both hypertension and diabetes in the Discharge Abstract Database.

RESULTS

We matched 63 483 patients between the APPROACH database and the Discharge Abstract Database. The validity of the Discharge Abstract Database for hypertension and diabetes remained mostly consistent over time. Between 2002 and 2013, sensitivity, specificity, positive predictive value and negative predictive value ranged from 66% to 87% for hypertension and from 81% to 98% for diabetes; the corresponding κ scores ranged from 0.50 to 0.62 and from 0.80 to 0.89. No significant differences in the validity of coding were found across age, sex or hospital location subgroups.

INTERPRETATION

The validity of coding for hypertension and diabetes in the Discharge Abstract Database remained fairly consistent between 2002 and 2013. Our findings support the use of the Discharge Abstract Database for hypertension and diabetes surveillance in hospital settings.

Validation of Diagnostic Groups Based on Health Care Utilization Data Should Adjust for Sampling Strategy

Supplemental Digital Content is available in the text.

Objective:

Valid measurement of outcomes such as disease prevalence using health care utilization data is fundamental to the implementation of a “learning health system.” Definitions of such outcomes can be complex, based on multiple diagnostic codes. The literature on validating such data demonstrates a lack of awareness of the need for a stratified sampling design and corresponding statistical methods. We propose a method for validating the measurement of diagnostic groups that have: (1) different prevalences of diagnostic codes within the group; and (2) low prevalence.

Methods:

We describe an estimation method whereby: (1) low-prevalence diagnostic codes are oversampled, and the positive predictive value (PPV) of the diagnostic group is estimated as a weighted average of the PPV of each diagnostic code; and (2) claims that fall within a low-prevalence diagnostic group are oversampled relative to claims that are not, and bias-adjusted estimators of sensitivity and specificity are generated.

Application:

We illustrate our proposed method using an example from population health surveillance in which diagnostic groups are applied to physician claims to identify cases of acute respiratory illness.

Conclusions:

Failure to account for the prevalence of each diagnostic code within a diagnostic group leads to the underestimation of the PPV, because low-prevalence diagnostic codes are more likely to be false positives. Failure to adjust for oversampling of claims that fall within the low-prevalence diagnostic group relative to those that do not leads to the overestimation of sensitivity and underestimation of specificity.

Concordance between patient self-reports and claims data on clinical diagnoses, medication use, and health system utilization in Taiwan

Purpose

The aim of this study was to evaluate the concordance between claims records in the National Health Insurance Research Database and patient self-reports on clinical diagnoses, medication use, and health system utilization.

Methods

In this study, we used the data of 15,574 participants collected from the 2005 Taiwan National Health Interview Survey. We assessed positive agreement, negative agreement, and Cohen's kappa statistics to examine the concordance between claims records and patient self-reports.

Results

Kappa values were 0.43, 0.64, and 0.61 for clinical diagnoses, medication use, and health system utilization, respectively. Using a strict algorithm to identify the clinical diagnoses recorded in claims records could improve the negative agreement; however, the effect on positive agreement and kappa was diverse across various conditions.

Conclusion

We found that the overall concordance between claims records in the National Health Insurance Research Database and patient self-reports in the Taiwan National Health Interview Survey was moderate for clinical diagnosis and substantial for both medication use and health system utilization.

Validity of self-reported hypertension: findings from the Thai Cohort Study compared to physician telephone interview

BACKGROUND

Surveys for chronic diseases, and large epidemiological studies of their determinants, often acquire data through self-report since it is feasible and efficient. We examined validity and associations of self-reported hypertension, as verified by physician telephone interview among participants in a large ongoing Thai Cohort Study (TCS).

METHODS

The TCS investigates the health-risk transition among distance learning Open University students living all over Thailand. It began in 2005 and at 4-year follow-up, 60 569 self-reported having or not having doctor diagnosed hypertension. Two hundred and forty participants were randomly selected from each of the "hypertension" and "normotension" self-report groups. A Thai physician conducted a structured telephone interview with the sampled participants and classified them as having hypertension or normotension. The sensitivity, specificity, positive and negative predictive value (PPV and NPV) and overall accuracy of self-report were calculated.

RESULTS

The sensitivity of self-reported hypertension was 82.4% and the specificity was 70.7%. As true prevalence was simulated to vary from 1% to 50% the overall accuracy of self-report varied little from 71% to 75%. High sensitivity and negative predictive value related to female gender, younger age (?40 years), higher education attainment and not visiting a physician in the last 12 months. High specificity and positive predictive value related to female gender, older age, higher education attainment and visiting a doctor in the previous year.

CONCLUSION

Self-report of hypertension had high sensitivity and good overall accuracy. This is acceptable for use in large studies of hypertension, and for estimating its population prevalence to help formulate health policy in Thailand.

Factors affecting the validity of self-reported data on health services from the community health survey in Korea

PURPOSE

As a follow-up for the validity study of Community Health Surveys (CHSs), the purpose of this study was to evaluate the factors affecting the accuracy of CHSs by investigating subjects' characteristics.

MATERIALS AND METHODS

We used data from 11,217 participants (aged 19 years or older) who had participated in the CHS, conducted by a local government in 2008 and analyzed the variables affecting the sensitivity and specificity of hospitalization and outpatient visit.

RESULTS

Multivariate logistic regression analysis showed that, factors related with the sensitivity of hospitalization and outpatient visit questions were gender, age, marital status, chronic diseases, medical checkup, the subjective health status and necessary medical services. Factors related with the specificity were gender, marital status, educational background, chronic diseases, medical checkup, alcohol consumption, necessary medical services and sadness.

CONCLUSION

This study revealed the subject-related factors associated with the validity of the CHS. Efforts to improve the sensitivity and the specificity from self-report questionnaires should consider how the characteristics of subjects may affect their responses.

Does a claims diagnosis of autism mean a true case?

The purpose of this study was to validate autism spectrum disorder cases identified through claims-based case identification algorithms against a clinical review of medical charts. Charts were reviewed for 432 children who fell into one of the three following groups: (a) more than or equal to two claims with an autism spectrum disorder diagnosis code (n = 182), (b) one claim with an autism spectrum disorder diagnosis code (n = 190), and (c) those who had no claims for autism spectrum disorder but had claims for other developmental or neurological conditions (n = 60). The algorithm-based diagnoses were compared with documented autism spectrum disorders in the medical charts. The algorithm requiring more than or equal to two claims for autism spectrum disorder generated a positive predictive value of 87.4%, which suggests that such an algorithm is a valid means to identify true autism spectrum disorder cases in claims data.

Predictors of congruency between self-reported hypertension status and measured blood pressure in the stroke belt

BACKGROUND

Few studies have comprehensively investigated the validity of self-reported hypertension (HTN) and assessed predictors of HTN status in the stroke belt. This study evaluates validity self-reporting as a tool to screen large study populations and determine predictors of congruency between self-reported HTN and clinical measures.

METHODS

Community Initiative to Eliminate Stroke project (n = 16,598) was conducted in two counties of North Carolina in 2004 to 2007, which included collection of self-reported data and clinical data of stroke-related risk factors. Congruency between self-reported HTN status and clinical measures was based on epidemiological parameters of sensitivity, specificity, and predictive values. McNemar's test and Kappa agreement levels assessed differences in congruency, while odds ratios and logistic regression determined significant predictors of congruency.

RESULTS

Sensitivity of self-reported HTN was low (33.3%), but specificity was high (89.5%). Prevalence of self-reported HTN was 16.15%. Kappa agreement between self-report and clinical measures for blood pressure was fair (k = 0.25). Females, whites, and young adults were most likely to be positively congruent, whereas individuals in high risk categories for total blood cholesterol, low density lipoproteins, triglycerides, and diabetes were least likely to accurately capture their HTN status.

CONCLUSION

Self-report HTN information should be used with caution as an epidemiological investigation tool.

Further validation that claims data are a useful tool for epidemiologic research on hypertension

BACKGROUND

The practice of using medical service claims in epidemiologic research on hypertension is becoming increasingly common, and several published studies have attempted to validate the diagnostic data contained therein. However, very few of those studies have had the benefit of using actual measured blood pressure as the gold standard. The goal of this study is to assess the validity of claims data in identifying hypertension cases and thereby clarify the benefits and limitations of using those data in studies of chronic disease etiology.

METHODS

Disease status was assigned to 19,150 employees at a U.S. manufacturing company where regular physical examinations are performed. We compared the presence of hypertension in the occupational medical charts against diagnoses obtained from administrative claims data.

RESULTS

After adjusting for potential confounders, those with measured blood pressure indicating stage 1 hypertension were 3.69 times more likely to have a claim than normotensives (95% CI: 3.12, 4.38) and those indicating stage 2 hypertension were 7.70 times more likely to have a claim than normotensives (95% CI: 6.36, 9.35). Comparing measured blood pressure values identified in the medical charts to the algorithms for diagnosis of hypertension from the claims data yielded sensitivity values of 43-61% and specificity values of 86-94%.

CONCLUSIONS

The medical service claims data were found to be highly specific, while sensitivity values varied by claims algorithm suggesting the possibility of under-ascertainment. Our analysis further demonstrates that such under-ascertainment is strongly skewed toward those cases that would be considered clinically borderline or mild.

Patient, physician, encounter, and billing characteristics predict the accuracy of syndromic surveillance case definitions

BACKGROUND

Syndromic surveillance systems are plagued by high false-positive rates. In chronic disease monitoring, investigators have identified several factors that predict the accuracy of case definitions based on diagnoses in administrative data, and some have even incorporated these predictors into novel case detection methods, resulting in a significant improvement in case definition accuracy. Based on findings from these studies, we sought to identify physician, patient, encounter, and billing characteristics associated with the positive predictive value (PPV) of case definitions for 5 syndromes (fever, gastrointestinal, neurological, rash, and respiratory (including influenza-like illness)).

METHODS

The study sample comprised 4,330 syndrome-positive visits from the claims of 1,098 randomly-selected physicians working in Quebec, Canada in 2005-2007. For each visit, physician-facilitated chart review was used to assess whether the same syndrome was present in the medical chart (gold standard). We used multivariate logistic regression analyses to estimate the association between claim-chart agreement about the presence of a syndrome and physician, patient, encounter, and billing characteristics.

RESULTS

The likelihood of the medical chart agreeing with the physician claim about the presence of a syndrome was higher when the treating physician had billed many visits for the same syndrome recently (ORper 10 visit, 1.05; 95% CI, 1.01-1.08), had a lower workload (ORper 10 claims, 0.93; 95% CI, 0.90-0.97), and when the patient was younger (ORper 5 years of age, 0.96; 95% CI, 0.94-0.97), and less socially deprived (ORmost versus least deprived, 0.76; 95% CI, 0.60-0.95).

CONCLUSIONS

Many physician, patient, encounter, and billing characteristics associated with the PPV of surveillance case definition are accessible to public health, and could be used to reduce false-positive alerts by surveillance systems, either by focusing on the data most likely to be accurate, or by adjusting the observed data for known biases in diagnosis reporting and performing surveillance using the adjusted values.

Are low wages risk factors for hypertension?

OBJECTIVE

Socio-economic status (SES) is strongly correlated with hypertension. But SES has several components, including income and correlations in cross-sectional data need not imply SES is a risk factor. This study investigates whether wages-the largest category within income-are risk factors.

METHODS

We analysed longitudinal, nationally representative US data from four waves (1999, 2001, 2003 and 2005) of the Panel Study of Income Dynamics. The overall sample was restricted to employed persons age 25-65 years, n = 17 295. Separate subsamples were constructed of persons within two age groups (25-44 and 45-65 years) and genders. Hypertension incidence was self-reported based on physician diagnosis. Our study was prospective since data from three base years (1999, 2001, 2003) were used to predict newly diagnosed hypertension for three subsequent years (2001, 2003, 2005). In separate analyses, data from the first base year were used to predict time-to-reporting hypertension. Logistic regressions with random effects and Cox proportional hazards regressions were run.

RESULTS

Negative and strongly statistically significant correlations between wages and hypertension were found both in logistic and Cox regressions, especially for subsamples containing the younger age group (25-44 years) and women. Correlations were stronger when three health variables-obesity, subjective measures of health and number of co-morbidities-were excluded from regressions. Doubling the wage was associated with 25-30% lower chances of hypertension for persons aged 25-44 years.

CONCLUSIONS

The strongest evidence for low wages being risk factors for hypertension among working people were for women and persons aged 25-44 years.

Diagnosed hypertension in Canada: incidence, prevalence and associated mortality

BACKGROUND

Hypertension is a leading risk factor for cardiovascular diseases. Our objectives were to examine the prevalence and incidence of diagnosed hypertension in Canada and compare mortality among people with and without diagnosed hypertension.

METHODS

We obtained data from linked health administrative databases from each province and territory for adults aged 20 years and older. We used a validated case definition to identify people with hypertension diagnosed between 1998/99 and 2007/08. We excluded pregnant women from the analysis.

RESULTS

This retrospective population-based study included more than 26 million people. In 2007/08, about 6 million adults (23.0%) were living with diagnosed hypertension and about 418,000 had a new diagnosis. The age-standardized prevalence increased significantly from 12.5% in 1998/99 to 19.6% in 2007/08, and the incidence decreased from 2.7 to 2.4 per 100. Among people aged 60 years and older, the prevalence was higher among women than among men, as was the incidence among people aged 75 years and older. The prevalence and incidence were highest in the Atlantic region. For all age groups, all-cause mortality was higher among adults with diagnosed hypertension than among those without diagnosed hypertension.

INTERPRETATION

The overall prevalence of diagnosed hypertension in Canada from 1998 to 2008 was high and increasing, whereas the incidence declined during the same period. These findings highlight the need to continue monitoring the effectiveness of efforts for managing hypertension and to enhance public health programs aimed at preventing hypertension.

Accuracy of administrative claims data for polypectomy

BACKGROUND

The frequency of polypectomy is an important indicator of quality assurance for population-based colorectal cancer screening programs. Although administrative databases of physician claims provide population-level data on the performance of polypectomy, the accuracy of the procedure codes has not been examined. We determined the level of agreement between physician claims for polypectomy and documentation of the procedure in endoscopy reports.

METHODS

We conducted a retrospective cohort study involving patients aged 50-80 years who underwent colonoscopy at seven study sites in Montréal, Que., between January and March 2007. We obtained data on physician claims for polypectomy from the Régie de l'Assurance Maladie du Québec (RAMQ) database. We evaluated the accuracy of the RAMQ data against information in the endoscopy reports.

RESULTS

We collected data on 689 patients who underwent colonoscopy during the study period. The sensitivity of physician claims for polypectomy in the administrative database was 84.7% (95% confidence interval [CI] 78.6%-89.4%), the specificity was 99.0% (95% CI 97.5%-99.6%), concordance was 95.1% (95% CI 93.1%-96.5%), and the kappa value was 0.87 (95% CI 0.83-0.91).

INTERPRETATION

Despite providing a reasonably accurate estimate of the frequency of polypectomy, physician claims underestimated the number of procedures performed by more than 15%. Such differences could affect conclusions regarding quality assurance if used to evaluate population-based screening programs for colorectal cancer. Even when a high level of accuracy is anticipated, validating physician claims data from administrative databases is recommended.

Comparing self-reported and measured high blood pressure and high cholesterol status using data from a large representative cohort study

OBJECTIVE

To examine the relationship between self-reported and clinical measurements for high blood pressure (HBP) and high cholesterol (HC) in a random population sample.

METHOD

A representative population sample of adults aged 18 years and over living in the north-west region of Adelaide (n=1537) were recruited to the biomedical cohort study in 2002/03. In the initial cross-sectional component of the study, self-reported HBP status and HC status were collected over the telephone. Clinical measures of blood pressure were obtained and fasting blood taken to determine cholesterol levels. In addition, data from a continuous chronic disease and risk factor surveillance system were used to assess the consistency of self-reported measures over time.

RESULT

Self-report of current HBP and HC showed >98% specificity for both, but sensitivity was low for HC (27.8%) and moderate for HBP (49.0%). Agreement between current self-report and clinical measures was moderate (kappa 0.55) for HBP and low (kappa 0.30) for HC. Demographic differences were found with younger people more likely to have lower sensitivity rates. Self-reported estimates for the surveillance system had not varied significantly over time.

CONCLUSION

Although self-reported measures are consistent over time there are major differences between the self-reported measures and the actual clinical measurements. Technical aspects associated with clinic measurements could explain some of the difference.

IMPLICATIONS

Monitoring of these broad population measures requires knowledge of the differences and limitations in population settings.

Regional Variation in Home Care Use in Manitoba

Previous studies report geographic variation in the use of home care services. In the province of Manitoba, home care is a core service that Manitoba' twelve regional health authorities (RHAs) are obligated to deliver. Manitoba' RHAs range from remote northern and rural southern regions to a major city, resulting in different challenges for delivering home care. Given this potential for inconsistent delivery and the previous findings of regional variation in other settings, the objective of this study was to measure and assess variation in the use of home care across Manitoba' RHAs. We used data from the Provincial Home Care Program' client registry, other health care administrative databases, and Vital Statistics. Home care use was measured using multiple indicators, including rates of population use, use after hospitalization, before entry to a long-term care facility, and before death. While some important differences emerged, overall we found comparable use of home care across Manitoba.

Comparing administrative and survey data for ascertaining cases of irritable bowel syndrome: a population-based investigation

Background

Administrative and survey data are two key data sources for population-based research about chronic disease. The objectives of this methodological paper are to: (1) estimate agreement between the two data sources for irritable bowel syndrome (IBS) and compare the results to those for inflammatory bowel disease (IBD); (2) compare the frequency of IBS-related diagnoses in administrative data for survey respondents with and without self-reported IBS, and (3) estimate IBS prevalence from both sources.

Methods

This retrospective cohort study used linked administrative and health survey data for 5,134 adults from the province of Manitoba, Canada. Diagnoses in hospital and physician administrative data were investigated for respondents with self-reported IBS, IBD, and no bowel disorder. Agreement between survey and administrative data was estimated using the κ statistic. The χ² statistic tested the association between the frequency of IBS-related diagnoses and self-reported IBS. Crude, sex-specific, and age-specific IBS prevalence estimates were calculated from both sources.

Results

Overall, 3.0% of the cohort had self-reported IBS, 0.8% had self-reported IBD, and 95.3% reported no bowel disorder. Agreement was poor to fair for IBS and substantially higher for IBD. The most frequent IBS-related diagnoses among the cohort were anxiety disorders (34.4%), symptoms of the abdomen and pelvis (26.9%), and diverticulitis of the intestine (10.6%). Crude IBS prevalence estimates from both sources were lower than those reported previously.

Conclusions

Poor agreement between administrative and survey data for IBS may account for differences in the results of health services and outcomes research using these sources. Further research is needed to identify the optimal method(s) to ascertain IBS cases in both data sources.

Does the Relationship Between Neighborhood Socioeconomic Status and Health Outcomes Persist Into Very Old Age? A Population-Based Study

Objective: The purpose of this article is (a) to extend previous research on the relationship between neighborhood socioeconomic status (SES) and health by considering a wide range of health-related measures derived from administrative health care records and (b) to explore whether this relationship persists into old age. Method: The study involved a complete cohort of community-dwelling residents in Winnipeg, Canada, who were 65 years or older in 2004/2005 ( N = 77,930). Health measures were derived from administrative claims data. Census data were used to derive neighborhood-level SES. Results: Multilevel logistic regressions indicated that, relative to individuals living in the most affluent areas, those in the poorest areas had significantly higher odds of having arthritis, diabetes, hypertension, congestive heart failure, ischemic heart disease, chronic obstructive pulmonary disease, depression, and stroke. Significant neighborhood income effects tended to be evident among individuals age 65 to 75 as well as those age 75+. Discussion: A wide range of health conditions among older adults are disproportionately clustered into the poorest areas. Programs and services should be designed to meet the needs of older adults of any age in such neighborhoods.

How accurate are self-reports? Analysis of self-reported health care utilization and absence when compared with administrative data

OBJECTIVE

To determine the accuracy of self-reported health care utilization and absence reported on health risk assessments against administrative claims and human resource records.

METHODS

Self-reported values of health care utilization and absenteeism were analyzed for concordance to administrative claims values. Percent agreement, Pearson's correlations, and multivariate logistic regression models examined the level of agreement and characteristics of participants with concordance.

RESULTS

Self-report and administrative data showed greater concordance for monthly compared with yearly health care utilization metrics. Percent agreement ranged from 30% to 99% with annual doctor visits having the lowest percent agreement. Younger people, males, those with higher education, and healthier individuals more accurately reported their health care utilization and absenteeism.

CONCLUSIONS

Self-reported health care utilization and absenteeism may be used as a proxy when medical claims and administrative data are unavailable, particularly for shorter recall periods.

Assessing validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions in a unique dually coded database

OBJECTIVE

The goal of this study was to assess the validity of the International Classification of Disease, 10th Version (ICD-10) administrative hospital discharge data and to determine whether there were improvements in the validity of coding for clinical conditions compared with ICD-9 Clinical Modification (ICD-9-CM) data.

METHODS

We reviewed 4,008 randomly selected charts for patients admitted from January 1 to June 30, 2003 at four teaching hospitals in Alberta, Canada to determine the presence or absence of 32 clinical conditions and to assess the agreement between ICD-10 data and chart data. We then re-coded the same charts using ICD-9-CM and determined the agreement between the ICD-9-CM data and chart data for recording those same conditions. The accuracy of ICD-10 data relative to chart data was compared with the accuracy of ICD-9-CM data relative to chart data.

RESULTS

Sensitivity values ranged from 9.3 to 83.1 percent for ICD-9-CM and from 12.7 to 80.8 percent for ICD-10 data. Positive predictive values ranged from 23.1 to 100 percent for ICD-9-CM and from 32.0 to 100 percent for ICD-10 data. Specificity and negative predictive values were consistently high for both ICD-9-CM and ICD-10 databases. Of the 32 conditions assessed, ICD-10 data had significantly higher sensitivity for one condition and lower sensitivity for seven conditions relative to ICD-9-CM data. The two databases had similar sensitivity values for the remaining 24 conditions.

CONCLUSIONS

The validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions was generally similar though validity differed between coding versions for some conditions. The implementation of ICD-10 coding has not significantly improved the quality of administrative data relative to ICD-9-CM. Future assessments like this one are needed because the validity of ICD-10 data may get better as coders gain experience with the new coding system.

Hypertension and happiness across nations

In surveys of well-being, countries such as Denmark and the Netherlands emerge as particularly happy while nations like Germany and Italy report lower levels of happiness. But are these kinds of findings credible? This paper provides some evidence that the answer is yes. Using data on 16 countries, it shows that happier nations report systematically lower levels of hypertension. As well as potentially validating the differences in measured happiness across nations, this suggests that blood-pressure readings might be valuable as part of a national well-being index. A new ranking of European nations' GHQ-N6 mental health scores is also given.

The relationship between work hours and utilization of general practitioners in four Canadian provinces

OBJECTIVE

To assess whether long work hours act as a barrier to accessing general practitioner (GP) services.

DATA SOURCES

Secondary data from the 1996/1997 National Population Health Survey (NPHS) and administrative health services utilization data from four Canadian provinces.

STUDY DESIGN

This study was cross-sectional, however, employment variables and GP utilization were reflective of the 12-month period preceding the NPHS interview date. Negative binomial regression was used to model the relationship between the number of GP visits in a 1-year period and employment-related variables while adjusting for other determinants of GP utilization including education, income, and health status.

DATA EXTRACTION METHODS

NPHS and administrative data were linked to create an analysis file.

PRINCIPAL FINDINGS

Subjects with long, standard work hours (>45 hours/week, with most hours during the day) had significantly lower GP utilization rates compared with full-time workers. White-collar workers with long work hours visited a GP significantly less often than white-collar workers with regular hours.

CONCLUSIONS

Long work hours may act as a nonfinancial barrier to accessing GP services independent of health status.

Accuracy of administrative databases in identifying patients with hypertension

BACKGROUND

Traditionally, the determination of the occurrence of hypertension in patients has relied on costly and time-consuming survey methods that do not allow patients to be followed over time.

OBJECTIVES

To determine the accuracy of using administrative claims data to identify rates of hypertension in a large population living in a single-payer health care system.

METHODS

Various definitions for hypertension using administrative claims databases were compared with 2 other reference standards: (1) data obtained from a random sample of primary care physician offices throughout the province, and (2) self-reported survey data from a national census.

RESULTS

A case-definition algorithm employing 2 outpatient physician billing claims for hypertension over a 3-year period had a sensitivity of 73% (95% confidence interval [CI] 69%-77%), a specificity of 95% (CI 93%-96%), a positive predictive value of 87% (CI 84%-90%), and a negative predictive value of 88% (CI 86%-90%) for detecting hypertensive adults compared with physician-assigned diagnoses. Compared with self-reported survey data, the algorithm had a sensitivity of 64% (CI 63%-66%), a specificity of 94%(CI 93%-94%), a positive predictive value of 77% (76%-78%), and negative predictive value of 89% (CI 88%-89%). When this algorithm was applied to the entire province of Ontario, the age- and sex-standardized prevalence of hypertension in adults older than 35 years increased from 20% in 1994 to 29% in 2002.

CONCLUSIONS

It is possible to use administrative data to accurately identify from a population sample those patients who have been diagnosed with hypertension. Given that administrative data are already routinely collected, their use is likely to be substantially less expensive compared with serial cross-sectional or cohort studies for surveillance of hypertension occurrence and outcomes over time in a large population.

Residential mobility of individuals with diagnosed schizophrenia: a comparison of single and multiple movers

BACKGROUND

Several studies have compared the residential mobility of individuals with schizophrenia to mobility of individuals with other mental disorders or with no mental disorders. Little research has been undertaken to describe differences between single (i.e., infrequent) and multiple (i.e., frequent) movers with schizophrenia, and the association between frequency of mobility and health and health service use.

METHODS

The data source is population-based administrative records from the province of Manitoba, Canada. Hospital separations and physicians claims are linked to health registration files to identify a cohort with diagnosed schizophrenia and track changes in residential postal code over time. Single movers (N = 736), who had only one postal code change in a 2.5-year observation period, are compared to multiple movers (N = 252), who had two or more postal code changes. Differences in demographic, socioeconomic, and geographic characteristics, measures of health service use, and the prevalence of several chronic diseases were examined using chi(2) tests, logistic regression, and generalized linear regression.

RESULTS

Multiple movers were significantly more likely to be young, live in socioeconomically disadvantaged neighborhoods, and reside in the urban core. The prevalence of a co-occurring substance use disorder and arthritis was higher for multiple than single movers. Use of acute and ambulatory care for schizophrenia, other mental disorders, as well as physical disorders was generally higher for multiple than single movers.

CONCLUSIONS

Frequency of mobility should be considered in the development of needs-based funding plans and service delivery interventions. Other opportunities to use record-linkage techniques to examine residential mobility are considered.

Can the quality of care in family practice be measured using administrative data?

OBJECTIVE

To explore the feasibility of using administrative data to develop process indicators for measuring quality in primary care.

DATA SOURCES/STUDY SETTING

The Population Health Research Data Repository (Repository) housed at the Manitoba Centre for Health Policy which includes physician claims, hospital discharge abstracts, pharmaceutical use (Drug Program Information Network (DPIN)), and the Manitoba Immunization Monitoring Program (MIMS) for all residents of Manitoba, Canada who used the health care system during the 2001/02 fiscal year. Family physicians were identified from the Physician Resource Database. Indicators were developed based on a literature review and focus group validation.

DATA COLLECTION/EXTRACTION METHODS

Data files were extracted from administrative data available in the Repository. We extracted data based on the ICD-9-CM codes and ATC-class drugs prescribed and then linked them to the Physician Resource Database. Physician practices were defined by allocating patients to their most responsible physician. Every family physician in Manitoba that met the inclusion criteria (having either 5 or 10 eligible patients depending on the indicator) was 'scored' on each indicator. Physicians were then grouped according to the proportion of the patients allocated to their practice who received the recommended care for the specific indicator.

PRINCIPAL FINDINGS

Using administrative health data we were able to develop and measure eight indicators of quality of care covering both preventive care services and chronic disease management. The number of eligible physicians and patients varied for each indicator as did the percent of patients with recommended care, per physician. For example, the childhood immunization indicator included 544 physicians who, on average, provided immunization for 65 percent of their patients.

CONCLUSIONS

Quality of care provided by family physicians can be measured using administrative data. Despite the limitations addressed in this paper, this work establishes a practical methodology to measure quality of care provided by family physicians that can be used for quality improvement initiatives.

Data quality in an information-rich environment: Canada as an example

This review evaluates the quality of available administrative data in the Canadian provinces, emphasizing the information needed to create integrated systems. We explicitly compare approaches to quality measurement, indicating where record linkage can and cannot substitute for more expensive record re-abstraction. Forty-nine original studies evaluating Canadian administrative data (registries, hospital abstracts, physician claims, and prescription drugs) are summarized in a structured manner. Registries, hospital abstracts, and physician files appear to be generally of satisfactory quality, though much work remains to be done. Data quality did not vary systematically among provinces. Primary data collection to check place of residence and longitudinal follow-up in provincial registries is needed. Promising initial checks of pharmaceutical data should be expanded. Because record linkage studies were ''conservative'' in reporting reliability, the reduction of time-consuming record re-abstraction appears feasible in many cases. Finally, expanding the scope of administrative data to study health, as well as health care, seems possible for some chronic conditions. The research potential of the information-rich environments being created highlights the importance of data quality.

Estimating the Incidence of Dementia: The Impact of Adjusting for Subject Attrition Using Health Care Utilization Data

PURPOSE

To estimate incidence rates for dementia and the impact of subject attrition on these rates.

METHODS

Crude, age- and gender-specific incidence rates of dementia and Alzheimer's disease were calculated using person-years analysis and Cox proportional hazard models in a population-based cohort study of 1952 adults aged 65+ years in Manitoba, Canada. Rates were standardized to the nondemented population using the direct method. Ratios of incidence rates comparing completers to subjects who had died, refused, or were unavailable for follow up were based on health care utilization data (available for all subject groups) and used to adjust rates for attrition.

RESULTS

Decedents had a significantly higher incidence of dementia than did subjects who completed the follow-up assessment. The incidence in subjects who refused or were unavailable at follow up was intermediate between decedents and completers. Adjusted for attrition, the standardized dementia incidence rate for community and institutional subjects was 25.3/1000 person-years, significantly higher than that based on follow-up assessments only (17.8/1000 person-years; 95% confidence interval: 14.3-21.4).

CONCLUSIONS

The impact of loss to follow up on incidence rates varies depending on the reason for subject attrition. Incidence studies of dementia should develop strategies to characterize and address subject attrition to avoid underestimating disease incidence.

Agreement Between Administrative Claims and the Medical Record in Identifying Patients With a Diagnosis of Hypertension

OBJECTIVE

This study evaluated the accuracy of 2 administrative claims-based selection rules to identify patients with hypertension (HTN) using medical records as the gold standard.

RESEARCH DESIGN

The claims database consisted of inpatient, outpatient, pharmacy, and eligibility claims for members of a single insurance company from January 2000 through March 2003. Medical records were abstracted for 258 matched patient pairs selected by Rule A (at least 1 HTN-related International Classification of Diseases, 9th Revision [ICD-9] claim) and 138 pairs selected by Rule B (at least 1 HTN-related ICD-9 and at least 1 HTN prescription claim) from 31 provider sites. Sensitivity and specificity of the 2 selection rules were computed using medical chart review as the gold standard for a diagnosis of HTN.

SUBJECTS

Of patients selected by Rule A, chart review identified 281 patients with and 235 patients without HTN. Of patients selected by Rule B, chart review identified 172 patients with and 104 patients without HTN.

RESULTS

The sensitivity and specificity was 70.8% and 74.9% for Rule A and 76.2% and 93.3% for Rule B. The kappa score was 0.45 for Rule A and 0.65 for Rule B.

CONCLUSION

To identify patients with HTN, a selection rule using both a diagnosis and prescription claim has greater sensitivity and specificity than a rule using a diagnosis claim only.

Predictors of receipt of a fluoroquinolone versus trimethoprim-sulfamethoxazole for treatment of acute pyelonephritis in women in Manitoba, Canada

PURPOSE

The increasing and comparatively high proportion of uropathogens in Canada resistant to trimethoprim-sulfamethoxazole (TMP-SMX) may be partially responsible for the increasing use of fluoroquinolones. A number of patient-specific variables have been identified as risk factors for infections caused by antibiotic-resistant pathogens. However, variables unrelated to need, have also been associated with receipt of broad-spectrum antibiotics. We identified patient variables associated with receipt of a fluoroquinolone versus TMP-SMX for treatment of acute pyelonephritis.

METHODS

Healthcare claims from the province of Manitoba, Canada for the period February 1996 to March 1999 were examined to identify episodes of pyelonephritis in non-pregnant females between 18 and 65 years of age treated with TMP-SMX or a fluoroquinolone. Patient variables were identified based on healthcare claims review and data from Statistics Canada. Logistic regression was used to model the probability of receipt of a fluoroquinolone.

RESULTS

A total of 1084 women met inclusion criteria; 653 treated with TMP-SMX and 431 treated with a fluoroquinolone. Age, income, rural residence, recent antibiotic use, recent hospitalization and presentation to an emergency room (ER) were positively associated with receipt of a fluoroquinolone.

CONCLUSIONS

Patient variables reportedly associated with an increased probability of resistant organisms (e.g., age, recent antibiotic use and recent hospitalization) were significantly associated with an increased probability of receipt of fluoroquinolones. However, variables unrelated to antibiotic resistance (e.g., income, rural residence and presentation to an ER) were also significantly associated with receipt of a fluoroquinolone.

First-line first? Trends in thiazide prescribing for hypertensive seniors

BACKGROUND

Evidence of reduced cardiovascular morbidity and mortality as well as cost support thiazide diuretics as the first-line choice for treatment of hypertension. The purpose of this study was to determine the proportion of senior hypertensives that received thiazide diuretics as first-line treatment, and to determine if cardiovascular and other potentially relevant comorbidities predict the choice of first-line therapy.

METHODS AND FINDINGS

British Columbia PharmaCare data were used to determine the cohort of seniors (residents aged 65 or older) who received their first reimbursed hypertension drug during the period 1993 to 2000. These individual records were linked to medical and hospital claims data using the British Columbia Linked Health Database to find the subset that had diagnoses indicating the presence of hypertension as well as cardiovascular and other relevant comorbidities. Rates of first-line thiazide prescribing as proportion of all first-line treatment were analysed, accounting for patient age, sex, overall clinical complexity, and potentially relevant comorbidities. For the period 1993 to 2000, 82,824 seniors who had diagnoses of hypertension were identified as new users of hypertension drugs. The overall rate at which thiazides were used as first-line treatment varied from 38% among senior hypertensives without any potentially relevant comorbidity to 9% among hypertensives with previous acute myocardial infarction. The rate of first-line thiazide diuretic prescribing for patients with and without potentially relevant comorbidities increased over the study period. Women were more likely than men, and older patients were more likely than younger, to receive first-line thiazide therapy.

CONCLUSIONS

Findings indicate that first-line prescribing practices for hypertension are not consistent with the evidence from randomized control trials measuring morbidity and mortality. The health and financial cost of not selecting the most effective and least costly therapeutic options are significant.

A longitudinal study of pediatric body mass index values predicted health in middle age

BACKGROUND AND OBJECTIVE

To characterize the use of pediatric body mass index (BMI) to predict obesity, overweight, and diseases in middle age.

METHODS

A longitudinal study of people born in a week in 1958 (n=12,327). The main outcome measures are obesity (BMI > or = 30) and overweight (BMI > or = 25) at age 33 and disease history self-reported at age 42. Receiver operating characteristic (ROC) analysis was performed using BMI measured at ages 7, 11, and 16 years as predictors.

RESULTS

BMI values measured at age 11 could predict obesity at age 33 with areas under ROC curve (AUC) of 0.78 for males and 0.80 for females (each P < .001). BMI values at age 11 predicted overweight with slightly smaller AUC (each P < .001). They could also predict history of diabetes and hypertension (AUC=0.60 and 0.56, respectively, each P < .01), both sexes pooled. Prediction based on BMI at age 7 was less satisfactory; that at 16 gave limited improvement. Cutoff points based on ROC curves, the international reference, and the 85th and 95th percentiles gave very different profiles of diagnostic features.

CONCLUSION

Pediatric BMI may predict adult obesity and overweight with a reasonable profile of sensitivity and specificity.

Validation of diagnostic codes within medical services claims

OBJECTIVES

Few studies have attempted to validate the diagnostic information contained within medical service claims data, and only a small proportion of these have attempted to do so using the medical chart as a gold standard. The goal of this study is to determine the sensitivity and specificity of medical services claims diagnoses for surveillance of 14 drug disease contraindications used in drug utilization review, the Charlson comorbidity index and the Johns Hopkins Adjusted Care Group Case-Mix profile (ADGs).

STUDY DESIGN AND SETTING

Diagnoses were abstracted from the medical charts of 14,980 patients, and were used as the "gold standard," against which diagnoses obtained from the administrative database for the same patients were compared.

RESULTS

Conditions associated with drug disease contraindications with the exception of hypertension and chronic obstructive pulmonary disease (COPD) showed a specificity of 90% or higher. Sensitivity of claims data was substantially lower, with glaucoma, hypertension, and diabetes being the most sensitive conditions at 76, 69, and 64%, respectively. Each of the 18 disease conditions contained in the Charlson comorbidity index showed high specificity, but sensitivity was more variable among conditions as well as by coding definitions. Although ADG specificity was also high, the vast majority of ADGs had sensitivities of less than 60%.

CONCLUSION

The administrative data was found to have diagnoses and conditions that were highly specific but that vary greatly by condition in terms of sensitivity. To appropriately obtain diagnostic profiles, it is recommended that data pertaining to all physician billings be used.

Delivering equitable care: comparing preventive services in Manitoba

OBJECTIVES

This study examined preventive care delivered in Manitoba during the 1990s by 3 different methods -childhood immunizations (by physicians and public health nurses under a government program), screening mammography (through a government program introduced in 1995), and cervical cancer screening (no program).

METHODS

Longitudinal administrative data, an immunization monitoring system, and Canadian census databases were used.

RESULTS

Cervical cancer screening rates remained static and showed strong socioeconomic differences; childhood immunization rates remained high with small socioeconomic gradients. The introduction of the Manitoba Breast Screening Program resulted in rising rates of screening and vanishing socioeconomic gradients.

CONCLUSIONS

Manitoba government programs in childhood immunization and screening mammography actively helped the provision of preventive care. Organized programs that target population groups, recognize barriers to access, and facilitate self-evaluation are critical for equitable delivery.

Using a population-based health information system to study child health

OBJECTIVE

This paper describes the population-based analyses of measures of child health status used throughout this supplement.

METHODS

The articles in this supplement examine health-related data for children 0 to 19 years. Most analyses cover the period from April 1, 1994 to March 31, 1999. Administrative and survey data were used to assess child health and well-being. For regional comparisons, data were broken down by subregions of Manitoba, called Regional Health Authorities (RHAs), and neighbourhoods of Winnipeg, called Winnipeg Community Areas (Winnipeg CAs). The premature mortality rate (PMR) was used as a proxy of the overall health of the population. All graphs comparing rates among RHAs and Winnipeg CAs rank these subregions in the same order, from lowest to highest PMR. Income was operationalized by dividing the province's population into urban and rural quintiles based upon household income. Other aspects of methodology are discussed.

RESULTS

Results are presented in the articles that follow this one.

CONCLUSION

The relationships between key child health indicators and geographic and socioeconomic factors for Manitoba children are discussed in the articles following this one.

Birthweight, childhood growth and hypertension in adulthood

BACKGROUND

Low birthweight (BW) and childhood growth have been hypothesized to be associated with an increased risk of hypertension in later life.

METHODS

We analysed data among 13,467 women with a recalled BW from the Shanghai Women's Health Study. Cases included those with a self-reported hypertension with ('confirmed cases') or without ('possible cases') antihypertensive medication(s) use. Logistic regression was used to derive adjusted odds ratios (OR) and 95% CI.

RESULTS

Birthweight was inversely associated with the odds of early onset (at age 20-40 years) hypertension in a dose response manner (P for trend = 0.01). This association is stronger for 'confirmed' hypertension (only OR for 'confirmed' hypertension are referred to subsequently). Being heavier or taller than average at 15 years of age were both related to elevated odds of early onset hypertension. Women who had a low BW but were heavier than average at age 15 were more than four times (OR = 4.63, 95% CI: 2.40-8.94) more likely to have an early onset hypertension, and those who had a low BW and became taller at 15 years of age had an OR of 1.87 (95% CI: 1.05-3.31). A significant interaction between BW and weight at age 15 was observed (P = 0.04).

CONCLUSION

Our study suggests that low BW, particularly if accompanied by accelerated childhood growth, may increase the risk of early onset hypertension in adulthood.

Hypertensive patients and their general practitioners

This study examined the relative contribution of hypertensive patients and their physicians in selecting total annual physician visit frequencies and made regional comparisons between two Canadian cities. We found that the frequency of physician visits was primarily influenced by physician referrals and physician practice patterns, which accounted for about 80 percent of the total explainable variance in physician visits. The relative contribution of other available patient and physician characteristics in determining visit frequency was rather small.

Administrative data for public health surveillance and planning

Electronically available administrative data are increasingly used by public health researchers and planners. The validity of the data source has been established, and its strengths and weaknesses relative to data abstracted from medical records and obtained via survey are documented. Administrative data are available from a variety of state, federal, and private sources and can, in many cases, be combined. As a tool for planning and surveillance, administrative data show great promise: They contain consistent elements, are available in a timely manner, and provide information about large numbers of individuals. Because they are available in an electronic format, they are relatively inexpensive to obtain and use. In the United States, however, there is no administrative data set covering the entire population. Although Medicare provides health care for an estimated 96% of the elderly, age 65 years and older, there is no comparable source for those under 65.

Monitoring health reform: a report card approach

During the past several years, budget cuts have forced hospitals in several countries to change the way they deliver care. Gilson (Gilson, L. (1998).

DISCUSSION

In defence and pursuit of equity. Social Science & Medicine, 47(12), 1891-1896) has argued that, while health reforms are designed to improve efficiency, they have considerable potential to harm equity in the delivery of health care services. It is essential to monitor the impact of health reforms, not only to ensure the balance between equity and efficiency, but also to determine the effect of reforms on such things as access to care and the quality of care delivered. This paper proposes a framework for monitoring these and other indicators that may be affected by health care reform. Application of this framework is illustrated with data from Winnipeg, Manitoba, Canada. Despite the closure of almost 24% of the hospital beds in Winnipeg between 1992 and 1996, access to care and quality of care remained generally unchanged. Improvements in efficiency occurred without harming the equitable delivery of health care services. Given our increasing understanding of the weak links between health care and health, improving efficiency within the health care system may actually be a prerequisite for addressing equity issues in health.

Health care seeking behavior following a health survey: impact on prevalence estimates of chronic diseases

This article addresses the time sequence between a population health survey and subsequent health care use and how this changes the incidence estimates of selected chronic diseases. A cardiovascular survey of a representative sample of the adult population of Manitoba, Canada was linked with the health insurance claims database. Of the 2792 subjects in the survey, 98% were linked successfully, using an encrypted personal health insurance number. Five years of physician claims data for the survey participants were reviewed including 18 months prior to and 42 months following the survey. Survey participants started seeking confirmation of possible hypertension as soon as they received blood pressure information at the interview. Confirmation of diabetes and elevated cholesterol were not completed until 3-4 months after participants had received the laboratory test results. As many as 4.6 times more new cases of hypertension per month, 5.1 times more cases of elevated cholesterol, and 3.3 times more cases of diabetes were diagnosed following the survey. Surveys designed to determine the prevalence of specific chronic diseases generate new cases within a short time afterwards, thus affecting the original prevalence estimates. The process of assessing the burden of disease in a population is dynamic rather than static, and comparisons across populations need to take into account the frequency and recency of past surveys.